Does Regulation Kill Jobs? 9780812209242

Citation preview

Does Regulation Kill Jobs?

This page intentionally left blank

Does Regulation Kill Jobs? Edited by

Cary Coglianese, Adam M. Finkel, and Christopher Carrigan

U N I V E R S I T Y O F P E N N S Y LVA N I A P R E S S PHIL ADELPHIA

Copyright © 2013 University of Pennsylvania Press All rights reserved. Except for brief quotations used for purposes of review or scholarly citation, none of this book may be reproduced in any form by any means without written permission from the publisher. Published by University of Pennsylvania Press Philadelphia, Pennsylvania 19104- 4112 www.upenn.edu /pennpress Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 Library of Congress Cataloging- in- Publication Data Does regulation kill jobs? / edited by Cary Coglianese, Adam M. Finkel, and Christopher Carrigan.—1st ed. p. cm. ISBN 978- 0- 8122- 4576- 9 (hardcover : alk. paper) Includes bibliographical references and index. 1. Trade regulation—United States. 2. Trade regulation—Economic aspects— United States. 3. Industrial laws and legislation—Economic aspects— United States. 4. Unemployment—United States. 5. United States— Economic conditions—2009–. I. Coglianese, Cary. II. Finkel, Adam M. III. Carrigan, Christopher KF1600.D47 2013 331'.13N72 2013020962

Contents

Preface 1

vii

The Jobs and Regulation Debate 1 Cary Coglianese and Christopher Carrigan

Evidence 2

Analyzing the Employment Impacts of Regulation 33 Richard D. Morgenstern

3

Do the Job Effects of Regulation Differ with the Competitive Environment? 51 Wayne B. Gray and Ronald J. Shadbegian

4

The Employment and Competitiveness Impacts of Power- Sector Regulations 70 Joseph E. Aldy and William A. Pizer

5

Environmental Regulatory Rigidity and Employment in the Electric Power Sector 89 Rolf Färe, Shawna Grosskopf, Carl A. Pasurka, Jr., and Ronald J. Shadbegian

Analytics 6

Toward Best Practices: Assessing the Effects of Regulation on Employment 111 Lisa A. Robinson

7

Emitting More Light than Heat: Lessons from Risk Assessment Controversies for the “Job-Killing Regulations” Debate 128 Adam M. Finkel

vi

Contents

8

Happiness, Health, and Leisure: Valuing the Nonconsumption Impacts of Unemployment 150 Matthew D. Adler

9

A Research Agenda for Improving the Treatment of Employment Impacts in Regulatory Impact Analysis 170 Ann E. Ferris and Al McGartland

10

Employment and Human Welfare: Why Does Benefit– Cost Analysis Seem Blind to Job Impacts? 190 Brian F. Mannix

Reform 11

Unemployment and Regulatory Policy 207 Jonathan S. Masur and Eric A. Posner

12

Reforming the Regulatory Process to Consider Employment and Other Macroeconomic Factors 223 Stuart Shapiro

13

Analysis to Inform Public Discourse on Jobs and Regulation 239 Michael A. Livermore and Jason A. Schwartz

14

Rationing Analysis of Job Losses and Gains: An Exercise in Domestic Comparative Law 256 E. Donald Elliott

Contributors Index

273

279

Acknowledg ments

289

Preface

Are regulations job killers or job creators? This question has dominated much public debate in the United States during the past several years as the nation has suffered sustained high levels of unemployment. Some politicians espouse the view that regulations are job killers, while others claim that regulations either have little negative effect or actually stimulate the creation of new industries and jobs. Although the debate over jobs and regulation often divides along party lines, politicians on both sides of the aisle share a common desire to improve economic conditions and lessen the hardship that unemployment imposes on individuals and their families. This book responds to that common desire by bringing together the work of leading scholars and practitioners to understand better how regulation affects employment and what regulatory agencies might do to improve their analysis of these employment impacts. Despite the obvious reasons for wanting to understand better whether regulation helps or hurts employment, neither regulatory analysts nor academic researchers have yet to develop the kind of evidentiary foundation needed to provide solid answers. Partly this is because the empirical relationship between regulation and employment is harder to untangle than it might seem at first glance. Intuitively it might seem obvious that regulation adversely affects employment. When regulation increases the cost of doing business, it drives up the cost of products and ser vices, reducing demand and thereby shrinking employers’ need for workers and the capacity to retain them. But just as intuitively, it might seem obvious that regulation can promote jobs. After all, one of the ways regulation increases the cost of doing business is by increasing the demand for goods and ser vices needed to comply with the law, thus creating additional demand for labor associated with installing and operating required equipment and implementing mandated protocols. Of course, it is also highly plausible that both intuitions have validity and that the same regulations that increase jobs for some individuals decrease them for others.

viii

Preface

Some empirical research has tested these various intuitions with respect to a limited number of regulatory domains by using several methods and sources of data. As the opening chapters in this book explain in detail, the results of past research have been informative even if still in ways somewhat limited. Given the economy’s complexity and dynamism, combined with regulation’s heterogeneity and expansiveness, more work is needed to produce firm, generalizable answers. This book seeks to help in filling this need. It adds to existing knowledge of regulation’s employment effects as well as aims to stimulate additional research and analysis by academic researchers as well as government analysts. Its chapters offer new empirical findings about the connection between regulation and jobs, substantial ideas about how economic analysis of regulations can better incorporate consideration of employment effects, and proposals for reforming the regulatory process to give employment its proper due in regulatory decision making. Of course, this book does not purport to settle definitively the debate between those who view regulation as job killing and those who view regulation as job creating. It does, however, aim toward narrowing the divide through the achievement of a better understanding of the true consequences of regulation on the level and quality of employment. In times of substantial economic stress, the public and its elected officials naturally expect that government agencies will give greater scrutiny to employment effects as they consider adopting new regulations or modifying existing ones. This book—the first of its kind to examine the relationship between regulation and jobs— offers guidance for ensuring that such heightened scrutiny can meaningfully contribute to improved regulatory analysis, design, and outcomes. Cary Coglianese

Chapter 1

The Jobs and Regulation Debate Cary Coglianese and Christopher Carrigan

The Great Recession wreaked havoc on employment in the United States. Even as the overall economy officially began to pick up by the middle of 2009, the American labor force still struggled to rebound. Month after month, millions of workers lost their jobs and millions more continued to look for new full-time work. Politicians responded to this great economic crisis by, among other things, blaming regulation (Coglianese 2012a). Some blamed the lack of adequate regulation for triggering the economic collapse in the first place, while others blamed regulation and its attendant burdens for hampering the pace of recovery. For those in the latter group, the phrase “ job-killing regulations” became a common rallying cry for a regulatory reform agenda. Still other politicians argued that strong regulations not only could prevent future economic, environmental, and public health disasters but would actually stimulate new jobs, forcing companies to innovate and creating so- called green jobs. Although ideological differences account for much of the polarized political debate over jobs and regulation in the United States, this debate fundamentally centers on an empirical question—namely, what impact regulation has on employment. This question can and should be approached with rigorous economic and policy analysis, and fortunately some important research has already addressed the empirical question. Nevertheless, uncertainty remains about how generalizable existing research findings are to today’s economy as well as exactly how to incorporate what is known about jobs and regulation into decision making about specific new regulations. Given the importance to society of having both effective regulation and available employment opportunities, we have assembled this volume to advance the search for a better understanding of how regulation affects jobs.

2

Cary Coglianese and Christopher Carrigan

In this opening chapter, we begin by showing in greater detail how the political debate over the economy has in recent years also turned into a debate over regulation, with partisans claiming that regulation either kills or creates jobs. Notwithstanding this political rhetoric, the existing empirical research suggests that regulation does relatively little to reduce or increase overall jobs in the United States. We consider here why, given that the published economics research does not provide a strong basis for believing that regulation affects overall employment levels, the political debate has nevertheless focused so much on regulation’s impact on jobs. We offer an account of the political economy of the jobs and regulation debate that emphasizes the distribution of job impacts and the greater responsiveness of the political system to relatively more certain, identifiable job losses than to less certain, unspecified job gains, even if in the aggregate the latter fully offset the former. Our aim is not merely to understand better the puzzling disconnect between politics and economics on this issue, but also to explain why both regulators and researchers ought to be more attentive to the kinds of analytic and empirical issues raised throughout this book. Only by developing better estimates of the real effects of regulation on employment can policy debate in the United States even hope to rise above the current polarized predicament where regulation’s effects on jobs are too often either superficially treated or overblown by officials on both ends of the ideological spectrum.

Jobs and Regulation on the Political Agenda The United States’ worst recession since the 1930s ushered in a deep and sustained period of job losses. Before the recession started in 2007, the national unemployment rate hovered at around 4.5 percent, but it quickly rose to over 7 percent by the end of 2008 and peaked at 10 percent in October 2009 (Bureau of Labor Statistics 2013a). Once the recession officially ended, unemployment took longer to rebound than in any previous recession, remaining at levels above 8 percent for more than three additional years (Bureau of Labor Statistics 2013a). As of February 2013, the United States still had 12 million persons out of work (Bureau of Labor Statistics 2013b). In addition, a substantial proportion of unemployed individuals had been out of work for up to a year or more. Prior to the recession, about 645,000 individuals could be counted as having been unemployed for a year or more, but by 2010 this number had risen to 4.5 million, the largest share of the U.S. labor force facing such longterm unemployment on record (Bureau of Labor Statistics 2010).1 The unemployment crisis prompted a heated political response. Republicans seized on the costs that regulations necessarily impose on

The Jobs and Regulation Debate

3

business and began repeatedly referring to regulations as “ job-killers” (Coglianese 2011), developing what one columnist referred to as “a seemingly immutable law of . . . rhetoric that the word ‘regulation’ can never appear unadorned by the essential adjective ‘ job-killing’ ” (Marcus 2012). In a Republican presidential primary debate in June 2011, Representative Michele Bachmann opined that the U.S. Environmental Protection Agency (EPA) “should really be renamed the job-killing organization of America” (CNN 2011). Another candidate for the Republican presidential nomination, former Utah Governor Jon Huntsman, called for “ending the EPA’s regulatory reign of terror” (Malcolm 2011), while yet another, Texas Governor Rick Perry, referred to a “cemetery for jobs at the EPA” (Broder and Galbraith 2011). The eventual Republican presidential nominee in 2012, former Massachusetts Governor Mitt Romney, made regulatory reform one of the key parts of his plan for restoring economic growth, lambasting what he saw as the government’s destruction of the American dream of economic prosperity “day by day, job-killing regulation by job-killing regulation” (Romney 2012). Even after President Obama’s reelection, Republicans continued to press their argument. In giving the Republican response to President Obama’s 2013 State of the Union address, for example, Senator Marco Rubio (R-Florida) disparaged the passage of “ job-killing laws” (Rubio 2013). Democrats, of course, had their own rhetorical playbook. Although President Obama (2011b) acknowledged that some regulations can be burdensome and even have a “chilling effect” on the economy, he also repeatedly defended the importance of regulation in protecting the public from economic and environmental disasters. Democrats used the words “common sense” instead of “ job-killing” in connection with regulation, defending the need for sensible rules to protect the public from the undesirable by-products of economic activity (Obama 2013a; Reid 2011). Democrats also continued to blame the lack of effective regulation for the economic crisis that triggered the recession (Coglianese 2012a; Obama 2012a), attacking the Republicans’ job-killing argument as a “myth” designed only to help them in “peddling a cure- all tonic of deregulation” (Reid 2011). Responding to the charges leveled specifically against environmental regulation, advocates of more stringent regulation adopted a countervailing rhetoric about “green jobs” (Middle Class Task Force 2009). The basic idea is that the imposition of regulations that call for the adoption of pollution control technology or techniques will support the development of new jobs in firms that produce the required technologies or the know-how to deploy the required techniques. Moreover, such regulations may create jobs within the affected fi rms, as when companies subject to new requirements need to hire additional staff to monitor

4

Cary Coglianese and Christopher Carrigan

compliance or when mandates induce changes to business operations that simply make those operations more labor intensive. Former EPA administrator Carol Browner defended the federal environmental agency by declaring that “the EPA creates opportunities [and] creates jobs” (Browner 2011). At the 2012 Democratic National Convention, former President William Clinton claimed that new federal fuel economy standards adopted by the Obama Administration would generate over 500,000 “good new jobs” over the next two decades (Clinton 2012). In defending his own first-term record, President Obama applauded his administration’s energy regulations for creating “tens of thousands of good American jobs” (Obama 2013b). Clearly, regulation and employment have become firmly linked in contemporary public discourse. That connection actually dates back decades. When Ronald Reagan ran for president in 1980, the United States had been experiencing a short recession—the first dip in a double- dip recession—that brought unemployment levels up from 5.7 percent in July 1979 to 7.8 percent by July 1980 (Bureau of Labor Statistics 2013a). On the campaign trail, Reagan vociferously criticized the Carter Administration for its economic policies, including its “continuing devotion to job-killing regulation” (Cannon 1980). By the 1990s, other politicians could be heard using the job-killing rhetoric—many of them California Republicans, like Reagan. In his first term as California’s governor, for example, Republican Pete Wilson blamed regulation for imposing “ jobkilling burdens” on his state’s businesses (Sacramento Bee, 19 December 1991; San Jose Mercury News, 14 November 1991). Wilson appointed former baseball commissioner Peter Ueberroth to chair a commission designed to develop recommendations to improve California’s economic competitiveness. Ueberroth had regulation in mind when he proclaimed in 1992 that “California has developed the most highly tuned, finely honed job-killing machine that this country has ever seen” (Stevenson 1992).2 Over the years, the “ job-killing” adjective has been used by others as well, such as when Senator Don Nickles (R-Oklahoma) called the ergonomics rule issued by the Clinton Administration’s Occupational Safety and Health Administration “the most intrusive, expensive and job-killing regulation ever handed down” by the agency (Salt Lake City Deseret News, 7 March 2001). Although claims about job killing are hardly new, Figure 1.1 clearly demonstrates how the intensity and frequency of these claims reached new heights during the most recent economic downturn. Not only did the specific phrase “ job-killing regulation” skyrocket in the media (Livermore and Schwartz this volume), but the general connection between jobs and regulation in the media followed a trend that closely tracked the increasing levels of unemployment. Figure 1.1 shows how

The Jobs and Regulation Debate

5

Figure 1.1. “Jobs” and “Regulation” in the Media, 2002–2012. Note: Media mentions were compiled through a LexisNexis database search of five newspapers, the Chicago Tribune, Los Angeles Times, New York Times, Wall Street Journal, and Washington Post, using the following search: regulation w/5 [jobs or employment or unemployment]. Unemployment rate data were collected from the Bureau of Labor Statistics web site.

the word “regulation” came to be increasingly accompanied by “ jobs” or “employment” in national newspapers over a five-year period ending in mid-2012—a trend indicative of the tight linkage between jobs and regulation in political debate. At the same time, the jobs and regulation debate has also manifested itself in some changes in regulatory policy. Perhaps the most striking change occurred at the state level when, on his first day in office in January 2013, Indiana’s new governor, Mike Pence, fulfilled a campaign promise and issued an executive order imposing a statewide moratorium on new regulations in order to “promot[e] job creation, economic development, and freedom” (Pence 2013). At the federal level, President Obama issued an executive order in 2011 expressly affirming that regulation needs to solve policy problems while also “promoting economic growth . . . and job creation” (Obama 2011a). In announcing the order, Obama called on agencies to review their existing regulations and change or repeal those that “stifle job creation and make our economy

6

Cary Coglianese and Christopher Carrigan

less competitive” (Obama 2011b). The President’s Council on Jobs and Competitiveness also issued a series of policy recommendations in early 2012 directed at accelerating employment growth—with regulatory reform being among its major proposals (Jobs Council 2012). Subsequently, President Obama issued another executive order on “reducing regulatory burdens” that directed agencies to “be especially careful not to impose unjustified regulatory requirements” (Obama 2012b). Congress also took steps to reduce perceived regulatory barriers to job growth. In the 112th Congress, the House of Representatives approved the Red Tape Reduction and Small Business Job Creation Act, a bill that would have operated at the federal level much like the Indiana governor’s executive order, imposing an across-the-board moratorium on federal regulations until the unemployment rate fell to 6 percent or lower. The House also passed another bill that would have required all major rules to be approved by Congress before they could take legal effect (Regulations From the Executive in Need of Scrutiny Act of 2011). Yet another bill passed that would have imposed on regulatory agencies a requirement to consider “estimated impacts on jobs” before issuing new regulations (Regulatory Accountability Act of 2011). Although the Democratically controlled Senate never approved any of these bills in the 112th Congress, regulatory reform legislation continued to be debated in the 113th Congress, again with job creation as the key stated objective (e.g., Regulations From the Executive in Need of Scrutiny Act of 2013; Regulatory Sunset and Review Act of 2013; Small Business Freedom of Commerce Act of 2013).

Jobs and Regulation in Economic Research Politicians’ heightened attention to regulation’s contribution to weak labor markets has intuitive appeal. Regulation imposes additional costs on firms, and these costs can in turn affect how many workers firms employ or how much they pay those workers. Basic microeconomic theory holds that when the cost of producing a product increases, the amount of that product that a firm will supply to the market at the existing price will decline. If the firm opts to charge more for its product, the price increases will in turn reduce sales, assuming demand is not completely inelastic (Hall 2013; Mankiw 2012). When output declines, so too does the need for the factors of production—including labor. Even if regulations require only fi xed capital investments that do not directly affect marginal costs, such mandated investments can still force financially struggling firms to close their doors, leaving their workers faced with the prospect of finding new employment.

The Jobs and Regulation Debate

7

Yet theory also predicts that regulations could increase employment. After all, regulation forces firms to incur increased costs in capital or labor (or both) (Berman and Bui 2001; Morgenstern et al. 2002). Any regulation-induced increases in labor costs mean that existing workers are getting paid more, that more workers are being employed, or that these two effects are occurring in tandem. For example, a regulation that requires automobile manufacturers to install catalytic converters or other pollution control devices on cars increases the demand for labor in producing the pollution control technology and installing the mandated devices.3 Predictions that regulation will have significant employment effects—positive or negative—would seem plausible given the size of the overall regulatory burden in the United States. The Office of Management and Budget (OMB) has reported that the estimated annual costs imposed by major regulations adopted from October 2002 through September 2012 totaled between $57 and $84 billion in 2001 dollars— hardly a trivial number in absolute terms (Office of Management and Budget 2013:12). In fiscal year 2012, just 14 rules together generated between $15 and $20 billion in estimated annual costs (Office of Management and Budget 2013:19). OMB estimates that the corresponding benefits of these regulations amply outweigh the costs, but the sheer magnitude of the costs at least reinforces the plausibility of the theoretical expectation that regulation discernibly affects employment. Despite this plausibility, it still remains an empirical question, given the alternative theoretical possibilities, as to whether regulatory mandates do cause employment to rise or fall. Researchers have yet to provide substantial support for either of the possible employment impacts that economic theory predicts, whether increases or decreases in jobs. The number of published studies rigorously examining the question is certainly not large, but to date the empirical work suggests that regulation plays relatively little role in affecting the aggregate number of jobs in the United States (Coglianese 2013; Morgenstern this volume). Studies generally find either no strong relationship at all or relatively modest effects of regulation on employment. Most of the research has focused on the employment effects of environmental regulation.4 In one of the earliest studies, Berman and Bui (2001) analyzed the impact on manufacturing jobs of local air pollution regulations adopted in Southern California. Comparing employment in firms located in that region over time as well as in comparable firms outside of Southern California, they found no substantive or statistically significant effects of local air pollution regulations on employment. Similarly, Morgenstern et al. (2002) evaluated whether reported

8

Cary Coglianese and Christopher Carrigan

spending by firms on environmental regulatory compliance correlated with changes in employment levels across those firms, finding no statistically significant changes in employment averaged across four industrial sectors from 1979 through 1991. Moreover, when analyzed separately, two of the four sectors actually showed small, statistically significant increases in jobs in the face of increased regulatory compliance spending. Using other data and a different study design, Greenstone (2002) found a decrease of an average of about 40,000 jobs per year in facilities located in “nonattainment areas,” that is, parts of the country declared to have “dirty” air and therefore subject to more stringent air pollution requirements under the Clean Air Act. However, because the observed employment changes were relative ones— derived from a comparison with areas in the country lacking more stringent controls—it is not known how much of Greenstone’s observed decrease reflects true job losses in the aggregate rather than a shift in jobs from dirtier areas of the country to cleaner ones. Greenstone (2002:1211) also observed that although the changes he found were “substantial,” they still amounted to a “modest 3.4 percent of total manufacturing sector employment.” More recent work has followed Greenstone’s approach of exploiting variation in the Clean Air Act’s air quality designations, comparing wages over time in cleaner (less regulated) versus dirtier (more regulated) air quality regions throughout the country. Walker (2011, forthcoming) found that overall employment in the more regulated sectors fell by about 15 percent— again relative to areas with less regulation— following the imposition of new clean air designations. The workers in these industries also reportedly saw on average a 20 percent reduction in the present value of their wages following new regulatory controls, with much of this decrease attributable to older, higher-paid workers who were laid off (Walker forthcoming). Although such an earnings effect is certainly nontrivial, Walker has characterized the loss as “relatively small” given that it was “two orders of magnitude below most estimates of the health benefits” of the law (Walker forthcoming). In other words, adding the estimated earnings loss to the computation of costs would make no difference in a benefit– cost assessment of existing air pollution regulation. Walker also did not include in his analysis any offsetting positive effects accruing to workers that gain jobs because of the imposition of new regulation. These major studies indicate that the relationship between regulation and jobs is far less pronounced than typically portrayed in political debate. The research has generated at most only tepid or mixed support for the proposition that regulation kills or creates jobs. Although the

The Jobs and Regulation Debate

9

results vary between positive and negative, statistically significant and insignificant, the studies do fairly consistently demonstrate that any effects of regulation are at most modest relative to the overall size of the labor market.5 That basic conclusion also finds support in additional research studying specific rules (Gray et al. 2011), using international data (Cole and Elliott 2007), employing alternative statistical techniques (Kahn and Mansur 2010), and considering policies for mitigating climate change (Deschenes 2012). In their chapter in this book, Gray and Shadbegian similarly find statistically significant but only “very small” job losses associated with regulation in certain manufacturing sectors. Aldy and Pizer, also in this book, estimate the downstream effects on employment in manufacturing firms caused by a substantial increase in electricity prices, an increase that itself might plausibly be caused by environmental regulation, finding a decline of only 0.2 percent in the level of employment. Data on “green jobs”—those generated by environmental regulation— tend to paint a similar picture of, at most, modest effects from regulation. Porter (2008) has argued that stringent environmental regulations force firms to innovate, thereby inducing gains in firms’ efficiency and competitiveness that offset, or even more than offset, the costs of regulatory compliance (see also Porter and van der Linde 1995). In addition to relying on a controversial assumption that without regulation firms are passing up profitable opportunities for innovation, Porter’s evidence for a regulatory “win–win” consisted primarily of case examples and did not systematically estimate employment effects. Palmer et al. (1995) challenged Porter’s hypothesis by referring to Census Bureau data showing that the cost savings firms reap from complying with environmental regulations amount to no more than 2 percent of firms’ overall regulatory compliance costs.6 Separately, the Bureau of Labor Statistics (2013c) has reported that the percentage of total employment in industries associated with the production of green goods and services accounted for just 2.6 percent of total public and private sector employment. These findings from the literature on environmental regulation’s impact on jobs are generally borne out by the more extensive literature on how minimum wage laws affect employment. Minimum wage requirements directly regulate a key feature of labor markets, so if any kind of regulation affects employment, it should presumably be these laws. For some time now, scholars have assumed that “minimum wage legislation reduces employment” (Sunstein 1993:56). A survey of over 100 studies beginning in the early 1990s concluded that the weight of the evidence supports the view that increasing the minimum wage reduces employment of low wage workers— but the authors of that same

10

Cary Coglianese and Christopher Carrigan

survey also noted that the research results on this question have “by no means always [been] statistically significant” (Neumark and Wascher 2007:121). By contrast, other more recent analyses and surveys of the literature on the effects of minimum wage laws have concluded that such laws have little impact on levels of employment (Dube et al. 2010; Schmitt 2013). Overall, what we know about the relationship between regulation and employment contrasts strikingly with the grandiose claims found in contemporary political debate about either dramatic job-killing or jobcreating effects of regulation. The empirical evidence actually provides little reason to expect that U.S. economic woes can be solved by reforming the regulatory process. Of course, this is not to deny that regulation does sometimes lead to some workers being laid off because of plant closures or slowdowns nor to deny that workers are sometimes hired to install and run new technologies or processes needed to comply with new regulations. But the picture that emerges is far removed from politicians’ emphatic rhetoric about both the job-killing nature of regulation as well as its ability to create lots of green jobs.

Why Politicians Link Regulation and Jobs A mismatch between political rhetoric and academic research should hardly be surprising. Political scientists and pundits often assume that politicians are motivated primarily by the drive to remain elected and that they favor taking symbolic gestures that allow them to claim credit and shift blame (Edelman 1967; Mayhew 1974). Targeting regulation as the source of either economic distress or salvation can certainly be a politically expedient gesture, even if not grounded in evidence (Carrigan and Coglianese 2012). After all, most politicians have few, if any, levers to control the fundamentals of the economy, especially in a period of sharp economic disequilibrium; however, they do have the power to issue, modify, and repeal regulations, thereby presenting an image to their constituents that something is actually being done. But one need not question entirely the sincerity of the politician who focuses on regulation’s impact on jobs. After all, the belief that regulation affects employment does have a basis in economic theory, and the empirical research that tests this belief is far from exhaustive. The data analyzed in the existing literature draw mainly from the 1980s and 1990s, and it is possible that regulation’s effects are different today, whether because firms can more easily outsource overseas, because the cumulative regulatory burden imposed on firms is quantitatively or qualitatively different today, or because regulation’s impacts on employment differ in periods of sustained economic downturns like the one the

The Jobs and Regulation Debate

11

United States recently experienced. In addition, existing research has also been limited to a few types of regulation, mostly labor and environmental policy. Gray and Shadbegian (this volume) report that regulation’s impact on jobs appears to be related to industry structure, suggesting the possibility that regulatory efforts in banking, health care, and other sectors could affect employment in ways that environmental regulation might not.7 We note these limitations in the existing literature not merely to present academic caveats but to suggest why it might appear reasonable for politicians to persist in their belief in regulation’s connection to jobs. The phenomenon at issue is, after all, complex; the research challenges in investigating it are daunting. Consider that during the fiveyear period leading up to the 2008 recession an average of 1.9 million workers were laid off or fired every month in the United States.8 With this much “normal” churning within labor markets, is it any wonder that it is difficult to determine with confidence how many layoffs a regulation, or a set of regulations, might cause? Researchers have a lot of statistical noise to penetrate. And even when they work through the noise, they cannot simply assume that jobs “lost” following the adoption of a regulation would have always been there in the absence of the regulation. Of course, the existing literature does not deny that regulation can affect employment, even if the overall net effects are insignificant or modest. As noted earlier, Morgenstern et al. (2002) found employment higher in two sectors in the face of increased spending on environmental regulation. Conversely, Greenstone (2002) and Walker (2011, forthcoming) showed relative declines in overall employment in areas with heightened levels of environmental controls. In other words, even if job losses in some areas of the country are cancelled out by gains in other areas (as the Morgenstern et al. [2002] results would appear to imply), regulation still can have tangible impacts in terms of job shifts. Some workers lose their jobs while others gain them. Even for the same workers, job shifts can occur when they move to new facilities or assume new responsibilities within the same firms, as well as when they take on new jobs in altogether different firms— jobs that may not necessarily pay as much as their former jobs. For workers and their families, job shifts caused by regulation have real consequences. Politicians care about these consequences. At a recent conference on regulatory reform, Senator Angus King (I-Maine) stated that “the driving issue for all politicians is jobs.” 9 Even if Senator King’s statement is an exaggeration, it may not be much of one. Politicians often treat jobs as possessing intrinsic value, defining—not just contributing to— individuals’ psychological, physical, and social well- being (Kalleberg

12

Cary Coglianese and Christopher Carrigan

2011). President William Clinton (2011:ix) has written: “Work is about more than making a living, as vital as that is. It’s fundamental to human dignity, to our sense of self-worth as useful, independent, free people.” Many years earlier, President Franklin D. Roosevelt declared that “the right to a useful and remunerative job” should be enshrined in a second, economic Bill of Rights (Roosevelt 1944). Political leaders from around the world have forged a Declaration of Human Rights (United Nations General Assembly 1948:Art. 23) that formally pronounces that “everyone has the right to work . . . and to protection against unemployment.” Politicians’ utmost concern for employment is not surprising, given how much their constituents value productive employment. Over the years, the Gallup organization has repeatedly asked survey respondents to assess what they believe is “the most important problem facing this country today” (see, e.g., Saad 2013). In polls asking this question from 1970 to 2013, the economy ranked as one of the top three problems 88 percent of the time (Figure 1.2), greatly outpacing even national defense, which ranked as a distant second and reached at least one of the top three spots in only 43 percent of the polls conducted. The priority the public gives to economic issues in Gallup’s national poll correlates closely with the unemployment rate at the time a poll is taken. As Figure 1.2 shows, economic issues rank as the top problem when unemployment is at its highest. Similarly, Davis and von Wachter (2011) have shown that as the unemployment rate increases nationally, workers’ perceived likelihood of losing their own jobs also increases. The level of public dissatisfaction with regulation also appears to increase with unemployment. As unemployment increased after the last financial crisis, the proportion of respondents reporting that government regulated business “too much” rose from 38 percent in 2007 to 50 percent in 2011 (Newport 2012)—the highest level of disaffection with regulation ever recorded (Carrigan and Coglianese 2012). Public attitudes obviously influence politicians’ incentives. Although economic conditions do not entirely determine politicians’ electoral fortunes (Bartels 2008; Fair 1978; Fiorina 1981; Healy and Malhotra 2013; Niemi et al. 1995; Tufte 1978), few politicians find it desirable to run for reelection in an economic climate of high unemployment. If nothing else, high unemployment leads politicians to create and foster a political narrative that either shifts blame or makes it look like they are taking action to reduce unemployment. Railing against regulators and their failings satisfies these political needs well (Carrigan and Coglianese 2012; Shapiro and Borie-Holtz forthcoming). Regulation also makes an advantageous target because it can be “fixed” without any major budgetary outlays on the part of the government, something that is

4.0%

Frequency of Top Economic Problem Rank 25 75 100 125 50 150 0

13

5.0% 6.0% 7.0% 7.5% Average U.S. Unemployment Rate

The Jobs and Regulation Debate

1

2 3 4 5 6 7 8 9 10 11 12 13 14 Rank Based on Gallup Poll Respondent Percentages Frequency of Top Economic Problem Rank Average U.S. Unemployment Rate

Figure 1.2. Public Ranking of Economic Problems, 1970–2013. Note: Using a database available from the Roper Center at the University of Connecticut, data were compiled from 263 Gallup polls from January 1970 through February 2013 asking respondents, “What do you think is the most important problem facing this country today?” Unemployment rate data were collected from the Bureau of Labor Statistics. For each Gallup poll, responses— economic and noneconomic—were ranked according to the percentage of respondents that named that problem, with the problem receiving the largest percentage being ranked one. Problems categorized as economic included “economy in general,” “unemployment,” “inflation,” “debt,” “recession,” and “wages,” as well as related terms. The top-ranked economic problem in each poll was used to compile the frequency of ranking across all polls. The average unemployment rate for each ranking was based on the months in which the top economic problem received that ranking.

especially helpful when periods of high unemployment combine with concerns about budget deficits and the size of the national debt. Most important, regulation really does affect some workers’ jobs— and politicians respond acutely to how these and other policy impacts are distributed. They care if factories in their districts lay workers off, even though factories in other politicians’ districts might hire more

14

Cary Coglianese and Christopher Carrigan

workers. “All politics is local,” the late House speaker Tip O’Neill famously opined (O’Neill and Hymel 1994). We have long known that impacts of public policy on employment can vary regionally and locally (Haveman and Krutilla 1968). Politicians are sensitive to these local employment effects, even if on net the aggregate impacts on employment across the country as a whole prove benign. Politicians, like most people, care more deeply about impacts that occur close to home. As President Harry S. Truman once stated, “It’s a recession when your neighbor loses his job; it’s a depression when you lose yours” (The Observer, 13 April 1958). By this mea sure, the Great Recession of 2008 spawned millions of depressions— but not ones distributed equally across every state or political district. After the national recession officially ended in 2009, 10 states still went on to suffer their highest rates of unemployment since the Bureau of Labor Statistics began tracking local unemployment in 1976 (Bureau of Labor Statistics 2013a). It is understandable that politicians in states like these will blame regulation for local labor market conditions, notwithstanding evidence showing that regulation has little to no net effect on job levels across the entire country.10 Politicians are also more likely to become activated about regulation’s “ job-killing” effects than about its job- generating potential. Unlike economists, who dispassionately count job losses the same as job gains when trying to tally the overall impacts of regulation in their empirical research, politicians at least implicitly treat job losses as weightier than job gains, even if the jobs pay the same. This is because job losses will often be more predictable and certain than job gains.11 The firms bearing the costs of new regulations already exist— as do jobs in those firms— and these impacted firms and their workers can be expected to mobilize politically. By contrast, job gains will often be more speculative, lacking identifiable firms and workers who could mobilize. When former President Clinton proclaimed that new fuel economy regulations would generate 500,000 new jobs over the next 20 years, no one could really say who specifically would land those jobs (nor even if these jobs would ever truly materialize). By contrast, when regulators propose placing new standards on coal-powered electricity plants, metal finishing plants, or trucking companies, the specific firms in the targeted sector can be assured that their costs of doing business will be affected. And the specific employees in these firms may reasonably wonder whether their own livelihoods will be threatened as well. Many politicians can identify with what Representative Jim Jordan (R-Ohio) once reported about regulation of the trucking industry: “I have heard from truck drivers who . . . tell me that the DOT [Department of Transportation] and the EPA are putting them out of business with their

The Jobs and Regulation Debate

15

multiple mandates” (U.S. House of Representatives, Committee on Oversight and Government Reform 2012:5). He and other legislators have undoubtedly heard from far fewer workers who will find new jobs in the future because of a DOT or EPA rule. In the end, politicians and social scientists are rather like the proverbial blind persons attentive to different parts of the elephant, looking at the connection between jobs and regulation in different ways. Regulation writ large may well have little or no net impact on aggregate employment. That is, job gains from regulation overall may well offset job losses across the entire economy. But this does not mean that individual regulations have no demonstrable or adverse effects on employment within specific regions, industries, and firms. Especially in democracies divided into smaller electoral districts, political leaders respond to individual and local impacts, and they respond to tangible losses more than they do to speculative gains, even when in the aggregate the negative and positive impacts of regulatory policies balance out across the entire nation. What might seem to many economists to be “mere” transfers of jobs can still palpably change real people’s lives by affecting their wage earnings, physical health, and psychological well-being (e.g., Moyle and Parkes 1999). These discrete effects, and the ways that they are distributed, matter to people and to their elected politicians. Politics, after all, is fundamentally about who gets what, when, and how— as well as about who loses what, when, and how (Lasswell 1958).

Implications for Regulatory Analysis Just as regulation’s impacts on jobs matter to citizens and their elected politicians, they should also presumably matter to appointed officials and their analysts within regulatory agencies. For many years, though, agency analysts have tended to ignore any job impacts of proposed regulations in their benefit– cost analyses (Shapiro this volume). Despite being instructed by executive order to consider “adverse effects” of proposed regulations on “productivity, employment, and competitiveness” (Clinton 1993), analysts have simply assumed either that employment effects are already implicitly accounted for in their benefit– cost analyses or that any separate employment effects are too transitory or small to change the outcome of these analyses (Masur and Posner this volume; Hall 2013). Analysts have often adopted a simplifying assumption of full employment (perhaps reasonably so), according to which any worker losing a job because of regulation could readily find another, comparable one elsewhere in the economy (Mannix this volume). With such an assumption, analysts in regulatory agencies have found it easier to focus on the most direct costs and benefits of regulation when

16

Cary Coglianese and Christopher Carrigan

calculating a proposed rule’s net benefits. They have acted as if their role is limited to determining whether the winners under a proposed regulation could in principle pay off the losers, not to worrying much about who the winners or losers might be. The failure to include employment explicitly in benefit– cost analyses of regulation does not derive from any overarching lack of concern about employment on the part of economists and policy analysts. On the contrary, agencies have sometimes tried to estimate the job effects of regulation separately, without incorporating them into their benefit– cost analyses (Ferris and McGartland this volume). Furthermore, in other policy realms, economists have actually undertaken extensive efforts to understand the macroeconomic factors that affect the level of employment in the economy as well as to analyze various policy options for lowering unemployment to its “natural” or “acceptable” rate. In any basic macroeconomics textbook, for example, controlling unemployment occupies a prominent place alongside managing inflation (Mankiw 2010). In practice, economists throughout the executive branch of government pay careful attention to unemployment and policy options to combat it. These economists just tend to work outside the traditional regulatory agencies and instead within other governmental entities such as the White House National Economic Council, the Council of Economic Advisors, and the Federal Reserve. Undoubtedly part of the reason analysts have neglected to itemize job effects in their regulatory benefit– cost analyses is that, as we have discussed, the empirical literature suggests that regulation in the aggregate does not seem to affect overall employment levels. The costs that regulations impose on firms may be sizable, but they are still quite small relative to the overall cost of doing business and do not appear to be the major driver affecting the competitiveness of U.S. industry (Jaffe et al. 1995). Yet the findings from the existing empirical research probably only partly explain why agencies do not incorporate job effects into their benefit– cost analyses of new regulations. After all, the principles of benefit– cost analysis do not say to exclude a specific kind of benefit or cost simply because it might be relatively small. A potentially more important reason for not including job effects in benefit– cost analysis is that doing so has been just too difficult— conceptually, analytically, and empirically (Bartik 2012). If it were easy to estimate and value job impacts reliably, far fewer agencies would hesitate to incorporate such effects into their analyses, especially given politicians’ interest in the connection between regulation and jobs. Still, when it is clear that a proposed regulation will kill or create an estimated number of jobs, particularly if the estimated number of jobs affected is substantial (Elliott this volume), it does make sense for the

The Jobs and Regulation Debate

17

promulgating agency to ensure these job losses are fully factored into its benefit– cost analysis. Unemployment brings with it not just a gain of “leisure” time for workers and a lowering of costs to employers; it can also impose negative consequences in terms of reduced future earnings potential, job search costs, social stigma, and negative physical and mental health effects (Davis and von Wachter 2011; Dooley et al. 1996; Frey and Stutzer 2002; Helliwell and Huang 2011; Sullivan and von Wachter 2009).12 Especially during a severe economic downturn, a regulation that results in layoffs can produce long spells of unemployment, which may cause disproportionate effects on income potential. Those out of work for extended periods can experience significant cuts in their preemployment earnings upon reentering the workforce (Congressional Budget Office 2004, 2007; von Wachter 2010). In effect, job losses caused by a regulation constitute a negative externality of that governmental action. At the same time that a regulation can serve to correct a market externality, thereby delivering benefits to society, the costs that the regulation imposes on firms can create their own externalities, over and above the opportunity costs associated with the resources devoted to complying with the regulation. The Bureau of Labor Statistics (2009:1) puts it this way: “When workers are unemployed, they, their families, and the country as a whole lose. Workers and their families lose wages, and the country loses the goods or ser vices that could have been produced. In addition, the purchasing power of these workers is lost, which can lead to unemployment for yet other workers.” To be complete, benefit– cost analyses of proposed regulations would need to take all of the indirect effects of job losses into account. When incorporating job effects into a benefit– cost analysis, the analyst must confront two questions. First, what will be the impact of the proposed regulation on jobs? That impact could be measured simply by the number of jobs, as it has been in much of the empirical research to date. But employment impacts could also be mea sured in terms of wages, job quality, or job fit. A job, after all, is not a (fungible) job. Job quality is at least partially determined by whether it is high paying or low paying (Acemoglu 2001); however, a “good” job also provides stability, security, and, to some extent, flexibility to its holder—not to mention it should also match well the skills and interests of the job holder (Kalleberg 2011; Tilly 1997). A given regulation may well make no difference in terms of the number of jobs, but it could still affect job pay, quality, or fit. The analyst needs to forecast how an individual regulation will affect the selected employment metric— a task that will seldom be easy. Predicting a regulation’s effects will often require making difficult long-term employment forecasts because regulations last for years and many important rules do not even take legal effect for a year or more after

18

Cary Coglianese and Christopher Carrigan

they are adopted (Robinson this volume). As the effects of regulation on employment are likely to be indirect, if not highly attenuated, regulatory officials may need to abandon their reliance on more tractable partial equilibrium models and work to develop dynamic general equilibrium models, an approach recently explored in industry- sponsored research (Smith et al. 2013). Of course, however they are estimated, employment forecasts need to include both negative effects (losses) and positive ones (gains). After the employment impact of a regulation has been determined, the second question for the benefit– cost analyst is: What is the monetary value of that impact? Actual earnings might initially seem to provide a basis, but when a firm lays off workers or reduces what it pays them, what the workers lose the employer reaps as a corresponding cost savings.13 What matters is valuing the real welfare effects to workers as they are forced to transition to new jobs (Arrow et al. 1996). Presumably that value should be less than current earnings (Bartik 2013). Separate from wages, the analyst could seek to estimate the impact on workers’ welfare by monetizing the ancillary effects of unemployment, such as adverse impacts on health (Adler this volume). Monetizing health effects sometimes generates moral objections (Ackerman and Heinzerling 2004), but well- accepted valuation practices that have been applied to quantify benefits in various policy realms, such as environmental or public health regulation, could be used to value the health effects of unemployment (Finkel this volume). Already, some have suggested that the full stream of ancillary effects from the loss of a single job should be valued around, or even somewhat more than, $100,000 per job in present value terms (Bartik 2013; Masur and Posner 2012). Bartik (2013) suggests that the welfare costs from regulation-induced job losses could amount to 10 percent to 20 percent of the other costs of the regulation conventionally included in a benefit– cost analysis. Of course, to the extent that a regulation also induces job gains, whether in other sectors or in other parts of the country, those positive effects would need to be included when making any complete valuation of job impacts. Still, if the labor impacts expected from a specific proposed regulation were indeed to add even 10 percent to its overall costs, knowing that might sometimes make a difference when public officials have to decide whether to proceed with that regulation— or whether to pursue other options, such as the use of market-based instruments that might potentially have both lower compliance costs and fewer detrimental employment effects (Färe et al. this volume). In the end, that is the purpose of regulatory analysis: to aid in decision making. Given the great concern elected lawmakers have expressed about regulation’s impacts on employment, regulatory analyses can

The Jobs and Regulation Debate

19

better advance public deliberation and decision making if they are more attentive to both the extent and value of regulation’s effects on employment (Elliott this volume; Livermore and Schwartz this volume). Politicians’ sensitivity about local effects also means that benefit– cost analysis of regulation would be more useful if it explicates how both the positive and negative employment effects will be distributed.14 Without more explicit inclusion of job effects into regulatory analysis, officials within agencies could very well be overly influenced by a political process that at times seems to place a nearly infinite value on jobs. Treating employment concerns as a trump card that blocks otherwise welfareenhancing regulation would be a mistake— but so too would it be a mistake to ignore the real employment-related externalities that are not accounted for in the typical benefit– cost analysis. If nothing else, the salience of the political debate over jobs and regulation makes it important to try to get the best possible estimates of both the impacts and value of employment effects.

About This Book The late economist Edward Gramlich once noted, in his leading textbook on benefit– cost analysis, that “the whole jobs issue is a potential alibi for large- scale fudging of numbers” (Gramlich 1990:227). For this reason, respectable economists and analysts have for years concluded that it is often better to make simplifying assumptions that in effect ignore public policy’s ancillary effects on jobs. Such an approach at least advances consistency, and it is certainly better than succumbing to political pressures by fudging numbers. But as Gramlich (1990:227) also noted, the analyst can play an important role in informing decision makers, not simply accepting or ignoring what might merely be politically expedient rationalizations: “Politicians are wont to try to obtain programs, and others to defend them, because they create jobs. At this point the benefit– cost analyst can ask some hard questions— are these temporary or permanent jobs, will the job gains here result in overall employment gains, or will other employment just go down, in which case using labor here is a real cost?” What Gramlich said in the context of government programs aiming to create jobs can also be said with respect to regulations that might either create or destroy jobs. The role of the regulatory analyst is to “ask some hard questions”— and to provide answers that can help decision makers. This vision of the analyst’s role explains the genesis of this book. We believe that the relationship between jobs and regulation deserves both better analysis by regulatory agencies in advance of their decisions as well as more retrospective research that can inform that analysis by

20

Cary Coglianese and Christopher Carrigan

identifying how regulations have affected employment after they have been implemented, how those effects have been distributed, and the conditions under which they have arisen (Coglianese 2012a; Coglianese and Bennear 2005; Greenstone 2009). Along with our coeditor, Adam Finkel, we have assembled an interdisciplinary group of regulatory scholars and analysts to give sustained attention to three vital questions raised by the jobs and regulation debate: Does regulation kill or create jobs? How should regulatory analysts investigate the job effects of regulation? How, if at all, should the regulatory process be reformed to give proper consideration to regulation’s impacts on employment to yield better policy results? The remainder of this book is divided into three parts, each corresponding to one of these three questions. The first part offers the reader a careful presentation of empirical evidence about regulation’s employment effects. In Chapter 2, Richard Morgenstern provides a foundation for the rest of the book by reviewing the existing research on regulation’s employment impacts as well as the welfare effects of unemployment gleaned from labor economists’ studies of mass layoffs. In Chapter 3, Wayne Gray and Ronald Shadbegian offer new data analysis on the relationship between employment and regulation and address a gap in the existing literature by investigating how differences in the competitiveness of different industrial sectors either accentuates or attenuates regulation’s employment effects. Joseph Aldy and William Pizer, in Chapter 4, focus on the relationship between upstream regulation and downstream employment by estimating the spillover effects on manufacturing from regulation-induced price increases in electricity. In Chapter 5, Rolf Färe, Shawna Grosskopf, Carl Pasurka, and Ronald Shadbegian model employment impacts under different regulatory approaches, comparing more rigid, traditional regulation with more flexible, market-based instruments. The second part of the book offers an in- depth treatment of many of the core conceptual and methodological issues that regulatory analysts will need to confront in seeking to improve their analyses of the employment effects of regulation. In Chapter 6, Lisa Robinson outlines nine important principles— or “best practices”—for agencies to follow when seeking to incorporate job impacts into their regulatory impact analyses. In Chapter 7, Adam Finkel translates and applies the lessons learned over the last 30 years in the scientific assessment of public health risks, concluding that analysts investigating employment effects would do well to replicate how health risk assessors have responded to challenges related to uncertainty, bias, and the estimation of secondorder effects. Matthew Adler, in Chapter 8, offers a model for incorporating into agency decision making the effects on individual psychological and physical well-being that can result from unemploy-

The Jobs and Regulation Debate

21

ment as well as some strategies for empirically mea sur ing these impacts. Ann Ferris and Al McGartland, in Chapter 9, explore issues that the EPA has encountered in studying employment effects and then advocate keeping jobs analyses separate from benefit– cost analyses, at least until economic theory and empirical research develop further. Finally, Brian Mannix maintains in Chapter 10 that, while the observable employment impacts of regulation may be important, they cannot simply be grafted onto the standard framework for benefit– cost analysis because, he further argues, such effects are already captured— albeit implicitly— in the standard computation of compliance costs. The third and final part entertains the possibility that the current regulatory process in the United States could be reformed in ways that would better ensure that federal agencies appropriately factored job impacts into their regulatory decision making. In Chapter 11, Jonathan Masur and Eric Posner defend and expand the argument that agencies should incorporate jobs impacts into their benefit– cost analyses (Masur and Posner 2012), recommending that agencies account for more than just first- order effects when making regulatory decisions. Stuart Shapiro, in Chapter 12, reviews how well regulatory agencies are currently doing in analyzing job impacts, concluding that the track record is abysmal and that a new, outside government entity should be charged with evaluating regulation’s effects on jobs. In Chapter 13, Michael Livermore and Jason Schwartz make the democratic case for agencies to conduct better assessments of employment impacts, arguing that such jobs analyses can usefully inform public deliberation regardless of whether they actually alter the outcomes of particular benefit– cost analyses. Finally, in Chapter 14, E. Donald Elliott argues that at the end of the day, the government needs to factor job effects into regulatory analysis when they may be significant either to decisions or to public debate and that experience with similar assessments in the United States and European Union provides a fruitful model for reforming regulatory practice.

Conclusion The impacts of regulation on employment— whether real or just alleged— will continue to matter to public policy decision makers, particularly in times of high unemployment. Although economists may persist in finding little or no aggregate net effect of regulation on jobs, politicians will continue to respond to localized and individual impacts as well as to the distribution of gains and losses. As long as some regulations affect some jobs, politicians will still either criticize or praise regulations for what they do to employment in their districts and states. The

22

Cary Coglianese and Christopher Carrigan

challenge for researchers and analysts is not merely to continue to test claims about how regulation writ large affects aggregate levels of jobs, but also to understand better which regulations have which specific effects on jobs and what are the conditions under which these effects occur. We hope this book can help move forward efforts to meet that challenge. Employment in the United States may have rebounded by the time many readers will encounter the pages of this book; we certainly hope it will have. With time, phrases like “ job- killing regulations” may even fade from the national political discourse. Yet even if economic renewal leads the debate over jobs and regulation to fall dormant for a time in Washington, D.C., it will undoubtedly persist in regulatory disputes at the state and local level and can be counted on to return to the national stage the next time the nation’s economy stalls and unemployment spikes for any sustained period. To ensure that policy analysis can better inform deliberation by the public and their leaders, researchers and analysts should seek to contribute by continuing to engage in the kind of work presented and addressed in the chapters of this book.

Acknowledgments We thank Mariah Ford, Ben Meltzer, Brady Sullivan, and Tim von Dulm for research assistance, Jen Evans and Matthew McCabe for help preparing the manuscript, and John Coglianese, Adam Finkel, Billy Pizer, and Brady Sullivan for their comments. We have also benefited from ideas shared by participants attending the two Penn Program on Regulation events from which this book grows, especially the Washington, D.C. workshop. Although the ground rules for that workshop prevent us from attributing ideas to specific individuals, several participants may notice an affinity between certain of our themes and comments they shared at these events. We gratefully acknowledge here their insights and inspiration.

Notes 1. By “on record,” we mean since 1967, when the Bureau of Labor Statistics began tracking unemployment of one year or more in duration. Although recordhigh, the number of long-term unemployed individuals only hints at the disproportionate burdens that the Great Recession has imposed on specific segments of U.S. society, including younger Americans, minorities, and the poor (Seefeldt and Graham 2013). 2. Other California Republican officials echoed these sentiments at the time (see, e.g., Carbone 1992; Fuentes 1992). 3. Morgenstern et al. (2002) further distinguish between “cost effects” and “factor- shift effects” that arise from regulation-induced increases in produc-

The Jobs and Regulation Debate

23

tion costs. Cost effects arise when, keeping the firm’s ratio of capital to labor the same, regulation increases costs for all factors of production, including capital and labor. Factor- shift effects occur when regulation is more or less labor intensive to implement. If a regulation leads to more labor-intensive operations (decreasing the capital–labor ratio), then wage increases or job gains result (or both). An increase in the capital–labor ratio would have the opposite effect. Any job effects resulting from either cost or factor- shift effects are, of course, distinct from job losses as well as wage cuts associated with the reduced demand for the more costly product. 4. The studies of environmental regulation we discuss in this section are ones that have the most direct mea sures of regulatory stringency, relying either on variation in actual rules or on fi rm- reported data on private sector spending on regulatory compliance. Other studies have attempted to discern various economic effects of regulation at the macro level by deploying proxies such as the size of government budgets, the number of pages of rules, or indices of regulatory burdens and then correlating these with overall macroeconomic indicators (e.g., Beard et al. 2011; Dawson and Seater 2013; Feldmann 2009; Jacobzone et al. 2010). Some of these studies report correlations between the deployed proxies and employment, but other studies using similar measures have found no effects (e.g., Sinclair and Vesey 2012). For a discussion of the use of proxies in regulatory research, see Coglianese (2012b). 5. This view is consistent with responses to a variety of surveys. From 1995 to 2013, for example, the Bureau of Labor Statistics (BLS) surveyed each business that incurred a “mass layoff”—that is, over 50 state unemployment insurance claims within five weeks. Since 2007, BLS has specifically asked these fi rms whether government regulations caused the layoffs—but only a small percentage of businesses has reported that regulation was a factor (Bureau of Labor Statistics 2012). Separately, randomized surveys have found that at most only about a quarter of small business owners view excessive government regulation as a pressing concern (American Sustainable Business Council et al. 2012; Dunkelberg and Wade 2011, 2013; Hall 2011). Similarly, only about a quarter of the respondents in a Wall Street Journal survey of about 50 economists pointed to “uncertainty about government policy” as a factor for the economy’s slow return of employment (Izzo 2011). 6. More recent empirical studies appear to show some support for the Porter hypothesis, suggesting that innovation spurred by regulation may take time to result in productivity gains (Ambec et al. 2013). We thank Adam Finkel for reminding us that even if the Porter hypothesis is true in some situations, employment could still go down because the cost- saving innovations induced by regulation might take the form of new technologies that eliminated some of the need for labor. 7. These results accord with others who have likewise found that the economic effects of regulations vary across sectors (e.g., Jorgenson and Wilcoxen 1990). 8. We used Bureau of Labor Statistics Job Openings and Labor Turnover Survey data at http://data.bls.gov/cgi-bin/dsrv as the source for the number of workers laid off or fired monthly. The average was computed from monthly total U.S. nonfarm layoffs and discharges, seasonally adjusted, over the 60-month period from 2003 through 2007. 9. Senator King made his comment while giving the luncheon address at the Progressive Policy Institute’s conference on “Regulating in the Digital Age,” held in Washington, D.C., on 9 May 2013.

24

Cary Coglianese and Christopher Carrigan

10. Invoking a hypothetical that will surely resonate with our academic readers, a legislative staff member expressed to one of us the reasonableness of politicians’ distributional concerns by imagining a university facing a tough decision that would affect the number of faculty positions. “Wouldn’t the provost want to know how different departments would fare under different options?” the staffer asked rhetorically. Even if the imaginary university’s decision resulted in no change in faculty appointments overall, university officials would presumably still find it relevant to know if the decision meant that the archeology department gained positions while the biology department lost positions (or vice versa). 11. Another possible explanation, from behavioral economics, might be that people feel the hurt of losing something more than the gain associated with getting that thing in the first place (Kahneman 2011). 12. On the other hand, Ruhm (2000) contends that overall mortality actually declines during periods of high unemployment, although the rate of suicide increases. 13. It is possible, of course, that the utility of the lost wages to the worker will not be counterbalanced perfectly by the utility connected to cost savings to the firm. However, the utility from the worker’s so- called leisure time would need to be factored in as well. Economists often use the reservation wage, or the earnings level at which a worker is indifferent between working and not working, to focus on the welfare or utility effects of policies that affect labor choices (Bartik 2012; Haveman and Farrow 2011). Another approach to valuation would be to multiply jobs lost times average unemployment benefits provided by the government. This would not represent a value in terms of economic welfare, but it might still be deemed relevant to public officials who must monitor the public fisc. 14. As Arrow et al. (1996:6) have noted, “While benefit– cost analysis should focus primarily on the overall relationship between benefits and costs, a good benefit– cost analysis will identify important distributional consequences of a policy.”

References Acemoglu, Daron (2001) “Good Jobs versus Bad Jobs,” Journal of Labor Economics 19: 1–21. Ackerman, Frank, and Lisa Heinzerling (2004) Priceless: On Knowing the Price of Everything and the Value of Nothing. New York: New Press. Ambec, Stefan, et al. (2013) “The Porter Hypothesis at 20: Can Environmental Regulation Enhance Innovation and Competitiveness?” Review of Environmental Economics and Policy 7:2–22. American Sustainable Business Council et al. (2012) Small Business Owners’ Opinions on Regulations and Job Creation (1 Feb.), http://asbcouncil.org/sites /default/files/files/Regulations _Poll _Report _FINAL .pdf. Arrow, Kenneth, et al. (1996) Benefit- Cost Analysis in Environmental, Health, and Safety Regulation. Washington, DC: AEI Press. Bartels, Larry M. (2008) Unequal Democracy: The Political Economy of the New Gilded Age. Princeton, NJ: Princeton University Press. Bartik, Timothy J. (2012) “Including Jobs in Benefit- Cost Analysis,” Annual Review of Resource Economics 4: 55–73.

The Jobs and Regulation Debate

25

———. (2013) “Social Costs of Jobs Lost Due to Environmental Regulations.” Upjohn Institute Working Paper 13-193, Kalamazoo, MI. Beard, T. Randolph, et al. (2011) “Regulatory Expenditures, Economic Growth and Jobs: An Empirical Study.” Phoenix Center Policy Bulletin No. 28, http://www.phoenix-center.org/PolicyBulletin/PCPB28Final.pdf. Berman, Eli, and Linda T. M. Bui (2001) “Environmental Regulation and Labor Demand: Evidence from the South Coast Air Basin,” Journal of Public Economics 79: 265– 95. Broder, John M., and Kate Galbraith (2011) “E.P.A. Is Longtime Favorite Target for Perry,” New York Times, 29 Sept., http://www.nytimes.com/2011/09/30/us /politics/epa-is -perrys -favorite -target.html. Browner, Carol (2011) “Job- Killing EPA— Carol Browner, October 19,” Colbert Report interview, http://www.colbertnation.com/the- colbert-report-videos /400166/october-19-2011/indecision-2012—job-killing- epa—carol- browner (accessed 10 March 2013). Bureau of Labor Statistics (2009) How the Government Measures Unemployment. Washington, DC: United States Department of Labor. ———. (2010) Issues in Labor Statistics. United States Department of Labor Summary 10-10, Washington, DC. ———. (2012) Extended Mass Layoffs in 2011. United States Department of Labor, Report 1039, Washington, DC. ———. (2013a) “Labor Force Statistics from the Current Population Survey, (Seas) Unemployment Rate,” United States Department of Labor, http:// data.bls.gov/timeseries/LNS14000000 (accessed 6 June 2013). ———. (2013b) “Labor Force Statistics from the Current Population Survey, (Seas) Unemployment Level,” United States Department of Labor, http:// data.bls.gov/timeseries/LNS13000000 (accessed 6 June 2013). ———. (2013c) Employment in Green Goods and Ser vices – 2011. United States Department of Labor USDL-13- 0476. Washington, DC. Cannon, Lou (1980) “Reagan, Ignoring Bush, Assails Carter’s Policies,” Washington Post, 19 May, sec. A, p. 8. Carbone, Leslie (1992) “Letter to the Editor,” California Journal Weekly, 1 Sept. Carrigan, Christopher, and Cary Coglianese (2012) “Oversight in Hindsight: Assessing the U.S. Regulatory System in the Wake of Calamity,” in C. Coglianese, ed., Regulatory Breakdown: The Crisis of Confidence in U.S. Regulation. Philadelphia: University of Pennsylvania Press. Clinton, William J. (1993) “Executive Order 12866: Regulatory Planning and Review,” Federal Register 58: 190. ———. (2011) Back to Work: Why We Need Smart Government for a Strong Economy. New York: Alfred A. Knopf. ———. (2012) “Transcript of Bill Clinton’s Speech to the Democratic National Convention,” New York Times, 5 Sept., http://www.nytimes.com/2012/09/05 /us/politics/transcript-of -bill -clintons -speech -to -the -democratic -national -convention.html. CNN (2011) “Republican Debate, June 13,” CNN transcript, http://tran scripts.cnn.com/TRANSCRIPTS/1106/13/se.02.html (accessed 10 March 2013). Coglianese, Cary (2011) “Tracking Down Killer Regulations,” RegBlog (19 June), https://www.law.upenn.edu/blogs/regblog/2011/06/tracking- down- killer -regulations.html (accessed 23 June 2013).

26

Cary Coglianese and Christopher Carrigan

———. (2012a) Regulatory Breakdown: The Crisis of Confidence in U.S. Regulation. Philadelphia: University of Pennsylvania Press. ———. (2012b) “Evaluating the Impact of Regulation and Regulatory Policy.” OECD Expert Paper No. 1 (August), http://www.oecd.org/gov/regulatory -policy/1_coglianese%20web.pdf. ———. (2013) “Regulation and Unemployment.” Penn-Wharton Public Policy Initiative Issue Brief (March), http://public.policy.wharton.edu/wp-content /uploads/2013/08/wpp1-issue-Brief-3-kfp1.pdf. Coglianese, Cary, and Lori Snyder Bennear (2005) “Program Evaluation of  Environmental Policies: Toward Evidence- Based Decision Making,” in G. Brewer and P. Stern, eds., Social and Behavioral Science Research Priorities for Environmental Decision Making. Washington, DC: National Academies Press. Cole, Matthew A., and Rob J. Elliott (2007) “Do Environmental Regulations Cost Jobs? An Industry- Level Analysis of the UK,” B.E. Journal of Economic Analysis and Policy 7: 1–25. Congressional Budget Office (2004) Family Income of Unemployment Insurance Recipients. Washington, DC: U.S. Congress. ———. (2007) Long-Term Unemployment. Washington, DC: U.S. Congress. Davis, Steven J., and Till M. von Wachter (2011) “Recessions and the Cost of Job Loss.” NBER Working Paper Series Paper 17638, Cambridge, MA. Dawson, John W., and John J. Seater (2013) “Federal Regulation and Aggregate Economic Growth,” Journal of Economic Growth 18: 137–77. Deschenes, Olivier (2012) “Climate Policy and Labor Markets” in D. Fullerton and C. Volfram, eds., The Design and Implementation of US Climate Policy. Chicago: University of Chicago Press. Dooley, David, et al. (1996) “Health and Unemployment,” Annual Review of Public Health 17: 449– 65. Dube, Arindrajit, et al. (2010) “Minimum Wage Effects Across State Borders: Estimates Using Contiguous Counties,” Review of Economics and Statistics 92: 945– 64. Dunkelberg, William C., and Holly Wade (2011) NFIB Small Business Economic Trends (August), http://www.nfib.com/Portals/0/PDF/sbet/sbet201108.pdf. ———. (2013) NFIB Small Business Economic Trends (June), http://www.nfib.com /Portals/0/PDF/sbet/sbet201306.pdf. Edelman, Murray (1967) The Symbolic Uses of Politics. Urbana: University of Illinois Press. Fair, Ray C. (1978) “The Effect of Economic Events on Votes for President,” Review of Economics and Statistics 60: 159–73. Feldmann, Horst (2009) “The Unemployment Effects of Labor Regulation Around the World,” Journal of Comparative Economics 37: 76– 90. Fiorina, Morris P. (1981) Retrospective Voting in American National Elections. New Haven, CT: Yale University Press. Frey, Bruno S., and Alois Stutzer (2002) “What Can Economists Learn from Happiness Research?” Journal of Economic Literature 40: 402–35. Fuentes, Thomas A. (1992) “Conroy Rejects Mediocre Rating, Says He ‘Hit Ground Running,’ ” Los Angeles Times, 1 Nov., http://articles .latimes .com /1992 -11 -01/local /me -1624 _1 _county -s -delegation -republican -men -and -women-willie -brown. Gramlich, Edward (1990) A Guide to Benefit- Cost Analysis 2nd ed. Englewood Cliffs, NJ: Prentice Hall.

The Jobs and Regulation Debate

27

Gray, Wayne, et al. (2011) “Do EPA Regulations Affect Labor Demand? Evidence from the Pulp and Paper Industry.” Unpublished paper, Clark University, Worcester, MA (23 May). Greenstone, Michael (2002) “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures,” Journal of Political Economy 110: 1175–219. ———. (2009) “Toward a Culture of Persistent Regulatory Experimentation and Evaluation,” in D. Moss and J. Cisternino, eds., New Perspectives on Regulation. Cambridge, MA: Tobin Project. Hall, Keith (2013) “The Employment Costs of Regulation.” Mercatus Center, George Mason University Working Paper Series Paper 13- 06, Arlington, VA. Hall, Kevin (2011) “Regulation, Taxes Aren’t Killing Small Business, Owners Say,” McClatchy Newspapers (Sept. 1), http://www.mcclatchydc.com/2011 /09/01/122865/regulations- taxes- arent- killing.html#ixzz1aDv1MwCt (accessed 23 June 2013). Haveman, Robert H., and Scott Farrow (2011) “Labor Expenditures and Benefit- Cost Accounting in Times of Unemployment,” Journal of Benefit- Cost Analysis 2: 1–7. Haveman, Robert H., and John V. Krutilla (1968) Unemployment, Idle Capacity, and the Evaluation of Public Expenditures: National and Regional Analyses. Baltimore, MD: Johns Hopkins University Press. Healy, Andrew, and Neil Malhotra (2013) “Retrospective Voting Reconsidered,” Annual Review of Political Science 16: 285–306. Helliwell, John F., and Haifang Huang (2011) “New Mea sures of the Costs of Unemployment: Evidence from the Subjective Well- Being of 2.3 Million Americans.” NBER Working Paper Series Paper 16829, Cambridge, MA. Izzo, Phil (2011) “Dearth of Demand Seen Behind Weak Hiring,” Wall Street Journal, 18 July, http://online.wsj.com/article/SB10001424052702303661904 576452181063763332.html. Jacobzone, Stéphane, et al. (2010) “Assessing the Impact of Regulatory Management Systems: Preliminary Statistical and Econometric Estimates.” OECD Working Papers on Public Governance 17, http://dx.doi.org/10.1787/5km fq1pch36h-en. Jaffe, Adam B., et al. (1995) “Environmental Regulation and the Competitiveness of the U.S. Manufacturing: What Does the Evidence Tell Us?” Journal of Economic Literature 33: 132– 63. Jobs Council (2012) Road Map to Renewal: 2011 Year- End Report. Washington, DC: President’s Council on Jobs and Competitiveness. Jorgenson, Dale W., and Peter J. Wilcoxen (1990) “Environmental Regulation and U.S. Economic Growth,” Rand Journal of Economics 21: 314– 40. Kahn, Matthew E., and Erin T. Mansur (2010) “How Do Energy Prices, and Labor and Environmental Regulations Affect Local Manufacturing Employment Dynamics? A Regression Discontinuity Approach.” NBER Working Paper Series Paper 16538, Cambridge, MA. Kahneman, Daniel (2011) Thinking Fast and Slow. New York: Farrar, Straus and Giroux. Kalleberg, Arne L (2011) Good Jobs, Bad Jobs: The Rise of Polarized and Precarious Employment Systems in the United States, 1970s to 2000s. New York: Russell Sage Foundation. Lasswell, Harold D. (1958) Politics: Who Gets What, When, How. New York: Meridian Books.

28

Cary Coglianese and Christopher Carrigan

Malcolm, Andrew (2011) “The Ames Republican Debate Transcript: Everything They Said That You Missed,” Los Angeles Times, 13 Aug., http://latimesblogs .latimes .com/washington/2011/08/ames -debate -romney -bachmann -paul -huntsman-santorum-gingrich-pawlenty-cain.html. Mankiw, N. Gregory (2010) Macroeconomics, 7th ed. New York: Worth Publishers. ———. (2012) Principles of Microeconomics. Mason, OH: South-Western, Cengage Learning. Marcus, Ruth (2012) “Bad Science Around ‘Job- Killing Regulations,’ ” Washington Post, 24 Apr., http://articles.washingtonpost .com/2012-04 -24/opinions /35451177 _1 _bad -science -regulation -political -economy -research -institute. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Mayhew, David (1974) Congress: The Electoral Connection. New Haven, CT: Yale University Press. Middle Class Task Force (2009) Green Jobs: A Pathway to a Strong Middle Class (Staff Report). Washington, DC: Office of the Vice President of the United States. Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry- Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. Moyle, Penny, and Katharine Parkes (1999) “The Effects of Transition Stress: A Relocation Study,” Journal of Organizational Behavior 20: 625– 46. Neumark, David, and William Wascher (2007) “Minimum Wages and Employment: A Review of Evidence from the New Minimum Wage Research.” NBER Working Paper Series Paper 12663, Cambridge, MA. Newport, Frank (2012) “Little Appetite in U.S. for More Gov’t Regulation of Business,” Gallup, 24 Sept., http://www.gallup.com/poll/157646/little -appetite -gov-regulation-business.aspx. Niemi, Richard G., et al. (1995) “State Economies and State Taxes: Do Voters Hold Governors Accountable?” American Journal of Political Science 39: 936– 57. Obama, Barack (2011a) “Executive Order 13563: Improving Regulation and Regulatory Review,” Federal Register 76: 14. ———. (2011b) “Toward a 21st- Century Regulatory System,” Wall Street J., 18 Jan., sec. A, p. 17. ———. (2012a) “Executive Order 13610: Identifying and Reducing Regulatory Burdens,” Federal Register 77: 93. ———. (2012a) “Full Transcript of Obama’s Speech on the Economy in Cleveland, Ohio,” Washington Post, 14 June, http://www.washingtonpost .com /politics/full -transcript -of -obamas -speech -on -the -economy -in -cleveland -ohio/2012/06/14/gJQAdY10cV_print.html. ———. (2012b) Executive Order 13610: Identifying and Reducing Regulatory Burdens. Federal Register, vol. 77, no. 93. Washington, DC: GPO. ———. (2013a) “Paring Back Unnecessary Regulations,” the White House, http://www.whitehouse.gov/economy/reform/paring- back- unnecessaryregulations (accessed 6 May 2013). ———. (2013b) “The 2013 State of the Union, February 12,” the White House, http://www.whitehouse.gov/state- of-the-union-2013 (accessed 10 March 2013). Office of Management and Budget (2013) 2013 Draft Report to Congress on the Benefits and Cost of Federal Regulations and Agency Compliance with the Unfunded Mandates Reform Act. Washington, DC: The White House.

The Jobs and Regulation Debate

29

O’Neill, Thomas P., and Gary Hymel (1994) All Politics Is Local and Other Rules of the Game. Holbrook, MA: Bob Adams. Palmer, Karen, et al. (1995) “Tightening Environmental Standards: The Benefit- Cost or the No- Cost Paradigm?” Journal of Economic Perspectives 9: 119–32. Pence, Michael R. (2013) “Executive Order 13- 03: Promoting Job Creation, Economic Development, and Freedom by Ordering a Moratorium on Regulations,” Executive Department, State of Indiana, http://www.in.gov/gov /2384.htm (accessed 10 March 2013). Porter, Michael E (2008) On Competition. Boston: Harvard Business School Publishing Porter, Michael E., and Claas van der Linde (1995) “Toward a New Conception of the Environment- Competitiveness Relationship,” Journal of Economic Perspectives 9: 97–118. Reid, Harry (2011) “Reid: Businesses, Economists Debunk Republican Myth of Job- Killing Regulations, November 15,” United States Senator for Nevada Harry Reid, http://www.reid.senate.gov/newsroom/pr_111511_debunkrepublicanmyth.cfm? (accessed 6 May 2013). Romney, Mitt (2012) “Remarks in Schaumburg, Illinois Following the Illinois Primary,” American Presidency Project, University of California, Santa Barbara, http://www.presidency.ucsb.edu/ws/index.php?pid=100244 (accessed 10 March 2013). Roosevelt, Franklin D. (1944) “State of the Union Message to Congress,” the White House, http://www.fdrlibrary.marist.edu/archives/address_text.html (accessed 17 June 2013). Rubio, Marco (2013) “Transcript: Marco Rubio’s Republican Response, February 13,” ABC News, http://abcnews.go.com/Politics/transcript- marcorubios- state-union-response/story?id=18484413 (accessed 10 March 2013). Ruhm, Christopher J. (2000) “Are Recessions Good for Your Health?” Quarterly Journal of Economics 115: 617– 50. Saad, Lydia (2013) “Economy Dominant in Obama’s Speech, Americans’ Priorities,” Gallup, 14 Feb., http://www.gallup.com/poll/160445/economy-domi nant-obama-speech-americans -priorities.aspx. Schmitt, John (2013) “Why Does the Minimum Wage Have No Discernible Effect on Employment?” Center for Economic and Policy Research, Washington, DC, February. Seefeldt, Kristin S. and John D. Graham (2013) America’s Poor and the Great Recession. Bloomington, IN: Indiana University Press. Shapiro, Stuart and Debra Borie-Holtz (forthcoming) The Politics of Regulatory Reform. New York: Routledge. Sinclair, Tara M., and Kathryn Vesey (2012) “Regulation, Jobs, and Economic Growth: An Empirical Analysis.” George Washington University Regulatory Studies Center Working Paper, http://research.columbian.gwu.edu/regu latorystudies /sites /default /files /u41 /032212 _sinclair _vesey _reg _ jobs _ growth.pdf. Smith, Anne E., et al. (2013) Estimating Employment Impacts of Regulations: A Review of EPA’s Methods for Its Air Rules. Report prepared by NERA Economic Consulting for the U.S. Chamber of Commerce, Washington, DC, February. Stevenson, Richard W. (1992) “Patching Up L.A.— A Corporate Blueprint,” New York Times, 9 Aug., http://www.nytimes.com/1992/08/09/business /patching-up -la-a-corporate -blueprint.html.

30

Cary Coglianese and Christopher Carrigan

Sunstein, Cass R. (1993) After the Rights Revolution: Reconceiving the Regulatory State. Cambridge, MA: Harvard University Press. Sullivan, Daniel, and Till von Wachter (2009) “Job Displacement and Mortality: An Analysis Using Administrative Data,” Quarterly J. of Economics 124: 1265– 306. Tilly, Chris (1997) “Arresting the Decline of Good Jobs in the USA?” Industrial Relations Journal 28: 269–74. Tufte, Edward R. (1978) Political Control of the Economy. Princeton, NJ: Princeton University Press. United Nations General Assembly (1948) The Universal Declaration of Human Rights (10 Dec.), http://www.un.org/en/documents/udhr/. U.S. House of Representatives, Committee on Oversight and Government Reform (2012) Continuing Oversight of Regulatory Impediments to Job Creation: Job Creators Still Buried by Red Tape. Serial No. 112-161. 112th Cong., 2d sess. Washington, DC: GPO. von Wachter, Till (2010) Long-Term Unemployment: Causes, Consequences and Solutions. Testimony before the Joint Economic Committee of the U.S. Congress, 29 April. Washington, DC. Walker, W. Reed (2011) “Environmental Regulation and Labor Reallocation: Evidence from the Clean Air Act,” American Economic Review: Papers and Proceedings 101: 442– 47. ———. (forthcoming) “The Transitional Costs of Sectoral Reallocation: Evidence from the Clean Air Act and the Workforce,” Quarterly. Journal of Economics 128, http://faculty.haas .berkeley.edu /rwalker/research /walker _transitional _costs _CAA .pdf.

Legislation Cited Red Tape Reduction and Small Business Job Creation Act, H.R. 4078, 112th Congress (2012). Regulations From the Executive in Need of Scrutiny Act of 2011, H.R. 10, 112th Congress (2011). Regulations From the Executive in Need of Scrutiny Act of 2013, H.R. 367, 113th Congress (2013). Regulations From the Executive in Need of Scrutiny Act of 2013, S. 15, 113th Congress (2013). Regulatory Accountability Act of 2011, H.R. 3010, 112th Congress (2011). Regulatory Sunset and Review Act of 2013, H.R. 309, 113th Congress (2013). Small Business Freedom of Commerce Act of 2013, H.R. 168, 113th Congress (2013).

Evidence

This page intentionally left blank

Chapter 2

Analyzing the Employment Impacts of Regulation Richard D. Morgenstern

Prior to the current economic downturn, neither the creation nor the destruction of jobs was a major concern about the effects of regulation. In contrast, the term “ job-killing regulations” has now become a rallying cry for regulatory opponents, just as the exaggerated claims of job creation have been embraced by regulatory advocates. The continuing high national unemployment rate has clearly amped up public debate on this issue and, correspondingly, the potential value of considering job impacts in Regulatory Impact Analyses (RIAs). Empirical assessment of the employment impacts of regulations is, in fact, a challenging task, due to data limitations as well as the inherent difficulty of disentangling regulatory effects from other economic changes. Although employment effects depend on the requirements of individual regulations, rules mandating installation of new equipment will likely have positive employment effects, at least in the short term. At the same time, they may impose additional long-term costs that could, in turn, lead to job losses. Unsurprisingly, a regulation with large positive net benefits may encourage the substitution of capital for labor-intensive activities, thereby leading to short- or long-term job losses. Determining the net impacts of these different factors is quite challenging. In fact, methodological and data problems abound in estimating the employment impacts of regulation. Arguably, the evidence about the impacts of environmental regulation—the area most studied for its job impacts—is quite limited on the magnitude and, in some cases, even the sign of the job impacts. Thus, a first- order question involves an assessment of our knowledge of the issue. Beyond the anecdotal, what is the nature of the evidence of regulation-induced job impacts of past rules? Do the available studies focus solely on plant-level changes in employment,

34

Richard D. Morgenstern

or do they account for job shifts to other plants or firms in the same or different industries, in both nearby and distant locations? Are regulatory job losses similar enough to the mass layoffs studied by labor economists to be treated interchangeably in evaluating regulatory impacts? A related set of questions involves the treatment of potential for job losses in RIAs. Currently, RIAs fully compliant with the relevant Office of Management and Budget (OMB) and agency guidelines effectively dodge the jobs issue by positing a full employment economy wherein few workers, on net, are displaced by regulation. Those individuals recognized to suffer job losses are assumed to quickly fi nd new ones at comparable wages in nearby firms, that is, with minimal welfare losses. In the limited instances where agencies actually estimate job losses associated with plant closure— typically in a separate Economic Impact Analysis (EIA)—the effects are treated as distributional in nature, different and separable from the assessment of industry costs, consumer welfare, and estimated benefits contained in RIAs. In effect, the potential impacts of regulation on employment are either ignored or relegated to the fringes of regulatory analysis. The current approach, which begs the question of how to evaluate competing regulatory options with the same net benefits but different job impacts, has been criticized on at least two grounds: •

•

In a period of continued high unemployment, displaced workers may not find other jobs quickly, especially if job search and retraining costs are high or if their skills reflect a large amount of firm- specific human capital that cannot be readily transferred to other applications. Even when they do find other jobs, workers who become unemployed have been found to suffer both short- and long-term income losses as well as declines in health and in life satisfaction. Arguably, if these effects could be credibly quantified and monetized, they should be considered in a cost–benefit framework, provided that doing so would not lead to double counting.

Once the full employment assumption is abandoned, there are also issues about the scope of the social costs to be considered in assessing regulation-induced job losses. Should one focus principally on lost wages and retraining costs, or is it proper to include increased alcoholism, family instability, or related impacts? Relying heavily on the labor economics literature, a recent paper by Masur and Posner (2012) developed a qualitative analysis of these broadly defined impacts, concluding that the true costs of job loss may exceed $100,000 per employed worker. Further, they argue that inclusion of such costs in RIAs could tip the balance against regulation in par ticu lar cases. Although the authors

Analyzing the Employment Impacts of Regulation

35

acknowledge that “it would be premature to use the $100,000 figure or any other figure in regulatory cost–benefit analyses,” their analysis has challenged the conventional wisdom on the proper treatment of potential job losses (or gains) in RIAs. In this chapter, I examine a range of issues associated with the regulation– jobs debate, focusing on the available empirical studies in the realm of environmental regulation. Existing research provides at best limited evidence that environmental regulation causes substantial net impacts on employment, but it is clear that the cost to individuals from job losses can be substantial. Especially in times of sustained high unemployment, the current assumption of full employment found within agency RIAs can be reasonably questioned, justifying further research so as to be able to incorporate job impacts into regulatory benefit– cost analyses.

Key Empirical Findings on Regulation- Induced Job Impacts At the outset, it is important to note that pollution abatement costs represent a relatively small fraction of total manufacturing revenues. U.S. Census data, from the Pollution Abatement Costs and Expenditures (PACE) survey, indicate that even for the most heavily regulated manufacturers, such as petroleum refining, the share of revenue devoted to pollution abatement costs reached a high of 2 percent in 1994. More typically, PACE expenditures have been in the range of 1.0–1.5 percent for pollution-intensive industries over the past 30 years. Claims that these self-reported PACE data understate true industry burdens in manufacturing industries are not broadly supported in the literature, although the evidence is somewhat mixed.1 Thus, on a prima facie basis there is not a strong case the pollution abatement costs should be a major determinant of employment levels. Turning to the direct studies on employment impacts of environmental regulations, it is useful to organize the empirical studies according to the principal methodology adopted: structural versus reduced-form modeling.2 Two widely cited papers that present a formal structural model of the underlying profit-maximizing/cost-minimizing activities of the firm are Berman and Bui (2001) and Morgenstern et al. (2002). Berman and Bui (2001) estimated the employment impacts of ozone and nitrogen oxide regulations in California’s South Coast Air Quality Management District for the years 1979–1992, a period when some of the most stringent air quality regulations were put in place. They developed a model that allows environmental regulation to operate via two separate mechanisms: the output elasticity of labor demand and the effect of pollution abatement activities on demand for variable factors of

36

Richard D. Morgenstern

production, including labor. Importantly, they demonstrated that the net effect of environmental regulation on employment predicted by this model is ambiguous on theoretical grounds. On the basis of their empirical analysis, however, the authors found that “while regulations do impose large costs they have a limited effect on employment” (Berman and Bui 2001:269). They concluded that local air quality regulation “probably increased labor demand slightly” (Berman and Bui 2001:293) but that “the employment effects of both compliance and increased stringency are fairly precisely estimated zeros, even when exit and dissuaded entry effects are included” (Berman and Bui 2001:269, emphasis added). Berman and Bui argued that the limited effects likely arose because (1) the regulations applied disproportionately to capitalintensive plants with relatively little employment, (2) the plants sold to local markets where competitors were subject to the same regulations (so that sales were relatively unaffected), and (3) abatement inputs served as complements to employment. In collaboration with Billy Pizer and Jhih- Shyang Shih, I developed a somewhat similar model to that of Berman and Bui, with the employment impacts disaggregated into three separate components (Morgenstern et al. 2002): •

•

•

Demand effect: Higher production costs raise market prices. Higher prices reduce consumption (and production), thereby reducing demand for labor within the regulated industry. As we noted, the “extent of this effect depends on the cost increase passed on to consumers as well as the demand elasticity of industry output” (Morgenstern et al. 2002:416). Cost effect: As production costs increase from added pollution abatement activities, plants use more of all inputs, including labor, to produce the same level of output. Factor- shift effect: Post-regulation technologies may be more or less labor intensive than conventional production, depending on any regulatory-induced change in the quantity of labor used per dollar of output.

Like Berman and Bui, we demonstrated that the net impact of regulation on employment is ambiguous on an ex ante basis. We used a sixyear (noncontinuous) panel data set, covering the years 1979–1991, which provided us with plant-level information on labor, capital, material, and energy inputs. We estimated our model for four highly polluting or regulated industries to examine the effect of higher abatement costs from regulation on employment, where regulation was mea sured by the self-reported PACE data collected by the U.S. Census Bureau.

Analyzing the Employment Impacts of Regulation

37

Using plant fi xed effects, we were able to account for the differences in a plant’s ability to adapt its production processes to stricter environmental requirements. We estimated demand elasticities using industry-level output and productivity data to calculate the effect of demand changes on employment. We found that increased abatement expenditures generally did not cause a significant change in employment across the four industries studied. On average, each additional $1 million of PACE spending resulted in a statistically insignificant net increase of 1.5 jobs in the regulated industries themselves. Looking at the demand, cost, and factor shift effects separately, we found positive employment effects in industries (such as petroleum and plastics) where environmental activities are labor-intensive and demand is relatively inelastic. Where the pollution abatement activities required by regulation are not labor intensive and where demand is elastic, negative effects are anticipated to occur and the (adverse) demand effect dominates. But we also found that in those industries where labor already represents a large share of production costs and where demand is relatively more elastic (such as steel and pulp and paper), there is little evidence of any statistically significant employment consequence. Given that structural models represent only part of the relevant literature, it is useful to consider papers by Henderson (1996), Becker and Henderson (1999), Greenstone (2002), and List et al. (2003), who have taken a reduced-form modeling approach. Focusing on the Clean Air Act (CAA), these studies examine the effect of attainment status, with respect to National Ambient Air Quality Standards (NAAQS), on employment growth and other metrics of economic activity. Because polluting plants already located in counties not in compliance with one or more NAAQS as well as those facilities seeking to locate in these counties are likely to face more stringent air pollution control requirements than others, the authors use this difference as a measure of incremental regulatory burden. The reduced-form studies all find some evidence that employment grows more slowly, that investment is lower, that fewer new polluting plants open, or that a combination of these effects occurs more frequently in nonattainment areas relative to attainment areas. For example, using a reduced-form model for plant-level growth, which includes plant- level, industry- by- period, and county- by- period fi xed effects, Greenstone (2002) found that in the first 15 years in which the CAA was in force (1972–1987), nonattainment counties lost 590,000 jobs, $37 billion in capital stock, and $75 billion (1987 dollars) of output in pollution- intensive industries relative to those counties that were in attainment with the CAA requirements. Importantly, although these

38

Richard D. Morgenstern

results indicate a movement of economic activity from nonattainment to attainment counties, they do not demonstrate a reduction in aggregate economic activity overall due to environmental regulation. Differences in relative growth rates across attainment and nonattainment areas are clearly an important issue, but it is not one that has generally been considered in the core benefit– cost analysis in RIAs.3 Should the changed circumstances of an auto worker who suffers a regulation-induced job loss in Detroit but is then reemployed at a comparable wage in Tennessee be counted as a job loss? The reduced-form models would count the reemployed worker as a job loss, while the structural models, which use industry-level employment as their frame of reference, would score it as no job impact from the regulation. Clearly, the individual has suffered a welfare loss by having to move 500 miles for the sole purpose of getting a comparable job. However, that is not the same welfare loss experienced by a person who loses the Detroit job and remains unemployed for an extended period. Not surprisingly, a hypothetical RIA that attempted to integrate the potential job losses into the central benefit– cost calculations would yield quite different answers depending on whether the reduced-form or structural studies were used. Arguably, reliance on the reduced-form analyses would result in overestimation of the number of jobs lost, which in turn would also lead to an overestimate of the social costs of the regulation. At the same time, the extrapolation from the structural models would suggest few or no jobs lost and hence generate an estimate of zero employment-related costs associated with regulation— clearly an underestimate— as it would exclude the transition costs for the hypothetical auto worker who regained a comparable job five hundred miles away. In addition to the actual relocation expenses, such transition costs could include lost wages (temporary or permanent) and a range of difficultto- quantify social costs, including depreciated human capital, increased alcoholism, and family instability. Beyond the different modeling approaches, some of the differences among the various studies arise from the alternative metrics used to proxy for regulation, that is, PACE expenditures versus nonattainment status. Regrettably, although each approach has its strengths and weaknesses, particularly possible mea surement error associated with PACE, there is no professional consensus on the preferred metric, from either a theoretical or empirical perspective. Research by Walker (2011) makes an important contribution to this literature by focusing on workers rather than industries as the unit of analysis. Specifically, he examined the previously unobserved economic outcomes such as unemployment and reduced labor force participation, which he labels nonemployment, as well as long-run earnings losses

Analyzing the Employment Impacts of Regulation

39

from regulation-induced job transitions, all associated with the CAA Amendments of 1990. Walker used linked worker- firm data in the United States to estimate the costs associated with workers transitioning from newly regulated industries to other sectors of the economy. Initially, he looked at industry- level pollution designations in a parallel framework to the reduced-form models. He then looked at employment in plants whose industry is considered “polluting” in nonattainment counties for the ten years after the CAA Amendments were enacted, controlling for a variety of industry and temporal fixed effects. Using panel variation induced by the 1990 CAA Amendments, Walker found that workers in newly regulated plants experienced more than $9 billion in (aggregate) foregone earnings for the years after the policy change. Most of these costs are driven by nonemployment and lower earnings in future employment, while earnings of workers who remain with their firm change little. He also noted that “relative to the estimated benefits of the 1990 CAAA, these one-time transitional costs are small . . . [although] . . . they far exceed the workforce compensation policies designed to mitigate some of these earnings losses” (Walker 2012:1). One jobs-related effect that none of the cited papers consider is the impact that regulation has on employment in industries that make substitute products. Demand for these products tends to increase as consumers respond to increased costs in the regulated industries. For example, more expensive virgin paper will cause a shift to recycled paper. The recycled paper industry, in turn, will employ more workers as sales increase. Similarly, employment in industries that are complements may decrease. Beyond these relatively direct effects, there are also indirect effects that environmental regulation may have on worker productivity. Greenstone et al. (2011) analyzed plant- level production data to estimate the effects of environmental regulations on manufacturing plants’ total factor productivity (TFP) levels. Their model, also of the reduced-form variety, distinguishes the county- level location of plants depending on pollutant- specific nonattainment and attainment categories. They found that among surviving polluting plants, a nonattainment designation is associated with a roughly 2.6 percent decline in TFP. Similarly, various studies have examined firms’ decisions to locate new plants or relocate existing plants in response to environmental regulations. The evidence is generally mixed. Using a 17-year panel, Keller and Levinson (2002) found that the stringency of environmental regulation (expressed in terms of PACE costs) has “small deterrent effects” on foreign direct investment (FDI) in states competing for such investment. In contrast, Xing and Kolstad (2002) analyzed the

40

Richard D. Morgenstern

spatial distribution of outbound U.S. FDI in six industries across 22 host countries. Using sulfur dioxide emissions as a mea sure of environmental stringency for the host countries, and treating this mea sure as endogenous, the authors found that the permissiveness of a country’s environmental regulations significantly predicted FDI originating from the United States— but only for industries emitting large amounts of pollution. Hanna (2010) mea sured the response of U.S.-based multinationals to FDI decisions to the CAA Amendments by using firm-level panel data over the period 1966–1999, controlling for firm- specific characteristics and industrial trends. Consistent with the notion that regulation causes firms to substitute foreign for domestic production, the author found that environmental regulation has led U.S.-based multinational companies to increase their foreign assets in polluting industries by 5.3 percent and their foreign output by 9 percent over the interval of the study. Further, Hanna found that these results were more robust for firms that manufactured within an industry for which imports had historically accounted for a large percentage of U.S. consumption. Brunnermeier and Levinson (2004), also using panel data, found statistically significant pollution haven effects of “reasonable magnitude.” Levinson and Taylor’s (2008) results for trade flows and environmental regulation are consistent with the other studies. In reviewing the studies directly focused on regulation-induced job loss, it is useful to first consider the differences between the structural and reduced-form estimates: •

•

•

The universe of included facilities differs somewhat across studies. All the reduced-form analyses are based on estimates derived from the entire manufacturing sector. In contrast, both Berman and Bui (2004) and Morgenstern et al. (2002) examined a small number of industries. Although both PACE and nonattainment metrics are imperfect proxies for regulatory burdens, it is possible that mea surement error is greater for PACE. If so, that could explain some of the differences between Morgenstern et al. (2002) and the structural estimates. However, the PACE– nonattainment distinction would not explain the Berman and Bui (2004) findings of negligible employment impacts, as they also analyzed effects of spatial differences in regulatory requirements. It is possible that misspecification of the cost models used in the structural analyses contributes to the differences in the findings. However, all the structural papers use relatively standard, well-tested

Analyzing the Employment Impacts of Regulation

•

41

cost functions, which makes this an unlikely, albeit not impossible, explanation for the differences. As Masur and Posner (2012) suggest, it is also possible that despite the failure to explicitly consider employment effects in the RIAs, environmental regulators actually craft their rules to avoid adverse employment impacts. Although this could, in principle, explain the results of the structural models, it seems unlikely that government’s “invisible hand” could consistently produce such results in the absence of any credible analysis of the impacts.

Notwithstanding the possible explanations for the differences among the papers, how can one summarize the preponderance of the evidence? In terms of direct, mea surable impact on employment at the plant level, it appears there is only limited evidence that environmental regulations lead to significant job loss. Beyond the direct analyses, there is evidence of adverse economic impacts of regulation based on metrics such as TFP and FDI. Clearly, over the long term, these latter impacts are potentially quite important.

The Labor Economics Literature on Mass Layoffs Compared with the limited empirical literature assessing the number of jobs lost (or gained) as a result of environmental regulations, the labor economics studies examining the impacts on workers of mass layoffs, especially at times of high national unemployment, are more robust, with more consistent results. At the same time, questions of relevance to issues of regulation-induced job loss plague the use of the mass layoff metric for studying such issues. This section reviews the recent evidence cited by Masur and Posner (2012) and others on the pecuniary and nonpecuniary impacts of mass layoffs, with particular emphasis on labor market conditions at the time of the initial unemployment. Conceptually, the categories of social cost of regulation-induced job loss are straightforward: the lost wages during the spell of unemployment (minus leisure benefits), relevant search and retraining costs, any long-term reduction in wages or diminished quality of working conditions in subsequent jobs, and a series of nonpecuniary losses, such as increased mortality, family strife, and related outcomes. Building on research by von Wachter et al. (2009), Davis and von Wachter (2011) explore the connection between economic downturns and permanent layoffs, including impacts on those employees with substantial tenures at their prior employers. Using Social Security and other records linked by the Employer Identification Number (EIN),

42

Richard D. Morgenstern

covering up to thirty years of earnings, they estimate the cumulative losses associated with mass layoffs and review the evidence indicating that job displacement leads to employment instability, health impairment, and other adverse consequences. According to Davis and von Wachter (2011), an employer must meet the following criteria for job losses to qualify as a mass layoff event in year y: • • • •

50 or more employees in y − 2; Employment shrinks by 30– 99 percent from y − 2 to y; Employment in year y − 2 is no more than 130 percent of y − 3; Employment in year y + 1 is less than 90 percent of employment in y − 2.

By allowing employment to shrink by no more than 99 percent, Davis and von Wachter avoid misclassifying as mass layoffs those cases where the company actually ceased operating. At the same time, the authors use the last two conditions to eliminate from the analysis transitory shifts in firm employment. Davis and von Wachter found that the present value of lifetime earnings fell by 12 percent for males let go after being with larger- sized companies for three or more years. Because of their displacement, men under age fi fty lost an estimated $77,557 (in 2000 dollars) computed over a twenty-year period using a 5 percent annual discount rate. The estimated losses were even more substantial for men with longer job tenure and smaller for women.4 Importantly, Davis and von Wachter also determined that the lost earnings of displaced workers increased significantly as the national unemployment rate rose. When the unemployment rate was below 6 percent, the earnings loss from displacement averaged only 1.4 years of wages. In contrast, if the unemployment rate was above 8 percent, the average earnings loss doubled, to 2.8 years of pre- displacement earnings. Overall, the Davis and von Wachter analysis indicates that a strong labor market can substantially buoy the future earnings potential—both in the medium- and long-term—of displaced workers. Moreover, since jobs are easier to find in tight labor markets, such conditions also increase the probability that recently unemployed workers will be back to work quickly. The appropriate categorization of these losses as true economic costs versus transfers is somewhat of an open issue. The actual harm may be less if, for example, some fraction of the lost wages represent abovemarket payments that may subsequently be captured by consumers or shareholders. Further, to the extent that firms and workers anticipate

Analyzing the Employment Impacts of Regulation

43

layoffs and firms pay workers wage premiums that compensate them for this risk, then the earnings decline will overstate the welfare loss. Although unemployment insurance has partially displaced compensating differentials, the latter effect may be quite modest. Empirically, of course, it is difficult to distinguish among these various situations. Drawing on data from the General Social Survey and Gallup polling, Davis and von Wachter (2011) also examined the relationship between anxieties about unemployment, cuts in wages, ease of job finding, and other labor market concerns, and actual labor market conditions. Unsurprisingly, the evidence they present indicates that such concerns track the state of affairs in the labor market quite closely and, moreover, that workers swiftly respond to worsening economic conditions. Specifically, using Gallup poll data, Davis and von Wachter show how workers’ fears about losing their jobs intensified after the severity of the financial crisis reached its climax in 2008 and 2009. 5 In sum, Davis and von Wachter’s results indicate that, for a large portion of the workforce, an important cost of an economic slump is the anxiety that such an event produces. Workers understand acutely the costs and implications of being laid off, particularly during declines. A paper by Sullivan and von Wachter (2009) used administrative data on quarterly employment and earnings matched to death records to estimate the effects of job displacement on mortality. They followed the incidence of mortality for the displaced workers up to twenty years after job loss and compared displaced workers’ mortality rates with those of similar workers who did not lose their job, thus allowing estimates of both the short- and long-term response of mortality rates to job displacement. Overall, they found that job displacement leads to a 15–20 percent increase in death rates during the twenty years following displacement. Although the results are not linked to a causal mechanism, they are robust to extensive controls for sorting and selection and are roughly consistent with other estimates of the effects of job loss on mortality. Translating the death rate increase into life expectancy, Sullivan and von Wachter show that a worker laid off from his job at age 40 can be expected to live 1.0–1.5 fewer years, assuming the death rate increase is sustained over the worker’s lifetime. Furthermore, the effect on expected mortality is heightened for those faced with larger earnings losses. Notably, the authors are able to rule out various selection effects, including the possibilities that the results are driven by displacement of particularly unhealthy employees or among workers exposed to relatively poor working conditions. In contrast, looking at outcomes across U.S. states, Ruhm (2000) reported that mortality declines during economic downturns, perhaps because workers have more time to invest in their health, reduce risky

44

Richard D. Morgenstern

health behaviors such as smoking, and experience fewer automobile and work-related accidents. Specifically, he found declines for eight of the 10 causes of death examined— one notable exception being suicide— and the decreases were most substantial for the causes and among those age groups where shifts would seem most reasonable theoretically. The results further show that the adverse health impacts of expansions can be negated if the upturn extends for a significant period of time. Further, through a review of micro- level data, Ruhm reports that physical activity diminishes, diet worsens, and incidence of smoking and obesity rise during periods of relative economic strength. Overall, this suggests that job loss could have some beneficial effect by reducing accidents and possibly risky health behavior. Comparing the Ruhm (2000) and the Sullivan and von Wachter (2009) papers, there are a number of important differences. First, the type of event analyzed by Sullivan and von Wachter is of a different nature than the aggregate variation used in Ruhm. Only a fraction of reallocation of employment in recessions occurs through job separations and mass layoffs, and these are often concentrated in par ticu lar regions, sectors, or groups of workers. Second, Sullivan and von Wachter analyzed effects associated with earnings losses exceeding what is typically observed in an economic downturn or in the career of an average worker. Thus, the beneficial effects from a lower rate of accidents or improved health behavior in a recession may be more than offset by the large negative effects of job loss on workers’ careers in the Sullivan and von Wachter sample. Turning to distributional and other impacts, a study of job loss in California during the 1990s by von Wachter and Handwerker (2009) estimated the relationship between education and job loss based on both survey data and administrative records. They found the relationship to be saucer- shaped in that workers who had earned a college degree did fare better than workers lacking even a high school degree—but workers with a high school degree or some college fared worst of all. Their findings suggest that more education can help workers recover from the earnings effects of losing a job, but mainly just for those who earn a college degree. Further, the authors found that, apart from education, losing a job is associated with a reduction in owning a home as well as being covered under private health insurance, and that these effects are probably at least in part explained by displaced workers’ longterm earnings loss. Adler (this volume) and Masur and Posner (2012) report a number of studies that examine the effect of various job-related events on people’s “subjective well-being,” defined as the score that they give themselves

Analyzing the Employment Impacts of Regulation

45

when asked by interviewers how happy they are. These studies uniformly show that unemployment reduces people’s happiness, even when controlling for lost earnings. Studies that considered the variation of impacts across different demographic groups and industry characteristics show, unsurprisingly, that the burdens of unemployment vary according to various worker characteristics, including age, sex, region, education, and experience. Based on this brief review, there appears to be fairly strong evidence that unemployment or involuntary declines in labor force participation, that is, job displacement, can have major adverse consequences on the affected individuals, including lost wages during the initial displacement, long-term wage loss, and a range of health and other nonpecuniary impacts. Importantly, the magnitude of these losses is positively related to the overall state of the economy, and the burdens appear to fall disproportionately on less educated, lower income workers.

Discussion A key issue for the present chapter concerns the extent to which the extensive labor economics literature on mass layoffs is directly applicable to the analysis of job loss due to environmental regulation. If the two can be readily linked, then the calculation of net welfare impacts would be relatively straightforward, and the estimated effects would be ripe for inclusion in RIAs. Upon inspection, however, it appears there are a number of challenges in making such a linkage. First, there is the difference in the definition of job loss used by environmental regulators and the one adopted in the labor economics literature. When estimated in EIAs, the term “ job losses” typically refers to estimates from model plant analyses indicating that one or more plants (or production lines within a plant) will become uneconomic as a result of the new rule. These EIAs generally do not address the possibility that some or all of the lost output associated with the closure will be picked up by other domestic plants. In contrast, the term “ job displacement,” as used by Davis and von Wachter (2011) and other labor economists, refers to individuals separated from large employers in mass layoffs during recessions. The mea surement of this job displacement is made via the EIN. As is well known, however, the same EIN may apply to multiple establishments, especially when the firm is a large entity. Thus, a worker who moves from one facility to another owned by the same firm would not be categorized as displaced according to this definition.

46

Richard D. Morgenstern

This distinction is critical, as there is abundant anecdotal evidence that a large proportion of workers “displaced” by environmental regulation move to other jobs in the same plant or firm. Although I am unaware of comprehensive data on job retention in these circumstances, the results from Berman and Bui (2001) as well as from Morgenstern et al. (2002) indicate that there is a high retention rate at the industry level. This exclusion of workers who are reemployed at the same plant or firm is a major barrier to the direct application of the findings from the labor economics literature to the issue of regulatory induced job loss. As an example, assume that the Environmental Protection Agency (EPA) estimates in an EIA that x workers are “displaced” as a result of a particular regulation. It is certainly plausible that some or all of these workers would be reemployed at the same plant or at least at the same firm, which was now expanding output elsewhere to make up for the reduced production at the closed facility. In such a case, the amount of meaningful job displacement would be negligible. By contrast, the labor economics studies focus exclusively on the individuals who separate completely from their long-term employer, as defined by the EIN, and move to another similarly paying job in a different firm, to a lower paying job in another firm, or possibly to no job at all (that is, they are unemployed or out of the labor force). If the labor economics studies were able to capture all the job movements, including those within the same firm, it would be plausible to use these studies to estimate the welfare losses associated with regulation. However, in the more likely case that the labor studies account for only a portion of the individuals identified by EPA as being displaced by regulation, use of the results from those studies in RIAs may seriously overstate the adverse impacts.

Conclusion Assessing the employment impacts of environmental regulation presents many challenges, due both to data limitations and to the difficulty of disentangling regulatory from other changes occurring at the same time, including the possibility of reorganization of production processes in response to the new regulations. Reviewing the studies directly addressing the regulation– employment nexus, there is only limited evidence of adverse national-level impacts, after netting out the within- country movements of jobs away from more stringently regulated geographic areas. The indirect studies, particularly those examining changes in FDI in response to domestic regulation, provide stronger evidence of adverse impacts, albeit at modest levels.

Analyzing the Employment Impacts of Regulation

47

As Masur and Posner (2012) have correctly noted, and as the labor economics literature fully supports, the cost of job loss is substantial, particularly in difficult economic times. The failure to take this into account in RIAs is clearly a legitimate issue. However, as with other difficult-to-measure benefits and costs that might be included in RIAs, one has to be cautious about getting ahead of the data and analytic support for quantifying such effects. Notwithstanding the evidence presented herein, we appear to have only a weak basis for quantifying or monetizing these effects. Interestingly, a recent paper by Bartik (2012) examined the quantitative importance of including such estimates in the benefit– cost analyses for the same set of 16 regulations referenced by Masur and Posner (2012). Bartik found that the inclusion of the social cost of the jobs lost would change overall present value costs of the rules examined from −0.7 percent to +166.5 percent. Across all 16 rules, the average impact, weighted by the present value of the estimated costs, is 1.6 percent. The cases where the inclusion of job loss has the largest percentage impacts are in the area of habitat protection. The largest impacts for EPA rules are 12.6 percent, in the case of industrial boiler air pollution standards. Thus, Bartik’s analysis does not support the contention that inclusion in RIAs of the social costs of job loss would tip the balance against regulation in very many cases. At the same time, there are various policy and research initiatives that could be undertaken in the near term. On the policy side, since the evidence on job movement from nonattainment to attainment counties is fairly strong, it may make sense to try to monetize the welfare losses associated with such movement. Although I am not aware of specific studies that attempt such estimates, they would seem to be quite tractable. On the research side, it is clear from the labor economics literature that those individuals subject to mass layoffs, defined as separation from their employer (that is, individuals who will no longer work for an employer with the same EIN), do indeed suffer both pecuniary and nonpecuniary losses. What is not known with any precision is the fate of workers estimated by EPA to be displaced by new rules—a critical component of any estimate of increased regulatory burdens. In particular, how many of these workers are reemployed at the same plant or firm? Presumably, this too is a tractable question, although probably more difficult to analyze than the more straightforward assessment of burdens imposed on workers who move across counties. In sum, the full employment assumption currently used in RIAs is not fully defensible, especially at times of high national unemployment. At the same time, additional research is required to develop credible

48

Richard D. Morgenstern

job displacement-related cost estimates suitable for inclusion in the core cost–benefit calculations of RIAs.

Acknowledgments I have benefited from helpful conversations with Adam Finkel, Arthur Fraas, Randall Lutter, and Ronald Shadbegian.

Notes 1. Morgenstern et al. (2001) and Gray and Shadbegian (2005) provide findings of little or no significant impacts on nonabatement activities. Gray and Shadbegian (1998) found some adverse impacts in the pulp and paper industry, including both environmental investment crowding out capital investment within a plant and firms shifting investment toward facilities facing less stringent abatement requirements. 2. Structural models are generally based on an underlying utility or profit function that is used to generate the estimating equation. Reduced-form models are not explicitly based on such functions. 3. Some or all of those 590,000 individuals who lost jobs in the nonattainment areas found work elsewhere, and some or all of the capital and output losses were undoubtedly offset at least partly by increases elsewhere in the country. The reduced-form models, however, are unable to provide estimates of such effects. 4. The earlier work by von Wachter et al. (2009), also based on Social Security records covering up to 30 years of earnings, confirms the larger range of estimates from previous studies based on single U.S. states and selected samples of workers. After accounting for rehiring of some part of the cohort, they found that workers losing their jobs as part of mass layoffs suffer an immediate wage loss of 30 percent. Even 20 years after the job loss, wage loss is still 20 percent. Using alternative specifications, including industry-year or firm-year effects, these results also hold for workers with weak prior job attachment and are strong and long-lasting for all age and industry groups examined. The authors calculated that over 20 years the average loss for a worker in their data set ranged from $110,000 to $140,000. 5. For similar findings, see Burgard et al. (2012).

References Bartik, Timothy J. (2012) “The Social Cost of Potential Job Losses Due to Environmental Regulations: How Job Losses’ Social Costs Compare to Lost Earnings and Overall Social Costs of Regulations.” W. E. Upjohn Institute for Employment Research, Kalamazoo, MI. Becker Randy A., and J. Vernon Henderson (1999) “Costs of Air Quality Regulation.” National Bureau of Economic Research Working Paper 7308, Cambridge, MA. Berman, Eli, and Linda T. Bui (2001) “Environmental Regulation and Productivity: Evidence from Oil Refineries,” Review of Economics and Statistics 83: 498– 510.

Analyzing the Employment Impacts of Regulation

49

Brunnermeier, Smita G., and Arik Levinson (2004) “Examining the Evidence on Environmental Regulations and Industry Location,” Journal of Environment and Development 13: 6– 41. Burgard, Sarah A., et al. (2012) “Perceived Job Insecurity and Health: The Michigan Recession and Recovery Study,” Journal of Occupational and Environmental Medicine: 1101– 6. Davis, Steven, and Till M. von Wachter (2011) “Recessions and the Cost of Job Loss.” National Bureau of Economic Research Working Paper 17638, Cambridge, MA, http://www.nber.org/papers/w17638.pdf (accessed 17 January 2013). Gray, Wayne, and Ronald Shadbegian (1998) “Environmental Regulation, Investment Timing, and Technology Choice,” Journal of Industrial Economics 46: 235– 56. ———. (2005) “Assessing Multidimensional Per for mance: Environmental and Economic Outcomes.” U.S. Census Bureau Center for Economic Studies Working Paper 05- 03, Washington, DC. Greenstone, Michael (2002) “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures,” Journal of Political Economy 110: 1175–219. Greenstone, Michael, et al. (2011) “The Effects of Environmental Regulation on the Competitiveness of U.S. Manufacturing.” U.S. Census Bureau Center for Economic Studies Paper CES-WP-11- 03, Washington, DC. Hanna, Rema (2010) “U.S. Environmental Regulation and FDI: Evidence from a Panel of U.S.- Based Multinational Firms,” American Economic Journal: Applied Economics 2: 158– 89. Henderson, J. Vernon (1996) “Effects of Air Quality Regulation,” American Economic Review 86: 789– 813. Keller, Wolfgang, and Arik Levinson (2002) “Pollution Abatement Costs and Foreign Direct Investment Inflows to U.S. States,” Review of Economics and Statistics 84: 691–703. Levinson, Arik, and M. Scott Taylor (2008) “Unmasking the Pollution Haven Effect,” International Economic Review 49: 223– 54. List, John A., et al. (2003) “Effects of Environmental Regulation on Manufacturing Plant Births: Evidence from a Propensity Score Matching Estimator,” Review of Economics and Statistics 85: 944– 52. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Morgenstern, Richard, et al. (2001) “The Cost of Environmental Protection,” Review of Economics and Statistics 83: 732–38. Morgenstern, Richard, et al. (2002) “Jobs vs. the Environment: An IndustryLevel Analysis,” Journal of Environmental Economics and Management 43: 412– 36. Ruhm, Christopher J. (2000) “Are Recessions Good for Your Health?” Quarterly Journal of Economics 115: 617– 50. Sullivan, Daniel, and Till von Wachter (2009) “Job Displacement and Mortality: An Analysis Using Administrative Data,” Quarterly Journal of Economics 124: 1265–1306. von Wachter, Till, and Elizabeth Weber Handwerker (2009) “Variation in the Cost of Job Loss by Worker Skill: Evidence Using Matched Data from California, 1991–2000.” Columbia University Working Paper, New York.

50

Richard D. Morgenstern

von Wachter, Till, et al. (2009) “Long-Term Earnings Losses due to Mass Layoffs During the 1982 Recession: An Analysis Using U.S. Administrative Data from 1974 to 2004.” Unpublished manuscript, Columbia University Economics Department, New York. Walker, W. Reed (2011) “Environmental Regulation and Labor Reallocation: Evidence from the Clean Air Act,” American Economic Review: Papers and Proceedings 101: 442– 47. ———. (2012) “The Transitional Costs of Sectoral Reallocation: Evidence from the Clean Air Act and the Workforce,” http://faculty.haas.berkeley.edu /rwalker/research/walker_transitional_costs_CAA.pdf (accessed 3 March 2013). Xing, Yuquing, and Charles Kolstad (2002) “Do Lax Environmental Regulations Attract Foreign Investment?” Environmental and Resource Economics 21: 1–22.

Chapter 3

Do the Job Effects of Regulation Differ with the Competitive Environment? Wayne B. Gray and Ronald J. Shadbegian

Prior to 1970, environmental regulation was primarily the responsibility of state and local agencies—for the most part with limited enforcement activity. After the formation of the U.S. Environmental Protection Agency (EPA) in the early 1970s and the passage of the Clean Air Act and Clean Water Act, the federal government replaced state and local agencies by taking over the lead role in regulating environmental quality, imposing more stringent regulations with correspondingly greater levels of enforcement. Since that time, the federal government has continually promulgated rules requiring U.S. manufacturing facilities to further decrease their emission levels. This increasing stringency of environmental regulation has caused U.S. manufacturing facilities to increase their level of pollution abatement expenditures. Even though environmental regulations have become more stringent over time, U.S. manufacturing facilities have faced only a moderate increase in their pollution abatement expenditures. U.S. Census data indicate that pollution abatement expenditures increased from roughly 0.3 percent of total manufacturing shipments in 1973 to only 0.4 percent in 2005. Even though pollution abatement expenditures are an extremely small percentage of the manufacturing sector’s total operating costs (even for the most highly regulated manufacturing industries), the widely shared assertion is that environmental regulations place a large burden on the manufacturing sector, contributing to large reductions in employment. The stock argument for this effect is that environmental regulations increase production costs, which raise prices and therefore decrease the demand for output, thus reducing employment (at least in a competitive market). This is the demand effect from Morgenstern et al. (2002) and discussed further in the preceding chapter by Morgenstern. It is also

52

Wayne B. Gray and Ronald J. Shadbegian

possible that more stringent environmental regulations may encourage manufacturing plants to adopt more efficient production technologies that are capital-intensive, thereby reducing employment. Although this effect might appear obvious at first, a detailed microeconomic analysis shows that it is not certain. Even if it is true that environmental regulation reduces output in the regulated industry, abating pollution may well require additional labor (for example, to operate and maintain pollution abatement capital). Furthermore, it is not impossible for pollution abatement technologies to be labor increasing (Berman and Bui 2001; Morgenstern et al. 2002). A standard benefit– cost analysis does not typically include a separate examination of employment and output impacts.1 However, these impacts are of particular concern now, given the current sustained levels of high unemployment. Executive Order 13563 stated recently that “our regulatory system must protect public health, welfare, safety, and our environment while promoting economic growth, innovation, competitiveness, and job creation” (Obama 2011:3821). It is understandable that policymakers are concerned that new, more stringent environmental regulations could lead to more unemployment, and hence it is important to test whether these concerns are justified. In this chapter, we analyze the impact of environmental regulation on employment in U.S. manufacturing.2 In particular, we examine how the effects of environmental regulatory pressure on an industry can differ, depending on the competitive environment in which the industry is operating. Recent EPA regulatory impact analyses (RIAs) have relied on Morgenstern et al. (2002) to estimate the impact environmental regulations have on employment in the regulated sector. 3 One of the limitations of Morgenstern et al. (2002) is that their analysis did not directly take into consideration the competitive environment in which the regulated plants are operating. Our study attempts to fi ll that gap. We systematically examine three mea sures of the competitive environment faced by an industry, each of which is expected to influence the extent to which environmental regulations affect employment. Import competition has been a key political concern for many years, raising the possibility of declining international competitiveness and increased imports, if increases in stringency for domestic firms are not matched with comparable increases in stringency for foreign producers. The extent of competition within the domestic industry may also influence the impact of regulation: if there are few producers, it may be easier for companies to raise prices in response to higher costs. Finally, the state of product demand in an industry may influence the impact of regulation. Growing demand makes it easier to raise prices to cover

Job Effects and the Competitive Environment

53

higher costs, while declining demand may squeeze some producers out of the market. Our analysis will relate differences in regulatory pressure across industries and over time to examine how industry employment and output vary as regulatory pressure varies. Doing the analysis at the industry level allows us to rely on publicly available data, so that other researchers and government analysts will have easy access to the necessary data to apply similar tools to their areas of interest. Working with industrylevel data also permits us to include a wider range of industries in the analysis with more complete coverage, rather than focusing on preparing detailed data for a plant- level analysis in one or a few industries. Finally, our primary interest is on the variation in the competitive environment at the industry level, which will help to inform EPA analyses, rather than plant-level variation. The results of this analysis will help answer a number of important questions that arise during the process of developing new environmental regulations: Is there any evidence that higher levels of regulation lead to losses in employment? Is there any evidence for different regulatory impacts under different competitive environments? Is competition from foreign imports more or less important than competition within the domestic industry? Does declining product demand in an industry increase the regulatory impact? This last question could be considered particularly relevant to the recent policy debate about whether to delay the creation of new regulations during recessions, although most new regulations do not become effective until 3 to 10 years after they are promulgated. Because we rely on publicly available measures of the industry’s competitive environment, regulatory agencies and policymakers could apply these results to predict which industries (if any) would be more sensitive to new regulations. One question that is beyond the scope of our chapter is what effect environmental regulation has on overall net employment in the U.S. economy or even just the manufacturing sector. It may be true that regulations impose costs on polluting plants and that these costs may affect plants’ competitive position and cause the loss of some jobs. However, at the same time, environmental regulations may also create jobs in other industries within the manufacturing sector (for example, firms producing scrubbers), as well as outside the manufacturing sector (for example, firms providing environmental consulting ser vices).4 Thus, calculating the net employment effect of regulation would require estimating job gains as well as job losses due to the regulation. In a fullemployment economy, we would expect that the number of jobs created by additional abatement spending would approximately equal the number of jobs lost in the regulated manufacturing industries.

54

Wayne B. Gray and Ronald J. Shadbegian

We perform our analysis using data from 1973 to 1994 at the industry level for U.S. manufacturing industries and consider the impact of regulation on both industry employment and output. Our analysis shows that higher levels of regulation lead to statistically significant but very small job losses. In particular, we find that a 10 percent increase in overall abatement costs relative to output (our mea sure of regulatory stringency) is associated with a loss of only 30 jobs in the average industry with 40,000 employees— so the overall effect of pollution abatement costs on employment in a typical industry is very small. We also find, as expected, that industries with more rapidly growing demand tend to have a smaller impact of regulation. Conversely, we unexpectedly find that employment in industries with higher concentration ratios seems to be more sensitive to regulation, while the sensitivity of employment to regulation does not seem to be significantly affected by an industry’s level of imports.

Environmental Regulation and Employment The question of how environmental regulations affect various economic mea sures of interest, including productivity, investment, and environmental per for mance, has been well studied, but very few papers specifically investigate the effect of environmental regulations on employment.5 Berman and Bui (2001) assembled a unique plant-level data set to estimate the effect air pollution regulations had on labor demand in the Los Angeles, CA, area— or the South Coast Air Quality Management District (SCAQMD). The data set they developed contains in- depth information on all the changes in environmental regulation, including adoption date, compliance date, date of increase in stringency, and the regulated pollutant for all the affected manufacturing plants in the SCAQMD. In their study, Berman and Bui found that new, more stringent air quality regulations promulgated between 1979 and 1992 did not decrease the demand for labor in Los Angeles but may have actually increased it by a small amount. Cole and Elliot (2007) estimated a similar model to that of Berman and Bui (2001) by using a panel data set consisting of 27 industries from the United Kingdom. They also found that environmental regulation had no statistically significant effect on employment. Morgenstern et al. (2002) estimated the impact of pollution abatement spending (their proxy for environmental regulation) on employment for four highly polluting and highly regulated sectors (pulp and paper, plastic, petroleum refining, and steel). Their empirical results also indicate that more stringent levels of regulation do not cause a significant change in employment. For example, in the steel and pulp and

Job Effects and the Competitive Environment

55

paper industries, additional abatement spending led to a statistically positive but quite small net increase in employment. More recently, Gray et al. (2013) analyzed how EPA’s Cluster Rule affected employment in the pulp and paper industry. Again their results show that more stringent regulations have only a small, generally insignificant impact on employment. Greenstone (2002) used a difference-in- differences model to examine the effect on employment of a county being designated by the EPA as out of attainment with respect to one or more of the criteria pollutants. Plants in counties that are designated nonattainment face stricter environmental regulations than those in counties that are in attainment. Greenstone found that nonattainment counties (relative to attainment ones) lost roughly 600,000 jobs during the 1972–1987 time period.6 To summarize, most research using plant- level data has not found large negative impacts of stricter environmental regulation on labor demand. Most studies have found no statistical effect, and some have even found small positive effects of regulation on employment.

A Model of Regulation’s Effect on Employment The popular belief is that environmental regulation reduces employment, because such regulation leads to increases in production costs, thereby reducing output. However, microeconomic analysis is not clear on this point. Although it is certainly possible for environmental regulation to reduce output, it is also possible for pollution abatement technologies to be labor enhancing. Therefore, we need a model that allows environmental regulation to impact the demand for labor through two channels: the output elasticity of labor demand and the marginal rate of technical substitution between labor demand and pollution abatement activity. Berman and Bui (2001) have derived such a model of regulation and labor demand with these two features based on the partial static equilibrium model (PSEM) of Brown and Christensen (1981). The PSEM of Brown and Christensen permits the levels of some “quasifi xed” factors to be set by exogenous constraints, instead of purely by cost minimization.7 In our case, we treat costs involved in pollution abatement, both capital and operating costs, as well as environmental regulatory variables as “quasi-fixed.” All other “productive” inputs are treated as variable factors. Assume that a perfectly competitive firm minimizes costs by choosing levels of the J variable inputs and K “quasi-fixed” inputs.8 We can write the variable cost function as follows: (1)

VC = F(Y, P 1, . . . , PJ , Z1 . . . , ZK).

56

Wayne B. Gray and Ronald J. Shadbegian

where Y is output, Pj are the prices of the variable factors, and Zk are the levels of the “quasi-fi xed” inputs. Applying Shephard’s lemma yields the following set of variable input factor demands as a function of output, prices, and the level of the “quasi-fixed” inputs: J

K

(2)

L = α + ρ yY +

∑

β kZ k +

∑γ P . j

j

j =1

k =1

The direct effect of regulation on L is (3)

dL dY = ρy + dR dR

K

∑ k =1

βk

dZ k + dR

J

∑γ j =1

dP j j

dR

= μ.

The first term in equation (3) captures how regulation affects the demand for labor via its effect on output. This “output” effect of environmental regulation is generally thought to be negative, though as Berman and Bui (2001) note, microeconomic theory gives no clear definite prediction. If compliance is accomplished by investment in abatement capital that decreases marginal costs, dY/dR could be positive. The second term captures the effect of regulation on labor demand via its effect on the demand for quasi-fi xed abatement activities, Zk , and the marginal rates of technical substitution between pollution abatement and labor. The change in demand for pollution abatement activity resulting from an increase in the stringency of regulation, dZk/dR, has to be positive. The signs of the βk coefficients are not known a priori—they depend on whether labor and pollution abatement activity are substitutes or complements. That is the primary reason that the sign of μ, the employment effect of regulation, cannot be predicted from theory alone. Finally, if input markets are competitive and the regulated industry is a small player in those markets, a change in regulatory stringency for the industry will not affect the price of its inputs, so the final term in equation (3) will drop out. Because of data limitations, Berman and Bui (2001) estimated the effect air pollution regulations have on labor demand in the SCAQMD between 1979 and 1992 with the following reduced- form version of equation (3): (4)

L = δ + μR.

We estimate an equation similar to Berman and Bui (2001) but augment (4) with several industry characteristics. These include our three mea sures of the industry’s competitive environment, which are de-

Job Effects and the Competitive Environment

57

scribed in more detail below: import competition, internal competition, and product demand. Finally, we also include interactions between each measure of the industry’s competitive environment and the pollution abatement cost mea sure. These interaction terms are the focus of our analysis, because they enable us to statistically test whether the impact of regulatory pressure on output and employment differs significantly across industries in ways that are related to our mea sures of the industry’s competitive environment.

Data To conduct our analysis, we need information on industry employment and output, a mea sure of environmental regulatory pressure, and various other industry characteristics. Our mea sures of industry economic activity are taken from the National Bureau of Economic Research and the U.S. Census Bureau’s Center for Economic Studies (NBER- CES) Manufacturing Industry Productivity database. This has been regularly updated since the 1970s and now contains annual data for 1958–2009.9 These data are based on the Census Bureau’s Annual Survey of Manufactures and include a variety of measures of industry activity. In our analysis, we focus on total employment in the industry and the total value of industry shipments as our dependent variables— expressed in log form as LN_EMPLOYMENT and LN_SHIPMENTS, respectively. To mea sure regulatory pressure, we rely on data from the Census’s Pollution Abatement Costs and Expenditures (PACE) survey, which was conducted annually between 1973 and 1994.10 Since 1994, the survey has been conducted on only two occasions: in 1999 and in 2005. The infrequent nature of these data and the substantially different questionnaire used in 1999 made it impractical to extend our analysis to these more recent years (Becker and Shadbegian 2005). The PACE survey measures both capital expenditures and operating costs for pollution abatement expenditures, and it provides data at the four- digit Standard Industrial Classification industry level for manufacturing industries. Our primary measure of the amount of regulatory pressure being directed toward an industry in a given year is ABATEMENT COST RATIO, the ratio of pollution abatement operating costs to the total value of shipments from the industry.11 This mea sure has been used frequently in prior research and incorporates the impact of past capital investments because depreciation of existing pollution abatement capital is included in the operating costs. Measures of the competitive environment faced by the industry come from a variety of sources. Import competition is measured by the ratio of imports to domestic production for that industry’s products. Annual

58

Wayne B. Gray and Ronald J. Shadbegian

data on imports at the four- digit industry level were taken from the NBER Trade Database initially prepared by Feenstra (1996) and more recently extended by Peter Schott (2008, 2010). These imports were then divided by the shipments from the domestic industry to form IMPORT RATIO. Our mea sure of the domestic competition within the industry is measured by the industry’s concentration ratio, specifically the share of total industry output accounted for by the eight largest firms. These data are prepared at the four- digit industry level by the Census Bureau every five years, based on the data collected during the Census of Manufactures. During the 1992 data collection, the Census Bureau prepared a collection of these concentration mea sures extending from 1947 to 1992 (U.S. Census Bureau 1992). The concentration ratios were interpolated between the census years (with the 1992 number being used for 1993–1994), generating the CONCENTRATION RATIO variable. Product demand is mea sured by combining an input– output matrix with a mea sure of industry activity, including final demand for the product by consumers. By seeing whether the purchasers of an industry’s output are themselves expanding or contracting, and by how much, we can predict whether overall demand for that industry’s output is rising or falling. We use the 1982 benchmark input– output matrix for the economy prepared by the Bureau of Economic Analysis (Bureau of Economic Analysis 1991). Purchases of the industry’s output by other manufacturing industries are linked to the growth rates of those other manufacturing industries (calculated from the NBER- CES database), while purchases by nonmanufacturing sectors (and purchases in final demand) are linked to the growth rates of those sectors, taken from BEA Industry Accounts data (Bureau of Economic Analysis 2004). This calculation results in DEMAND GROWTH, our mea sure of overall demand for the industry’s output. In addition to these mea sures of industry competitive pressures, we also include three other industry characteristics that might have an impact on industry employment, all taken from the NBER- CES data set. Two price indexes are included (in logarithmic form). LN_ENERGY PRICE is an index of the price of energy faced by the industry (based on the mix of different energy types in the industry’s energy consumption, applied to price indexes for those energy types). LN_MATERIALS PRICE is an index of the price of materials faced by the industry (based on data from input– output tables on the industry’s consumption of different material inputs, applied to price indexes for those materials). There is also a measure of the current intensity of capital spending activity in the industry, INVESTMENT RATIO, calculated as the ratio of real capital investment in the current year to the real capital stock of the industry.

Job Effects and the Competitive Environment

59

There are a total of 457 manufacturing industries in the NBER- CES data set, based on the Standard Industrial Classification (SIC) system.12 Of these, 29 industries are located within the apparel sector (SIC 23), which is not included in the PACE survey (it has many small establishments with little or no pollution abatement expenses). Another 29 industries had missing values for the data needed to calculate the product demand or import ratio mea sures. Our final- analysis data set contains data for 399 manufacturing industries over the 1973–1994 period, a total of 8,778 industry-year observations. In addition to the year-by-year definitions of our three mea sures of industry competitive pressures, we also calculate their average values for each industry over the time period. Those industry- average values are used for the interactions with ABATEMENT COST RATIO, designed to test whether industries that tend to face more (or less) of each sort of competitive pressure have their employment more (or less) affected by environmental regulatory pressures. Our main estimating equation is

(5)

LN_EMPLOYMENTit = α i + μ * ABATEMENT COST RATIOit + ∑ ρk * X ik * ABATEMENT COST RATIOit + ∑ βk * X itk + ∑ λ j * Witj + δ t + ε it .

Here Xk refers to the three variables mea sur ing the industry competitive environment (CONCENTRATION RATIO, IMPORT RATIO, and DEMAND GROWTH). Their long-run average values (Xl) for the industry are interacted with ABATEMENT COST RATIO, while their timevarying values (Xit) are allowed to affect employment directly. Wj refers to the additional industry characteristics included in the model without any interaction terms (LN_ENERGY PRICE, LN_MATERIALS PRICE, and INVESTMENT RATIO). Note that the model includes fi xed industry effects (αi) and year effects (δt) as well as the error term. Our focus will be on the impact of abatement costs on employment (μ) and especially any changes in that impact due to the industry competitive environment (ρk ).

Results Table 3.1 presents some descriptive statistics for our data set. The average industry-year observation has 40,000 employees, $5 billion in shipments, and $25 million in pollution abatement operating costs, with large variation in those three characteristics across the observations. The table also provides information about the distribution of the

Table 3.1. Summary Statistics (8,778 observations, 399 industries, 1973–1994)

Variable EMPLOYMENT SHIPMENTS PAOC ABATEMENT COST RATIO CONCENTRATION RATIO DEMAND GROWTH IMPORTS IMPORT RATIO INVESTMENT RATIO Variables in Logs, for Regressions LN_EMPLOYMENT LN_SHIPMENTS LN_ENERGY PRICE LN_MATERIALS PRICE

M

SD

39.714 4,935.958 24.909 0.0042 0.537 0.056 597.527 0.169 0.073

54.606 10,913.590 117.808 0.0065 0.227 0.0532 2,515.923 0.447 0.040

M

SD

3.139 7.870 −0.195 −0.144

1.019 1.135 0.415 0.301

Description Employment (000s) Value of shipments ($ million) Pollution abatement operating costs ($ million) PAOC/((vshipt−1 + vshipt−2)/2) Concentration ratio (top 8 firms) Demand growth Imports ($ million) IMPORTS/((vshipt−1 + vshipt−2)/2) Investment/capital stock Description log(EMPLOYMENT) log(SHIPMENTS) log(energy price index) log(materials price index)

Distribution of Industry- Average Variables Variable ABATEMENT COST RATIO DEMAND GROWTH CONCENTRATION RATIO IMPORT RATIO

25%

50%

75%

95%

0.0014 0.0400 0.3650 0.0357

0.0022 0.0613 0.5405 0.0766

0.0092 0.0753 0.6964 0.1687

0.0190 0.0827 0.9386 0.6069

Job Effects and the Competitive Environment

61

industry- average values of key industry characteristics. The abatement cost ratio shows a skewed distribution, with most industries having relatively low abatement costs; three- quarters of the industries spend less than 1 percent of shipments on abatement, and the mean observation (0.4 percent) is twice as large as the median (0.2 percent). The import ratio is even more skewed, with the mean value (16.9 percent) being larger than three- quarters of the industry averages. The concentration ratio and demand growth mea sures are more evenly balanced across industries. These distributions will be helpful in the following section as we interpret our models, with a focus on the coefficients on the abatement cost ratio and the interactions between it and the industry characteristics. Table 3.2 presents the results from our basic models of industry employment. In interpreting these results, it is important to note that these are fi xed- effect models (including both industry and year dummies). This substantially reduces the amount of variation used in the model but allows us to control for fixed differences across industries not captured by our other control variables. The impact of each explanatory variable on employment depends only on the variation in that variable within the industry over time: for example, the import ratio coefficient does not depend on whether a given industry always has high or low imports but only on whether an industry’s import ratio for this par ticu lar year is higher or lower than that industry’s average import ratio over the whole time period. In addition, the year dummies control for anything specific to a particular year, so if all industries had higher imports in later years, that variation would be captured by the year dummies, so that only higher (or lower) than average growth in imports would be counted. Nearly all of the control variables are statistically significant. Of those variables whose signs are predictable, both the investment rate and import ratio have the expected signs, with lower investment rates and higher import ratios being associated with lower employment. Higher concentration ratios are also associated with lower employment, perhaps reflecting a tendency of fi rms in more concentrated industries to restrict output in order to raise prices (more concentrated industries also tend to be more capital intensive). Demand growth has an unexpected negative coefficient but is not statistically significant. The focus of our analysis is on the abatement cost ratio and especially on its interactions with other industry characteristics. In Model 1, we see that higher abatement costs are associated with significantly lower employment. To interpret the magnitude of this coefficient, we need to refer back to the numbers in Table 3.1. The median industry has an

Table 3.2. Determinants of LN_EMPLOYMENT

Variables ABATEMENT COST RATIO ABATEMENT COST RATIO × DEMAND GROWTH ABATEMENT COST RATIO × CONCENTRATION RATIO ABATEMENT COST RATIO × IMPORT RATIO DEMAND GROWTH CONCENTRATION RATIO IMPORT RATIO LN_ENERGY PRICE LN_MATERIALS PRICE INVESTMENT RATIO

(1)

(2)

(3)

(4)

(5)

−3.369*** (−4.69)

−11.24*** (−9.81) 209.9*** (8.79)

14.45*** (6.78)

−3.647*** (−4.58)

4.852 (1.92) 171.5*** (6.95) −20.96*** (−7.14) 0.173 (0.13) −0.0352 (−0.56) −0.455*** (−8.53) −0.143*** (−18.35) 0.107*** (3.65) −0.109*** (−4.56) 1.682*** (24.94)

−0.0814 (−1.28) −0.472*** (−8.80) −0.142*** (−18.66) 0.102*** (3.44) −0.104*** (−4.30) 1.702*** (25.07)

Controls: Industry Year N R2 Note: t statistics are in parentheses; ***p < .001.

X X 8778 0.212

−0.0253 (−0.40) −0.458*** (−8.57) −0.146*** (−19.21) 0.102*** (3.46) −0.107*** (−4.45) 1.693*** (25.04) X X 8778 0.220

−25.41*** (−8.88) −0.0813 (−1.29) −0.465*** (−8.71) −0.140*** (−18.35) 0.108*** (3.68) −0.107*** (−4.46) 1.687*** (24.96) X X 8778 0.220

1.060 (0.80) −0.0797 (−1.26) −0.471*** (−8.78) −0.144*** (−18.30) 0.102*** (3.44) −0.104*** (−4.32) 1.701*** (25.05) X X 8778 0.212

X X 8778 0.224

Job Effects and the Competitive Environment

63

ABATEMENT COST RATIO value of 0.22 percent, so a 10 percent increase in its overall abatement costs would lead to a loss of roughly 30 jobs in the average industry with 40,000 employees. The overall effect of pollution abatement costs on employment in a typical industry is very small.13 Looking at the interaction terms in the remaining models, we see that their coefficients are consistent in significance and magnitude regardless of whether they are included individually (Models 2– 4) or all three together (Model 5). As expected, industries with more rapidly growing demand tend to have a smaller impact of abatement costs on employment (with the positive interaction term reducing the negative ABATEMENT COST RATIO term from Model 1). What is surprising is that employment in industries with higher concentration ratios seems to be more sensitive to abatement costs (negative interaction term). Industries with higher imports also show an unexpected coefficient (employment being less sensitive to ABATEMENT COST RATIO), but that coefficient is not statistically significant. To roughly quantify these effects (using Model 5) for the 10 percent increase in abatement costs mentioned earlier, an industry with 1 percentage point higher demand growth would lose 15 fewer jobs, and an industry with 10 percentage points higher concentration would lose 19 more jobs, relative to the 30 jobs in the earlier example. Table 3.3 presents similar models, this time looking at the determinants of industry shipments. Not surprisingly, the coefficients are quite similar to those in Table 3.2, in terms of signs and significance (and usually in magnitude). The only shift in sign comes for demand growth, which now has the expected positive sign, although it remains statistically insignificant. Abatement costs are negatively related to output, with this impact being smaller for industries with more growth in product demand and (surprisingly) being greater for more concentrated industries, while industry imports do not seem to affect the sensitivity of industry shipments to abatement costs.14 As we noted earlier, the pollution abatement cost ratio is distributed unevenly across industries, with most industries spending only a small fraction of their total costs on pollution abatement. In Table 3.4, we restrict our analysis to employment models for those industries whose average ABATEMENT COST RATIO value exceeds 0.4 percent, corresponding roughly to the top one- quarter of the sample (102 industries). The coefficients on abatement cost are similar to those for the full sample but slightly larger, suggesting that these high-regulation industries may be somewhat more sensitive to abatement costs overall than the typical industry, as well as facing higher abatement costs than average. In this case, a 10 percent increase in abatement costs, using the

Table 3.3. Determinants of LN_SHIPMENTS

Variables ABATEMENT COST RATIO ABATEMENT COST RATIO × DEMAND GROWTH ABATEMENT COST RATIO × CONCENTRATION RATIO ABATEMENT COST RATIO × IMPORT RATIO DEMAND GROWTH CONCENTRATION IMPORT RATIO LN_ENERGY PRICE LN_MATERIALS PRICE INVESTMENT RATIO Controls: Industry Year N R2

(1)

(2)

(3)

(4)

(5)

−4.678*** (−5.23)

−13.13*** (−9.19) 225.4*** (7.57)

7.125** (2.68)

−5.357*** (−5.40)

−4.315 (−1.37) 200.8*** (6.51) −11.73** (−3.20) 1.276 (0.77) 0.0796 (1.01) −0.123 (−1.85) −0.121*** (−12.36) 0.326*** (8.86) −1.290*** (−43.11) 1.930*** (22.90)

0.0239 (0.30) −0.141* (−2.11) −0.117*** (−12.26) 0.323*** (8.74) −1.285*** (−42.79) 1.948*** (23.02) X X 8778 0.329

Note: t statistics are in parentheses; *p < .05, **p < .01, ***p < .001.

0.0842 (1.06) −0.126 (−1.89) −0.121*** (−12.71) 0.323*** (8.78) −1.288*** (−43.05) 1.937*** (22.98) X X 8778 0.334

−16.83*** (−4.70) 0.0240 (0.30) −0.136* (−2.04) −0.115*** (−12.07) 0.327*** (8.87) −1.287*** (−42.91) 1.938*** (22.93) X X 8778 0.331

2.590 (1.58) 0.0279 (0.35) −0.139* (−2.07) −0.120*** (−12.28) 0.323*** (8.75) −1.286*** (−42.82) 1.945*** (22.99) X X 8778 0.329

X X 8778 0.334

Table 3.4. High Abatement Cost Industries—Assorted Models

Variables ABATEMENT COST RATIO ABATEMENT COST RATIO × DEMAND GROWTH ABATEMENT COST RATIO × CONCENTRATION RATIO ABATEMENT COST RATIO × IMPORT RATIO DEMAND GROWTH CONCENTRATION RATIO IMPORT RATIO LN_ENERGY PRICE LN_MATERIALS PRICES INVESTMENT RATIO Controls: Industry Year N R2

(1) LN_EMPLOYMENT

(2) LN_EMPLOYMENT

(3) LN_SHIPMENTS

(4) LN_SHIPMENTS

−4.054*** (−5.35)

−1.803 (−0.66) 161.3*** (6.24) −10.12** (−3.17) −0.383 (−0.28) 0.000772 (0.01) −0.257** (−2.69) −0.126*** (−8.10) −0.165*** (−3.40) 0.0960* (2.03) 0.650*** (6.76)

−5.411*** (−5.78)

−9.798** (−2.90) 193.2*** (6.03) −2.311 (−0.58) −1.265 (−0.75) 0.324* (2.53) 0.183 (1.55) −0.113*** (−5.84) −0.0993 (−1.65) −0.375*** (−6.38) 0.534*** (4.48)

−0.0737 (−0.71) −0.282** (−2.91) −0.120*** (−8.43) −0.181*** (−3.71) 0.101* (2.11) 0.667*** (6.86) X X 2244 0.270

Note: t statistics are in parentheses; *p < .05, **p < .01, ***p < .001.

X X 2244 0.289

0.233 (1.82) 0.163 (1.37) −0.106*** (−6.06) −0.107 (−1.76) −0.372*** (−6.28) 0.547*** (4.56)

X

X

2244 0.154

2244 0.169

66

Wayne B. Gray and Ronald J. Shadbegian

median ABATEMENT COST RATIO value of 0.75 percent in this sample, would be predicted to lead to a loss of 122 jobs in an industry employing 40,000 workers (these high-regulation industries are somewhat smaller than other manufacturing industries, averaging under 30,000 workers, so the actual predicted job loss would be 87). The signs and magnitudes on the demand and concentration interactions are similar to those found in earlier models, although the import interaction is now negative for both employment and shipments, though not signifi cant. Relative to the example of 122 jobs lost, an industry with 1 percentage point higher demand growth would lose 49 fewer jobs, and an industry with 10 percentage points higher concentration would lose 30 more jobs.

Conclusions and Future Work In this chapter, we have used a multiple regression framework to analyze the impact of environmental regulation on employment in U.S. manufacturing. In particular, we examined how the effects of environmental regulatory pressure on an industry can differ, depending on the competitive environment in which the industry is operating. We performed our analysis using data from 1973 to 1994 at the industry level for U.S. manufacturing industries and considered the impact of regulation on both industry output and industry employment. Our analysis shows that higher levels of regulation are associated with statistically significant but quantitatively very small job losses. In par ticular, we find that a 10 percent increase in overall abatement costs relative to output (our measure of regulatory stringency) is associated with a loss of roughly 30 jobs in the average industry with 40,000 employees, so the overall effect of pollution abatement costs on employment in a typical industry is very small. This effect is somewhat larger when we focus only on highregulation industries. We also find, as expected, that industries with more rapidly growing demand tend to have a smaller impact of regulation on employment. Conversely, we unexpectedly find that employment in industries with higher concentration ratios seems to be more sensitive to regulation, while the sensitivity of employment to regulation is not significantly affected by an industry’s level of imports. One question that is beyond the scope of our chapter is what effect environmental regulation has on overall net employment in the U.S. economy. We estimate the impact abatement costs have only on employment in the manufacturing sector and not on the jobs created in the manufacturing sector by the abatement expenditures. We would expect that the number of jobs created by additional abatement spending would

Job Effects and the Competitive Environment

67

approximately equal the number of jobs lost in the regulated manufacturing industries.

Acknowledgments The opinions and conclusions expressed are those of the authors and not the U.S. Environmental Protection Agency.

Notes 1. Of course, to the extent that labor costs are part of total costs, then they are included in a benefit– cost analysis. 2. For an attempt to monetize the value of regulation-induced job gains or losses, see Masur and Posner (2012, this volume). 3. Berman and Bui (2001) developed a similar theoretical model. 4. For more information on the size of the environmental protection sector, see Nestor and Pasurka (1995) and Bezdek et al. (2008). 5. For instance, studies have examined the effect of environmental regulation on productivity (e.g., Berman and Bui 2001; Boyd and McClelland 1999; Färe et. al. 1986; Shadbegian and Gray 2005, 2006), investment (e.g., Gray and Shadbegian 1998; Greenstone 2002), and environmental per for mance (e.g., Laplante and Rilstone 1996; Magat and Viscusi 1990; Shadbegian and Gray 2003, 2006). 6. Note that this, unlike the above studies, is a gross effect and not a net effect. Thus, Greenstone’s results do not mean that there is less aggregate employment due to environmental regulation. They simply suggest that the relative growth rate of employment in some sectors may differ between attainment and nonattainment areas. 7. This approach allows one to model the fi rm’s behavior with a variable cost function that is minimized with respect to a subset of the factor inputs conditional on both output and the levels of the “quasi-fi xed” factors. 8. Because our empirical model is estimated with industry-level data, our theoretical model should be considered a representative firm model. 9. The data are available at Bartelsman et al. (2000) and described in detail in Bartelsman and Gray (1996). 10. The PACE survey was not conducted in 1987 for budgetary reasons. We interpolated a 1987 value, using the 1986 and 1988 values, to generate a complete sample of years. 11. To avoid concerns about the denominator of this ratio (shipments) also serving as a dependent variable, we used the average shipments from the two preceding years rather than this year’s shipments. Thus, 1977 abatement costs are divided by the average shipments for 1975 and 1976. This same adjustment was applied to the denominator of the import penetration ratio described below. 12. The United States is currently using the North American Industry Classification System (NAICS) for industry definitions, but the change from SIC to NAICS industry definitions occurred in 1997, so the NAICS codes are not relevant for the time period being studied here.

68

Wayne B. Gray and Ronald J. Shadbegian

13. Multiplying the ABATEMENT COST RATIO change of 0.00022 (10 percent of 0.22 percent) by the estimated coefficient of −3.369 yields a percentage reduction in employment of 0.074 percent. 14. We tried two alternative models. In one model, we expressed each variable in terms of the difference between its current value and its value from five years before. In the other model, we dropped the industry dummies from the model, allowing fi xed industry characteristics to influence the coefficients. In  both cases, the patterns of signs and significance are similar to those shown in Tables 3.2 and 3.3, although the estimated coefficients are somewhat smaller. These results are available upon request.

References Bartelsman, Eric J., and Wayne Gray (1996) “The NBER Manufacturing Productivity Database.” NBER Technical Working Paper 205, Cambridge, MA. Bartelsman, Eric J., et al. (2000) “NBER- CES Manufacturing Industry Database,” National Bureau of Economic Research, http://www.nber.org/nberces /nbprod96.htm (accessed 8 January 2013). Becker, Randy A., and Ronald J. Shadbegian (2005) “A Change of PACE: A Comparison of the 1994 and 1999 Pollution Abatement Costs and Expenditures Survey,” Journal of Economic and Social Measurement 30: 63– 95. Berman, Eli, and Linda T. M. Bui (2001) “Environmental Regulation and Labor Demand: Evidence from the South Coast Air Basin,” Journal of Public Economics 79: 265– 95. Bezdek, Roger H., et al. (2008) “Environmental Protection, the Economy, and Jobs: National and Regional Analyses,” Journal of Environmental Management 86: 63–79. Boyd, Gale A., and John D. McClelland (1999) “The Impact of Environmental Constraints on Productivity Improvement in Integrated Paper Plants,” Journal of Environmental Economics and Management 38: 121– 42. Brown, Randall S., and Laurits Ray Christensen (1981) “Estimating the Elasticities of Substitution in a Model of Partial Static Equilibrium: An Application to US Agriculture 1947 to 1974,” in E. R. Berndt and B. C. Field, eds., Modeling and Measuring Natural Resource Substitution. Cambridge, MA: MIT Press. Bureau of Economic Analysis (1991) “1982 Interindustry Transactions, Margins, and Direct Requirements Coefficients,” Bureau of Economic Analysis, http://www.bea.gov/industry/zip/ndn0025.zip (accessed 8 January 2013). ———. (2004) “BEA Industry Accounts Data,” Bureau of Economic Analysis, http://www.bea.gov/industry/xls/GDPbyInd_VA_SIC.xls (accessed 8 January 2013). Cole, Matthew A., and Rob J. Elliott (2007) “Do Environmental Regulations Cost Jobs? An Industry- Level Analysis of the UK,” B.E. Journal of Economic Analysis & Policy 7: 28. Färe, Rolf, et al. (1986) “Effects on Relative Efficiency in Electric Power Generation due to Environmental Controls,” Resources and Energy 8: 167– 84. Feenstra, Robert C. (1996) “U.S. Imports, 1972–1994: Data and Concordances.” NBER Working Paper 5515, Cambridge, MA. Gray, Wayne B., and Ronald J. Shadbegian (1998) “Environmental Regulation, Investment Timing, and Technology Choice,” Journal of Industrial Economics 46: 235– 56.

Job Effects and the Competitive Environment

69

Gray, Wayne B., et al. (2013) “Do EPA Regulations Affect Labor Demand? Evidence from the Pulp and Paper Industry.” U.S. Census Bureau Center for Economic Studies Working Paper 13–39. Greenstone, Michael (2002) “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures,” Journal of Political Economy 110: 1175–1219. Laplante, Benoît, and Paul Rilstone (1996) “Environmental Inspections and Emissions of the Pulp and Paper Industry in Quebec,” Journal of Environmental Economics and Management 31: 19–36. Magat, Wesley A., and W. Kip Viscusi (1990) “Effectiveness of the EPA’s Regulatory Enforcement: The Case of Industrial Effluent Standards,” Journal of Law and Economics 33: 331– 60. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry- Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. Nestor, Deborah, and Carl Pasurka, Jr. (1995) U.S. Environmental Protection Industry: A Proposed Framework for Assessment. PB—96-134309/XAB; EPA—230 /R-95/011. Washington, DC: U.S. Environmental Protection Agency. Obama, Barack (2011) Executive Order 13563 of January 18, 2011: Improving Regulation and Regulatory Review, 76 Federal Register 3821. Schott, Peter K. (2008) “Schott’s International Economics Resource Page: Trade Data and Concordances,” http://faculty.som.yale.edu/peterschott /sub_international.htm (accessed 8 January 2013). ———. (2010) “U.S. Manufacturing Exports and Imports by SIC or NAICS Category and Partner Country, 1972 to 2005,” http://faculty.som.yale.edu/ peterschott/files/research/data/sic_naics_trade_20100504.pdf (accessed 8 January 2013). Shadbegian, Ronald J., and Wayne B. Gray (2003) “What Determines Environmental Per formance at Paper Mills? The Roles of Abatement Spending, Regulation, and Efficiency,” Topics in Economic Analysis and Policy 3, 15 (Nov.), http:// www.bepress.com/bejeap/topics/vol3/iss1/art15. ———. (2005) “Pollution Abatement Expenditures and Plant- Level Productivity: A Production Function Approach,” Ecological Economics 54: 196–208. ———. (2006) “Assessing Multi- Dimensional Per for mance: Environmental and Economic Outcomes,” Journal of Productivity Analysis 26: 213–34. U.S. Census Bureau (1992) “Concentration,” http://www.census.gov/epcd/www /concentration92- 47.xls (accessed 8 January 2013).

Chapter 4

The Employment and Competitiveness Impacts of Power-Sector Regulations Joseph E. Aldy and William A. Pizer

In the debate over environmental regulations, a principal concern is the potential impact on employment in the more energy-intensive U.S. manufacturing industries. Although the academic literature and agency practice in regulatory impact analyses have estimated the direct effect of manufacturing- sector environmental regulations on employment, these literatures have been largely silent on the indirect effects of powersector environmental regulations (the exception being general equilibrium analyses of power- sector policies; Rausch and Mowers 2012). Yet the extensive array of power- sector regulations on the horizon could increase electricity rates manufacturing firms face. This would increase domestic production costs, and eventually prices charged to customers must rise, causing a decline in domestic sales as well. This production decline may include, in part, a shift of economic activity and jobs overseas to key trading partners, if they do not face comparable regulation. This chapter focuses on estimating the magnitude of both gross manufacturing- sector employment impacts and net competitiveness impacts in the power sector under environmental regulations. We define such competitiveness effects as the adverse business impacts related to a domestic regulatory policy in the absence of regulation on international competitors. It is the harm domestic firms bear because they face a higher price on factors of production, in this case electricity, than their foreign competitors, specifically owing to differences in the regulatory regimes faced by firms participating in a given market. Some of these domestic firms have limited pricing power for manufactured commodity-like goods that compete in a global market, and this inhibits their ability to pass through the costs of a domestic regulatory policy.

Employment and Competitiveness Impacts of Regulations

71

This definition of competitiveness highlights that a portion of the regulatory impact on U.S. industry is unrelated to foreign trade. Even if all U.S. trading partners were to implement environmental regulations identical to those in the United States, some emission- and energyintensive firms in the United States could still bear substantial costs. In the face of a new domestic regulatory program, the costs of investing in new technologies to reduce a firm’s emissions, along with declines in the consumption and production of emissions- and energy-intensive goods, are distinct from domestic producers losing market share or profits solely because foreign competitors do not face similar regulation. We believe this distinction is important because foreign competition and the loss of jobs overseas are frequently cited in arguments against environmental regulation. As an illustration of the need for this kind of analysis of power- sector regulatory impacts on manufacturing employment, consider the Environmental Protection Agency’s 2011 Cross- State Air Pollution Rule (CSAPR). This rule imposes pollution control requirements on about 3,700 fossil-fuel-fired generating units that could increase the cost of producing electricity.1 As a part of the rule, the U.S. EPA (2011) published a regulatory impact analysis that characterizes the benefits and costs of the regulation. The U.S. EPA (2011:286) notes, however, that it has “not quantified the rule’s effects on all labor in other sectors not regulated by this rule.” Although the agency has not quantified the employment impacts beyond the power sector, it recognizes that this is a “notable” impact when describing the overall labor impacts of the regulation: We expect ongoing employment impacts on regulated and non-regulated entities for a variety of reasons. These include labor changes in the regulated entities resulting from shifts in demand for fuels, increased demand for materials and the labor required to provide them to operate pollution control equipment, reductions in employment resulting from coal retirements, and reductions in other industries due to slight projected increases in the price of electricity and natural gas. The most notable of the ones we are unable to estimate are the impacts on employment as a result of the increase in electricity and other energy prices in the economy. (U.S. EPA 2011:295)

The empirical methods employed in this chapter could explicitly address this kind of omission in regulatory impact analysis.

Overview of the State of the U.S. Manufacturing Sector The sensitivity of American manufacturing to energy prices varies across industries because of the significant heterogeneity in the energy

72

Joseph E. Aldy and William A. Pizer

required to produce a dollar of output. For example, the manufacturing of hydraulic cement is approximately 100 times more energyintensive than making cigarettes, and it is about 50 times more energy- intensive than the manufacturing of telephones. The chemicals, primary metals, pulp and paper, and stone, glass, and clay (including cement) industries consume slightly more than half of all energy used in the manufacturing sector (U.S. Energy Information Agency multiple dates). These energy- intensive industries’ share of energy in the manufacturing sector has remained fairly steady since the early 1990s, although primary metals (for example, steel, aluminum) have experienced a modest decline, reflecting their declining share of manufacturing output over time. Their shares of the manufacturing sector’s production— about 16 percent— have likewise remained steady since the early 1990s (Bureau of Economic Analysis n.d.), and they illustrate the relative energy intensity of their production. The energy-intensive industries’ share of employment has followed a similar pattern since the early 1990s, consistently comprising about 20 percent of the manufacturing sector’s payrolls (Bureau of Labor Statistics n.d.). Of energy inputs, electricity expenditures comprise a majority of energy expenditures in the manufacturing sector. In 2001, the mean and median electricity cost share of total energy costs were 0.64 and 0.68, respectively. Eighty-two percent of manufacturing industries had a majority of energy costs coming from electricity expenditures in 2001. Over the past three decades, the energy intensity of the U.S. manufacturing sector has improved, with much of this improvement occurring as a result of the oil shocks– induced price increases in the 1970s and early 1980s. Petroleum consumption in the broader industrial sector peaked in 1979 and fell nearly one- quarter through 2010 even as total industrial production was 75 percent greater in 2010 than in 1979 (Council of Economic Advisers 2012; U.S. Energy Information Administration 2011). The declining energy intensity of output reflects changing production techniques and innovation in manufacturing. Steel production has shifted from blast oven furnace (BOF) production, which comprised 70 percent of U.S. steel output in 1985, to electric arc furnace (EAF) techniques, which made up 55 percent of production in 2005 (Office of Technology Assessment 1985; U.S. EPA 2007). This transition affects the energy impact of steel production, as BOF allows for cogeneration of heat, whereas EAF requires large amounts of electricity and thus makes it more sensitive to the costs of power- sector regulations. The energy intensity of U.S. aluminum production has declined by 61 per-

Employment and Competitiveness Impacts of Regulations

73

cent over the past 40 years, reflecting technological improvements and the growth of the recycling sector, which requires less energy than manufacturing from raw materials (U.S. Department of Energy 2007). The resilience of the paper industry to the energy price shocks of the 1970s and early 1980s reflects its atypical position as a major source of power: the industry fulfills roughly half of its own energy needs via biomass cogeneration. This partially buffers the industry from electricity price shocks and could even yield a benefit if paper mills can sell power to the grid when electricity rates are higher. Cement production has shifted toward dry process cement, which requires less direct energy but more electricity than wet process cement, as the proportion of U.S. kilns using the dry process increased from 38 to 70 percent over the 1975–2001 period (Hanle et al. 2004). A few snapshots of the energy- intensive manufacturing sector reveal industries that grow slower than the rest of the U.S. economy and, through technological change and competitive pressures, have reduced payrolls over the past few decades. Most energy- intensive industries— iron and steel, aluminum, paper, cement, glass, and industrial chemicals industries— have experienced declines in payrolls on the order of 40 percent or more since 1983 (cement is the outlier, with an increase of 20 percent). This has occurred while some industries— such as iron and steel, glass, and cement—have experienced production increases of 40  percent or more, while other industries— aluminum, paper, and chemicals—have witnessed flat or modest growth in output. The slow growth in U.S. manufacturing output reflects two phenomena: slow demand growth and increasing international competition. Just as energy-intensive industries responded to high energy prices by economizing on their use of energy in production, downstream users of these industries’ goods have found ways to efficiently use less of these energy-intensive inputs in the production of their final goods. Some of this may reflect changes in quality—for example, steel used in automobiles becoming stronger over time—that allows downstream producers to use less of the energy-intensive good. Some of this may reflect opportunities for substitution—for example, aluminum or plastic substituting for steel in automobile manufacture. U.S. energy-intensive manufacturers’ share of the domestic market has also declined over time. Although net imports can vary significantly from year to year, they do show an increasing trend in recent years for most energy- intensive industries (United Nations Statistics Division n.d.). Thus, any assessment of the employment and competitiveness impacts in manufacturing should account for the baseline of significantly lower employment

74

Joseph E. Aldy and William A. Pizer

levels today than a generation ago and increasing competition from foreign suppliers.

The Impacts of Environmental Regulation on Manufacturing A substantial research literature has addressed the question of whether and how environmental regulations adversely affect American industry. Studies including Schmalensee (1993), Rutledge and Vogan (1994), Jaffe et al. (1995), and Morgenstern et al. (2001) have all focused on the burden of environmental regulation on industry. Yet, although the overall impact on costs, employment, or production is of interest in its own right, most of the literature has framed the issue in terms of trade competition with other jurisdictions. That is, there is some notion that lowering consumption (and production) of polluting goods may be okay, but simply shifting their production abroad is not. Indeed, much of the current policy debate reflects issues raised by theoretical analyses suggesting that environmental policy could create so- called pollution havens in developing countries: “The conventional wisdom is that environmental regulations impose significant costs, slow productivity growth, and thereby hinder the ability of U.S. firms to compete in international markets. This loss of competitiveness is believed to be reflected in declining exports, increasing imports, and a long-term movement of manufacturing capacity from the United States to other countries, particularly in ‘pollution-intensive’ industries” (Jaffe et al. 1995:133). Evaluating this conventional wisdom requires a careful examination of a simple empirical question: Do firms lose market share in response to domestic environmental policies, either by relocating their manufacturing activity to or by facing lower- cost competition from countries with lax environmental policies? Addressing this question necessitates an assessment of the broader context about firm location. A variety of factors may mitigate or dominate the effect of environmental regulatory costs in determining manufacturing location decisions. First, the availability of relevant factors of production, such as appropriately skilled labor, natural resources, and capital, can play a more significant role than pollution control costs. Pollution- intensive industries tend to be capital-intensive, so capital abundance in developed countries may outweigh the impacts of environmental regulations (Antweiler et al. 2001). Second, transportation costs may discourage relocation to countries far from the major markets for manufactured goods. Ederington et al. (2005) found that transportation costs diminish the impact of pollution abatement costs on net imports: an industry with high transport costs (for example, at the 80th

Employment and Competitiveness Impacts of Regulations

75

percentile in the manufacturing sector) experiences a percentage increase in net imports equal to about 20 percent of the impact for an industry with average transport costs (for example, at the 50th percentile in the manufacturing sector). Firms with a significant share of their investments in large, fixed physical structures also appear to move activity less in response to environmental regulations (Ederington et al. 2005). Proximity to firms that produce inputs or purchase outputs—for example, industrial parks and related forms of so- called agglomeration economies—also discourages relocation (Jeppesen et al. 2002). These factors all determine whether an industry is “footloose,” that is, sufficiently mobile that a small change in production costs, such as from an environmental regulation, could drive some firms to relocate to other countries. Because the most pollution-intensive industries tend to be relatively immobile by these measures of “footlooseness,” the empirical literature typically finds quite limited impacts of environmental regulations on international competitiveness. Levinson and Taylor (2008) showed that U.S. pollution abatement costs in the 1970s and 1980s increased net imports in the manufacturing sector from Mexico and Canada. The estimated increase in net imports from pollution abatement roughly equaled about 10 percent of the total increase in bilateral trade for both Mexico and Canada, suggesting that other factors played much more substantial roles in the evolution of trade among the North American trading partners. An extensive literature on the competitiveness effects of variation in environmental policies across the U.S. states has shown more significant impacts on domestic firm and employment relocation resulting from variation in the stringency of environmental regulations (Greenstone 2002; Henderson 1996). The larger domestic competitiveness impacts may reflect the fact that labor costs and availability of capital do not vary much across the U.S. states and transportation costs are less important, relative to the international context. In the context of carbon pricing policy in the European Union (EU) and potential carbon pricing in the United States, a wave of papers has addressed the associated competitiveness impacts of climate change policies. Given the prospective nature of these analyses, the scholars have undertaken detailed accounting exercises or employed models to simulate the effects of carbon prices on output and related impacts. The accounting-based papers focus on energy-intensive sectors and infer a percentage cost increase from a carbon price at varying proportions of free permit allocation by using data on average cost, electricity use (assuming some level of pass-through), and direct carbon dioxide (CO2) emissions. They then make assumptions about demand elasticities to infer changes in production or employment. Reinaud (2005) examined impacts under a €10 per ton CO2 price (modeled after the EU

76

Joseph E. Aldy and William A. Pizer

Emission Trading Scheme [EU ETS]). She estimated that before accounting for any free allocation, energy- intensive industries would experience cost increases ranging from 1.5 percent for EAF steel to 18.6 percent for cement. Applying her assumptions of price elasticity of demand and maintenance of profitability margins, output declines, ranging from 2.3 percent for EAF steel to 12 percent for BOF steel. McKinsey and Company and Ecofys (2006) performed a similar EU ETS–based analysis at €20 per ton CO2. When properly scaled to a comparable carbon price, assuming linear costs, the McKinsey numbers are of a magnitude similar to Reinaud’s. Carbon Trust (2008) used a similar approach to that of the Reinaud and McKinsey studies in an evaluation of the UK manufacturing sector. Like McKinsey, the Carbon Trust analysis assumes a €20 per ton CO2 allowance price modeled on the EU ETS. This carbon price would increase the production costs in lime, cement, and iron and steel by more than 25 percent in the United Kingdom. Aluminum, inorganic chemicals, and pulp and paper would experience cost impacts on the order of 10 percent at €20 per ton. Ho et al. (2008) used a model-based approach to simulate the output, consumption, and trade impacts of a $10 per ton CO2 price implemented unilaterally in the United States. They simulated short- term impacts when firms have little opportunity to change production inputs and invest in new, low- carbon capital (in a partial equilibrium analysis) and long-term impacts that account for all adjustments to the CO2 price (in a general equilibrium model). They found that the CO2 price drives down manufacturing output by 1.3 percent in chemicals and plastics, 1.1 percent in primary metals, and 0.9 percent in nonmetallic minerals. Slightly more than half of the decline in chemicals and plastics production is offset by an increase in net imports from countries that are not implementing greenhouse gas emission mitigation policies. Primary metals would experience a 0.46 percent competitiveness effect and nonmetallic minerals a 0.42 percent effect. These results show that the reduction in output is represented by a larger drop in domestic consumption than in an increase in net imports.

Employment and Competitiveness: Measures, Data, and Methods We estimate the impacts of power- sector environmental regulations on manufacturing employment and competitiveness by drawing on the historic effects electricity prices have on these measures. Using regression analysis, we separately estimate the effects of the price of electricity on employment, production, and consumption over the 1986–1994

Employment and Competitiveness Impacts of Regulations

77

period for more than 400 manufacturing industries. We then use these statistically estimated relationships to simulate the gross employment effect as well as a measure of the net or competitiveness employment effect of higher electricity rates resulting from new regulations, such as carbon pricing or air quality rules. Gross employment impacts are of interest in their own right as an important economic indicator of the effect of new environmental regulations. That is, how much does employment change because of a regulation? The analysis is straightforward empirically: we simply estimate the electricity price– employment relationship while controlling for other important factors, as detailed below. To estimate the competitiveness employment impacts of electricity regulations—that is, the change in employment arising from the lack of similar regulation among our trading partners—we focus on electricity price impacts on production, net imports, and consumption, where we define consumption as production plus net imports. Changes in the consumption of manufactured goods reflect the impact of an environmental regulation that would occur independent of any trade effects. This is not a competitiveness effect. Therefore, we look at the residual effect on production arising from changes in net imports as the competitiveness effect (Aldy and Pizer 2009). Because we can decompose a gross production effect into a net competitiveness effect, we can use the ratio of the competitiveness to gross production effects to adjust the gross effect on employment to estimate the employment competitiveness effect. We used detailed data from more than 400 industries in the manufacturing sector. For example, our analysis includes 11 classifications within the iron and steel sector. We undertake our analysis with the relevant data in logarithms because these sectors differ vastly in size. Our models estimate and predict changes in percent terms, rather than in actual dollars or jobs, which facilitate comparability even among different- sized sectors. The reported impacts represent percent changes from the respective base levels for each of the more than 400 industries. We estimate employment, production, and consumption as a function of energy prices and other factors over the 1986–1994 period by using a model of the following form:2 Yit = αi + αt + ƒ(priceit; β ) + δ ′Xit + εit . In the model, Yit represents an industry and year- specific outcome measure—the natural logarithm of employment, value of shipments, and consumption; the α s are fixed effects for industries (i) and years (t); priceit is the natural logarithm of the average electricity cost in 1987 dollars; and X it is a vector of additional determinants of the industry

78

Joseph E. Aldy and William A. Pizer

outcome measures, including tariffs and factor intensity variables (to estimate the returns to human capital and physical capital). Table 4.1 presents the summary statistics for the three outcome mea sures used as dependent variables and all independent variables used to estimate the regression functions (except for year and industry fixed effects). The key results for our simulation are the estimated functions f for each of the different outcome variables. The function of price we specify in our regressions is a piecewise linear spline function that estimates distinct electricity price elasticities for different components of the distribution of industries based on their average energy intensity in the sample period. We focus on a seven- segment spline that estimates electricity price elasticities for industries in the first, second, third, and fourth quintiles of the energy intensity distribution, as well as those in the eighth decile (80th to 90th percentile) and 19th and 20th semi- deciles (90th to 95th percentile and greater than 95th percentile). Aldy and Pizer (2009) provide further details in a technical appendix on the empirical framework.

Table 4.1. Summary Statistics of Variables Employed in Various Regression Models

Variable Dependent variables

ln(employment) ln(value of shipments) ln(value of shipments + net imports)

Independent variables

ln(electricity price) Ratio of energy costs to value of shipments Average tariff rate (percent) Physical capital intensity Human capital intensity

M (SD) 3.09 (1.11) 7.76 (1.23) 7.81 (1.19) 0.72 (0.86) 0.026 (0.035) 5.95 (6.12) 0.58 (0.15) 0.14 (0.069)

Note: Means and standard deviations for 400+ industries over 1974–1994, representing 8,519 observations. Each regression used in this analysis employs one of the dependent variables and includes all independent variables in this table plus year and industry fi xed effects. The text describes how electricity prices are specified as a spline function of the energy intensity variable.

Employment and Competitiveness Impacts of Regulations

79

The assumption that outcomes depend on prices in a flexible way based on the function f is a critical distinction in our analyses. When we estimate the electricity price– employment relationship for the entire manufacturing sector without accounting for the energy intensity of industrial output, for example, we find no statistically meaningful effect. Decomposing this relationship as a function of energy intensity helps to illustrate the interesting variation across the manufacturing sector. For example, one might expect that firms in relatively energy-lean textiles could respond differently to a 10 percent increase in electricity prices than relatively energy-intensive steel firms. Our analysis allows us to estimate the energy price- competitiveness mea sure relationships for distinct components of the manufacturing sector as a function of their energy intensity.

Results of Empirical Analysis Figure 4.1 presents the distribution by energy intensity of the nearly 450 industries in the manufacturing sector. Figures 4.2 through 4.4 then present the basic results that we discuss in this and the following section. Figure 4.2 shows the estimated gross effect of prices on employment across a range of energy intensities—the estimated function f for

Figure 4.1. Distribution of More than 400 Industry Classifications by Energy Intensity. Note: Constructed by authors from U.S. Census Bureau Annual Survey of Manufactures (multiple dates).

80

Joseph E. Aldy and William A. Pizer

Figure 4.2. Estimated Gross Employment Elasticity. Note: See Aldy and Pizer (2009). CI = confidence interval.

employment. In particular, we can see that there is no statistically significant effect on employment for the roughly 80 percent industries in our sample with the lowest energy intensities (below 3– 4 percent). Above that level, we see employment facing a −0.2 to −0.3 elasticity with respect to electricity prices. As electricity prices rise 5 percent, employment declines 1–1.5 percent. As noted at the outset, however, this is a gross effect on employment. Some of that impact arises from declining consumption of energyintensive goods. Some also arises from a shift in production overseas, even as consumption remains unchanged. To estimate the net competitiveness effect on employment, we estimate the effect of prices on consumption and production and then take the ratio of the net import (production – consumption) effects to production effects. That ratio— the share of production effects that can be attributed to competitiveness (net imports)— is shown in Figure 4.3. For the roughly 60 percent of industries in our sample with an energy intensity above 1–2 percent, the estimate of competitiveness as a share of output effects ranges between 20– 40 percent. Industries with very small energy intensities have large shares, in some cases exceeding 100 percent, although it is impor-

Employment and Competitiveness Impacts of Regulations

81

Figure 4.3. Ratio of Change in Net Imports to Change in Total Production for a Change in Electricity Prices. Note: Constructed by authors based on independently estimated statistical results for production and competitiveness effects presented in Aldy and Pizer (2009).

tant to note that neither the production nor the consumption elasticities used to construct this ratio are statistically different from zero in this part of the energy intensity distribution. We combine Figures 4.2 and 4.3 into an estimate of the net competitiveness effect on employment in Figure 4.4. While the 80 percent of our sample with the lowest energy intensities should have an insignificant effect (from Figure 4.2), the elasticity for the upper 20 percent (above 3– 4 percent) varies between −0.05 and −0.1. That is, a 5 percent rise in electricity prices would be expected to lower employment by 0.25– 0.5 percent owing to increased imports from jurisdictions not facing similar price increases. This is what we define as the competitiveness effect.

Simulation of an Environmental Regulation: Carbon Pricing After conducting the statistical analysis described in the preceding section, we use the estimated relationships between electricity prices and our industry impact– competitiveness measures—which vary with energy

82

Joseph E. Aldy and William A. Pizer

Figure 4.4. Net Employment Elasticity (Product of Estimates in Figures 4.2 and 4.3).

intensity—to simulate the effects of a carbon pricing policy. We assume that the carbon pricing program yields a $15 per ton CO2 price in the first year of the program’s operation. Although we focus in this simulation on a carbon price—that is, a carbon tax or a cap- and-trade program—it is important to recognize that this analysis could also be applied in a straightforward manner to power- sector carbon dioxide per for mance standards as well as power- sector renewable portfolio and clean energy standards. Recent modeling work by the U.S. Energy Information Administration (2008) indicates that a $15 per ton CO2 price would increase the cost of electricity in the industrial sector by about 8 percent. This CO2 price— as an increase in electricity prices—represents approximately a one standard deviation increase in electricity prices given the historic price variation we observe. It would strain the credibility of our approach to use an effective price change that exceeded the values used to estimate the model parameters. Extrapolating impacts for higher CO2 prices is beyond the scope of this analysis as it would reflect an out- ofsample prediction.3 This carbon price is similar to allowance prices expected at the start of cap- and- trade programs proposed in recent legislation, including EPA’s (2009) estimate of a $13 per ton CO2 price under the Waxman-Markey Bill (2009), EPA’s (2010) estimate of a $17 per ton CO2 price under the American Power Act (2010) (that is, draft legislation from Senators Kerry and Lieberman), as well as the first-year

Employment and Competitiveness Impacts of Regulations

83

carbon tax of $15 per ton CO2 in a 2009 Republican- sponsored carbon tax bill (Raise Wages, Cut Carbon Act of 2009). On the basis of these estimated model parameters, this electricity price increase then drives the competitiveness and employment impacts in our simulation. Table 4.2 shows the impacts of a $15 per ton CO2 price on employment for all manufacturing and for specific sectors of the most energyintensive industries. Focusing on the first column, the manufacturing average change in gross employment at $15 per ton CO2 is −0.2 percent. The energy-intensive industries of iron and steel, aluminum, pulp and paper, cement, glass, and industrial chemicals would experience employment declines of −0.4 to −2.3 percent. Below the reported percentage changes in employment are the estimated changes in the level of employment by industry, which assumes that the employment level in a no-policy counterfactual is on par with 2010 employment data for these industries. The total gross employment change in the energy-intensive industries is a loss of about 10,000 jobs. These estimates reflect the product of the 7.4 percent increase in electricity prices (resulting from the $15 per ton CO2 price) and the elasticity appropriate for a given industry based on its energy intensity (see Figure 4.2). Each reported industry is an average of the constituent 6- digit NAICS industry estiTable 4.2. Predicted Impacts of a $15 per ton CO2 Price on Various Manufacturing Sectors (Percentages and Number of Workers)

Industry Industrial chemicals Paper Iron and steel Aluminum Cement Bulk glass Manufacturing average

Gross Employment −1.6% −5,500 −2.0% −2,200 −1.1% −2,000 −1.0% −700 −0.4% −700 −2.3% −200 −0.2%

Net Imports / Production

Net Employment (Competitiveness)

0.34

−0.5% −1,900 −0.6% −600 −0.4% −700 −0.3% −200 −0.2% −300 −0.4% −40 −0.1%

0.28 0.35 0.32 0.49 0.18 0.54

Note: Constructed based on our statistical analysis and the change in electricity prices predicted under a carbon pricing policy in U.S. Energy Information Administration (2008). Impacts are based on 2001 industry energy intensity, weighted by 2010 employment reported in the Quarterly Census of Employment and Wages among constituent six- digit NAICS industries based on a crosswalk from four- digit SIC industries listed in appendix 4 of Aldy and Pizer (2009). Number of worker estimates rounded to nearest hundred.

84

Joseph E. Aldy and William A. Pizer

mates weighted by their 2010 employment; the manufacturing average is an average across all 400-plus industries. We can convert to an estimate of the net competitiveness effect by multiplying the gross effect in column one by the estimate of the competitiveness share of the total production effect from Figure 4.3. Column two in Table 4.2 presents the share of the decline of production as a result of an increase in net imports, which varies by the energy intensity of the industry (see Figure 4.3), while column three reports the product of columns one and two. As expected, the net effects are uniformly smaller. All of the industry-level effects are under 1 percent, and the average for manufacturing as a whole is under 0.1 percent. It is worth noting that some more narrowly defined industries would experience impacts outside this range. The total net employment effect is about 4,000 lost jobs. These results suggest that consumers of energy-intensive goods do not respond to higher energy prices by consuming a lot more imports. Instead, they economize on their use of these higher-priced manufactured goods, perhaps by using less of the good in the manufacture of their finished products or by substituting with other, less energyintensive materials. Consumers appear to pursue only partial substitution with imports, suggesting that the imported versions of domestically produced goods may be imperfect substitutes. Other determinants of trade flows— such as transport costs, tariffs, and so forth— may limit the substitution possibilities.

Conclusion To illustrate the potential impacts of power- sector regulations on manufacturing-sector employment and competitiveness, we have estimated the historical relationship between electricity prices and employment and competitiveness (mea sured as the share of the output effect attributable to increased net imports). On the basis of our empirical model drawing from more than 400 manufacturing industries, we have simulated the impacts of a power- sector carbon pricing policy. We found estimated gross employment impacts on the order of 0.2 percent for the entire manufacturing sector and on the order of 1 to 2 percent for energy- intensive manufacturing as a result of a $15 per ton CO2 price in the power sector. The manufacturing- sector competitiveness impacts are less than 0.1 percent for the sector on average and under 1 percent for energy-intensive industries at $15 per ton CO2. These carbon pricing impacts also provide a sense of the magnitude of the manufacturing employment impacts of CSAPR, as noted in the introduction. Because U.S. EPA (2011) estimates electricity rate impacts

Employment and Competitiveness Impacts of Regulations

85

of about 2.2 percent, or approximately 30 percent of the price increase under the $15 per ton CO2 scenario, we would estimate a reduction in energy-intensive manufacturing employment of less than one-half of 1 percent, and competitiveness impacts on the order of 0.1 to 0.2 percent. Given the annual volatility in the economic characteristics of the manufacturing sector, especially energy- intensive manufacturing, the impacts for carbon pricing and CSAPR are likely to be swamped by other drivers in these industries. For example, the less than 1 percent competitiveness impacts resulting from a $15 per ton carbon pricing policy for energy-intensive manufacturing industries are swamped by the average percentage change in production that ranges from about 6 to 9 percent annually for these industries (Aldy and Pizer 2009). There are a variety of limitations to these estimates. First, they cannot credibly simulate the impacts of electricity price increases resulting from power- sector regulations much in excess of about 8 percent. Historical experience simply does not tell us what might happen when prices go higher— because we have not seen an isolated, equivalent change in energy prices in available data.4 Second, our estimates represent near-term impacts over one year to a few years. Unfortunately, volatility in energy prices and the confounding nature of other events make it difficult to estimate long- term impacts. Arguably with more time to adjust, U.S. industry could fare better (if it can reduce energy usage) or worse (if it has more time to move operations). Third, even with our disaggregated data and flexible model, we still cannot flexibly capture all of the features relevant for every industry in every international trading situation. The effects for some firms and sectors could be different than what we have estimated. Fourth, in using historical data, we are necessarily assuming that the past is a useful guide to future behavior. To the extent that there have been or will be substantial institutional changes, this assumption is flawed. Finally, our analysis has focused on the historic influence on net imports arising from domestic energy price increases as a mea sure of the difference between U.S.- only versus global action. To the extent net imports change significantly even with global regulatory action, our estimates will not capture these effects. The policy debate on employment and competitiveness impacts of environmental regulations could benefit from additional research. Given the spatial concentration of some industries as well as the heterogeneity in electricity prices and electricity price impacts resulting from environmental regulations, further work focused on regional impacts could enlighten the policy debate. In addition, alternative statistical identification strategies— such as through instrumental variables and regression discontinuity— could enhance the robustness of the findings reported here. Finally, work focused on other developed countries— such as the

86

Joseph E. Aldy and William A. Pizer

Eu ropean Union in the context of carbon taxes and cap- and-trade— could illuminate the discussion about the employment and competitiveness impacts of climate change policies.

Acknowledgments We benefited from excellent research assistance provided by Evan Herrnstadt. Thanks to workshop participants and Chris Carrigan, Cary Coglianese, and Adam Finkel for helpful comments on an earlier version. This chapter draws on research supported by the Pew Center on Global Climate Change and Resources for the Future.

Notes 1. Note that the U.S. Court of Appeals vacated this rule and remanded it to the EPA on August 21, 2012 (EME Homer City Generation v. EPA 2012). 2. We have also undertaken analyses over the 1974–2001 period. We have excluded the 1974–1985 period because it was the era of high energy prices and dramatic restructuring throughout the manufacturing sector. Our primary analysis concludes with 1994 because of data limitations on net imports. It is not possible to create industry- specific series for net imports after 1994. We extend our employment analyses through 2001 in sensitivity analyses. We find that our results are robust to the choice of time period. See Aldy and Pizer (2009) for more details. 3. We do not know if these relationships are linear over a small or large range of carbon prices, and if the relationship becomes nonlinear, theory cannot clarify whether the relationship would become convex or concave. 4. It is important to note that our analysis identifies the effect of electricity prices on employment and competitiveness mea sures after controlling for economy-wide factors (through year fi xed effects) and time-invariant industryspecific factors (through industry fi xed effects). It is the residual variation after accounting for time-varying economy-wide factors and time-invariant industry factors that drives our results.

References Aldy, Joseph E., and William A. Pizer (2009) The Competitiveness Impacts of Climate Change Mitigation Policies. Arlington, VA: Pew Center on Global Climate Change. Antweiler, Werner, et al. (2001) “Is Free Trade Good for the Environment?” American Economic Review 91: 877– 908. Bureau of Economic Analysis (n.d.) “State Annual Personal Income,” U.S. Department of Commerce, http://www.bea.gov/regional/spi/ (accessed 8 July 2008). Bureau of Labor Statistics (n.d.) “Output and Employment Database,” U.S. Department of Labor, http://www.bls.gov/emp/empind2.htm (accessed 20 June 2008). Carbon Trust (2008) EU ETS Impacts on Profitability and Trade: A Sector by Sector Analysis. London: Carbon Trust.

Employment and Competitiveness Impacts of Regulations

87

Council of Economic Advisers (2012) Economic Report of the President. Washington, DC: U.S. Government Printing Office. Ederington, Josh, et al. (2005) “Footloose and Pollution- Free,” Review of Economics and Statistics 87: 92– 99. Greenstone, Michael (2002) “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures,” Journal of Political Economy 110: 1175–1219. Hanle, Lisa, et al. (2004) “CO2 Emissions Profile of the Cement Industry.” Presented at the 13th International Emission Inventory Conference, Clearwater, FL (8–10 June). Henderson, J. Vernon (1996) “Effects of Air Quality Regulation,” American Economic Review 86: 789– 813. Ho, Mun S., et al. (2008) “The Impact of Carbon Price Policies on U.S. Industry.” Resources for the Future, RFF DP 08-37, Washington, DC. Jaffe, Adam B., et al. (1995) “Environmental Regulation and the Competitiveness of U.S. Manufacturing: What Does the Evidence Tell Us?” Journal of Economic Literature 33: 132– 63. Jeppesen, Tim, et al. (2002) “Environmental Regulations and New Plant Location Decisions: Evidence from a Meta- Analysis,” Journal of Regional Science 42: 19– 49. Levinson, Arik, and M. Scott Taylor (2008) “Unmasking the Pollution Haven Effect,” International Economic Review 49: 223– 54. McKinsey and Company, and Ecofys (2006) “EU ETS Review: Report on International Competitiveness,” http://ec.europa.eu/clima/policies/ets/docs /report_int_competitiveness_20061222_en.pdf (accessed 6 January 2013). Morgenstern, Richard, et al. (2001) “The Cost of Environmental Protection,” Review of Economics and Statistics 83: 732–38. Office of Technology Assessment (1985) Strategic Materials: Technologies to Reduce U.S. Import Vulnerability. Washington, DC: U.S. Congress. Rausch, Sebastian, and Matthew Mowers (2012) “Distributional and Efficiency Impacts of Clean and Renewable Energy Standards for Electricity.” MIT Joint Program Report #225, Cambridge, MA. Reinaud, Julia (2005) “Industrial Competitiveness Under the European Union Emissions Trading Scheme.” International Energy Agency, IEA Information Paper, Paris. Rutledge, Gary L., and Christine R. Vogan (1994) “Pollution Abatement and Control Expenditures 1972–1992,” Survey of Current Business 74: 36– 49. Schmalensee, Richard (1993) “The Costs of Environmental Protection,” in M. B. Kotowski, ed., Balancing Economic Growth and Environmental Goods. Washington, DC: American Council for Capital Formation. United Nations Statistics Division (n.d.) “United Nations Comtrade Database,” http://comtrade.un.org/db/ (accessed 6 January 2013). U.S. Census Bureau (multiple dates) Annual Survey of Manufactures. Various editions. Washington, DC: U.S. Census Bureau. U.S. Department of Energy (2007) Aluminum Industry Profile. Washington, DC: Office of Energy Efficiency and Renewable Energy. U.S. Energy Information Administration (2008) Energy Market and Economic Impacts of S.2191, the Lieberman-Warner Climate Security Act of 2007. Washington, DC: U.S. Department of Energy. ———. (2011) Annual Energy Review 2011. Washington, DC: Department of Energy.

88

Joseph E. Aldy and William A. Pizer

———. (multiple dates) Manufacturing Energy Consumption Survey. Various editions. Washington, DC: U.S. Energy Information Administration. U.S. Environmental Protection Agency (2007) Energy Trends in Selected Manufacturing Sectors. Washington, DC: U.S. Environmental Protection Agency. ———. (2009) EPA Analysis of the American Clean Energy and Security Act of 2009 H.R. 2454 in the 111th Congress, http://www.epa.gov/climatechange/Downloads /EPAactivities/HR2454_Analysis.pdf (accessed 6 January 2013). ———. (2010) EPA Analysis of the American Power Act in the 111th Congress, http://www.epa.gov/climatechange/Downloads/EPAactivities/EPA_APA _ Analysis_6-14-10.pdf (accessed 6 January 2013). ———. (2011) Regulatory Impact Analysis for the Federal Implementation Plans to Reduce Interstate Transport of Fine Particulate Matter and Ozone in 27 States; Correction of SIP Approvals for 22 States. Washington, DC: U.S. Environmental Protection Agency.

Case Cited EME Homer City Generation v. EPA, No. 11-1302 (D.C. Cir. Aug. 21, 2012).

Legislation Cited American Power Act, Discussion Draft, 111th Congress (2010), http://kerry .senate.gov/imo/media/doc/APAbill3.pdf (accessed 2 January 2012). Raise Wages, Cut Carbon Act of 2009, H.R. 2380, 111th Congress (2009). Waxman-Markey Bill, H.R. 2454, 111th Congress (2009).

Chapter 5

Environmental Regulatory Rigidity and Employment in the Electric Power Sector Rolf Färe, Shawna Grosskopf, Carl A. Pasurka, Jr., and Ronald J. Shadbegian

In recent decades, there has been recurring interest in the extent to which implementing environmental regulations to reduce bad outputs (for example, sulfur dioxide, or SO2 emissions) adversely affects the ability of an economy to produce its marketed good outputs (see Pasurka 2008). This concern has emerged because with a fi xed technology and fi xed inputs (for example, labor and capital), reductions in bad outputs are achieved at the cost of reduced production of good outputs (for example, electricity generation). A recent variation of this topic asks about the effect of environmental regulations on employment. When a government implements a tariff reduction or increases the stringency of regulations on bad output production, there is a reallocation of resources within a society. For example, a reduction in tariffs results in a reallocation of inputs from the import- competing industry to the export industry. In addition, more stringent environmental restrictions can result in a reallocation of inputs within an establishment (from good output production to pollution abatement), among establishments within an industry, and among industries within an economy. In a long-run model, economists assume that increasing the stringency of environmental regulations has no net effect on aggregate employment.1 When assessing the welfare and distributional consequences of a policy, a long-run analysis assumes that a plant will move from one point on its production possibilities frontier (PPF) to another point on the production frontier. However, there will still be changes in employment

90

Färe, Grosskopf, Pasurka, and Shadbegian

among plants within an industry and among industries in an economy. Although this perspective assumes there is no long-run unemployment associated with the policy, 2 nevertheless, there will be short-run adjustment costs borne by capital and labor. The long-run frontier assumes no rigidities in the economy. However, if there are short-run rigidities that affect labor mobility— for example, wages are rigid in a downward direction or there is a lack of input mobility (due to industry- specific skills)—then the adjustment path to reduced bad output production involves short-run unemployment that results in production occurring inside the production frontier. In the present volume, Masur and Posner (this volume) express interest in quantifying the adjustment costs associated with these rigidities. Regulations aimed at reducing bad output can take two forms: command- and- control (CAC) or market-based regulation. CAC regulations are prescriptive in nature and typically designate how much an individual plant is allowed to emit or what types of control equipment it must use to meet such requirements. Conversely, market-based regulation creates an incentive for the private sector to incorporate pollution abatement into production decisions and to innovate in such a way as to continually search for the least costly method of abatement. Marketbased regulation can take several forms, including a cap- and-trade (or tradable permits) program in which the government sets the level of aggregate emissions (for example, of SO2) and emission allowances are then distributed to polluters who may buy or sell them. More flexible than CAC, market- based regulations capitalize on the heterogeneity of abatement costs across polluters to reduce aggregate pollution effi ciently. In this chapter, we introduce a new rigidity via CAC regulations—the lack of mobility (achieved by a cap- and-trade program, for example) in bad output production among plants when bad output production is regulated. Unlike Masur and Posner (this volume), our model does not attempt to estimate the adjustment costs associated with short-run unemployment due to changes in environmental regulations. Instead, we focus on the inefficiency (overemployment) associated with bad output immobility. When analyzing the effect of pollution abatement on good output production, we model the environmental technology to illustrate the joint production of good and bad outputs when production of the bad output is regulated.3 An important advantage of joint production models is that they do not require information on which input costs should be assigned to good output production and which input costs are assigned to pollution abatement (see Pasurka 2013).4

Environmental Regulatory Rigidity and Employment

91

In order to assess the cost of regulations on bad output production, it is necessary to model both the regulated and unregulated technologies. For the regulated technology, it is costly for a plant to reduce its bad outputs, while the unregulated technology allows the plant to freely dispose (that is, ignore) its bad outputs. In Figure 5.1, the regulated technology is represented by the production frontier, 0abcdefG, which is the environmental technology.5 For a fi xed technology and vector of inputs, the reduction in bad output production, which is the output of pollution abatement, comes at the cost of reduced good output production.6 In Figure 5.1, point f represents the maximum production of the good output if the industry is allowed to freely dispose of its bad output. Hence, Hf represents the unregulated PPF.7 Identifying the good output reduction associated with bad output immobility requires us to model two regulated technologies. In addition to the frontier that depicts the trade- offs under a perfectly functioning y (good)

H

f

e d c

k

b a

j

P(x)

0

G

b (bad)

Figure 5.1. Command- and- Control and Tradable Permit Production Frontiers (Good Output- Based Perspective).

92

Färe, Grosskopf, Pasurka, and Shadbegian

tradable permit (TP) system, we specify a CAC frontier that is constructed assuming plants are restricted to their observed production of bad outputs. Unlike the TP PPF, which assumes that bad output production is mobile, the CAC PPF assumes that bad output production is immobile among plants. The effect of bad output immobility among plants on the industry PPF is depicted in Figure 5.1. The TP PPF (0abcdefG) represents the industry frontier when bad output production is mobile among plants. The CAC PPF (0jckfG) represents the industry frontier when bad output production is immobile among plants. The industry can produce different combinations of good and bad outputs along the two frontiers. The TP (0abcdefG) and CAC (0jckfG) frontiers have three points in common. First, the two PPFs coincide when no bad output is produced— point 0. Second, the PPFs overlap when the bad output is unregulated—point f. Finally, the two PPFs coincide at point c—the original point when there is no incentive to reallocate bad outputs via trades and move from the observed allocation of bad outputs. Otherwise, the CAC frontier lies inside the TP frontier. For a fi xed technology and input vector, bad output immobility reduces good output production. From this perspective, immobility of bad output production among plants has the same qualitative effect on the production frontier as immobile labor. Färe et al. (forthcoming) calculated foregone good output production due to imperfect bad output mobility under the SO2 tradable permit program. They identified the reduced good output by calculating the difference between good output production for a sample of coalfired electric power plants (EPPs) when bad outputs were perfectly mobile and when bad output production was immobile. Hence, they were comparing good output production for the TP frontier with good output production for the CAC frontier. Therefore, the Färe et al. (forthcoming) model represents the (good) output-based perspective on the cost of regulatory rigidity. This chapter also focuses on the costs of bad output immobility resulting from regulations on bad outputs produced by an EPP. However, instead of an output-based technology, we specify a labor-based technology to calculate the employment effects of regulatory rigidity.8 Instead of maximizing good output production and identifying potential increases in good output production, we minimize labor use and identify regulatory rigidity via potential reductions in employment. This chapter represents the first use of a model of the joint production of good and bad outputs to calculate the employment effects of regulatory rigidity. After specifying production technologies with and

Environmental Regulatory Rigidity and Employment

93

without bad output mobility, we operationalize our model using a sample of coal-fired electric power plants from 1995 through 2005.

The Model The next step in our analysis uses the unregulated and regulated technologies to determine employment associated with different levels of bad output mobility. For this labor- based approach, we construct unregulated and regulated isoquants and determine the minimum level of labor necessary for each EPP to produce its observed good and bad outputs. All isoquants, whether they reflect the unregulated or regulated technology, are drawn for a given level of good output production. In addition, the regulated isoquants— both CAC and TP— are constructed for a given level of bad output production. In order to identify the decrease in labor demand associated with bad output mobility, we model two regulated technologies— the TP and the CAC. The difference in employment between the CAC regulated technology and the unregulated technology is the employment change associated with observed pollution abatement efforts. Comparing the results when bad outputs are immobile (CAC) with the results when bad outputs are mobile (tradable permits) among EPPs provides a mea sure of the employment effect of regulatory rigidity. For the unregulated technology, there is a fixed amount of capital, labor, and fuel available to each EPP. Our model calculates the minimum (that is, technically efficient) level of labor required for each EPP to operate on its unregulated isoquant. Hence, the model calculates an EPP’s minimum amount of labor that combined with its observed capital and fuel consumption allows the EPP to produce at least its observed level of good output. For the CAC isoquant, bad output production is immobile among EPPs. Our model calculates the minimum level of labor required for each EPP to operate on the isoquant associated with its observed output. Hence, the model seeks the minimum amount of labor that combined with its observed capital and fuel consumption would allow the EPP to produce at least its observed level of good output and produce no more than its observed level of bad output. The summation of the efficient level of labor of each EPP yields the efficient level of industry employment under CAC. The difference in minimum employment between the unregulated and CAC isoquants represents the change in employment associated with the observed regulatory strategy. The increase in labor required to reduce bad output production, while maintaining the original good output production, corresponds to the cost

94

Färe, Grosskopf, Pasurka, and Shadbegian

effect set forth by Morgenstern et al. (2002) and Morgenstern (this volume).9 For the TP isoquants, we allow bad output production to be mobile among the EPPs in the sample, while holding the total bad output production of all EPPs in the sample to be no greater than the observed level of industry bad output production. Our model calculates the minimum level of labor required for the industry to be on its regulated isoquants. Hence, the model determines the minimum amount of labor for each EPP that combined with the observed capital and fuel consumption of each EPP will allow the industry to produce at least the observed level of good output and produce no more than the observed level of bad output of the industry. The difference in employment between the unregulated and TP isoquants represents the change in employment associated with implementing a perfectly functioning tradable permit system. The labor-based perspective is demonstrated by the isoquants in Figure 5.2. The TP isoquants represent efficient input combinations when EPPs are allowed to reallocate bad output production, while the CAC isoquants represent input combinations when EPPs are not allowed to reallocate bad output production. Isoquant CAC A = TPA represents the case when there is no incentive to reallocate bad outputs via trades. Isoquant CACC = TPC represents the unregulated technology (that is, bad output production can be ignored). For all other output vectors, the CAC isoquant lies above the TP isoquant. Hence, CACB requires more labor than TPB. Point m, which lies above the isoquants, represents an inefficient EPP. The linear programming (LP) models allow us to calculate the difference shown by the dashed vertical line between point m and the “isoquants.” The distance represents the potential reduction in labor use by the EPP, while producing the observed output vector. Formal specifications of the CAC (Model 1) and TP (Model 2) abatement strategies can be found in the Appendix to this chapter. Because no factor mobility is permitted, these models depict the very short run. In Model 1, EPPs are not allowed to find a more efficient allocation of bad output production than their observed level of bad output production. In Model 2, EPPs are allowed to find the most efficient allocation of bad output production. Hence, Model 1 can be viewed as mea sur ing the technical efficiency of EPPs subject to a CAC regulatory system, whereas Model 2 mea sures the technical efficiency of EPPs in a regulatory system with tradable permits. If a CAC regulatory system exists, Model 1 identifies the minimum level of labor for each EPP under the CAC system, and Model 2 calculates the potential employment if an efficient TP system is implemented. If a TP system exists, Model 1 identifies the minimum level of labor for each EPP with the observed level of bad outputs produced under the TP

Environmental Regulatory Rigidity and Employment

95

Labor

• m CACA = TPA CACB TPB

CACC = TPC

0

Nonlabor Inputs

Figure 5.2. Command- and- Control and Tradable Permit Isoquants (LaborBased Perspective).

system, and Model 2 calculates the potential minimum employment under the existing TP system by eliminating rigidities that affect bad output mobility. The difference in minimum employment between the CAC and TP represents the employment changes associated with regulatory rigidity. For a fi xed technology and output vector, less flexible (that is, less efficient) regulations lead to increased labor demand. Therefore, we anticipate that the minimum levels of employment will be smallest for the unregulated frontier and highest for the CAC model.

Empirical Example: Background, Data, and Results Title IV of the 1990 Clean Air Act Amendments significantly changed the way EPPs were regulated in the United States. Before Title IV, EPPs

96

Färe, Grosskopf, Pasurka, and Shadbegian

were regulated via a CAC regime that set the allowable sulfur content of the coal used at each individual EPP. Title IV created a cap-and-trade program that put a cap on total SO2 emissions, allocated allowances among EPPs equal to that cap, and allowed EPPs to freely trade these allowances among their own units, to sell them to other EPPs, or to bank them for future use. The only requirement faced by an EPP under the trading program is that at the end of the year it must hold one allowance for each ton of SO2 emitted that year. Thus, the allowance trading program established by Title IV provides much greater flexibility to achieve any given emission standard because EPPs that face high marginal abatement costs may purchase SO2 permits from EPPs that face lower marginal abatement costs rather than having to install expensive pollution abatement equipment (for example, flue- gas desulfurization) or use more expensive inputs (for example, low sulfur coal). The ultimate goal of Title IV was to reduce total SO2 emission levels to roughly 9 million tons by 2010, about half of the 1980 level. The reduction was accomplished in two phases. Phase I (1995–1999) targeted the dirtiest 110 power plants (with 263 generating units). These generating units, called the Table A units, were required to decrease their collective SO2 emissions to 7.2 million tons per year starting in 1995, 6.9 million tons per year in 1996, and then 5.8 million tons per year from 1997 to 1999. Phase II, which started in 2000, required the smaller generators— generators that have an output capacity between 25 and 100 megawatts— to join the cap- and-trade system.10 In addition to compelling the smaller and cleaner units to join the cap- and-trade program, the Table A units were obligated to make further reductions in their SO2 emissions—reducing their overall emissions by another 3.4 million tons, down to 2.4 million tons by 2010. To analyze the effects of EPP regulation under the Clean Air Act, we implement the joint production model developed in the previous section of this chapter. We solve the LP problems by using a balanced panel data set of 80 coal-fired power plants from 1995 to 2005. Table 5.1 presents summary statistics of the data for 1995 and 2005.11 The sample size is dictated by the data available to produce a balanced panel for 1995–2005. It should be noted that these 80 coal- fired power plants are not necessarily representative of the Phase I and II coal-fired power plants that were regulated by Title IV; therefore the results presented here should be taken only as illustrative. The technology modeled in this study consists of one good output, “net electrical generation” (kilowatt hours), and one bad output, SO2. The inputs are the capital stock (in constant dollars), the number of employees, and the heat content (in Btu) of coal, oil, and natural gas consumed at each power plant (there

Environmental Regulatory Rigidity and Employment

97

Table 5.1. Summary Statistics (80 coal-fired power plants)

1995 Electricity SO2 Capital stock Employees Heat content of coal Heat content of oil Heat content of gas 2005 Electricity SO2 Capital stock Employees Heat content of coal Heat content of oil Heat content of gas

Units

M

Sample SD

Maximum

kWh (in millions) short tons (in thousands) dollars (in millions, 1973 $) workers Btu (in billions) Btu (in billions) Btu (in billions)

5,711 40

4,866 40

20,222 192

167 2

290

195

863

57

214 57,064 109 93

136 47,174 116 284

578 193,574 514 2,083

42 2,255 0 0

6,647 34

5,249 33

22,338 186

176 1

332

230

1,009

59

172 66,877 108 71

104 51,036 129 157

468 215,802 738 911

28 2,297 0 0

kWh (in millions) short tons (in thousands) dollars (in millions, 1973 $) workers Btu (in billions) Btu (in billions) Btu (in billions)

Minimum

are separate constraints for each of the fuels). The power plants may consume coal, oil, or natural gas; however, in order to model a homogeneous production technology, coal must provide at least 95 percent of the Btu of fuels consumed by each plant in each year. FERC Form 1 is the source of labor and capital data for private electric power plants (Federal Energy Regulatory Commission 2012), while the EIA- 412 survey is the source of labor and capital data for public power plants (U.S. Energy Information Administration n.d.).12 The U.S. Department of Energy (DOE) halted the EIA- 412 survey after 2003; however, the Tennessee Valley Authority voluntarily posted 2004–2006 data for its electric power plants online (Tennessee Valley Authority n.d.). The EIA767 survey (U.S. Energy Information Administration 2006) is the source of information about fuel consumption, the sulfur content of coal and oil consumed, and net generation of electricity. The SO2 emission data are collected by the U.S. EPA Continuous Emissions Monitoring System (CEMS). A number of plants consume fuels other than coal, oil, and natural gas (for example, petroleum coke, blast furnace gas, coal- oil mixture, fuel oil #2, methanol, propane, wood and wood waste, and refuse,

98

Färe, Grosskopf, Pasurka, and Shadbegian

bagasse, and other nonwood waste). In this study, any plant whose consumption of fuels other than coal, oil, and natural gas represented more than 0.0001 percent of its total fuel consumption (in Btu) is excluded. We ignore consumption of fuels other than coal, oil, and natural gas when these fuels represent less than 0.0001 percent of a plant’s fuel consumption. Although our models generate the minimum level of labor for each power plant in each year, we present only annual results for the average plant in our sample. Columns 2 and 3 of Table 5.2 report the results for a modified version of the LP problem specified in Model 1 in which power plants are allowed to ignore their SO2 emissions (that is, the unregulated technology).13 In terms of Model 1, the bad output constraint is excluded when EPPs are allowed to freely dispose of their SO2 emissions.14 The minimum employment when SO2 emissions are freely disposable ranges from 37.5 percent of the observed employment in 1995 to 46.5 percent in 2005, with an average of 41.9 percent during the years 1995–2005.15 Columns 4 and 5 report the results for the LP problem specified in Model 1. The difference between the minimum quantities of labor reported in Column 4 relative to Column 2 is the change in employment associated with pollution abatement under CAC. The percentage increase in minimum employment under CAC (column 4) relative to when SO2 emissions are freely disposable (Column 2) ranges between 19.0 percent in 1997 and 27.8 percent in 2004, with an average of 22.8 percent during the period 1995–2005. This increase in labor required to reduce bad output production while maintaining the original good output production corresponds to the cost effect of Morgenstern et al. (2002) and Morgenstern (this volume). Columns 6 and 7 report the results for the LP problem specified in Model 2, which allows for tradable permits. The difference between the minimum quantities of labor reported in column 6 relative to column 2 is the change in employment associated with pollution abatement with perfectly mobile bad outputs, that is, when EPPs obtain the most efficient allocation of SO2 emissions. Because SO2 permits were traded by some plants in our sample during the years 1995–1999 and all plants in our sample from 2000 to 2005, our results do not necessarily reflect the change in employment associated with a TP system. Instead, our results show the change in employment when electric power plants obtain the most efficient allocation of SO2 emissions. The increase in minimum employment under a TP system (column 6) relative to the unregulated case (Column 2) ranges from 0 percent during 1996–2001 and 2003 to 5.5 percent in 2005, with an average of 0.7 percent in the years 1995– 2005. This difference represents the opportunity cost of reducing the

Table 5.2. Employment Under Different Regulatory Strategies (Average Number of Plant Employees)

Unregulated Technology (ignore bad outputs)

Command- and- Control Technology (immobile bad outputs)

Tradable Permit Technology (mobile bad outputs)

Year

Observed Labor (1)

Minimum Labor (2)

Difference (3) = (1) − (2)

Minimum Labor (4)

Difference (5) = (1) −(4)

Minimum Labor (6)

Difference (7) = (1) − (6)

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

213.98 196.46 188.33 189.63 174.50 170.49 172.96 169.51 165.64 164.01 171.96

80.14 82.86 75.60 77.52 74.13 75.47 71.43 70.15 69.72 71.45 79.93

133.83 113.60 112.73 112.11 100.37 95.02 101.53 99.36 95.92 92.56 92.03

98.99 102.43 89.85 92.93 91.36 91.29 87.98 85.05 86.59 91.29 99.77

114.98 94.03 98.47 96.70 83.14 79.20 84.98 84.46 79.05 72.73 72.20

80.65 82.86 75.60 77.52 74.13 75.47 71.43 70.16 69.72 72.68 84.34

133.32 113.6 112.73 112.11 100.37 95.02 101.53 99.35 95.92 91.33 87.62

Arithmetic means (1995–2005) Arithmetic means (1995–2005)— capital congested

179.77

75.31

104.46

92.50

87.27

75.87

103.90

179.77

131.76

141.30

132.33

100

Färe, Grosskopf, Pasurka, and Shadbegian

bad output with an efficient TP system. One potential explanation for the small difference is that the frontiers for each year are contemporaneous frontiers. As a result, the unregulated technology is constructed from observations generated in an environment in which the bad outputs are already regulated. Hence, the unregulated frontier can be thought of as the least regulated frontier. If all plants are subject to similar regulations, the potential improvement in efficiency of a TP system in which all potential trades are executed may be relatively small when compared with the unregulated frontier. For this reason, it is not surprising that the decrease in the minimum quantity of labor under a TP system (Column 6) relative to the CAC case (Column 4) ranges from 15.5 percent in 2005 to 20.4 percent in 2004, with an average decrease of 18.0 percent in the years 1995–2005. The lower minimum employment represents the reduced employment associated with the movement to a system with no rigidities in bad output mobility. This difference indicates the reduction in labor demand— relative to the observed labor use—if a perfectly efficient TP system could be implemented. Using the same data set used in this chapter, Färe et al. (forthcoming) modified Model 1 and Model 2 to calculate the difference in good output production between the TP technology and CAC technology, which is the foregone good output associated with bad output (that is, SO2) immobility. Whereas we find an average annual decrease of 18 percent in employment associated with perfect bad output mobility in 1995–2005, Färe et al. (forthcoming) found a 1.94 percent average annual increase in good output production associated with perfect bad output mobility in 1995–2005.

Conclusion Although stricter environmental SO2 regulations reduce bad output production, improve ambient air quality, and improve the quality of life for the citizens of a society, the costs resulting from the policy remain an important consideration. In this chapter, we have outlined how improved efficiency in a regulatory system achieved with market-based regulatory instruments will reduce labor requirements in the regulated sector when compared with CAC. By increasing the efficiency of regulations by implementing a TP system, we demonstrated that it is possible to produce the same level of good and bad outputs with less labor. Failure to execute all beneficial trades results in a less efficient outcome with either lower good output production for a given input vector or more labor required to produce a

Environmental Regulatory Rigidity and Employment

101

given output vector. It is not surprising that the model with the least mobility of bad output production requires the most labor (that is, is the least efficient). Although a more efficient regulatory system reduces labor demand, the labor previously employed is available to produce more of the good output or be used for additional reductions in bad output production. In addition to modeling the effects of environmental regulatory rigidities, our model can be modified to calculate short- run and longrun pollution abatement costs— the foregone good output associated with pollution abatement— resulting from short-run rigidities that limit mobility in the labor market (see Masur and Posner this volume). By comparing the short-run and long- run results, it is possible to determine the adjustment costs associated with labor immobility among establishments within an industry or among industries in the economy. The model presented here could be modified to calculate the factorshift effect by observing changes in employment resulting from changes in labor intensities over time. In order to mea sure the demand effect of regulations on employment (see Morgenstern et al. 2002; Morgenstern this volume), it is necessary to extend our model to allow input mobility to respond to changes in demand for the output of the electric power industry.

Appendix The Activity Analysis Model In order to implement our model, we specify the environmental technology. This technology assumes weak disposability and null- jointness of the outputs. The latter concept tells us that producing good outputs requires producing bad outputs as well. We assume there are k = 1, . . . , K observations (EPPs) of inputs and outputs. Inputs other than labor are denoted by x 
N+ , and labor is L ≥ 0. The good or desirable output is y ≥ 0, and the bad or undesirable output is b ≥ 0. Thus, for each EPP we have a vector (xk , Lk , yk , bk), with k = 1, . . . , K observations. Our production model is represented by its output set P(x, L) = {(y, b): (x, L) can produce (y, b)} that is expressed as an Activity Analysis or Data Envelopment Analysis (DEA) model, namely

102

Färe, Grosskopf, Pasurka, and Shadbegian K


z y

k k


y,

k=1 K

z b

k k


b,

k=1 K

z x

k kn


xn ,

n = 1, . . . , N,

k=1 K

z L k

k


L,

k=1

zk
0 ,

k = 1, . . . , K.

For each combination of nonlabor and labor inputs (x, L), the output set P(x, L) is the combinations of good and bad outputs (y, b) that can be produced by that vector. The intensity variables zk are the weights assigned to each of the K observations that are used to construct the isoquant. This model has a good output that is strongly disposable together with the good and bad outputs (y, b) being jointly weakly disposable.16 This is formally stated as (y, b) ∈ P(x, L), y′ ≤ y ⇒ (y′, b) ∈ P(x, L), which imposes strong disposability on the good output, and (y, b) ∈ P(x, L), 0 ≤ θ ≤ 1 ⇒ (θy, θb) ∈ P(x, L), which imposes weak disposability on the good and bad outputs. Imposing weak disposability on the good and bad outputs allows us to model the opportunity cost of reducing the bad outputs (that is, the quantity of good output that must be foregone).17 This allows us to model the technology when bad outputs are regulated. All inputs are strongly disposable, and the model exhibits constant returns to scale, since the zk is just nonnegative. The specification of the bad output constraint, which was introduced in Färe et al. (forthcoming), yields a P(x, L) that is not bounded. The model can be easily adjusted to account for P(x, L) not being bounded. The advantage of this specification of the bad output con-

Environmental Regulatory Rigidity and Employment

103

straint is that it restricts the shadow price of the bad output to be nonpositive. We operationalize our model by specifying a modified input-based environmental technology. Unlike the traditional input- based DEA model that identifies technical inefficiency of an EPP by measur ing its potential percentage reduction in all inputs while maintaining its observed production of good and bad outputs, our modified model identifies technical inefficiency by calculating the minimum level of labor that combined with the observed levels of nonlabor inputs (capital and fuels) allows an EPP to maintain its observed production of good and bad outputs. Model 1: Labor Requirement with Command-and- Control Model 1,18 which is the CAC version of our model, determines the minimum level of labor required by each power plant needed to produce the observed level of good and bad outputs with the observed level of nonlabor inputs (capital, coal, oil, and natural gas). The Model 1 LP problem is equivalent to specifying a separate LP problem for each EPP that minimizes its labor use with its observed level of nonlabor inputs (capital, coal, oil, and natural gas) and produces its observed level of good and bad outputs. minL1 + ⋅ ⋅ ⋅ +LK

(A.2)

EPP 1 K

s.t.

∑z y

1 k k

≥ y1 ,

k=1 K

∑z b

1 k k

≤ b1 ,

k=1 K

∑z x

1 k kn

≤ x1n ,

n = 1, . . . , N,

k=1 K

∑z L

1 k k

≤ L1 ,

k=1

z k1 ≥ 0, . . .

k = 1, . . . , K.

104

Färe, Grosskopf, Pasurka, and Shadbegian

EPP K K

∑z

K y k k

≥ yK ,

K b k k

≤ bK ,

k=1 K

∑z k=1 K

∑z

K x k kn

≤ x Kn ,

n = 1, . . . , N,

k=1 K

∑z

K L k k

≤ LK ,

k=1

z kK ≥ 0,

k = 1, . . . , K.

Model 2: Labor Requirement with Tradable Permits Model 2 is the TP version of our model that determines the minimum level of labor required by each EPP to produce the observed level of good and bad outputs with the observed level of nonlabor inputs.19 Unlike Model 1, Model 2 allows both labor and the bad outputs to be variable. However, total bad output production must not exceed the observed level of bad output production by the industry. minL1 + ⋅ ⋅ ⋅ +LK

(A.3)

EPP 1 K

s.t.

∑z y

1 k k

≥ y1 ,

k=1 K

∑z b

1 k k

≤ b1 ,

k=1 K

∑z x

1 k kn

≤ x1n ,

n = 1, . . . , N,

k=1 K

∑z L

1 k k

≤ L1 ,

k=1

z k1 ≥ 0, . . .

k = 1, . . . , K.

Environmental Regulatory Rigidity and Employment

105

EPP K K

∑z

K y k k

≥ yK ,

K b k k

≤ bK ,

k=1 K

∑z k=1 K

∑z

K x k kn

≤ x Kn ,

n = 1, . . . , N,

k=1 K

∑z

K L k k

≤ LK ,

k=1

z kK ≥ 0,

k = 1, . . . , K.

Aggregate Bad Output b 1 + ⋅ ⋅ ⋅ +b K ≤

K

∑b . k

k=1

Note that when we impose weak disposability on capital (that is, model the existence of capital congestion), the constraint on capital in Model 1 and Model 2 is changed from “≤” to “=”.

Acknowledgments We wish to thank David Evans for his assistance with data acquisition and Curtis Carlson for providing his capital stock and employment data. Any errors, opinions, or conclusions are the authors’ and should not be attributed to the U.S. Environmental Protection Agency.

Notes 1. A change in aggregate employment requires a change in the mix of work and leisure within the society. 2. The same assumption is maintained when conducting benefit– cost analyses. 3. Pollution abatement does not eliminate the undesirable by- products of  good output production. Instead, abatement activities transform the byproducts from— for example— one medium (air) to another medium (solid waste), where it constitutes a reduced threat to human health and the environment. 4. Färe et al. (2007a) provide a formal discussion of the environmental technology. The environmental technology is specified as a linear programming (LP) problem. Formal specification of the models can be found in the Appendix.

106

Färe, Grosskopf, Pasurka, and Shadbegian

The piecewise linear PPFs depicted in Figure 5.1 are derived from Baldwin et al. (1980). 5. Lowercase letters represent observations. 6. For example, one strategy for reducing bad outputs involves reassigning inputs from good output production to pollution abatement. 7. This depiction of the unregulated frontier, which ignores bad outputs, corrects Färe et al. (2007b). 8. With constant returns to scale, an output-based mea sure of technical efficiency that expands all outputs proportionately with a given input vector yields the same efficiency results as an input-based efficiency mea sure that contracts all inputs proportionately for a given output vector. 9. An input- based Data Envelopment Analysis (DEA) model that minimizes all inputs would exactly match the cost effect. 10. Some of these smaller generators opted into Phase I under the “substitution” and “compensation” provisions (see Ellerman et al. 1997). 11. The LP problems are calculated using GAMS/MINOS. The data and GAMS programs are available from Carl Pasurka upon request. 12. Data on the cost of plant and equipment for years prior to 1981 were published in annual reports from the Federal Power Commission and the Energy Information Administration. The Utility Data Institute (1999) provides the cost of plant and equipment data for 1981–1997. Data for (1) the cost of plant and equipment and (2) employment collected by FERC Form 1 for 1998–2005 and EIA- 423 for 1998–2003 can be downloaded from the respective websites. Although both surveys collect data on the historical cost of plant and equipment, they do not collect data on investment expenditures. Hence, changes in the value of plant and equipment reflect the value of additional plant and equipment plus the value of retired plant and equipment. For this study, we assume changes in cost of plant reflect net investment (NI), which is the same assumption employed by Yaisawarng and Klein (1994:453n30) and Carlson et al. (2000:1322). We then convert the historical cost of plant data to constant dollar values by using the Handy-Whitman Index (HWI; see Whitman, Requardt and Associates, LLP 2006). The net constant dollar capital stock (CS) for year n is calculated in the following manner: n

CSn =


HWI . t =1

NIt

t

In the first year of its operation, the net investment of a power plant is equivalent to the total value of its plant and equipment. 13. The results are sensitive to how the best-practice frontiers are constructed. In our model, the production frontier is constructed using contemporaneous technology (i.e., the frontier is constructed from observations in the current period). 14. Because plants already faced regulations prior to 1995, what we refer to as our “unregulated” technology is, in fact, a least regulated technology. 15. The last row of Table 5.2 shows the results when the constraint on capital is modified to allow the isoquants to be backward bending (i.e., capital is congested). Based on the results, congestion in capital appears to be the key factor explaining the high level of inefficiency. The existence of congestion in capital suggests the EPP employs too much capital. This can be viewed as an extreme version of the Averch-Johnson effect. If we model capital as being congested,

Environmental Regulatory Rigidity and Employment

107

the minimum employment when SO2 emissions are freely disposable averaged 73.3 percent of the observed employment in 1995–2005. In addition, the difference between columns 4 and 2, the difference between columns 6 and 2, and the difference between columns 6 and 4 are all less when we assume capital is congested. 16. (y, b) ∈ P(x, L), (x′, L′) ≥ (x, L) ⇒ (y, b) ∈ P(x′, L′) defines strong disposability of the inputs; while P(λx, λL) ⇒ λP(x, L), λ ≥ 0, defines constant returns to scale. 17. Weak disposability is the proportional reduction in all good and bad outputs. Shephard (1970) introduced this concept. 18. We use a superscript k for labor, e.g. Lk , to denote that it is a variable and that we optimize over it. 19. Here the bad output is also a variable.

References Baldwin, Robert E., et al. (1980) “Welfare Effects on the United States of a Significant Multilateral Tariff Reduction,” Journal of International Economics 10: 405–23. Carlson, Curtis, et al. (2000) “Sulfur Dioxide Control by Electric Utilities: What Are the Gains from Trade?” Journal of Political Economy 108: 1292–1326. Ellerman, A. Denny, et al. (1997) Emissions Trading Under the U.S. Acid Rain Program: Evaluation of Compliance Costs and Allowance Market Per for mance. Cambridge, MA: MIT Center for Energy and Environmental Policy Research. Färe, Rolf, et al. (2007a) “Environmental Production Functions and Environmental Directional Distance Functions,” Energy 32: 1055– 66. Färe, Rolf, et al. (2007b) “Pollution Abatement Activities and Traditional Productivity,” Ecological Economics 62: 673– 82. Färe, Rolf, et al. (forthcoming) “The Impact of CO2, NOx, and SO2 Regulation on Electricity Production.” Resource and Energy Economics. Federal Energy Regulatory Commission (2012) “Form 1—Electric Utility Annual Report,” http://www.ferc.gov/docs-filing/forms/form-1/data.asp (accessed 7 January 2013). Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry-Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. Pasurka, Carl A. (2008) “Pollution Abatement and Competitiveness: A Review of Data and Analyses,” Review of Environmental Economics and Policy 2: 194–218. ———. (2013) “Adventures in Modeling Good and Bad Outputs: A 30-Year Retrospective.” in Aaron Mamula and John Walden, eds., Proceedings of the National Marine Fisheries Ser vice Workshop (Santa Cruz, CA (June 11–12, 2012), NOAA Technical Memorandium NMFS, NOAA-TM-NMFS- SWFSC- 503. Shephard, Ronald W. (1970) Theory of Cost and Production Functions. Princeton, NJ: Princeton University Press. Tennessee Valley Authority (n.d.) “Financial Reports,” http://www.tva.gov /fi nance/reports/index.htm (accessed 7 January 2013). U.S. Energy Information Administration (2006) “Form EIA-767 Historical Data Files,” http://www.eia.gov/cneaf/electricity/page/eia767/ (accessed 7 January 2013).

108

Färe, Grosskopf, Pasurka, and Shadbegian

———. (n.d.) “Form EIA- 412 Data,” http://www.eia.gov/cneaf/electricity /page/eia412.html (accessed 7 January 2013). Utility Data Institute (1999) North America Electric Power Data. CD- ROM. New York: McGraw-Hill. Whitman, Requardt and Associates, LLP (2006) “The Handy-Whitman Index of Public Utility Construction Costs: Trends of Construction Costs,” Bulletin No. 163, Baltimore, MD. Yaisawarng, Suthathip, and J. Douglass Klein (1994) “The Effects of Sulfur Dioxide Controls on Productivity Change in the U.S. Electric Power Industry,” Review of Economics and Statistics 76: 447– 60.

Analytics

This page intentionally left blank

Chapter 6

Toward Best Practices: Assessing the Effects of Regulation on Employment Lisa A. Robinson

A poorly performing economy, accompanied by congressional proposals to curb regulation and new research on regulation’s effects on employment, has led to increased interest in incorporating employment impacts into regulatory analysis. Both the extent to which such impacts are estimated and the approaches used to estimate them vary greatly due in part to a lack of consensus on how to best conduct such assessments.1 As a result, there is a growing need to develop analytic approaches that are consistent with a welfare economics framework, reflect well- conducted empirical research, and meet the information needs of policymakers and other stakeholders. Such “best practices” are typically identified through careful review of theory, previous analyses, and scholarly research, as well as substantial discussion among experts. This chapter takes an initial step toward identifying best practices for analyzing the impacts of regulations on employment, proposing general principles for further consideration by both regulatory analysts and policymakers. Under Executive Order 12866 (Clinton 1993), as supplemented by Executive Order 13563 (Obama 2011), federal agencies are required to assess the costs and benefits of a regulatory action if it is expected to be “significant.” They must also assess regulatory and nonregulatory alternatives to the proposed action if it is expected to be “economically significant.”2 Executive Order 12866 defines both costs and benefits broadly, including adverse effects on employment as one of several impacts to be considered. Executive Order 13563 lists the promotion of job creation as one of the goals of the U.S. regulatory system. Despite these executive orders, until recently employment effects were primarily considered temporary transitional impacts and received relatively little attention in regulatory analyses. The U.S. Office

112

Lisa A. Robinson

of Management and Budget’s (OMB 2003) Circular A-4 provides general government- wide guidance for regulatory analysis but does not specifically address employment impacts. Federal agencies have now started to investigate how to best assess the effects of regulation on jobs, given continuing high unemployment rates. In its Draft 2012 Report to Congress (OMB 2012), OMB solicited public comment on whether and how to conduct employment impact analysis, encouraging agencies to assess such impacts while recognizing that such analysis is difficult and complex. The U.S. Environmental Protection Agency (EPA) is the only federal agency that has developed more comprehensive guidance, which is not surprising given that it promulgates a substantial proportion of the regulations subject to the executive orders’ analytic requirements. EPA’s 2010 economic analysis guidelines follow the usual convention of separating assessment of social costs (as part of the benefit– cost analysis) from assessment of their distribution (in an economic impact analysis) (EPA 2010). Employment effects are primarily discussed as part of the latter, consistent with the standard assumption of full employment. Under this assumption, any unemployment associated with a new regulation is expected to be temporary as workers transition to new jobs and wages adjust as needed. Thus unemployment is currently treated largely as a short-term distributional impact, related to who bears the costs of a regulation as labor shifts across industries and geographic areas, not as a significant determinant of a regulation’s net benefits. EPA reserves a yet-to-be- completed appendix for discussion of how to address structural unemployment in benefit– cost analysis. In this chapter, I recommend consideration of nine principles that fit into the existing framework for regulatory analysis and could be feasibly implemented more broadly in the near term. I narrowly focus on job creation and job loss (both short-term and long-term) but note that it is also important to consider other characteristics of employment (such as effects on wage levels, quality of work, and the number of people seeking work), other micro- and macroeconomic mea sures (such as returns to capital, location decisions, incentives for innovation, and overall production), and nonmarket outcomes (such as improved health and environmental quality). Before turning to the recommendations, I discuss some key definitions and conceptual distinctions.

Some Definitions and Distinctions My recommendations refer to several key terms and concepts that are not always used consistently, so I offer the following definitions to help clarify what follows.

Toward Best Practices

113

Benefits Versus Costs The distinction between “benefits” and “costs” is somewhat arbitrary, because decreased benefits can be categorized as costs and decreased costs can be categorized as benefits. I use the term “costs” to refer to the reallocation of real resources associated with regulatory compliance, including direct compliance expenditures, any offsetting savings, and resulting impacts on market supply and demand. The value of a real resource is determined by its opportunity cost (its value in its best or most beneficial use), often referred to as its “social” cost. Thus the costs of a regulation are the net opportunity costs of forgone goods and services that result when regulatory compliance reallocates resources away from what would otherwise be produced and consumed. I use the term “benefits” to refer to the purpose for which such resource reallocation is required— the goals the regulation is intended to achieve. For regulations subject to the executive order analytic requirements, these goals often include improving human health or safety or environmental quality. In some cases, these outcomes can be valued using market data; in others, nonmarket valuation methods (based on analysis of related market goods or surveys) are needed. Benefit–Cost Analysis Versus Distributional Analysis The distinction between benefit– cost analysis and distributional analysis is more principled. As traditionally practiced, benefit– cost analysis focuses on determining which policy, if any, would be most economically efficient—that is, which policy is likely to lead to the largest net benefits on a national level. In contrast, distributional analysis focuses on how costs and benefits are allocated across different demographic groups or sectors, including any transfers between groups. Although benefit– cost analysis and distributional analysis rely on much of the same information, they differ in how the impacts are disaggregated and reported. Although both types of analysis provide important and useful information, U.S. regulatory decisions are rarely, if ever, based solely on the analytic results. Numerous other factors influence these decisions, such as the need to adhere to statutory and other legal requirements, concerns about impacts that cannot easily be quantified, and the overall policy goals of the administration. Regulatory Impact Analysis Versus Economic Impact Analysis The term “regulatory impact analysis” (RIA) is commonly used to refer to the analyses required by Executive Orders 12866 (Clinton 1993) and

114

Lisa A. Robinson

13563 (Obama 2011) and OMB’s Circular A-4 (OMB 2003). Although the benefit– cost analysis is a major component of an RIA, it also addresses other topics. The term “economic impact analysis” (EIA) usually refers to an analysis of the distribution of regulatory costs, provided as a chapter in the RIA or as a separate document. An EIA may focus on how costs are distributed across individual firms, within or across industrial sectors, or throughout the economy (EPA 2010). Currently, employment impacts are generally estimated as part of an EIA (although the term EIA is not always used) rather than in the benefit–cost analysis, because, under the assumption of full employment, the effects are expected to be largely distributional rather than involving net changes in social welfare.

Nine Proposed Principles I propose nine general principles for estimating employment impacts of regulation, meant to apply to economically significant regulations subject to the analytic requirements of the regulatory review executive orders. These principles are intended as practical advice that fits within the current framework for regulatory analysis, that can be implemented more broadly in the near term at a relatively low cost without substantial additional research. Given space constraints, I reference other sources that provide more detailed, technical information. The ability to quantify and value employment impacts is currently significantly constrained by the need for more primary research; as the research base evolves, these recommendations should be revisited and refined.3 For analyses of employment impacts to be useful for decision making, agencies’ documentation must be transparent. It must clearly explain the data, analytic methods, and results, as well as the strengths and limitations of the overall analysis. The documentation must also explore related uncertainties (including nonquantified effects) both qualitatively and quantitatively and discuss the implications for decision making.4 Such transparency and recognition of uncertainty is particularly important given the substantial challenges associated with assessing employment impacts. Recommendation 1: Address employment impacts when salient, tailoring the analysis to their likely significance. The purpose of regulatory analysis is to inform decision makers and the general public about the relative merits of alternative interventions (including no action). To be useful, the analysis must provide information that is salient, addressing impacts of concern to decision makers

Toward Best Practices

115

and other interested parties. At a time when unemployment rates are high and numerous bills to curb regulation are being considered by Congress, assessing the number of jobs that may be gained or lost seems essential. A finding that employment impacts are likely to be negligible may be just as important as a finding that they may be significantly negative or positive. When unemployment rates are low, these impacts may be smaller and less important to the debate, reducing the need for such analysis. The analysis of employment impacts need not be extensive. As discussed under Recommendation 2, simple screening may suffice for many regulations. If screening analysis indicates that the regulation is not likely to noticeably affect employment, all that may be needed is a short statement that summarizes the supporting evidence. Where more substantial impacts are expected, the next step will often involve estimating the number of jobs gained or lost and describing their characteristics as part of an EIA. When quantifiable job losses or gains are likely to be significant, analysts should also consider whether and how to value related impacts in the benefit– cost analysis, as discussed under Recommendation 6. Recommendation 2: Use screening analysis to determine the need for detailed assessment and to guide next steps. Regulatory analyses are usually conducted under tight time and budget constraints, and they require thinking carefully about where best to target analytic resources. Screening analysis is an important tool for guiding these decisions. Such analysis can involve simple comparisons; for example, analysts may compare compliance costs to total production costs to determine whether they appear large enough potentially to affect employment. It can also involve “what if” analysis that applies simple assumptions to estimate the possible size of the impacts. For example, analysts can use high- end parameter values to explore whether noticeable changes in employment are likely under extreme assumptions. A finding of negligible impacts in this case may suggest that no further analysis is needed. Although data gaps may mean that such assessments are at times largely qualitative, such exploration nonetheless may provide useful insights for decision making. A review of any relevant research literature will aid in this screening as well as support subsequent work. Because prior analyses may yield estimates of employment impacts for similar policies that could be adapted to the present context, agencies should consider developing easily searchable databases of relevant studies and data sources. They should also develop criteria for determining whether a source is of

116

Lisa A. Robinson

sufficient quality and suitable for use in a particular regulatory context, analogous to the criteria now recommended for transferring benefit values (for example, as in EPA 2010).5 For some regulations, the screening analysis will be adequate to conclude that no further work is needed. If more detailed assessment is warranted, then the screening results can be used to identify priorities for further research as well as to determine how to best allocate analytic time and resources. In some cases, better understanding of other cost or benefit categories will be more important than improving the analysis of employment impacts; in others, further exploring employment impacts will be a high priority.6 The research base is presently insufficient to provide specific indicators or thresholds that could be used to determine the extent to which detailed assessment of employment impacts is desirable. Although development of such indicators or thresholds may become possible as the research base evolves, they are likely to vary depending on the nature of the industry and the regulation as well as market conditions. In the interim, analysts should carefully describe the criteria applied in each particular case to determine whether the screening analysis is adequate or whether further assessment is desirable. Recommendation 3: Be realistic about baseline employment conditions and associated uncertainties. Regulatory impacts must be mea sured in comparison to a baseline that predicts conditions in the absence of the intervention over time. Specifying this baseline can be challenging, in part because most regulations will not be implemented immediately. Regulations are usually first proposed for public comment, then revised before final promulgation (which may take several years), and may not become fully effective until long after they are issued. Thus current employment conditions may not be good indicators of likely conditions during the transition period nor at the point when industry must fully comply with the requirements.7 Predicting near- and long-term changes in employment levels is notoriously difficult. When uncertainty in the baseline assumptions is expected to significantly affect the results, analysts should assess the regulatory options in comparison to alternative baselines. Given current economic conditions, at present it may be reasonable to consider one baseline that assumes that the economy will return to full employment by the time the regulation is implemented and a second baseline that assumes continuing high unemployment.

Toward Best Practices

117

In addition, employment conditions in the par ticu lar industries or geographic areas most directly affected by the regulation may not reflect conditions in the overall economy. These conditions may influence factors such as the characteristics of affected jobs and individuals, the number of jobs gained or lost, the duration of the transition period, the likelihood that new positions will be filled by individuals who would be otherwise unemployed, and, in the case of job losses, the likelihood that unemployment will be long-term. Such factors will need to be considered in predicting responses to the regulation as well as in describing baseline conditions, because they affect the extent to which any unemployed workers will find it difficult to transition to different jobs as well as the extent to which employers will find it difficult to fill any new jobs created. Recommendation 4: Address increases and decreases as well as net changes in employment, recognizing the complexity of the impacts. Regulatory compliance can lead to temporary or long-term increases and decreases in employment in the affected industries, in related industries, and in the overall economy.8 Some observers simply assume that regulation will increase production costs in the affected industries and that these costs will be passed onto consumers as increased prices, which will in turn lead to decreases in the quantities demanded and thereby reduce employment. This scenario is flawed, however, because it fails to consider the underlying characteristics of the affected firms and markets as well as the many ways in which they may respond to a regulation (see Morgenstern this volume). For example, Morgenstern et al. (2002) noted that industry and market conditions may limit the extent to which costs are passed onto consumers as well as the extent to which price changes affect demand. In addition, as production costs rise, more labor and other inputs may be required to produce the same output. Compliance activities also may be more or less labor intensive than pre-regulatory production. As discussed in Berman and Bui (2001), regulation is likely to have a greater impact on employment when it affects labor-intensive industries. Thus regulation may cause employment in the affected industries to increase, decrease, or remain the same. When a regulation is expected to lead to both employment gains and losses, analysts should report these increases and decreases as well as the net effect where feasible. Those who may lose their positions may differ in significant respects from those who may gain employment, affecting, for example, their ability to easily transition to new jobs. In addition, reporting only the net result masks the magnitude of the

118

Lisa A. Robinson

underlying changes: a loss of one job balanced by two jobs gained would appear the same as a loss of 10,000 jobs counterbalanced by an increase of 10,001 jobs. The likely timing and duration of job gains and losses also should be estimated to the extent possible. Employment impacts may include a short-term component, as firms make the initial investments needed for compliance, and a longer-term component, as firms undertake continued operation and maintenance and the market adjusts. In some cases, the “short”-term effects of transition may last for many years. Impacts may also vary over time as the industry becomes accustomed to the requirements and innovates. Because employment in the affected industries may decline or grow for various reasons, analysts will need to carefully consider the interactions between the regulation and other factors that influence employment trends. Compliance requirements also may indirectly affect employment in related industries. These may include, for example, those that are the suppliers or customers of the regulated industry, those that provide substitutes for or complements to the regulated industries’ products, and those that produce the goods or ser vices needed for regulatory compliance. The screening analysis discussed under Recommendation 2 can aid agencies in determining the extent to which various impacts can and should be investigated for a particular regulation and in describing the rationale for focusing on certain types of effects. Recommendation 5: Consider the effects of regulatory benefits on employment. Analyses of job losses and gains typically focus on the effects of regulatory compliance costs. Regulatory benefits may also affect employment.9 In particular, many economically significant regulations lead to substantial health benefits, enhancing productivity by reducing illnessor injury-related absences from work as well as decreased effectiveness while at work and reducing work years lost to premature mortality.10 The extent to which increased productivity will lead to changes in employment is uncertain, however. Rising productivity may increase the income of those affected and their demand for goods and ser vices, promoting overall economic growth and associated increases in employment. Yet improved productivity also decreases the number of workers required to achieve a given level of output. Better health may also lead workers to allocate more time to leisure, reducing hours worked but potentially enhancing their sense of well-being. Further review of the available research literature is needed to explore how changes in health and longevity contribute to employment.

Toward Best Practices

119

Any employment impacts associated with regulatory benefits are likely to be more diffuse than cost-related employment impacts and may affect different members of the population. For example, regulatory benefits may include small reductions in mortality risks throughout a large population, increasing its productivity, while regulatory costs may lead to layoffs at a few individual firms. Thus understanding the distribution of these impacts as well as their potential magnitude is desirable. Recommendation 6: Consider the implications of employment gains and losses for benefit– cost analysis. Incorporating the value of employment gains and losses into regulatory benefit– cost analysis involves four steps: (1) quantifying the number of jobs gained or lost; (2) characterizing the positions and individuals affected; (3) valuing the gains and losses; and (4) assessing related uncertainties. Each of these steps raises a number of complex and difficult issues, full discussion of which would require substantially more space than available here. Below I focus primarily on the third step, valuation. I first introduce the valuation framework, then discuss labor as a production input, transition costs, involuntary unemployment, and nonpecuniary effects. The screening analysis described under Recommendation 2 can be used to identify regulations with employmentrelated impacts that should be included in the benefit– cost analysis. If employment gains and losses cannot be reliably quantified (see Recommendation 4), analysts should instead discuss the potential implications qualitatively. Consistent with the focus of this book, this discussion largely addresses job gains and losses. However, it is important to recognize that regulation may affect earnings even for individuals who continue to be employed in the same positions. For example, tax interaction effects in the labor market mean that regulatory implementation may depress real wages, regardless of whether identifiable jobs are gained or lost (Goulder 2000). Valuation framework. Welfare economics provides the framework for benefit– cost analysis. Within this framework, labor can be assessed from two perspectives.11 First, it is a resource (“human capital”) used along with other resources to produce goods and ser vices. The standard  assumption is that workers are paid the value of their marginal product, so that compensation can be used to value productivity. From this perspective, loss of employment may allow skills to deteriorate and decrease human capital, while employment gains may have the opposite effect.12

120

Lisa A. Robinson

Second, labor is a source of individual satisfaction (“utility”) largely because it provides income that can be used to buy goods and ser vices. Individuals choose how to allocate their time between full- or part-time market production (paid work), nonmarket production (for example, household, child care, and volunteer work), and various leisure pursuits, based on their willingness to trade off income for the satisfaction they gain from uncompensated activities. As discussed under Recommendation 3, regulatory analysis involves comparing conditions without and with the intervention over time. Thus the value of a job created or lost due to a regulation should be compared with the value of what the individual would be doing otherwise, including paid work (part- or full- time), nonmarket work, or leisure activities— as well as job search and involuntary idleness in the case of undesired unemployment. The comparison should address the affected individuals’ predicted employment status over time without and with the regulation, rather than solely focusing on immediate or shortterm consequences. Labor as a productive input. The value of labor as an input to production is already incorporated into regulatory analysis when estimating costs and may also be reflected in benefit values. For example, if regulatory compliance requires that firms use scrubbers to reduce air pollutant emissions, then the cost estimates will include the labor used to install and maintain the scrubbers. On the benefit side, if the regulation improves human health and longevity, individual willingness to pay for these improvements would presumably reflect the impact on one’s ability to work as well as other effects (see Recommendation 5). Transition costs. Regulatory analyses often compare baseline conditions in the absence of intervention with full implementation of the regulation; that is, they are frequently static analyses of two equilibrium conditions (see Ferris and McGartland this volume). Thus they commonly exclude the transition costs associated with the movement of workers across jobs or between employment and voluntary or involuntary unemployment. These costs may include those related to job search, training, and relocation, as well as lost earnings while involuntarily unemployed (net of the value of the activities in which they are otherwise engaged). A more dynamic view would include the social costs and benefits that accrue over the transition period. The rationale for exclusion appears to be the belief that these costs are likely to be relatively small and difficult to measure. Given constrained time and resources, excluding such costs reflects analysts’ need to focus on assessing those impacts that are most important to decision makers and other stakeholders. It would be desirable, however, to review the research literature to estimate the

Toward Best Practices

121

potential range of these costs under differing conditions (including characteristics of the regulation and affected industries as well as the overall unemployment rate), so as to determine when transition costs may be significant and provide estimates that could be included more routinely in regulatory analysis. Involuntary unemployment. Regulations may lead to involuntary unemployment regardless of the overall employment rate due to mismatches between the skills and locations of affected workers. While workers with highly demanded skills may find it easy to transition to new employment regardless of overall labor market conditions, a high unemployment rate is likely to make this transition more difficult for many. The full effect of regulation-induced job gains and losses on longterm involuntary unemployment is not currently included in regulatory benefit– cost analyses. To value these effects, analysts need to first predict the extent to which new jobs are likely to be fi lled by those who would otherwise be involuntarily unemployed and the extent to which job losses are likely to lead to long-term involuntary unemployment. They would then need to determine the value of these impacts, that is, their opportunity costs mea sured using market data or nonmarket methods to estimate individual willingness to pay for the changes. Because more theoretical and empirical work on these issues is needed, in the near term any such assessment should be considered illustrative and reported in a separate section of the benefit– cost analysis, so as to highlight related uncertainties and allow readers to clearly distinguish net benefits with and without these values. Analysts also may find it useful to conduct a break- even analysis, estimating the amount that employment impacts would need to contribute to the costs or benefits of the regulation to change the relative rankings of the policy options (including no action). Decision makers and other interested parties can then review the results to consider whether employment impacts within this range seem possible. Recent research provides approaches that can be used to value employment impacts in these illustrative analyses. For policies that increase employment, Bartik (2012) proposes an approach that allows analysts to predict the extent that new jobs will be filled by those currently unemployed and to value the effects of this employment. He suggests two mea sures of value: one based on adjustment of the reservation wage (that is, the lowest wage at which a worker is indifferent between employment and unemployment) and a second based on adjustment of expected earnings. It appears that both approaches can be feasibly implemented based on available data. Bartik also provides information on associated uncertainties that should be discussed in presenting the analytic results. For policies that decrease employment, analysts could

122

Lisa A. Robinson

explore inverting the Bartik approach so as to value employment losses rather than gains. Another option is to apply estimates from research on the effects of unemployment on income, physical and emotional health, and other factors. However, as discussed in Masur and Posner (2012), most studies focus on mass layoffs, the effects of which may differ significantly from the employment effects of many regulations. In addition, these studies do not provide a method for estimating whether a particular regulation is likely to lead to short- or long-term involuntary unemployment or lead to employment of individuals who would otherwise be unemployed. Analysts should exercise extreme caution in applying these estimates, which may not be appropriate for many regulations. Nonpecuniary effects. The values individuals place on employment or unemployment may not be fully reflected in wages or other costs, due to market imperfections and other factors. Often the only way to capture these values is to use nonmarket valuation methods. These include revealed preference studies, which estimate the value of nonmarket goods based on the value of market goods, using statistical methods to control for other factors that affect their value. Alternatively, stated preference studies may be used that rely on surveys to elicit individual willingness to pay for the outcome of concern. Both methods have been used extensively to estimate the value of time use, including numerous recreation and transportation studies (Aguiar et al. 2012; Palmquist et al. 2010; Small 2012). Many of these studies include information on the value of work time relative to other activities. This literature could be reviewed to provide further insights into the value of regulation-induced employment changes. Adler (this volume) discusses related issues in more detail. Recommendation 7: Place employment impacts in context. It is very difficult to determine the policy implications of a statement that a regulation will create or eliminate X jobs. Regardless of whether employment impacts are addressed solely in a screening analysis, in an EIA, or in a benefit– cost analysis, they need to be placed in context so that decision makers understand their significance in comparison to other regulatory impacts as well as in comparison to the employment situation more generally. Information on the characteristics of those affected is also needed. Analysts should begin by clearly defining what they mean by losses or gains in “employment.” It may not be possible to define these terms consistently across analyses, due to limitations in the available research, but it is important to be clear about what is being assessed in each particu-

Toward Best Practices

123

lar case. For example, are analysts counting the number of individuals employed full- and part-time? Or are they computing full-time- equivalent employment? Are job losses and gains estimated at the plant level, firm level, or industry-wide? In what industries and geographic areas and over what time period do the estimates apply? In the case of job losses, to what extent does the analysis address the likelihood of reemployment elsewhere? If new jobs are created, does the analysis consider whether they are filled by workers previously otherwise employed or by those who were unemployed? Analysts should also provide information on the magnitude of the impacts relative to other mea sures. For example, a regulation that adversely affects employment may have other benefits that more than compensate, and a regulation that leads to job losses or gains that appear large when expressed as an absolute value may be viewed differently if compared with overall employment levels or to other policies or economic events that affect jobs. Consideration of cumulative effects across regulations may also be useful. Where job losses or gains appear significant, participants in the decision- making process are likely to be interested in more than the number of jobs affected and any related social costs or benefits. To the extent possible, analysts should provide information on job characteristics, including those that may make it particularly easy or difficult to fill new positions or for those who lose their jobs to find other employment. For example, an individual who loses a job in one industry and quickly gains similar employment nearby at the same wage rate will be of less concern than an individual who suffers long-term unemployment or takes a new job at a significantly lower wage. Other characteristics of those affected should also be discussed; in par ticular, job gains and losses among low-income or otherwise disadvantaged groups will be of greater concern than such gains and losses imposed on more affluent groups. In addition, wages and the quality of employment matter and should also be addressed.13 An analysis that assesses the effects of regulation on the overall income distribution, and on economic growth more generally, ultimately may be more informative than an analysis that considers only employment. Recommendation 8: Assess the costs and benefits of any policy changes designed to address employment impacts. If employment impacts are expected to be significant enough that the agency proposes to address them by changing the regulation or by implementing other policies, then the costs and benefits of these additional actions should be assessed as well. In cases where a regulation is

124

Lisa A. Robinson

expected to lead to job losses, it may be less costly and more effective to mitigate these effects through policies that address unemployment directly (including tax and income support as well as job search and training programs), rather than by altering a regulation that focuses on reducing environment, health, or safety risks or other goals. Similarly, policies that directly focus on job creation are likely to be a more efficient approach to increasing employment than changing regulations designed primarily for other purposes. Recommendation 9: Include employment impacts in plans for retrospective analysis. Under Executive Orders 13563 and 13610 (Obama 2011, 2012), agencies are required to conduct retrospective review of their regulations, including both their individual and cumulative impacts. These reviews are focused primarily on identifying opportunities for regulatory reform. However, retrospective analysis also can provide important insights for ex ante assessment, as illustrated by the research now used to inform analysis of employment impacts that was conducted from an ex post perspective.14 Additional retrospective assessment that addresses more recent regulatory impacts could contribute substantially to the development of improved analytic practices.

Conclusion In this chapter, I have proposed nine principles for further consideration as agencies begin to develop “best practices” for assessing the impacts of regulation on employment. All regulatory impact analyses require making some simplifying assumptions because of gaps in the research, limitations in the available modeling tools, and time and resource constraints. Many of these assumptions affect all aspects of the analysis, not solely the estimates of employment impacts. However, assessing employment impacts is particularly difficult because of the complexity of the effects and the need for additional research. At minimum, analysts should disclose the assumptions used and discuss their implications. Quantitative sensitivity analysis or probabilistic uncertainty analysis (for example, using Monte Carlo models) is useful whenever possible to supplement qualitative discussion.

Acknowledgments My views were substantially influenced by many thoughtful comments on drafts of this chapter from Don Arbuckle, Jennifer Baxter, Cary

Toward Best Practices

125

Coglianese, Leslie Genova, David Greenberg, Randy Lutter, James Neumann, Jason Price, Arden Rowell, David Weimer, and participants in the Penn Program on Regulation’s September 2012 conference, “Regulation’s Impact on Jobs: Economic Analysis and Institutional Reform.”

Notes 1. Available empirical research and approaches currently used by federal agencies to assess employment impacts are reviewed in Masur and Posner (2012) and U.S. Office of Management and Budget (2012) as well as elsewhere in this book. 2. Under Executive Order 12866, an “economically significant” regulation is one that may have annual economic impacts of $100 million or more (or other adverse economic effects), while a “significant” regulation may have other major impacts. 3. See Ferris and McGartland (this volume) and Morgenstern (this volume) for more discussion of research needs. 4. See Finkel (this volume) for detailed discussion of uncertainty analysis. 5. In developing criteria for quality and applicability, analysts should consider the extent to which market conditions have changed since the data were collected as well as the characteristics of the policy impacts considered. Many previous studies of employment impacts rely on relatively old data and look at relatively major changes (e.g., all pollution abatement expenditures or all rules promulgated under the Clean Air Act) rather than considering the impact of an individual regulation under market conditions more similar to the present. 6. Elliott (this volume) discusses other factors that could be considered in determining the extent to which employment impacts should be assessed. 7. At times, the affected industries will anticipate regulatory changes and begin adapting their practices in advance of when a rule is issued. Where significant, this advance compliance should be taken into account in the analysis. 8. These impacts can be estimated using a variety of different approaches, including input– output, partial equilibrium, and computable general equilibrium models. U.S. Environmental Protection Agency (2010) discusses the strengths and limitations of alternative approaches in various contexts. When employment impacts are addressed in both the distributional analysis and the benefit– cost analysis, differing modeling approaches may at times be appropriate. However, analysts will need to clearly describe the rationale for and the implications of these differences. 9. Although benefit valuation can be challenging, ideally the benefit– cost analysis would apply estimates that reflect how those affected value these changes in productivity as well as other attributes. Thus the value of productivity impacts resulting from health or other improvements is presumably already incorporated in the benefit– cost analysis. Recommendation 5 instead addresses whether these benefits might noticeably affect the number of jobs available. 10. Improvements in health also decrease the demand for medical treatment, which if significant could potentially affect employment in related fields. 11. This simple overview largely ignores the effects of market distortions, including taxes, unemployment insurance, minimum wage requirements, union bargaining power, and so forth, as discussed in many basic economics texts. These factors complicate the relationship between compensation and productivity as

126

Lisa A. Robinson

well as individual decisions regarding the trade- off between paid work and other activities. In addition, individuals may value work time above or below the amount of compensation received. For example, they may find work particularly satisfying or may spend more time working than they prefer because of market constraints (such as a required 40-hour work week). 12. Human capital may also improve or deteriorate because of the effects of employment and unemployment on physical and emotional health and overall life satisfaction. 13. Wages may change for those who are continually employed in the regulated industry, not solely for those who lose or gain jobs as a result of the regulation. 14. Coglianese and Bennear (2005) discuss retrospective analysis in more detail.

References Aguiar, Mark, et al. (2012) “Recent Developments in the Economics of Time Use,” Annual Review of Economics 4: 373– 97. Bartik, Timothy J. (2012) “Including Jobs in Benefit– Cost Analysis,” Annual R eview of Resource Economics 4: 55–73. Berman, Eli, and Linda T. M. Bui (2001) “Environmental Regulation and Labor Demand: Evidence from the South Coast Basin,” Journal of Public Economics 79: 265– 95. Clinton, William J. (1993) “Executive Order 12866: Regulatory Planning and Review,” Federal Register 58: 51735– 44. Coglianese, Cary, and Lori D. Snyder Bennear (2005) “Program Evaluation of Environmental Policies: Toward Evidence- Based Decision Making,” in G. D. Brewer and P. C. Stern, eds., Decision Making for the Environment: Social and Behavioral Science Research Priorities. Washington, DC: National Academies Press. Goulder, Lawrence H. (2000) “Environmental Policy Making in a Second- Best Setting,” in R. N. Stavins, ed., Economics of the Environment: Selected Readings, 4th ed. New York: W. W. Norton. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry- Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. Obama, Barack (2011) “Executive Order 13563: Improving Regulation and Regulatory Review,” Federal Register 76: 3821–23. ———. (2012) “Executive Order 13610: Identifying and Reducing Regulatory Burdens,” Federal Register 77: 28469–70. Palmquist, Raymond B., et al. (2010) “Short Run Constraints and the Increasing Marginal Value of Time in Recreation,” Environmental and Resource Economics 46: 19– 41. Small, Kenneth A. (2012) “Valuation of Travel Time,” Economics of Transportation 1: 2–14. U.S. Environmental Protection Agency (2010) Guidelines for Preparing Economic Analyses. EPA-240- R-10- 001. Washington, DC: U.S. Environmental Protection Agency.

Toward Best Practices

127

U.S. Office of Management and Budget (2003) Circular A-4: Regulatory Analysis, the White House, http://www.whitehouse.gov/omb/circulars_a004_a- 4 (accessed 24 January 2013). ———. (2012) Draft 2012 Report to Congress on the Benefits and Costs of Federal Regulations and Unfunded Mandates on State, Local, and Tribal Entities, the White House, http://www.whitehouse.gov/sites/default/fi les/omb/oira /draft_2012_cost_benefit_report.pdf (accessed 24 January 2013).

Chapter 7

Emitting More Light than Heat Lessons from Risk Assessment Controversies for the “Job-Killing Regulations” Debate Adam M. Finkel

Although we can choose to think descriptively, quantitatively, or both when we evaluate the pros and cons of whether and how to attack a hazard to health, safety, or the environment, both political leaders and the public increasingly expect that numbers will play a central role. And not just any numbers will do. As our faculties for collecting data, discerning causal relationships, and refining empirical models continue to improve, regulatory analysts are struggling to provide “high- quality quantification.” Those responsible for developing, supporting, or criticizing estimates of regulatory costs in general, and of the effects of regulation on jobs in particular, can either rise to or dodge the challenges of analyzing in the following ways: • •

• • • •

thoroughly, neither omitting important categories nor illuminating a small part of the problem while ignoring the larger whole; humbly, by acknowledging uncertainty in all inputs to analysis and quantifying (to the extent feasible) the uncertainty in the final estimate; transparently, by enumerating all major assumptions that underlie the estimate; objectively, by seeking a mathematically unbiased estimate of the quantity being estimated;1 logically, by counting each effect exactly once, not twice or more; responsively, by presenting cost and benefit information that allows individuals who bear disproportionate costs or face disproportionate risks to see their own situation as well as that of the whole population averaged together; and

Emitting More Light than Heat •

129

carefully, by keeping separate the empirical estimates of effect from  the subjective valuation used to make commensurate those effects that are in different units or that occur at different times in the future.

These are all challenges that quantitative risk assessment (QRA) for human health and environmental harms has already confronted, and surmounted with varying degrees of success, over the past several decades. In a larger project supported by the National Science Foundation (Finkel et al. 2006), of which this chapter and volume are parts, my colleagues and I have taken a wide-ranging look at how risk assessors involved in the regulatory process may approach many aspects of their work with more rigor, transparency, and attention to uncertainty than do their colleagues who conduct cost analysis in support of regulatory decision making. As a trailing- edge subcategory of regulatory cost analysis, the study of the effects of regulation on layoffs and unemployment ( job impact analysis, or JIA) demonstrates even greater contrasts with the more established methods of QRA. But despite the very different raw materials used by JIA and QRA (production functions and elasticities in the former; toxicologic dose-response relationships and environmental fate- and-transport models in the latter), the controversies that risk assessment has faced down and defused have many close, sometimes exact, parallels to the challenges involved in the quest to improve regulatory economics in general and JIA in particular. This chapter will make explicit many thematic connections between the two spheres of analysis, in order to suggest lessons that job impact analysts can glean from the achievements— and the unfinished business— of QRA. I regard as fundamental to this volume the premise that both analysts and decision makers care about the magnitude of the employment impacts of regulation because we want (or should want) to give them proportionate attention—treating them as neither infinitely important nor absolutely trivial but rather considering them to be as important as their magnitude and perhaps also their distribution warrants. Choosing not to estimate any effects on employment, or declaring that it is too difficult to do so or that it is improper to make them part of the overarching cost–benefit ledger, really leaves us with only two options: to treat them as too unimportant to affect any decision or to treat them as so important as to trump any other aspects of decisions. So without more rigor in how we estimate the job impacts of regulation, we are in effect stuck with the same “precautionary principle” that in my opinion has sometimes stymied environmental health policy— and for which QRA is a principled alternative (Montague and Finkel 2007). As Livermore and Schwartz make clear (this volume), advocates of polar

130

Adam M. Finkel

positions on certain regulations have simultaneously asserted either that we cannot afford to regulate due to mass layoffs (especially during a recession) or that we cannot afford not to regulate and turn our back on masses of new “green jobs” (especially during a recession). Of course, this just gives us two dueling formulations of the “we have every reason to assume the worst, and we have an urgent duty to prevent the worst from occurring” (Bush 2002) attitude of precaution, one that can be applied with equal ferocity either to avoid the “worst” of regulatory costs or to avoid the “worst” outcome of leaving workers unemployed when a regulatory stimulus could help them. Just as QRA has offered a controversial but, in my view, a generally preferable alternative to “pure precaution,” so too we need a rigorous brand of JIA. Through controversy can come methodological improvement, and thence better decisions: that is my theme, both looking retrospectively at QRA and looking prospectively at JIA and at regulatory cost analysis more broadly.

A Stylized Example To foreshadow the major section that follows, consider how what might be tempting to regard as the end of a risk assessment is really only its beginning. Consider a hypothetical proposal to reduce by 10 parts per billion (ppb) the enforceable standard for the amount of ground- level ozone to which 300 million Americans are exposed. Because we can set the exposure goal, can precisely estimate the number of beneficiaries, and can use epidemiology and toxicology to model the relationship between ozone concentration and mortality from chronic respiratory disease, it is tempting to say summarily (in this hypothetical) that “the benefits of this intervention are the 1,234 lives we can save, which amounts to a monetary benefit of $8.074 billion” [1,234 times a value for a statistical life (VSL) of $6.543 million]. But this is no longer seen as a sufficient description of risk or benefit. As I will detail later in this chapter, risk assessors now routinely recognize the need to surround with various major elaborations the false precision, overaggregation, and tunnel vision of a single benefit number. Now let us walk briefly around to the other side of the cost–benefit mirror and consider cost estimation. Is it good enough to say summarily that “the 10 ppb across-the-board reduction in ozone will cost $1.234 billion to effectuate,” given that each of the complexities above applies to economics in its own way? And for the specific purposes of this volume, is it good enough to conclude that “in addition to the direct costs,

Emitting More Light than Heat

131

the regulation will destroy 123 jobs, which we value at $12.3 million using an estimate of $100,000 per job”? In all three arenas—risk, cost, and job impacts—we can choose to ask a better set of questions and provide richer answers. In the latter two arenas, though, we have been aiming our sights too low.

Realized (and As-Yet-Unrealized) Improvements in QRA: Parallels to Cost Estimation and Job Impact Analysis To draw parallels between QRA and the current state of play in the assessment of regulatory costs and in the analysis of employment impacts of regulation, I begin by discussing in detail three major controversies in the history of QRA that have (generally) resulted in improved methods and results.2 In job- loss discussions, scholars and stakeholders sometimes use words such as “impacts” and “effects” to mean the number of jobs lost (or gained), sometimes to mean the toll of each job loss on individual human welfare, and sometimes to mean the product of the two. I will adopt the last definition and hence will discuss many of the parallels between QRA and JIA twice— once with respect to the estimation of total outcomes (“assessment”) and once with respect to the monetization of each outcome (“valuation”). Reducing Excessive “Conservatism” in Estimation Procedures Risk Assessment. The accusation that QRA systematically and signifi cantly exaggerates risk has been in many ways the central and most virulent controversy surrounding the field for nearly 30 years. From the unremarkable observation that some steps in a typical QRA might inject some bias into the final risk estimate (for example, that the average person drinks two liters of water per day, when probably 80 or 90 percent of the population drinks somewhat less than this amount), critics posited a “vicious circle” of massive exaggeration, which purportedly, in turn, both fueled public fears and increased the psychological and budgetary incentives for agency risk assessors to ratchet up their “conservative” stance (Breyer 1993; Kuran and Sunstein 1999; Nichols and Zeckhauser 1986; Office of Management and Budget 1991).3 I have written extensively about the cherry-picking, the conflation of uncertainty (criticizing agencies for preferring risk overestimation to underestimation) and variability (criticizing them for trying to protect real people with aboveaverage exposures), and other flaws in this line of attack (Finkel 1989, 1995, 1997). Nevertheless, the agencies have (over-?) reacted to their critics by making various changes so that their risk- estimation procedures

132

Adam M. Finkel

are now less “conservative,” including (1) declaring that body weight to the 3 ⁄4 power, rather than to the 2 ⁄ 3 power, should be used to scale exposures from laboratory animals to humans (U.S. Environmental Protection Agency 1992); (2) relaxing several of the exposure assumptions that formerly defined the “maximally exposed individual” (Goldstein 1989); (3) changing some “default” adjustment factors in noncancer risk assessment for interspecies and intraspecies variation in susceptibility from tenfold to threefold; and (4) becoming much more receptive to alternative theories of nonlinear and threshold “modes of action” for carcinogens (National Academy of Sciences 2008:ch. 6). Risk Valuation. Although the VSL estimates that emerge from revealed-preference studies of wage premiums tend to be severalfold larger than those from stated-preference surveys (Cropper et al. 2011), there have been few if any calls for agencies to reduce any “exaggeration” in their VSL estimates. However, some observers, such as Cox (2012), have criticized agencies for attributing “life- saving” to cases where the regulatory interventions would tend only to prolong the lives of those helped by days or weeks; a contrary view is that stated-preference VSL estimates tend not to decline with the age of the respondent (Krupnick 2007). Cost Assessment. In contrast to the furor about QRA, practitioners and observers of cost–benefit analysis seem generally blasé about the possibility of systematic “conservatism” in the estimation of regulatory cost, despite a substantial and fairly one- sided literature documenting that costs are indeed routinely overestimated (Goodstein and Hodges 1997; Harrington et al. 2000; Office of Technology Assessment 1995; Pizer and Kopp 2003; Putnam, Hayes and Bartlett, Inc., 1980). This literature also provides a litany of theoretical and practical reasons why such bias is natural, citing economies of scale, technological learning, strategic incentives, general- equilibrium effects (Hazilla and Kopp 1990), and other factors that tend to “inflate” cost estimates either if ignored or if included uncritically into the analysis (Finkel forthcoming). To my knowledge, the only specific change to the Environmental Protection Agency’s (EPA) analytic practices to address this bias so far has been the recent incorporation of a modest correction factor for technological learning into several major cost–benefit analyses, such as the study of the past and future costs of the Clean Air Act (U.S. Environmental Protection Agency 2011a). Job-Loss Assessment. In the polarized “dueling estimate” controversies like the ones Livermore and Schwartz recount in their chapter in this volume, where extreme claims of mass layoffs vie with extreme claims of

Emitting More Light than Heat

133

“green job” creation, serious discussions of the sources and magnitude of over- or underestimation cannot occur. In the few cases where JIAs have emerged from within the agencies, we can offer theoretical reasons why they might underestimate the number of jobs lost: they may be looking too narrowly only at the sectors directly impacted, ignoring multiplier effects that ripple through the economy and depress other sectors’ hiring, or failing to account for firms that overreact to compliance costs and reduce labor inputs they had lived with in the preregulatory state of the world. A mirror-image set of factors could instead make the job-loss number an overestimate: the agencies may well be ignoring the new hiring the regulation will spur elsewhere in the economy, or they may fail to account for firms that will absorb compliance costs preferentially in ways that do not require layoffs. And, of course, because the number of jobs lost is tightly coupled to the estimated costs of the regulation itself, any upward or downward biases in the estimation of cost will propagate through to the estimate of job loss. But is any reasonable estimate, positive or negative, of the net jobs impacted by a regulation inherently biased high or low? Because JIA is in its infancy as compared with QRA, we cannot satisfactorily answer such a question. Indeed, we cannot even ascribe a particular bias to the failure to consider job impacts at all, given that plausible estimates of the net impact of a regulation or regulatory program on hiring nationwide may conclude that there are net losses or net gains. Job-Loss Valuation. Other authors in this volume engage in a lively debate about whether the figure of $100,000 in lifetime lost wages that Masur and Posner (2012) have suggested is biased low or high. It could certainly be an underestimate, in that it does not account for the “nonpecuniary costs,” including increased mortality risk and depression, that Masur and Posner say might add as much as $160,000 more to their figure. Conversely, any valuation method that does not consider the reservation wage will overstate the true welfare loss associated with a period of unemployment, and it is also possible (see Morgenstern in this volume) that the epidemiology of mortality and depression is not so straightforward or so dire as Masur and Posner conclude. Lessons Learned. The evolution of concern over “conservatism” in risk assessment points to several traps that JIA should avoid. One danger involves stubbornly clinging to dubious assumptions and data in the face of compelling, or merely relentless, criticism. Had QRA not prominently relaxed some of its less tenable procedures, it might not have survived the congressional regulatory “reform” proposals of the 1990s in its current form. But there is no guarantee that “modernizing” one or more steps in the analytic method will actually reduce bias. JIA must

134

Adam M. Finkel

also be wary of being hijacked by critics who see its output only through “with-us- or-against-us?” lenses. Acknowledging and Quantifying Parameter Uncertainty in Analytic Inputs Risk Assessment. For more than 25 years, risk assessors have increasingly construed their job as not merely providing a point estimate of risk but as giving decision makers and the public a depiction of uncertainty— the range and relative likelihoods of the possible true values of risk (Cullen and Frey 1999; Morgan and Henrion 1990; Roberts 1990; Thompson et al. 1992). Although techniques of uncertainty analysis can be as simple as providing a pair of bounding estimates for two scenarios and their relative likelihood, agencies such as EPA have long been using Monte Carlo techniques to derive probability density functions for risk.4 Contemporary risk assessments now often include elaborate elicitations to array the plausible scientific theories that might govern a hazard’s behavior in the environment or its interaction with human biochemistry and their relative likelihoods of being correct (Industrial Economics, Inc. 2006). Or they may include Monte Carlo analyses using empirical data to update and refine Bayesian prior distributions (Occupational Safety and Health Administration 1997:1544– 59). Risk Valuation. An uncertain estimate of risk can readily be combined with an uncertain estimate of the value of risk reduction to yield a distribution of plausible values of benefit. Indeed, in recent years EPA has begun to provide a range for the VSL (reflecting the divergence of mean estimates from several dozen different studies) as well as a standard error for the grand mean of all the studies considered. Various agencies have also added a rudimentary bow to uncertainty in the valuation of future benefits by showing the results with an alternative discount rate to the (usually higher) one that the Office of Management and Budget’s Office of Information and Regulatory Affairs requires as the default value. Cost Estimation. In a forthcoming monograph (Finkel forthcoming), I review the use patterns of uncertainty analysis by regulatory economists between 1990 and the present and conclude that for most of the advances risk assessors have first introduced and later made routine, regulatory economists have lagged behind by roughly 15 years. In particular, I found that virtually all of the EPA regulatory analyses praised by various researchers as having provided ranges for cost as a bow toward uncertainty (Hahn and Dudley 2007) in fact provided only “quasi-ranges” that instead of bounding possible true values for total cost actually represented two incompatible estimates (often coming from two different as-

Emitting More Light than Heat

135

sumptions about the stringency of regulatory enforcement) connected by a hyphen. Even as recently as 2011, EPA has continued to present major (billion- dollar) regulatory cost estimates as point estimates with several significant figures of “accuracy,” even in regulatory impact analyses that at the same time contain hundreds of pages of uncertainty analysis for risk. Or, EPA presents a point estimate for cost surrounded merely by a “rule of thumb” for imprecision, such as the “cost-plus- or-minus-20percent” conclusion in its 2004 regulatory impact analysis for nonroad diesel engine emissions control (U.S. Environmental Protection Agency 2004). As a result, most of the charts and tables in EPA regulatory impact analyses that purport to reveal uncertainty in net benefits in fact represent a distribution for benefits from which it has subtracted a point estimate of cost.5 Other agencies, however, notably the U.S. Food and Drug Administration (2004:56886) and National Highway Traffic Safety Administration (2007:VIII–20), have begun to conduct elaborate Monte Carlo analyses of uncertainty in cost and hence of net benefit. Job-Loss Assessment. Official agency estimates of job impacts, when they are made at all (see Shapiro this volume), tend to ignore or give short shrift to uncertainty. An interesting exception is the EPA boiler MACT rule (U.S. Environmental Protection Agency 2011b:4- 6–4-13). Here EPA took the standard errors from the seminal study of Morgenstern et al. (2002) and produced a central estimate of 1,400 jobs gained— but with confidence bounds from 5,400 jobs gained to 2,600 jobs lost (whether these were intended to represent the 95th, 97.5th, or some other percentile interval was not mentioned). Unfortunately, EPA applied these standard errors, which apply to coefficients on functions for “ job changes per million dollars of regulatory expenditures,” to its cost figure of $2.4 billion that it presented with zero uncertainty— so the given range for job impacts is surely overconfident despite how broad it already is. Just as with the tendency to compare uncertain total benefits with “arrogant” estimates of total cost, this regulatory impact analysis demonstrates that any uncertainty estimate is only as useful as its weakest link, and if the weaker link is “upstream” from a variable of intense interest, it becomes very difficult to give an honest portrayal of uncertainty. Job- Loss Valuation. These analyses also tend to end with point estimates rather than ranges or distributions that would reflect the many unanswered questions about the length of time displaced workers remain unemployed, the proper value of the reservation wage from which welfare loss due to unemployment should be estimated, the effect of unemployment on mortality risk and quality of life (Adler this volume), and other factors. As with many of the prominent point estimates of “lives lost” in the early days of agency risk assessment, researchers and critics tend to proffer starkly different point estimates, rather than

136

Adam M. Finkel

trying to ground both the stalking horse and their alternative in terms of relative plausibility. One recent exception is found in Bartik (2012), in which he bounds the elasticity of unemployment in order to derive a range for the welfare losses from a demand shock that might result from a major regulation. Lessons Learned. Risk assessors have embraced uncertainty not only because the scientific method teaches that false precision is a form of error but because honest portrayals of uncertainty have helped to defuse and moderate controversy. When the only way to explain two divergent estimates is by proclaiming that one’s own position is right and the alternative is wrong, energy tends to go into posturing and strategic obfuscation. Uncertainty, in contrast, is the context within which it is possible to explain that a competing estimate is part of a legitimate continuum of possibilities but is a less likely or less reasonable guide to sound decision making (or both)—and to explain why. But in order for uncertainty characterization to facilitate productive dialogue, analysts cannot be so determined not to overpromise that they settle for “underconfident” pronouncements. For example, EPA began responding circa 1990 to criticisms that its cancer risk estimates were too “conservative” by adding the new caveat “the true estimate could be as low as zero”— but in many cases zero was simply an implausible lower bound given the observed data. JIA is fortunate to have on hand most of the raw material needed to produce reasonable (that is, not farcical) characterizations of uncertainty. Estimates of the effect of new compliance costs on consumer prices, of the effect of consumer prices on aggregate demand, and of the changes in labor needed to maximize profits when demand drops, can all be estimated in a way that preserves rather than discards their standard errors. With these in hand, Monte Carlo techniques are tailor-made for expressing the uncertainty in the employment impact as a function of the uncertainties at each link in the causal chain. Balancing the Ledger by Estimating Effects Net of Trade- offs, Not Just First- Order Effects Risk Assessment. One of the most vexing aspects of risk assessment is the tendency for risk-reducing actions to have multiple consequences, ones that are often sufficiently large to palpably reduce the true benefits of the targeted action, to surpass them and turn net benefits into net harms, or to complement them and yield even more beneficial results. Unfortunately, the natural tendency toward myopia (“that’s not my department”), and the sometimes-mistaken mental model that “higherorder” consequences must involve smaller and smaller ripples compared with the primary ones, delayed the progress of risk assessment toward a

Emitting More Light than Heat

137

fuller accounting of net effects. But as with the other parallels described in this chapter, “risk-versus-risk” effects are now far more routinely acknowledged than are the analogous “cost-versus- cost” and “ job-versusjob” effects. Beginning with the influential book by Graham and Wiener (1995), regulatory agency risk assessments have steadily incorporated more information about the offsetting reactions of risk-reduction interventions. I suggest here that the most relevant aspect of any purported action– reaction effect, for both risk assessment and regulatory economics, is the attenuation of cause and effect, or how many immediate steps or assumptions must be bridged for the trade- off to actually occur as advertised. In order of increasing attenuation, the following are brief examples of five distinct levels or types of trade- offs: (1) Inherent and inevitable. Ground-level ozone harms the human respiratory system but at the same time also absorbs ultraviolet light that otherwise might trigger a skin cancer. Pursuant to the Supreme Court decision in Whitman v. American Trucking Associations, Inc. (2001), EPA was obligated to consider the net effect of reducing ground- level ozone on mortality. (2) “Adverse substitution” that is hard to avoid. Although the magnitude of this effect is subject to ongoing dispute, the substitution of lighter materials is the most reliable way to increase automobile fuel efficiency, and these materials may tend to make cars less crashworthy and thus cause more highway fatalities as one price of decreasing pollution. (3) Inherent but amenable to a technological win–win. The first generation of automobile airbags saved the lives of many adults but also injured or killed a smaller number of children and small adults. However, identifying the trade- off led to design changes that have greatly reduced the latter consequence.6 Hence, it would have been unfortunate had the government shied away altogether from requiring airbags on the grounds that they offered smaller net life- saving benefits. (4) Adverse substitution that is needlessly risky. As EPA and states and localities tighten controls on perchloroethylene in dry cleaning, some cleaners are turning to safer solvents such as supercritical carbon dioxide, but others are turning to the more potent neurotoxin and carcinogen 1-bromopropane (Finkel 2010; Urbina 2013). As a result, the magnitude and even the sign of the net effect is wholly under the control of the regulated parties.

138

Adam M. Finkel

(5) Indirect and causally ambiguous. The theory that expenditures on regulations can so impoverish consumers as to increase their mortality risks via a “poorer is riskier” effect has been largely discarded as a guide to policy choice, for reasons I will discuss below (Rascoff and Revesz 2002). Cost Assessment. The major limitation of the standard partialequilibrium approach to estimating regulatory costs as composed primarily of compliance costs is that it does not consider the effects elsewhere in the economy of increased compliance spending (Pizer and Kopp 2003). There are theoretical arguments why general- equilibrium costs could either add to compliance costs, partially offset them (Palmer et al. 1995), or offset them by such a degree as to make the total economic costs of a given regulation negative (Porter and van der Linde 1995). But it remains the case that, although risk assessment more routinely “follows the risk” to see what second- order effects might accompany it, regulatory economics less often “follows the cost.” Job-Loss Assessment. JIA currently suffers greatly from the tendency to declare victory on net effects, having looked at only the proverbial tip and not the rest of the iceberg. Morgenstern et al. (2002) have done a great ser vice by pioneering the idea that in addition to the obvious “demand effect” of environmental compliance on jobs (higher production costs may raise market prices, hence reducing consumer demand, hence reducing the need for labor), there are two other effects that can plausibly counteract (or in one case, could instead reinforce) this job-loss effect. The “cost- effect” is of the opposite sign (firms need more labor to maintain a given level of output than before), and the “factor-shift effect” is of indeterminate sign (the new production processes needed to comply with the regulation may turn out to be more or less labor-intensive per unit of output).7 For a very simple example of a countervailing factor- shift effect, consider that one clear effect of the infamous withdrawal of the growth hormone Alar from the U.S. apple crop in 1989 was to increase the demand for labor, as Alar had helped the apples to ripen all at once. But even the sum of these three effects represents jobs lost (or gained) only by the firms most directly and adversely affected by the regulation. It is not, as EPA suggested when it used the three coefficients from Morgenstern et al. (2002) in its boiler MACT regulatory impact analysis, a way of “decomposing the overall effect of a regulation on employment into three subcomponents” (U.S. Environmental Protection Agency 2011b:49, emphasis added). Using the metaphor of a set of three concentric circles rippling outward from the directly affected firms, consider how much more broadly the “indirect” (but very salient) impacts on employment really reach:

Emitting More Light than Heat •

•

•

139

They may reach firms that produce the same product more efficiently than the affected fi rms, allowing them to gain market share, and hence be able to increase employment (or wages). The same could be true of firms in “attainment” counties, whose potential offsetting labor gains Greenstone (2002) could not tease out (Morgenstern this volume). They may reach firms that produce a different product competing with the affected product(s), allowing them to benefit from increased demand. For example, although media coverage (and an unsuccessful product- disparagement lawsuit) related to the Alar controversy focused exclusively on the roughly $100 million in lost revenue that growers of Red Delicious and Macintosh apples experienced in the first years after the withdrawal, revenue from other varieties of apple (Fuji, Gala, and so forth) that did not depend on Alar increased by more than $1 billion during the same period (Finkel 1995). Of course, firms that produce a product that complements (rather than substitutes for) the most affected product may well experience job losses in tandem with the firms represented in the JIA. Employment impacts may reach firms that produce the goods and ser vices needed to comply with the regulation, who should experience the need for additional labor to supply that demand. I am unaware of any first principles under which the amount of “green hiring” for a given pollution- control project should be greater than or less than the number of jobs lost when other firms essentially have to purchase these new goods, but the gross number of new jobs is certainly greater than zero. Indeed, I observe that several of the authors in this volume emphasize that we generally do not “create jobs by breaking windows,” as Bastiat explained 160 years ago (Ferris and McGartland this volume; Mannix this volume; Morgenstern this volume). But to exactly the same extent that position is true, neither will we “kill” jobs by inventing an unbreakable window. We should either count both sides of the ledger or neither.

Job-Loss Valuation. Although I see no obvious lessons for job-loss valuation in the arena of estimating the value of risk reductions net of offsetting risks, there are important net effects from job impacts that are not fully considered in contemporary attempts to assign value to the welfare toll of unemployment. Several of the authors in this volume grapple with how to do better than making simplistic extreme assumptions— namely, either assuming that a laid- off worker loses the full value of her salary (but by that logic the net societal effect would be zero, as the employer gains back that full value) or assuming that she loses nothing because the wage rate was the equilibrium value of her labor (but by that

140

Adam M. Finkel

logic, none of the nonpecuniary harms associated with unemployment could be counted). The reservation wage seems to be the appropriate way to think about the pecuniary portion of the welfare toll. That is, just as when a product is banned the consumer loses only the surplus value he assigned to it (anyone but the “marginal consumer” will lose more utility than he gains if he recoups the purchase price instead of owning the product), so when a worker is laid off, he loses a salary but has restored to him the value of his work week (and to call this “leisure” seems to trivialize the importance of also being able to look for alternative work). The disparate studies of unemployment’s effects on mortality and morbidity referenced in this volume also speak to the importance of considering both harmful and beneficial effects. And there also seem to be unappreciated net effects of how we value the ancillary costs of unemployment. The attention being paid (so far only in theory) to the “transition costs” of relocation and retraining also needs to ground these estimates in their proper general- equilibrium context: to the movingvan company and the headhunter, these “costs” appear as benefits, though perhaps by definition of smaller magnitude than the costs to the worker. Lessons Learned. The thoughtful and mea sured expansion of QRA to include many, but not all, net effects offers several valuable lessons for JIA: •

•

Analysts need the freedom to choose principled and consistent boundaries between second- order effects that they will and will not consider. A full treatment of this issue is beyond the scope of this chapter, but in the spectrum of five kinds of these effects introduced above, I would certainly recommend that risk assessors and regulatory economists routinely consider the first two categories and eschew the fifth. Specifically with respect to JIA, I think the second- order effects of unemployment on depression, suicide, and the like ought to be considered (as long as the favorable effects of reemployment on longevity and quality of life are also considered). However, the effects of lost wages on mortality are too tenuous to be included at this time.8 Analysts should carefully consider the counterfactual. It is easy to imagine that a truck driver might be injured or killed transporting contaminated soil from a hazardous waste site (Hoskin et al. 1994). But should those “statistical fatalities” be counted against the expected lives saved from a national program to clean such sites? Only if the program is large enough to increase the national production of trucks; otherwise, this would be a net effect of zero, as fewer accidents would presumably occur attributable to all other uses for trucking.

Emitting More Light than Heat •

•

141

Analysts should consider reinforcing effects (“co-benefits”) as well as countervailing effects. Rascoff and Revesz (2002) thoroughly dissect the bias of regulatory critics to badger agencies to fi nd reasons to offset benefits. Dudley (2012), however, makes a persuasive theoretical case that agencies should be careful not to claim the same cobenefits in more than one regulatory impact analysis. Analysts must not avoid confronting the problem of volition. In risk– risk trade- off analysis, the fourth category I discussed above is essentially the “Washington Monument fallacy”— from the apocryphal story about the Interior Department responding to a proposed budget cut by asserting that the first expenditure it would “have” to cut would be the maintenance and staffing of the monument. Like it or not, it is the analyst’s job to model what she thinks will happen when actors with free will respond to regulation, not necessarily what those actors say will happen, perhaps in order to forestall or weaken regulation. The parallel is glaring to one side of the job loss– gain ledger, where regulated firms have every incentive to assert that rather than reduce dividends, executive salaries, or inefficiencies in production, the only way they can respond to additional compliance costs is to pass 100 percent of them to their workers in the form of layoffs.

Other Parallels There exist many other close analogies between the evolution of risk assessment and the challenges facing job impact analysis and cost estimation more generally. Any plan for improving these economic aspects of cost–benefit analysis ought to also consider these other parallels. Risk assessment has evolved a system for disclosing major “default” assumptions, for making transparent the scientific and policy rationales for choosing them, and for considering evidence that would replace a given assumption with a specific alternative in cases where the default is implausible. Both of the major reviews of QRA conducted by the National Academy of Sciences (NAS) strongly recommended that the agencies continue to use defaults but better articulate the standard of evidence required to supplant them (National Academy of Sciences 1994, 2009). The 2009 NAS study recommended a “clearly superior” standard, so that alternatives will not be rejected for lack of “proof,” but neither will they be accepted without compelling evidence behind them. Ferris and McGartland (this volume) call for a “unified theory” of employment impacts, which is another way of saying “a collection of well-reasoned default assumptions about how regulation can affect social welfare through employment changes.” Risk assessment is struggling to acknowledge important “missing defaults” and to admit that some of the strongest assumptions occur when analysts claim

142

Adam M. Finkel

they are making no assumptions. Chapter 6 in the NAS (2009) report offers a list of six steps in the risk assessment process where scientifically implausible assumptions result from the refusal to make actual judgments. For example, by claiming to not have enough information to account for the genetic and environmental variation in susceptibility to cancer in the human population, assessors in effect treat all humans as identical and may fail to protect up to half of the population (Finkel 2013).9 In economic analysis and JIA, a reluctance to apply what is known about the diminishing marginal utility of money results in overstating the change in welfare when higher- income persons bear disproportionate costs (or face unemployment) and in underestimating the same when lower- income persons are so affected. Risk assessment has begun to “harmonize” a major dichotomy and allow two major categories of human health effect to be estimated in the common currency of probability. For decades, the agencies have generated risk estimates for carcinogens but only “bright lines” of safety or harm for noncarcinogens. This stymies benefits assessment, as the Reference Concentrations (RfCs) for noncarcinogens allow analysts only to consider the number of people whose exposures will be reduced from “unsafe” to “safe” and not to estimate the changing incidence of any health effect or to consider at all the benefits of exposure reductions that begin and end either above the RfC or below it. The 2009 NAS report summarized a vast literature on methods for deriving continuous population dose-response functions for noncarcinogens and recommended that EPA “unify” across the two paradigms. In JIA, a decision to keep job impacts in “natural units” (number of layoffs) rather than monetizing them would similarly marginalize them, just as noncancer health effects have awaited a full seat at the table on the risk side. Furthermore, within the employment- effects category, only modeling plant closures and job losses could similarly marginalize the less dramatic (but possibly greater) welfare effects of decreases in wages among those who retain their jobs at firms burdened by new costs. Risk assessment has greatly increased its attention to the full spectrum of variations in human exposure. The EPA Exposure Factors Handbook (U.S. Environmental Protection Agency 2011c) and its companion document for children (U.S. Environmental Protection Agency 2008), together contain over 2,000 pages of information that allows analysts and policymakers to consider both the tail and the center of virtually any distribution of human behavior, intake, activity, location, and exposure to hazards. Of the many unexplored analogies in JIA, one example would be the quality of jobs affected. Just as a regulation can reduce total population exposure but change its distribution in untoward or fortuitous ways, so too can it either increase low- quality jobs at the expense

Emitting More Light than Heat

143

of higher- quality ones, or vice versa, for a constant change in total employment. Risk assessment has a long and productive history of developing new assays and test methods, both to evaluate proposed interventions and to corroborate or dispute previous predictions with hindsight. Just as risk scientists have come up with new ways to test substances more efficiently and reliably (National Academy of Sciences 2007), calibrate across the divide between toxicology and epidemiology (Allen et al. 1988), and refi ne the judgments of experts to account for biases and knowledge gaps (Walker et al. 2001), economists should develop (among other things) survey methods to assess the mean and distribution of lost worker surplus when a layoff occurs and ways to look back at past predictions to see whether history has borne these predictions out. For example, Masur and Posner (2012:594– 95) have noted that EPA said 16 years ago that “5,711 jobs” would be lost in the pulp and paper sector because of one of its rules. Were that many jobs actually lost?

Policy Prescriptions to Link Unemployment and Regulation In addition to the many parallels between risk assessment and job impact analysis, experiences from the risk- science domain in translating the findings of analysis into sensible regulatory decisions and into sensible legislation to govern the regulatory process could offer a variety of helpful insights. Because it is so timely, I will mention just one such parallel here. The story goes— and it has gone as far as the 2012 passage of a bill by the House of Representatives (Red Tape Reduction and Small Business Job Creation Act 2012) that because regulations increase unemployment, we should impose a moratorium on all regulations until the unemployment rate goes below some number such as 6 percent. This is, of course, an extremely blunt instrument in response to a potential problem. Individual regulations may cause job loss, but there are three truisms applicable to the relationship between regulations in general and national employment levels: (1) many other macroeconomic factors affect national employment levels; (2) regulations are highly variable from each other with respect to their effects on jobs; and (3) regulations also confer benefits. Consider, however, that individual chemical substances are known to cause cancer, and these three truisms also apply, with just a few word changes, to the relationship between chemical production and national cancer rates: (1) national cancer rates depend strongly on many other exogenous factors; (2) there is enormous substance-to- substance variation in carcinogenic potency (indeed, many substances are not

144

Adam M. Finkel

carcinogenic in the least); and (3) chemicals also confer benefits. Therefore, I propose, with tongue in cheek, that we consider a national moratorium on chemical production until the national cancer rate dips below 20 percent (from its current 35 percent). But even this is not a fully drawn analogy to a one- size- fits- all regulatory moratorium. Others in this volume suggest that individual regulations, and perhaps regulations in general, can increase net employment. In that case, a better analogy would be a proposed moratorium on the production of chemotherapeutic agents until the national cancer rate dips below 20 percent. Here the policy overreaction would ensure that the problem will get worse and worse and that the moratorium could never be lifted because it is so wrongheaded.

Conclusion Although I have worked as a regulatory decision official and as the director of a regulatory enforcement program, I am foremost an analyst, and so I resist the temptation to conclude that substantial effects on human welfare are too important (or too recalcitrant to analysis) to handle analytically. Having been “laid off” once in the past (Carey 2005), I have to admit that we risk assessors sometimes ascribe large monetary benefits to averting health effects whose real impact on the individual may not be as wrenching as job loss is. Worse yet, I am concerned about the needless expense and delay involved in analyzing failure modes, exposures, potencies, compliance costs, and other “quantifiable” aspects of riskcost decisions, only to make a decision orthogonal to all the analysis, dictated by a laser-like focus on jobs alone, especially if the job impact is merely asserted rather than carefully estimated. In an economic downturn, everyone wants to be on the side of jobs and on the side of the “ job creators,” whoever they are. The fact that we cannot begin to agree about the past— did the auto bailout money “save” a million jobs, or would these workers still be employed (in postbankruptcy auto companies or elsewhere) absent the intervention?—makes it very daunting to imagine agreement about the future, such as in asking how many jobs an individual new regulation would destroy or create. As difficult as risk assessment can be, with arguments persisting for decades about which physical or biochemical laws are at work, the arguments about job creation or destruction rise or fall over assumptions about human behavior, which is far more inscrutable. If when discussing the effect of raising tax rates on the wealthy, we still cannot agree about whether the income effect (those affected will work more hours to recoup their previous wealth position; Diamond and Saez 2011) or the substitution effect (they will work less because the tax makes it rela-

Emitting More Light than Heat

145

tively more attractive to spend time elsewhere; Mathur et al. 2012) is larger than the other, it seems a vain hope that consensus will be reached on how firms will make hiring decisions in response to compliance costs and to new markets for compliance technology. Nevertheless, the history of one- step-back-two- steps-forward in the quality of (and broad expert and public comfort with) quantitative risk assessment suggests that the controversy over job impacts may yet yield to careful analysis— and that we need better ways to consider these impacts more sensibly while we await such potential answers from analysis.

Acknowledgments I appreciate the helpful comments of Arden Rowell on a draft of this chapter and those of many participants at the Penn Program on Regulation conference.

Notes 1. I belabor these last five words deliberately, to express my strong belief that there is nothing “biased” about explicitly presenting an estimate other than the expected value of a cost, risk, or any other quantity (Finkel 1995). An upperbound estimate that has a 95 percent chance of being at or above the true value of an uncertain quantity can be an ideal guide to decision making and communication (if it is, in fact, a 95th percentile estimate), or it can be a terrible guide (if it is outside any bounds of possibility or is a central estimate falsely presented as an upper bound). 2. Later in this chapter, I enumerate five other parallels but do not treat them in detail, due to space considerations. 3. In this context, “conservative” does not have its traditional political connotation but simply means to choose an estimator greater than the central tendency. Of course, being “conservative” in estimating (or valuing) risk or benefit tends to result in more stringent regulation, while “conservatism” in estimating or valuing cost or job impacts leads to less stringent regulation. 4. For an early example, see U.S. Environmental Protection Agency (1986), especially at pages 1640– 53. For a hierarchy containing 10 degrees of increasing sophistication in uncertainty analysis, see Finkel (forthcoming). 5. This half-measure is guaranteed to understate uncertainty in net benefit and may well depict a truly ambiguous net benefit as having a definite sign. For example, which is more expensive, an apple that costs somewhere between 60 and 80 cents or an orange that costs between 70 cents and $1.00? The truth is that the “net cost” of buying an orange is somewhere between 40 cents (100 – 60) and “negative 10 cents” (70 – 80); the sign is ambiguous. But if you quantify the range for the apple’s price but present only the mean (85 cents) for the orange’s price, the net cost appears to be unambiguously positive— at least 5 cents (85 – 80) and as much as 25 cents (85 – 60). 6. Similarly, it is possible to eat less fish containing high levels of methyl mercury and still reap the benefits of the omega- 3 fatty acids found in fish oil (Ginsberg and Toal 2009).

146

Adam M. Finkel

7. In the Morgenstern et al. (2002) analysis, the factor- shift effect worked opposite to the demand effect, but the authors state that this may not always be the case. 8. The “richer is safer” theory suffers from all manner of flaws, including reverse causation (any correlation between wealth and mortality may be driven more by sick people losing income than by poor people dying young), the ecological fallacy (even if richer people as a class have a lower mortality risk, it does not necessarily follow that any individual person whose wealth increases will live longer), and nonlinearity (a given total amount of regulatory cost could either increase or decrease net longevity, depending on whether the rich, whose risk cannot be further decreased by more wealth, pay more or less than their proportional share of the costs). For these and other reasons, various congressional regulatory reform bills in the 1990s steadily retreated from requiring such analysis. The initial “Contract with America” bill (Job Creation and Wage Enhancement Act 1995) would have required agencies to include an analysis “of any significant substitution risks to human health.” By the time Senator Carl Levin sponsored the last such bill in that decade (Regulatory Improvement Act 1999), the analogous section defined “substitution risk” as one “expected to result from a regulatory option [but that] shall not include risks attributable to the effect of an option on the income of individuals.” 9. The NAS (2009) reviewed a significant literature on this issue and concluded that an upward adjustment of from tenfold to fi fty-fold to account for susceptible subpopulations would be much more realistic than the current tacit assumption of zero adjustment.

References Allen, Bruce C., et al. (1988) “Correlation Between Carcinogenic Potency of Chemicals in Animals and Humans,” Risk Analysis 8: 531– 44. Bartik, Timothy J. (2012) “Including Jobs in Benefit– Cost Analysis,” Annual Review of Resource Economics 4: 55–73. Breyer, Stephen (1993) Breaking the Vicious Circle: Toward Effective Risk Regulation. Cambridge, MA: Harvard University Press. Bush, George W. (2002) “Remarks by the President on Iraq: Speech at the Cincinnati Museum Center,” the White House (7 Oct.), http://www.gwu. edu/~nsarchiv/NSA EBB/NSA EBB80/new/doc%2012/President%20 Bush%20Outlines%20Iraqi%20Threat.htm (accessed 3 March 2013). Carey, John (2005) “The Unrecognized Epidemic,” 3931 Business Week 41– 42 (2 May). Cox, Louis A., Jr. (2012) “Reassessing the Human Health Benefits from Cleaner Air,” Risk Analysis 32: 816–29. Cropper, Maureen, et al. (2011) “Valuing Mortality Risk Reductions: Progress and Challenges,” Annual Review of Resource Economics 3: 313–36. Cullen, Alison C., and H. Christopher Frey (1999) Probabilistic Techniques in Exposure Assessment: A Handbook for Dealing with Variability and Uncertainty in Models and Inputs. New York: Plenum Press. Diamond, Peter, and Emmanuel Saez (2011) “The Case for a Progressive Tax: From Basic Research to Policy Recommendations.” CESinfo Working Paper No. 3548, Munich. Dudley, Susan (2012) “Perpetuating Puffery: An Analysis of the Composition of OMB’s Reported Benefits of Regulation,” Business Economics 47: 165–76.

Emitting More Light than Heat

147

Finkel, Adam M. (1989) “Is Risk Assessment Really Too ‘Conservative’? Revising the Revisionists,” Columbia Journal of Environmental Law 14: 427– 67. ———. (1995) “A Second Opinion on an Environmental Misdiagnosis: The Risky Prescriptions of Breaking the Vicious Circle,” New York University Environmental Law Journal 3: 295–381. ———. (1997) “Disconnect Brain and Repeat After Me: ‘Risk Assessment Is Too Conservative,’ ” in E. Bingham and D. P. Rall, eds., Preventive Strategies for Living in a Chemical World: Annals of the New York Academy of Sciences, vol. 837. New York: New York Academy of Sciences. ———. (2010) “Increased Toxicity and Carcinogenicity of n-Propyl Bromide (1-Bromopropane) Relative to Perchloroethylene.” Report to the Philadelphia Department of Public Health (17 June), http://www.phila.gov/health/pdfs /airmanagement/Finkel%20 Perc%20nPB.pdf (accessed 15 September 2013). ———. (2013). “Protecting the Cancer Susceptibility Curve,” Environmental Health Perspectives 121: A238. ———. (forthcoming) The Cost of Nothing over the Value of Everything: The Failure of Regulatory Economics to Keep Pace with Improvements in Quantitative Risk Analysis. Finkel, Adam M., et al. (2006) “Transferring to Regulatory Economics the RiskAnalysis Approaches to Uncertainty, Interindividual Variability, and Other Phenomena.” National Science Foundation, Proposal #0624218 (funded as project #0756539). Ginsberg, Gary L., and Brian F. Toal (2009) “Quantitative Approach for Incorporating Methyl Mercury Risks and Omega- 3 Fatty Acid Benefits in Developing Species- Specific Fish Consumption Advice,” Environmental Health Perspectives 117: 267–75. Goldstein, Bernard D. (1989) “The Maximally Exposed Individual: An Inappropriate Basis for Public Health Decision-Making,” Environmental Forum 6: 13–16. Goodstein, Eban, and Hart Hodges (1997) “Polluted Data: Overestimating Environmental Costs,” American Prospect (November– December): 64– 69. Graham, John D., and J. B. Wiener (1995) Risk Versus Risk: Tradeoffs in Protecting Health and the Environment. Cambridge, MA: Harvard University Press. Greenstone, Michael. (2002) “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufacturers,” Journal of Political Economy 110: 1175–1219. Hahn, Robert W., and Patrick M. Dudley (2007) “How Well Does the U.S. Government Do Benefit– Cost Analysis?” Review of Environmental Economics and Policy 1: 192–211. Harrington, Winston, et al. (2000) “On the Accuracy of Regulatory Cost Estimates,” Journal of Policy Analysis and Management 19: 297–322. Hazilla, Michael, and R. J. Kopp (1990) “Social Cost of Environmental Quality Regulations: A General Equilibrium Analysis,” Journal of Political Economy 98: 853–73. Hoskin, Alan F., et al. (1994) “Estimated Risk of Occupational Fatalities Associated with Hazardous Waste Site Remediation,” Risk Analysis 14: 1011–17. Industrial Economics, Inc. (2006) Expanded Expert Judgment Assessment of the Concentration- Response Relationship Between PM 2.5 Exposure and Mortality. Washington, DC: U.S. Environmental Protection Agency, http://www.epa.gov /ttn/ecas/regdata/Uncertainty/pm_ee_report.pdf (accessed 3 March 2013).

148

Adam M. Finkel

Krupnick, Alan (2007) “Mortality- Risk Valuation and Age: Stated Preference Evidence,” Review of Environmental Economics and Policy 1: 261– 82. Kuran, Timur, and Cass R. Sunstein (1999) “Availability Cascades and Risk Regulation,” Stanford Law Review 51: 683–768. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Mathur, Aparna, et al. (2012) “Should the Top Marginal Income Tax Rate Be 73 Percent?” Tax Notes 137 (19 November): 905–15. Montague, Peter, and Adam Finkel (2007) “Two Friends Debate Risk Assessment and Precaution,” Rachel’s Democracy and Health News (16 August), http://www.rachel.org/?q=en/newsletters/rachels_news/920#Two- Friends -Debate- Risk- Assessment- and- Precaution (accessed 15 February 2013). Morgan, M. G., and Max Henrion (1990). Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. New York: Cambridge University Press. Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry- Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. National Academy of Sciences (1994) Science and Judgment in Risk Assessment. Washington, DC: National Academy Press. ———. (2007) Toxicity Testing in the 21st Century: A Vision and a Strategy. Washington, DC: National Academy Press. ———. (2009) Science and Decisions: Advancing Risk Assessment. Washington, DC: National Academy Press. National Highway Traffic Safety Administration (2007) Final Regulatory Impact Analysis: Electronic Stability Control Systems. Washington, DC: Office of Regulatory Analysis and Evaluation. Nichols, Albert L., and Richard J. Zeckhauser (1986) “The Perils of Prudence: How Conservative Risk Estimates Distort Regulation,” Regulation 13–24 (November/December). Occupational Safety and Health Administration (1997) “Occupational Exposure to Methylene Chloride: Final Rule,” 62 Federal Register 1494–1619. Office of Management and Budget (1991) “Current Regulatory Issues in Risk Assessment and Risk Management,” in Regulatory Program of the United States, April 1, 1990– March 31, 1991. Washington, DC: Office of Management and Budget. Office of Technology Assessment (1995) Gauging Control Technology and Regulatory Impacts in Occupational Safety and Health: An Appraisal of OSHA’s Analytic Approach. Washington, DC: Office of Technology Assessment. Palmer, K., et al. (1995) “Tightening Environmental Standards: The Benefit- Cost or the No- Cost Paradigm?” Journal of Economic Perspectives 9: 119– 32. Pizer, William A., and Raymond Kopp (2003) “Calculating the Costs of Environmental Regulation.” Resources for the Future Discussion Paper 03- 06, Washington, DC. Porter, Michael E., and Claas van der Linde (1995) “Toward a New Conception of the Environment- Competitiveness Relationship,” Journal of Economic Perspectives 9: 97–118. Putnam, Hayes, and Bartlett, Inc. (1980) Comparisons of Estimated and Actual Pollution Control Capital Expenditures for Selected Industries. Washington, DC: U.S. Environmental Protection Agency.

Emitting More Light than Heat

149

Rascoff, Samuel, and Richard Revesz (2002) “The Biases of Risk Tradeoff Analysis: Towards Parity in Environmental and Health- and- Safety Regulation,” University of Chicago Law Rev. 69: 1763–1836. Roberts, Leslie (1990) “Risk Assessors Taken to Task,” Science 247 (9 March): 1173. Thompson, K. M., et al. (1992) “Monte Carlo Techniques for Quantitative Uncertainty Analysis in Public Health Risk Assessments,” 12 Risk Analysis 53– 64. Urbina, Ian (2013) “As OSHA Emphasizes Safety, Long-Term Health Risks Fester,” New York Times, March 30, page A1. U.S. Environmental Protection Agency (1986) “Hazardous Waste Management System: Land Disposal Restrictions,” Federal Register 51: 1602–1766. ———. (1992) “A Cross- Species Scaling Factor for Carcinogen Risk Assessment Based on Equivalence of Mg/Kg0.75/Day,” Federal Register 57: 24152–73. ———. (2004) Final Regulatory Analysis: Control of Emissions from Nonroad Diesel Engines. Office of Transportation and Air Quality, EPA420- R-04- 007, Washington, DC. ———. (2008) Child- Specific Exposure Factors Handbook. Washington, DC: U.S. Environmental Protection Agency. ———. (2011a) The Benefits and Costs of the Clean Air Act from 1990 to 2020. Washington, DC: U.S. Environmental Protection Agency. ———. (2011b) Regulatory Impact Analysis: National Emission Standards for Hazardous Air Pollutants for Industrial, Commercial, and Institutional Boilers and Process Heaters, http://www.epa.gov/ttnecas1/regdata/RIAs/boilersriafinal110221 _psg.pdf (accessed 3 March 2013). ———. (2011c) Exposure Factors Handbook: 2011 Edition. Washington, DC: U.S. Environmental Protection Agency. U.S. Food and Drug Administration (2004) “Prevention of Salmonella Enteritidis in Shell Eggs During Production: Proposed Rule,” 69 Federal Register 56824 –906. Walker, Katherine D., et al. (2001) “Use of Expert Judgment in Exposure Assessment: Part 1; Characterization of Personal Exposure to Benzene,” 11 J. of Exposure Analysis and Environmental Epidemiology 308–22.

Legislation Cited Job Creation and Wage Enhancement Act of 1995, H.R. 9, 104th Congress (1995). Regulatory Improvement Act of 1999, S. 746, 106th Congress (1999). Red Tape Reduction and Small Business Job Creation Act of 2012, H.R. 4078, 112th Congress (2012).

Case Cited Whitman v. American Trucking Assns., Inc., 531 U.S. 457 (2001).

Chapter 8

Happiness, Health, and Leisure Valuing the Nonconsumption Impacts of Unemployment Matthew D. Adler

The income effects of unemployment have been much studied (Davis and von Wachter 2011). Reduced income, in turn, means reduced consumption: the value (at market prices) of the goods and ser vices that an individual purchases and utilizes. However, it is clear that unemployment also has substantial nonconsumption effects. First, unemployment increases leisure— a benefit (Krueger and Mueller 2012). But unemployment can have significant nonconsumption effects beyond changes to leisure—in particular, effects on psychological well- being and physical health. Some of the correlation between unemployment and poor psychological or physical health may be due to the fact that these conditions reduce employability; but the evidence also suggests a causal arrow from unemployment to poor psychological and physical health. What accounts for this nexus? The psychological and health effects of unemployment may, in part, be the result of diminished consumption: the unemployed have less to spend on making themselves happy and healthy. But there is evidence of a more direct causal pathway. It is plausible to think that— even holding constant income—being jobless tends to produce setbacks to mental and physical health. In this chapter, I discuss how to incorporate the nonconsumption impacts of unemployment (including psychological and health effects, as well as leisure changes) into regulatory cost– benefit analysis (CBA). Doing so is important—because these effects are substantial—but poses large conceptual and empirical challenges. I begin by reviewing the evidence on the linkage between unemployment and, respectively, psychological ill fare and poor health. CBA

Happiness, Health, and Leisure

151

takes account of well- being changes by converting them into monetary equivalents (so- called compensating or equivalent variations). We thus need a model of individual utility that will provide a fi rm, theoretical basis for calculating an individual’s monetary equivalent for the nonconsumption effects of unemployment. I propose such a model and review the data sources that might be used to estimate these monetary equivalents. Although the utility model presented here shows that, in principle, monetary equivalents for the nonconsumption impacts of unemployment do exist, estimating these monetary values is by no means straightforward. Predicting how many jobs will be added or lost as a result of a given regulation can also be difficult. This important problem, discussed in other chapters in this book, is ignored here. My topic is not prediction but valuation: to the extent that a given regulation is predicted to change net employment (over some period), how should CBA practitioners value the resultant nonconsumption effects on the workers whose employment status changes? In other words, what is the monetary equivalent of a particular individual for moving between unemployment and employment (at a par ticu lar wage)?1 Masur and Posner (2012:618), in a recent but already quite influential paper, suggest that this monetary equivalent might be as high as $160,000. The aim of this chapter is to develop the utility-theoretic underpinnings for such a number and to discuss the difficulties in estimating what the number really is. It should be noted that the justification for CBA is controversial. On one view, CBA implements a Kaldor-Hicks test. A different view sees CBA as a proxy for a utilitarian or nonutilitarian social welfare function— a view that may argue for adjusting monetary equivalents with distributive weights (Adler 2012). The analysis here is independent of this controversy. The revised model of utility proposed in this chapter, and its attendant definition of monetary equivalent, fits smoothly with both the Kaldor-Hicks and social-welfare-function defenses of CBA.2

The Psychological and Health Effects of Unemployment: Some Evidence Psychological Effects A large literature in psychology and economics has grown up around “happiness” surveys (Adler 2013).3 These surveys ask respondents to quantify their current “happiness” or “life satisfaction” via questions such as “All things considered, how satisfied would you say you are with your life these days? Please tell me on a scale from 1 to 10, where 1

152

Matthew D. Adler

means very dissatisfied and 10 means very satisfied.” Survey questions are also posed about other attributes, allowing an estimation of the relative contribution of different attributes to individual happiness. The happiness impact of unemployment has been extensively studied by this literature (Blanchflower and Oswald 2004, 2011; Carroll 2007; Clark 2010; Clark et al. 2010; Dolan et al. 2008; Helliwell and Huang 2011; Knabe and Rätzel 2011a, 2011b; Ochsen and Welsch 2011; Winkelmann 2009). A general finding is that being unemployed correlates with reduced happiness. Formally, multiple regression is undertaken, with happiness or life satisfaction as the dependent variable and a host of individual attributes as independent variables—including employment status. The coefficient on this variable is consistently found to be nonzero. Although the link between psychological well-being and unemployment is now being studied with particular intensity by happiness scholars, that connection has long been of interest to psychologists. Paul and Moser’s (2009) meta-analysis examined hundreds of studies comparing the mental health of employed and unemployed individuals, with mental health measured using either a happiness scale or a scale of distress, depression, anxiety, self- esteem, or psychosomatic symptoms. Paul and Moser found a statistically significant association between unemployment and lower measures of psychological well-being on all the scales used. To be sure, correlation does not establish causation. Unemployed individuals might express lower levels of happiness because their dispositional unhappiness tends to produce job loss. However, a number of the happiness studies cited above have examined longitudinal data, which allow researchers to control for the possibility that an unhappy disposition causes both unemployment and lower stated happiness (by inserting so- called individual fi xed effects in the regression equation). Such studies, in estimating an individual’s happiness as a function of  her attributes, continue to fi nd a significant coefficient on employment status— suggesting that job loss causes unhappiness. More generally, Paul and Moser— in their meta- analysis of 86 longitudinal studies— found a reduction in psychological well-being associated with the move from employment to unemployment and an improvement associated with the move from unemployment to employment, both statistically significant. The longitudinal data just described help rebut the hypothesis that the unemployment– unhappiness association is merely the result of chronically unhappy individuals losing their jobs— but do not definitively establish causation, as it is possible that some temporary source of a worker’s unhappiness outside of work causes job loss. However, clever

Happiness, Health, and Leisure

153

longitudinal designs can control for that possibility. In par ticu lar, the fact that workers in mass layoffs tend to become unhappier is very strong evidence of a causal arrow from unemployment to unhappiness (Paul and Moser 2009:277). Several other facts about unemployment and psychological wellbeing emerge from this research literature. First, although individuals “hedonically adapt” to many other sources of happiness or unhappiness, this seems not to be true of unemployment (Lucas et al. 2004).4 Second, the psychological impact of unemployment is substantial, in the sense of effect size rather than statistical significance (Paul and Moser 2009). In a number of the above- cited happiness studies, the effect of unemployment on happiness is the same order of magnitude as large changes in income or major life events such as divorce or the death of a spouse. Third, the psychological impact of unemployment is mediated by social comparisons. It seems that unemployed individuals in areas with high unemployment tend to be happier (Clark 2003). Finally, the psychological impacts of unemployment seem to be partly independent of the effect of unemployment on income. This emerges from the happiness studies. In those studies, both employment status and income are regularly coded as independent variables. Employment status retains a statistically significant coefficient (both in cross- sectional studies and in longitudinal studies, including with individual fi xed effects). In other words, if we move an individual from employment to unemployment and hold constant his income, his happiness tends to decrease.5 It might be suggested that an association between current unemployment and current unhappiness, holding current income constant, does not show an independent pathway from unemployment to unhappiness. Perhaps unemployment is merely an indicator of permanent income. As between two individuals with different employment histories but identical lifetime income paths, are there happiness differences? This is a topic for further study— but, in a happiness regression with longitudinal data, Knabe and Rätzel (2011b) found a statistically significant coefficient on employment status even taking account of permanent income. To be clear: income changes are one mechanism by which unemployment causes unhappiness. What the empirical evidence suggests is that this is not the only such mechanism. But why would there be a direct pathway from unemployment to unhappiness? The psychological literature suggests various plausible “stories” (McKee- Ryan et al. 2005). “Work- role centrality” is one: many see work as a key component of a meaningful life. Relatedly, strong social norms in various cultures encourage work, and the unemployed may suffer a loss of social status. On a day-to- day basis, work structures time:

154

Matthew D. Adler

the unemployed may feel “at loose ends.” The unemployed person no longer has a social circle of coworkers— social interaction being one source of happiness. The process of job searching can be stressful and discouraging. Finally, the individual currently unemployed, whatever her current income, may be anxious about future income. Physical Health Although relatively few cross- sectional studies of unemployment and poor physical health have been undertaken, they do find an association (McKee- Ryan et al. 2005: table 3). Numerous studies find that unemployed workers are at higher risk of suicide (Wanberg 2012). Both kinds of findings, however, might be explained by the selection of unhealthy or suicidal workers into unemployment. Studies more probative on the question of causality fall into two groups. First, some longitudinal survey studies look to changes in an individual’s self-rated health or her reports of specific disease conditions after job loss. On balance, these studies fi nd a worsening in the unemployed worker’s health that cannot be fully explained by selection effects (Wanberg 2012). Yet more dramatic evidence of the causal effect of unemployment on health comes from recent longitudinal work using objective health indicators. Matching Pennsylvania employment records from the 1970s and 1980s with Social Security administration death records through 2006, Sullivan and von Wachter (2009) found that high-tenure male workers terminated in mass-layoff events had a much higher annual mortality risk during the years immediately following job loss, as compared with controls, and an elevated risk even decades following the layoff. In a 15-year longitudinal study of approximately 5,000 young males, JanickiDeverts et al. (2008) found that workers who were unemployed at the 10-year examination, or during the three previous years, tended to have elevated levels of C-reactive protein (CRP: a marker of inflammation linked to increased risk of diabetes, hypertension, and cardiac disease) at year 15, even controlling for CRP at the seven-year examination and for other factors. Maier et al. (2006) conducted sequential physical exams on recently unemployed workers who were healthy at the baseline of the study— finding an increase in cortisol during the first year of unemployment. Classen and Dunn (2012), looking at monthly, state- bystate data, found an association between the proportion of workers terminated in mass layoffs and the suicide rate. Is there an independent pathway from unemployment to physical health? Or is the causal connection just mediated by income? Janicki-

Happiness, Health, and Leisure

155

Deverts et al. (2008) found that the association between unemployment during years 7–10 and elevated CRP at year 15 was independent of income at year 15. On the other hand, Sullivan and von Wachter (2009) found that most (if not all) of the increased mortality risk incurred by the Pennsylvania workers terminated in mass layoffs could be accounted for by a reduction in their average long-run earnings and an increase in the variability of their earnings. More research on this topic is clearly needed. However, it is hard to see how there could be an independent pathway from unemployment to psychological but not physical well-being—given the physical health effects of stress and anxiety. At a bare minimum, such pathways are plausible. They are theoretically well supported, and there is substantial (although not yet conclusive) evidence in their favor. Our models of individual utility should allow for the reasonable possibility of a causal impact of unemployment on health and psychological well-being that is independent of changes in income (and leisure).

The Disutility of Unemployment: The Classical Model and Beyond Recall that our focus is on valuing the employment impacts of regulatory policy at the level of each individual worker. A regulatory policy has the effect, inter alia, of causing some worker to change her employment status. She goes from being employed at a certain wage to being unemployed (or vice versa). What is her monetary equivalent for that change? In this section, I discuss the limitations of the classical leisureconsumption utility model in valuing the health and psychological impacts of unemployment and then present a revised model. This revised model, in turn, is used to defi ne a monetary equivalent for unemployment—the estimation of which is discussed in the next section of this chapter. The revised model that I present is somewhat more technical than the rest of this chapter, and some readers may prefer to skip it. Its essence, though, is straightforward. In the revised model, individuals have preferences over health, happiness, consumption, and leisure. Consumption and leisure are not merely the objects of preference but have a second role: they are inputs into health and psychological “production functions,” along with additional individual attributes, in par ticu lar an individual’s employment status. These production functions represent the causal nexus from an individual’s various attributes (including employment status) to her health and psychological well- being. The monetary equivalent for unemployment is the sum of two terms: the change in consumption that results from unemployment,

156

Matthew D. Adler

plus the monetary equivalent for the nonconsumption effects of unemployment (MENCEU). But before developing this model further, I explain why a revised model is needed. Limitations of the Classical Utility Model In economics, an individual “utility” function is a mathematical tool that represents the extent to which someone’s preferences are satisfied. If some individual (for example, Tim) prefers one bundle of attributes to a second, then Tim’s utility function—if well constructed—will be such that it assigns a higher number to the first bundle than the second. In turn, Tim’s utility function can be used to assign a monetary equivalent to changes in his attributes. The classical view of utility in labor economics says that an individual’s utility function has two inputs: her consumption and her leisure (Cahuc and Zylberberg 2004:ch.1). This parsimonious and mathematically tractable function has, no doubt, been extremely fruitful. However, the consumption-leisure utility function has a key flaw, if our ambition is to arrive at an adequate valuation of the change in someone’s employment status, for purposes of CBA. Assume, first, that Tim cares not only about his consumption of marketed goods and the amount of leisure he enjoys but also about his health and his psychological well-being. Although he prefers to have more goods and more leisure, he also prefers to be in better health and to be at a higher level of psychological well- being. Ceteris paribus (for any given bundle of marketed goods and leisure), he prefers improvements in his physical health and he also prefers to feel happier, to experience more plea sure and less pain, to have a stronger sense of satisfaction with his life, and so forth. Such preferences for health and psychological well-being are quite intelligible, and the empirical evidence suggests that individuals often (if not always) possess them (Adler 2013). Assume, second, that consumption and leisure are not the only causal determinants of Tim’s health and psychological well-being. Tim’s nonconsumption and nonleisure attributes can, independently, affect how healthy he is and how happy he feels. In particular, by moving Tim from an employment state to the condition of being unemployed, we tend to reduce his health and psychological well- being, quite apart from the effect of unemployment on consumption and leisure. The literature reviewed earlier suggested that such a pathway is at least quite plausible.6 Imagine, now, that we move Tim from a condition of employment to a condition of unemployment but hold constant his consumption and

Happiness, Health, and Leisure

157

leisure. In the unemployment state, his foregone wages are replaced by nonwage income, and the time he would otherwise be working is spent in home production, job search, or other nonleisure activities. This thought experiment is not meant to be realistic but instead to shine a clear, theoretical, spotlight on the limitations of the classical labor-utility model. According to that model, Tim must be indifferent between the two conditions. Utility is just a function (in that model) of leisure and consumption, and by hypothesis those are equal in Tim’s particular case. But Tim need not be indifferent between unemployment and employment, even if consumption and leisure are equal in the two conditions. Because of the independent causal pathway running from nonconsumption and nonleisure attributes to health and happiness, Tim may well be less healthy and happy when unemployed. And because he has a preference for health and psychological well- being as well as for consuming marketed goods and for leisure, he is not indifferent. He prefers to be employed, and his utility function— if adequate to reflect his preferences— should assign the employment state a higher utility number. Can this limitation in the classical model be repaired by moving to a multiperiod model, such that an individual’s utility over his lifetime is a function of his leisure and consumption in each period (for example, each year)? No. The classical model here says that Tim must be indifferent between two identical lifetime streams of leisure and consumption, even if those streams differ in the health and psychological well-being Tim would experience in various periods (an empirical possibility, given the independent causal pathway). But if Tim cares about health and happiness as such, a more accurate utility function will assign a higher number to the stream associated with a higher level of these attributes.7 A Revised Model Thus a new model is needed that will not suffer the limitations of the classical model. For simplicity, the model presented here is a one-period model, and it does not include uncertainty, but it could easily be refined in both of these directions. In “consumer theory,” individual utility is solely a function of marketed goods and ser vices. Let m be a vector of such goods and ser vices. Then ui(.), the “direct” utility function of individual i, takes the form ui(m). The individual has a corresponding “indirect” utility function vi(.), defined as follows. If p is the vector of prices of marketed goods and ser vices, and c is the total consumption of individual i,8 then vi(c, p)

158

Matthew D. Adler

is the maximum direct utility that individual i can realize, at those prices and with that total consumption. We can then define monetary equivalents for changes in prices or consumption. Assume that, in the status quo, individual i faces prices p and has consumption c. A policy alters prices and consumption, to p* and c*, respectively. Then his monetary equivalent for the policy is the amount Δc such that vi(c* − Δc, p*) = vi(c, p). As already discussed, labor economists see leisure along with consumption as a determinant of individual utility. Let the direct utility function ui(.) now take the form ui(m, l), where l is individual leisure. Indirect utility, in turn, is now vi(c, l, p): the maximum achievable value of utility from leisure and marketed goods and ser vices, given total consumption c, leisure l, and prices p. The traditional utility function typically seen in labor economics, with utility = vi(c, l), is in turn a simplification of vi(c, l, p), achieved by holding prices constant or assuming a single consumption good. The direct utility function captures individuals’ “intrinsic” preferences— the attributes that individuals prefer “as such.” The indirect utility function captures “instrumental” preferences— the attributes that individuals prefer because these lead to intrinsically preferred attributes. My revised utility model allows individuals to have intrinsic preferences for health and psychological states. For simplicity, I assume that all the aspects of an individual’s health that he cares about can be summarized in a single number h. Similarly, the psychological states that the individual prefers or disprefers (affects, feelings of satisfaction, a sense of happiness, perceptions, emotions, and so forth) can be summarized in a single number e, for “experience.” Thus the direct utility function takes the form ui(m, l, h, e). For short, I will refer to the level of e as the individual’s “psychological well-being,” that is, the extent to which his preferences regarding his psychological states are realized. In order to define the corresponding indirect utility function, we need first to introduce some additional constructs and symbolism. Let B be “background attributes.” B encompasses all the causal precursors of an individual’s health and psychological states that the modeler wants to include— both properties of that individual and features of his society— other than the individual’s consumption, his leisure, and his employment status. Let E be an employment status indicator variable, with E = 1 indicating that the individual is employed and E = 0 indicating unemployment. A “health production function” h(.) takes as its arguments the goods consumed by an individual, his leisure, his employment status, and background attributes and maps these onto the individual’s health level h. Similarly, a “psychological production function” e(.) maps those arguments onto the individual’s level of psycho-

Happiness, Health, and Leisure

159

logical well-being e. Thus h(.) takes the form h(m, l, B, E), and e(.) the form e(m, l, B, E). Unemployment can now be modeled as an independent drag on psychological well-being and health, meaning that e(m, l, B, 0) is less than e(m, l, B, 1) and h(m, l, B, 0) is less than h(m, l, B, 1) for any given bundle of consumption goods, leisure time, and nonemployment background attributes. A moment’s reflection suggests that a further refinement is needed. If employment, as such, tended to increase health and happiness, rational individuals might be willing to accept employment at negative wages. And yet we do not observe wage contracts with negative wages (although we do observe volunteering). Plausibly, the pathway from employment status to health and psychological well-being involves the wage level w. An individual who had to pay for the privilege of work might well feel stigmatized and seriously undervalued (even though he would still reap time- structure benefits from employment). Conversely, an increase in wage rate is a marker of improved social status, and our model should allow for such increase to have a psychological and health benefit, ceteris paribus. With this refinement, the health production function takes the form h(m, l, B, E, w)—health is now a function of marketed goods and services, leisure, employment status, the wage rate (if employed), and background attributes B. A similar structure is imposed on the psychological production function e(.). We can now define the indirect utility function. Let vi(.) = vi(c, p, l, B, E, w), namely, the maximal value of utility from marketed goods and ser vices, leisure, desired psychological states, and health that can be realized given total consumption c, prices p, leisure l, employment status E, wage rate w (if employed), and background attributes B. This model gives us the apparatus for calculating a monetary equivalent for unemployment. Imagine that an individual i with a particular set of background attributes B and wage rate w becomes unemployed. His consumption changes from c to c* (reflecting lost wages and the receipt of unemployment benefits); his leisure changes from l to l*. Then his monetary equivalent for unemployment is defined implicitly as Δc, such that vi(c, p, l, B, 1, w) = vi(c* − Δc, p, l*, B, 0).9 More precisely, Δc is the compensating variation for unemployment. One can also define an equivalent variation— although this would be yet more complicated to estimate than the compensating variation.10 Note that the monetary equivalent Δc can be expressed as the sum of two terms. One term is the actual consumption change (from c to c*). The second term (call it Δc +) is the monetary equivalent for the “pure” case in which c* = c, that is, consumption is held constant and

160

Matthew D. Adler

unemployment produces only nonconsumption (leisure, health, and happiness) changes. I will refer to Δc + as the monetary equivalent for the nonconsumption effects of unemployment (MENCEU). Δc, in turn, is the sum of the actual consumption change (c* − c) and Δc +.11 The model of utility proposed here may seem idiosyncratic, but it in fact has at least four precedents in the economics literature. First, health economists often use health as one of the inputs into the direct utility function. And health scholars often use “health” to mean psychological as well as physical functioning. The direct utility function I propose, ui(m, l, h, e), can be seen as supplementing consumption goods and leisure with two health arguments, one physical and the other psychological. Second, Becker and Rayo (2008), in commenting on the Easterlin paradox (that the increase in average incomes over time does not seem to have increased average happiness), have suggested that rational individuals may have preferences both for happiness and for other attributes. Third, health production functions are common in health economics (Freeman 2003:ch. 10). Finally, Akerlof (1980), in modeling unemployment, has incorporated compliance with social norms as a separate entry into the utility function and has argued that those norms take account of the wage at which the worker is employed (as compared with a social reference value; see Clark 2003).

Estimating MENCEU I have defined MENCEU—the monetary equivalent for the nonconsumption effects of unemployment—within the context of a revised model of individual utility. Further refinements of the model are certainly possible, but it seems a reasonable starting point in defining monetary equivalents for unemployment that take account of preferences for health and psychological well- being and of the complex causal determinants of these attributes. The monetary equivalent for unemployment is, in turn, the sum of the direct consumption losses from unemployment and MENCEU. Unfortunately, although MENCEU is perfectly well defined within this model, estimating MENCEU by using available data sources is not straightforward. Here I briefly discuss the use of revealed- preference data, as well as various kinds of survey data (stated-preference, happiness, and quality-adjusted life year [QALY] surveys), as sources of information about MENCEU. Space limitations preclude a longer discussion, but these brief remarks should at least persuade the reader that substantially more research is needed before plausible estimates of MENCEU can be formulated.

Happiness, Health, and Leisure

161

Behavioral Data Behavioral (“revealed preference”) data are widely used by CBA practitioners to estimate monetary equivalents. In par ticular, within the context of the classical leisure- consumption utility model, MENCEU can be estimated from the market supply curve of labor, observed as the market wage changes. Why? From the market supply curve, we can gain information about workers’ “reservation wages” (Bartik 2012). Moreover, within the classical leisure- consumption utility model, the only nonconsumption impact of unemployment is increased leisure, and MENCEU therefore turns out to be equal to the reservation wage times labor hours. By contrast, MENCEU in the context of the revised utility model I have presented is not equal to the reservation wage times labor hours.12 It is not clear, therefore, how to estimate MENCEU from the market supply curve once health and psychological well-being are included as nonconsumption effects of unemployment. Stated Preference Surveys The “stated preference” methodology, now widespread in environmental economics (albeit still controversial), infers respondents’ monetary equivalents by posing survey questions asking how the respondents would make tradeoffs between income or consumption and nonmonetary attributes. MENCEU is the monetary equivalent for a package of psychological and nonpsychological (health and leisure) changes, and indeed there are precedents in stated-preference research for valuing such packages. Stated willingness to pay to reduce a cancer risk is intended to capture both the physical health and mortality impact of cancer and the experiential costs (suffering, anticipatory anxiety) that go along with having cancer or being at risk thereof (Hammitt and Haninger 2010). Similarly, the “fear of crime” literature has employed stated-preference surveys to value the psychological as well as physical and financial effects of being a crime victim or facing an increased risk of crime (Cohen et al. 2004). How might the stated-preference methodology be used to estimate MENCEU? Consider taking an individual employed at some wage, with a certain level of income, leisure, and background attributes, and asking him: “Imagine becoming unemployed, but with unemployment benefits making up for the loss in income. A benefit of unemployment is free time. However, you will experience increased stress, leading to unhappiness and poorer health. Given this combination of changes, how much more in unemployment benefits would you require in order to be just as

162

Matthew D. Adler

well off as you are now?” Or we take an unemployed individual with some income (for example, from unemployment benefits) and ask: “Imagine becoming employed, with no change to your net income, but with less stress and thus more happiness and better health. How much more in unemployment benefits would you need to be just as well off as you would be then?” Monetary equivalents, for purposes of CBA, should generally be based on individuals’ rational and well-informed preferences. Thus answers to the sorts of questions posed in the previous paragraph are evidence of MENCEU only if respondents meet these idealizing conditions. But they may not. Employed individuals may not fully grasp what it feels like to be unemployed (a kind of informational failure). Unemployed individuals may grasp that, but if sufficiently depressed or anxious they may not be reasoning about monetary valuations in a calm and thoughtful way (a kind of rationality failure). Moreover, recall that MENCEU is calculated by taking account of the role of consumption and leisure in producing health and happiness. Respondents to stated-preference surveys may well be poor “hedonic forecasters”: they may fail to grasp the happiness impact of changes to consumption or leisure (a kind of informational failure). Finally, determining MENCEU requires a complex, optimizing calculation: it is the change in consumption just sufficient to counterbalance the leisure, health, and psychological well-being impacts of unemployment, taking account of preferences for all items and the particular form of the health and psychological production functions. Respondents may find it very difficult to perform this calculation (a kind of rationality failure). Are failures of rationality and information an especially grave problem for using stated-preference surveys to estimate MENCEU—as compared with stated-preference estimates of the value of statistical life or other cognitively complex policy impacts? Perhaps not. And perhaps debiasing and information- provision techniques that have had some success in other areas could be used in arriving at reasonable, surveybased estimates of MENCEU. Still, there are serious difficulties to be overcome. Happiness Surveys Happiness surveys are increasingly used to estimate monetary equivalents for purposes of CBA. Imagine that the coefficient on income in a happiness regression is y: an increase in annual income of $1 tends to increase expressed happiness or life satisfaction by y. And imagine that the coefficient on some other individual characteristic (whether scalar

Happiness, Health, and Leisure

163

or dichotomous) is g. An increase in one unit of that good, or a shift from not having to having the good, tends to increase stated happiness or life satisfaction by g. Then the monetary equivalent for one unit of the good, or for having it (in the dichotomous case), can— seemingly— be estimated as g/y (Adler 2013). The happiness- survey approach has, specifically, been used repeatedly to arrive at a monetary equivalent for unemployment (see Caroll 2007 and studies cited therein; Knabe and Rätzel 2011b). The Masur and Posner (2012) estimate of MENCEU is based, in part, on happiness surveys. Happiness surveys provide strong evidence of the impact of unemployment on psychological well-being. My discussion in the first part of this chapter relied heavily on such surveys. But they do not offer good evidence of MENCEU. Why not? MENCEU, as defined, reflects individual preferences both for psychological well-being and for other items: consumption, leisure, physical health. The happiness- survey-based monetary equivalent for unemployment tells us the change in income (and thus the change in consumption) with the same happiness impact as moving someone from employment to unemployment. But that quantity is not the same as MENCEU. For example, someone might be hedonically insensitive to consumption, but have a strong preference for consumption, not merely happiness. In that case, the happiness- surveybased monetary equivalent for unemployment will be an overestimate of MENCEU. Money, in such a case, is hedonically useless for the individual, but it is still a source of utility in the sense of preference satisfaction: it still can buy things that the individual wants.13 What about using happiness surveys as evidence of the happiness component of MENCEU? Might MENCEU be approximated as the sum of the monetary equivalent for the health and leisure changes that occur in unemployment, plus the monetary equivalent for the psychological well- being changes— with the latter estimated using happiness surveys? Again, the problem arises that a monetary equivalent for the psychological changes occurring in unemployment—if used for CBA— should reflect preferences for both psychological well-being and consumption. Happiness surveys do not evidence such preferences. Instead, they evidence something different, namely, the relative casual contribution of unemployment, income, and other attributes to happiness. The objection here to happiness surveys as evidence of MENCEU transcends the specifics of the revised utility model I have offered in this chapter. A yet better model— although one further removed from standard utility functions used in economics—might see individual preferences for consuming marketed goods as being wholly derived

164

Matthew D. Adler

(“instrumental”). In such a model, consumption is merely an input: to happiness, to health, and to the achievement of various goals that individuals have. Individuals never prefer consumption as such, intrinsically, but only because of what consumption yields. Still, happiness surveys would remain a problematic basis for estimating MENCEU. MENCEU would reflect the role of consumption as a precursor to the multiplicity of attributes that individuals prefer (happiness, health, and goal fulfillment), not merely happiness. QALY-to-Dollar Trade- offs Much scholarship in health policy mea sures health states on a QALY scale, with zero signifying a health state equivalent to premature death and one signifying perfect health. These numbers are ascertained via surveys that pose time trade- off or standard gamble questions or simply ask individuals to assign a given health state a number between zero and one. Many such surveys have been undertaken, attaching QALY numbers not merely to physical diseases but also to setbacks to psychological health (depression, anxiety). The QALY value of a health state, then, is this zero- to- one value multiplied by the duration of the state (Freeman 2003:ch. 10). It is tempting to think that we can ascertain monetary equivalents (in the CBA sense) for health changes by applying a QALY-to- dollar conversion factor to the associated change in QALY values. Indeed, a burgeoning literature addresses the viability of doing this (Baker et al. 2011). QALY- to- dollar conversions have, specifically, been used to monetize the well-being loss associated with fear, anxiety, and other psychological conditions (Dolan and Peasgood 2007). Might QALY- to- dollar conversions be employed to estimate MENCEU? Perhaps. One would translate the change in health and psychological well- being associated with unemployment onto a QALY scale, and then multiply by the conversion factor. But such an approach ignores leisure impacts. Moreover, the psychological states to which QALY values have been assigned are negative states, departures below a neutral level of no affect, rather than positive emotions such as joy, contentment, “flow,” or ebullience. Why? It is much easier to conceptualize the top point of the zero– one scale as a condition of no physical or mental disease rather than a condition of no physical disease and maximal positive happiness. Insofar as the psychological impact of unemployment is both the loss of positive happiness and an increase in negative affective states, QALY-to- dollar conversions focused on the latter will understate the monetary equivalent. Finally, as a general matter, the QALY-to- dollar approach may provide an inaccurate mea sure of someone’s monetary

Happiness, Health, and Leisure

165

equivalent (compensating or equivalent variation) in the CBA sense (Freeman 2003:ch. 10).

Conclusion This chapter is, ultimately, a plea for research. We do not yet know enough to assign a monetary value to unemployment. The standard leisureconsumption utility model is not adequate to take account of the direct pathway from unemployment to poor health and psychological well-being. We need more scholarly work in arriving at a new model—the revised model I have presented here is simply one possibility—and in determining how to use surveys or behavioral evidence to estimate the monetary equivalent for the health, psychological, and leisure changes that occur as a result of being unemployed. Decades of scholarship were needed to arrive at our current understanding of the valuation and estimation of regulatory effects on the environment. A similar research effort may well be needed for the valuation of employment. In particular, as I have discussed, happiness surveys provide evidence of the causal impact of unemployment on happiness, but the now-popular method of using such surveys to estimate a monetary equivalent for unemployment is problematic—by ignoring preferences for attributes other than happiness. It might be objected that if regulation has little net impact on jobs, research into the valuation of employment impacts is unnecessary. But a regulation may arrive at a zero net employment impact by shifting jobs from those with larger to those with smaller monetary equivalents (and thus have a nonzero CBA impact). Until we refine our understanding of how to value unemployment at the worker level, no firm conclusions can be reached about the appropriate role of employment impacts in regulatory CBA.

Notes 1. The employment status of a given worker may also have third-party effects. For example, layoffs may cause those who continue in employment to feel less secure in their jobs (Clark et al. 2010). A full CBA would take account of such effects; but, given space constraints, I will focus here on valuing the fi rst- party effects of unemployment. 2. Some writing on employment impacts and CBA characterizes such impacts as “distributional”— suggesting that a Kaldor- Hicks framework would ignore them. But why should we presume that the monetary- equivalent costs incurred by individual workers who become unemployed as a result of regulation—whether calculated using the classical leisure- consumption utility model or a different utility model— are, in general, precisely or roughly counterbalanced by benefits or costs to other economic actors?

166

Matthew D. Adler

3. Happiness surveys are also called “subjective well- being” surveys. I will use the term “happiness” to mean what surveys in the happiness/subjective wellbeing literature typically ask about (namely, a feeling of happiness or a sense of satisfaction with life) and “psychological well- being” to refer more generally to any mental states (affects, cognitions, perceptions, and so forth) the occurrence of which the individual prefers or disprefers. 4. Relatedly, being unemployed has a “scarring” effect, reducing the individual’s psychological well- being even after reemployment (Knabe and Rätzel 2011b). 5. Nonwage sources of income that unemployed individuals might receive include unemployment insurance benefits, investment income, or (if “income” is understood as the individual’s share of household income) spousal wages. Helliwell and Huang (2011) found that the psychological impact of unemployment is not moderated by the level of available unemployment benefits— consistent with the direct, income-independent pathway from unemployment to unhappiness. 6. Admittedly, the empirical studies reviewed earlier that examine the unemployment–unhappiness nexus and that control for income do not control for leisure. This evidence thus does not rule out the hypothesis that the psychological and health impacts of unemployment are mediated by leisure changes, rather than by unemployment per se. But this hypothesis is very implausible— given the evidence that unemployment increases leisure. The hypothesis would imply that increasing the leisure of employed workers, holding constant their income, tends to reduce (not increase) their psychological well-being and health. 7. It might be objected that the impact of unemployment on health and psychological well- being can be captured by a lifetime model in which future income is uncertain, and unemployment creates anxiety about future income. Lifetime utility is a function of consumption and leisure in each period; individuals in each period have probabilistic beliefs about the future flow of consumption and leisure; unemployment reduces expected lifetime utility (as calculated in a given period) by changing such beliefs. However, this model fails to capture an intrinsic preference for psychological well- being and health apart from consumption and leisure; nor does it capture non-anxiety-mediated linkages from unemployment to poorer psychological and physical health (such as social stigma, social isolation, and the lack of daily time structure). 8. The vector product of m and p. 9. If Δc is negative (as is presumably the case), this means that consumption must be increased to offset the effects of unemployment. The wage rate with unemployment is 0, so need not be included with E = 0. 10. An individual’s equivalent variation for unemployment, relative to the state of being employed at wage w, is the change in income while employed that makes her just as well off as unemployed. But because our utility model incorporates a direct pathway from an individual’s wage to her health and happiness, the definition of equivalent variation requires that w be held constant and income varied. It would seem more cognitively demanding for the respondent to a stated-preference survey to imagine that, than to imagine changes to income while unemployed (e.g., changes in the level of unemployment benefits). 11. Let Δc+ be such that vi(c, p, l, B, 1, w) = vi(c − Δc+, p, l*, B, 0). Putting this equation together with the equation in the previous paragraph, it follows that vi(c* − Δc, p, l*, B, 0) = vi(c − Δc+, p, l*, B, 0), or that c* − Δc = c − Δc+, that is, that Δc = (c* − c) + Δc+.

Happiness, Health, and Leisure

167

12. For purposes of this footnote only, let Δc + denote the individual’s monetary equivalent for the nonconsumption effects of unemployment in the sense of the equivalent rather than compensating variation. Consider first the classical model, such that vi(.) = vi(c, p, l). Assume that the individual, if unemployed, would have consumption c* and leisure l*. If employed, his nonconsumption attributes, that is, leisure, would change to l. Then Δc + is implicitly defined by vi(c* + Δc +, p, l) = vi(c*, p, l*). Individual i’s reservation wage w res is the wage that would make him indifferent between employment and unemployment, holding fi xed his nonwage income. Also note that if the individual were employed at a given wage, he works l* – l hours, so that his wage income is the wage multiplied by (l* – l). Thus w res is such that vi(c* + w res(l* – l), p, l) = vi(c*, p, l*). So it follows immediately that, within the classical model, w res(l* – l) = Δc +. Consider now the richer utility model proposed in this chapter. If unemployed, the individual would have leisure l* and consumption c*. If employed, his wage would be w—the market wage for his skills— and his leisure would be l. Δc + is now such that vi(c* + Δc +, p, l, B, 1, w) = vi(c*, p, l*, B, 0). The concept of reservation wage remains the same: the wage that makes the individual indifferent between employment and unemployment, holding fixed nonwage income. But that quantity, w res, must now satisfy the equation vi(c* + w res(l *– l), p, l, B, 1, w res) = vi(c*, p, l*, B, 0). It no longer follows that w res(l* – l) = Δc +. The problem is that wages are no longer merely a source of consumption but now are both such a source and a separate argument in the utility function (so as to allow for the direct causal impact of the wage level on health and psychological well- being). 13. Abusing notation slightly, let e(c, p, l, B, E, w) be the level of psychological well-being that would be realized by an individual with consumption c, prices p, leisure l, background attributes B, employment status E, and wage w. Let Δz be the monetary equivalent for unemployment estimated from happiness surveys, that is, the change in consumption with the same psychological impact as moving from employment to unemployment. Then Δz is such that e(c, p, l, B, 1, w) = e(c − Δz, p, l*, B, 0). MENCEU, recall, is Δc+ such that vi(c, p, l, B, 1, w) = vi(c − Δc +, p, l*, B, 0). Because vi(.), the indirect utility function, reflects preferences for nonpsychological as well as psychological attributes, there is no reason to think that vi(.) is ordinally equivalent to e(.), and thus no reason to think that Δz = Δc +.

References Adler, Matthew (2012) Well- Being and Fair Distribution: Beyond Cost- Benefit Analysis. New York: Oxford University Press. ———. (2013) “Happiness Surveys and Public Policy,” Duke Law Journal 62: 1509–1601. Akerlof, George (1980) “A Theory of Social Custom, of Which Unemployment May Be One Consequence,” Quarterly Journal of Economics 94: 749–75. Baker, Rachel et al. (2011) “Searchers vs. Surveyors in Estimating the Monetary Value of a QALY,” Health Economics, Policy and Law 6: 435– 47. Bartik, Timothy (2012) “Including Jobs in Benefit– Cost Analysis,” Annual Review of Resource Economics 4: 55–73. Becker, Gary, and Luis Rayo (2008) “Comments on ‘Economic Growth and Subjective Well-being: Reassessing the Easterlin Paradox,’ ” Brookings Papers on Economic Activity 88– 95 (Spring).

168

Matthew D. Adler

Blanchflower, David, and Andrew Oswald (2004) “Well- Being over Time in Britain and the USA,” Journal of Public Economics 88: 1359– 86. ———. (2011) “International Happiness: A New View on the Mea sure of Per formance,” Academy of Management Perspectives 25: 6–22. Cahuc, Pierre, and André Zylberberg (2004) Labor Economics. Cambridge, MA: MIT Press. Carroll, Nick (2007) “Unemployment and Psychological Well- Being,” Economic Record 83: 287– 302. Clark, Andrew (2003) “Unemployment as a Social Norm: Psychological Evidence from Panel Data,” Journal of Labor Economics 21: 323– 51. ———. (2010) “Work, Jobs, and Well- Being Across the Millennium,” in E. Diener et al., eds., 436– 64. International Differences in Well-Being. New York: Oxford University Press. Clark, Andrew, et al. (2010) “Boon or Bane? Others’ Unemployment, WellBeing and Job Insecurity,” Labour Economics 17: 52– 61. Classen, Timothy, and Richard Dunn (2012) “The Effect of Job Loss and Unemployment Duration on Suicide Risk in the United States,” Health Economics 21: 338– 50. Cohen, Mark, et al. (2004) “Willingness- to- Pay for Crime Control Programs,” Criminology 42: 89–109. Davis, Steven, and Till von Wachter (2011) “Recessions and the Costs of Job Loss.” NBER Working Paper No. 17638, Cambridge, MA. Dolan, Paul, and Tessa Peasgood (2007) “Estimating the Economic and Social Costs of the Fear of Crime,” British Journal of Criminology 47: 121–32. Dolan, Paul, et al. (2008) “Do We Really Know What Makes Us Happy?” Journal of Economic Psychology 29: 94–122. Freeman, A. Myrick (2003) The Measurement of Environmental and Resource Values, 2nd ed. Washington, DC: Resources for the Future. Hammitt, James, and Kevin Haninger (2010) “Valuing Fatal Risks to Children and Adults,” Journal of Risk and Uncertainty 40: 57– 83. Helliwell, John F., and Haifang Huang (2011) “New Mea sures of the Costs of Unemployment.” NBER Working Paper No. 16829, Cambridge, MA. Janicki- Deverts, Denise, et al. (2008) “History of Unemployment Predicts Future Elevations in C-Reactive Protein Among Male Participants in the Coronary Artery Risk Development in Young Adults (CARDIA) Study,” Annals of Behavioral Medicine 36: 176– 85. Knabe, Andreas, and Steffen Rätzel (2011a) “Scarring or Scaring?,” Economica 78: 283– 93. ———. (2011b) “Quantifying the Psychological Costs of Unemployment,” Applied Economics 43: 2751– 63. Krueger, Alan, and Andreas Mueller (2012) “Time Use, Emotional Well- Being, and Unemployment,” American Economic Review: Papers and Proceedings 102: 594– 99. Lucas, Richard, et al. (2004) “Unemployment Alters the Set Point for Life Satisfaction,” Psychological Science 15: 8–13. Maier, Richard, et al. (2006) “Effects of Short- and Long-Term Unemployment on Physical Work Capacity and on Serum Cortisol,” International Archives of Occupational and Environmental Health 79: 193– 98. Masur, Jonathan, and Eric Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634.

Happiness, Health, and Leisure

169

McKee- Ryan, Frances, et al. (2005) “Psychological and Physical Well- Being During Unemployment,” Journal of Applied Psychology 90: 53–76. Ochsen, Carsten, and Heinz Welsch (2011) “The Social Costs of Unemployment,” Applied Economics 43: 3999– 4005. Paul, Karsten, and Klaus Moser (2009) “Unemployment Impairs Mental Health,” Journal of Vocational Behavior 74: 264– 82. Sullivan, Daniel, and Till von Wachter (2009) “Job Displacement and Mortality,” Quarterly Journal of Economics 124: 1265–1306. Wanberg, Connie (2012) “The Individual Experience of Unemployment,” Annual Review of Psychology 63: 369– 96. Winkelmann, Rainer (2009) “Unemployment, Social Capital, and Subjective Well- Being,” Journal of Happiness Studies 10: 421–30.

Chapter 9

A Research Agenda for Improving the Treatment of Employment Impacts in Regulatory Impact Analysis Ann E. Ferris and Al McGartland

Benefit–cost analysis (BCA) is one of the dominant paradigms for evaluating regulatory decisions. In 2011, President Obama reaffirmed BCA’s role with Executive Order 13563 (Obama 2012). Not surprisingly, new political appointees and other senior policy officials are always anxious to learn about BCA and how economists conduct it. If economists are the ones “scoring” policy proposals, decision makers naturally want to understand the methods for assessing benefits and costs. Some are surprised to learn that economists have a well- defined conceptual underpinning to BCA, dating back 75 years (Hicks 1940; Kaldor 1939). They do not change the conceptual framework and the implementation rules on the basis of policymakers’ preferences. Three key elements of BCA bear emphasis. First, economists do not assign values on the basis of what they think, or what policymakers think, the value should be. Instead, they apply a consumer sovereignty principle and use the best available economic methods to assess values on the basis of what people would be willing to pay for changes brought about by policy. Second, BCA provides an efficiency test, not an equity test. Economists do not value a dollar of benefits going to those worse off any differently than a dollar of benefits going to the wealthy. Finally, transfers are neither benefits nor costs. For example, if people are expected to live longer because of a policy change, the increased Social Security payments they will receive are not considered benefits (to the recipients) or costs (to the government)—they are simply transfers. Recently, policymakers, stakeholders, and the public more broadly have focused attention on how BCA accounts for employment impacts.

A Research Agenda for Improving the Treatment of Employment Impacts

171

The term “employment impacts” has various meanings throughout the literature, but if any changes in employment from a regulation are to be incorporated into BCA, these changes must be estimated in terms of their net aggregate monetized value. This aggregate value is composed of two parts: employment change, as mea sured by the net number of jobs affected by the policy, and the monetized value of those affected jobs. Invariably, newcomers to BCA simply believe that job creation is a benefit and that job loss carries a cost. Yet in assessing benefits and costs, economists have not agreed with this layperson’s view (Hopkins 1992; Jaffe et al. 1995:133). At EPA’s National Center for Environmental Economics (NCEE), we have briefed policymakers specifically on this point over the years and have generally supported the traditional economists’ view that employment should not be included in BCA. This is a difficult conclusion for noneconomists to accept, particularly when the economy is underperforming and unemployment remains unacceptably high. There is, of course, a simple defense of the traditional position. BCA generally assumes a full- employment economy, where labor is shifted toward producing cleaner air and water, meaning that labor is no longer available to produce other products, with no overall net change in jobs. Bastiat’s (1850) broken window fallacy is extremely helpful to explain economists’ traditional position. When the shopkeeper’s window is broken, customers console the shopkeeper by advising that his expenditure on the window repair will provide work for the glazier. Bastiat points out that the customers count only “that which can be seen.” They do not see that the shopkeeper would have otherwise used his savings for new equipment, expansion, or consumption, thereby putting others to work. The broken window did not add “net” jobs to the economy. But even Bastiat’s story needs additional elaboration and considerations. What if the shopkeeper would have spent his savings on a vacation in England? Or what if he would not have spent his savings at all for many years, even when the town was experiencing an abnormally high unemployment rate? Although Bastiat’s story is valid, it rests on a set of assumptions that make it not easily generalizable. Bastiat’s framework asks us to consider both the benefits and costs of actions—whether obvious or hidden— even if they may ultimately offset each other such that their net effect is negligible. In the broken window example, the presumed gain in employment roughly equals the unobserved loss elsewhere because of the shift in economic activity. Therefore, it can be helpful to examine the offsetting employment changes that are presumed to underlie the traditional BCA approach under conditions of full employment. However,

172

Ann E. Ferris and Al McGartland

there has been surprisingly little conceptual work to guide BCA on this issue (Haveman and Farrow 2011; V. K. Smith 2012). Whether a regulation’s effects on employment are classified as benefits and costs— or whether they should instead just be identified and estimated separately as employment impacts— matters to regulatory economists and agency decision makers. If employment changes are classified as benefits and costs, rather than as impacts, that will affect the final determination of “net” benefits and perhaps may influence the preference of various policy outcomes. In that vein, incorrect accounting and classification of these impacts could lead to poor policy decisions. Clarity is needed about whether employment impacts, mea sured either as changes in jobs or earnings, may be categorized as changes in social welfare. Providing clear and appropriate social welfare estimates of employment impacts becomes even more complicated given the need to take into account general economic conditions at the time of the analysis, including modeling open economy dynamics. Moreover, economists currently lack the models and data to quantify these impacts well for specific regulations. It is our strong contention, and the main point of this chapter, that additional research is needed in a number of areas to resolve the theoretical treatment of jobs within BCA and to improve economists’ ability to estimate and value the employment impacts from regulations. The Great Recession of 2007–2009 and the subsequent sluggish recovery have challenged the underlying standard assumption that employment changes can be ignored in BCA because labor use is a “cost” and the wage rate reflects that cost. Furthermore, policymakers and the public have clearly grown very interested in knowing more about the employment implications of regulations. We face an acute need for better methods to assess these employment impacts, as well as a need to revisit the traditional approach and reconsider in depth the theoretic foundations for whether employment impacts should be included directly within BCA.

Conceptual Treatment of Employment Within BCA Economics has yet to provide a widely accepted unified theory that incorporates employment impacts into BCA. We illustrate the need for such a unified theory by summarizing approaches that have been used or recommended by economists in dealing with employment impacts in the context of regulatory analysis. Although these approaches are not mutually exclusive, they necessarily imply certain characteristics about the macroeconomy and how markets respond to regulation. Just as macroeconomists rely on macreconomic models to answer questions about the impacts of fiscal and monetary policy, and microeconomists use

A Research Agenda for Improving the Treatment of Employment Impacts

173

micro models to estimate impacts of policies on particular industries or groups of individuals, the literature offers multiple ways to approach estimating the employment impacts of regulation. Equilibria Matter Schumpeter (1942) observed that the economy is continually in the process of creative destruction, moving efficiently toward new equilibria, where the economy is stabilized, no longer buffeted by changes or shocks to the system. The equilibrium concept is primarily conceptual, rather than practical: although economists use equilibria as a kind of benchmark for the status of the economy, in practice, economies are constantly in flux. The dominant approach in BCA is to compare old and new fullemployment equilibria; the transitions and transition costs are generally ignored. Based on the principle of efficient markets, economic theory instructs that regulation should address the market failure (for example, pollution externalities) by adjusting prices and then step back to let the private market use those prices to make production decisions. For example, if BCA is to be a “private market test,” one could argue that the costs weighted by the private sector in making production decisions are what matter. If private firms do not incorporate the social cost of worker displacement (or the benefits of reduced worker displacement) into their decision making because labor markets already exist and serve some of these functions, then perhaps BCA should not either. Of course, private businesses are required to consider displaced worker costs, at least in part, through the unemployment insurance program. But the connection of these costs to the social costs of a displaced worker is far from clear. If instead of a private market test using the “equilibria- at- fullemployment” approach, analysts were to value employment impacts within BCA alongside environmental quality, they may encounter a situation where the BCA concludes that policymakers should (or should not) regulate on the basis of employment impacts instead of environmental quality. With employment continuously varying, due to Schumpeter’s creative destruction, including employment impacts within the BCA may rely on employment estimates that will change over time and therefore drive variation in net benefits, completely independent of environmental policy targets. These employment impacts within BCA also have the potential to be double counted if we include measures for both workers and firms in terms of the labor portion of their compliance costs. Another potential result from departing from the equilibria- at-fullemployment approach could be that the regulating agency would

174

Ann E. Ferris and Al McGartland

design regulations requiring labor-intensive solutions, instead of costeffective ones. For example, regulations that require fi rms to clean up chemical spills using manual labor instead of power equipment may “cost” more, but the employment estimates may tilt the net benefits to favor such a labor-intensive option. In addition, including employment impacts in BCA may lead to another unintended consequence: heavily polluting entities would have the most to gain by becoming more laborintensive in order to reduce the relative stringency of their environmental regulation, all else held equal. Transition Costs Matter Under a transition costs approach, the costs that displaced workers must incur in finding new employment are included in BCA. Similarly, the job search costs not incurred when workers find jobs as a result of a regulation are included as benefits. The lost productivity, both in the short run and possibly the long run, as well as associated transition costs, including lost earnings and retraining costs, should be counted— but lost productivity should not be double counted in terms of both workers and firms. Conversely, if a regulation generates demand for labor—for example, it calls for building and operating pollution abatement equipment— the transition costs of previously unemployed workers may be mitigated as they fi nd new employment in building or operating the required equipment. These transition costs are likely to depend on the state of the labor market, that is, whether the economy is at or near full employment. With this approach, these costs should be included in BCA because workers have a willingness to pay to avoid layoffs and associated transition costs in finding new employment. Similarly, unemployed workers have a willingness to pay to find new employment more quickly, avoiding transition costs associated with searching for new work. Even though there are employment placement ser vices, job training programs, and unemployment insurance, there is a missing market for avoiding layoff risk—no insurance market exists for this particular risk, so analysts cannot observe market prices. The question arises of whether and how to incorporate the willingness to pay to avoid transition costs from layoffs within BCA. Opportunity Costs Matter A third approach considers how to account for the value of labor when the market wage is not an accurate mea sure of the opportunity cost of time, that is, for unemployed workers. Some have argued that unem-

A Research Agenda for Improving the Treatment of Employment Impacts

175

ployed workers have a lower opportunity cost than employed workers. If a regulation leads to unemployed workers being hired to build and operate pollution control equipment, for example, the opportunity costs for society are lower because these workers were not productively employed prior to the regulation. The symmetrical argument for workers displaced by regulations is also relevant and may overlap with the transition costs approach. Opportunity costs are also likely to depend on the state of labor markets, for example, how far away is the economy from full employment. Delving into this research area gets messy very quickly; although there is long- standing consensus that the market wage is the appropriate measure for the opportunity cost of time for employed workers, the value of time is unobserved for unemployed workers. Labor economists have shown that a reservation wage—the lowest wage an unemployed worker will accept to go back to work—is an appropriate measure. However, reservation wages are unobserved and difficult to estimate. Labor Markets Matter According to a fourth approach, employment impacts should be incorporated into BCA depending on the state of labor markets. When the economy faces high unemployment, when specific geographic areas confront long-term unemployment, or when other labor market disequilibria exist, BCA may appropriately include the aggregate monetized value of net employment changes. Otherwise, when the economy is closer to, or at, full employment, employment impacts should not be included directly within BCA. This perspective assumes that spillovers related to unemployment (or employment) flows can vary with labor market dynamics and are potentially greater (in absolute value) the further from full employment. Depending on the size of the existing pool of unemployed workers, a regulation that results in either displacing workers or hiring workers will change the size of the pool. Such changes can impact the time it takes for other unemployed workers to fi nd employment. A larger pool likely increases the average time for unemployed workers to find employment, whereas a smaller pool may decrease their search time.

EPA’s Economics: BCA and Employment Impacts Since it began to use BCA in the 1980s, EPA has adopted the textbook interpretation of BCA. In explaining BCA to policymakers, we like to draw analogies to the private sector. Adam Smith (1776) showed that the private market can allocate and distribute scarce resources efficiently.

176

Ann E. Ferris and Al McGartland

His “invisible hand” magically and automatically allocates scarce resources to their most valued use. For example, a shoemaker uses society’s scarce resources (labor, capital, energy, and raw materials) to make shoes. If the shoemaker cannot sell the shoes for more than the costs of these inputs, he goes out of business. In such a case, the shoemaker was an inefficient user of society’s scarce resources, and the discipline of the marketplace reallocates these resources to more efficient uses. Society cannot, of course, sell public goods like clean air in the private market. However, BCA asks the question: if society could capture the willingness to pay of consumers for clean air, would the private market produce it? Could it make a profit doing so? The difficulty arises when considering how to apply the discipline of the private market to publicly provided or mandated commodities. Economists, after all, must develop values for the commodities on the basis of consumers’ preferences (not their own). BCA rests on the so- called potential Pareto efficiency test, subsequently formulated as the Kaldor-Hicks compensation criterion—namely, would those made better off by the policy be willing to fully compensate those made worse off (Boardman et al. 2006; Hicks 1940; Kaldor 1939; U.S. Environmental Protection Agency 2010)? This seemingly simple criterion has spurred a great deal of research on the value of environmental commodities as well as estimating the costs of environmental protection. Theory has helped economists separate real costs and benefits from transfers and to develop rules for avoiding double counting of benefits and costs under a default assumption of full employment. Despite the various rationales articulated in the previous section of this chapter for including in BCA employment impacts and, for that matter, impacts on capital and other productive resources, they are not currently included in BCA conducted by agencies such as the EPA. The dominant approach to date has been for regulatory analysis to draw on comparative statics, to focus on long-run equilibria, and to ignore the transition costs. BCA, as deployed by EPA, is a positive exercise—not a normative one. We use consumers’ willingness to pay, not bureaucrats’ notion of value (U.S. Environmental Protection Agency 2010). We also recognize that BCA says nothing about equity and fairness— although in conducting a BCA, those groups that bear the costs and enjoy the benefits can be informatively identified. However, even though we omit employment impacts from BCA, this does not mean EPA ignores the issue. On the contrary, EPA does include these impacts within a separate economic impact analysis whenever they are relevant to decision makers (U.S. Environmental Protection Agency 2010). Most economically significant regulations are informed by an employment impact analysis.

A Research Agenda for Improving the Treatment of Employment Impacts

177

At EPA, we attempt to be as comprehensive as practicable when estimating both the positive and negative economic impacts, including changes in employment, of new regulations. We first make a qualitative assessment of employment impacts when appropriate, such as for rules with economic effects of over $100 million annually. What we do next depends on whether appropriate methods and high- quality data are available. Given our commitment to relying on the best available science, we proceed with further analysis only when we can reliably estimate the net number of job changes associated with the policy. The employment analysis for a recent EPA regulation, the Mercury and Air Toxics Standards (U.S. Environmental Protection Agency 2011), provides an example of a case where we were able to quantify employment impacts in terms of the impact on net jobs. However, as there is no consensus yet in the literature for monetizing the value of net employment changes due to regulation, particularly in terms of valuing unemployed workers’ time and transition costs, the monetization of the value of affected jobs focuses solely on market wages for different categories of employed individuals as computed in our estimates of compliance costs for the affected sector. EPA’s (2010) Guidelines for Preparing Economic Analyses state that employment impacts are not relevant for BCA in most cases. In rare cases, when structural unemployment may affect social costs or benefits, employment impacts may be included within a BCA. However, the literature is still developing on this issue (U.S. Environmental Protection Agency 2011). Our view is that generally employment impacts are more appropriately described within a separate economic impact analysis, rather than incorporated into the BCA itself. This assertion is based, in part, on the traditional approach to employment within BCA: in a fullemployment economy, a regulation induces labor to shift between sectors, with costless transitions to new market equilibria, including with no net change in employment levels overall. This is generally the clearest assumption for a long-run analysis, but obviously it can gloss over shortrun changes. In addition, although there are a limited number of currently available methods to quantify the number of net employment changes, there is less available research to draw from on the appropriate methods for monetizing the value of those jobs, in terms of social welfare changes. The primary issue here, as we already value employed workers’ time at their market wage, is how best to value unemployed workers’ time. We have already mentioned that recent unacceptably high unemployment rates have motivated a reexamination of the full employment assumption. In addition, new research indicates that transitions, particularly when unemployment is high, are not costless (Davis and von

178

Ann E. Ferris and Al McGartland

Wachter 2011; Krueger and Mueller 2012; von Wachter et al. 2009). At the same time, economic activity is constantly shifting as prices change and the economy moves continuously from equilibria to equilibria, in an efficient manner. As Schumpeter (1942:83) stated, “Creative Destruction is the essential fact about capitalism.” Job flows in the economy are nonnegligible and continuous, even across business cycles (Davis et al. 1996). In order to address and to consider incorporating employment considerations into the traditional approach, we are in need of a unified economic theory addressing incremental employment changes associated with a particular policy and their aggregate, monetized value within the context of BCA. Our current approach primarily relies on employment impact analyses, separate from BCA, but within the regulatory impact analysis. These two components of economic analysis— impact analyses and BCA— are among many factors that influence policy design, which also include statutory instruction, institutional feasibility, technical feasibility, enforceability, ethics (e.g., distributive justice, environmental justice), and sustainability. Economic analysis informs policymakers but does not alone determine the regulatory process or outcomes. By presenting economic impacts both in a BCA and separate impact analyses, we can provide the best possible analysis without conditioning on the other factors that affect policy design. This separation also allows policymakers flexibility to pay close attention to economic impacts, particularly employment impacts, by having them emphasized in a separate document, rather than folded within the BCA as a line item. More important, separation allows policymakers to analyze each component of the regulatory analysis and to determine, themselves, how to assemble the pieces into their policy decisions. The alternative would require economists to impose how those pieces, specifically employment impacts, would be assembled into net benefits, thereby potentially diminishing a key decision-making function of policymakers. Finally, this separation allows politically appointed policymakers to base their decisions on the best available analysis, using the best available science rather than politically motivated studies of dubious quality. In a recent essay, V. Kerry Smith (2012) highlighted the issue of employment impacts and BCA as one for regulatory agencies to pursue further. Smith asserted that regulatory agencies should consider the state of the economy, specifically unemployment, in their benefit– cost analyses. Responding in part to the Office of Management and Budget’s March 2012 memo requesting comments on how federal agencies should assess the impacts of regulations on employment (Office of Management and Budget 2012), Smith’s (2012:314) recommendation is

A Research Agenda for Improving the Treatment of Employment Impacts

179

“to start with first principles and consider how large- scale policies should be evaluated within models that allow for unemployment as a structural feature of the economic system.” We agree that this should be considered, and any new analytical practices should be based on first principles of economics.

What Research Is Needed? Our assessment of the available literature leads us to conclude that the primary research need is conceptual (V. K. Smith 2012:314). To fully consider whether it is appropriate to move away from our current approach of keeping employment impacts analysis separate from BCA, we need more research on theoretical approaches to BCA that consider employment. The current body of literature is small and varies widely, particularly in regard to the scope used to approach employment impacts (Bartik 2012; Haveman and Farrow 2011; Livermore et al. 2012; Masur and Posner 2012; V. K. Smith 2012). The primary emphasis in the literature is on empirical studies, while there are few discussions of the underlying theory. There is no consensus yet on whether we should include employment changes within BCA and, if so, how best to implement that inclusion. The few papers that have discussed the theory underlying BCA in the context of employment impacts are not close to consensus. Some, like Arrow et al. (1996), suggest transition costs as the appropriate inclusion, while others, like Haveman and Farrow (2011), focus on adjusting opportunity costs during times of high unemployment. In order to move forward, it is crucial to have some consensus on the economic theory first, to provide a foundation for the empirical work that follows. Therefore, our priority is the development of a unified theory of employment impacts within BCA. However, we need both theory and empirical evidence to move forward. In line with V. Kerry Smith (2012), we propose a two- step research program. First, theory must be developed to determine whether we should be moving down this path to include employment impacts directly within BCA. What aspects of employment impacts could, or should, be counted as part of social benefits or costs? Second, we must consider how to apply this theory within BCA. For example, how would we estimate the opportunity cost of unemployed labor in order to provide a monetized value for the aggregate net employment changes? Would the valuation of employment impacts be included within BCA only under certain conditions of long-term unemployment? The theoretical foundation we are calling for would help inform the  research agenda for empirical assessment of how employment is

180

Ann E. Ferris and Al McGartland

affected by regulation. If the theory suggests that transitional costs are important, we may need information and data on reservation wages, transfer payments to the unemployed, and length of unemployment spells to properly include employment impacts in BCA. The theoretical literature offers few specifics on how to proceed. It would be ideal if this unified theory were available before embarking on further empirical studies.

Existing Empirical Research As we develop regulatory impact analyses at EPA, we rely on the best available science. However, so far the vast majority of the available empirical research on employment impacts focuses on estimating employment changes in terms of the number of jobs, while very few studies propose how we should place a monetary value on those jobs, in terms of social welfare mea sures. Again, although there is consensus in the literature that the value of a currently employed worker’s time is best given by her market wage, there is no straightforward way to mea sure similar values for unemployed workers. Of course, analysis focused on only one of the two pieces, due to lack of availability of the other, is still valuable and informative on its own. Recognizing these limitations in the available literature, EPA is currently supporting new economic research on theory and methods for employment analysis of environmental regulations. Earlier in this chapter we presented four theoretical approaches to the treatment of employment effects that we labeled “equilibria matter,” “transition costs matter,” “opportunity costs matter,” and “labor markets matter.” For each of these approaches, we now turn to a consideration of the available empirical literature. Equilibria First, in reference to the approach that equilibria matter, Hopkins (1992) spells out the economic argument that underlies the traditional approach: jobs are included only as costs, as a component of compliance costs, directly within BCA. Baumol and Oates (1988:255) focus on employment impacts as being primarily distributional and informative to regulatory analysis, but they draw a line between BCA and a separate impacts analysis. The traditional approach of analyzing employment impacts separately from BCA is further described in Gramlich (1990) and Boardman et al. (2006:299). In the economics literature, most employment impact studies are partial equilibrium studies of a particular industry or region. For ex-

A Research Agenda for Improving the Treatment of Employment Impacts

181

ample, Gray et al. (2013) analyzed how EPA’s Cluster Rule affected employment in the pulp and paper industry. The 2001 Cluster Rule was the first integrated, multimedia regulation imposed on a single industry, and stringency varied across plants. Using establishment-level data from the Census of Manufacturers and Annual Survey of Manufacturers at the U.S. Census Bureau from 1992 to 2007, Gray et al. found evidence of small employment declines (on the order of 3–7 percent), sometimes statistically significant, at a subset of the plants covered by the rule, measured by the difference between plants that faced more stringent regulations versus those that did not. Likewise, Greenstone (2002) takes a broader scope, estimating relative changes in manufacturing employment, also relying on a methodology that estimates policy- induced employment changes in air pollution nonattainment counties relative to a comparison group of attainment counties. It is important to emphasize that the employment estimates he reports are relative differences between the groups, and as such cannot be categorized as absolute losses or gains. However, environmental regulation may induce a number of economic effects within affected industries, with differing impacts on numbers of jobs. Some papers focus on the regulated industry and incorporate a net approach, which considers both the positive and negative employment changes, rather than looking only at gross impacts (Berman and Bui 2001; Morgenstern et al. 2002). This approach recognizes that it is crucial to assess all aspects of a regulation, including the negative and positive impacts. A regulation can be seen as a shift in spending—from morepolluting to less-polluting sources. From this perspective, at a minimum analysts should capture the net change in employment. Berman and Bui (2001) outlined a theoretic model to estimate the net employment impacts of environmental regulations within regulated industries. Their model allows for regulation to affect employment through two separate mechanisms: the output elasticity of labor demand and the effect of pollution abatement activities on demand for variable factors, combined with the marginal rates of technical substitution between abatement activity and variable factors. Neoclassical economic theory predicts that the output effect is, in most cases, negative, while the direction of the second, substitution effect is indeterminate, making the overall net effect ambiguous. Berman and Bui (2001:269) applied their model to an estimation of employment impacts of very stringent local air quality regulations in southern California and found that employment effects were “fairly precisely estimated zeros.” Morgenstern et al. (2002) outlined a similar theoretic framework, for net employment impacts. They empirically estimated their model for four

182

Ann E. Ferris and Al McGartland

heavily polluting industries and found insignificant impacts of environmental regulation on employment within regulated industries. Transition Costs From the perspective of transition costs, analysts could potentially itemize and estimate different short-run transition costs, including “relocation or retraining costs, long-term productivity effects, and any negative effects on psychological or physical health resulting from long-term unemployment” (Livermore et al. 2012:9; see also Jaffe et al. 1995:table 3). Arrow et al. (1996:7– 8) opined that if a regulation results in economic spillovers that lead to “significant job losses or increased costs to a specific industry in a local economy,” those spillovers can be considered within BCA but should be weighed against any other positive spillovers. They asserted that regulation does not usually affect the general employment level and may have only a “very minor effect on either wages or employment in the industry to which it applies” (Arrow et al. 1996:8). Even conditional on that statement, they suggested that an estimate of the “transition costs of employees who are forced to switch jobs because of the regulation is the appropriate measure for use in benefit cost analysis” (Arrow et al. 1996:8). However, to date, there is almost no published literature estimating these transition costs resulting from environmental regulations. Opportunity Costs When it comes to adjusting the opportunity costs of unemployment for use within BCA, Haveman and Farrow (2011) proposed multiple approaches to valuing newly hired labor for public projects. Other researchers have also proposed approaches to adjusting social benefits or costs by using an adjusted value for the opportunity cost of unemployed workers, instead of the market wage. Boardman et al. (2006:99–101) described several alternative measures for estimating the opportunity cost of unemployed labor, while Masur and Posner (2012) proposed a single value for all unemployed workers for use in regulatory analysis. Livermore et al. (2012) discussed some of the factors that may contribute to variation in opportunity costs for unemployed workers, including high unemployment rates and long unemployment durations. Bartik (2012) proposed two approaches to estimating the social benefits of hiring the unemployed on the basis of adjustments to reservation wages or adjusted earnings. Yet even within the labor economics literature, aside from any consideration of BCA, there is broad recognition of the difficulties in estimating reservation wages (Shimer and

A Research Agenda for Improving the Treatment of Employment Impacts

183

Werning 2007). V. Kerry Smith (2012:314) supports Bartik’s approach in using the “reservation wage as the appropriate welfare mea sure for considering the effects of policy in the context of involuntary unemployment,” but he makes the point that any such adjustments to social benefits or costs for unemployment “must follow from a consistent macro model of the process.” In addition, Smith (2012:314) suggested that any such adjustments should also consider other possible effects of regulation that may impact welfare, apart from employment, including any activities affecting “the labor/leisure tradeoff, the consumption choices of market goods, and the size of the tradeoffs people can make (given their incomes) for non-market goods.” This is a tall order, particularly given the difficulties discussed above in these first approaches to valuing employment within BCA. Labor Markets Although the literature suggests that labor market equilibria matter, it has yet to reach any consensus on specifics (Bartik 2012; Haveman and Farrow 2011; Livermore et al. 2012; Masur and Posner 2012; V. K. Smith 2012). Haveman and Farrow (2011) pointed out that the traditional approach to BCA works well in times of full employment, but they proposed that in times of high unemployment, social benefits may outweigh the social costs of public investments. In a setting with imperfect labor markets, Livermore et al. (2012) pointed out that, in addition to spillover effects from labor market dynamics, short-term transition costs, including relocation and retraining costs, may be exacerbated.

Areas of Improvement Even with our current state of insufficient theory and empirical evidence to fit employment impacts within BCA, there are still several more tractable issues in estimating employment impacts that can be targeted for improvement. Temporal Issues with Regulations The time period for compliance with a given regulation is typically lengthier in practice than assumed in economic analysis. Most studies use the date that the regulation was promulgated as the date by which the regulation is fully in effect. However, EPA regulations often build in compliance or phase-in periods that can average a few years. Under the Clean Air Act, for example, compliance periods can range from immediate (some New Source Performance Standards) to lengthy (National

184

Ann E. Ferris and Al McGartland

Table 9.1. Compliance Timelines for Economically Significant EPA Regulations Issued in 2012

Regulation Joint Rule Making to Establish 2017 and Later Model Year Light Duty Vehicle (LDV) GHG Emissions and CAFE Standards Review of the National Ambient Air Quality Standards for Particulate Matter Criteria and Standards for Cooling Water Intake Structures (Notice of Data Availability) Reconsideration of Final National Emission Standards for Hazardous Air Pollutants for Reciprocating Internal Combustion Engines (RICE) Petroleum Refineries—New Source Per for mance Standards (NSPS)— Subparts J and Ja Regulation of Fuels and Fuel Additives: 2013 Biomass- Based Diesel Renewable Fuel Volume Oil and Natural Gas Sector— New Source Per for mance Standards and National Emission Standards for Hazardous Air Pollutants

Publication Date

Compliance Date(s)

Years Until Compliance

10/15/2012

2017 and later model years

5 years

12/14/2012

2020 (Extensions to 2025 are possible) 2018

8–13 years

6/7/2012

Immediately, May 2013, October 2013

0–1 year

9/2/2012

2017

3– 5 years

9/27/2012

2013

0–1 year

8/16/2012

2015

6/11/2012

5 years

3 years

Sources: We selected all economically significant EPA regulations in 2012 as identified and reviewed by the Office of Information and Regulatory Affairs (2012). For each rule’s publication and compliance dates, we relied on the following web pages accessed on 23 August 2013: LDV rule (http://epa.gov/otaq/climate/regs -light-duty.htm); particulate matter NAAQS (http://www.epa .gov/pm/actions.html); Cooling Water Intake rule (http://yosemite.epa.gov/opei/rulegate.nsf /byrin/2040 -ae95?opendocument #4); RICE rule (http://www.epa.gov/ttn/atw/rice/ricepg.html and http://www.epa.gov/ttn/atw/rice/20120717riceqaupdate.pdf); Petroleum Refineries NSPS (http://www.epa.gov/ttnatw01/nsps/petrefnsps/petrefnspspg.html); Biomass- Based Diesel (http://www.epa.gov/otaq/fuels/renewablefuels/regulations.htm); Oil and Gas Sector rule (http://www.epa.gov/airquality/oilandgas/actions.html).

Ambient Air Quality Standards [NAAQS] implementation). To illustrate, Table 9.1 provides summary data on the compliance dates for economically significant regulations issued by EPA during 2012. For this sample of recent EPA regulations, only two rules take effect within two years. The remainder of the regulations average about three to five years until compliance is required, with the particulate matter

A Research Agenda for Improving the Treatment of Employment Impacts

185

NAAQS as a relative outlier with at least eight years because of the lengthy process of designating attainment and nonattainment areas and developing State Implementation Plans. As discussed in EPA (2012), employment impacts of an NAAQS are incredibly difficult to quantify, as states choose their own sets of policies and regulations to meet the ambient air quality targets, and state regulations are often in effect for a period of time before they become federally enforceable. Furthermore, it is frequently the case that compliance dates are further extended because of litigation and subsequent regulatory action. Transition Costs EPA’s current approach implicitly assumes that transition costs are de minimis— approximately zero. Yet if those transition costs are substantial, it would be informative to include that information as part of a regulatory analysis. It is even possible that transition costs can be, in some cases, positive, such as when environmental regulations induce creation of new pollution control technologies (Livermore et al. 2012). Recent research suggests that transition costs may not be negligible and may be larger when the unemployment rate is higher (Davis and von Wachter 2011; von Wachter et al. 2009). More research is needed on employment transition costs from environmental regulations, in particular, on the conditions under which these costs, in whole or in part, can reasonably be included as social costs or benefits. Some of the conditions mentioned as potentially important include the existence of macro stabilization policies that address employment, concentration of employment impacts within a sector or local area, labor market conditions related to the business cycle and changes in aggregate demand, and the availability of adjustment assistance programs (Baumol and Oates 1988; Goodstein 1994). General Equilibrium Versus Partial Equilibrium In practice, agencies generally assess the costs and benefits only in the regulated market and ignore downstream and upstream effects. These spillovers are generally considered to be relatively small for most regulations. Yet in the past, some employment assessments have used input– output models or macroeconomic models to assess changes in numbers of jobs across the economy. We should seek to learn more about whether input– output models are appropriate (macro) models to assess employment. Indeed, one might ask whether there are any models that can accurately trace expenditures and employment across the national economy in the short run (disequilibrium). Regulations cause a shift in

186

Ann E. Ferris and Al McGartland

expenditures toward pollution abatement equipment and operation. Can models predict what expenditures are reduced to fund pollution abatement? Given this “shift” in expenditures, is it plausible to argue to a first- order approximation that increases in employment are approximately equal to the decreases in employment? Finally, improved environmental quality may indirectly affect labor demand and supply. The health benefits from environmental policies may result in a healthier, more productive labor force, resulting in fewer lost sick days and increased labor force participation. High Unemployment Some authors have proposed that analysis of employment impacts and BCA proceed differently during times of high unemployment because the opportunity costs of labor may change with economic conditions (Haveman and Farrow 2011; V. K. Smith 2012). We suggest that more work in this area is necessary to determine the economic conditions under which approaches should differ and to develop a professional consensus regarding this approach. Monetizing Employment Impacts Because the available literature focuses almost entirely on estimating changes in employment, rather than on how to monetize the value of those employment changes within BCA, there is a need for more research on valuation. The few studies that have engaged this topic so far discuss various aspects of social welfare that could be considered and how assumptions for each may produce vastly different estimates for the social value of unemployment. Haveman and Farrow (2011) outlined three approaches to valuing labor during times of high unemployment: an opportunity cost approach, a worker surplus approach, and a gross flows approach. Masur and Posner (2012) proposed, as a simplifying heuristic, a single estimate of the value of the social cost of unemployment of $100,000 per worker, estimated in terms of foregone earnings. However, they also acknowledged that the social value should be broader than foregone earnings, and at the same time they described many factors that induce heterogeneity into such an estimate (for example, unionization, compensating differentials, subjective well- being, how layoffs are defined, and industry conditions). Bartik (2012) proposed two approaches to valuing the social benefits of job creation during times of high unemployment, relying on reservation wages or adjusted earnings. These papers illustrate the current difficulties in describing economic conditions under which the social value of employment may

A Research Agenda for Improving the Treatment of Employment Impacts

187

differ from the market wage, determining which components of the social value are relevant and estimable with some degree of precision by researchers, and estimating reservation wages. We recommend more research in this area, as monetization of the value of net changes in employment is one of two key components needed to incorporate employment impacts into BCA.

Conclusion We have outlined a two- step research agenda for improving the treatment of employment impacts in regulatory impact analysis. Both scholars and economic analysts need a unified theory about how to incorporate employment impacts into BCA. The first step of the research agenda calls for researchers to rely on first principles to develop this muchneeded theory. The second step calls for more empirical research, building on this theoretical foundation. We have identified areas for improvement within the empirical literature on employment impacts, including better attention to temporal aspects of regulatory implementation, transition costs for workers, general equilibrium analysis, the treatment of employment impacts during periods of high unemployment, and the need to identify ways of valuing the employment impacts of regulation.

Acknowledgments Any opinions and conclusions expressed herein are those of the authors and do not necessarily represent the views of the U.S. Environmental Protection Agency. The authors want to thank Cary Coglianese, Adam Finkel, participants at the Penn Program on Regulation’s conference, “Regulation’s Impact on Jobs,” and colleagues at EPA’s National Center for Environmental Economics for insights and helpful comments.

References Arrow, Kenneth J., et al. (1996) Benefit– Cost Analysis in Environmental, Health, and Safety Regulation— A Statement of Principles. Washington, DC: American Enterprise Institute, the Annapolis Center, and Resources for the Future, http://regulation2point0.org/wp- content/uploads/downloads/2010/08/Be nefit- Cost- Analysis- in- Environmental- Health- and- Safety- Regulation.pdf (accessed 7 September 2012). Bartik, Timothy J. (2012) “Including Jobs in Benefit– Cost Analysis,” Annual Review of Resource Economics 4: 55–73. Bastiat, Frederic (1850) Selected Essays on Political Economy (1995). Irvington- onHudson, NY: Foundation for Economic Education, http://www.econlib.org /library/Bastiat/basEss1.html (accessed 13 February 2013).

188

Ann E. Ferris and Al McGartland

Baumol, William. J., and Wallace E. Oates (1988) The Theory of Environmental Policy. 2nd ed. New York: Cambridge University Press. Berman, Eli, and Linda T. M. Bui (2001)“Environmental Regulation and Labor Demand: Evidence from the South Coast Air Basin,” Journal of Public Economics 79: 265– 95. Boardman, Anthony, et al. (2006) Cost– Benefit Analysis: Concepts and Practice. 3rd ed. Upper Saddle River, NJ: Prentice Hall. Davis, Steven J., and Till von Wachter (2011) “Recessions and the Cost of Job Loss,” Brookings Papers on Economic Activity (Fall): 1–72, http://muse.jhu.edu /journals/brookings_papers_on_economic_activity/v2011/2011.2.davis. html (accessed 13 February 2013). Davis, Steven J., et al. (1996) Job Creation and Destruction. Cambridge, MA: MIT Press. Goodstein, Eban (1994) Jobs and the Environment: The Myth of a National Trade- Off. Washington, DC: Economic Policy Institute. Gramlich, Edward M. (1990) A Guide to Benefit– Cost Analysis. Prospect Heights, IL: Waveland Press. Gray, Wayne B., et al. (2013) “Do EPA Regulations Affect Labor Demand? Evidence from the Pulp and Paper Industry.” U.S. Census Bureau, Center for Economic Studies Working Paper 13–39. Greenstone, Michael (2002) “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures,” Journal of Political Economy 110: 1175–1219. Haveman, Robert H., and Scott Farrow (2011) “Labor Expenditures and Benefit– Cost Accounting in Times of Unemployment,” Journal of Benefit– Cost Analysis 2: 7, http://www.degruyter.com/view/j/jbca.2011.2.2/jbca.2011.2.2 .1081/jbca.2011.2.2.1081.xml (accessed 3 March 2013). Hicks, John R. (1940) “The Valuation of the Social Income,” Economica 7: 105–24. Hopkins, Thomas D. (1992) “Regulation and Jobs: Sorting out the Consequences.” Center for the Study of American Business, Washington University, Occasional Paper 117, St. Louis, MO. Jaffe, Adam B., et al. (1995) “Environmental Regulation and the Competitiveness of U.S. Manufacturing: What Does the Evidence Tell Us?” Journal of Economic Literature 33: 132– 63. Kaldor, Nicholas (1939) “Welfare Propositions of Economics and Interpersonal Comparisons of Utility,” Economic Journal 49: 549– 52. Krueger, Alan B., and Andreas I. Mueller (2012) “Time Use, Emotional WellBeing, and Unemployment: Evidence from Longitudinal Data,” American Economic Review 102: 594– 99. Livermore, Michael A., et al. (2012) “The Regulatory Red Herring: The Role of Job Impact Analyses in Environmental Policy Debates,” Institute for Policy Integrity, New York University School of Law, http://policyintegrity. org/files/publications/Regulatory_Red_Herring.pdf (accessed 3 March 2013) Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry- Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. Obama, Barack (2011) “Executive Order 13563: Improving Regulation and Regulatory Review: Section 1: General Principles of Regulation,” Federal Register 76: 3821–23.

A Research Agenda for Improving the Treatment of Employment Impacts

189

Office of Information and Regulatory Affairs (2012) “Info from OIRA: Office of Information and Regulatory Affairs (OIRA) Economically Significant Executive Order Reviews Between January 01, 2012 to December 31, 2012: Environmental Protection Agency,” http://www.reginfo.gov/public/do/eoHist Rev iewSearch (accessed 23 August 2013). Office of Management and Budget (2012) Draft 2012 Report to Congress on the Benefits and Costs of Federal Regulations and Unfunded Mandates on State, Local, and Tribal Entities. Washington, DC: U.S. Office of Management and Budget, http://www.whitehouse.gov/sites/default/fi les/omb/oira/draft _2012 _cost _ benefit _report.pdf. Schumpeter, Joseph (1942) Capitalism, Socialism, and Democracy. New York: Harper and Brothers. Shimer, Robert, and Ivan Werning (2007) “Reservation Wages and Unemployment Insurance,” Quarterly Journal of Economics 122: 1145– 85. Smith, Adam (1776) An Inquiry into the Nature and Causes of the Wealth of Nations (1977). Chicago: University of Chicago Press. Smith, V. Kerry (2012) “Reflections: In Search of Crosswalks Between Macroeconomics and Environmental Economics,” Review of Environmental Economics and Policy 6: 298–317. U.S. Environmental Protection Agency (2010) Guidelines for Preparing Economic Analyses, Washington, DC: U.S. Environmental Protection Agency, http:// yosemite.epa.gov/ee/epa/eed.nsf/pages/guidelines.html (accessed 14 February 2013). ———. (2011) Regulatory Impact Analysis for the Final Mercury and Air Toxics Standards. Washington, DC: U.S. Environmental Protection Agency, http://www .epa.gov/ttn/ecas/regdata/RIAs/matsriafinal.pdf. ———. (2012) Regulatory Impact Analysis for the Proposed Revisions to the National Ambient Air Quality Standards for Particulate Matter. Washington, DC: U.S. Environmental Protection Agency, http://www.epa.gov/ttn/ecas/regdata /RIAs/PMRIACombinedFile_Bookmarked.pdf. von Wachter, Till, et al. (2009) “Long-Term Earnings Losses due to Mass Layoffs During the 1982 Recession: An Analysis Using U.S. Administrative Data from 1974 to 2004.” Columbia University Department of Economics Discussion Paper Series DP0910- 07, New York, www.columbia.edu/~vw2112/papers /mass_layoffs_1982.pdf (accessed 13 February 2013).

Chapter 10

Employment and Human Welfare Why Does Benefit–Cost Analysis Seem Blind to Job Impacts? Brian F. Mannix

When economists perform a benefit– cost analysis (BCA) of a public project or policy, they think of it as summarizing the real effects of the decision on public welfare: the well-being of real people, mea sured according to people’s own preferences. The methods they use are specifically designed to look past the familiar aggregate statistics of economic activity— the national income, gross domestic product (GDP), and so forth— to isolate net changes in economic surplus, which is a dollardenominated aggregate estimate of how much better or worse off individual consumers can be expected to feel. After all, economic activity itself has no intrinsic value; it exists only to satisfy human wants. And that is how public decisions should be evaluated, too: not by how much activity they generate, but by how well they satisfy human desires, including those— like caring about whether there are whales in the ocean—that may not correspond to much observable economic activity at all. To a layperson, however, this can all look too obscure and reductionist. Once the economist has added up all the “compensating variations” (mea sures of willingness to pay), the net benefits of a project take the form of disembodied dollars— they appear in no budget, cannot be spent, and are “owned” by no one in particular. As a result, economists often find themselves lonely advocates for choosing policies that maximize net benefits. Adding together what each individual cares about produces, it seems, a result that no one cares about very much. In contrast, people who know nothing else about the economy know, and care about, the level of unemployment. It is the most widely re-

Employment and Human Welfare

191

ported summary statistic on business cycles and the per for mance of the economy, and it gets enormous political attention—particularly during election seasons when unemployment is high and economic growth has seemingly stalled. In times of economic stress, national elections may hinge, as much as anything else, on the voters’ perceptions about which candidates will be most effective in reducing unemployment. So it is not surprising that the first question senior policymakers ask economists about a public decision is very often: What will be the effect on jobs? And the answer from economists is often “zero,” or “negligible,” or “ambiguous,” or “that is the wrong question.” This chapter will explore some of the reasons why the human welfare metric, as it is typically calculated in a BCA, appears to be insensitive to the employment effects that loom so large in the perceptions of the public and its elected representatives. It argues that, to a first approximation, employment effects are already counted in a BCA as a component of compliance costs— so that BCA is in fact not insensitive to job impacts. It is just not hypersensitive to them. Of course, no BCA is ever complete, so it is always possible that some job-related welfare effects are omitted, just as it is likely that some other welfare effects unrelated to employment are also omitted. But any attempt to include additional categories of welfare effects must confront the problem of potentially counting these effects more than once. This chapter concludes that, in most cases, employment effects should be treated as they traditionally have been treated— implicitly part of the calculation of compliance costs— and that some of the proposed alternatives to the status quo would result in double counting. It would be helpful, however, if economists could do a better job of educating the public about what, exactly, compliance costs represent. If people understood that these are not simply a “cost of doing business” but real welfare changes experienced by the public, then benefit– cost analysis would be a far more informative tool than it is today. In addition, there may be employment-related distributional considerations that are important and deserve analytical attention, but these are outside the scope, or at least beyond the reach, of the BCA methodological framework.

Origins of Benefit– Cost Analysis The use of BCA in governmental decision making traces its origins to the Rivers and Harbors Act of 1902, wherein Congress first began asking for economic analysis of civil engineering projects (Lipton et al. 1995). Although the federal interest in maintaining and improving navigable waters was not in question, Congress was confronted with intractable and competing demands for federal funding of local navigational

192

Brian F. Mannix

improvements: river and harbor dredging and the construction of dams, levees, locks, and other infrastructure. Local advocates touted the benefits of these projects, but costs were often understated and benefits overstated or counted more than once. Advocates for a project would typically use the volume of trade through a port as the mea sure of its benefit, notwithstanding that there are many other costs associated with those goods and that other ports may be competing for the very same trade as well as for the same federal dollars. Throughout the twentieth century, the theory and methodology of BCA developed in parallel with welfare economics. BCA gradually acquired both a rigorous theoretical foundation (embodied today in the Kaldor-Hicks criterion that welfare gains to the winners should exceed welfare losses to the losers) and a rich portfolio of mathematical optimization methods, many of which developed as part of military operations research during World War II. From the very beginning, however, BCA was intended to serve as an antidote to the well-known tendency of the government to be overresponsive to concentrated, organized, and loud interests, at the expense of the general welfare. BCA is designed to provide an objective, unbiased, empirical assessment of all of the welfare effects, positive and negative, on whomever they fall, of a public project or policy choice. It strives to be complete—including, with appropriate weights, all of a decision’s consequences: remote as well as proximate, indirect as well as direct, diluted as well as concentrated, delayed as well as immediate, improbable as well as probable, unintentional as well as intentional. The flip side of this completeness principle is that a BCA needs to take great pains to avoid double counting any benefits or costs. In an economy where everything is connected to everything else, it is all too easy to mistakenly count something twice. The analyst has to be particularly sensitive to the various ways that the benefits or costs ascribed to a regulation may also be incorporated into prices: a cleaner waterfront will raise property values, a more efficient car will command a premium, a safer job will not pay quite as much as a more hazardous one. These price effects are important indicators of value, but they also complicate the task of counting everything exactly once. Given this challenging task, the only way that BCA is even remotely feasible is by looking at the economy through a specialized filter, one that aggressively removes offsetting effects while allowing a clear view of those things that are not expected to be in balance. In this respect, the accounting of welfare changes within a BCA is very different from the accounting that appears in more familiar macroeconomic metrics, such as those in the National Income and Product Accounts (NIPA), which mea sure economic activity. The double- entry

Employment and Human Welfare

193

accounting in the NIPA ensures that there is a debit for every credit, so that the accounts are always in balance. In contrast, a BCA ignores the debits and credits that mea sure economic exchange and instead seeks to identify the economic surplus. If I rent, for $1, a movie that was worth $3 to me, the NIPA measures $1 of expenditure and $1 of production; it ignores the $2 “consumer surplus” altogether. In contrast, a BCA looking at the same transaction would count only the $2 surplus and would ignore the $1 of production and consumption. For this reason, there may be very little overlap between the approach taken in NIPA, which seeks to capture exchange, and the approach taken in a BCA, which proceeds by excluding symmetrical exchanges in order to capture residual differences. Should we allow motorboats on a quiet mountain lake? The answer will be determined by asking not how we can maximize GDP but how we can maximize the net welfare of those who despise motorboats as well as those who enjoy them. In this calculation, the motorboats themselves, and all of the capital and labor they embody and all of the complexities of the supply chain that produced them, will be largely irrelevant—they are presumed to be worth, at the margin, what they cost to produce.

What Is Seen and What Is Not Seen Another aspect of looking at the economy through the BCA lens is that it prevents objects in the foreground from dominating the picture and obscuring those in the background that may be equally important. As Frédéric Bastiat explained in his 1850 essay, What Is Seen and What Is Not Seen: “There is only one difference between a bad economist and a good one: the bad economist confines himself to the visible effect; the good economist takes into account both the effect that can be seen and those effects that must be foreseen” (Bastiat 1850:1.2). That admonishment is a pretty good statement of the central problem of BCA and is all the more remarkable because it was written a good half- century before BCA began to take its modern form. In his essay, Bastiat skewers various economic sophisms, beginning with the “broken window fallacy.” He notes, “Everybody has to make a living. What would become of the glaziers if no one ever broke a window?” (Bastiat 1850:1.6). After examining the flows of money and the associated opportunity costs, Bastiat concluded that the loss of a window represents a real cost, one that has no corresponding “ job creation” benefit—not even a small one: “And if we were to take into consideration what is not seen, because it is a negative factor, as well as what is seen, because it is a positive factor, we should understand that there is no benefit to industry in general or to national employment as a whole, whether windows are broken or

194

Brian F. Mannix

not broken” (Bastiat 1850:1.13). He then extends his analysis to the purported job- creation benefits of public works: “The sophism that I am attacking in this essay is all the more dangerous when applied to public works, since it serves to justify the most foolishly prodigal enterprises. When a railroad or a bridge has real utility, it suffices to rely on this fact in arguing in its favor. But if one cannot do this, what does one do? One has recourse to this mumbo jumbo: ‘We must create jobs for the workers’ ” (Bastiat 1850:1.87). He continues by observing that There is an article in the Constitution [of the Second French Republic] which states: “Society assists and encourages the development of labor . . . through the establishment by the state, the departments, and the municipalities, of appropriate public works to employ idle hands.” As a temporary mea sure in a time of crisis, during a severe winter, this intervention on the part of the taxpayer could have good effects. It acts in the same way as insurance. It adds nothing to the number of jobs nor to total wages, but it takes labor and wages from ordinary times and doles them out, at a loss it is true, in difficult times. As a permanent, general, systematic measure, it is nothing but a ruinous hoax, an impossibility, a contradiction, which makes a great show of the little work that it has stimulated, which is what is seen, and conceals the much larger amount of work that it has precluded, which is what is not seen. (Bastiat 1850:1.91–1.94)

Note that Bastiat acknowledges (in 1850 no less) that there may be countercyclical benefits associated with public works in “time of crisis” but cautions that this is only a temporary effect and not a net positive one. Bastiat also turns to the subject of regulation, applying his method to rebut the economic arguments that were being used to promote two regulatory programs that he regarded as especially pernicious. The first was protectionism, a topic he had tackled five years earlier in his famous spoof, the “Candlemakers’ Petition Against the Sun” (Bastiat 1845). The second was what, across the Channel, was called Luddism— a movement that opposed technological progress because of the fear that machines would displace workers: “The theoretical objection that is raised against this inclination is the same in both cases. In one as in the other, the reproach is made that it apparently makes for a scarcity of jobs. However, its actual effect is not to make jobs scarce, but to free men’s labor for other jobs” (Bastiat 1850:1.162). In reaching his conclusions, Bastiat sorts through the various benefits and costs that flow from an action by using a method that is much the same as we use today: he assumes that in a voluntary exchange, the benefits and costs are in equilibrium for both the buyer and the seller, reflecting both the individuals’ preferences and their budget constraints: “[Private ser vices] are in the domain of the voluntary, i.e., of individual responsibility. Each gives and receives what he wishes, or what he can,

Employment and Human Welfare

195

after bargaining. These ser vices are always presumed to have a real utility, exactly mea sured by their comparative value” (Bastiat 1850:1.96). This presumption— that the value, at the margin, of voluntary exchanges is reflected in the prices and wages at which they take place—is what a modern economist would call the equilibrium assumption, and Bastiat’s results depend on it. The question is: More than 160 years later, why do economists practicing BCA today still cling to that same assumption? Should they?

The Equilibrium Assumption in Benefit– Cost Analysis In an economy where everyone is fully informed, markets are complete and atomistic (no monopolies), property rights are well defined and secure, government stays within its proper boundaries, and no other market imperfections disturb the good order, there would be little for BCA to do. Public works and regulations either would not be needed or would have to be justified on some grounds (such as distributional grounds) other than the promise of positive net benefits. With the market in perfect equilibrium, perturbations could produce only Pareto disimprovements, with negative net benefits. Similarly, in an economy that is pervasively dysfunctional, in which such voluntary transactions as do take place take the form only of barter, so that markets lack a reliable medium of exchange, BCA would be of little use. Here the BCA analyst would be at sea, without reference points with which to gauge the value of things. Benefit–cost analysis of government projects is both useful and feasible in a functioning market economy that is almost, but not quite, ideal. There must be some market failure to be put under the microscope, to see what distortions it might cause in prices or incentives and what their ramifications might be. A public works project or a regulation, if well designed, may produce net benefits by correcting such distortions and compensating for the market failure. But those distortions and corrections can be observed only against a background in which prices are generally accurate and people’s unconstrained choices can be presumed to reflect their honest preferences. It would be meaningless to apply BCA to an economy in which prices were broadly inaccurate, just as it would be meaningless to search for a flaw in an amorphous crystal. This is the reason that BCA uses the equilibrium assumption. It is not because economists are stubbornly ignoring the latest economic news. Rather, the equilibrium assumption allows the BCA analyst to recognize that markets will adjust to a government decision in countless ways (and will adjust differently in response to Option A versus Option B). It defines exactly how much they will adjust (to the point where

196

Brian F. Mannix

gains from trade are exhausted). It allows the analyst to ignore those effects that can be expected to cancel each other out (as in Bastiat’s voluntary exchanges) while isolating the relatively small subset of effects that are not expected to cancel each other out. For example, if a regulation requires that a widget factory install pollution control equipment to reduce emissions, the expectation is that this will produce cleaner air— something that consumers value. At the same time, it will cause the economy to produce less of other things that consumers value. Which things? We do not know. But we can estimate what those things are worth, because we know the price of the pollution control equipment, and we know that it therefore is diverting that amount of resources away from the production of other good things—not necessarily widgets. Or, close to equivalently, we can say that, one way or another, consumers will have to pay the cost of the pollution control equipment and will find that they have less money to spend on other things. The equilibrium assumption allows the analyst to say that the price paid for the pollution control equipment is a reasonable mea sure of the welfare loss to consumers, as well as the losses that fall on investors and employees—notwithstanding the fact that, between the expenditure by the widget factory and the change in human welfare, there are countless intervening elasticities that interact in ways far too complex to model. Prices will rise or fall; fi rms may close or open; workers may be laid off or be employed; land, capital, or patents may increase or decrease in value as a result. In the end, the equilibrium assumption allows us to use the compliance costs of businesses, which we are able to mea sure, as a proxy for the myriad welfare effects that inevitably occur downstream, and which are too complex to mea sure. It is important to keep in mind that no corporation has ever experienced a welfare loss. The only reason to count business compliance costs in a BCA is to represent the welfare of those consumers, investors, and employees who do experience welfare losses. If the equilibrium assumption is relaxed to the point where prices (for materials, labor, capital, and so forth) become unreliable, then the connection between the regulatory mandate and the ultimate effect on consumer welfare becomes impenetrably opaque. BCA is not alone in relying on the equilibrium assumption to this degree. Consider the basic tools of the microeconomist, supply and demand curves. We plot on the x- axis just one of the many goods (or, sometimes, bads) in the consumer’s market basket; the others are relegated to the y-axis, where they are summarized by a dollar value. We learn in Economics 101 that this dollar value represents “all other goods and ser vices,” which can be placed on a single axis only because we assume

Employment and Human Welfare

197

that those goods and ser vices are worth, at the margin, what people pay for them. Yes, it is a heroic assumption, but it is also a very powerful and useful one. It makes microeconomics possible; that is, it allows us to study the properties of a single market that is embedded in a larger market economy, without having to model the whole thing. Absent some version of the equilibrium assumption, most of microeconomics would be an intractable mess.

The Equilibrium Assumption and Employment Although an explicit treatment of employment effects is rarely found in a typical BCA, it is not accurate to say that BCA ignores employment effects; rather, it treats employment as a voluntary transaction, which takes place at a price mutually agreeable to the employer and the employee. As such, a job is neither a cost nor a benefit—both parties to the transaction experience both costs and benefits that, at the margin, are expected to be in balance. Hence small changes in the level of employment would be expected to generate zero net benefits. At the same time, the level of employment quickly finds its equilibrium level, as any jobs lost or gained in consequence of a policy decision will likely be offset elsewhere in the economy. So a typical BCA calculates the net benefits of a policy choice under the assumption that the welfare effects of employment are captured in the wages paid and received. Any additional job-related contribution to the net benefits is not just zero, but zero times zero. The first zero is the effect on quantity: the net change in the level of employment is expected to be small. Even if it is not, however, the second zero is the value assigned to that change: the difference between the market price of labor and its economic value at the margin. A strong case would need to be made to move both of these numbers significantly away from zero before the monetized value of employment effects would begin to matter enough to change the outcome of a BCA. Strictly speaking, neither of these numbers will be exactly zero in any particular case. For example, if we assume that jobs created in one place (manufacturing energy- efficient light bulbs, for example) will be offset by job losses elsewhere in the economy, we have to explain how this occurs. Similarly, if the effect of direct job losses is, in Bastiat’s words, “to free men’s labor for other jobs,” then that signal has to be transmitted somehow to the larger labor market. Price—the level of wage demands—is the mechanism that transmits these signals. A government decision (or, for that matter, a private decision) that creates jobs will reduce the quantity of unemployment and will thereby drive up wages demanded, destroying jobs someplace else. And a decision that destroys jobs

198

Brian F. Mannix

directly will drive down wages and thereby encourage job creation elsewhere. Such adjustments necessarily involve some friction. The assumption made is that, in a market economy governed by functioning wages and prices, these effects are relatively small. When a regulation affects the employment transaction directly, the ceteris paribus equilibrium assumption would not apply: different options may produce very different effects. A minimum wage, for example, will change the level of employment as well as the level of wages; empirical studies can help determine what the effects are, and a BCA could not realistically assume that regulated wages were in equilibrium. More broadly, it may be the case that taxes on employment (especially payroll taxes) or some regulatory requirements (such as pension and health care provisions) alter wages in a way that produces systemic distortions significant enough to warrant an adjustment in routine BCAs. Something similar is done when a BCA uses the “excess burden of taxation” (U.S. Office of Management and Budget 1992:sec. 11) for government expenditures, or when it uses the “shadow price of capital” (Lind 1982) for private investment. Outside of an analytic framework that systematically accounts for wage distortions, however, a BCA would have no objective basis for rejecting the assumption that wages are an accurate reflection of the payer’s and payee’s preferences. As far as the “first zero” is concerned, the degree to which regulations may cause more than a transient change in the level of employment is an empirical question, beyond the scope of this chapter; suffice it say that the evidence is mixed. At a coarse cross- sectional level, no clear pattern emerges, at least not anecdotally. In a hyperregulated totalitarian society such as North Korea, public welfare is orders of magnitude below the level in South Korea. But although reliable data are hard to come by in North Korea, the level of unemployment presumably is not as starkly divergent as overall welfare is. In completely unregulated failed states such as Somalia, public welfare is also at horrifyingly low levels. But people are probably about as busy there as they are in Switzerland. Formal levels of employment may seem very low, but that is just a statistical problem; the informal sector is all that there is. Regulation (and such fundamental legal institutions as property rights and the rule of law) may be responsible for dramatic swings in productivity; but employment seems to find its own level, more or less independent of the level of regulation. Excessive regulation may wreak such havoc on the economy that we find ourselves back in the Stone Age; but in the Stone Age, everyone had a job. A BCA analyst would face similar challenges if tasked with determining how the value of a job might differ from the value that the employee and employer place on it. Even during a recession, the natural assump-

Employment and Human Welfare

199

tion is that voluntary transactions reveal true preferences. It may be the case that displaced employees suffer in various ways from the reduced income and perhaps suffer long-term welfare losses. This, of course, is why they were working in the first place—people generally are aware of the connection between income and economic welfare. It is implausible to think that a BCA analyst will have a better idea of what the individual is feeling than the individual does. At the same time, the wages that workers do not receive when they are laid off represent a benefit to others in the economy—perhaps to consumers who have more income to buy other goods and ser vices they desire, or perhaps to retirees whose savings were invested in companies who have found it necessary to reduce costs. How is the BCA analyst to decide which of these groups to favor at the expense of others? If that analyst were to put a thumb on the scale in order to “correct” some wages or prices, in which direction shall it be tilted? Better to resist such temptations altogether. If BCA stands for anything, it is for the proposition that analysis is far more informative when it is grounded in rigorous, objective observations about the revealed preferences of the public. When a BCA has been done properly, it accounts for the lost output that results from directing resources away from consumer goods and ser vices to, say, pollution control. The reduced output of market goods represents a welfare loss to somebody; we just do not know to whom. Perhaps a particular family bears a disproportionate share of the lost welfare, but that will not necessarily change the total. It may be fair or unfair, and it may be a very important thing for policymakers to know about—but it is not a net change in the benefits and costs. For better or for worse, the BCA metric is indifferent to who experiences gains and losses; it is interested only in the sum.

Reconciling the BCA Model with Empirical Observations A growing empirical literature, including research in this book, has identified statistically significant correlations between regulatory actions and changes in the level of employment and between employment status and human welfare. Are these observations sufficient to refute the equilibrium assumption, and should we now modify BCA to incorporate the observed welfare effects? To answer this question, we need to ask what predictions flow out of the BCA framework that might be inconsistent with the empirical results. And it is not clear what those might be. Certainly the traditional (going back to Bastiat) framework does not suggest that there will be no direct (and therefore observable) job losses—merely that they will effectively be offset by indirect (and thus more difficult to observe) job

200

Brian F. Mannix

gains elsewhere. And the traditional framework does not suggest that there will be no observable welfare effects—including health effects, psychological effects, and “nonconsumptive” effects— among workers, investors, and consumers. Rather, it suggests that business compliance costs, which have no direct welfare effects associated with them, can be used as a proxy for the multitude of welfare effects that we know will flow from them. Masur and Posner (this volume) note, correctly, that compliance costs may be passed through to investors, to consumers, or to workers. Then they point to empirical findings of welfare losses borne by workers and their families and argue that none of these costs have ever been included in a standard BCA. But it is also true that retirees who see their life savings disappear suffer serious welfare consequences, and none of those costs, per se, have ever been included in a standard BCA. And consumers who see the cost of heating their homes go up suffer welfare consequences, and some of them die without ever knowing why. None of these costs have ever been included in a standard BCA either. None of these have been included because including them would be double counting; the compliance cost is a stand-in for all of those things. Because it is such a complex problem, we do not have very good information about the ultimate incidence of regulatory compliance costs on the population. It seems likely that the distribution of costs across income levels is much more regressive than our income tax, but it is not something that can be easily mea sured. We can be pretty confident that consumers bear the lion’s share of the costs, simply because demand curves tend to be much steeper than supply curves, which causes compliance costs to translate into lost consumer surplus. The costs that fall on consumers tend to be highly diluted, but at the same time consumers are at a disadvantage in that they are poorly organized when it comes to influencing public policy decisions. Labor likely bears a smaller share of costs, but it is also the most concentrated share, and individuals can find their lives devastated by a single regulatory decision. Investors tend to be more diversified, so that the regulatory costs falling on them will often be diluted, but this does not apply to homeowners, landowners, and small business owners. Moreover, investors who are retirees may be relying on a lifetime of savings and may have limited options when they find their savings suddenly depleted. In any case, whether they are concentrated or diluted, the final incidence of regulatory compliance costs represents real losses to real people. Lest we get calloused with the recitation of dollar values, it can be useful to think about what form these losses may take. Using a rough value of $50 million per life to represent the income effect on mortality,1 for example, we can estimate that, for every $1 billion in regulatory compli-

Employment and Human Welfare

201

ance costs, 20 people will die prematurely. We do not know who, and we do not know how. But we can predict a statistical average of 20 premature deaths somewhere in the economy, whether among consumers, workers, or investors. Of course, we hope that these deaths will be more than offset by the life-prolonging benefits that the regulation generates. But, too often, people fail to realize that in a BCA, both benefits and costs are cut from the same cloth. This illustrates a major weakness of BCA as a communication tool. By using compliance costs as a proxy for downstream welfare effects, it allows people— including government decision makers— to believe the fiction that compliance costs somehow will be “absorbed by business.” The truth is, when it comes to welfare changes, business does not absorb anything. The costs of regulation fall on real people, and these costs can be painful and even deadly. So empirical observations of death and despair suffered by displaced workers and their families (or by retirees living on their depleted savings, or by consumers unable to make ends meet) are entirely consistent with BCA as it is practiced today. That is what we mean when we talk about compliance costs. Such empirical studies are an important reminder that welfare changes, whether on the cost or the benefit side of the ledger, are not just abstract dollars.

Conclusion Let us stipulate that the economy today is not at a full- employment equilibrium, that government policies have an important influence on this, and that it is properly a matter in need of urgent attention at the highest levels of government. Excessive or ill- considered regulations may well be a significant contributing factor. Still, it is not obvious that the use of the equilibrium assumption in the benefit– cost analysis of individual regulations is part of the problem, and it is even less obvious that setting aside that assumption would be a constructive part of the solution. Without an analytical framework that allows us to cata log welfare effects once and only once, it would be foolish to make ad hoc additions and subtractions to the benefits and costs of the standard BCA model. It may be helpful to think of the equilibrium assumption, not as a factual assertion to be refuted, but as a methodological tool that is approximately correct and that has the considerable virtue of facilitating the division of labor. It allows one governmental department, agency, or office to focus on a particular problem, while “assuming away” other market distortions or public policy problems— even ones that are important and that some other unit of government may be working on. Does

202

Brian F. Mannix

that sound like different offices might be working at cross-purposes? Sometimes they do, and it is something to be watchful for. But consider the genius of the private economy, which produces vast increases in human welfare, even while each of the billions of economic entities within it are working on narrow suboptimizations with limited information. It seems likely that government agencies will be more successful, both individually and collectively, if they try to mimic that aspect of private actors, rather than to approach each problem as if it contained every problem. One of the core techniques of operations research, the simplex method, takes an otherwise intractable search for a global optimum in a system of linear equations and renders it computable by following a series of more narrow stepwise suboptimizations (Dantzig 1963). If this sounds perhaps too formal as a metaphor for BCA, then policy incrementalism, or “muddling through,” sounds perhaps too informal (Lindblom 1959, 1979). The general point of both metaphors is that the easiest path to a global optimum may involve solving one problem at a time, using the equilibrium assumption to isolate it, study its effects, and test its solutions, and using BCA to explain how solving this problem can make us better off.

Notes 1. This number is used for illustrative purposes and is meant to represent the income effect on mortality, also known as the “health-wealth” effect. It is significantly larger than the more familiar “willingness to pay” or “value of life” metric, which is an estimate of the substitution effect. See, for example, Keeney (1990).

References Bastiat, Frederick (1845) “Candlemakers’ Petition,” in Economic Sophisms (1996). Irvington- on-Hudson, NY: Foundation for Economic Education, http://www .econlib.org/library/Bastiat/basSoph3.html (accessed 4 March 2013). Bastiat, Frederick (1850) “What Is Seen and What Is Not Seen,” in Selected Essays on Political Economy (1995). Irvington- on- Hudson, NY: Foundation for Economic Education, http://www.econlib.org/library/Bastiat/basEss1.html (accessed 4 March 2013). Dantzig, George (1963) Linear Programming and Extensions. Santa Monica, CA: RAND, http://www.rand.org/pubs/reports/R366.html (accessed 16 January 2013). Keeney, Ralph L. (1990) “Mortality Risks Induced by Economic Expenditures,” 10 Risk Analysis 147– 59. Lind, Robert, ed. (1982) Discounting for Time and Risk in Energy Policy. Washington, DC: Resources for the Future.

Employment and Human Welfare

203

Lindblom, Charles E. (1959) “The Science of ‘Muddling Through,’ ” Public Administration Review 19: 79– 88. ———. (1979) “Still Muddling, Not Yet Through,” Public Administration Review 39: 517–26. Lipton, Douglas W., et al. (1995) Economic Valuation of Natural Resources: A Handbook for Coastal Resources Policymakers. NOAA Coastal Ocean Program Decision Analysis Series No. 5. Washington DC: U.S. Department of Commerce, http://www.mdsg.umd.edu/programs/extension/valuation/handbook.htm (accessed 16 January 2013). U.S. Office of Management and Budget (1992) OMB Circular A-94, Revised. http://www.whitehouse.gov/omb/circulars_a094 (accessed 16 January 2013).

This page intentionally left blank

Reform

This page intentionally left blank

Chapter 11

Unemployment and Regulatory Policy Jonathan S. Masur and Eric A. Posner

Unemployment is generally thought to be a problem that is best addressed with fiscal and monetary policy, not with regulations. But regulatory agencies have long tried to calculate and respond to the possible unemployment effects of regulations. Some statutes require agencies to use so- called feasibility analysis, according to which the agencies should regulate up until there is significant job loss (Masur and Posner 2010). Even when statutes do not contain this requirement, the Obama administration has asked agencies to conduct “ job loss” analysis, under which agencies estimate the unemployment effects of proposed regulations and disregard regulations that would cause excessive job loss. The job loss analysis is conducted separately from a cost– benefit analysis (CBA), which the regulation must also pass. In a recent article, “Regulation, Unemployment, and Cost– Benefit Analysis,” we argued that regulatory agencies should incorporate the costs of unemployment into cost– benefit analyses of proposed regulations (Masur and Posner 2012). Feasibility analysis and job loss analysis make little sense because they do not specify the threshold at which job loss is excessive and do not explicitly make trade- offs between unemployment effects and social gains. Our article appeared at a politically contentious time. Republicans had been arguing that regulation causes unemployment and blamed the high rate of unemployment at that time on the Obama administration’s regulatory agenda. They proposed a bill that would ban new regulation when the unemployment rate exceeds 6 percent (Regulatory Freeze for Jobs Act of 2012). Our article received some modest media attention (“Clause and Effect” 2011) and was cited in the 2012 draft OMB report, which sought advice from commentators as to whether cost–benefit analysis should incorporate unemployment costs and, if so, how it should do so (Office of Management and Budget 2012).

208

Jonathan S. Masur and Eric A. Posner

In this chapter, we respond to some important questions and critiques that commentators have offered in response to our article. We also discuss some broader issues raised by the debate about the incorporation of unemployment costs into CBA, including the role of “second- order” or remote costs and benefits and the treatment of the ex ante incentives of regulation. To begin, we briefly recapitulate our earlier argument.

A Recapitulation When regulators conduct cost–benefit analysis, they typically compare the benefits of a proposed regulation and the compliance costs. On the usual understanding of CBA, compliance costs are borne by shareholders and consumers. If the firm can pass on the costs to consumers in the form of higher prices, it will do so. If not, shareholders will see a reduction in profits. If compliance costs force the firm to reduce production and lay off workers, then the workers will normally suffer losses as well. While unemployed, they will lose whatever surplus they enjoyed while working; they may incur search costs; and, if industry- specific human capital accounted for some of their wages, and they cannot find a job in the same industry, then they will suffer the destruction of some of their human capital. Economists have generally ignored the losses to workers, probably because they assume the losses will be small relative to the major costs and benefits of the regulation. But recent empirical work suggests that the loss to each worker who becomes unemployed may be large. The most likely explanation is that industry- (or firm-) specific human capital accounts for a large portion of wages in many regulated industries. The lost human capital of workers should be counted in any cost– benefit analysis just as the loss of physical assets. From a normative standpoint, there is no reason to prefer a regulation that causes a machine to lose its value because its output can no longer be marketed to a regulation that causes a worker’s skills to lose their value because the output that depends on those skills can no longer be marketed. To account for the loss to workers, CBA should incorporate the loss of human capital—which can be roughly approximated as the difference between the wage that the worker earns in his original position and the wage that he earns in the next-best position outside the industry (or outside of a position that requires that human capital). In our earlier work, we also estimated the social costs of lost jobs. Workers who are laid off lose an average of $100,000 in wages over the course of their lifetimes. These wage losses could simply reflect that laid- off employees were being paid above-market wages in their previous jobs, in which case the wage loss does not represent a true social

Unemployment and Regulatory Policy

209

cost. But they could also represent the loss of firm- or industry- specific capital and thus the loss of employee productivity. In such a case, the wage losses would be genuine social costs. Regardless of the loss of wages, unemployment also carries with it significant hedonic and transition costs. Workers who lose their jobs typically suffer substantial hedonic penalties—the loss of a great deal of happiness—that persists even after they have found new employment. They are also more likely to suffer a variety of health problems and to die earlier, even controlling for their lower income. In addition, when a worker is laid off she must often bear significant costs in the process of finding new employment. These include the costs of searching for a new job, retraining costs, and the costs of moving to a new location. Finally, the worker’s dependents typically suffer costs and losses from the worker’s unemployment. None of these types of costs have ever been included in standard CBA. In many instances, the inclusion of unemployment costs will not (and should not) alter a regulatory decision. For instance, in recent work, Walker (2011:28) calculated that the Clean Air Act created unemploymentrelated costs of approximately $9 billion. Compared with benefits estimated to be between $160 billion and $1.6 trillion (Walker 2011:28), these additional costs do not seem terribly significant. However, including unemployment costs in CBA could nonetheless affect the outcome of many regulatory analyses. In our article, we conservatively estimated the costs of unemployment at $100,000 per worker over the worker’s lifetime, or slightly more than $3,000 per year. (Regulation that leads to greater employment would produce equally significant benefits.) We then recalculated a CBA, performed by EPA, on a regulation limiting the use of chlorinated compounds in the pulp and paper industries, to include the costs of unemployment. EPA projected that the regulation would create $159.5 million in net lifetime benefits but would also result in more than 5,711 lost jobs, a fact that EPA calculated and acknowledged but did not include in its CBA. Recalculating the cost–benefit analysis to include the costs of these lost jobs reduces the lifetime net benefits of the regulation from $159.5 million to –$411.6 million. This does not necessarily mean that the agency should not have regulated at all. But it suggests that the agency should have considered a less stringent regulation, one that would produce smaller benefits but would also have led to less unemployment.

Objections The Double- Counting Problem David Driesen (2012) has argued that our approach would involve counting the costs to workers twice. Driesen correctly notes that when

210

Jonathan S. Masur and Eric A. Posner

agencies calculate the costs of regulation, they look at the firm’s compliance costs and do not consider how those costs are distributed. For example, suppose that it costs a firm $100,000 to comply with a new regulation. The firm dictates how that cost is distributed. If demand for the firm’s product is inelastic, the firm will pass on the cost to consumers. If not, the firm will either cut dividends or reduce wages (or fire workers). It is possible that the firm will simply fire a worker whose wage is $100,000. If the worker does not obtain a new job, then the worker bears the full cost of the regulation. Thus, it would be wrong to add the $100,000 cost to the worker to the $100,000 compliance cost. This analysis is incorrect. To see why, observe that if a firm fires a worker who earns $100,000, the firm will not actually save $100,000. That would be the case only if the worker earns $0 of revenue for the firm; but if that were true, the firm would have fired the worker long before the regulation was issued. Firms employ workers only when their value is greater than their cost; thus, firing a worker would not enable the firm to raise funds to finance compliance costs, as Driesen argues, but instead put the firm even farther in the hole. Put another way, Driesen assumes a world in which the firm fires a worker earning $100,000 and nothing else changes. But this worker must have been producing something for the firm, or else the worker would have been fired long ago. This lost productivity is precisely the social cost that can accompany regulation-induced unemployment. Driesen confuses accounting costs and economic costs. By assuming away the loss of production, Driesen assumes away the social cost we seek to measure. Inefficient Industries and the Buggy Whip Factory Problem Another objection is that, if our analysis is correct, it would be proper to refrain from regulating (or fully regulating) industries that produce net social costs because of harmful externalities. Imagine a coal-fired power plant that produces a great deal of pollution. Assume that the negative externality is greater than the compliance cost, which we will assume will be taken out of consumer surplus— so that the factory produces energy whose social value to consumers is less than the social loss from pollution. Meanwhile, a number of workers have industry- specific skills; if the plant is shut down, they will lose their jobs and earn lower wages in other industries. Should the lower wages make a difference to regulatory analysis? To make the answer as clear as possible, suppose that the negative externality is worth 100, the cost of compliance is 90 (and hence the consumer’s surplus, as we assume), and the loss to the workers is 20.1 In addition, imagine a second firm that uses robots instead of workers.

Unemployment and Regulatory Policy

211

Those robots are worth 20 but would lose their value if the regulation is issued. But if that is true, then compliance costs go up to 110, and the regulation would fail a cost– benefit analysis. The same point can be made about the firm with workers. The difference is that the workers enjoy the value of their human capital (which they recover in the form of a wage premium), while the firm enjoys the value of the robots. There is no reason to treat the cases differently, but that would be the result if unemployment costs are not incorporated into CBA. The odd- seeming result is that regulators would refrain from issuing a regulation that prevents a coal power plant from spewing out pollution where the cost of the pollution is greater than the value of the electricity for consumers, just in order that the power plant workers continue to hold jobs that do not seem to be socially productive. But this impression that the jobs are not socially productive is an artifice of the way in which standard CBA counts costs. The coal is not valueless; it provides electric power. Some of the value of that power is captured by consumers, some by the owner of the power plant, and some by the workers. Failing to count the benefits captured by the workers results in undercounting the overall benefits. From an economic perspective, the factory should remain open because the net social benefits are positive, and it is arbitrary how those benefits are distributed among workers, consumers, and shareholders. A skeptic might then wonder whether the same analysis that we apply here to regulation should be applied to market-based unemployment as well. For instance, prior to the invention of the automobile, some workers made a living producing buggy whips. This production was socially beneficial until the automobile was invented and demand for buggy whips collapsed. Should the government have propped up buggy whip producers in order to avoid unemployment in that industry? The answer is that it should not have. If the market is functioning properly, the fact that buggy whip factories close and buggy whip manufacturers are laid off indicates that the social costs of producing buggy whips exceed the social benefits. There is no surplus to divide between consumers and producers. If there were some surplus, consumers, shareholders, and workers would negotiate some division of the surplus. The price of buggy whips might fall, returns to buggy whip shareholders might fall, and wages might fall as well, but buggy whips would continue to be produced. A well-functioning market is the mechanism that determines whether an activity produces net social benefits or costs. No cost–benefit test is necessary; the market itself performs this test. This is the reason that cost–benefit analysis is necessary when the government regulates: the regulation is standing in as a corrective to a poorly functioning market, and thus the market itself cannot perform the cost–benefit test.

212

Jonathan S. Masur and Eric A. Posner

Thus far we have been discussing only traditional command-andcontrol regulation, in which government mandates that firms install a particular safety device, or discontinue use of a hazardous chemical, or even shut down entirely. There are alternatives to this type of regulation, including the imposition of Pigouvian taxes.2 A Pigouvian tax is designed to correct a market failure by forcing some actor to internalize a cost that it is creating. It thus replicates a well-functioning market by forcing this actor to account for all of its externalities. Accordingly, CBA is not necessary— and thus unemployment costs are irrelevant—when the government imposes a Pigouvian tax for the same reasons that it is unnecessary when a market is functioning properly. That is, imagine a firm that creates an externality of 10 per unit of a good that it produces. Before the government bans production of the good (via commandand- control regulation), it should conduct a CBA to determine whether such a ban would be cost–benefit justified. But if the government instead simply imposed a Pigouvian tax of 10 on each unit of that good, thereby forcing the producers and consumers of the good to internalize the externality, it would not need to conduct a CBA. If the market for the good is otherwise efficient, the good will be produced and consumed in optimal amounts once there are no externalities. Should unemployment costs be calculated in determining the Pigouvian tax itself? The answer is no. Once the firm is forced by the tax to internalize its externalities, it will make the proper decision whether to continue to produce (and how much), taking into account the benefits from production for consumers, shareholders, and workers. The difference between CBA and the Pigouvian tax is that CBA takes into account the costs to the firm, not just the benefits to third parties, and those costs include potential costs to workers, as well as to shareholders and consumers. When the government calculates a Pigouvian tax, it must ignore all of those costs. Wages: Rents Versus Returns Another question we have received is whether wage losses incurred by workers would not be recovered by shareholders in the form of higher profits or consumers in the form of lower prices, in which case there is no social loss. If that is the case, then wage losses incurred by workers should not play a role in CBA. The answer is that the wage loss is a social loss only if the worker was earning a wage premium because of human capital— and thus because of his productivity—rather than because the worker was capturing rents. The most common source of such rents is the union wage premium. If

Unemployment and Regulatory Policy

213

unionized workers receive above-market wages of, say, $110, lose their jobs as a result of regulation, and then obtain identical or similar nonunionized jobs that pay $100, then it would be wrong to include the $10 wage loss in the cost–benefit analysis. The regulation is simply eliminating the union, putting downward pressure on wages until they reach market levels. In some cases, the affected workers may have no specialized skills. For instance, if a unionized steel mill is forced to close, there will be some workers whose skills have nothing to do with steel manufacturing— accountants, to name one example—who will be laid off. These workers possess no industry- or firm- specific skills and can find new jobs in a variety of other industries. If they take nonunionized accountant jobs in high-tech chip factories, their wages might decrease. But this decrease would not represent a social loss. However, as we discuss in Masur and Posner (2012), the empirical literature so far suggests that long- term wage losses endured by many workers cannot be attributed for the most part to the loss of rents, leaving the loss of human capital as the most likely explanation. In any event, the regulator would need to take this into account, although this has the politically sensitive implication that regulators would count wage losses in unionized industries less than wage losses in nonunionized industries when conducting the cost–benefit analysis. This implicates a broader point about what CBA is really mea sur ing when it counts wage losses. The socially valuable aspect of employment is not the wage paid by the firm to the employee—that is just a transfer. Rather, it is the worker’s productivity that is socially valuable. If a skilled worker who had been trained to make steel can no longer find work in a steel mill, that worker’s potential productivity has been lost. If the steel worker can only find a job flipping burgers at McDonald’s, the lost productivity is the difference between the value of the individual’s work in the steel mill and the value of the work at McDonald’s. This is why we describe the loss as a loss of human capital. The worker’s human capital—the skills that she has acquired, the education and training that she has received— are what create the potential for her to be a productive steel worker. There is no “direct” mea sure of the worker’s lost productivity; the only way to mea sure the value of the productivity is to calculate what it is worth in the market. This is why it is the worker’s market wage that matters for CBA. In a competitive marketplace, workers will be paid the value of their productivity, and thus the wage loss that results from unemployment will be the clearest mea sure of the lost productivity. If workers are earning above-market rents, then they are earning wages in excess of their productivity. It would overstate the social loss from

214

Jonathan S. Masur and Eric A. Posner

unemployment if CBA were to count their entire wage loss on the cost side of the ledger.3

Broader Issues in Cost– Benefit Methodology Our argument for counting unemployment costs in CBA raises larger questions about the methodological choices that underlie textbook cost–benefit analysis. Two issues are of particular importance. First, CBA has traditionally drawn a line between “first- order” and “second- order” costs and benefits, calculating the former and largely eschewing the latter. Second, CBA is predominantly an ex post analysis, rather than an ex ante one, in the sense that it takes all economic activity preceding the regulation as given and asks only whether the regulation will produce net costs or benefits. That is, CBA examines whether a regulation will improve welfare from the status quo, not whether that regulation would create efficient ex ante incentives for parties to anticipate regulation and prepare accordingly. We suggest here that it might be time to reconsider these methodological choices. There is no principled distinction between first- and second- order costs and benefits, and CBA should not make distinctions on that basis. More speculatively, policymakers should consider reformulating CBA to take greater account of ex ante incentives and pay less attention to ex post benefits and costs. Such a move could result in increased welfare over the long term. First- and Second- Order Costs Any type of regulation will produce some number of initial consequences followed by a potentially infinite set of subsequent ripple effects. Consider the Department of Transportation’s (DOT’s) recent regulation raising fuel economy standards for automobiles and light trucks. The regulation described will, at the outset, force automobile manufacturers to produce more fuel- efficient automobiles. Consumers will pay more for these cars, but they will also consume less gasoline per mile driven, making it less expensive to operate the vehicles. These are the most direct consequences of the regulation, but they are hardly the only ones. Consumers will also purchase fewer cars because of the higher price, which will lead to unemployment as car manufacturers lay off excess workers. At the same time, the need for more fuel- efficient engines may lead to economic development and job growth in industries related to engine technology. The higher price of automobiles may lead some consumers to shift away from driving and toward public transportation. Other consumers who can still afford cars might find driving less expensive than before (because of their improved fuel

Unemployment and Regulatory Policy

215

economy) and choose to drive more. DOT estimated that there would be fewer cars on the road but that the cars that remained would be driven more miles (Average Fuel Economy Standards Passenger Cars and Light Trucks Model Year 2011 2009). DOT believed that the total number of automobile-miles driven would increase, and this in turn would increase automobile fatalities and delays due to roadway congestion. Every additional roadside fatality created (or avoided) will also have direct effects on the families of the victims, and so forth. For that matter, increases or decreases in the demand for gasoline among American drivers might affect the price of gasoline, the rate of global warming, or even American foreign policy, all of which could produce a variety of tertiary effects. Most of these hypothetical ramifications are what might be called “second- order” effects of the regulation (Institute for Policy Integrity 2012:10). This is to distinguish them from what one might describe as the regulation’s “primary” effects: the costs of improving fuel economy and the reduction in automobile operating costs. As a general matter, CBA does not attempt to account for second- order effects (Baram et al. 1998:19), though there are notable exceptions to that rule. For instance, the CBA that accompanied DOT’s fuel economy rule included an analysis of a variety of that regulation’s effects, including the regulation’s impact on global climate change, as we discussed at length in earlier work (Masur and Posner 2011). Several commentators have suggested to us that it is inconsistent to advocate including unemployment costs in CBA without similarly incorporating all other “second- order” costs and benefits. At the outset, it is worth noting the slippery nature of these categories. If we think of a regulation’s compliance cost as reducing the joint surplus that consumers and producers share, then unemployment costs are just one manifestation of the reduction in producer surplus—part of the overall compliance cost. Most practitioners of CBA would classify this as a “first- order” cost. On the other hand, the causal chain that leads from regulation to unemployment involves a number of steps. Regulation makes a product more expensive to produce; elasticity in the market for that product causes the product’s price to rise; that rise in price reduces the quantity of that product that consumers will purchase; the reduction in quantity causes firms to produce less; the drop in production renders some of the firm’s workers irrelevant; and the firm responds by firing those workers. Accordingly, many commentators would consider unemployment costs a second- order (or thirdorder) effect. We need not wrestle this question to the ground, however, because we do not believe that there is any conceptual importance to these categories. There is nothing magical about the number of causal steps between a regulation and its effect. Policymakers must find some

216

Jonathan S. Masur and Eric A. Posner

other principle to separate those costs and benefits that should be calculated and included in CBA from those that should not. This problem is common to several areas of law. Any tortious act or breach of contract can create an extended series of secondary and tertiary effects. For example, a driver who runs his car into a pedestrian will not only injure the pedestrian, forcing the pedestrian to incur medical costs, but also cause the pedestrian to miss work, which may put a greater burden on his colleagues or cause the employer to lose profits, and also affect family members in various ways and possibly their work and relationships with other people. The accident might use emergency personnel who are then unable to rescue someone else. And so on. Faced with the need to draw some line, tort and contract law usually address this problem by holding the tortfeasor or the breaching party responsible only for the direct and foreseeable effects of her action.4 The theory behind this approach in tort and contracts is threefold: (1) it is only fair to hold a party responsible for the foreseeable consequences of her actions; (2) it is impossible to give individuals incentives to take precautions against unforeseeable harms; and (3) unforeseeable harms are likely to be significantly smaller and harder to measure than direct harms (Posner 2007:1936), or they may be offset by benefits,5 so that including them will create excessive error costs (Lawson v. Safeway Inc. 2010; Palsgraf v. Long Island R. Co. 1928). The first two of these rationales do not apply to cost–benefit analysis, which is not concerned with individual responsibility or with ex ante incentives (more on that point later). Instead, CBA practitioners have largely seized on the third rationale: that second- order costs and benefits are much smaller and more difficult to mea sure (Baram et al. 1998:19; Goldberg and Epstein 2005:511). This may be true as a general matter, but there are a number of “second- order” costs and benefits that could be large enough to alter the outcome of a CBA in a variety of regulations. We demonstrated this with respect to reductions in carbon outputs and climate change (Masur and Posner 2011) and then unemployment (Masur and Posner 2012).6 It is not enough, then, for CBA to declare that certain costs and benefits are “second- order” and ignore them. Nor should CBA incorporate second- order costs and benefits on an ad hoc basis, randomly including some but not others. Rather, practitioners of CBA should be guided entirely by pragmatic concerns: (1) is a par ticu lar cost or benefit projected to be relatively large in comparison with the overall net costs and benefits of the regulation; and (2) can that cost or benefit be mea sured with any degree of accuracy? When the answer to both questions is “yes,” CBA should mea sure and include the cost or benefit within the overall calculation. We believe that the answer to both questions is likely

Unemployment and Regulatory Policy

217

to be “yes” for a wide range of regulations that affect employment, but additional research needs to be done before we can know for sure. Ex Post Versus Ex Ante Underlying the debate regarding unemployment and other secondorder costs is the fact that CBA is fundamentally an ex post procedure, rather than an ex ante one. That is, CBA asks whether, under current conditions, a regulation or project will produce net benefits or net costs. It is not designed to create correct ex ante incentives. When a regulator uses cost–benefit analysis to determine whether an industry excessively pollutes, it will usually not consider whether a regulation will have negative ex ante effects, for example, on the decisions of people to live near factories (which increases the amount of harm). In other words, it may be better from a social standpoint for people not to move near dirty factories, in which case the pollution causes no harm, than for people to move near dirty factories that are then shut down by a regulator. Tort law addresses this incentive problem by depriving people who “come to the nuisance” of a remedy under certain circumstances, but cost–benefit analysis does not. In a similar vein, regulators often impose stricter regulations on new plants than on existing plants even though regulating existing plants with equal stringency could have desirable ex ante incentives on firms to anticipate future regulation. Indeed, without better regulatory parity, firms may overinvest in plants in the hope that high compliance costs could block future regulation. Similar points can be made about regulation that takes into account unemployment costs. It is socially beneficial when workers anticipate which industries will be most socially valuable in the future and seek employment and training in those industries. From the perspective of social welfare, it might be preferable if a worker entering the labor force in 1995 sought training in the solar power industry rather than the coalgenerated power industry (to the extent that such substitutions are possible). CBA does not attempt to create additional incentives, beyond those already generated by the market, for workers to make welfare- enhancing choices. It asks only whether a regulation or project would enhance welfare compared with the status quo, and the status quo already subsumes these types of individual decisions. But as a result workers might enter marginal industries, which, under our approach, may be saved rather than regulated out of existence. In this sense, our approach takes CBA as a given; it was not our purpose to reconsider CBA’s foundations. But this perverse result is a reminder that CBA is very much a second- best procedure. The task of generating efficient ex ante incentives might be too complicated for

218

Jonathan S. Masur and Eric A. Posner

policymakers employing CBA.7 Yet it is worth at least considering whether cost–benefit analysis could be altered to focus on ex ante incentives as well as ex post consequences. A full- scale treatment of this subject is beyond the scope of this chapter, but one can imagine the general shape that a reformulated CBA would take. The crucial questions for regulators would be whether particular regulatory costs and benefits were foreseeable (that is, whether the cost of anticipating the regulation was less than social benefits) at an earlier point in time and whether individuals and firms failed to anticipate the costs and benefits. If, for instance, a future regulatory cost was foreseeable and a fi rm failed to account for it, it may be that the cost should not be included in the cost– benefit calculation. Thus, cost–benefit analysis might accept the possibility of welfare-reducing regulation in the interest of creating incentives for affected parties to take precautions ex ante due to the possibility of regulation.8 Suppose, for instance, that a firm must decide in 2012 whether to install a coal-fired power plant or a solar plant. The coal-fired plant will cost 100 and the solar plant will cost 120, but suppose it is well understood that the coal-fired plant will create a negative externality worth 30 due to climate change. The firm is not yet required to internalize the climate change externality because carbon emissions are not regulated. But it is certainly foreseeable that coal-fired power might eventually be banned if its costs exceed its benefits. Suppose that the firm installs the coal- fired plant, and 20 years later the EPA considers banning coalburning power plants. It seems clear that firms should be prohibited from installing additional coal-fired plants because the costs of burning coal (100 + 30) exceed the costs of using solar power (120). Should the EPA additionally close existing coal-burning plants and force firms to install solar in their place? The costs of doing so (120 minus the cost to install a new solar plant) would seem to exceed the benefits (30). But perhaps policymakers should adopt an ex ante approach and refuse to include the cost of switching to solar in the cost–benefit analysis. After all, in our example, the firm could have anticipated that the costs of coal power would exceed the benefits and that coal might someday be regulated. Perpetuating the status quo by allowing the existing coal plant to continue operating would encourage firms to ignore these costs ex ante. Conversely, forcing the firm to switch to solar despite the fact that such a move is welfare diminishing might provide the firm with the proper ex ante incentives to install solar power in the first place. This would effectively align CBA with the literature on transition relief, which has counseled against grandfathering preexisting harmful uses where doing so would diminish incentives for firms to anticipate legal change (Kaplow 1986; Levmore 1999; Masur and Nash 2010; Shavell 2008).

Unemployment and Regulatory Policy

219

The same analysis would apply to unemployment costs. Regulators would need to ask whether workers who enter an industry could foresee at reasonable cost that it would at a later time be optimal for the government to regulate the industry or increase the level of regulation so that the workers’ industry-specific skills would lose their value. We suspect that, in the case of workers, the answer to this question will almost always be no; and so we do not see ex ante effects as posing a significant challenge to our proposal that standard CBA incorporate unemployment costs. The literature on legal transitions and grandfathering is vast and raises significant questions that would need to be addressed before policymakers could consider adopting “ex ante”- style CBA wholesale. Among the many problems this approach would create is the difficulty in gauging which costs and benefits were foreseeable at some point in the past. Nonetheless, this very brief sketch should make clear that a cost– benefit methodology focused on ex ante incentives could function very differently than standard CBA and could conceivably produce greater social benefits in the long term.

Conclusion Ultimately, it is impossible to know in advance what impact unemployment costs will have on regulatory decision making. It may be that most regulations will resemble the Clean Air Act, in that the net benefits so far exceed the costs that unemployment costs represent little more than a rounding error. Or it may be that many regulations, like the EPA pulp and paper regulation we analyzed in prior work, will create sufficiently high unemployment costs that the agencies involved should consider less stringent regulation. It is even possible that existing “ job loss analysis” overweights unemployment so that if unemployment costs were instead properly incorporated into cost–benefit analysis, stricter rather than weaker regulation would result. The only way to know for certain is for regulatory agencies to begin incorporating unemployment costs into their CBAs, as we have argued. Moreover, unemployment costs are the canary in the cost– benefit analysis coal mine. For decades agencies have ignored what might be a significant source of costs (or, in some cases, benefits) on the assumption that they were at most second- order. That assumption has been proven false, at least in some cases. This raises the question of what other important “second- order” costs and benefits agencies have ignored and whether those costs and benefits might be substantial enough to affect regulatory outcomes. The benefits of reduced carbon emissions, which agencies began to consider only in 2009, are an example of such a benefit (Masur and Posner 2011). Agencies that conduct cost–benefit analysis

220

Jonathan S. Masur and Eric A. Posner

should engage in a comprehensive search for such substantial but uncounted costs and benefits. Even more generally, agencies should reconsider some of the assumptions that have underlain cost–benefit analysis from the beginning. Agencies might be capable of creating incentives for private actors to anticipate efficient regulations (and adopt efficient behaviors earlier) by adjusting the way that they conduct cost–benefit analysis—for instance, by counting only certain costs and benefits. This would represent a substantial departure from traditional CBA, and perhaps it is ill advised. But the weaknesses of cost–benefit analysis that we have already exposed suggest that further changes could be in order.

Acknowledgments The authors would like to thank Cary Coglianese, Adam Finkel, Howard Chang, Shi-Ling Hsu, William Hubbard, Prasad Krishnamurthy, Michael Livermore, Tom Miles, Jonathan Nash, Paul Noe, Arden Rowell, and participants in the Penn Program on Regulation’s conference “Regulation’s Impact on Jobs” for helpful comments on drafts of this chapter. This research was funded in part by the David and Celia Hilliard Fund.

Notes 1. As we noted in Masur and Posner (2012) and describe here as well, the loss to the workers is mea sured by taking their current market wage and subtracting the market wage that they will earn in their next job and any transition costs (moving costs, retraining costs, etc.). In Masur and Posner (2012), we conservatively estimated that amount at $100,000 per worker over the worker’s lifetime. 2. We thank Howard Chang for raising this issue. 3. We note that the above discussion centers principally around the costs of unemployment generated by lost productivity. In addition, as we explained in Masur and Posner (2012), there may be significant transition costs as unemployed workers are forced to search for new employment, seek job training and education, or move to another location in search of a job. These costs have nothing to do with the wages that the workers were previously earning, and so they are separate from the discussion in this section. 4. Restatement (Second) of Contracts § 351 (1) (1981) (“Damages are not recoverable for loss that the party in breach did not have reason to foresee as a probable result of the breach when the contract was made”). See also Wood v. Foremost Ins. Co. (2007) and Dillon v. Legg (1968). 5. This is illustrated by the economic loss rule, which does not allow businesses that lose profits from a tort to collect damages for those lost profits. The theory is that the “lost” customers will patronize other businesses, so there is no or little net economic loss. 6. The regulations we examined in our 2011 article on the social cost of carbon did not cease to be cost– benefit justified under different carbon valuations (Masur and Posner 2011). Yet that does not mean that they were optimal, either.

Unemployment and Regulatory Policy

221

The larger point is that the mere fact that the benefits of a regulation will exceed the costs does not mean that an agency should promulgate precisely that regulation. Rather, agencies should search for regulations that maximize benefits net of costs, a point that Adam Finkel explores in greater detail in his chapter in this volume. 7. At a minimum, it is important that CBA be impervious to strategic behavior on the part of potential targets of regulation. If CBA can easily be manipulated, it is of little use to policy analysis. In earlier work (Masur and Posner 2012), we described why including unemployment costs would not expose CBA to significant strategic manipulation by regulated parties. 8. We discussed a version of this point and noted its limitations in Masur and Posner (2012).

References Baram, Michael, et al. (1998) “Transgenic Agriculture: Biosafety and International Trade,” Boston University Journal of Science and Technology 4: 4– 51. “Clause and Effect: The Business Cycle Matters When Assessing the Cost of New Regulations” (2011), The Economist, http://www.economist.com/node /21534767 (accessed 4 January 2013). Driesen, David (2012) “Job Loss and the Incoherent Expansion of Cost– Benefit Analysis,” http://works.bepress.com/context/david_driesen/article/1003 /type/native/viewcontent (accessed 4 January 2013). Goldberg, Victor P., and Richard A. Epstein (2005) “Introductory Remarks: Some Reflections on Two- Sided Markets and Pricing,” 2005 Columbia Business Law Review: 509–13. Institute for Policy Integrity (2012) “The Regulatory Red Herring: The Role of Job Impact Analyses in Environmental Policy Debates,” http://policyintegrity.org/files/publications/Regulatory_Red_Herring.pdf (accessed 4 January 2013). Kaplow, Louis (1986) “An Economic Analysis of Legal Transitions,” Harvard Law Review 99: 509– 617. Levmore, Saul (1999) “Changes, Anticipations, and Reparations,” Columbia Law Review 99: 1657. Masur, Jonathan S., and Jonathan R. Nash (2010) “The Institutional Dynamics of Transition Relief,” New York University Law Review 85: 391. Masur, Jonathan S., and Eric A. Posner (2010). “Against Feasibility Analysis,” University of Chicago Law Review 77: 657–716. Masur, Jonathan S., and Eric A. Posner (2011) “Climate Regulation and the Limits of Cost– Benefit Analysis,” California Law Review 99: 1557–1600. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Office of Management and Budget (2012) Draft 2012 Report to Congress on the Benefits and Costs of Federal Regulations and Unfunded Mandates on State, Local, and Tribal Entities. Washington, DC: Office of Management and Budget, http://www.whitehouse.gov/sites/default/fi les/omb/oira/draft_2012_cost _benefit_report.pdf (accessed 4 January 2013). Posner, Eric A. (2007) “Climate Change and International Human Rights Litigation: A Critical Appraisal,” University of Pennsylvania Law Review 155: 1925– 45.

222

Jonathan S. Masur and Eric A. Posner

Shavell, Steven (2008) “On Optimal Legal Change, Past Behavior, and Grandfathering,” Journal of Legal Studies 37: 37– 85. Walker, W. Reed (2011) “The Transitional Costs of Sectoral Reallocation: Evidence from the Clean Air Act and the Workforce.” U.S. Census Bureau Center for Economic Studies Paper No. CES-WP- 12- 02, http://papers.ssrn.com /sol3/papers.cfm?abstract_id=2000069 (accessed 4 January 2013).

Cases Cited Dillon v. Legg, 68 Cal. 2d 728, 739 (1968). Lawson v. Safeway Inc., 191 Cal. App. 4th 400, 417, 119 Cal. Rptr. 3d 366, 379 (2010). Palsgraf v. Long Island R. Co., 248 N.Y. 339, 345, 162 N.E. 99 (1928). Wood v. Foremost Ins. Co., 477 F.3d 1027, 1029 (8th Cir. 2007).

Legislation and Regulation Cited Average Fuel Economy Standards Passenger Cars and Light Trucks Model Year 2011, 74 Fed. Reg. 14,196, 14,412 (Mar. 30, 2009) (codified at 49 C.F.R. pts. 523, 531, 533, 534, 536, and 537). Regulatory Freeze for Jobs Act of 2012, H.R. 4708, 112th Congress (2012). Restatement (Second) of Contracts § 351 (1) (1981).

Chapter 12

Reforming the Regulatory Process to Consider Employment and Other Macroeconomic Factors Stuart Shapiro

The issue of unemployment is never far off of the national radar. During the Great Recession, it was at the forefront of policy debates, with both parties claiming unique solutions to lingering high unemployment rates. In the past few years, the role of regulation has also become central in the debate over jobs— and jobs have become central in the debate over regulation. Whereas the rhetoric used to oppose new regulations had earlier focused on costs, since the onset of the Great Recession oppositional rhetoric has increasingly focused on employment (Livermore et al. 2012). Every Republican presidential contender in the lead-up to the 2012 election cited regulation as a cause for the slow recovery.1 President Obama halted a prominent environmental regulation tightening ozone standards reportedly in part because of anticipated effects on employment (see Hargreaves 2011; Sunstein 2011). The president also issued several sets of instructions to agencies to consider employment effects in their regulatory decision making (Obama 2012, 2011a, 2011b).2 Increased attention to the effects of regulation on employment does not arise only from political calculations. Recent research has shown that the cost of long-term unemployment is larger than had been previously understood (Sullivan and von Wachter 2009). Long-term unemployment also has public health effects and effects on the families of the unemployed (Blanchflower and Oswald 2004). In a sustained economic downturn, the likelihood increases that workers who lose their jobs will become members of the long-term unemployed. Any regulatory system that aspires to understand the costs and benefits of regulation should

224

Stuart Shapiro

understand its costs and benefits to the larger economy, particularly its effects on employment. But how do we gain such an understanding? As I discuss in this chapter, agencies have long been required by executive order to assess the effect of their regulatory decisions on the larger economy. But in examining 56 recent Regulatory Impact Analyses (RIAs), I show that agencies have fallen woefully short of doing so in a meaningful way. This lack of attention could be due to negligence on the part of regulatory agencies. It could also be a rational response to a question that has no easy answers. Indeed, the literature on regulatory impacts on the broader economy produces conflicting results. Despite limited evidence that requirements for more detailed analysis have any impact on agency decisions, the political response has largely been to exhort agencies to consider more thoroughly the impacts of each of their regulations or to pass new legislation requiring them to do so. This transactional approach, requiring enhanced analysis for each regulation with potential impacts, is fraught with difficulties. Continuing to require additional analyses by agencies is an approach whose efficacy is belied by the history of regulatory analysis and is unlikely to produce results that will meaningfully affect agency decisions. The decision to consider the employment effects of regulations should not depend on the political demand to do so or on random factors unconnected to the magnitude of the impact. Finally, it is highly plausible that while individual regulations have limited effects, cumulatively they make it harder for businesses to profitably operate. If it is the cumulative effect of regulations that affects jobs, we should analyze these effects cumulatively, either instead of assessing them individually or in addition to doing so. In short, any procedural solution should divorce the question of regulation’s impact on jobs from the transactional nature of regulatory review. I propose delegating the task of analyzing the effects of regulation on employment and on other macroeconomic variables to an agency outside the regulatory process. Such an agency could do this analysis in a meaningful way. Such an agency could also conduct analyses on other problems that have bedev iled regulatory policy, such as distributional impacts and retrospective analysis. Such an agency could produce a series of analyses that could actually improve policymaking.

Regulation and Its Impacts In the 2012 election season, Republican candidates took it as an article of faith that regulation has a negative effect on employment levels. That viewpoint has undeniable theoretical appeal. Regulations typically re-

Reforming the Regulatory Process to Consider Employment

225

strict the decisions of business as to how to invest their resources. Any such restriction is likely to mean that fewer resources are available for other business expenditures, including the wages paid to labor. It also follows that businesses may be less likely to innovate and become less productive when faced with additional resource constraints. However, as other chapters in this volume have recounted, this simple explanation misses a fair amount about the effects of regulation. The direction and the magnitude of the effect of regulations on employment depend on factors that vary by the state of the economy, the industry affected, and numerous other factors, creating substantial uncertainty over whether and how much any given regulation might affect employment. Given this uncertainty about the effects of regulation, how do agencies attempt to assess prospectively the impact of their particular regulations on employment and other macroeconomic variables? Under Executive Order 12866, agencies must include effects of their regulation on employment, productivity, and competitiveness in an RIA for any regulation with an annual economic effect of more than $100 million. Given that the federal government issues dozens of such regulations, agency RIAs should be a fertile source of examination of the effect of regulation on jobs and other macroeconomic variables. Only two other studies have looked at multiple RIAs to discern how agencies approach the question, and their results are not encouraging. Masur and Posner (2012) described 14 regulations that contained an analysis of the impact on employment. The RIAs, dating from 1994 to 2011, largely contained point estimates of the jobs lost, with estimates ranging from a U.S. Environmental Protection Agency (EPA) regulation affecting the construction industry and costing more than 7,000 jobs to several regulations predicted to lead to only negligible job losses or even to be offset by job gains in other industries. Masur and Posner (2012) took agencies to task for sloppy analysis.3 They argued that currently the analysis of job impacts by agencies is ad hoc and incoherent and that it rarely includes a discussion of why a certain level of job loss is acceptable but greater losses are not. They called on agencies to better integrate the employment effects of their regulations into their analyses and to disaggregate the impacts so that industryspecific effects could be captured. They argued that this is critical because, when an industry shrinks so that workers must search for jobs outside the industry, those workers’ industry- specific human capital is destroyed. A report for the Institute for Policy Integrity considered several approaches used by EPA to analyze the effect of their regulations on jobs (Livermore et al. 2012). The report’s authors noted that agencies often neglect to include job analyses because jobs created or lost are often not

226

Stuart Shapiro

considered in the standard cost–benefit analysis, based on the assumption that labor markets are relatively efficient, meaning that the costs associated with layoffs should be transitory. The report concluded that each of the methods that agencies use to assess job losses have problems, but it stopped short of seconding the recommendation by Masur and Posner (2012) for more detailed analysis. The work by Masur and Posner (2012) and Livermore et al. (2012) is extremely helpful in assessing current agency practice. They both highlight the challenge of conducting a job impact analysis for each regulation. Neither set of authors, however, tries to differentiate between RIAs that include a look at jobs and those that do not. One might hope that agencies follow some logic in deciding whether to quantify the macroeconomic effects of regulation. In the next section, I compare RIAs containing such an analysis with those that do not in order to discern whether such a consistent pattern exists.

Regulatory Impact Analysis: Where We Are Now Agencies have been required to conduct RIAs since 1981, when President Reagan issued Executive Order 12291. There have been many studies of the quality of these RIAs (Ellig and McLaughlin 2011; Hahn and Dudley 2007; Hahn and Tetlock 2008; McLaughlin et al. 2012). In general, existing studies have concluded that the quality of RIAs has been uneven at best, with many of them missing information considered basic to using an RIA to assist in regulatory decision making. These studies have largely been concerned with whether the agency evaluated alternatives to its preferred policy, appropriately addressed discounting and uncertainty and thoroughly monetized benefits. None of them focused on the current executive order’s requirement to assess the impact of regulations on employment, productivity, and competitiveness. In two earlier works, John Morrall and I attempted to place in context the amount of information provided by RIAs. In one study, we found that RIAs containing greater amounts of information were not associated with regulations that had greater net benefits. Instead, political factors correlated best with the net benefits of a rule. That is, net benefits were highest for rules with low political salience promulgated near the end of a presidential administration (Shapiro and Morrall 2012). In the other study, we found that the factor that best correlated with the amount of information in an RIA was the time spent developing the RIA. More time spent both by agencies producing RIAs and by the Office of Information and Regulatory Affairs (OIRA) reviewing

Reforming the Regulatory Process to Consider Employment

227

RIAs meant the analysis contained more information, although the direction of causality, if any, is unclear (Shapiro and Morrall forthcoming). In this section, I build on the data used in my two papers with Morrall. The data set I rely on here consists of 56 regulations and their supporting RIAs. The data set comprises all economically significant regulations (as defined by Executive Order 12866) described in the most recent three OMB Reports to Congress on the Costs and Benefits of Regulations (2008–2010) that were covered in our previous work.4 For each of these regulations, we had collected basic information on the rule and the RIA (net benefits, date of promulgation, agency). For this chapter, I went back to the RIAs and added information on whether each analysis contained information on employment, productivity, and innovation effects. I then added to this database all subsequent economically significant rules listed in the 2012 Draft Report to Congress. I gave each RIA one of three scores for its treatment of employment, productivity, and innovation. An RIA received a “0” if the factor was not mentioned at all, a “1” if the factor was mentioned but not quantified, and a “2” if the measure was quantified.5 I did not attempt to assess the role that each factor played in the selection of each regulatory alternative, as I agree with Masur and Posner’s (2012) conclusion that there is little (if any) evidence that agencies consider macroeconomic impacts in their selection of a regulatory course of action.6 As Table 12.1 shows, employment impacts were quantified in fewer than 20 percent of cases, and other macroeconomic impacts were rarely quantified at all. Given the focus of the rhetorical debate on the effect of regulation on jobs, it is not surprising that agencies attempted to quantify employment effects more often than impacts on productivity and innovation. However, even for employment, fewer than 20 percent of the RIAs provided an estimate of the impact of the regulation. For both employment and innovation, a majority of RIAs did not even mention the fact that the regulation might have such an impact. For productivity, a bare majority did so, but many analyses merely consisted of assertions that the Table 12.1. Agency Consideration of Macroeconomic Effects in RIAs

RIA Scoring

Employment

Productivity

Innovation

0 1 2

33 12 11

27 23 6

40 16 0

Source: Compiled from 56 Regulatory Impact Analyses (RIAs) produced by agencies for economically significant rule makings listed in OMB reports to Congress as noted in the text.

228

Stuart Shapiro

regulations would have no effect on productivity or would improve productivity by improving public health. Looking at the eleven regulations for which employment impacts were quantified, it is clear that even in these regulations there are significant issues with the estimates and their use. •

•

•

•

•

Seven of the 11 RIAs were from the Department of Energy. DOE assesses the effect on the overall economy and on the affected industry. In most of its analyses, DOE asserts that because the affected industry is capital- intensive, a shift of consumer spending to other industries will result in increases in employment. Masur and Posner (2012:597) describe the DOE approach as rife with problems largely because DOE fails to examine whether consumer spending will actually shift to more labor-intensive industries. Regardless of the quality of the analysis, it does not play an obvious or well- defined role in DOE policy decisions, as the department says it takes employment impacts into account without monetizing the economic value of such impacts (U.S. Department of Energy 2007). In its rule to require warnings on cigarette packages, the U.S. Food and Drug Administration (FDA) argued that losses of employment in the cigarette industry would be offset by employment gains due to improvements in public health (FDA 2011). But the employment effects were not included in the assessment of costs and benefits by FDA. EPA issued its Clean Air Interstate Rule in 2011 and relied on Morgenstern et al. (2002) in assessing the rule’s employment impacts. EPA calculated the net employment effect as somewhere in between a loss of 1,000 jobs and a gain of 3,000 jobs (EPA 2011). These impacts were small compared with the multibillion- dollar environmental benefits calculated by EPA, so presumably did not impact the agency’s ultimate decision. The other EPA regulation in my data set was the Corporate Average Fuel Economy (CAFE) rule issued jointly with the National Highway Traffic Safety Administration (NHTSA) within the Department of Transportation (NHTSA 2009). Masur and Posner (2012) cover the assessment of employment impacts in this rule extensively. They point out that EPA and DOT, while estimating the job impacts of various alternatives, did not monetize the costs of job loss or give any indication of why some amounts of job loss are tolerable and others would not be. The final rule in my dataset where employment impacts were quantified was a U.S. Department of Homeland Security (DHS) regulation implementing its Western Hemisphere Travel Initiative (DHS 2008).

Reforming the Regulatory Process to Consider Employment

229

Although DHS showed that the effect on jobs would be lessened under the department’s preferred alternative, this did not appear to have driven DHS officials’ decision making about their choice of regulatory alternative. In none of these 11 highest- scoring rules did a calculated jobs impact appear to play a role in the regulatory decision or in the estimation of net benefits. In addition, there are many legitimate questions about the quality of the calculation of the impacts on employment. But again, these are the only analyses that even attempt to quantify employment impacts. It is clear that under the current institutional framework, despite an executive order requirement, job impact analyses are rarely if ever pursued with any analytical rigor, and formal estimates of job impacts are playing virtually no role in regulatory decisions. A further question might be asked about the 11 rules with an RIA containing an estimate of employment impacts: what, if anything, distinguished them from the other 45 rules?7 The factor most strongly correlated with whether an agency quantified employment effects was the identity of the agency. As noted above, seven of the 11 rules were from DOE and involved the setting of energy efficiency standards.8 The DOE model used to examine the economic effects of its energy efficiency standards included an assessment of employment effects. Because DOE rules comprised such a high proportion of the regulations that evaluated employment impacts at all, it is difficult to examine other determinants of quantification of employment impacts. For example, there is a statistically significant difference (using a one- sided t test at the 5 percent level of significance) between the probability that an RIA quantified employment effects and the probability that the RIA was required by the Unfunded Mandates Reform Act (UMRA) as opposed to being merely required by Executive Order 12866. However, this is because all of the DOE rules fell under the UMRA requirements. More interesting are the factors that were not correlated with the quantification of employment effects. Rules with higher total costs were no more likely to have RIAs with quantified employment impacts than were rules with lower costs. Indeed, of the six rules with estimated costs of more than $1 billion, only one even mentioned the possible impact on employment. Political factors also had no correlation with treatment of employment—neither the volume of public comments nor the regulation’s status as a so- called midnight rule was related to the agency’s decision to examine regulatory impacts on employment. The decision to quantify employment impacts appears to be ad hoc at best. Moreover, with the possible exception of the CAFE standards, it appears that no

230

Stuart Shapiro

agency took employment impacts into account in selecting among several regulatory alternatives.

So What Do We Do? Agencies are clearly not giving much consideration to the employment (or other macroeconomic) effects of their regulatory decisions. Assuming for the moment that they should, there is an obvious solution to this problem: Congress could simply require agencies to do more detailed analysis of employment or other effects as part of their RIAs. Yet as H. L. Mencken famously observed, for every problem there is a solution that is simple, neat, and wrong. A congressional requirement for enhanced regulatory analysis of this type would certainly be just such a solution. Agencies’ existing analyses of job impacts show why adding a statutory requirement for jobs analyses would be misguided. Despite already being required by executive order to consider the effects of regulation on employment and other macroeconomic factors, agencies often do not do so, and when they do, those considerations seem to have no impact on their regulatory decisions. This does not necessarily reflect poorly on agencies; it is merely rational behavior to confi ne oneself to qualitative analysis when required to estimate impacts that are surrounded by considerable uncertainty. Scholars have frequently commented on how the uncertainty in cost–benefit analysis generally leads to analyses that are more often used to justify policy decisions that are already made, rather than to inform those decisions (Wagner 2009). And given the very different conclusions reached in the empirical assessments of employment impacts of regulation (ranging from large net decreases in jobs to large net increases therein), the uncertainty in measur ing job impacts is likely greater than the uncertainty in mea sur ing compliance costs and possibly public health benefits. It is tempting to argue that a requirement embodied in legislation would be more likely to be obeyed by agencies. However, as one practitioner has noted, requiring job impact statements in regulations would condemn consideration of job impacts to “stumble into their dotage in the Federal Register on the concluding pages of rules as humiliated, featureless, grey boilerplate” (Arbuckle 2012). Indeed, numerous studies have noted the limited salutary impact of statutory requirements to consider the impact of regulations on small businesses (Batkins and Brannon 2012), on states and localities (Sharkey 2012), on the collection of information (Shapiro 2013), and on other particular entities or outcomes (Wagner 2009). There is every reason to believe that a statutory requirement to consider employment impacts would be treated similarly. As other chapters

Reforming the Regulatory Process to Consider Employment

231

in this volume note, there is considerable uncertainty about the macroeconomic effects of regulations (e.g., Aldy and Pizer this volume; Morgenstern this volume). A variety of well- constructed studies show positive effects while others show negative effects on employment levels and on other macroeconomic variables. Given this level of uncertainty, it seems unreasonable to require agencies to assess macroeconomic impacts of individual rules. If some of the best economists in the country have been unable to resolve this question, agency economists with limited resources and limited time are unlikely to be able to produce results that will assist in regulatory decision making. Furthermore, all the existing studies of the employment effects of regulations exhibit at least one of two qualities. They are either ex post examinations of regulatory impacts, or they are ex ante estimates of an entire category of regulations, such as regulations to address climate change (see Ho et al. 2008). In other words, none are ex ante estimates of the impact of an individual regulation. A new requirement that agencies do job impact analyses of individual regulations would require agencies to do the types of analysis that virtually no one else has done— or at least done well. Imposing a statutory requirement is a recipe for results that have little support and are easily manipulable to support the regulatory option already chosen by the agency. It is the barrage of new regulations that is most often cited in political rhetoric as destroying jobs.9 Any mea sure to study job impacts should therefore focus on this barrage rather than requiring agencies to conduct analyses of specific, individual rules that are likely to amount to little more than mere guesswork. Masur and Posner (2012) made a convincing case that the effects of unemployment can sometimes be serious enough to merit seeking to understand better the impacts of regulation on employment. But the question of macroeconomic impacts mirrors many other questions that bedev il the understanding of regulations. Ex post evaluation of costs and benefits has long been called for by scholars and advocates (Coglianese and Bennear 2005; Greenstone 2009), but despite President Obama’s Executive Order 13563, it has never had a central institutional home and consistent presence in the federal government. Distributional consequences of regulations are significantly understudied despite numerous proposals to better understand them (Adler 2012; Farrow 2011). Indeed, the employment impacts of a regulation could be seen as a subset of better understanding the distributional consequences of that regulation.10 All of these questions are too important to be treated summarily by agencies pursuant to a requirement that they be considered as part of the regulatory process. In contrast, a better idea would be to create a new office of regulatory analysis that would be devoted to the study of employment impacts

232

Stuart Shapiro

of regulation and to other questions that are better separated from the political nature of regulatory review. Promulgating regulation is as much (or more) a political process as a policymaking process (Croley 1998). The most challenging empirical and analytical questions are better addressed away from the glare of politics. A new office would not be in lieu of OIRA but a complement to it. Nor would it be a replacement for current requirements that agencies conduct RIAs. Although there have been calls for moving the regulatory review and regulatory analysis processes out of the executive branch (or duplicating it elsewhere) from members of both the academy (Greenstone 2009) and Congress (REINS Act 2011), Livermore and Revesz (forthcoming) outline the practical difficulties with such a proposal. They go on to argue that a neutral body would be better suited to conducting retrospective reviews of regulation with the goal of informing the public and Congress on the efficiency of the regulatory system. Building on this suggestion, I propose that both employment impacts and the distributional consequences of regulation could also be studied by such an office that conducts retrospective reviews. Such an office would ideally be created either as an independent agency, an organization with public and private funding such as the Health Effects Institute, or a congressional agency, such as the Congressional Budget Office or the Government Accountability Office, with a strong emphasis on its independence. This new agency could evaluate the macroeconomic impacts and distributional consequences of regulation in broad policy areas rather than (or in addition to) individual regulations. Examining the effect of a set of regulations— such as all energy efficiency standards, all regulations that address climate change, or all regulations by a particular agency such as the Occupational Safety and Health Administration or FDA—is likely to be a useful enterprise. Looking at broader policy areas rather than individual regulations is the approach taken by many of the better academic studies to date, and doing so obviously provides a better understanding of the cumulative costs and benefits of regulation. Tasking an independent agency with periodically studying regulatory policy impacts on employment and other variables would also allow for the longer-term development of standards for studies such as codifying an approach to estimating how firms will react to a regulation and whether this reaction will lead to layoffs. The agency could make decisions that could be replicated across policy areas about how to examine the distributional effects of lost employment or increased innovation. It may also be in a position to offer advice about best practices to minimize the negative employment impacts of a regulation or set of regulations.

Reforming the Regulatory Process to Consider Employment

233

Finally, although this chapter has focused on the impacts of regulation on the larger economy, any outside or independent agency retrospectively analyzing multiple regulations in a policy area would also need to include an analysis of the benefits as well as the costs. Revesz and Livermore (2008) have criticized the proportionately greater focus on costs in RIAs and have argued that if an outside independent agency is going to examine a set of regulations addressing (say) climate change, it should predict the aggregate effect of these regulations. In addition to better evaluating the collection of regulations, doing so would help identify overlapping requirements where the same benefit is achieved by two separate regulations or when one regulation thwarts a benefit believed to have been achieved by a different regulation. There are, of course, disadvantages to this approach. Moving any type of analysis of regulations outside of agencies means that agency expertise will not be applied to the analysis to the same extent that it normally would inside the agency (although analysts working for the new agency would still be able to consult with agency officials). Also, divorcing evaluation questions from the agency that produces RIAs and issues regulations may make it less likely that these analyses will be used to make regulatory decisions and choose between policy options. However, based on the research above and the years of experience with required ex ante analyses, it is hard to argue that keeping ex post review within the lead agency will ensure that it influences policy decisions either. The advantages of having an independent agency tasked with complex analytical tasks outweigh the disadvantages. An independent assessment of the effect of agency regulations on the economy, done in a manner divorced from the decision to promulgate a particular regulation, will inform various parties in the regulatory process. Congress and the president can use the results to direct agencies toward or away from particular regulations through statutory changes or budgetary policy. Agencies can better base their regulatory decisions on the effects of previous regulations, both in improving public health and in affecting employment levels and other economic factors. Interest groups and the public will be able to use the information to better assess regulatory policy and to inform their input to pending regulatory decisions. Realistically, the improved knowledge about the relationship between regulation and the broader economy that would arise from this proposal for independent, broad-based regulatory evaluations would by necessity be incremental. Academic uncertainty surrounding the job effects and other impacts of regulation will not be resolved overnight. No agency, no matter how well funded, independent, or talented, will be able to produce definitive responses to these questions in the short run. But more

234

Stuart Shapiro

attention to evaluation research applied in a systematic way has the potential to gradually improve our regulatory decision making.

Conclusion In difficult economic times, attention naturally focuses on everything that affects the state of the economy. In the stagflation era of the late 1970s, Ronald Reagan focused voters’ attention on the role of government regulation in harming the greater economy. The result was the first president who arguably received an electoral mandate to scale back regulation. However, in the decade that followed Reagan’s presidency, attention to the relationship between regulation and macroeconomic variables faded as the economy boomed. With the Great Recession in 2008, political rhetoric returned to the impact of regulation on the economy, particularly on employment. Republican presidential candidates and members of Congress have blamed regulation for a slow recovery from the recession. Today much more research exists on the relationship between the economy and regulation (particularly environmental regulation) than existed during the Reagan era. And the research supports conclusions less strident than those reflected in political rhetoric. Some studies have shown significant negative effects of regulation on employment levels, while others have found ambiguous and even positive effects. This ambiguity makes it difficult to recommend a change in the way we write regulations. Requiring agencies to focus more on macroeconomic impacts of their regulatory decisions is likely to produce few changes in the actual decisions. Current practice shows that even with an executive order in place that requires agencies to consider job impacts, such analysis occurs in only an ad hoc and cursory manner. Strengthening the requirement, such as by embedding it in legislation, may increase the frequency with which agencies purport to conduct jobs impact analysis but is unlikely to increase its quality. Given the varying conclusions in the academic literature, agencies will likely be able to find support for both regulatory and deregulatory alternatives. At the same time, the need to examine regulatory impacts on employment may be clearer now than it has been in the past. The literature indicates that the social costs of unemployment may be greater than even imagined. Research has made it clear that long- term unemployment has impacts on personal well- being (Blanchflower and Oswald 2004) and public health (Sullivan and von Wachter 2009). Many regulations with employment effects are likely to affect workers in manufacturing and extraction industries who may find it hardest to find new positions and therefore will end up among the long-term unemployed.

Reforming the Regulatory Process to Consider Employment

235

In this chapter, I propose a process for studying the job effects of regulation divorced from the promulgation of individual regulations. Infused with politics, the regulatory process within agencies is not a good place for looking at big and challenging empirical questions. Even the quantification of direct benefits and compliance costs for regulations has proven to be a challenge (although one that has been met with some success). In addition, there seems little reason to believe that the individual regulation is the best unit to base an assessment of regulatory impacts on employment, productivity, innovation, and other variables. Even critics of regulation often cite the barrage of regulation as the problem for businesses. If this barrage really exists, then it is the barrage that should be examined rather than the individual regulation. Moreover, by placing responsibility for the assessment of macroeconomic effects within a separate agency with some independence, important analytical questions can be studied in an environment more conducive to the production of meaningful answers. Divorced from the transactional nature of regulatory politics, the agency charged with producing retrospective evaluations will feel less pressure to produce particular conclusions. The findings from such an independent agency’s evaluation research will also not provide a vehicle for challenging or holding up the development of individual regulations. Questions about the effect of regulations on employment are undoubtedly important. But they are not trump cards to be used to stop the regulatory process in its tracks.

Acknowledgments I would like to thank Anne Gowen, Don Arbuckle, Arden Rowell, Adam Finkel, Joseph Seneca, Chris Carrigan, Richard Morgenstern, Cary Coglianese, and William Rodgers for helpful comments on drafts of this chapter and Anneliese Haines for research assistance. All errors are my own.

Notes 1. In the December 10, 2011, Republican presidential debate, each of the candidates listed eliminating burdensome regulations as one of the steps to take to improve the economy. 2. In a 2011 op- ed in the Wall Street Journal, President Obama said that sometimes government “rules have gotten out of balance, placing unreasonable burdens on business—burdens that have stifled innovation and have had a chilling effect on growth and jobs” (Obama 2011c). 3. It should be noted that some agencies (including EPA for some rules issued under the Clean Air Act) are restricted in the extent to which economic impacts, including those on employment, can influence their regulatory decisions. 4. We collected information on economically significant rules that were supported by RIAs that included quantified benefits and costs.

236

Stuart Shapiro

5. As such, this is not a mea sure of quality like those in the works cited above. It is more of a modified “on/off” mea sure of whether these impacts were mentioned at all, either qualitatively or with some numerical estimate. 6. Masur and Posner (2012) do not assess the accuracy of the estimates per se, but their discussion of the limited justification for the job impacts indicates that such estimates are likely not replicable. 7. I also looked at regulations that received a combined score of three (or higher) across the three macroeconomic variables, but the analysis was virtually identical to the one described in the text for employment (with one exception, described in the next note). 8. Only one difference of note emerges when one examines rules that scored at least three points when adding the scores for each macroeconomic variable: the Department of Energy’s dominance is slightly less stark. Six of the 13 rules with three total points or higher were from DOE. The others were split across four agencies (DOT, EPA, the Department of Health and Human Ser vices, and DHS). 9. See, e.g., Representative Eric Cantor (2012), Representative Randy Forbes (2011), and Senator Jim DeMint (Herszenhorn 2010). 10. Labor market scholars have focused on the distributional consequences of policies that impact employment (see Walker 2012).

References Adler, Matthew (2012) Well Being and Fair Distribution: Beyond Cost– Benefit Analysis. New York: Oxford University Press. Arbuckle, Donald R. (2012) “Regulation’s Impact on Jobs,” RegBlog (26 November), https://www.law.upenn.edu/blogs/regblog/2012/11/26- arbuckle - jobs.html (accessed 23 January 2013). Batkins, Sam, and Ike Brannon (2012) “Small Business Regulation: A Case Study and Options for Reform,” Regulation 35: 6–7. Blanchflower, David G., and Andrew J. Oswald (2004) “Well- Being over Time in Britain and the USA,” Journal of Public Economics 88: 1359– 86. Cantor, Eric (2012) “Red Tape Is Stifl ing Job Creation,” U.S. News and World Reports, http://www.usnews.com/opinion/articles/2012/07/26/eric- cantor -red-tape-is- strangling- job- creation (accessed 23 January 2013). Coglianese, Cary, and Lori Bennear (2005) “Program Evaluation of Environmental Policies: Toward Evidence- Based Decision Making,” in G. D Brewer and P. C. Stern, eds., Decision Making for the Environment: Social and Behavioral Science Research Priorities. Washington, DC: National Academies Press. Croley, Steven (1998) “Theories of Regulation: Incorporating the Administrative Process,” Columbia Law Review 98: 1–168. Ellig, Jerry, and Patrick McLaughlin (2011) “The Quality and Use of Regulatory Analysis in 2008,” 32 Risk Analysis 855– 80. Farrow, Scott (2011) “Incorporating Equity in Regulatory and Benefit– Cost Analysis Using Risk Based Preferences,” Risk Analysis 31: 902–7. Forbes, J. Randy (2011) “Relieving Regulatory Burdens on Small Businesses,” Randy’s Blog, http://forbes.house.gov/news/documentsingle.aspx? DocumentID =240904 (12 May) (accessed 23 January 2013). Greenstone, Michael (2009) “Toward a Culture of Persistent Regulatory Experimentation and Evaluation,” in D. Moss and J. Cisternino, eds., New Perspectives on Regulation. Cambridge, MA: The Tobin Project.

Reforming the Regulatory Process to Consider Employment

237

Hahn, Robert W., and Patrick M. Dudley (2007) “How Well Does the U.S. Government Do Benefit– Cost Analysis?,” Review of Environmental Economics and Policy 1: 192–211. Hahn, Robert W., and Paul C. Tetlock (2008) “Has Economic Analysis Improved Regulatory Decisions?” Journal of Economic Perspectives 22: 67– 84. Hargreaves, Steve (2011) “Obama Backs off Tough Clean Air Regulation,” CNN Money, http://money.cnn.com/2011/09/02/news/economy/regulations/in dex.htm (accessed 23 January 2013). Herszenhorn, David (2010) “DeMint Wants Law to Rein in Regulations,” New York Times, http://thecaucus.blogs.nytimes.com/2010/09/22/demint-wantslaw-to-rein-in-regulations/ (accessed 23 January 2013). Ho, Mun S., et al. (2008) “Impact of Carbon Price Policies on U.S. Industry,” Resources for the Future Discussion Paper No. DP 08- 37, Washington, DC. Livermore, Michael, et al. (2012) “The Regulatory Red Herring: The Role of Job Impact Analyses in Environmental Policy Debates.” Institute for Policy Integrity, New York. Livermore, Michael, and Richard Revesz (forthcoming) “Regulatory Review Capture and Agency Inaction,” Georgetown Law Journal. Masur, Jonathan S., and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. McLaughlin, Patrick, et al. (2012) “Continuity Change and Priorities: The Quality and Use of Regulatory Analysis of Regulatory Analysis Across U.S. Administrations,” Regulation and Governance (13 Aug), http://onlinelibrary .wiley.com/doi/10.1111/j.1748- 5991.2012.01149.x/pdf (accessed 23 January 2013). Morgenstern, Richard D., et al. (2002) “Jobs Versus the Environment: An Industry- Level Perspective,” Journal of Environmental Economics and Management 43: 412–36. Obama, Barack (2011a) “Executive Order 13563: Improving Regulation and Regulatory Review,” the White House, http://www.whitehouse.gov/thepress- office/2011/01/18/improving- regulation- and- regulatory- review - executive- order (accessed 23 January 2013). ———. (2011b) “Memorandum of January 18, 2011: Regulatory Flexibility, Small Business, and Job Creation,” Federal Register 76: 3827–29. ———. (2011c) “Toward a 21st Century Regulatory System,” Wall Street Journal, 18 Jan., sec. A, p. 17. ———. (2012) “Executive Order 13610: Identifying and Reducing Regulatory Burdens,” the White House, http://www.whitehouse.gov/the- press- office /2012/05/10/executive- order-identifying- and-reducing-regulatory- burdens (accessed 23 January 2013). Revesz, Richard, and Michael Livermore (2008) Retaking Rationality: How Cost– Benefit Analysis Can Better Protect the Environment and Our Health. New York: Oxford University Press. Shapiro, Stuart (2013) “The Paperwork Reduction Act: Benefits, Costs, and Directions for Reform,” Government Information Quarterly, http://www.science direct.com/science/article/pii/S0740624X13000087 (accessed 8 March 2013). Shapiro, Stuart, and John Morrall (2012) “The Triumph of Regulatory Politics: Benefit– Cost Analysis and Political Salience,” Regulation and Governance 6: 189–206.

238

Stuart Shapiro

Shapiro, Stuart, and John Morrall (forthcoming) “Does Haste Make Waste? How Long Does It Take to Do a Good Regulatory Impact Analysis?” Administration and Society. Sharkey, Catherine (2012) “Inside Agency Preemption,” Michigan Law Review 110: 521– 96. Sullivan, Daniel, and Till von Wachter (2009) “Job Displacement and Mortality: An Analysis Using Administrative Data,” Quarterly Journal of Economics 124: 1265–1306. Sunstein, Cass (2011) “Letter to EPA Administrator Jackson,” Executive Office of the President, http://www.whitehouse.gov/sites/default/fi les/ozone _ national_ambient_air_quality_standards_letter.pdf (accessed 23 January 2013). Wagner, Wendy (2009) “The CAIR RIA: Advocacy Dressed up as Policy Analysis,” in W. Harrington et al., eds., Reforming Regulatory Impact Analysis. Washington, DC: Resources for the Future Press. Walker, Reed (2012) “The Transitional Costs of Sectoral Reallocation: Evidence from the Clean Air Act and the Workforce.” U.S. Census Bureau Center for Economic Studies Paper No. CES-WP- 12- 02, Washington, DC.

Legislation and Regulations Cited National Highway Traffic Safety Administration (2009) “Average Fuel Economy Standards Passenger Cars and Light Trucks Model Year 2011,” Federal Register 74: 14196– 456. REINS Act, H.R. 10, 112th Congress (2011). U.S. Department of Energy (2007) “Energy Conservation Standards for Residential Furnaces and Boilers,” Federal Register 72: 65136–70. U.S. Department of Homeland Security (2008) “Documents Required for Travelers Departing from or Arriving in the United States at Sea and Land Ports- of- Entry from Within the Western Hemi sphere,” Federal Register 73: 18384– 420. U.S. Environmental Protection Agency (2011) “Federal Implementation Plans: Interstate Transport of Fine Particulate Matter and Ozone and Correction of SIP Approvals,” Federal Register 76: 48208– 483. U.S. Food and Drug Administration (2011) “Required Warnings for Cigarette Packages and Advertisements,” Federal Register 76: 36628–777.

Chapter 13

Analysis to Inform Public Discourse on Jobs and Regulation Michael A. Livermore and Jason A. Schwartz

Despite the fact that job impact analysis poses steep challenges and is unlikely to substantially alter most regulatory choices, there are good reasons to integrate employment effects into cost– benefit analysis of federal rule makings. Cost–benefit analysis not only offers government decision makers a technocratic tool to identify efficient policies; it also informs public debate about the social effects of regulation. This broader deliberative purpose justifies committing federal agencies’ analytical resources to investigate employment effects. Cost–benefit analysis is often a complex and time- consuming task: it can involve engineering studies and technology forecasts, detailed scientific models and dose-response curves, and careful economic surveys to value health or environmental gains. As several authors in this book have noted, broadening the scope of cost–benefit analysis to include job effects will only multiply the analytical challenges. Devoting scarce resources to this daunting enterprise demands strong justification. Cost–benefit analysis customarily concentrates on those regulatory effects that could tip the scales for or against a policy alternative, by changing its comparative efficiency (U.S. Office of Management and Budget 2003:26). Yet employment effects may seldom be significant enough to alter any given rule’s efficiency calculus. The most sophisticated empirical assessments to date typically calculate employment impacts as a small fraction of overall regulatory costs and benefits. Except for rare cases where a rule’s marginal net benefits are very modest, regulatory efficiency may not be a good reason for agencies to quantify and monetize job impacts.

240

Michael A. Livermore and Jason A. Schwartz

In this chapter, we argue that, nevertheless, informing the public debate may justify the substantial effort required for job impact analysis. By placing employment effects in context, highlighting distributional impacts, and clarifying rules’ cumulative effects on employment, job impact analysis can contribute to a more productive democratic conversation about regulatory policy. First, job impact analysis can benefit public debates by adding realism and empirical underpinnings to estimates of employment impacts and by contextualizing those effects within a rule’s broader costs and benefits. In the wake of the 2008 financial crisis and 2010 midterm elections, which returned divided government to Washington, DC, the public dialogue on environmental regulations has become clouded with rhetoric and the misleading use of empirical studies. Monetizing job impacts may provide clarity, by enabling the public to assess a rule’s net social benefits in light of employment effects. To maximize this positive influence on the public debate, agencies should follow best practices for transparent, accessible analysis. Second, job impact analysis can highlight salient distributional consequences. Even if a rule’s job gains and losses cancel out, leaving no discernible efficiency impact, certain communities may experience disproportionate employment burdens. Although fully incorporating such effects into cost–benefit analysis raises weighty methodological and ethical questions, job impact analysis can at least identify and quantify these effects. Finally, job impact analysis can clarify rules’ cumulative effects on employment. Even if an individual rule’s net job impact is too minor to alter its efficiency, cumulatively a whole suite of regulatory policies may generate significant employment effects. Some advocates calling for moratoria on rule makings have pointed to cumulative employment effects for support, especially following the recent recession. Analysis could quantify and monetize cumulative employment effects—if any— and compare them with the cumulative net benefits of regulation, to help determine whether such moratoria are justified. For agencies, the prospect of devoting considerable time to a difficult task unlikely to significantly affect policy outcomes may seem wasteful. Job impact analysis will have opportunity costs by diverting agency resources from other vital analytical questions. Yet analysts must not overlook their work’s considerable benefits to the public debate. Administrative agencies are ultimately responsible to the American people. Even if job impact analysis is not necessary to identify actions that efficiently promote the public interest, such analysis may be necessary to justify regulatory choices to the public—a task that is essential in a democratic society.

Analysis to Inform Public Discourse on Jobs and Regulation

241

Employment Considerations May Not Typically Affect Rules’ Efficiency Both the academic literature and recent examples from agencies’ impact analyses suggest that, for many rules—perhaps the vast majority—the efficiency effects associated with changes in employment are markedly smaller than other regulatory impacts, such as public health improvements or compliance costs. Quantifying and monetizing job impacts, therefore, may contribute little to the goal of maximizing regulatory efficiency. In theory, employment can have an important effect on efficiency. Under neoclassical economic assumptions, a dynamic labor market means that displaced workers will move from one firm or sector to another in response to job openings, and wages will adjust to restore employment levels. Employment transitions seemingly produce no lasting social costs or benefits; such shifts instead appear distributional, moving labor between industries and regions. But labor markets do not always operate so smoothly. If workers have difficulty finding replacement employment, the social costs could be considerable, entailing relocation and training expenses, loss of skills and lifetime wages, and even psychological and health impacts; short-term and long-term unemployment may present very different costs (Institute for Policy Integrity 2012:5– 9). These transition effects can be monetized using standard cost–benefit methodologies (Masur and Posner 2012). By the same token, regulations could spur transition benefits, especially during recessions, by promoting the hiring of underutilized workers: underutilized workers have lower opportunity costs than employed workers, meaning the social cost of using that labor is substantially below the wage; analyses can reflect this transition benefit by adjusting compliance cost estimates downward. Rehiring underutilized workers could also generate other social benefits as workers regain skills and see increased lifetime wages. Masur and Posner (this volume) believe regulatory efficiency justifies including employment in cost–benefit analyses. They identify one example where including the social costs of job losses could have counseled an agency to opt for a less stringent rule. The U.S. Environmental Protection Agency (EPA) forecasted that regulating chlorinated compounds in the paper industry would generate $159.5 million in lifetime net benefits but would also cause 5,711 lost jobs; Masur and Posner’s (2012) estimate of unemployment costs ($3,000 per worker per year) would have erased those net benefits. A less stringent rule or an alternative regulatory design that would have fewer jobs consequences would seem to be justified in light of this analysis.

242

Michael A. Livermore and Jason A. Schwartz

But that example may be anomalous. Many empirical studies raise serious questions about whether regulations will usually affect employment to a degree that would justify substantial changes in rule design. For example, Gray and Shadbegian (this volume) suggest that a calculation of net economy-wide employment effects would likely cancel itself out; Aldy and Pizer (this volume) estimate the employment impact of one major environmental regulation on energy-intensive manufacturing as less than 1 percent, an effect “likely to be swamped by other drivers.” Other studies that ambitiously try to monetize job impacts have found comparatively small effects: Walker (2011:28–29) estimated that the 1990 Clean Air Act Amendments may have generated $9 billion in unemployment-related costs, but these one-time transition costs are small compared with total net benefits as high as $1.6 trillion. After surveying the body of literature to date, Morgenstern (this volume) concludes “there is only limited evidence of adverse national-level impacts.” Where EPA has analyzed job impacts, the message has been that employment effects are not substantial enough to alter regulatory effi ciency. Following its own Guidelines for Preparing Economic Analyses, EPA restricts job impacts to its distributional analyses, because transitional employment costs usually “would not appreciably alter the quantitative conclusions” of its cost– benefit analyses (U.S. Environmental Protection Agency 2000:121–22; see also U.S. Environmental Protection Agency 2010:8- 6). Three recent, controversial rules provide useful illustrations: •

•

Boiler MACT Rule: In 2010, EPA proposed controls for hazardous air pollutants, such as mercury, from industrial boilers. EPA estimated the rule would generate between $25.2 and $65.5 billion in annual net benefits, including up to 8,000 premature deaths avoided per year. By comparison, the agency estimated a net employment effect on the regulated industry of between −4,000 and +8,300 jobs, with a central estimate of +2,100 (U.S. Environmental Protection Agency 2011c). Cross- State Air Pollution (Transport) Rule: In 2011, EPA finalized a regulation to curb upwind sources from impairing air quality in downwind states. EPA estimated the rule would generate between $120 and $280 billion in annual net benefits, including up to 34,000 premature deaths avoided per year. By comparison, the agency estimated a onetime increase of +2,230 compliance-related job years; plus an annual net effect on the regulated industry of between −1,000 and +3,000 jobs, with a central estimate of +700 (U.S. Environmental Protection Agency 2011a). (In 2012, the U.S. Court of Appeals for the D.C. Circuit overturned the rule on unrelated grounds.)

Analysis to Inform Public Discourse on Jobs and Regulation •

243

Mercury and Air Toxics (MATS) Rule: In 2012, EPA finalized a regulation of hazardous air pollutants from utilities. EPA estimated the rule would generate between $27 and $80 billion in annual net benefits, including up to 11,000 premature deaths avoided per year. By comparison, the agency estimated a onetime increase of +46,000 compliance-related job years; plus an annual net effect on the regulated industry of between −15,000 and +30,000 jobs, with a central estimate of +8,000 (U.S. Environmental Protection Agency 2011b). (As of this writing, the rule is being litigated in the D.C. Circuit.)

Given these rules’ large net benefits, EPA easily could have concluded that anticipated employment effects simply seemed “too small to warrant further analysis” (U.S. Environmental Protection Agency 2000:169). For example, even EPA’s worst- case estimate for the Boiler MACT Rule, a potential loss of 4,000 jobs, is small in the context of a regulation with a minimum of $25.2 billion in annual net benefits. Only if each job lost entailed on average $6.3 million per year in transitional efficiency costs (a number that is not far off from the value of each statistical life) would the rule no longer be cost–benefit justified. Even using industry estimates of employment consequences would not radically change this calculus. For example, the American Coalition for Clean Coal Electricity (ACCCE) (2011) published an extreme prediction that the Transport and MATS Rules together would terminate 1.4 million job years over the course of seven years— an implausible forecast, as discussed below. Nevertheless, EPA puts the combined annual net benefits for the two rules between $147 and $360 billion. Multiplying the low end of that range over seven years roughly approximates minimum net benefits: $1.03 trillion. Only if each of ACCCE’s estimated lost job years entailed an average of $735,000 in costs would the rules no longer be cost–benefit justified. For comparison, Masur and Posner’s (2012) preliminary estimate of costs per lost job year is $3,000. These examples reasonably suggest that employment effects—in the broader context of other regulatory costs and benefits— are often sufficiently modest not to warrant resource-intensive analysis on efficiency grounds alone.

Employment Effects in the Political Arena The 2010 midterm election elevated a Republican House majority deeply skeptical of EPA’s regulations. Simultaneously, persistent unemployment from the 2008 financial crash continued to plague the country. This mix created a political atmosphere in which competing claims about employment impacts began to dominate environmental policy debates. Industry trade organizations and other opponents of regulation argued that

244

Michael A. Livermore and Jason A. Schwartz

increasing production costs would trigger layoffs; environmental groups and other proponents countered that stronger protections could spark businesses to hire new workers for compliance or in industries producing green substitutes. To bolster their positions, advocates on both sides have promoted economic studies that purport to examine the employment effects of regulations. These “advocacy analyses” are often deployed in the political arena without disclosing underlying assumptions or limitations. As a result, the public debate about jobs and regulation has grown confusing and shrill. By infrequently conducting analyses of employment effects, and particularly by not explicitly monetizing these effects, EPA may have missed an opportunity to correct this confused political rhetoric and reduce the misleading use of outside economic analyses. Keeping job impacts separate from cost–benefit analysis gives the public no guidance on how to compare employment changes against other regulatory effects. Thus, when advocacy analyses generate outsized estimates of job impacts, both proponents and opponents of regulation gravitate toward the numbers that fit their preferred narratives, and they wield those numbers like a trump card in the policy debate. Agencies can help restore rationality to these debates by demonstrating to the public the hallmarks of wellexecuted analysis, as well as how to weigh job impacts against other costs and benefits. Regulatory impact analysis is more than an internal decision tool focused on maximizing efficiency; it is also a disclosure document, capable of enriching public debate by providing clarity to controversial issues. Political Attempts to Link Environmental Protection to Employment Claims that regulations cause unemployment have been a staple of political discourse for decades. Ronald Reagan may have first coined the phrase “ job-killing regulation” as a presidential candidate (Cannon 1980), but the sentiment recently resurfaced with a vengeance: since 2007, “ job-killing regulation” underwent a 174-fold increase in usage in U.S. newspapers, from just four appearances per year to nearly 700 in 2011 (Institute for Policy Integrity 2013a). Environmental regulations bear the brunt of that negative attention. “EPA” has appeared alongside “ job-killing” in 701 recent news articles; labor-regulating agencies clocked only 126 mentions (Institute for Policy Integrity 2013a). Representative Darrell Issa (R-California), chair of the House Committee on Oversight and Government Reform, commissioned a series of reports on “regulatory impediments to job creation,” which asked businesses to identify rules

Analysis to Inform Public Discourse on Jobs and Regulation

245

that threatened employment. Of the 145 “impediments” listed, EPA actions comprised nearly half (U.S. House of Representatives 2011a). During the 112th congressional session, legislators introduced numerous bills to directly or indirectly weaken EPA’s regulatory authority, in the name of job protection (Institute for Policy Integrity 2012). Adding more analyses and audits to the rule-making process became the leading proposal in the House Republican Plan for America’s Job Creators (House Republican Conference 2011). At least 16 bills proposed draconian, across-the-board moratoria on rule making. One bill, the Regulatory Freeze for Jobs Act, sought to block all new significant regulations until unemployment dropped below 6 percent—and it passed the House in July 2012, although it did not progress in the Senate (Institute for Policy Integrity 2013b). Supporters of environmental protections also capitalized on employment arguments to bolster their agenda. When EPA took its first steps to regulate greenhouse gases, Administrator Jackson promised, “this pollution problem has a solution— one that will create millions of green jobs.” At the 2012 Democratic National Convention, former president Clinton praised the potential of new vehicle efficiency standards to “bring us another half a million good new jobs.” And congressional Democrats have cited employment to back new environmental policies on everything from renewable electricity standards to e-waste recycling. All this political attention on employment effects has been fueled by economic reports commissioned by advocacy groups—reports that proliferated in the vacuum left by federal agencies’ relative inattention to job impacts. A few case studies will demonstrate how participants in regulatory debates simply pick the advocacy analysis that justifies the position they already support and present those job impact estimates in definitive, raw numbers, without discussion of assumptions, uncertainties, or limitations. The policy debate over jobs and regulation has become clouded with rhetoric; federal agencies, therefore, should step in with rigorous, balanced, transparent analysis to provide clarity. Case Studies: How Advocates’ Job Analyses Shape Environmental Policy Debates Nearly every recent controversial environmental proposal has ignited a debate over employment effects. Few studies used to support either side rely on validated, well- executed models, and even less frequently do the political debates mention potential limitations. Job impact analysis can enrich the regulatory process: using the right models to report appropriately limited results can help inform public debate and decision

246

Michael A. Livermore and Jason A. Schwartz

making. But the wrong models, used to answer the wrong questions, reported without caveat, only obscure the important trade- offs at stake. Boiler MACT Rule. In August 2010, two industry groups released reports predicting that the Boiler MACT Rule would jeopardize hundreds of thousands of jobs. First, the Council of Industrial Boiler Owners (CIBO) estimated that the rule would put “at risk” upward of 800,000 jobs. “At risk” was never clearly defined, but it did not mean the job would necessarily be eliminated (Council of Industrial Boilermakers 2010). The study did not acknowledge that the rule might simultaneously lead to job gains elsewhere in the economy. The CIBO report used the IMPLAN model, a type of input– output simulation that holds prices constant and tracks goods flowing through the economy: one sector’s output is another’s input (Berck and Hoffman 2002:137). Because of their fi xed- price approach, these simulations cannot reflect long-term, structural economic changes or account for substitutions between goods. Consequently, these models tend to overstate employment effects, and some analysts have advised that “IMPLAN is not readily suitable for forecasting the effects of public policy changes” (Morgan 2010:5; see also U.S. Environmental Protection Agency 2011b:6- 6). The CIBO study never discussed those limitations, conducted no sensitivity analysis, and only partially disclosed assumptions and data. That same month, the American Forest and Paper Association (AFPA) released an analysis it commissioned on how the Boiler MACT Rule might affect the pulp and paper industries (Fisher International 2010). This study reported that nearly 90,000 jobs were “at risk” in mills and through indirect effects. The study simply estimated compliance costs for each mill type and assumed that an increase to a mill’s production costs of 12.5 percent or more would place all of the workers at that mill “at risk.” Both studies received considerable media and congressional attention. Eighty- five percent of news articles mentioning the Boiler MACT Rule also referenced jobs or employment (Institute for Policy Integrity 2013a). The two studies’ predictions were cited repeatedly at congressional hearings and in press releases by political opponents of the rule. In August and September 2010, directly after the release of the reports, over 150 members of Congress signed letters urging EPA to balance the Boiler MACT Rule against job preservation. Later the AFPA study was cited prominently in the House committee report on H.R. 2250, a bill to delay the rule, which passed the House in October 2011 (although it did not advance in the Senate). The results of the two analyses were routinely reported as definitive raw numbers, without discussion of assumptions or limitations.

Analysis to Inform Public Discourse on Jobs and Regulation

247

Although the precise influence of these reports and the surrounding political attention is hard to gauge, EPA did significantly reevaluate its initial proposal. In March 2011, EPA issued a fi nal rule, having scaled back its proposed standards; a few months later, EPA suspended the rule’s effective date, and the agency finalized even further modifications in December 2012. Transport and MATS Rules. To curb upwind sources from impairing air quality in downwind states, EPA proposed its “Transport Rule” in 2010 and finalized it in 2011. EPA also proposed the “Mercury and Air Toxics Rule” in 2011 to regulate hazardous air pollutants from utilities. Together, EPA estimated these two rules would deliver annual net benefits between $147 and $360 billion, including up to 45,000 premature deaths avoided per year (U.S. Environmental Protection Agency 2011a, 2011b). The political debate over these rules pitted two advocacy analyses against each other. A report commissioned by the ACCCE estimated that the two combined rules would trigger a 1.4 million job- year loss, while a Political Economy Research Institute (PERI) study predicted the same two rules would spur a 1.4 million job-year gain (American Coalition for Clean Coal Electricity 2011; Ceres 2011). The PERI report ran IMPLAN, the same input– output model whose limitations were discussed above. The ACCCE report instead used REMI, a computable general equilibrium (CGE) model that allows for fluctuating prices and more complex sectorial interactions. Unfortunately, the main strength of CGE models— complexity—is also their chief disadvantage (Xie and Saltzman 2000). A CGE model comprises multiple equations solved simultaneously: values for key parameters are often left to the analyst’s best guesses, raising concerns about accuracy and transparency (Bandara 1991:31; Berck and Hoffman 2002:146). Neither PERI’s nor ACCCE’s study included any sensitivity analysis. None of those modeling choices or assumptions appeared when the two studies were cited repeatedly in congressional hearings (U.S. House of Representatives 2011b). Utility lobbyists widely circulated the ACCCE study in support of legislation to block both rules. A joint resolution of disapproval for the MATS Rule fell short of passing the Senate by just five votes. Proponents of the rules were just as quick to cite only the PERI study.

Using Jobs Analysis to Inform Public Debates over Jobs and Regulation Regulatory impact analysis is more than an internal decision tool focused on maximizing efficiency; it is also a disclosure document, capable of

248

Michael A. Livermore and Jason A. Schwartz

enriching public debate by providing clarity to controversial issues. OMB’s Circular A-4 posits that “good regulatory analysis is designed to  inform the public” (2003:2). EPA’s own Guidelines echo this sentiment, calling economic analysis an “important resource for the public” (2010:11–12). Economic analyses have shaped several public debates. Cole (2009:13) cites the national ambient air quality standards in the 1990s: even though EPA could not legally rely on cost– benefit analysis in setting the standards, publicizing the regulation’s large net benefits “play[ed] a valuable role in the political process by muting potential opposition [to] welfare- enhancing regulations.” Elliott (this volume) calls the Congressional Budget Office a model of success, where transparent budgetary analyses help advocates and opponents of legislation begin the debate from a common starting place. Finkel (this volume) draws from the world of risk assessment, pointing out that better analysis and quantification of controversial issues such as uncertainty have helped settle some debates. Following these examples, more quantification of job impacts may help provide clarity in the political fights over regulation. Estimating job impacts and integrating them into cost–benefit analysis can offer some clarity in the shrill, politicized debate over environmental regulations and their employment effects. Adding Realism and Context to Public Debates over Jobs and Regulation By providing expert estimates of employment impacts based on the best available models, agencies can restore realism to the conversation, countering the more extravagant claims of advocates. For example, in the debate over the Transport and MATS Rules, industry supporters predicted 1.4 million job-year losses, while environmentalist supporters forecasted 1.4 million job-year gains. EPA’s own analysis showed far more modest— and realistic—impacts, in the tens of thousands of jobs gained or lost. The agency’s analysis serves as a check on more outlandish estimates, and observers who lack technical expertise in employment models— such as journalists and congressional staff— can use the agency’s numbers to guide their interpretation of reports publicized by advocates. Those three strikingly divergent estimates also demonstrate the influence of modeling and data choices on results. All models are subject to limitations. To maintain their credibility for unbiased analysis, agencies must transparently communicate those limitations to the public. Analysts should disclose all assumptions and data sources, describing

Analysis to Inform Public Discourse on Jobs and Regulation

249

how realistic the assumptions are and how complete the data sets are. Reports should acknowledge uncertainty and include a sensitivity analysis to identify how results change under different assumptions and model structures. All final results should clearly indicate any limitations or weaknesses in the model. Policymakers should rely only on studies that meet these criteria, and even then only after fully acknowledging potential limitations. Analysts, politicians, and commentators should all avoid translating the complex model outputs into a single, often very misleading, sound bite about the “ jobs” created or lost by regulation. In addition, by monetizing a rule’s positive or negative job effects, agencies can contextualize those impacts within the rule’s broader economic consequences, helping to clarify their relative importance and reducing their potential to act as a trump against other regulatory effects. Expressing all costs and benefits in terms of a single, comparative metric— dollars—allows the public to place job effects, compliance costs, health gains, and environmental improvements on the same scale and to see the trade- offs implicated by the policy choice. Presenting each regulatory consequence (whether positive or negative) in both monetary terms and natural units, such as employment years, lives saved, or injuries avoided, may be the most transparent way to inform public debate (Keohane 2009:50– 51). Conversely, given how volatile the jobs issue is, incorporating employment impacts could risk politicizing cost–benefit analysis. Instead of leading the public to accept a common set of facts, access to more information may simply magnify people’s tendency to cherry-pick data to fit their beliefs (Silver 2012:4). One study found that greater general knowledge of science increased the likelihood that strong political partisans would disagree about global warming (Kahan et al. 2012). Both Elliott (this volume) and Shapiro (this volume) caution that agencies’ cost–benefit analyses have become advocacy documents, susceptible to manipulation and therefore viewed with increasing skepticism by the public. This danger could be even more pronounced for monetizing job impacts, because the limited methodologies available will make estimates quite uncertain. That risk does not mean that informing the public debate is not worthwhile. But it may counsel in favor of approaches to estimating job impacts that minimize the chance of politicization. Both Elliott (this volume) and Shapiro (this volume) raise the possibility of delegating analysis to some independent entity—like the Bureau of Labor Statistics— that has employment expertise and can generate transparent job estimates consistently across agency proposals. There may be other ways for agencies to increase the transparency and reliability of their job impact analyses without relinquishing the task to some independent entity. For

250

Michael A. Livermore and Jason A. Schwartz

example, simply developing tools to better communicate the job impacts, and particularly the uncertainty around efficiency effects, may be a solution. After all, if the goal of analysis is to inform public debate, precise estimates may be less important than order- of-magnitude judgments coupled with “a more honest [presentation] of the deep uncertainties involved” (Keohane 2009:48). Highlighting Important Distributional Effects Even if regulations do not generate significant employment effects from an efficiency perspective, important distributional considerations may exist. For example, a rule may help affluent customers but harm lowincome workers. So long as aggregate benefits outweigh harms, standard cost–benefit analysis would show such a rule to be efficient. Only subsequent distributional analysis would scrutinize exactly who benefits and who is burdened. By identifying and quantifying distributional effects, agencies can guide the debate over how to mitigate any important, negative distributional consequences. As EPA’s own Guidelines note, “even if total employment does not change, localized changes in employment may interest policymakers” (2000:153). Some scholars propose incorporating distributional impacts directly into cost– benefit analysis, through tools such as equity weighting. Adler (2012) argues that applying the social welfare function to evaluate public policy requires sensitivity to distributional impacts. The United Kingdom’s Green Book on evaluating government programs calls for explicitly weighing costs and benefits more heavily if they accrue to lower-income households and discounting effects on higher-income households (H.M. Trea sury 2011:24). However, these techniques remain controversial. Not only does calculating equity weights entail considerable methodological complexity, but monetizing distributional effects raises thorny normative questions about the appropriate allocation of wealth (Organization for Economic Co- operation and Development 2006:234) and about income’s impact on individualized valuations of costs and benefits (Livermore and Rosenberg 2013). But even without the use of equity weights, agencies can still identify and quantify distributional impacts to inform debate over appropriate policy responses. Altering or revoking rules in response to distributional concerns will rarely be optimal. In general, achieving redistributional goals through tax-and-transfer will often be a more efficient tool than altering regulatory stringency (Kaplow and Shavell 1994). Changing rules in a bid to minimize transition costs may also result in large, unintended efficiency

Analysis to Inform Public Discourse on Jobs and Regulation

251

losses, as “grandfathering” under the Clean Air Act amply demonstrates (Revesz and Westfahl Kong 2011). Careful analysis of both distributional and efficiency effects of policy alternatives could help decision makers avoid such inadvertent outcomes. Furthermore, over the course of many rules, distributional effects may cancel themselves out: those hurt by some rules may benefit under others (Livermore and Rosenberg 2013). On the other hand, if distributional analysis uncovers certain groups consistently on the losing side of policy, steps can and should be taken to improve the fairness of the regulatory process (Livermore and Rosenberg 2013). Notwithstanding these concerns, if regulation generates acute, employment-related harms to small populations, direct compensation through unemployment insurance, job retraining, or economic development funding may be appropriate. Identifying and quantifying distributional effects will help policymakers determine how much mitigation is necessary and where to direct it. Cumulative Effects and Regulatory Moratoria Individual environmental regulations are very unlikely to fundamentally alter labor supply or demand at a national scale. It is somewhat more conceivable, however, that a suite of regulatory policies could together affect aggregate employment. Moreover, existing methodologies for estimating employment impacts may be able to assess policies only with very large, economy-wide effects and may often not be fit to analyze each regulation individually (Berck and Hoffman 2002:146, 154; Institute for Policy Integrity 2012). Cumulative regulatory effects may, therefore, be a more appropriate target of analysis. Recent White House guidance instructs agencies to “consider cumulative effects . . . as part of their analysis of particular rules” (Sunstein 2012). However, a broader assessment of cumulative effects, not attached to any individual rule analysis, may be a more natural approach. If an individual rule analysis uncovered cumulative employment losses, most agencies could respond only by decreasing the stringency of that particular rule. Yet that rule may not be the optimal intervention point. The government possesses numerous tools for a more effective response to unemployment: monetary policy, such as increasing the money supply; fiscal policy, such as increasing government spending; or redistributive tax policy, such as expanding the earned income tax credit. Unless such tools are deployed in a coordinated, government- wide fashion, agencies risk making unnecessary or ineffectual changes to regulatory policy (Masur and Posner 2012). Additionally, some sets of policies

252

Michael A. Livermore and Jason A. Schwartz

may implicate multiple agencies. A coordinated approach to assessing cumulative effects is, therefore, essential, whether that means delegating to some independent agency (Shapiro this volume) or relying on the Office of Information and Regulatory Affairs to coordinate the analytical contributions of multiple agencies. Examining cumulative employment effects will allow government to tailor appropriate policy responses. For example, Walker noted that the 1990 Clean Air Act Amendments included transition assistance: $50 million allocated to aid workers displaced by environmental regulations. However, Walker estimated that the monetized cost of unemployment from those regulations actually totaled $9 billion. Arguably, Walker’s analysis supports his conclusion that the government’s response to cumulative employment effects was “severely underfunded” (2011:29). Examining cumulative effects can further inform the public dialogue about broader regulatory policies, particularly the debate over rule-making moratoria. Most advocacy analyses of employment effects study only one or two rules; more recently, a handful have attempted to measure cumulative effects from a larger number of regulations (see, for example, American Coalition for Clean Coal Electricity 2012). Both types of studies have been cited in support of legislative proposals to institute regulatory moratoria for fi xed periods or until unemployment drops below certain levels. These proposed moratoria are sometimes based on inaccurate assumptions about the business cycle. Although the social costs of individual layoffs may increase during recessions, induced hiring is also more likely to draw from the pool of underutilized workers during such times, which could increase social benefits. In addition, rule-making moratoria typically apply only to rules currently under development; yet agencies require substantial time to finalize rules and even longer to start implementing them (Ferris and McGartland this volume). Regulatory policy changes are a slow and fumbling way to react to recessions; monetary and fiscal policies are much quicker and more precise. Finally, industry has often already anticipated the future requirements of ongoing rule makings (Masur and Posner this volume); a moratorium that upsets the anticipated regulatory timeline may actually exacerbate uncertainty about future costs, which may hinder certain investments (Dixit and Pindyck 1994). By studying cumulative effects, agencies can provide a valuable counterpoint to advocacy analyses. After reviewing information about the cumulative employment effect of regulation, the uncertain employment value of halting the rule-making process, and the net social benefits as-

Analysis to Inform Public Discourse on Jobs and Regulation

253

sociated with proposed regulations, the public can more readily judge the desirability of regulatory moratoria.

Conclusion Even if maximizing regulatory efficiency and rigorously assessing distributional impacts are not sufficient justifications to incorporate job impacts into cost– benefit analysis, helping to inform public debates about regulatory policy may be. Job impact analysis is not an alternative to cost– benefit analysis. Rather, employment effects should become part of cost– benefit analysis, based on traditional economic principles. By translating employment effects into the common metric of dollars, agencies can help the public weigh their importance alongside all other regulatory effects. Distributional and cumulative impacts should be identified as well, to guide policymakers in selecting appropriate responses. All results should be reported with full disclosure of assumptions and limitations: honest communication about uncertainty may be more important than precise estimates. Given how hard job impact analysis is, these recommendations may seem ambitious. But agencies are asked to assess many hard-to-monetize values, such as human lives and climate change damages. Because employment has become so central to the public debates, agencies should view the analytic difficulties as a challenge in their efforts to press forward in developing more rigorous, balanced, and transparent analysis.

References Adler, Matthew (2012) Well- Being and Fair Distribution: Beyond Cost– Benefit Analysis. New York: Oxford University Press. American Coalition for Clean Coal Electricity (2011) “Proposed CATR + MACT,” America’s Power, http://www.americaspower.org/sites/default/files /NERA_CATR_MACT_29.pdf (accessed 24 January 2013). American Coalition for Clean Coal Electricity (2012) “Economic Implications of Recent and Anticipated EPA Regulations Affecting the Electricity Sector,” http:// www.nera.com/nera-files/PUB_ACCCE_1012.pdf (accessed 24 January 2013). Bandara, Jayatilleke (1991) “Computable General Equilibrium Models for Development Policy Analysis in LDCs,” Journal of Economic Surveys 5: 3– 69. Berck, Peter, and Sandra Hoffman (2002) “Assessing the Employment Impacts of Environmental and Natural Resource Policy,” Environment and Resource Economics 22: 133– 56. Cannon, Lou (1980) “Reagan, Ignoring Bush, Assails Carter’s Policies,” Washington Post, 20 May, sec. A, p. 8. Ceres (2011) “New Jobs— Cleaner Air: Employment Effects Under Planned Changes to the EPA’s Air Pollution Rules,” http://www.ceres.org/resources /reports/new- jobs- cleaner- air (accessed 24 January 2013).

254

Michael A. Livermore and Jason A. Schwartz

Cole, Daniel (2009) “Regulatory Cost– Benefit Analysis and Collective Action,” Institute for Policy Integrity Working Paper no. 2009/1, New York. Council of Industrial Boilermakers (2010) “Economic Impact of the Proposed EPA Boiler/Process Heater MACT Rule on Industrial, Commercial, and Institutional Boiler and Process Heater Operators,” http://www.cibo.org /pubs/boilermact_jobsstudy.pdf (accessed 24 January 2013). Dixit, Avinash K., and Robert S. Pindyck (1994) Investment Under Uncertainty. Princeton, NJ: Princeton University Press. Fisher International (2010) “Economic Impact of Pending Air Regulations on the US Pulp and Paper Industry,” American Forest and Paper Association, http://www.bipac.net/mwv/Fisher_Cumulative_Air_Burden_PandP_ Job _ Impact_Summary_August_2010.pdf (accessed 6 March 2013). H.M. Trea sury (2011) The Green Book: Appraisal and Evaluation in Central Government. London: HMSO. House Republican Conference (2011) House Republican Plan for America’s Job Creators, http://www.gop.gov/resources/library/documents/jobs/theplan.pdf (accessed 6 March 2013). Institute for Policy Integrity (2012) “The Regulatory Red Herring: The Role of Job Impact Analyses in Environmental Policy Debates,” http://policyintegrity.org/files/publications/Regulatory_Red_Herring.pdf (accessed 24 January 2013). ———. (2013a) Fact Sheet: Jobs and Regulation in the Media— By the Numbers. New York: Institute for Policy Integrity. ———. (2013b) Fact Sheet: Regulatory Moratoria. New York: Institute for Policy Integrity. Kahan, Dan, et al. (2012) “The Polarizing Impact of Science Literacy and Numeracy on Perceived Climate Change Risks,” 2 Nature Climate Change (27 May), http://www.nature.com/nclimate/journal/v2/n10/full/nclimate1547.html. Kaplow, Louis, and Steven Shavell (1994) “Why the Legal System Is Less Efficient than the Income Tax in Redistributing Income,” Journal of Legal Studies 23: 667– 81. Keohane, Nathaniel O. (2009) “The Technocratic and Democratic Functions of the CAIR Regulatory Analysis,” in W. Harrington et al., eds., Reforming Regulatory Impact Analysis. Washington, DC: Resources for the Future. Livermore, Michael, and Jennifer Rosenberg (2013) “The Shape of Distributional Analysis,” in Michael A. Livermore and Richard L. Revesz, eds., The Globalization of Cost– Benefit Analysis in Environmental Policy. New York: Oxford University Press. Masur, Jonathan, and Eric A. Posner (2012) “Regulation, Unemployment, and Cost– Benefit Analysis,” Virginia Law Review 98: 579– 634. Morgan, Jonathan Q. (2010) “Analyzing the Costs and Benefits of Economic Development Projects,” UNC School of Government Community and Economic Development Bulletin no. 7 (April), http://sogpubs.unc.edu//electronicversions /pdfs/cedb7.pdf (accessed 6 March 2013). Organization for Economic Co- operation and Development (2006) Cost– Benefit Analysis and the Environment: Recent Developments. Paris: OECD Publishing. Revesz, Richard L., and Allison L. Westfahl Kong (2011) “Regulatory Change and Optimal Transition Relief,” Northwestern University Law Review 105: 1581– 1633. Silver, Nate (2012) The Signal and the Noise: Why So Many Predictions Fail— But Some Don’t. New York: Penguin Press.

Analysis to Inform Public Discourse on Jobs and Regulation

255

Sunstein, Cass (2012) “OIRA Memorandum on Cumulative Effects of Regulation,” The White House, http://www.whitehouse.gov/sites/default/fi les /omb/assets/inforeg/cumulative- effects- guidance.pdf (accessed 24 January 2013). U.S. Environmental Protection Agency (2000) Guidelines for Preparing Economic Analyses. EPA-240- R-00- 003. Washington, DC: U.S. Environmental Protection Agency. ———. (2010) Guidelines for Preparing Economic Analyses. EPA-240- R-01- 001. Washington, DC: U.S. Environmental Protection Agency. ———. (2011a) Regulatory Impact Analysis for the Federal Implementation Plans to Reduce Interstate Transport of Fine Particular Matter and Ozone in 27 States. EPA-HQOAR-2009- 0491. Washington, DC: U.S. Environmental Protection Agency. ———. (2011b) Regulatory Impact Analysis for the Final Mercury and Air Toxics Standards. EPA- 425/R-11- 011. Washington, DC: U.S. Environmental Protection Agency. ———. (2011c) Regulatory Impact Results for the Reconsideration Proposal for National Emission Standards for Hazardous Air Pollutants for Industrial, Commercial, and Institutional Boilers and Process Heaters at Major Sources. Washington, DC: U.S. Environmental Protection Agency. U.S. House of Representatives, Committee on Oversight and Government Reform (2011a) Preliminary Staff Report: Assessing Regulatory Impediments to Job Creation. 112th Cong., Washington, DC. ———. (2011b) Lights Out: How EPA Regulations Threaten Affordable Power and Job Creation. 112th Cong., Washington, DC. U.S. Office of Management and Budget (2003) Circular A-4: Regulatory Analysis, http://www.whitehouse.gov/omb/circulars_a004_a- 4 (accessed 24 January 2013). Walker, W. Reed (2011) “The Transitional Costs of Sectoral Reallocation: Evidence from the Clean Air Act and the Workforce.” Columbia University Job Market Paper, New York. Xie, Jian, and Sidney Saltzman (2000) “Environmental Policy Analysis: An Environmental Computable General Equilibrium Approach for Developing Countries,” Journal of Policy Modeling 22: 453– 89.

Chapter 14

Rationing Analysis of Job Losses and Gains An Exercise in Domestic Comparative Law E. Donald Elliott

The issue of whether to quantify, monetize, and include the job losses (or gains) in the benefit– cost analysis conducted prior to adopting major rules is hardly unique in American law. Many analogous situations exist in which a policymaker must decide in advance of conducting an analysis which factors to include and which to leave out. This is called “scoping,” and the American legal system has experience with it in the regulations created by the Council on Environmental Quality (CEQ) to implement the National Environmental Policy Act (NEPA), as well as in scientific risk assessments. I call the process of looking for analogous policy problems and learning from them “domestic comparative law.” The essential idea is to apply the methods of legal precedents to policy questions, by way of analogy to similar situations with which the law has coped successfully. We may gain guidance for how to scope Regulatory Impact Analyses (RIAs) and when to include quantitative, monetized estimates of net job losses or gains by comparison with other areas in the U.S. legal system. The “precedents” that I draw on are scientific risk assessments, environmental justice, scoping under NEPA, civil procedure, and analyses of “green jobs” created by government subsidies for renewable energy. The common lesson—the “common law,” if you will, of how to ration analysis—is that the analyst ought to spend resources analyzing a problem if but only if one judges that the additional analysis is likely to affect the decision, or its legitimacy or public dialogue, by more than the costs of doing the analysis. Having abstracted this principle from the “precedents,” I then apply it to the issue at hand. The phrase “ job-killing regulations”

Rationing Analysis of Job Losses and Gains

257

has recently become prominent in our political discourse (Corum 2012; Howard 2012), but there is scant empirical evidence to confirm or refute it. That in itself is a strong reason to analyze the effects of significant regulations on jobs. Because I am trained as a lawyer, the domestic comparative law approach comes naturally to me: what lawyers and judges do is to abstract principles from precedents and apply them from one situation to another (Levi 1949). There are five key steps in legal thinking: (1) perceiving resemblances between past situations and the situation at issue; (2) determining that the salient features of the two situations are sufficiently similar; (3) deciding that the precedent is “successful” in the sense that it is working tolerably well and should not be “overruled”; (4) abstracting the key lessons or principles from the precedent; and (5) applying those principles to the situation at hand. All five steps involve judgment (Elliott 1984) and cannot be reduced to an algorithm or programmed into a computer (D’Amato 1977). In suggesting that we should approach the problem of quantifying job losses or gains in RIAs in the same way that parallel problems have been solved in environmental impact analysis, or risk assessment, or civil procedure, I am not suggesting that an agency like the Office of Information and Regulatory Affairs (OIRA) is bound to follow the decisions made elsewhere in the way that a lower court is bound to follow the decisions of a higher court. Rather, it is more like the way that the courts in one state look to the decisions of courts in other jurisdictions for guidance, taking the other state’s decisions for whatever value that it may find in them. Common law precedents are a storehouse of accumulated wisdom as to how others have balanced competing values (Elliott 1984). Domestic comparative law is a method to tap into that wisdom and experience.

Context-Dependent Analysis of Job Losses (or Gains) Whether to include anticipated net job losses (or gains) from a proposed rule in an agency’s benefit– cost analysis depends on the circumstances: in some situations, anticipated net job losses (or gains) from a new regulation may be a material factor that may tip the decision one way or the other. In those situations, net anticipated job losses (or gains) should be weighed in the balance and quantified and monetized to the extent possible to factor into the benefit– cost analysis. In other situations—perhaps even in most situations—net job losses or gains are not likely to affect the decision, and in those situations, job losses or gains can generally be relegated to qualitative analysis and passing mention. There are, however, two caveats that counsel in favor of quantification of job losses

258

E. Donald Elliott

even in situations where they are unlikely to affect aggregate social welfare in a significant way: (1) when we anticipate that there may be significant distributional effects on particular communities and (2) when quantitative analysis might make significant contributions to policy debates, such as the current controversy over “ job-killing regulations.” My position that job losses (or gains) should be quantified and included in the benefit– cost analysis in some cases and not in others was stimulated by an observation by Richard Williams who, in pointing out that some rules, such as those mandating full-body inspections by the Transportation Security Administration at airports, have significant effects on personal liberty, asked whether we should analyze those effects as well as effects on jobs.1 That is indeed true for some rules but not equally true for all rules. An Environmental Protection Agency (EPA) rule mandating utilities to control their air pollution does not impinge on personal liberty interests to nearly the same extent (although liberty interests to conduct a business in the way one chooses are not entirely irrelevant there either; they are just less weighty). The relative importance of analyzing a proposed policy in terms of a particular value will vary from one rule to another because values are implicated to varying degrees in different situations (Eisenberg 1976). This is a familiar problem in law; judges find that they cannot define a single procedure for all situations but must tailor what process is due to the nature of the conflicting interests involved (Mathews v. Eldridge 1976).

Job Losses or Gains May Have Significant Effects on Aggregate Social Welfare The approach of quantifying the effects of job losses (or gains) for some rules but not for others requires the agency preparing the RIA to take a preliminary peek at the probable effects on jobs and to make a judgment of whether it is likely to be a material factor that may affect the outcome. Renn and Elliott (2011:256) discuss a similar problem, stating, “the practical question that every regulator must ask is, ‘Shall I act to address this particular problem now, basing my decision on what is currently known (or, more accurately, believed to be known)? Or shall I instead defer action until a later date, when more may be known, but at the cost of what occurs in the meantime?’ ” Fortunately, anticipating whether the results of a full jobs analysis is likely to be a material factor in the decision is not as difficult as it sounds; something similar is often done in connection with the “green jobs” debate about renewable energy in Europe, as I will discuss later. Every new regulation will create some new jobs in implementing and complying with it, as well as potentially destroy or move abroad some

Rationing Analysis of Job Losses and Gains

259

other jobs at facilities that cannot comply with the new regulation economically. Consequently, the focus should generally be on the net jobs created or destroyed, but there are some exceptions, which are also discussed later, in which the particular jobs created or destroyed do matter. Executive Order 12291, which in 1981 established the OIRA review process, focused primarily on estimating “net social benefit.” If we can see in advance that a new regulation may destroy significantly more domestic2 jobs than it creates— or, inversely, that it is likely to create significantly more domestic jobs than it destroys—then we should analyze those effects along with all of the other economic effects that go into computing net social benefit.3 The Supreme Court long ago opined that Congress did not intend to protect the health of workers by putting their employers out of business (Industrial Union Dept., AFL- CIO v. Amer. Petroleum Inst. 1980), and the adverse physical and mental health effects of being unemployed are well documented (Linn et al. 1985). Whether the effects on jobs are significant enough to warrant including them in the benefit– cost analysis will vary. A regulation with huge anticipated social benefits may swamp the costs of job losses; conversely, where benefits and compliance costs are relatively close, effects on jobs could tip the balance one way or the other.

Job Losses May Also Have Significant Distributional Effects Later versions of the OIRA regulatory review process in Executive Orders 12866 and 13563 broadened the focus to include distributional effects.4 A regulation that creates a net gain in aggregate social welfare may nonetheless have devastating effects on subgroups that bear a disproportionate share of the burdens. The concept of “environmental justice” is based on the idea that a fair distribution of burdens should also be part of our regulatory calculus. So, too, we should pay attention to distributional effects if job losses are concentrated on par ticular communities. This is especially important when the communities affected offer few alternative economic opportunities, because the costs of transitioning to other employment will be especially large. Thus, the distributional analysis of job effects must take into account not just net jobs but also the kinds and locations of the particular jobs lost, as well as those created. Even in situations in which distributional effects do not weigh heavily enough in the balance to alter the decision, nonetheless significant and concentrated adverse effects on our fellow citizens should be identified and considered. One of the lessons from other areas of law is that when government decisions adversely affect individuals, the decision-making

260

E. Donald Elliott

process should not only reach the right result, but it should also respect “dignitary values” for those who may be adversely affected (Mashaw 1976:47). Social science research has shown that people are more likely to accept as legitimate government decisions that adversely affect them if they feel that the decisions were made using fair procedures and took their interests and positions into account (Tyler 1990). Therefore, if job losses are concentrated so that they disproportionately affect par ticular communities or groups, they should be identified and weighed, even if they do not tip the balance of social welfare. However, as elsewhere, these effects should be considered only where the effects on communities are anticipated to be significant.

Policy Controversies May Warrant Additional Analysis of Job Losses and Gains RIAs are intended not merely to inform decision makers but also to provide information to inform policy debate and facilitate public understanding and acceptance. This broader public education function is well understood for NEPA, which requires agencies undertaking major projects that would affect the environment to conduct environmental impact assessments (CEQ 1978). The Supreme Court has said that NEPA “gives the public the assurance that the agency has . . . considered environmental concerns . . . and perhaps more significantly, [NEPA] provides a springboard for public comment” (Robertson v. Methow Valley Citizens Council 1989). But the comparable functions are often less appreciated for the OIRA RIA process. Public policy debates are often informed by the information in an RIA. For example, former OIRA administrator Susan Dudley, now a professor at George Washington University, recently criticized EPA’s “Mercury and Air Toxics Rule,” which regulates pollution from electric utility power plants. Her primary criticism was that 99.996 percent of the benefits claimed for the rule result not from controlling mercury but from controlling other conventional (nontoxic) pollutants as “co-benefits” from the pollution controls installed to control the mercury. She also criticized the rule on the grounds that the prime beneficiaries are not children as claimed by EPA in its press releases but people in their eighties (Dudley 2011). The point for present purposes is not whether Dudley was right but rather that her comments were based entirely on information made available through the RIA. It would have been impossible to have had that policy debate without the information in the RIA. In some circumstances, political and public policy debates may focus significantly on the alleged effects of proposed public policies on job losses or gains. Two examples are the claim by Ross Perot in the 1992

Rationing Analysis of Job Losses and Gains

261

presidential debates that the NAFTA agreement with Mexico would result in significant losses of U.S. jobs—“a giant sucking sound” (Perot 1992)— and the claim more recently by President Obama in his 2011 State of the Union speech that investments by government in stimulating renewable energy would create the green jobs of the future (Obama 2011). In situations in which analysis of the effect on jobs might contribute significantly to public debate, it may be appropriate to do a quantitative analysis of job losses or gains on a discretionary basis even if they are not anticipated to affect the decision, so as to inform public understanding and policy debate.

Rationing Analysis Based on the Anticipated Value of Information If resources were unlimited or analysis were costless, we could analyze every conceivable ripple effect of every policy choice. A naive popular economic journalist in the 1940s, Henry Hazlitt, advocated the comprehensive analysis approach, demanding that we analyze all of the secondary effects of every policy choice (Hazlitt 1946). Hazlitt even presumed to call comprehensive analysis the central “lesson” of economics (1946:3). “The art of economics,” he opined, “consists in looking not merely at the immediate but at the longer effects of any act or policy; it consists in tracing the consequences of that policy not merely for one group but for all groups” (1946:5). A laudable goal perhaps, but tracing the secondary consequences of a proposed policy for all groups also has costs— a point that modern economics also teaches. Not the least of the potential costs is “paralysis by analysis,”5 thinking too much and being paralyzed from acting, a concept that goes back as far as Aesop’s fable The Fox and the Cat and Shakespeare’s Hamlet. Consequently, modern approaches to the problem of how much to analyze have focused on what in economics and risk assessment is called the “value of information approach” (VOI; Finkel 1990; Finkel and Evans 1987). The idea is that a rational policymaker wants to minimize the sum of error costs and transaction costs—that is, to improve the accuracy of prediction but only if the benefits of doing so are not outweighed by the costs, including delay. As Adam Finkel puts it, “The cornerstone of VOI analysis is the common- sense notion that one should never spend more money to study a problem than the losses one would incur . . . by taking one’s best guess and ‘letting the chips fall where they may’ ” (1990:63). The central insight is that we should spend scarce resources analyzing something only if we believe that the value to improving the decision will be greater than the costs and delays

262

E. Donald Elliott

from doing the analysis. As the OIRA review process focuses on aggregate social welfare and significant distribution effects, it follows that we should analyze net job gains or losses if we predict that such analysis may affect the decision (or if there are other reasons to do so, such as informing significant policy debates). Risk analysts have learned the same lesson. Reflecting on twenty-five years of experience doing scientific risk assessments, the National Research Council of the National Academy of Sciences counseled in a 2009 report that risk analysis should not be an “end in itself” but “decisiondriven,” in the sense that risk assessment should be viewed as a method for evaluating the relative merits of various options for managing risk rather than an end in itself. . . . The process of planning risk assessment and ensuring that its level and complexity are consistent with the needs to inform decision- making can be thought of as the “design” of risk assessment. The committee encourages EPA to focus greater attention on design in the formative stages of risk assessment, specifically on planning and scoping and problem formulation. (National Research Council 2009:5, emphasis added)

This is also how economists have approached questions of devising legal procedures: “The basic idea [behind normative welfare economics] is to choose the rule that maximizes social welfare. . . . In the procedure setting [context], social costs include the parties’ private litigation costs, the public costs of the court system, and the error costs created by incorrect decisions. The aim is to choose a rule that minimizes the sum of all these costs” (Bone 2003:5– 6, emphasis added). But how is one to know in advance of doing the analysis what value it may have to “avoiding error costs”? Admittedly, under the VOI approach, the analyst must estimate in advance of doing the full- scale jobs analysis whether doing the analysis is likely to have a significant effect on either aggregate social welfare or distributional consequences or policy debates. Fortunately, this is possible, as is described in the next two sections.

Social Welfare Analysis Should Focus on Net Job Losses or Gains The problem that Hazlitt called considering the “secondary consequences” of a policy was identified a century earlier as “that which is not seen” by the French economist Frédéric Bastiat in his 1850 essay Ce qu’on voit et ce qu’on ne voit pas (That Which Is Seen and That Which Is Unseen). In his famous parable of the broken window (also known as “the glazier’s fallacy”), Bastiat exposed the fallacy that by breaking windows

Rationing Analysis of Job Losses and Gains

263

we create employment for the glazier who repairs them or, more broadly, that hurricanes, natural disasters, or wars stimulate the economy. Bastiat argued that if the money spent repairing the broken window had not been spent that way, it would instead have been spent on something else, such as buying shoes: Have you ever witnessed the anger of the good shopkeeper, James B., when his careless son happened to break a square of glass? . . . [T]he spectators offered the unfortunate owner this invariable consolation— “It is an ill wind that blows nobody good. Everybody must live, and what would become of the glaziers if panes of glass were never broken?” . . . Suppose it cost six francs to repair the damage, and you say that the accident brings six francs to the glazier’s trade—that it encourages that trade to the amount of six francs— I grant it; I have not a word to say against it; you reason justly. The glazier comes, performs his task, receives his six francs, rubs his hands, and, in his heart, blesses the careless child. All this is that which is seen. But if, on the other hand, you come to the conclusion, as is too often the case, that it is a good thing to break windows, that it causes money to circulate, and that the encouragement of industry in general will be the result of it, you will oblige me to call out, “Stop there! Your theory is confined to that which is seen; it takes no account of that which is not seen.” It is not seen that as our shopkeeper has spent six francs upon one thing, he cannot spend them upon another. It is not seen that if he had not had a window to replace, he would, perhaps, have replaced his old shoes, or added another book to his library. In short, he would have employed his six francs in some way, which this accident has prevented. (Bastiat 1850:51)

Bastiat’s point is that every expenditure, whether mandated by regulation or not, creates employment for someone, but at the opportunity cost of diverting spending from elsewhere in the economy. Bastiat’s parable of the broken window is the beginning of the economic theory of “general equilibrium analysis,” which focuses not only on immediate effects of a policy but also on the net effects of jobs in the economy as a whole. Other things being equal, a policy should have net positive effects on jobs in the economy as a whole if it reduces more damage than it costs, because this increases the social pie that is available to  spend and create jobs elsewhere. But other things are not always equal. When a new EPA rule is promulgated, some facilities that were teetering on the brink of closing may be forced to close, thereby causing the workers they employ to become unemployed, with potentially devastating personal consequences. This is the immediate and obvious effect of the new policy; those put out of work are an obvious consequence of

264

E. Donald Elliott

the new policy. Politicians and the news media see their suffering, and this carries substantial weight in our politics. But would they have been put out of work anyway by some later adverse event if the regulation had not been adopted? And what about those individuals elsewhere in the economy who will benefit from the new rule, both in terms of the jobs created to comply with the new rule but also those unseen beneficiaries who will not suffer adverse effects on their health from the pollution avoided? And what about the money spent to comply with the new rule: if it had been spent elsewhere, would it have contributed more or less to aggregate social welfare? And even if the new regulation does contribute to some net social benefit, what about its distributional consequences? Perhaps glaziers are richer than shoemakers, and by creating employment for one rather than the other we are worsening income inequalities in society? Or alternatively, maybe we are justly rewarding those whose efforts contribute more to the common good? These effects are difficult to tease out, but the policy debates in Europe over whether government policies designed to encourage renewable energy create jobs or destroy them show that it can be done. Some argue that government support for renewable energy actually destroys more traditional jobs than it creates “green jobs” (Álvarez et al. 2009). The European green jobs studies also go on to consider distributional effects, arguing that that the green jobs created by subsidizing green energy are temporary, involve lower skills, pay less, and are more likely to be abroad. For example, a study by Gabriel Calzada Álvarez and his colleagues at Universidad Rey Juan Carlos in Spain concluded: “Spain’s experience reveals with high confidence, by two different methods, that the U.S. should expect a loss of at least 2.2 jobs on average, or about 9 jobs lost for every 4 created [by investments in renewable energy]” (2009:1). They then go on to assert that about two-thirds of the green jobs created were just to set up the energy source: in construction, fabrication, installation, marketing, and administration. Only 10 percent of the green jobs created, they claim, were permanent jobs actually operating and maintaining the renewable sources of energy. (Álvarez et al. 2009:25, n.54; see also Álvarez 2009). Other studies of the effects of government investments in renewable energy on jobs in Germany (Frondel et al. 2009) and in Italy, Denmark, Germany, and Spain (Lavecchia and Stagnaro 2010) reach similar conclusions. But, of course, not everyone agrees (see Lantz and Tegen 2009; Shakir et al. 2009). The point for present purposes is not who is right in this particular policy debate. What is important is that the policy debate exists and that it is being conducted based on studies and analyses that project the net

Rationing Analysis of Job Losses and Gains

265

effects of particular policies on jobs and also the nature and distribution of jobs created or destroyed. It has not proved intractable for modestly funded think tanks on both the right and the left in Europe to evaluate the aggregate and distributional effects on jobs of government investments in renewable energy.6 Several different methods have been used in the green jobs debate in Europe, but the usual one (used in both the Álvarez et al. [2009] study in Spain and the Frondel et al. [2009] study at the RheinischWestfälisches Institut für Wirtschaftsforschung in Germany) is to start from the number of jobs that are created on average for a given amount of expenditure in the economy as a whole. One then estimates whether the proposed policy is anticipated to create substantially more or fewer jobs than those that would be created on average by spending the same amount of money elsewhere in the economy. The critics dispute the details of this method (arguing, for example, that it would be more accurate to use data for the jobs created by expenditures in a particular industry rather than for the economy as a whole), but they do not dispute the basic methodology of comparing jobs anticipated from one kind of spending as opposed to another (Lantz and Tegen 2009). Comparing the jobs that will be created by a new policy against the average for spending the same amount of money elsewhere in the economy is not perfect, because it ignores a number of downstream effects, but it is good enough for a first- order approximation to predict whether the net jobs issue is significant and should be analyzed further. The renewable energy studies also go on to look at distributional effects by considering what kinds of jobs will be created and whether they are foreign or domestic. These brief and not very expensive analyses of whether government support for renewable energy projects produces or costs net jobs suggest that a preliminary scaling analysis of the net effects of a proposed policy on jobs is possible. It is already being done at several different think tanks and universities in Europe. Using similar techniques, it should be possible to estimate in the United States whether a par ticu lar regulation may reasonably be expected to have either aggregate employment effects sufficiently substantial that it might affect the cost–benefit balance or significant distributional effects that should be considered.

Who Should Analyze Job Effects When Significant? Unfortunately, RIAs are advocacy documents; they are prepared by the agency proposing a regulation in order to justify it. As a result, claims of huge benefits and low costs in RIAs have come to be regarded with

266

E. Donald Elliott

increasing skepticism. For an example, we return to Susan Dudley’s criticism of the EPA’s Mercury and Air Toxics Rule: The only “direct” benefits EPA presents in support of the rule amount to at most $6 million per year—the upper bound value it places on reducing exposure to toxic mercury emissions, which EPA predicts will raise even the most highly exposed children’s IQ by 2/10ths of one percent (or .002 IQ points). Scientists suggest even these very small effects greatly overstate the likely effects of the rules on children’s health, arguing that EPA “systematically ignored evidence and clinical studies” in developing these health estimates. . . . [T] hrough what is essentially an accounting trick, EPA calculates almost all of its monetary benefits for this rule from particle reductions well below the levels it has established as safe. Contrary to EPA’s claim that the rule will provide par ticu lar benefits to children, the premature deaths EPA says will be averted are modeled to accrue to people with an average age of 80 years, who would live weeks or months longer, if at all, as a result of the regulations. This modeling is also suspect, because EPA assumes causality where none can be explained, and makes other assumptions that overstate effects. (Dudley 2011, emphasis added)

The claim that EPA is using “accounting tricks” and “upper bound values” to overstate the anticipated benefits from its proposed rule is strong language. But it should not be surprising that, after living for over thirty years under OIRA review, agencies have learned how to “game the system” and present data in a way that is helpful to achieving the agency’s goals and objectives. The benefits side of the equation is particularly subject to manipulation, because it is uniquely within the agency’s expertise. Typically, the benefits side of the benefit– cost evaluation consists of the health effects or other harms allegedly to be avoided by an agency’s proposed rule (Katzen 2006). Historically, OIRA has had weak expertise to second- guess agencies on the issues of toxicology and epidemiology that underlie their benefits estimates. OIRA attempted to standardize the way that agencies do risk assessment— and thereby how they calculate benefits— in its 2006 proposed “Risk Bulletin” (U.S. Office of Management and Budget 2006). The draft bulletin was withdrawn in 2007 after its “one- size-fits- all” approach was criticized in a peer review by the National Academy of Sciences (National Research Council 2007). Today individual agencies—and, indeed, individual programs within agencies— decide for themselves how to do risk assessments, under very general “principles” promulgated by OIRA (Dudley and Hays 2007). The technique of regulating agency computations of benefits by promulgating guidelines dictating how they do risk assessments has not been successful to ensure that estimates of benefits are credible.

Rationing Analysis of Job Losses and Gains

267

Is there a better way that might provide more credible information about costs and benefits? One successful model from which we might learn is the Congressional Budget Office (CBO). Prior to the creation of CBO, estimated costs for proposed legislation were provided by individual members of Congress introducing bills by using whatever methods they chose. This system was notoriously inaccurate, with actual budgetary effects often exceeding the predictions made at the time of enactment by orders of magnitude. To cite but one (in)famous example: “When Medicare Part A was enacted in 1965, costs were projected to rise to $9 billion by 1990, but actual costs reached $67 billion by 1990” (Edwards 2003). CBO was created by the Congressional Budget and Impoundment Control Act (1974) to provide independent, nonpartisan, and transparent estimates of the budgetary effects of proposed legislation. CBO is not perfect; politicians have learned ways to game the CBO scoring system, primarily by building in unrealistic provisions that are almost certain to be changed later.7 But CBO has been a great success at standardizing policy discourse about the budgetary effects of legislation. Today advocates and opponents of legislation begin the debate from the common starting place of the CBO estimates, rather than making up their own facts, as they formerly did. “Everyone is entitled to his own opinion, but not his own facts” is an aphorism often attributed to the late senator Daniel Patrick Moynihan, but it actually had many precursors (Doyle et al. 2012). The idea that productive policy debate first requires getting the facts straight goes back to Confucius, who taught 2,500 years ago that “rectification of terms”— getting the facts straight to clarify the debate—was the single most important thing that one could do to improve government ( Jennings 1900:68). Today a powerful narrative is taking hold in many quarters that “ jobkilling regulation” is responsible for many of our country’s economic woes (Howard 2012). For example, in his unsuccessful campaign for the presidency in 2012, Mitt Romney pledged to “put an end to the jobkilling regulations imposed by the Obama administration” (Corum 2012). It may or may not be true that excessive regulation is a significant cause of unemployment in the United States, but we ought to find out whether this is true. One way to do so is to analyze in the OIRA process whether regulations are actually “killing jobs,” as the critics charge, or creating them, as the proponents claim. According to Adam Smith (1776:650), “Science is the great antidote to the poison of enthusiasm and superstition.” An independent, credible, and transparent system for quantifying the net job gains or losses from regulations may be important to confirm or put the lie to the increasingly prominent claim in our

268

E. Donald Elliott

public discourse that regulation kills jobs or, more likely, to clarify that some regulations do destroy jobs while others do not. Perhaps an independent, nonpartisan system for estimating job effects along the lines of the CBO would be more credible than leaving this task to the agencies involved in proposing the rule. Agencies dealing with a particular subject matter such as the environment or transportation would not appear to have any special expertise in estimating the job effects of their proposed rules. An independent, nonpartisan agency with expertise in the area, such as the Bureau of Labor Statistics (BLS), should be better able to evaluate the net jobs to be gained or lost from proposed rules where preliminary analysis shows that the effect on jobs may be significant. An additional advantage to the CBO or BLS approach would be that job estimates could be made on a consistent basis across agencies.

Domestic Comparative Law and Analysis of Jobs Effects of Regulations In this chapter, I have employed a method that I call domestic comparative law analysis. The idea has been to draw on analogies elsewhere in the law to gain purchase on the problem at hand. This approach suggests that we should not always quantify and monetize job losses or gains and include them in the cost– benefit analysis. Rather, we should do so only selectively where we believe that job losses or gains may be significant enough to tip the analysis one way or the other or where there are special reasons to do so even though the outcome is unlikely to be affected. Examples of such special situations include where job gains or losses fall disproportionately on discrete communities or where analysis of job losses or gains may contribute to broader policy debates.

Notes 1. Statement of Richard Williams, “Beyond Compliance Costs: The Other Economic Impacts of Regulation Conference,” Penn Program on Regulation, June 8, 2012, Washington, D.C. 2. Whether U.S. policymakers should be concerned about improving social welfare in the less developed world is a difficult moral dilemma that is outside the scope of this chapter. 3. As a matter of theoretical economics, it might be argued that the anticipated costs of job losses from future regulations are already priced into the wages that workers charge for their ser vices (Mannix this volume). However, by that same logic, the compliance costs for a new regulation should also already have been discounted into the prices that companies were charging for the products (Calabresi 1991). Why analyze one but not the other?

Rationing Analysis of Job Losses and Gains

269

4. For a primer on what is currently included in RIAs and a summary of the relevant legal authorities, see U.S. Office of Management and Budget (2011). 5. For an exploration of the trope and its history in administrative law, see Coglianese (2008). 6. Admittedly, the green jobs analyses in Europe focus on a policy that has already been implemented, rather than on one that is merely proposed. However, few rules analyzed by OIRA have no precursors whatsoever. EPA rules, for example, generally prescribe technologies that have been demonstrated elsewhere, so it should be possible to study employment effects in a sample of early adopters. 7. For an example, see the analysis by the Republican staff of the Senate Budget Committee that claims that the Obama administration used “budget gimmicks” to manipulate CBO into providing an unrealistically low estimate of the actual budgetary effects of the Affordable Care Act (also known as Obamacare) (U.S. Senate Committee on the Budget: Republicans 2012).

References Álvarez, Gabriel Calzada (2009) “Testimony before the House Select Committee on Energy Independence and Global Warming.” Select Committee on Energy Independence and Global Warming. http://globalwarming.markey .house.gov/files/HRG/092409Solar/calzada.pdf. Álvarez, Gabriel Calzada et al. (2009) “Study of the Effects on Employment of Public Aid to Renewable Energy Sources.” Instituto Juan de Mariana. http://www .juandemariana.org/pdf/090327-employment-public-aid-renewable.pdf. Bastiat, Frédéric (1850) “That Which Is Seen, and That Which Is Not Seen,” in Economic Sophisms: Translated from the 5th Edition of the French by Patrick James Stirling. Princeton, N.J.: Van Nostrand, 1964. Bone, Robert G. (2003) Civil Procedure: The Economics of Civil Procedure. New York: Foundation Press. Calabresi, Guido (1991) “The Pointlessness of Pareto: Carrying Coase Further.” Yale Law Journal 100: 1211–1237. Coglianese, Cary (2008) “The Rhetoric and Reality of Regulatory Reform.” Yale Journal on Regulation 25: 85– 96. Corum, Jonathan (2012) “ ‘I Will,’ He Said: Romney’s Campaign Promises.” New York Times, September 6. D’Amato, Anthony (1977) “Can/Should Computers Replace Judges?” Georgia Law Review 11: 1277–1301. Doyle, Charles C., et al., eds. (2012) The Dictionary of Modern Proverbs. New Haven, Conn.: Yale University Press. Dudley, Susan (2011) “EPA Misrepresents Mercury Rule Benefits.” The Absurd Report. http://www.theabsurdreport.com/2011/epa-misrepresents -mercury -rule -benefits -by-susan-dudley/. Dudley, Susan, and Sharon Hays (2007) “Memorandum for the Heads of Executive Departments and Agencies: Updated Principles for Risk Analysis.” Executive Office of the President. http://www.whitehouse.gov/sites/default /files/omb/assets/regulatory_matters _pdf/m07-24.pdf. Edwards, Chris (2003) “Government Schemes Cost More than Promised.” Tax and Budget Bulletin 17. http://www.cato.org/pubs/tbb/tbb -0309 -17.pdf.

270

E. Donald Elliott

Eisenberg, Melvin (1976) “Private Ordering through Negotiation: Dispute Settlement and Rulemaking.” Harvard Law Review 89: 637– 681. Elliott, E. Donald (1984) “Holmes and Evolution: Legal Process as Artificial Intelligence.” Journal of Legal Studies 13: 113–146. Finkel, Adam (1990) Confronting Uncertainty in Risk Management: A Guide for Decision-Makers. Washington, D.C.: Resources for the Future. Finkel, Adam, and J. S. Evans (1987) “Evaluating the Benefits of Uncertainty Reduction in Environmental Health Risk Management.” Journal of the Air Pollution Control Association 37: 1164–1171. Frondel, Manuel et al. (2009) “Economic Impacts from the Promotion of Renewable Energies: The German Experience.” Rheinisch-Westfälisches Institut für Wirtschaftsforschung. www.instituteforenergyresearch.org/germany /Germany_Study_ - _FINAL .pdf. Hazlitt, Henry (1946) Economics in One Lesson: The Shortest and Surest Way to Understand Basic Economics. New York: Three Rivers Press. Howard, Philip K. (2012) “Want More Jobs? Clean up Our Messy Regulatory System.” The Atlantic, September. Jennings, William, trans. (1900) The Wisdom of Confucius. New York: Colonial Press. Katzen, Sally (2006) “Cost– Benefit Analysis: Where Should We Go from Here?” Fordham Urban Law Journal 33: 1313–1319. Lantz, Eric, and Suzanne Tegen (2009, August) “NREL Response to the Report Study of the Effects on Employment of Public Aid to Renewable Energy Sources from King Juan Carlos University (Spain).” National Renewable Energy Laboratory White Paper, NREL/TP- 6A2- 46261. http://www.nrel.gov/docs/fy09osti /46261.pdf. Lavecchia, Luciano, and Carlo Stagnaro (2010) “Are Green Jobs Real Jobs? A Comparative Analysis of Denmark, Italy, Germany and Spain.” Istituto Bruno Leoni. http://docs.wind -watch.org/IBL -Green -Jobs -May -2010.pdf. Levi, Edward H. (1949) An Introduction to Legal Reasoning. Chicago: University of Chicago Press. Linn, Margaret, Richard Sandifer, and Shayna Stein (1985) “Effects of Unemployment on Mental and Physical Health.” American Journal of Public Health 75: 502– 506. Mashaw, Jerry L. (1976) “The Supreme Court’s Due Process Calculus for Administrative Adjudication in Mathews v. Eldridge: Three Factors in Search of a Theory of Value.” University of Chicago Law Review 44: 28– 59. National Research Council (2007) “Appendix B: OMB Proposed Risk Assessment Bulletin.” In Scientific Review of the Proposed Risk Assessment Bulletin from the Office of Management and Budget. Washington, D.C.: National Academies Press. ———. (2009) Science and Decisions: Advancing Risk Assessment. Washington, D.C.: National Academies Press. Obama, Barack (2011) “State of the Union: Winning the Future.” The White House. http://www.whitehouse.gov/state -of-the -union-2011. Perot, Ross (1992) “Giant Sucking Sound— Ross Perot 1992 Presidential Debate,” http://www.youtube.com/watch?v=Rkgx1C _S6ls. Renn, Ortwin, and E. Donald Elliott (2011) “Chemicals,” in J. Wiener, et al., eds., The Reality of Precaution: Comparing Risk Regulation in the United States and Europe. Washington, D.C.: RFF Press.

Rationing Analysis of Job Losses and Gains

271

Shakir, Faiz, et al. (2009) “The Reality of Green Jobs.” Huffington Post. http:// www.huffi ngtonpost .com/the -progress -report/the -reality -of-green -jobs _b _193806.html. Smith, Adam (1776) An Inquiry into the Nature and Causes of the Wealth of Nations. An Electronic Classics Series Publication. http://www2.hn.psu.edu/faculty /jmanis/adam-smith/wealth-nations.pdf. Tyler, Tom R. (1990). Why People Obey the Law. New Haven, Conn.: Yale University Press. U.S. Office of Management and Budget (2006) “Proposed Risk Assessment Bulletin.” http://www.whitehouse.gov/sites/default/fi les/omb/assets/omb /inforeg/proposed _risk _assessment _bulletin _010906.pdf. ———. (2011) Regulatory Impact Analysis: A Primer. http://www.whitehouse.gov /sites/default /fi les/omb/inforeg /regpol/circular -a -4 _regulatory -impact -analysis -a-primer.pdf. U.S. Senate Committee on the Budget: Republicans (2012) “SBC Analysis of President Obama’s Health Law.” http://budget .senate.gov/republican /public /index .cfm /budget -background ?ID=9594b951 -2651 -48b5 -8805 -8709ce7685c5.

Cases Cited Industrial Union Dept., AFL- CIO v. Amer. Petroleum Inst., 448 U.S. 607 (1980). Mathews v. Eldridge, 424 U.S. 319 (1976). Robertson v. Methow Valley Citizens Council, 490 U.S. 332 (1989).

Statutes Cited Congressional Budget and Impoundment Control Act of 1974, 2 U.S.C. § 601 (1974). National Environmental Policy Act of 1969 (NEPA), 42 U.S.C § 4331 (1970).

Regulation Cited Council on Environmental Quality (1978), NEPA and Agency Planning, 40 C.F.R. § 1501.7 Scoping, 43 Federal Register 55992 (Nov. 29).

This page intentionally left blank

Contributors

Matthew D. Adler

Matthew D. Adler is Richard A. Horvitz Professor of Law and Professor of Economics, Philosophy and Public Policy at Duke University. His scholarship lies at the intersection of public law, welfare economics, and moral philosophy. He is the author of Well-Being and Fair Distribution: Beyond Cost– Benefit Analysis and, with Eric Posner, New Foundations of Cost– Benefit Analysis. His current research focuses on ways to incorporate distributive considerations and more sophisticated mea sures of human well-being into policy analysis. Joseph E. Aldy

Joseph E. Aldy is Assistant Professor of Public Policy at the Harvard Kennedy School, a Nonresident Fellow at Resources for the Future, a Faculty Research Fellow at the National Bureau of Economic Research, and the Faculty Chair of the HKS Regulatory Policy Program. His research focuses on climate change policy, energy policy, and mortality risk valuation. In 2009–2010, he served as the Special Assistant to the President for Energy and Environment at the White House. Christopher Carrigan

Christopher Carrigan is Assistant Professor of Public Policy and Public Administration at the George Washington University Trachtenberg School of Public Policy and Public Administration. His research examines the impacts of organizational characteristics and political forces on the behavior of government regulatory agencies. Previously a Regulation Fellow with the Penn Program on Regulation at the University of Pennsylvania Law School, he received his PhD from the Harvard Kennedy School and an MBA from the University of Chicago.

274

Contributors

Cary Coglianese

Cary Coglianese is the Edward B. Shils Professor of Law and Professor of Political Science and Director of the Penn Program on Regulation at the University of Pennsylvania Law School. He specializes in the study of regulation and regulatory processes, with a particular emphasis on empirical evaluation of alternative regulatory strategies. He is founder and faculty advisor to RegBlog.org and editor of Regulatory Breakdown: The Crisis of Confidence in U.S. Regulation, also published by the University of Pennsylvania Press. E. Donald Elliott

E. Donald Elliott is Professor (adj) of Law, Yale Law School, where he has taught since 1981. He also practices law as Senior Of Counsel, Covington & Burling LLP in Washington, DC. Elliott writes about administrative and environmental law and serves as a Senior Fellow of the Administrative Conference of the United States. He was General Counsel of the U.S. Environmental Protection Agency from 1989 to 1991 and its liaison to the Office of Information and Regulatory Affairs. Rolf Färe

Rolf Färe is Professor of Economics and Agricultural and Resource Economics at Oregon State University. His research is rooted in the areas of production and duality theory, which is documented in 12 books, over 200 refereed journal articles, and over 50 contributions to books. He is an Institute for Scientific Information (ISI) most highly cited scholar in the area of economics and finance and also serves on the editorial boards of two journals. Ann E. Ferris

Ann E. Ferris is an economist with the U.S. Environmental Protection Agency. Ferris previously served as the Deputy Associate Director for Energy and Climate Change on the White House Council on Environmental Quality and as a Senior Research Associate at the Federal Reserve Bank of Boston. She completed her undergraduate training in economics at the University of Chicago and received her PhD in economics from the University of Michigan. Adam M. Finkel

Adam M. Finkel is a Senior Fellow at the University of Pennsylvania Law School and Executive Director of the Penn Program on Regulation. A lead-

Contributors

275

ing expert in risk assessment and cost– benefit analysis, especially for toxic chemicals, he was a senior regulatory and enforcement official at the U.S. Occupational Safety and Health Administration from 1995 to 2005 and has been a professor at the University of Medicine and Dentistry of New Jersey School of Public Health and at Princeton University. Wayne B. Gray

Wayne B. Gray is the John T. Croteau Professor of Economics at Clark University, a Research Associate at the National Bureau of Economic Research, and the Director of the Boston Census Research Data Center. His research focuses on regulatory effectiveness and the economic impact of government regulation of environmental and workplace hazards, including effects on employment, productivity, investment, and plant location. Shawna Grosskopf

Shawna Grosskopf is Professor Emerita of Economics at Oregon State University. Her research includes work in per for mance mea surement with applications in environmental productivity, public sector per for mance, education, and health. She serves as Associate Editor for the Journal of Productivity Analysis and is on the editorial board of Health Care Management Science. She is listed in Who’s Who in Economics and is an ISI most highly cited scholar in economics and finance. Michael A. Livermore

Michael A. Livermore is Associate Professor of Law at the University of Virginia School of Law. His research focuses on environmental regulation, cost–benefit analysis, and executive review of agency decision making. He was the founding executive director of the Institute for Policy Integrity at New York University School of Law, a think tank dedicated to improving the quality of government decision making through advocacy and scholarship. He received his J.D. from NYU Law and clerked for Judge Harry T. Edwards on the U.S. Court of Appeals for the D.C. Circuit. Brian F. Mannix

Brian F. Mannix is currently the president of Buckland Mill Associates and a visiting scholar at George Washington University. From 2005 to 2009, he served as the U.S. Environmental Protection Agency’s Associate Administrator for Policy, Economics, and Innovation. Earlier, he served as Deputy Secretary of Natural Resources for the Commonwealth of Virginia. From 1987 to 1989, he was the managing editor of Regulation magazine at the American

276

Contributors

Enterprise Institute. He received his undergraduate, master’s, and public policy degrees from Harvard University. Jonathan S. Masur

Jonathan S. Masur is Deputy Dean and Professor of Law at the University of Chicago Law School, where he has taught since 2007. Prior to joining the faculty he served as a Bigelow Fellow and Lecturer in Law and as a law clerk to Judge Richard Posner of the U.S. Court of Appeals for the Seventh Circuit. He received his law degree from Harvard Law School in 2003. Al McGartland

Al McGartland is Director of the National Center for Environmental Economics at the U.S. Environmental Protection Agency, where he advises policy officials on the economics of environmental policies. The National Center for Environmental Economics issues EPA’s Guidelines for Preparing Economic Analyses and conducts research on the economics of the environment. Prior to EPA, McGartland worked at the Office of Information and Regulatory Affairs in the Office of Management and Budget. He received his PhD from the University of Maryland. Richard D. Morgenstern

Richard D. Morgenstern is Senior Fellow at Resources for the Future in Washington, DC, where he studies the economics of environmental policy. His work on climate change and air quality focuses on the design of cost- effective regulatory instruments. He has served in the State Department, where he participated in the Kyoto Protocol negotiations, and in the U.S. Environmental Protection Agency, where he oversaw policy, planning, and evaluation. He holds a PhD in economics from the University of Michigan. Carl A. Pasurka, Jr.

Carl A. Pasurka, Jr., is an economist with the National Center for Environmental Economics at the U.S. Environmental Protection Agency and an Adjunct Professor in the School of Public Policy at George Mason University. His current research focuses on cost and productivity issues associated with environmental regulations. In addition, he has served as an Associate Editor for the Journal of Environmental Economics and Management (1994–1996) and recently started serving as an Associate Editor of Energy Economics. William A. Pizer

William A. Pizer is Associate Professor at the Sanford School at Duke University, where he is also Faculty Fellow at the Nicholas Institute for Environ-

Contributors

277

mental Policy Solutions. His current research examines how to promote clean energy through private- sector investments, how environmental policy can affect production costs and competitiveness, and how the design of market-based environmental policies can be improved. From 2008 to 2011, he was Deputy Assistant Secretary for Environment and Energy at the U.S. Department of the Trea sury. Eric A. Posner

Eric A. Posner is Kirkland & Ellis Distinguished Ser vice Professor of Law and Aaron Director Research Scholar at the University of Chicago. He has written or edited over a dozen books on topics such as cost– benefit analysis, international law, and social norms. His current research covers immigration law and foreign relations law. He received his law degree from Harvard University and his undergraduate and master’s degrees from Yale University. Lisa A. Robinson

Lisa A. Robinson is a Senior Fellow at the Harvard Kennedy School’s Mossavar- Rahmani Center for Business and Government and an Affiliated Fellow of its Regulatory Policy Program. She is also a Research Associate at the Harvard School of Public Health, where she is affiliated with the Center for Health Decision Science and the Center for Risk Analysis. She specializes in the economic analysis of environmental, health, and safety regulations. Jason A. Schwartz

Jason A. Schwartz is Adjunct Professor and Legal Director of the Institute for Policy Integrity at New York University School of Law. He graduated magna cum laude from New York University School of Law and worked previously as a public policy associate at Pillsbury Winthrop. He has authored publications on climate change and state regulatory policy, such as 52 Experiments with Regulatory Review: The Political and Economic Inputs into State Rulemaking and The Road Ahead: EPA’s Options and Obligations for Greenhouse Gas Regulation. Ronald J. Shadbegian

Ronald J. Shadbegian is an economist at EPA’s National Center for Environmental Economics and an Adjunct Assistant Professor of Economics for Georgetown University’s Economics Department and Public Policy Institute. His research focuses on the costs of complying with environmental regulations and on the regulatory effectiveness and economic impact of

278

Contributors

environmental regulations, including studies on employment, productivity, investment, environmental per for mance, and technological change. He received his PhD in economics from Clark University in 1991. Stuart Shapiro

Stuart Shapiro is Associate Professor and Director of the Program in Public Policy at the Bloustein School of Planning and Policy at Rutgers University. He has written extensively about the U.S. regulatory process and the role of cost–benefit analysis in regulatory decision making. Prior to joining the faculty at Rutgers, he worked for five years at the Office of Information and Regulatory Affairs as a desk officer and assistant branch chief.

Index

Abatement cost ratio, 57, 59, 60, 61– 66, 62, 64, 65, 68n13. See also Pollution abatement ACCCE (American Coalition for Clean Coal Electricity), 243, 247 Activity Analysis Model, 101–103 Adler, Matthew D., 20, 44, 122, 250 “Adverse substitution” trade- offs, 137 Advocacy analyses, 244–247, 252, 265–266 AFPA (American Forest and Paper Association), 246 Akerlof, George, 160 Alar controversy, 138–139 Aldy, Joseph, 9, 20, 78, 242 Aluminum production, 72–73 Álvarez, Gabriel Calzada, 264–265 American Coalition for Clean Coal Electricity (ACCCE), 243, 247 American Forest and Paper Association (AFPA), 246 Analytics. See Regulatory analysis Anxiety and mass layoffs, 43 Arrow, Kenneth J., 24n14, 179, 182 Assessment, QRA parallels to JIA, 129–141 Assumptions: best practices and, 115, 116, 119; carbon pricing and, 75–76; clinging to, 133; disclosing major “default,” 141–142; electricity price and, 79; enumerating, 128; equilibrium, 195–199, 201–202; full employment, 15, 34– 35, 47, 112, 171, 176, 177; green jobs and, 9; simplifying, 19, 124, 139; transparency and, 248–249 Bachmann, Michele, 3 Bartik, Timothy J., 18, 47, 121–122, 136, 182, 186

Bastiat, Frederic, 139, 171, 193–195, 196, 197, 199, 262–263 Baumol, William. J., 180 BCA (benefit- cost analysis), 16–19, 24n14, 38, 113, 125nn8,9, 170–172; areas of improvement and, 183–187; conceptual treatment of employment within, 15, 172–175; EPA and, 175–179; existing empirical research and, 180–183; implications of employment gains and losses for, 119–122; rationing of analysis and, 257–258; research needed and, 179–180. See also CBA (cost- benefit analysis); Regulatory analysis; Welfare costs/effects, BCAs and Becker, Gary, 160 Behavioral data, 161 Benefits: analysis of, 233; best practices and, 118–119, 125n9; co-, 141; defined, 113; net of costs, 221n6. See also BCA (benefit- cost analysis) Berman, Eli, 7, 35– 36, 40, 46, 54, 55– 56, 117, 181 Best practices, 111–127; benefit- cost analysis and jobs and, 119–122; benefits effects and, 118–119, 125n9; complexity and, 17–18, 117–118, 125n8; context of employment impacts and, 122–123; definitions and distinctions in, 112–114; indirect effects and, 118; job increases and decreases and, 117–118; mitigating policies and, 123–124; retrospective analysis and, 124; salient information and, 114–115; screening analysis and, 115–116, 118, 119, 125n5; timing and duration of job gains/losses and, 118; uncertainties and, 116–117, 125n7

280

Index

Bias in research, 128, 131, 132–133 Biomass cogeneration, 73 BLS (Bureau of Labor Statistics), 9, 13, 14, 17, 23n5, 249, 268 Boardman, Anthony, 180, 182 BOF (blast oven furnace production), 72, 76 Boiler MACT Rule (EPA), 135, 138, 242–243, 246–247 Broken window fallacy, 139, 171, 193–194, 262–264 Brown, Randall S., 55 Browner, Carol, 4 Brunnermeier, Smita, 40 Buggy whip factory problem, 210–212 Bui, Linda T.M., 7, 35– 36, 40, 46, 54, 55– 56, 117, 181 Bureau of Labor Statistics (BLS), 9, 13, 14, 17, 23n5, 249, 268 CAC (command- and- control) regulation, 90, 212; Clean Air Act and, 95– 96; in regulatory rigidity study, 93– 95, 95, 98, 99, 100, 103 CAFE (Corporate Average Fuel Economy) Rule (EPA), 228 California, 4, 7, 35– 36, 44 Cap- and-trade, 82, 86, 90, 96. See also carbon pricing policy Capital abundance, 74 Capital congestion, 105, 106–107n15 Capital- labor ratio, 22–23n3 Carbon pricing policy, 75–76, 81– 85, 86n3 Carbon Trust, 76 CBA (cost- benefit analysis): advocacy and, 249; conservatism and, 132, 133, 145n3; delegation to independent entity of, 249, 252; issues in methodology of, 214–219; nonconsumption impacts and, 150–151, 156, 161–162, 165; uncertainty and, 230–231; unemployment and, 207–214, 219. See also BCA (benefit- cost analysis); Regulatory analysis CBO (Congressional Budget Office), 248, 267, 268 Cement production, 73 Census Bureau, 9, 36, 57, 58, 181 CGE (computable general equilibrium) model, 247

Christensen, Laurits Ray, 55 CIBO (Council of Industrial Boiler Owners), 246 Classen, Timothy, 154 Clean Air Act (CAA) and amendments, 8, 37, 39, 40, 183; Interstate Rule of, 228; Title IV of, 95– 96; transition costs and, 242, 252 Climate change policies, 75–76 Clinton, William, 4, 12, 14, 245; Executive Orders of, 15, 111, 113 Cluster Rule (EPA), 55, 181 CO2 emissions. See Carbon pricing policy Cogeneration: biomass, 73; BOF and, 72 Cole, Daniel, 248 Cole, Matthew A., 54 Command- and- control (CAC) regulation, 90, 212; Clean Air Act and, 95– 96; in regulatory rigidity study, 93– 95, 95, 98, 99, 100, 103 Competition: domestic, 52, 58, 75–76; import (international), 52, 57– 58, 73–76 Competitive environment of industry, 51– 69; data for study of, 53– 54, 57– 59; future work on, 66– 67; model for study of, 55– 57, 67nn7,8, 68n14; results of study of, 59– 66; studies on employment and regulation and, 54– 55 Competitive impacts in power- sector, 70– 88; carbon pricing simulation and, 81– 85; competitive effects and, 70–71, 75–76, 80, 80– 81, 81, 84; future work on, 85– 86; mea sures, data and methods in study of, 76–79, 78, 86n2; results of study of, 79, 79– 81, 80; U.S. manufacturing and, 71–76 Complexity, research and, 11, 117–118, 119 Compliance costs, 133, 138, 191, 196, 200–201, 208, 210–211 Compliance timelines, 183–185, 184 Concentration ratio, 54, 58– 59, 60, 61– 66, 62, 64, 65 Congressional Budget and Impoundment Control Act (1974), 267 Congressional Budget Office (CBO), 248, 267, 268 Conservatism in estimating procedures, 131–134, 145n3 Construction industry, 225 Consumer theory, 157

Index Consumption-leisure utility function, 155–157, 166n6 Context- dependent analysis, 257–258 Corporate Average Fuel Economy (CAFE) Rule (EPA), 228 Cost effects, 22–23n3, 36– 37, 93– 94, 98, 106n9, 138 Cost estimation: controversies and, 129–130, 132; parameter uncertainty and, 134–135; reducing excessive conservatism in, 131–134; trade- offs and, 138 Costs: defined, 113; first- and secondorder, 140, 214–217. See also CBA (cost- benefit analysis) Council of Industrial Boiler Owners (CIBO), 246 Creative destruction, 173, 178 Cross- State Air Pollution (Transport) Rule (CSAPR) (EPA), 71, 84– 85, 86n1, 242, 243, 247, 248 CRP (C-reactive protein), 154–155 CSAPR (Cross- State Air Pollution Rule) (Transport Rule) (EPA), 71, 84– 85, 86n1, 242, 243, 247, 248 Cumulative effects, 123, 224, 240, 251–253 Data: approaches to analyzing, 128–129, 145n1; in environmental regulation study, 53– 54, 57– 59; in power- sector studies, 76–79, 96– 98, 106n12; revealed preference, 161; single benefit number and, 130–131; traps to avoid regarding, 133–134; worker studies versus industry-level, 38– 39 Data Envelopment Analysis (DEA), input- based, 101–103, 106nn8,9 Davis, Steven J., 12, 41– 43, 45 Debate about jobs and regulation, 1– 30, 223, 234; analysis to inform, 239–255, 260; changes in policy and, 5– 6; Congress and, 6; economic research and, 6–10; linking of regulation and jobs in, 10–15; political agendas and, 2– 6, 243–247; regulatory analysis and, 15–19 Declaration of Human Rights, 12 Demand effect, 36– 37, 51, 101, 138, 146n7 Demand growth, 58– 59, 60, 61– 66, 62, 64, 65, 73 Democrats, 3– 4, 245

281

Department of Energy (DOE), 97, 228–229, 236n8 Department of Homeland Security (DHS), 228–229 Department of Transportation (DOT), 14, 214–215, 228 DHS (Department of Homeland Security), 228–229 Distributional analysis, 113, 125n8, 250–251, 259 Distributional impacts/consequences, 89, 112, 165n2, 231–232, 240, 250, 253, 259–260, 265 Disutility, 155–160 DOE (Department of Energy), 97, 228–229, 236n8 Domestic comparative law, 256–271; context- dependent analysis and, 257–258; credibility problems and, 265–268; distributional effects and, 259–260; policy controversies and, 260–261; social welfare and, 258–259, 262–265; value of information and, 261–262 Domestic competition. See Competition DOT (Department of Transportation), 14, 214–215, 228 Double counting: employment impacts and, 34, 173–174, 176, 191, 192, 200; problem of, in CBAs, 209–210 Driesen, David, 209–210 Dudley, Susan, 141, 260, 266 Dunn, Richard, 154 EAF (electric arc furnace techniques), 72, 76 Earnings: impact of unemployment on, 17–18, 24n13, 119, 120; mass layoffs and, 42– 43, 48n4 Easterlin paradox, 160 Ecofys, 76 Economic Impact Analyses (EIAs), 34, 45– 46, 113–114, 115, 122 Economic issues, public and, 12, 13 Economic loss rule, 220n5 Economic research, 6–10; key empirical findings in, 35– 41; limits of, 10–11, 33–35; studies of mass layoffs and, 41– 45 Economic theories/models: macro-, 16, 172–173, 183, 185–186; micro-, 6–7, 55– 56, 196–197; neoclassical, 181, 241

282

Index

Ederington, Josh, 74–75 Education, job loss and, 44 EIAs (Economic Impact Analyses), 34, 45– 46, 113–114, 115, 122 Electric arc furnace techniques (EAF), 72, 76 Electricity: energy expenditures, 72; increases in rates of, 70, 72; prices of, as mea surement, 76– 84, 81, 82, 86n4. See also Power- sector regulation Electric power plants (EPPs), 92– 96, 98, 103–105, 106n15 Elliott, E. Donald, 21, 125n6, 248, 249, 258 Elliott, Rob J., 54 Employment: as dependent variable, 57– 59, 60, 61– 66, 62, 64; efficiency and, 241–243; elasticity and gross effects of, 77, 80, 80– 81, 82, 83; equilibrium assumption and, 197–199 Employment impacts, 2, 6–10, 23n5, 33– 50, 125n6, 224–226; BCAs and, 15, 16, 171–172, 190–203; conceptual treatment of, 172–175; cumulative effects and, 224, 240, 242, 251–253; environmental regulation and, 51– 69; EPA and, 175–179; high unemployment and, 186; industry characteristics and, 11, 117; key empirical fi ndings on, 6–10, 35– 41; long- term forecasts, 17–18; monetizing, 67n2, 186–187, 242; new office to study, 231–234; nonconsumption, 150–169; power sector regulation and, 70–108; public debate on, 239–255; research available on, 180–183; research needed on, 179–180, 183–187; studies of mass layoffs and, 41– 45; unifying theory for, 141, 172, 178, 179–180. See also Unemployment Energy expenditures, 72 Energy intensity, 71–73, 79, 79– 84 Environmental justice, 256, 259 Environmental Protection Agency (EPA). See EPA (Environmental Protection Agency) Environmental regulation, 51– 69; alternative metrics used to study, 38, 40; claims against, 3; competitive environments and, 51– 69; data for study of, 53– 54, 57– 59; federal versus state role in, 51; future work on, 66– 67; indirect effects and, 39– 40; model for

studying, 55– 57, 67nn7,8, 68n14; reduced- form model studies on, 37– 38, 39, 40; regulatory rigidity and, 89–108; results of study of, 59– 66; structural model studies on, 35– 37, 38, 40– 41; studies on employment and, 54– 55; worker versus industry studies on, 38– 39. See also Power- sector regulation Environmental technology, 90– 91, 101, 103, 105n4 EPA (Environmental Protection Agency): analysis guidelines and practices of, 112, 132, 176–178, 242–243, 250; BCAs and, 175–179; Boiler MACT Rule of, 135, 138, 242–243, 246–247; Cluster Rule of, 55, 181; compliance timelines of, 183–185, 184; construction industry and, 225; Corporate Average Fuel Economy (CAFE) Rule and, 228; cost estimation and, 134–135, 145n5; Cross- State Air Pollution (Transport) Rule (CSAPR) of, 71, 84– 85, 86n1, 242, 243, 247, 248; Exposure Factors Handbook of, 142; job-loss linked to regulation by, 3– 4, 47, 244– 345; Mercury and Air Toxics (MATS) Rule of, 177, 243, 247, 248, 260, 266; national ambient air quality standards of, 248; National Center for Environmental Economics (NCEE) of, 171; pulp and paper industry and, 55, 143, 181, 209, 241, 246; RIAs and, 46, 52, 228; Transport Rule of, 242, 243, 247, 248. See also Clean Air Act (CAA) and amendments EPPs (electric power plants), 92– 96, 98, 103–105, 106n15 Equilibria- at- full- employment approach, 173, 176, 177–178, 180–186 Equilibrium assumption, 195–199, 201–202 Equity, BCAs and, 176 Estimation procedures. See Cost estimation EU ETS (EU Emission Trading Scheme), 76 Evaluation. See Retrospective analysis Evidence. See Employment impacts Ex ante incentives, CBAs and, 214, 216, 217–219 Executive Order 12291, 226, 259 Executive Order 12866, 111, 113, 125n2, 225, 259

Index

283

Executive Order 13563, 52, 111, 113–114, 124, 170, 231, 259 Executive Order 13610, 124 Ex post evaluation. See Retrospective analysis Ex post procedure, CBA as, 214, 217–219 Externalities: job loss as negative, 17, 19; market failure and, 173, 212

Ho, Mun S., 76 Hopkins, Thomas D., 180 House Republican Plan for America’s Job Creators (House Republican Conference 2011), 245 Human capital, 34, 119, 126n12, 208, 211, 212, 213, 225 Huntsman, Jon, 3

Factor- shift effects, 22–23n3, 36– 37, 101, 138, 146n7 Fairness, BCAs and, 176 Färe, Rolf, 20, 92, 100, 102, 105n4 Farrow, Scott, 179, 182, 183, 186 FDA (Food and Drug Administration), 228, 232 Ferris, Ann, 21, 141 Finkel, Adam M., 20, 23n6, 220–221n6, 248, 261 First- and second- order costs, 140, 214–217, 219 Foreign direct investment (FDI), 39– 40, 41 Fuel economy standards, 4, 14, 214–215 Full employment assumption, 15, 34– 35, 47, 112, 171, 176, 177

Immobility, bad output, 90– 92; empirical example and, 95–101; model to study, 93– 95 IMPLAN model, 246, 247 Import (international) competition, 52, 57– 58, 73–76 Import ratio, 58– 59, 60, 61, 62, 64, 65 Incentives in market- based regulation, 90 Income effect, 144, 200, 202n1 Indirect effects, 17, 39– 40, 118, 138–139 Industry- level data, 38– 39, 53– 54, 57– 59, 67n8 Input- based Data Envelopment Analysis (DEA), 101–103, 106nn8,9 Institute for Policy Integrity, 225–226 Investment ratio mea sures, 58– 59, 60, 62, 64, 65 “Invisible hand” (Smith), 176 Issa, Darrell, 244–245

Gallup polls, economic issues and, 12, 13, 43 General- equilibrium costs, 138, 140 Graham, John D., 137 Gramlich, Edward, 19, 180 Grandfathering, transitions and, 218–219 Gray, Wayne B., 9, 11, 20, 48n1, 55, 181, 242 Great Recession, 1, 2, 14, 172, 223, 234 Greenhouse gas emissions, 76 Green jobs, 3– 4, 9, 129–130, 245, 269n6 Greenstone, Michael, 8, 11, 37– 38, 39, 55, 67n6, 139, 181 Handwerker, Elizabeth Weber, 44 Hanna, Rema, 40 Happiness: surveys, 151–152, 162–164, 165, 166n3; and unemployment, 45, 152–153, 166nn4,5 Haveman, Robert H., 179, 182, 183, 186 Hazlitt, Henry, 261, 262 Health effects, 18, 118, 125n10, 150, 154–155, 166n7, 223, 259, 266 “Health- wealth” effect, 202n1. See also “Richer is safer” theory

Jackson, Lisa, 245 Janicki- Deverts, Denise, 154–155 JIA ( job impact analysis), 129–130, 142; QRA parallels to, 131–141 Job creation: Bastiat on, 193–194; regulation and, 111, 124, 197–198, 244; rhetoric and, 5– 6, 33, 52, 133, 144, 171, 186 Job displacement, 42– 46 Job impact analysis (JIA), 129–130, 142; QRA parallels to, 131–141 “Job-killing regulation”: effects of versus effects of job gains, 14; linked to environmental protection, 244–245; as rhetorical phrase, 1, 3– 5, 5, 22, 33, 244, 256–257, 258, 267 Job- loss assessment and valuation: estimation procedures and, 132–133; parameter uncertainty and, 135–136; rationing analysis of, 256–271; trade- offs and, 138–140. See also Employment impacts

284

Index

Jobs, value of, 12, 16, 18, 24n13, 47, 67n2, 119–122, 126n11, 131, 135–136, 139–140, 150–165, 170–171, 174–178, 180, 182, 186–187, 195, 197–198, 208–210, 213–214, 220n3, 228–229, 247 Jordan, Jim, 14

Labor: as productive input, 120; valuation framework and, 119–120 Labor- based technology/perspective, 92– 94, 95 Labor demand: regulation impact study and, 35– 36, 54– 56, 181; regulation rigidity study and, 93, 95, 100–101 Labor markets, 6, 9, 11, 42, 173, 175, 183, 226, 241 Labor Requirement with Command- andControl Model, 103–104 Labor Requirement with Tradable Permits Model, 104–105 Leisure, utility models and, 155–160 Levinson, Arik, 39, 40, 75 Livermore, Michael A., 21, 129, 132, 182, 183, 225–226, 232–233 Location decisions, 39, 74–75 LP (linear programming models) problem, 94, 96, 98, 103, 105n4 Luddism, 194

Masur, Jonathan S.: adjustment costs of rigidities and, 90; employment impacts and, 21, 34– 35, 41, 44– 45, 47, 125n1; reform and, 225–226, 228, 231, 241; regulatory analysis and, 122, 133, 143, 151, 163, 182, 186, 200, 213, 220nn1,3, 227, 236n6, 243 MATS (Mercury and Air Toxics) Rule (EPA), 177, 243, 247, 248, 260, 266 McGartland, Al, 21, 141 McKinsey and Company, 76 MENCEU (monetary equivalent for the nonconsumption effects of unemployment), 156, 160–165, 167n13 Mercury and Air Toxics (MATS) Rule (EPA), 177, 243, 247, 248, 260, 266 Microeconomic theory, 6–7, 55– 56, 196–197 Minimum wage laws, 9–10, 198 Monetary equivalent for the nonconsumption effects of unemployment (MENCEU), 156, 160–165, 167n13 Monetary value of unemployment. See Earnings; Jobs, value of; Valuation Monte Carlo techniques, 134–136 Moratoria, regulatory, 5, 6, 143–144, 240, 245, 251–253 Morgenstern, Richard D.: cost- and factor- shift- effects and, 22n3, 51, 94, 98, 138, 146n7; employment impacts and, 7– 8, 11, 20, 35, 36, 40, 46, 48n1, 52, 54– 55, 74, 117, 135, 181–182, 228, 242 Morrall, John, 226–227 Mortality, 24n12, 41, 43– 44, 118, 119, 130, 133, 135, 137, 138, 140, 146n8, 154–155, 161, 200–201, 202n1 Moser, Klaus, 152

Macroeconomic effects, 225, 227, 227, 231, 232 Macroeconomic theories/models, 16, 172–173, 183, 185–186 Maier, Richard, 154 Manufacturing, 51– 54, 57– 59, 66– 67, 71–76 Mannix, Brian F., 21 Market- based regulation, 18, 90 Markets, well and poorly functioning, 211–212 Mass layoffs, studies of, 20, 41– 45, 48n4, 122, 154–155

NAAQS (National Ambient Air Quality Standards), 37, 184–185 National Academy of Sciences (NAS): report, 141–142, 146n9; Research Council of, 262, 266 National Ambient Air Quality Standards (NAAQS), 37, 184–185 National Environmental Policy Act (NEPA), 256, 260 National Highway Traffic Safety Administration (NHTSA), 228 National Income and Product Accounts (NIPA), 192–193

Kaldor- Hicks criterion, 151, 165n2, 176, 192 Keller, Wolfgang, 39 King, Angus, 11, 23n9 Knabe, Andreas, 153 Kolstad, Charles, 39– 40

Index Neoclassical economic theory, 181, 241 NEPA (National Environmental Policy Act), 256, 260 Net effects, 11, 84, 181, 263, 265; trade- offs and, 136–141 NHTSA (National Highway Traffic Safety Administration), 228 Nickles, Don, 4 NIPA (National Income and Product Accounts), 192–193 Nonconsumption impacts, 150–169; disutility models and, 155–160; estimating MENCEU and, 156, 160–165, 167n13; psychological and health effects and, 151–155 Nonpecuniary effects, 41, 45, 119, 122 Oates, Wallace E., 180 Obama, Barack: Executive Orders of, 5– 6, 52, 111, 113–114, 124, 170, 231; on jobs and regulation, 3– 4, 207, 223, 235n2, 261, 267, 269n7 OIRA (Office of Information and Regulatory Affairs), 226–227, 232, 252, 257, 259, 260, 262, 266, 267, 269n6 OMB (Office of Management and Budget), 7, 34, 112, 114, 134, 207, 227 O’Neill, Tip, 14 Opportunity costs, 17, 113, 121, 174–175, 179, 182–183, 186, 193, 240, 241 Output- based technology/perspective, 91, 92, 106n8 PACE (Pollution Abatement Costs and Expenditures), 35– 37, 38, 40, 57, 59, 67n10 Palmer, Karen, 9 Parameter uncertainty, 134–136, 145n5 Partial static equilibrium model (PSEM), 55 Paul, Karsten, 152 Pence, Mike, 5 PERI (Political Economy Research Institute), 247 Perot, Ross, 260–261 Perry, Rick, 3 Petroleum consumption, 72 Pigouvian taxes, 212 Pizer, William, 9, 20, 36, 78, 242 Point estimates, 134, 135

285

Policy. See Regulatory policy Political agendas and jobs, 2– 6, 243–247 Political concerns about employment, 11–14 Political Economy Research Institute (PERI), 247 Pollution abatement, 35– 37, 48n1, 51, 52, 54, 55– 57, 63, 66, 74, 75, 90– 91, 93, 96, 98, 101, 105n3, 106n6, 174, 181, 185–186. See also PACE (Pollution and Abatement Costs and Expenditures) Pollution Abatement Costs and Expenditures (PACE), 35, 37, 38, 40, 57, 59, 67n10 Pollution havens, 40, 74 Porter, Michael E., 9, 23n6 Posner, Eric A.: adjustment costs of rigidities and, 90; employment impacts and, 25, 34– 35, 41, 44– 45, 47, 125n1; reform and, 225–226, 228, 231, 241; regulatory analysis and, 122, 133, 143, 151, 163, 182, 186, 200, 213, 220nn1,3, 227, 236n6, 243 Potential Pareto efficiency test, 176 Power- sector regulation, 70–108; carbon pricing simulation and, 81– 85; future work on, 85– 86; mea sures, data and methods in study of, 76–79, 78, 86n2; regulatory rigidity and, 89–108; results of study of, 79, 79– 81, 80; U.S. manufacturing and, 71–76. See also EPPs (electric power plants). PPFs (production possibilities frontiers), 89– 92, 91, 106n13 President’s Council on Jobs and Competitiveness, 6 Price, regulation and, 197 Product demand, 52– 54, 58, 59, 63 Production possibilities frontiers (PPFs), 89– 92, 91, 106n13 Productivity, 67n5, 74, 118, 119, 125nn9,11, 174, 182, 198, 209, 210, 212, 213–214, 220n3, 227–228 Productivity levels, total factor (TFP), 39, 41 Program evaluation. See Retrospective analysis Protectionism, 194 PSEM (partial static equilibrium model), 55 Psychological effects, 151–154

286

Index

Public attitudes toward employment, 12–13, 13 Pulp and paper industries, 39, 48n1, 72–73, 143; EPA and, 55, 143, 181, 209, 219, 241, 246 QALY (quality- adjusted life year)-to- dollar trade- offs, 164–165 QRA (quantitative risk assessment), 129–141; estimation procedures and, 131–134; pa rameter uncertainty and, 134–136; trade- offs net effects and, 136–141 “Quasi- fi xed” factors, 55– 56, 67n7 Rascoff, Samuel, 141 Rätzel, Steffen, 153 Rayo, Luis, 160 Reagan, Ronald, 4, 226, 234, 244 Reallocation of resources, 89, 94, 113 Recycling sector, 73 Reduced- form model studies, 35, 37– 38, 39, 40, 48nn2,3 Reform of regulatory process, 223–238; CBAs and, 207–222; domestic comparative law and, 256–271; impacts and, 224–226; informing public discourse and, 239–255; new regulatory analyses office and, 230–234; RIAs and, 225–230 Regulation, impacts of. See Employment impacts Regulatory analysis: challenge of, 33– 35, 45– 48; decision making and, 18–19, 114; failure to include employment in, 15–16, 34– 35, 71, 170–179; implications for, 15–19; jobs and job quality and, 17–18, 142–143; monetary value and, 18–19, 47; proposed office of, 231–234; public debate, 239–255; rationing of, 256–271. See also BCA (benefit- cost analysis); Best practices; CBA (costbenefit analysis); Nonconsumption impacts; Reform of regulatory process; Research agenda; RIAs (Regulatory Impact Analyses); Risk assessment; Welfare costs/effects, BCAs and Regulatory Freeze for Jobs Act, 245 Regulatory Impact Analyses (RIAs). See RIAs (Regulatory Impact Analyses) Regulatory moratoria, 5, 6, 143–144, 240, 245, 251–253

Regulatory policy, unemployment and, 207–222; CBAs to- date and, 208–209; double- counting problem and, 209–217; issues in methodology and, 214–219; moratoria and, 5, 6, 143–144, 240, 245, 251–253; policy controversies and, 260–261 Regulatory rigidity, 89–108; bad output immobility and, 90– 92; empirical example of, 95–100, 97, 99; models for analyzing, 93– 95, 101–105; outputbased perspective on, 91, 92 Reinaud, Julia, 75–76 Reinforcing effects (“co- benefits”), 141 Relocation of inputs, 38, 74–75 REMI model, 247 Renewable energy studies, 261, 264–265 Renn, Ortwin, 258 Rents versus returns, 212–214 Republicans, 2– 3, 4, 207, 223, 224, 234, 245, 269n7 Research agenda, 125n3, 170–189; areas for improvement in, 183–187; existing empirical research and, 180–183; need for unifying theory and, 172, 178, 179–180; traditional BCAs and, 171–179 Reservation wage, 24n13, 121, 133, 135, 140, 161, 167n12, 175, 182–183 Retrospective analysis, 19–20, 124, 126n14, 224, 231–233, 235 Revesz, Richard, 141, 232–233 RIAs (Regulatory Impact Analyses), 33– 35, 38, 45– 48, 52; as advocacy document, 265–266; defined, 113–114; as disclosure document, 244, 247–253; domestic comparative law and, 256–257, 258, 260; reform of regulatory process and, 225–230, 227, 233 “Richer is safer” theory, 146n8 Risk assessment, 128–149, 261; “decisiondriven,” 262; estimation procedures and, 131–134; net of trade- off effects and, 136–141; other parallels to, 141–143; parameter uncertainty and, 134–136, 248; policy prescriptions and, 143–144, 266; QRA parallels and, 131–141; stylized example of, 130–131 Risk reduction, 134, 136, 137 Risk valuation, 132, 134 “Risk-versus-risk” effects, 137

Index Rivers and Harbors Act (1902), 191 Romney, Mitt, 3, 267 Roosevelt, Franklin D., 12 Rubio, Marco, 3 Ruhm, Christopher J., 24n12, 43– 44 “Scarring” effect, 166n4 Schumpeter, Joseph, 173, 178 Schwartz, Jason A., 21, 129, 132 Scoping, 256, 262 Second- order costs, 140, 214–217, 219 Shadbegian, Ronald, 9, 11, 20, 48n1, 242 Shapiro, Stuart, 21, 226–227, 249 Shih, Jhih- Shyang, 36 Shipments, as dependent variable, 57– 58, 60, 61– 65, 64, 65, 67n11 Smith, Adam, 175–176, 267 Smith, V. Kerry, 178–179, 182–183 SO2 emissions, 89, 96– 98, 97, 106–107n15 Social costs: BCAs and, 112, 113, 120, 123, 173, 177, 183, 185; extent of, 234, 241, 262; Morgenstern and, 208–209, 210, 211; regulation and, 34, 38, 41, 47, 252; rents and, 213–214. See also Welfare costs/effects Stated preference surveys, 122, 132, 161–162, 166n10 Steel production, 72 Structural model studies, 35– 37, 38, 40– 41 Substitution effect, 144–145, 181, 202n1 Substitution of products, 39, 73; adverse, 137 Suicide, 24n12, 44, 140, 154 Sullivan, Daniel, 43– 44, 154–155 Taylor, M. Scott, 40, 75 TFP (total factor productivity levels), 39, 41 Third-party effects, 165n1 Tort and contract law, 216, 217 TP (tradable permit system) regulation, 92– 95, 95, 98, 99, 100, 104 Trade- offs: net effects of, 136–141; QALY-to- dollar, 164–165; types of, 137–138 Transition costs, 38, 120–121, 140, 173, 209, 220nn1,3, 241–242; Clean Air Act and, 242, 250–251, 252; research and, 174, 176, 177, 179, 182, 185

287

Transparency: best practices and, 114; informing public debate and, 240, 247–250; risk assessment and, 128–129, 141 Transportation, relocation and costs of, 74 Transport Rule (EPA), 71, 84– 85, 86n1, 242, 243, 247, 248Truman, Harry S., 14 Ueberroth, Peter, 4 Uncertainty: agency estimates of job impacts and, 135; government policy and, 23n5; informing public debate and, 249; parameter, 134–136, 145n5; research and, 1, 114, 116, 118, 119, 121, 125n4, 128, 131 Unemployment: disutility models and, 155–160; effects of long- term, 223–224; estimating MENCEU and, 160–165; Great Recession and, 1, 2; high, 186; increase of, 12; mass layoffs and, 41– 45; minimum wage laws and, 9–10; political concern about, 11–14; psychological and health effects and, 150–155; public attitude toward, 12, 190–191; regional and local variations of, 13–14; regulatory policy and, 207–222; valuation of, 17–19, 24n13, 119–122 Unemployment insurance, 23n5, 43, 125n11, 166n5, 173, 174, 251 Unfunded Mandates Reform Act (UMRA), 229 Union wage premium, 212 Unregulated technology, 91– 95, 98, 99, 100, 106n14 Utility function, 156–160 Utility model, classical, 156–160 Valuation: of employment impacts, 122, 179, 186–187, 197–199; of nonconsumption impacts, 151, 155–160; of risk, 132, 134; of unemployment, 17–19, 24n13, 119–122, 133, 249. See also Jobs, value of. Value of information approach (VOI), 261–262 Von Wachter, Till M., 12, 41– 44, 45, 48n4, 154–155 VSL (value for a statistical life), 130, 132, 134

288

Index

Wages: impact of regulation on, 8, 126n13, 197–199; impact of unemployment on, 17–18; mass layoffs and, 42– 43, 48n4; rents versus returns and, 212–214; reservation, 24n13, 121, 133, 135, 140, 161, 167n12, 175, 182–183 Walker, W. Reed, 8, 11, 38– 39, 209, 242, 252 “Washington Monument fallacy,” 141 Wealth, effects on health, 140, 146n8, 200, 202n1 Welfare costs/effects: analysis and, 119–122, 139–140, 142; domestic comparative law and, 258–259, 262–265; employment impacts and, 18, 24n13, 34, 38, 43– 45. See also Social costs

Welfare costs/effects, BCAs and, 190–203; empirical observations and, 199–201; equilibrium assumption and, 195–199; invisible effects and, 193–195; origin of BCAs and, 191–193 Welfare economics: BCAs and, 192; social welfare function and, 151; valuation framework and, 119–120, 125–126nn11, 12 Wiener, Jonathan B., 137 Williams, Richard, 258 Wilson, Pete, 4 Worker losses, 201n1, 210–211; rents versus returns and, 212–214 Worker versus industry studies, 38– 39 Work-role centrality, 153–154 Xing, Yuquing, 39– 40

Acknowledgments

Producing this book has been a team effort. Not only did three editors collaborate in planning, writing, and editing, but we brought together 19 thoughtful scholars and practitioners to author chapters. We thank first and foremost these authors. Their willingness to participate in this project— and meet tight deadlines—has made editing this book a privilege. We believe that their evidence and insights will ensure the attainment of this book’s goal of helping elevate public discussion about regulatory policy and stimulating more research and better analysis of the employment effects of regulation. We also thank the many other experts who attended the two events that the Penn Program on Regulation organized in 2012 around the issues in this book: a workshop held in Washington, D.C., and a conference held at the University of Pennsylvania Law School. We appreciate the comments and contributions of all the participants at these events; they have profoundly informed and improved the pages of this book. We note our particular gratitude for the speakers, commentators, and reviewers at these two events, including Don Arbuckle, Linda Bui, Susan Dudley, Mickey Edwards, John Graham, Sandra Hoffmann, Adriana Kugler, Robert O’Connor, Arden Rowell, Sarah Stafford, Michael Toffel, Wendy Wagner, Reed Walker, David Weimer, Richard Williams, and Clifford Winston. We note as well our special thanks to V. Kerry Smith, not only for his thoughtful participation in the Penn Law conference organized around the chapters of this book but also for extending an invitation to us to participate in an EPA- sponsored workshop he organized on regulation and employment. The University of Pennsylvania Law School, under the extraordinary leadership of Dean Michael Fitts, provided both the soil in which the seed of this project was planted and the institutional support that made the project bloom. It is simply not possible to thank Dean Fitts enough for all he has done for Penn Law. We hope that this project will be seen

290

Acknowledgments

as another positive manifestation of his commitment to infusing legal education and scholarship with interdisciplinary insights. We also thank the extraordinary administrative team at Penn Law. Anna Gavin handled everything from workshop logistics to manuscript preparation with her usual knack for making tireless dedication look effortless. Bailey White and other members of the Faculty Support Staff headed by Silvana Burgese provided crucial support with the Penn Law conference. Jen Evans, Kelly Farraday, Paul Maneri, and Matthew McCabe provided much appreciated assistance with manuscript preparation. Benjamin Meltzer gave careful attention to citations, and he and other members of the superb Biddle Law Library staff ably led by Paul George helped in tracking down materials whenever needed. Penn Law students Valerie Baron and Alisa Melekhina took helpful notes at our Washington, D.C., workshop, and Brady Sullivan, now a Penn alum, continued to outperform as a research assistant. In addition to the support provided by the University of Pennsylvania Law School, this undertaking was only possible thanks to generous financial support from additional sources. We greatly appreciate the support of Allen J. Model and the Leo Model Foundation, and we are pleased that this book was made part of a larger project supported by the National Science Foundation’s Decision, Risk, and Management Sciences program, under Grant 0756539, on “Transferring to Regulatory Economics the Risk-Analysis Approaches to Uncertainty, Interindividual Variability, and Other Phenomena.” We are grateful to the team at the University of Pennsylvania Press for giving this work a home, and we offer particular thanks to our editor, Bill Finan, for his key assistance during all phases of publication. Michael Haggett and his capable team at Westchester Publishing Ser vices proved instrumental in moving the production of the book forward in a timely manner. Finally, we dedicate this book to workers everywhere who have suffered harm from ill- designed, poorly managed, or insufficiently effective regulation. We hope that with more rigorous attention to the issues presented throughout this book, political leaders and civil servants will be able to make smarter decisions that better protect workers and their family members.