Valuing Food Safety And Nutrition [1 ed.] 9780367212957, 9780429267031, 9781000017441, 9781000004083, 9780367215767

Citation preview

Valuing Food Safety and Nutrition

Edited by Julie A. Caswell

Valuing Food Safety and Nutrition

Valuing Food Safety and Nutrition EDITED BY

Julie A. Caswell

~

'l Routledqe

i~

Taylor

Frani'~,oup

LONDON AND NEW YORK

For my parents, Catherine and Merle Caswell

First published 1995 by Westview Press Published 2018 by Routledge 52 Vanderbilt Avenue, New York, NY 10017 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Routledge is an imprint of the Taylor & Francis Group, an informa business Copyright © 1995 by Taylor & Francis All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data Valuing food safety and nutrition I [edited by] Julie A. Caswell. p. cm. Includes bibliographical references. ISBN 0-8133-8944-5 (alk. paper) I. Food adulteration and inspection. 2. Food industry and tradeSafety measures. 3. Nutrition. I. Caswell, Julie A. TX531.V34 1995 363. I 9'2~c0

ISBN 13: 978-0-367-21295-7 (hbk)

95-15133 CIP

Contents ix

List of Tables and Figures Preface Acknowledgments About the Contributors

xv

xix

xxi

PART ONE Setting the Stage: R981'Ch Perspectives and Theoretical Models 1

Valuing Food Safety and Nutrition: The Research Needs, Eileen 0. van Ravenswaay

3

2

Self-Protection, Risk Information, and Ex Ante Values of Food Safety and Nutrition, Young Sook Eom

27

3

Mitigation, Product Substitution, and Consumer Valuation of Undesirable Foodbome Effects, Robert D. Weaver

51

4

Information Issues for Principals and Agents in the "Market" for Food Safety and Nutrition, Michael D. Weiss

69

PART TWO A Comparison of Valuation Methodologies 5

Contingent Valuation of Health Risk Reductions for Shellfish Products, C.-T. Jordan Lin and J. Walter Milon

6

Experimental Auctions to Measure Willingness to Pay for Food Safety, John A. Fox, Jason F. Shogren, Dermot J. Hayes, and James B. Kliebenstein

115

Using Conjoint Analysis to Assess Consumers' Acceptance of pST-Supplemented Pork, Catherine Halbrendt, John Pesek, April Parsons, and Robert Lindner

129

7

v

83

vi

8

Contents

Health Risk Concern of Households vs. Food Processors: Estimation of Hedonic Prices in Fats and Oils,

Dong-Kyoon Kim and Wen S. Chern 9

Valuation by the Cost of Illness Method: The Social Costs of Escherichia coli 0157:H7 Foodbome Disease,

Tanya Robens and Suzanne Marks

10 11

Valuing Food Safety: Which Approaches to Use?

155

173

Maureen L. Cropper

201

Thoughts About Different Methods to Value Food Safety and Nutrition, Ann Fisher

211

PART THREE

A Closer Look at Performing Contingent Valuation 12

Using Contingent Valuation to Value Food Safety: A Case Study of Grapefruit and Pesticide Residues, Jean C. Buzby,

Jerry R. Skees, and Richard C. Ready 13

Anglers' Willingness to Pay for Information About Chemical Residues in Sport Fish: Design of a CV Questionnaire,

Douglas J. Krieger and John P. Hoehn 14

Using Contingent Valuation Methods to Value the Health Risks from Pesticide Residues When Risks Are Ambiguous,

Eileen 0. van Ravenswaay and Jennifer Wohl 15

16

Contingent Valuation of Consumers' Willingness to Purchase Pork with Lower Saturated Fat, Catherine Halbrendt,

219

251

287

Lesa Sterling, Sue Snider, and Gail Santoro

319

Criteria for Evaluating Results Obtained from Contingent Valuation Methods, Richard B. Belzer and Richard P. Theroux

341

PART FOUR

Inputs to Valuation Studies 17

Determining Foodbome Illness in the United States: A Step Toward Valuation, Thomas E. Steahr

365

Contents

vii

18

Measuring the Food Safety Risk of Pesticides, Kelly A. Day, Betsey A. Kuhn, and Ann M. Vandeman

393

19

Adding Nutritional Quality to Analysis of Meat Demand, Guijing Wang and Wen S. Chern

411

20

Irradiation and Food Safety: Consumer Attitudes and Awareness, Sukant K. Misra, Stanley M. Fletcher, and Chung L. Huang

435

About the Book and Editor

451

Tables and Figures Tables 5.1 5.2

Descriptive Statistics for WTP Responses by Format and Hypothesis Tests Between WTP Responses Ordinary Least Squares Estimates of the Willingness to Pay Model (Outliers Excluded)

92

95

6.1 6.2

Differences in Mean Bids at Trial 1 and Trial 20 Frequency Distribution of Bids

121 123

7.1 7.2 7.3 7.4

Porlc Product Profiles Used for Evaluation by Respondents Shopping Centers Chosen for Data Collection Socio-Demographics of Study Respondents Information Sheet Read to Respondents Before Conducting Survey Belief in Organiution Reports on Product Safety Mean Rating Values for Survey Product Profiles Estimated Coajoint Model for Parameters Affecting Product Ratings Hypothetical pST Products with Equivalent Rating as Current Technology at 20 Percent Fat Reduction

134 135 136

7.5 7.6 7.7 7.8 8.1 8.2

9.1 9.2 9.3

Estimated Coefficients from Selected Moving Regression Runs, Cooking and Salad Fat and Oils Estimated Coefficients from the Entire Sample With and Without Cholesterol Information Variable, Baking and Frying Fats and Oils Assumptions Used to Estimate Annual Cost of Illness for Foodbome Disease, 1992 Annual Incidence of Escherichia coli 0157:H7 and Severity of Illness, U.S. Medical Costs of Acute Illness from Escherichia coli 0157:H7 by Severity Category, 1992 Dollars, Typical Year ix

137 139 140 141 142 165 167 176 180 182

x

Tables and Figures

9.7 9.8

Productivity Loss During Acute Illness from Escherichia coli 0157:H7 by Severity Category, 1992 Dollars, New Cases in Typical Year Summary of Costs for Disease Caused by Escherichia coli 0157:H7, 1992 Medical Costs and Productivity Losses Estimated for Selected Foodborne Pathogens, 1992 Strengths and Weaknesses of Different Valuation Methods Uses for Food Safety Valuation Methods

191 195 198

11.1

Criteria for Evaluating Valuation Methodologies

213

12.1

Techniques Followed in Study to Increase Participation and Reduce Nonresponse Bias Response Rates for Four Versions of the Consumer Mail Survey Results from the Two Regressions Using WTP as the Dependent Variable

230

13.1

Pretest WTP Means

268

14.1 14.2 14.3

Perceived Chance of a Health Problem Sureness About Health Risk Perceived Reduction in the Risks from Pesticide Residues When Foods Meet Federal Standards for Pesticide Residues or Are Produced Without Pesticides Attitudes Toward the Government and the Scientific Community Willingness to Buy Labeled and Unlabeled Apples

297 297

9.4

9.5 9.6

12.2 12.3

14.4 14.5 15.1 15.2 15.3

Profile of Respondents Willingness to Pay for LSF Pork Estimated Influence of Socio-Demographic Characteristics on WTC and WTP for LSF Fresh Pork

17.1

Patients Discharged from Hospitals by Category of Foodborne Disease, List A, United States, 1990 Patients Discharged from Hospitals by Category of Foodborne Disease, List B, United States, 1990 Patients Discharged from Hospitals by Category of Foodborne Disease, List A, by Days of Care, Average Length of Stay, and Average Annual Hospital Costs, United States, 1990

17.2 17.3

186 188

223 226

298 300 301 322 325 327 371 378

382

Tables and Figures

17.4 17.S 17.6 17.7 17.8

Patients Discharged from Hospitals with Principal Diagnosis as Foodbome Disease, List A, United States, 1990 Frequency of Foodbome Disease, Physician's Office Visits and Hospital Discharge Certificates for 1990 and Appearance on Death Certificate for 1986, United States Selected Demographic Characteristics, Physician Office Visits with Foodbome Illness, List A, United States, 1990 Selected Demographic Characteristics, Patients Discharged from Hospitals with Foodbome Illness, List A, United States, 1990 Selected Demographic Characteristics, Deaths from Foodhome Illness, List A, United States, 1986

18.1

Chemicals Detected in the Total Diet Study, by Relative Risk Category

19.1

Average Food Consumption Per Capita Per Week and Prices for Different Household Groups Means of Nutritional Quality Indexes of Meat and Related Products Means of Nutritional Quality Indexes of Meats for Different Consumer Groups by Educational Levels Scaled Condition Indexes and Variance-Decomposition Proportions (INQs of Beef) Scaled Condition Indexes and Variance-Decomposition Proportions (INQs of Beef) Scaled Condition Indexes and Variance-Decomposition Proportions (INQs of Beef) Regression Results of Price-Quality Functions Results of Auxiliary Regressions and Selected Regression Results of Price-Quality Functions

19.2 19.3 19.4 19.S 19.6 19.7 19.8 20.1 20.2 20.3 20.4 20.S 20.6

Sample Characteristics of Survey Respondents Consumer Perceptions of Suggested Food Safety Concerns Consumer Concern About Specific Irradiation Ramification Consumer Perception About the Necessity of Irradiation Cross-Tabulation of Consumer Awareness Index by Economic and Demographic Characteristics (Chi-Square Contingency Test) Cross-Tabulation of Consumer Perceptions of Irradiation by Economic and Demographic Characteristics, and Awareness Index (Chi-Square Contingency Test)

xi

383 385 387 388 389 401 420 422 423

425 426 426 427 430 438 439 440 441

444 446

xii

Tables and Figures

20.7

Consumers' Confidence in Comments Made About Food Safety by Various Groups of Individuals 20.Al Complete Text for Question on Factual Statements to Measure Consumer Awareness of Irradiation 20.A2 Complete Text for Question on Arguments in Favor of Irradiation 20.A3 Complete Text for Question Pertaining to Advantages of Irradiation

448 453 454

455

Figures 6.1 6.2 6.3

Subjective Assessments of the Annual Risk of Contracting Salmonellosis Average Willingness to Pay for Reduced Salmonella Risk Average Willingness to Pay to Avoid bST Milk

120 120 125

7.1

Predicted Ratings

142

8.1

Cholesterol Information Index (CHOLS and CHOL), 1950-1990 The Composition of Fatty Acids by Product Estimates of Implicit Prices of Nutrient Fats for Cooking and Salad Dressing Use Estimates of Implicit Prices of Nutrient Fats for Baking and Frying Use Estimated Expenditure Shares of Fatty Acids in Cooking and Salad Fat and Oils Estimated Expenditure Shares of Fatty Acids in Baking and Frying Fats and Oils

8.2 8.3 8.4 8.5 8.6

161 163 166 167 168 169

9.1

While 98 Percent Survive an E. coli 0157:H7 Infection, It Can Cause Severe Outcomes

174

13.1 13.2 13.3 13.4 13.5 13.6

Initial Wording of WTP Questions Final Description of Current Advisory Intermediate Form of WTP Questions Final Form of WTP Question Behavioral Change Questions Pattern of Survey Response

264 265 266 267 271 274

14.1

Health Effects Perceived to Be Associated with Pesticide Residues in Food

299

Tables and Figures

18.1 18.2 18.3

Pesticide Residues Found in a Market Basket of Food Pesticide Use Per Acre, 1965-1991 Number of Pesticides Detected in the Total Diet Study, 1965-1991

xiii

402 404 406

Preface The delivery of safety and nutritional quality is an increasingly important measure of the performance of the food production, manufacturing, and distribution system. Survey research suggests that consumers are interested in and have a demand for products with improved safety and nutritional attributes or characteristics (e.g., lower contamination levels for foodbome pathogens, lower fat content). At the same time, there is public demand for such attributes, particularly as their availability is linked to improved health status and lower health care costs. This book's intent is to increase our understanding of the economics of food safety by addressing the unresolved question of how much value consumers and society place on food safety and nutritional quality. Quantifying the value (ultimately in dollar terms) of food safety and nutritional quality is a central ingredient in private and public decision-making regarding the types of products offered in the market and of regulations passed and enforced. The core questions are what are the benefits of improved food safety and nutrition; who receives them; who is willing to pay for them; how much are they willing to pay; and how do the benefits measure up against the costs of improvements? Valuation (measurement of benefits) for food safety and nutrition attributes poses real difficulties, as van Ravenswaay discusses in detail in her introductory first chapter of this book. In terms used by economists, the valuation difficulties arise because safety and nutrition attributes are often nonmarket goods and associated with market failures due in large part to a lack of sufficient information for market participants. Because of the complex nature of the good, valuation can focus on improving product attributes, information about attributes and substitute products, or both. Van Ravenswaay's Chapter 1 provides a road map to the valuation questions asked, difficulties in answering them, and research approaches that can be applied. Economists are seeking to achieve a better understanding of markets for food safety and nutrition through use of formal models that emphasize the most important aspects of and interactions in these markets. Part 1 (Chapters 2-4) of the book carries forward this modeling effort. Eom presents a formal economic model of consumer self-protection, use of information, and valuation, while Weaver's model explores relationships between the possibility of mitigating risks, product substitution, and consumer valuation. Weiss discusses the promise of employing principal-agent theory in research on the economics of food safety. xv

xvi

Preface

There is a great demand for valuation research both for public (e.g., to support regulatory rule-making) and private (e.g., to judge the market viability of new products) purposes. Valuation of food safety and nutritional quality is a newly developing field. Relatively few studies, employing varying methodologies and reporting standards, have been published. Given the scattered nature of the research to date, a major goal of this book is to report a broad range of current research and lay a foundation for future work. Part 2 of the book begins to lay this foundation by presenting a side-by-side comparison of the five major methodologies now in use for valuation research on food safety and nutrition. The authors of chapters in this part describe the methodology they used; discuss their efforts to design a study that would yield reliable results; honestly evaluate the strengths, weaknesses, and possible pitfalls of their own and other methodologies; and report valuation results. Here and throughout the book the complete texts of surveys, where used as part of the research design, are included in chapter appendixes in order to facilitate evaluation and give maximum impetus to further research. Chapters 5-9 provide the side-by-side comparison of the five valuation methodologies. The first three chapters use methodologies that collect primary valuation data from consumers. Lin and Milon employ a contingent valuation survey to assess the value consumers would attach to lowering health risks associated with shellfish products. Their findings highlight the importance to the results of the information context in which the valuation is done. Fox et al. report on their use of an experimental auction to value consumers' willingness to pay for reduced risk from pathogens in meat products. They note the method's strength in eliciting values and the safeguards needed to generate valid results. Halbrendt et al. use conjoint analysis methodology to assess consumer demand for pork products with lower fat levels attained by use of pST in production, illustrating a valuation technique frequently employed in marketing research. The last two valuation methodologies compared employ secondary data in their experimental designs. Kim and Chem use the hedonic method and market data on prices and demand to estimate the value households and food processors place on the health-related characteristics of fats and oils. They note the advantages of using secondary data for valuation research. Finally, Roberts and Marks employ the cost of illness method to value the social costs of Escherichia coli 0157:H7 in the United States, providing an in-depth discussion of the relative advantages of cost of illness and willingness to pay methodologies. Part 2 also provides two expert overviews and comparisons of the five methodologies by Cropper and Fisher. Cropper emphasizes the crucial importance of understanding exactly what is being valued in order to obtain reliable and interpretable results. For example, what are the health consequences that survey respondents are valuing and do they interpret them the same as the survey designers? And what averting behaviors are respondents considering and incorporating in their valuation estimate? Fisher addresses the advantages and

Preface

xvii

costs of different methodologies in light of the question: "How important is it to know 'the real answer'?" Of particular aid to users and designers of valuation research is a table of criteria for evaluating valuation methodologies that Fisher provides. Among the methodologies in use for valuation of food safety and nutrition, contingent valuation has received a great deal of recent attention and has been one of the most widely employed approaches. To further lay a foundation for future research, Part 3 (Chapters 12-16) of the book takes a closer look at the requirements for performing well-designed contingent valuations in this area. In particular, the chapters focus on issues of survey design. Buzby et al. pay special attention to specification of the relative risk from grapefruit treated with two different pesticides; the decrease in risk they ask consumers to value; and alternative payment vehicles to allow comparison of consumers' responses. Krieger and Hoehn focus on the specification of alternative scenarios to be valued by respondents to their survey on the value anglers would attach to improved advisories on chemical residues in sport fish. They candidly report how much they learned from very thorough pretesting of their survey instrument. Van Ravenswaay and Wohl carry the specification theme forward by detailing their approach to clearly communicating risk scenarios to survey respondents when the risks involved are ambiguous. Halbrendt et al. then present results from a contingent valuation survey of consumers' willingness to pay for lower fat pork. Part 3 ends on a note of caution from Belzer and Theroux, who lay out a strict set of requirements they believe studies using the contingent valuation methodology must meet in order to be suitable for use in policy-making decisions. Part 4 (Chapters 17-20) of the book goes further behind the scenes in laying the foundation for valuation research by discussing the development of important inputs to valuation research, many of which are often not in place. Steahr develops estimates based on available national data bases of the incidence of foodbome illness in the United States. He highlights that current reporting systems are not specific and differentiated enough to be used for reliable incidence data, contributing to the difficulty in conducting precise cost of illness studies. Day et al. develop an overall measure of food safety risk from pesticides, focusing on difficulties in pinpointing risk levels given incomplete underlying scientific information. Wang and Chem develop nutritional quality indexes for meat, arguing that demand studies are not reliable if they do not incorporate changes in the quality of meat consumed by different groups over time. Finally, Misra et al. report on a survey of consumer knowledge of and attitudes toward use of irradiation technology to affect food safety and quality. A striking finding relevant to all valuation research is the lack of knowledge about this technology they found among the survey respondents. Their results underline again the importance in valuation studies of presenting a clear context to respondents before eliciting a valuation estimate.

xviii

Preface

Information on the value to consumers and society of improved food safety and nutrition is increasingly in demand from food marketers, government regulators, and consumers themselves. As this new field of valuation work develops, it is important that research designers and users share insights, critically compare methodologies, cross-verify results, and work together to improve the quality and usefulness of their valuation estimates. It is hoped that this book, by its design, lays the foundation for that effort. Julie A. Caswell University of Massachusetts, Amherst

Acknowledgments The chapters included in this book originated as presentations at a major research conference titled Valuing Food Sqfety and Nutrition held in Washington, D.C. in June 1993. The conference was organized by Regional Research Project NE-165: Private Strategies, Public Policies, and Food System Performance in recognition that safety and nutritional quality are increasingly important measures of the performance of the food production, manufacturing, and distribution system. The Organizing Committee for the conference included Helen Jensen, Department of Economics, Iowa State University; Tanya Roberts, Economic Research Service, U.S. Department of Agriculture; Eileen van Ravenswaay, Department of Agricultural Economics, Michigan State University; Ronald Cotterill, Director, Food Marketing Policy Center, Department of Agricultural and Resource Economics, University of Connecticut; and the editor. Administrative support for the conference was provided by the Food Marketing Policy Center, University of Connecticut. Funding support was provided by the Food Marketing Policy Center, University of Connecticut; the Cooperative State Research Service, U.S. Department of Agriculture; and the Farm Foundation. The editor wishes to thank the book's contributors for their hard work in preparing high-quality chapters that clearly and directly address important food safety and nutrition valuation issues. Special thanks go to Darleen Slysz who prepared the book manuscript with unsurpassed competence and unmatched good spirits, and to Eileen Keegan who prepared the figures. Finally, special thanks go to Richard Rogers and the Department of Resource Economics, University of Massachusetts at Amherst, for their support of this project. J.A.C.

xix

About the Contributors Richard B. Seber is an economist, Office of Information and Regulatory Affairs, Office of Management and Budget. Jean C. Buzby is a research specialist, Department of Agricultural Economics, University of Kentucky. Wen S. Chern is a professor, Department of Agricultural Economics and Rural Sociology, The Ohio State University. Maureen L. Cropper is an associate professor, Department of Economics, University of Maryland and fellow, Resources for the Future. Kelly A. Day is an economist, Resources and Technology Division, Economic Research Service, U.S. Department of Agriculture. Young Sook Eom is an assistant professor, Department of Economics, Clark University. Ann Fisher is a senior research associate, Department of Agricultural Economics and Rural Sociology, The Pennsylvania State University. Stanley M. Fletcher is a professor, Department of Agricultural and Applied Economics, University of Georgia. John A. Fox is an assistant professor, Department of Agricultural Economics, Kansas State University. Catherine Halbrendt is an associate professor, Department of Food and Resource Economics, University of Delaware. Dermot J. Bayes is an associate professor, Department of Economics, Iowa State University. John P. Hoehn is an associate professor, Department of Agricultural Economics, Michigan State University. Chung L. Huang is a professor, Department of Agricultural and Applied Economics, University of Georgia. Dong-Kyoon Kim is manager, Department of International Research, Korea Research Center. James B. Kliebenstein is a professor, Department of Economics, Iowa State University. Douglas J. Krieger is a graduate assistant, Department of Agricultural Economics, Michigan State University. Betsey A. Kuhn is an economist, Resources and Technology Division, Economic Research Service, U.S. Department of Agriculture. C.-T. Jordan Lin is a research assistant professor, Food and Resource Economics Department, University of Florida. He is currently on assignment with the Economic Research Service, U.S. Department of Agriculture. xxi

xx ii

About the Contributors

Robert Lindner is a professor, Department of Agricultural Economics, University of Western Australia. Stmll1l1e Marks is an economist, Economic Research Service, U.S. Department of Agriculture. J. Walter Milon is a professor, Food and Resource Economics Department, University of Florida. Sukant K. Misra is an assistant professor, Department of Agricultural Economics, Texas Tech University. April Parsons is a former research assistant, Department of Food and Resource Economics, University of Delaware. John Pesek is a senior programmer/analyst, Department of Food and· Resource Economics, University of Delaware. Richard C. Ready is an assistant professor, Department of Agricultural Economics, University of Kentucky. Tanya Roberts is leader of the Food Safety and Regulation Section, Economic Research Service, U.S. Department of Agriculture. Gail Santoro is a market analyst, E. I. DuPont de Nemours & Company. Jason F. Shogren is an associate professor, Department of Economics, Iowa State University. Jerry R. Skees is a professor, Department of Agricultural Economics, University of Kentucky. Sue Snider is a food and nutrition specialist, Cooperative Extension System, University of Delaware. Thomas E. Steahr is a professor, Department of Agricultural and Resource Economics, University of Connecticut. Lesa Sterling is an associate professor, Department of Animal Science and Agricultural Biochemistry, University of Delaware. Richard P. Theroux is an economist, Office of Information and Regulatory Affairs, Office of Management and Budget. Ann M. Vandeman is an economist, Resources and Technology Division, Economic Research Service, U.S. Department of Agriculture. Eileen 0. van Ravenswaay is a professor, Department of Agricultural Economics, Michigan State University. Guijing Wang is a postdoctoral associate, Department of Agricultural and Applied Economics, University of Georgia. Robert D. Weaver is an associate professor, Department of Agricultural Economics, The Pennsylvania State University. Michael D. Weiss is an agricultural economist, Economic Research Service, U.S. Department of Agriculture. Jennifer Wohl is a Ph.D. candidate, Department of Agricultural Economics, Michigan State University.

PART ONE

Setting the Stage: Research Perspectives and Theoretical Models

1 Valuing Food Safety and Nutrition: The Research Needs Eileen 0. van Ravenswaay 1 Systematic assessment of research needs requires an understanding of the key questions which define a research field, an assessment of the extent to which existing research has answered those key questions, and an analysis of what research remains to be done. However, it is not so clear what questions define the research field we have come to call "valuation of food safety and nutrition." There are a growing number of studies that employ some type of valuation method to address food safety and nutrition issues, but they are not directed at some commonly accepted and widely recognized set of research questions. Like most research fields, valuation of food safety and nutrition has developed incrementally with no initial defining agenda of research questions. Because it is a relatively new field, it is still in the process of identifying core research questions. This state of affairs gives this author a rather large scope of discretion in staking out an agenda of needed research on valuation of food safety and nutrition. But it also presents a formidable task since it means exploring a much larger territory than my own comparatively narrow and familiar research area. One obvious approach for identifying valuation research needs would be to identify current issues in food safety and nutrition policy and develop implications for valuation research needs. However, I chose not to take this approach because, by the time an issue is getting serious political or judicial attention, the time for conducting valuation research is usually already passed. It takes years to produce reliable, valid, and accurate research results, so production of these results needs to be timed to occur well before, not after, the period of public attention to policy issues. Thus, a list of current policy issues tells us more about what valuation research results are currently useful rather than what valuation research we should conduct next. 3

4

Eileen 0. van Ravenswaay

What valuation researchers need is a reasonably reliable way of anticipating potential policy issues. That way, when research is needed in the policy process, it will be available off the shelf. This chapter presents my attempt to develop a framework for identifying potential food safety and nutrition policy issues and the implications for valuation research. I do this by raising exploratory questions about how social institutions, such as unregulated markets, might fail to provide goods related to food safety and nutrition, and how the nature of these institutional failures change over time with changes in production and consumption technologies. Because changes in institutional failures create resource allocation problems, they are potential policy issues. By examining the extent to which existing valuation research has addressed these potential policy issues, I identify needs for research on valuation of food safety and nutrition. The first section begins by examining the development of valuation research. The section argues that valuation research developed primarily to inform analysis of public policies, and that a central task of that research has been to estimate the demand for nonmarket goods (i.e., goods which unregulated markets fail to provide for some reason). However, nonmarket goods can become market goods through a combination of public policy and private actions, thus creating valuation issues for policy and marketing research. The second and third sections examine what these nonmarket goods are in the case of food safety and nutrition. Section two argues that there are three classes of activities that households engage in to produce health and that certain types of information about food safety and nutrition, as well as food attributes, are inputs to these activities. Section three raises questions about how unregulated markets might fail to provide information and food attributes, and how these failures might change in response to developments in production and consumption technologies, thus creating valuation research issues. The section also points out the extent to which these valuation issues have been examined by economists. The fourth section identifies some methodological valuation research needs that underlie the broader set of food safety and nutrition policy issues. The concluding section summarizes future research directions.

The Purpose of Valuation Research The concept of "valuation research" grew out of the field of cost-benefit analysis. Beginning in the 1960s, it became increasingly common for economists to conduct cost-benefit analyses of proposed changes in government regulations, programs, and policies (Gramlich 1981). The objective of this analysis was to improve the allocation of public resources by providing policy makers with information about the economic impacts of policy change. Specifically, it was hoped that this type of analysis would improve economic efficiency by encouraging adoption of policies that would give the greatest net

Valuing Food Safety and Nutrition: The Research Needs

5

benefits. Cost-benefit analysis was thus employed to analyze the consequences of policy alternatives that policy makers were considering adopting and to identify how existing policies might be improved. As cost-benefit analysis began to be conducted more widely, it became apparent that while it was relatively obvious how the costs of policy change should be measured (though hardly simple to actually measure them), it was not so evident how benefits should be measured. The problem stemmed from the fact that while there were organized markets for most policy inputs, many policy outputs such as improved environmental quality and safety are goods for which no primary markets exist. With no primary market data, demand curves, and, hence, changes in consumer welfare resulting from a change in supply, could not be estimated. Consequently, other methods for estimating the demand for nonmarket goods needed to be developed if cost-benefit analysis was to provide meaningful information to policy makers. This research came to be known as benefit assessment or nonmarket valuation research, and, since in most cases the private sector had substantially less incentive to accurately document the benefits than the costs of policy change, it was funded primarily by tax dollars. The majority of nonmarket valuation research grew out of the health, safety, and environmental legislation of the 1960s and the concern with regulatory reform in the 1970s (Asch 1988, Bailey 1980, Bentkover et al. 1986, Braden and Kolstad 1991, Ferguson and LeVeen 1981, Lave 1981). Some of these policies directly reduced human health risks by requiring private and public goods to meet certain health and safety standards while others required the provision of information which individuals could use to make their own health and safety decisions. The branch of this research of greatest relevance to food safety and nutrition focused on estimating the value of preventing deaths or disease in a given population and to some extent, the value of reduced morbidity and symptom days (Cropper and Freeman 1991, Fisher et al. 1989, Viscusi 1993). Applications of these methods to food safety began appearing in the early 1970s (Ricardo-Campbell 1974, Lave 1981), and recent public attention has substantiatly increased the amount of research in this area. In contrast, valuation research on nutrition policy focused primarily on assessing changes in nutrient intakes and estimating the marginal value of specific nutrients (Capps and Schmitz 1991), rather than on valuing the health effects a change in nutrient intake might produce. Studies on the value of health and safety programs sought to value the ex ante reduction in death or disease within a specified population. The earliest studies used the human capital approach which involved using an estimate of cost savings as an estimate of the value of saving lives. Cost savings were approximated by the present value of foregone earnings net of consumption, plus estimated medical expenses associated with illness or injury. This approach was widely criticized for defining benefits solely in terms of cost savings and for its narrow definition of the relevant costs saved. The approach has since been

6

Eileen 0. van Ravenswaay

modified to incorporate the value of lost leisure time, the opportunity cost of lost household production, and other intangibles. This broader definition of cost savings is commonly referred to as the cost-of-illness approach to estimating the value of health and safety improvements. The cost-of-illness approach has the advantage of being based of actuarial data. A more theoretically correct measure of the value of health and safety programs is people's willingness to pay for reduced risk of death or illness in a specified population. This is a preferred measure because it represents the full value to the individual of health and safety improvements. Willingness to pay measures have been developed using hedonic and contingent valuation methods. These methods involve observing choices made in an actual or constructed market, and inferring the value of risk reduction based on those choices. This brief overview of the development of the field of valuation of reduced risk reveals several key questions that valuation research has sought to answer. The unifying question has been an interest in measuring the benefits (and sometimes the social costs) of changes in government regulations and programs. Answering this question has involved answering questions about what benefits are produced by government policies and how they should be estimated. Because public goods are not directly exchanged in private markets, alternative methods needed to be developed to estimate the demand for these nonmarket goods, including exploiting market data on complements and substitutes for those goods or simulating markets for the goods themselves or their complements or substitutes. Although research on valuing health risk reductions has focused on providing benefit assessments for cost-benefit analysis, there are two other research areas that use methods similar to some nonmarket valuation methods. One of these research areas is program evaluation. Program evaluation addresses the question of the effectiveness of public programs and the extent to which public policy intent is or could be achieved by particular programs. For example, a number of public safety programs, including food safety and nutrition, produce safety and nutritional information for the public. A body of research, some of which includes risk communication, has been conducted in order to test the effectiveness of various types and formats of information delivery (National Academy of Sciences 1989, Magat and Viscusi 1992). This research frequently uses survey methods in which information is presented to subjects and the effectiveness of the information in changing risk perceptions and subsequent risk management behavior is assessed. This methodology is quite similar to methods that might be used to estimate the benefits of providing risk information. A second research area with similarities to nonmarket valuation research is product marketing research. In fact, some of the techniques used to simulate or construct markets for public goods were initially drawn from the literature on marketing research for new products. In particular, there are parallels between what has come to be known as conjoint analysis in marketing research and

Valuing Food Safety and Nutrition: The Research Needs

7

contingent valuation in benefits assessment. Both methods seek to assess consumer behavior toward goods that do not exist in current markets. However, these two methods seek to answer very different questions. In marketing research, the questions are what product configuration needs to be produced for different market niches, what is the market demand for a product, and how is demand affected by factors such as consumer concern about food safety and nutrition. In contrast, valuation research seeks to estimate welfare changes from the demand curves for nonmarket goods. For example, marketing research might ask what the demand is for pesticide-free foods, but policy research would ask what welfare changes would result from an increase in the supply of pesticide-free foods. Similarly, marketing research might seek an estimate of the demand for fat-free foods, whereas policy research might ask what welfare changes would result from improved information on the fat content of foods. Both types of research might involve estimating the demand for pesticide-free or fat-free food, but would use different models and derive different types of estimates. There is the potential for confusion when these different research areas and objectives are not clearly delineated. For example, research on willingness to pay for reduced risk from pesticide residues in food would be useful to estimate the consumer health benefits that might be associated with reduced pesticide use, but that same research design may not be appropriate for market research aimed at estimating the potential market for pesticide-free foods because of differences in the way the product is designed and assumptions about the amount of advertising effort needed to achieve a given level of product awareness. There is also the point that welfare analysis seeks a measure of how consumers would value a given risk reduction if they were fully informed about the "true" risk, whereas marketing research seeks an understanding of the actual knowledge level and practices of consumers. This does not mean that welfare analysis would not involve measuring consumers' revealed preferences for risks and their valuation of perceived risk reductions, but it does mean that the valuation of a public policy change should be based on objective measures of risk reduction. Should valuation research encompass all three types of research identified in this section? Valuation research as it is applied to environmental regulation, for example, would probably not include all three types of research (Braden and Kolstad 1991). However, food safety and nutrition issues differ from environmental issues in terms of the types of institutional failures involved. In particular, the benefits of food safety and nutrition policy are a joint product of public and private action. The benefits of food safety and nutrition policies depend on the ultimate changes in food producer and food consumer behavior. Thus, there are complementarities between the three research areas, so perhaps a more encompassing view of valuation research may be in order. Nonetheless, there is the potential for confusion in assessing the progress that has been made in answering the key research questions that each of the three areas of research

8

Eileen 0. van Ravenswaay

has sought to answer. Rather than settle the question, this author hopes only to have alerted readers to a potential source of difficulty in assessing the contributions made by various studies available in the literature on food safety and nutrition.

Role of Food Safety and Nutrition in Household Health Producnon If one essential task of valuation research is to estimate the demand for the nonmarket goods produced by food safety and nutrition policies, then a valuation research agenda for food safety and nutrition may be described by identifying the goods these policies produce. Alas, there seems to be little agreement about the nature of these goods in the literature, so this section proposes a preliminary framework. Obviously, food safety and nutrition policies affect human health, but they do so differently. To see how they are different, let us examine a model of household health production. In such a model, it is assumed that an individual derives utility from health directly, by avoiding pain and suffering, and indirectly, by having increased time available to enjoy leisure and consumer goods. Health also affects an individual's budget constraint by reducing the portion of time available to earn income. Thus, individuals or households may be viewed as engaged in the production of healthy states (Cropper and Freeman 1991). The production of healthy states involves three classes of activities: health maintenance, health protection, and health rehabilitation. Health maintenance involves activities which ensure optimum development and functioning of the physical systems and parts of the human body. These activities include adequate food consumption, sleep, exercise, and the like. Health protection involves actions to protect the body from damage from external hazards such as harmful levels of exposure to microbes, chemical toxins, radiation, electricity, and extreme temperatures or forces. Protection activities may be of three different (1) averting hazard formation (e.g., sanitation, pasteurization, types: refrigeration, and heat), (2) avoiding hazard exposure (e.g., destroying contaminated food; reducing consumption of fat, cholesterol, or calories), and (3) increasing resistance to hazards (e.g., vaccinations). Health rehabilitation involves activities to mitigate damage or alleviate symptoms or incapacitation. These mitigating activities include the use of drugs, surgery, mechanical aids, and special diets or treatments. Health rehabilitation activities may be undertaken to mitigate the damage from inherited conditions or the damage that occurs when health is not adequately maintained or protected. This damage may vary in terms of its severity, duration, reversibility, and time of onset as well as in terms of the painfulness of the symptoms and degree of incapacitation the damage causes.

Valuing Food Safety and Nutrition: The Research Needs

9

Note that these three classes of activities involved in household production of healthy states affect the utility of the household differently, and thus have implications for welfare analysis and marketing research. The first two classes of activities, health maintenance and protection, affect current expenditures and future reductions in utility and increases in time constraints. Health rehabilitation activities affect current expenditures, increases in current and future utility, and decreases in current and future time constraints. The level of a household's health protection activities over time can vary with exogenously determined changes in huards in the household's environment, whereas household health maintenance can vary with changes over time in household characteristics such as age. A household's health rehabilitation activities vary with a household's choices about health maintenance and protection as well as household characteristics and exogenous changes in ha7.ard exposure. Finally, there are substitution possibilities and potential complementarities across the three activities. Food safety and nutrition fit into the household's health production activities in different ways. Food safety relates to a household's health protection and rehabilitation activities. For example, food preparation choices can affect exposure to pathogens, and illness caused by foodbome pathogens may need medical treatment. In contrast, nutrition may relate to health maintenance, health protection, or health rehabilitation activities of a household. For example, consumption of an adequate level of nutrients is necessary to maintain optimum health, but too much consumption of nutrients may harm health. In addition, some types of nutrient consumption may be necessary to reverse health problems, such as in the case of malnutrition. It is important to distinguish among these different types of household activities involved in the production of healthy states because they require different physical and informational inputs. For example, health maintenance activities of households include obtaining certain types of food attribute inputs whereas health protection activities include avoiding other types of food attributes. Health rehabilitation activities may involve obtaining some and avoiding other food attributes. To obtain desired food attributes, the household must possess two types of information: the type and level of attribute intake necessary to maintain or rehabilitate health and the marginal attribute contribution of each food. To avoid undesirable food attributes, the individual must possess three types of information: the type and level of attribute intake that may impair health, the marginal attribute contribution of each food, and actions that may be taken to reduce the marginal attribute contribution (e.g., washing or cooking). The next section examines how unregulated markets may fail in providing desired levels of food attributes and informational inputs, thus creating policy and food marketing issues which valuation research can 'help to

address.

Eileen 0. van Ravenswaay

10 Market Failures

Economic logic bolds that coercive public action may be justified if voluntary actions in unregulated markets fail to produce a desired good and government can redress the market failure (i.e., when government failures such as rentseeking and bureaucratic supply are not so great as to preclude redress of market failure). However, insofar as individuals possess adequate income, markets do not fail to provide food. What they sometimes fail to provide is the information or food attributes the household requires to produce health maintenance, protection, and rehabilitation. Information includes knowledge of bow particular food attributes contribute to maintaining and protecting health, as well as information about the marginal attribute contribution of specific foods. This section explores what some of these market failures might be, identifies the associated nonmarket goods for valuation research, and discusses bow changes in food production, food consumption, and scientific understanding of risks may create new valuation research issues.

Health Maintenance and Market Fallure Two types of information are needed for a household to choose its food consumption and preparation activities so as to maintain optimum health. One type of information is nutritional requirements for achieving a particular health state. The second is which foods in the market meet these nutritional requirements. These two types of information are complements since both are required to affect food choice and preparation. This section examines potential market failures in the production of information or food attributes that contribute to health maintenance. These market failures identify valuation research opportunities. Nutritional Requirements. In order to maintain health, households need to know which nutrients need to be consumed and what level of consumption results in health maintenance. This information may be more or less detailed and accurate. At a minimum, a household needs to know what foods must be consumed and in what quantities to produce health. An example of this type of information is the food pyramid developed by the USDA. More detailed information involves knowledge of specific nutrient requirements such as calories, carbohydrates, proteins, fats, vitamins, and minerals. Will an unregulated market produce and distribute the level of detailed information that · consumers are willing to pay for? good attributes. Once public bas information needs Human nutritional not diminish the quality does household one by information this produced, use of is a nonrival information needs nutritional so others, to or quantity available specific, so product not often is information needs nutritional good. Moreover,

Valuing Food Safety and NuJrition: The Research Needs

11

competitive pressures would be unlikely to encourage the production of this information. In addition, to the extent that the accuracy of the information cannot be verified by users, what incentives are there to produce accurate information? Indeed, there may be incentives for unregulated food sellers to produce inaccurate nutritional needs information to the extent that the demand for their product is affected negatively or positively by the information. The benefits of verification of nutritional requirements information are a nonmarket good because it is hard to exclude nonpayers from enjoying them. These points suggest that the market may fail to produce the efficient level of information on human nutritional needs, and thus the value of publicly sponsored research should get some attention from valuation researchers. To what extent is the nutritional requirements information being produced by public and private dollars? What safeguards are there to ensure accuracy of research results, and what is the value of accuracy? The value of verification of the healthfulness of nutrients such as beta carotene, vitamin B, and others is an area virtually untouched by economists. If levels of these and other nutrients are in question, what would be the value of additional public research? If public funds are not used, what incentives do private research providers face? Given continuing increases in health care costs, such questions should become more important in the future. However, to the best of my knowledge, economists have not examined the value of producing nutritional needs information and whether current levels of research funding are efficient. Another aspect of this issue is the distribution of nutritional needs information and its use by households. Obviously, only those households that receive the information can benefit from it. This raises the question of the value of publicly funded or mandated programs to increase understanding of nutritional needs by households. For example, how much benefit is to be gained from consumers knowing about the food pyramid or more detailed knowledge of nutritional requirements? Some research has been directed at these issues, particularly in terms of nutritional educational programs targeted at poor households (Senauer 1982). To establish the value of nutritional needs education programs, valuation research should determine the extent to which consumers' knowledge of nutritional needs affects food demands. For example, have consumers increased their consumption of foods high in fiber (Ippolito and Mathios 1991), beta carotene, and other compounds that have been found to provide health benefits? This type of research can be useful in marketing food products. For example, research on the marginal valuation of nutrients by households can help food producers design products that better meet consumers' preferences. There has been some work on this latter question (Capps and Sclimitz 1991, Eastwood et al. 1986, Ladd and Suvannunt 1976, Morgan et al. 1979), but more is needed, for example, on antioxidants, fiber, and other health-promoting nutrients. However, in the U.S., over-nutrition rather than under-nutrition may be the more relevant issue, and is discussed below.

12

Eileen 0. van Ravenswaay

Marginal Nutrient Content of Foods. The second piece of information households need to make wise food consumption and preparation decisions that maintain health is the marginal nutrient contribution of each food. Can the unregulated market be expected to produce and distribute this information? The production of nutrient content information has both public and private good aspects. To the extent that all raw ingredients (milk, meat, grains, vegetables) possess essentially the same nutrient content, this information has public good characteristics, and, indeed this kind of information has been publicly available to both producers and consumers. But the long held assumption that different varieties of a raw food do not differ substantially in nutritional content may no longer hold. For example, organic food is alleged to have higher nutritional content than conventionally grown foods. Animal drugs, such as pST, alter the leanness of pork. Biotechnology may result in foods that differ greatly in nutritional content. In these instances, would an unregulated market produce the nutritional information and would it be accurate? The same type of incentive problems would probably apply here as discussed earlier. Producers would have the incentive to overstate the nutrient level if they perceived it to increase the demand for the variety. Thus, there is a potential research issue on the value of nutritional content even for raw foods. Once the nutritional content of raw foods is produced, the question still remains whether the unregulated market would provide the information on raw products or whether consumers might be willing to pay for the information to appear on individual products. To the extent that varieties of raw foods do not vary in terms of nutrient content, provision of this information on raw foods has a public good quality. If one variety contains the same information as all others, and more than one variety is sold, there is no incentive for producers of the other varieties to incur the cost of the label. Thus, we would not expect nutrient labeling of raw varieties to occur. To the extent that consumers purchase particular foods in order to ensure that they consume minimum nutritional requirements, such information clearly has value. Consumers can buy books with the nutritional values listed for various raw foods. But it is unknown what consumers would be willing to pay for this information on each raw food package or how this value might change with the development of new varieties with altered nutritional characteristics In contrast, the nutrient content of processed foods depends on the ingredients, and that information is known only to the food processor. The theory of asymmetric information suggests that unregulated processors that provide high levels of desirable nutrients would have an incentive to provide this information, whereas unregulated processors producing foods with low levels of such nutrients would have the incentive to omit or inflate the information. Thus, consumers could to some extent distinguish between those foods that do or do not provide essential nutrients. Moreover, competitive pressures among firms would increase the amount and truthfulness of nutrient labeling of food (Ippolito

Valuing Food Safety and Nutrition: The Research Needs

13

and Mathios 1991). However, comparison between foods and the construction of a nutritious diet is difficult if producers do not use similar formats or can use misleading information. Thus, a public requirement of a particular format and enforcement of accuracy appears to be beneficial. The extent of the benefits of required labels is still a debatable question among researchers (Caswell 1992, Caswell and Padberg 1992, Daly 1976, Ippolito and Mathios 1991, Jacoby et al. 1977, Padberg 1992, Russo et al. 1986, Zarkin and Anderson 1992). The federal Nutrition Labeling and Education Act of 1990 has mandated standardiz.ed and complete nutrition labeling, and regulations promulgated under this act included a benefit-cost analysis. These labels require information on both daily nutritional requirements and the marginal nutrient contribution of a serving of the labeled product. The presumption in the previous paragraphs is that a fully informed, competent adult makes better health maintenance decisions than one who is not. Obviously, labeling of the marginal nutritional content of foods is not useful or valuable to people who are illiterate, unable to process the information, or are ignorant of the role of nutrients in protecting health. To the extent that a significant portion of the population falls into these categories, the regulated market fails in that individuals do not make the nutrient consumption decisions they would make if they were fully informed. In this case, either public programs to educate people about nutritional needs must be offered or the nutrient content of foods must be regulated. Indeed, U.S. food regulations still require nutrient fortification of bread, milk, and salt. Whether such regulations are still worth the cost is a potential valuation issue.

Health Protecti.on and Market Failure Three types of information are needed for a household to choose its food consumption and preparation activities so as to protect health from external haz.anls. One type of information is about potential haz.ards. For example, what substances are potentially hazardous to health and at what level of exposure'? The second type of information is which foods in the market contain hazardous attributes and to what extent. The third type of information is actions such as washing or cooking that households may take to avoid foodbome haz.anls. The first two types of information are complements, whereas the third type of information may in some cases be a substitute for the other two (e.g., thoroughly cook all chicken and pork). This section examines potential market failures in production of these three types of information and their associated food attributes. Health HatArds. In an unregulated market, households need to have information about the potential harmful effects of microbes, chemicals, and nutrients in order to make food consumption and preparation decisions that protect health. This information would include the effect of exposure to bacteria

14

Eileen 0. van Ravenswaay

such as Salmonella and environmental contaminants such as mercury in fish, and the level of exposure to nutrients such as fat, cholesterol, salt, sugar, caffeine, and alcohol that may harm health. Can the unregulated market be expected to produce this health effects information? AB it turns out, there are two different answers to this question, depending on the proprietary nature of the substance. Microbes, environmental contaminants, and excessive food nutrients are not proprietary. They may be present in foods or other vectors of human exposure, and food nutrients may be present in a wide range of foods. Thus, no one individual-producer or consumer-would be willing to produce the hamrd information when others enjoy the benefits for free, so production of information on the hamrds from microbes, environmental contaminants, and excessive intake of food nutrients is a nonmarket good. This naturally raises questions about the value of public production of food huard information. There appears to have been little work on the value of public funding of research designed to produce health huard information (Hammitt and Cave 1991 is one of the few studies), yet there would appear to be great gains from analyzing the value of public investment in this type of research. In contrast, pesticides, animal drugs, and food additives are proprietary. They are developed and sold for intentional use in food production. Thus, the question is whether the unregulated manufacturer has an incentive to develop health effects information. The answer is they do, but only up to a point. To the extent that exposure to any of these substances has acute health effects, human experience with any product containing amounts sufficient to produce acute effects would quickly spoil the market, assuming, of course, that tort laws would apply to recoupment of damages. However, to the extent that exposure would produce substantially delayed health effects for which it would be extremely difficult to hold manufacturers accountable for damages, unregulated manufacturers would not have the incentive to produce the health risk information. Indeed, this is why government regulations requiring producers to develop this information are in place. An interesting and important question is to what extent this form of regulation actually produces reliable information. There have been media reports of fraud in animal testing of the chronic effects of pesticides, for example, suggesting that it would be useful to conduct research on the value of controls which might reduce fraud or negligence. Even if the information on negative health effects is produced, it does not mean that it will be delivered to consumers or that they will be able to understand it. One of the key questions this raises is how effective the provision of information on the negative health effects of nutrients has been. Some studies have examined the effect of nutrition education programs on perceptions of the relationship between diet and health problems such as cardiovascular disease (Heimbach 1981, Levy et al. 1985). Economic research has focused on measuring the effects of negative information about nutrients such as fat and

Valuing Food Safety and Nutrition: The Research Needs

15

cholesterol on food consumption (Brown and Schrader 1990, Capps and Schmitz 1991, Chang and Kinnucan 1991, Ippolito and Mathios 1991, Potier 1987, Potier and Fruao 1991). I am not aware of any estimates of the value of public provision of negative health effects information, however, with the exception of programs targeted at the poor. Marginal Har.arrl Concentration. Suppose that health effects information were produced for both proprietary and nonproprietary food constituents. Can unregulated food markets be expected to distribute that information to consumers'? The answer to this question hinges on the usefulness of the information to the consumer, which in turn depends on the availability of information about whether and to what extent particular foods contain each of the potentially harmful constituents. In other words, it is not particularly useful to know that Salmonella may make you ill if you cannot determine which foods are contaminated with Salmonella. Thus, the question becomes whether unregulated food markets will produce and provide information about the extent to which particular foods contain microbial contaminants, chemical additives, and potentially harmful nutrients. The theory of asymmetric information suggests that the answer to this question is yes when consumers can link health problems to the food producer and no when consumers cannot make the link. In other words, since tort law will ensure that victims who can identify the guilty party will be able to make them pay, most accidents will be deterred, assuming that liability is sufficiently large. The question of how liability affects firm behavior has been addressed by some economists in the case of food safety (Caswell and Johnson 1991). However, estimates of the value to consumers of changes in liability for food safety are yet to be developed. Given recent cases of bacterial contamination of hamburger and dairy products this topic would seem to be a fruitful line for valuation research. One interesting question that arises in this area is the extent to which actual guilty parties can be identified given the increasing complexity of the food system. Because the answer to the question is no in the case where health effects are delayed, hard to prove, or too minor to justify individual legal action, government involvement in information provision may be justifiable. However, there is the question of whether the type of government involvement should be to require food labeling or to require the establishment and enforcement of standards. The distinction between these two types of approaches to dealing with food safety and nutrition issues is important because they produce different types of goods, and, thus different types of benefits. Labeling requirements produce consumer information, and the benefits of this public action are determined by the value of the information to consumers. In contrast the establishment and enforcement of safety standards involves the public production of health protection, and the benefits of this public action are determined by the value of this health protection action to households. This value depends on the

16

Eileen 0. van Ravenswaay

health protection practices households would take in the absence of government regulation. The question of whether the appropriate role of government is with respect to information or providing health protection actions has been answered differently for potentially harmful nutrients than it has been for food safety haurds. Nutrients must be labeled, but foods with potentially haurdous constituents have been subject to performance standards, process standards, and bans. The main reason for the difference in treatment appears to lie with the fact that potentially harmful nutrients are necessary for health at some level. Because of asymmetric information problems, standards for nutrient labeling have been established (Caswell and Padberg 1992). An interesting question here is to what extent such labeling can be expected to be honest. Will adequate private enforcement of truth-in-labeling be forthcoming? If not, how much should be invested in public enforcement? To my knowledge, few studies have been conducted on this issue (Ippolito and Mathias 1991). While there has been some research on consumer utilization of label information, only a few studies have sought to examine the value of that information (Morgan et al. 1979, Padberg 1992, Caswell and Padberg 1992). In contrast, most public programs with respect to microbes and chemicals provide health protection services (often called risk reduction) rather than requiring producers to provide consumers with information about product safety. The main justification for this approach is that it would be too costly to enforce the accuracy of the safety claims compared to directly controlling food handling practices and use of certain inputs such as pesticides. Consequently, a number of studies have sought to evaluate the benefits of public safety standards (Eom 1991, Gold and van Ravenswaay 1984, Hammitt 1986, Rae 1987, Roberts 1983, 1988, 1989, Roberts and Frenkel 1990, Roberts and Pinner 1990, van Ravenswaay 1992, van Ravenswaay and Hoehn 1991a, 1991b, 1991c). In the case of microbes, these standards are process or manufacturing standards that protect against microbe formation (note however that home food handling practices are addressed through public education programs). In the case of chemicals, these standards are generally tolerance levels that establish the safe dose. Other studies have sought to examine potential market implications of safety concerns (Jolly 1991, Malone 1990, McGuirk et al. 1990, Misra et al. 1991, Ott 1990, Ott et al. 1991, Weaver et al. 1992). Producer benefits also result from any increase in consumer food demand that may result from public provision of health protection services. The increased demand results in a larger producer surplus. For example, limits on haurds in food establish consumer confidence in the safety of the food supply and increase demands for foods where scandals would have otherwise occurred. In fact, modem federal food safety laws are often attributed to efforts of the meat industry to restore demand following publication of Upton Sinclair's novel

Valuing Food Safety and Nutrition: The Research Needs

17

The Jungle (Okun 1986, Sporleder et al. 1983). Several studies have attempted to estimate the impact of a loss of confidence on food demand in particular

markets and estimate the impact on sales. For example, time series analysis of market demand data covering the period preceding and following food scares have sought to determine the impact of government actions on sales, and in some cases, consumer and producer welfare (Brown 1969, Foster and Just 1989, Johnson 1988, SchuJrer et al. 1983, Smith et al. 1988, Swartz and Strand 1981, van Ravenswaay and Hoehn 1991a). Purchase intention data from surveys have been used to estimate the potential impact on product demand of utilizing new animal drugs such as bST and pST (Douthitt 1990, Florkowski et al. 1989, Halbrendt et al. 1989, Halbrendt et al. 1990, Halbrendt et al. 1991, Kaiser et al. 1992, Preston et al. 1991). Many improvements on these methods are still needed to increase their usefulness in both welfare analysis and marketing research (see section below). Future research needs will depend on developments in food production and processing such as bioengineering. Avoiding Hazards. It may be possible for consumers to reduce the marginal hamrd concentration in foods. For example, pathogens in food such as Salmonella or Trichina may be destroyed by thorough cooking and exposure to those pathogens prior to cooking may be reduced by proper food handling. Fat may be removed during cooking. Some pathogens or pesticide residues on raw fruits and vegetables may be reduced by thorough washing. Moreover, some hamrds in food may be created in the household and, thus, avoided by modified food handling. For example, refrigeration may be used to avoid growth of harmful bacteria, and proper heating can reduce bacterial contamination of home-canned and -frozen foods. Production and provision of information on how a household may take actions to avoid and prevent ha7.ards in food have public good attributes, unless they are associated with the purchase of hamrd-reducing goods by the household. In the latter case, private incentives to produce and/or provide such information exist because the information boosts demand for the hamrd-reducing good (e.g., use of bleach to sanitize cooking surfaces which reduces exposure to foodbome bacteria). Thus, unregulated production and distribution of this information may or may not be inefficient. This leads to questions about whether public production and distribution of this information is efficient, and, thus, to valuation research issues.

Implications of Changes in Food Production, Food Consumption, and Science By identifying how unregulated markets might fail to provide information and food attributes that households need to maintain or protect health, researchers can identify generic categories of questions that valuation of food safety and nutrition might address. However, the specific research issues relevant at any point in time depend on changes in food production and consumption

18

Eileen 0. van Ravenswaay

technologies, as well as scientific understanding of the relationship between food attributes and health (Roberts and van Ravenswaay 1989). New food production technologies may eliminate some existing hazards. For example, genetic engineering may make some pesticides unnecessary or change the nutrient content of some foods. Likewise developments in sensing and computing technologies may create new options for households in making dietary and food selection choices. But new hazards may be created as well as food markets become increasingly global and food more highly processed. These trends may make some existing food safety and nutrition programs unnecessary and new programs desirable, as well as alter the demand for food attributes and products. Both types of changes have implications for the type of policy and food marketing issues that valuation research should address. Some of these changes can be anticipated by examining developments in food production methods in particular industries and in food import trends. Other changes can be anticipated by examining developments in food consumption trends. Changes in science will also create new research issues. Much is still to be learned about the relation between diet and health. The possible positive and negative effects of various nutrients on health is still being investigated. Understanding of microbiological pathogens is still being developed. The range of health effects of chemicals is still full of mysteries. As discoveries are made, policy and food marketing implications will follow. These policy and food marketing implications can be anticipated by examining developments in knowledge of health effects, incidence of food contamination, and human exposure in each of the different areas comprising food safety and nutrition. Research Methods

mues

This section briefly reviews some of the methodological issues that underlie valuation of food safety and nutrition. In general, these issues are different depending on whether valuation research is addressing government regulation of market information and food attribute provision, government provision of nonmarket goods, or private marketing of information and food attributes. One set of issues in examining valuation of improved information is the separation between information on health effects and information on the marginal attribute contribution of individual foods. The value of the latter type of information depends on possession of the former type of information. For example, requiring nutritional labeling does little good if consumers know nothing about the effects of nutrients on health. However, much of the public campaign to educate consumers (e.g., the food pyramid) has sidestepped these issues by directing consumers to consume certain types of food (fruits and vegetables) and avoid others (red meat and fried foods). In other words, these

Valuing Food Safety and Nutrition: The Research Needs

19

programs instruct consumers on what types of health maintenance and health protection actions to take, rather than provide information that consumers can use to reach their own conclusions about which actions are best. Thus, there may be a connection in some consumers' minds between certain health risks and particular foods, rather than between food health risks and particular food attributes. This makes measurement of the benefits of improved nutrition labeling difficult since consumer behavior will be affected by both government actions. An important methodological issue in examining the value of public health protection services is estimating the value of avoiding the illness or death of statistically anonymous individuals in the protected population. One question is whether it is reasonable to use savings in ex post costs such as pain and suffering, lost wages, and the medical expenses of obtaining the maximum possible cure to approximate an ex ante value of preventing illness. A second question is whether reliable measures of the ex ante value of preventing illness or death can be obtained from hedonic or contingent valuation studies. A key issue here is what health risk perception motivates health protection actions by households in the actual or contingent markets. For many food safety and nutrition issues, morbidity is more an issue than mortality. Yet the economic valuation literature has concentrated on mortality (Fisher et al. 1989, Cropper and Freeman 1991, Viscosi 1993). Research in the food safety and nutrition area will have to develop new methods to address morbidity risks. Another important methodological issue in examining the value of public health protection services is identifying the range of benefits they provide. For example, reduced pesticide residues might have environmental benefits as well as food safety benefits. Public health protection services lower the cost to the household of providing its own health protection services. Reduced bacterial or pesticide contamination of foods means that households can dispense with certain preventative practices. Public health protection services also have benefits to food producers by increasing public confidence in the safety of their products. Measurement of these benefits involves examining the effect of public programs on changes in risk perceptions that ultimately affect food demand. Valuation research conducted to inform marketing research seeks to answer questions about market demand for certain product attributes (e.g., reduced pesticide residues, fat, or cholesterol) or methods of production (e.g., organic or biotechnology). Marketing research questions may focus on estimating willingness to pay for these attributes, the size of the market, the type of product desired by food consumers, or the identity of different consumer segments. Demand for product attributes may be motivated by an underlying consumer choice problem regarding health protection services or it may be motivated by a number of other factors. For example, the demand for organic food may be affected by product quality, travel costs, environmental concerns, or health concerns related to both pesticide residues and nutrients. Demand for milk from

Eileen 0. van Ravenswaay

20

cows treated with bST may be affected by concerns about animal welfare, the survival of small family farms, or health. Demand for reduced pesticide residues may be motivated by concerns about the environment, farm worker safety, or health. Because demand for product attributes may be based on a number of factors, it is difficult to specify an underlying household production or consumer choice model. However, given the nature of the marketing questions such research seeks to answer, it may be enough to learn, for example, that consumers are or are not willing to pay for an increase or decrease in a particular food attribute, and not necessary to know why consumers are willing to pay. In this case, a model of consumer choice of product attributes may be adequate, especially if consumer tradeoffs among attributes, such as quality versus safety, are of interest. Market research on food safety and nutrition issues currently falls into several categories. The largest category examines possible consumer reaction to new technologies such as irradiation, new animal drug use (e.g., bST and pST), and transgenic plants and animals. Another segment looks at the impact of concerns about fat and cholesterol on market demand for certain foods. Yet a third type of research has examined market demand for organic foods. Contingent valuation and related methods have been used to project consumer reaction to new food technologies, but little consensus has been achieved yet in how to design appropriate hypothetical market structures.

Conclusions Systematic assessment of research needs requires an understanding of the key questions which define a research field and the extent to which existing research has answered those questions. This chapter sought to develop a description of the key questions that research on the valuation of food safety and nutrition should address and discussed some of the research that has been conducted to answer these questions. The key questions were identified by first examining the general questions that existing valuation research have sought to answer. Three types of research areas were identified: (1) benefit-cost analysis, (2) program evaluation, and (3) food demand and marketing analysis. The first area has concentrated on estimating the benefits of changes in food safety and nutrition policy. The second research area has focused on estimating the effectiveness of existing programs. The third area has looked at the impacts of changes in health and risk information on demand for existing and new food products. Next, the chapter attempted to identify the types of issues that lead to the need for research in these three areas. The concept of market failures was used to identify generic ways in which our economic institutions create food safety and nutrition issues for consumers. To identify specific types of market failure,

Valuing Food Safety and Nutrition: The Research Needs

21

the chapter examined the household production framework that has been used by researchers to understand how human health concerns impact the economic behavior of households. Three different types of health production activities were distinguished: health maintenance, health protection, and health rehabilitation. Each of these type of activities affects the utility of the household differently. More importantly, each of the activities requires different types of inputs in terms of information about food or actual physical attributes of food. Unregulated markets can fail to efficiently provide these informational and physical inputs, thus, creating research issues on valuation of food safety and nutrition. For example, health maintenance activities require households to obtain certain types of food attributes, while health protection activities require the avoidance of certain food attributes. To obtain or avoid food attributes the household must have information about the health effects of food attributes needed to maintain or protect health and the marginal attribute concentration of each food. To avoid undesirable food attributes, the individual also needs information on actions that may be taken to reduce the marginal attribute concentration. The unregulated market may fail to provide efficient levels of food attributes or information. Market failure can occur because of the public good aspects of information as well as asymmetric information between producers and consumers. The specific ways that markets fail depend on the extent to which varieties of similar foods vary in attribute content and the degree to which a food is processed before reaching the consumer. The specific nature of market failures change over time as consumption and production technologies change. Furthermore, changes in scientific information can change the type of activities that households need to carry out in order to maintain or protect health. Thus, to identify future research needs, economists need to anticipate how changes in consumption, production, and science may affect the performance of food markets. To address the different categories of valuation research, methodological improvements are needed. One methodological issue is determining the type of good that we want to value. For example, the value of some types of information depends on the availability of other types of information. Another methodological issue is the validity of willingness to pay measures when consumers are not fully informed. For consumer behavior in actual or contingent markets to reveal willingness to pay, we need to be able to measure the food attributes that consumers believe they are valuing rather than assume we know what they are valuing. This is particularly important in consumer decisions involving health risks, and, thus, a key issue for valuation research on food safety and nutrition.

Note 1. The author wishes to thank Julie Caswell for helpful comments on this chapter.

Eileen 0. van Ravenswaay

22

References Asch, Peter. 1988. Consumer Safety Regulation: Puning a Price on Life and Limb. New York, NY: Oxford University Press. Bailey, Martin J. 1980. Reducing Risks to Life: Measurement of the Benefits. Washington, D.C.: American Enterprise Institute. Bentkover, Judith D., Vincent T. Covello, and Jeryl Mumpower. 1986. Benefits Assessment: 'Ihe State of the Art. Boston, MA: D. Reidel Publishing Company. Braden, John B. and Charles D. Kolstad, eds. 1991. Measuring the Demand for Environmental Quality. New York, NY: Elsevier Science Publishers B.V. (NorthHolland). Brown, Deborah and L. Schrader. 1990. Information on Cholesterol and Falling Shell Egg Consumption. American Journal of Agricultural Economics 72(1):48-55. Brown, Joseph D. 1969. Effect of a Health Hazard Scare on Consumer Demand. American Journal of Agricultural Economics 51:676-678. Capps, Oral, Jr. and J. D. Schmitz. 1991. A Recognition of Health and Nutritional Factors in Food Demand Analysis. Western Journal of Agricultural Economics 16(1):21-35. Caswell, Julie A. 1992. Current Information Levels on Food Labels. American Journal of Agricultural Economics 74(5):1196-1201. Caswell, Julie A. and Gary V. Johnson. 1991. Firm Strategic Response to Food Safety and Nutrition Regulation. In Economics of Food Safety, ed. Julie A. Caswell, 273297. New York, NY: Elsevier Science Publishing Company, Inc. Caswell, Julie A. and Daniel I. Padberg. 1992. Toward a More Comprehensive Theory of Food Labels. American Journal of Agricultural Economics 74(2):460-468. Chang, Hui-Shung and Henry W. Kinnucan. 1991. Advertising, Information, and Product Quality. American Journal of Agricultural Economics 73(4):1195-1202. Cropper, Maureen L. and A. Myrick Freeman III. 1991. Environmental Health Effects. In Measuring the Demand for Environmental Quality, eds. John B. Braden and Charles D. Kolstad, 165-211. New York, NY: Elsevier Science Publishers B.V. (North-Holland). Daly, P. 1976. The Response of Consumers to Nutritional Labeling. Journal of Consumer Affairs 10:170-178. Douthitt, Robin. 1990. Biotechnology and Consumer Choice in the Market Place: Should There be Mandatory Product Labeling? A Case Study of Bovine Somatotropin and Wisconsin Dairy Products. Presented at the Second International Conference on Research in the Consumer Interest, Snowbird, Utah (August 9-11, 1990). University of Wisconsin, Madison, Department of Consumer Science. Eastwood, David B., J. B. Brooker, and D. E. Terry. 1986. Household Nutrient Demand: Use of Characteristics Theory and a Common Attribute Model. Southern Journal of Agricultural Economics 17:235-246. Eom, Young Sook. 1991. Pesticide Residues and Averting Behavior: A Conceptual Framework for Analysis. Division of Economics and Business, North Carolina State University.

Valuing Food Safety and Nutrition: The Research Needs

23

Ferguson, Allen and E. P. LeVeen, eds. 1981. The Benefits of Health and Safety Regulation. Cambridge, MA: Ballinger Publishing Company. Fisher, Ann, L. G. Chestnut, and D. M. Violette. 1989. The Value of Reducing Risks of Death: A Note on New Evidence. Journal of Policy Analysis and Management 8(1):88-100. Florkowski, W. J., C. L. Huang, and Brian Goggin. 1989. Attitudes Towards Porcine Somatotropin: A Consumer Survey of the Atlanta Metropolitan Area. Georgia Agricultural Experiment Station, College of Agriculture, University of Georgia, Research Report 570. Foster, William and Richard E. Just. 1989. Measuring Welfare Effects of Product Contamination with Consumer Uncertainty. Journal ofEnvironmental &onomics and Management 17:266-283. Gold, Marion S. and Eileen 0. van Ravenswaay. 1984. Methods for Assessing the &onomic Benefits of Food Safety Regulations: A Case Study of PCBs in Fish. East Lansing, MI: Michigan State University, Department of Agricultural Economics, Agricultural Economic Report No. 460. Gramlich, Edward M. 1981. Benefit-Cost Analysis of Government Programs. Englewood Cliffs, NJ: Prentice-Hall, Inc. Halbrendt, Catherine, L. Sterling, C. Gempesaw, W. Florkowski, and C. L. Huang. 1989. Public Attitudes in the Northeast Region Toward Recombinant Porcine Somatropin. Journal of Food Distribution Research 20(1):153-164. Halbrendt, Catherine, W. Florkowski, L. Sterling, and C. Huang. 1990. Socioeconomic Determinants of Attitudes Toward the Use of Bioengineered Products in Food Production. Department of Food and Resource Economics, University of Delaware. Halbrendt, Catherine, Conrado Gempesaw, Richard Bacon, and Lesa Sterling. 1991. Public Perceptions of Food Safety in Animal-Food Products. Journal ofAgribusiness 9(1):85-96. Hammitt, James. 1986. Organic Carrots: Consumer Willingness to Pay to Reduce Food Borne Risks. Santa Monica, CA: The RAND Corporation R-3447-EPA. Hammitt, James K. and Jonathan A. K. Cave. 1991. Research Planning for Food Safety: A Value-of-Information Approach. Santa Monica, CA: The RAND Corporation. Heimbach, James T. 1981. Yesterday, Today and Tomorrow: Consumer Perceptions of Food Safety. Washington, D.C.: Division of Consumer Studies, Bureau of Foods, U.S. Food and Drug Administration. Ippolito, Pauline M. and A. D. Mathios. 1991. Information, Advertising, and Health: A Study of the Cereal Market. In &onomics of Food Safety, ed. Julie A. Caswell, 211-246. New York, NY: Elsevier Science Publishing Company, Inc. Jacoby, Jacob, R. W. Chestnut, and W. Silberman. 1977. Consumer Use and Comprehension of Nutritional Information. Journal of Consumer Research 4:119-128. Johnson, F. Reed. 1988. Economic Cost of Misinforming about Risk: The EDB Scare and the Media. Journal of Risk Analysis 8:261-269. Jolly, Desmond A. 1991. Differences Between Buyers and Nonbuyers of Organic Produce and Willingness to Pay Organic Price Premiums. Journal of Agribusiness 9(1):97-111.

Eileen 0. van Ravenswaay

24

Kaiser, Harry M., C. W. Scherer, and D. M. Barbano. 1992. Consumer Perceptions and Attitudes Towards Bovine Somatotropin. Northeastern Journal of Agricultural and Resource Economics 21(1):10-20. Ladd, George W. and Veraphol Suvannunt. 1976. A Model of Consumer Goods Characteristics. American Journal of Agricultural Economics 58:504-510. Lave, Lester B. 1981. The Strategy of Social Regulation: Decision Frameworks for Policy. Washington, D.C.: The Brookings Institution. Levy, Alan S., 0. Mathews, M. Stephenson, J.E. Tenney, and R. E. Schucker. 1985. The Impact of a Nutrition Program on Food Purchases. Journal of Public Policy and

Alarketing4:1-13.

Magat, Wesley A. and W. Kip Viscusi. 1992. Informational Approaches to Regulation. Cambridge, MA: MIT Press. Malone, John. 1990. Consumer Willingness to Purchase and Pay More for Potential Benefits of Irradiated Food Products. Agribusiness: An International Journal 6(2):163-178. McGuirk, Anya M., W. P. Preston, and A. McCormick. 1990. Toward the Development of Marketing Strategies for Food Safety Attributes. Agribusiness: An International Journal 6:297-308. Misra, Sukant K., C. L. Huang, and Stephen Ott. 1991. Consumer Wtllingness to Pay for Pesticide-Free Fresh Produce. Western Journal of Agricultural Economics 16(2):218-227. Morgan, Karen J., E. J. Metzen, and S. R. Johnson. 1979. A Hedonic Index of Breakfast Cereals. Journal of Consumer Affairs 6:67-75. National Academy of Sciences, National Research Council. 1989. Improving Risk Communication. Washington, D.C.: National Academy Press. Okun, Mitchell. 1986. Fair Play in the Alarketplace: The First Batlle for Pure Food and Drugs. De Kalb, IL: Northern Illinois University Press. Ott, Stephen L. 1990. Supermarket Shoppers' Pesticide Concerns and Wtllingness to Purchase Certified Pesticide Residue-Free Produce. Agribusiness: An International Journal 6(6):593-602. Ott, Stephen L., C. L. Huang, and S. K. Misra. 1991. Consumer Perceptions of Risks from Pesticide Residues and Demand for Certification of Residue-Free Produce. In Economics of Food Safety, ed. Julie A. Caswell, 175-188. New York, NY: Elsevier Science Publishing Company, Inc. Padberg, Daniel. 1992. Nutrition Labeling as a Policy Instrument. American Journal of Agricultural Economics 74(5):1208-1212. Preston, W. P., A. M. McGuirk, and G. M. Jones. 1991. Consumer Reaction to the Introduction of Bovine Somatotropin. In Economics of Food Safety, ed. Julie A. Caswell, 189-210. New York, NY: Elsevier Science Publishing Company, Inc. Puller, Daniel S. 1987. The Effect of Health Information on Shell Egg Consumption. University of California, Berkeley, Department of Agricultural and Resource Economics Working Paper. Puller, Daniel S. and Elizabeth Frazao. 1991. Assessing the Effects of Diet/Health Awareness on the Consumption and Composition of Fat Intake. In Economics of Food Safety, ed. Julie A. Caswell, 247-270. New York, NY: Elsevier Science Publishing Company, Inc.

Valuing Food Safety and Nutrition: The Research Needs

25

Rae, Douglas. 1987. Risks of Consuming Pesticide and Fungicide Additives: Perceptions and Behavior of Organic Food Consumers. Final Report to the U.S. Environmental Protection Agency Benefits Staff. (Douglas Rae, 36 Gage Street, Needham, MA 02192.) Ricardo-Campbell, Rita. 1974. Food Safety Regulation: A Study of the Use and Limitations of Cost-Benefit Analysis. Washington, D.C.: American Enterprise Institute for Public Policy Research. Roberts, Tanya. 1983. Benefit Analysis of Selected Slaughterhouse Meat Inspection Practices. University of Wisconsin, Agricultural Experiment Station, NC-117 Working Papers Series. Roberts, Tanya. 1988. Salmonellosis Control: Estimated Economic Costs. Journal of Poultry Science 67:936-943. Roberts, Tanya. 1989. Human Illness Cost of Foodborne Bacteria. American Journal of Agricultural Economics 71(2):468-474. Roberts, Tanya and Eileen van Ravenswaay. 1989. Effects of New Scientific Knowledge on Food Safety Policy. Agricultural-Food Policy Review: U.S. Agricultural Policies in a Changing World. Washington, D.C.: Economic Research Service, U.S. Department of Agriculture, Agricultural Economic Report No. 620. Roberts, Tanya and R. Pinner. 1990. Economic Impact of Disease Caused by Listeria monocytogenes. In Foodborne Listeriosis, eds. A. J. Miller, J. L. Smith, and G. A. Somkuti, 137-149. Amsterdam, The Netherlands: Elsevier Science. Roberts, Tanya and J. K. Frenkel. 1990. Estimating Income Losses and Other Preventable Costs Caused by Congenital Toxoplasmosis in People in the United States. Journal of the American Veterinary Medicine Association 96(2):249-256. Russo, J. Edward, R. Staelin, C. A. Nolan, G. J. Russell, and B. L. Metcalf. 1986. Nutrition Information in the Supermarket. Journal of Consumer Research 13:48-69. Schuker, R. E., R. C. Stokes, M. L Stewart, and D. P. Henderson. 1983. The Impact of the Saccharin Warning Label on Sales of Diet Soft Drinks in Supermarkets. Journal of Public Policy and Marketing 2:46-56. Senauer, Benjamin. 1982. The Current Status of Food and Nutrition Policy and the Food Programs. American Journal of Agricultural Economics 64(5):1009-1016. Smith, M. E., E. 0. van Ravenswaay, and S. R. Thompson. 1988. Sales Loss Determination in Food Contamination Incidents: An Application to Milk Bans in Hawaii. American Journal of Agricultural Economics 70(3):513-520. Sporleder, Thomas L., Carol Kramer, and Donald J. Epp. 1983. Food Safety. In Federal Marketing Programs in Agriculture: Issues and Options, eds. Walter J. Armbruster, Dennis R. Henderson, and Ronald D. Knutson, 269-304. Danville, IL: The Interstate Printers and Publishers. Swartz, D. G. and I. E. Strand, Jr. 1981. Avoidance Costs Associated with Imperfect Information: The Case of Kepone. Land Economics 57:139-150. van Ravenswaay, Eileen. 1992. Public Perceptions of Food Safety: Implications for Emerging Agricultural Technologies. In Volume 2: A New Technological Era for

American Agriculture-OTA Commissioned Background Papers, Part E: Food Safety and Quality, ed. U.S. Office of Technology Assessment. Springfield, VA: National

Technical Information Service. van Ravenswaay, Eileen 0. and John P. Hoehn. 1991a. The Impact of Health Risk on

26

Eileen 0. van Ravenswaay

Food Demand: A Case Study of Alar and Apples. In Economics of Food Safety, ed. Julie A. Caswell, 155-174. New York, NY: Elsevier Science Publishing Company, Inc. van Ravenswaay, Eileen 0. and John P. Hoehn. 1991b. Contingent Valuation and Food Safety: The Case of Pesticide Residues in Food. East Lansing, MI: Department of Agricultural Economics, Michigan State University, Staff Paper No. 91-13. van Ravenswaay, Eileen 0. and John P. Hoehn. 1991c. Willingness to Pay for Reducing Pesticide Residues in Food: Results of a Nationwide Survey. East Lansing, MI: Department of Agricultural Economics, Michigan State University, Staff Paper No. 91-18. van Ravenswaay, Eileen 0., Jennifer Wohl, and John P. Hoehn. 1992. Michigan Consumers' Perceptions of Pesticide Residues in Food. East Lansing, Ml: Department of Agricultural Economics, Michigan State University, Staff Paper No. 92-56. Viscusi, W. Kip. 1993. The Value of Risks to Life and Health. Journal of &onomic Literature 31(4):1912-1946. Weaver, Robert D., David Evans, and A. E. Luloff. 1992. Pesticide Use in Tomato Production: Consumer Concerns and Willingness to Pay. Agribusiness: An International Journal 8(2):131-142. Zarkin, Gary A. and Donald W. Anderson. 1992. Consumer and Producer Responses to Nutrition Labeling Changes. American Journal of Agricullural &onomics 74(5): 1202-1207.

2 Self-Protection, Risk Information, and Ex Ante Values of Food Safety and Nutrition Young Sook Eom 1 Over the past two decades, consumers have become increasingly concerned about the safety of available food and the potential linkage between diet and health. Traditionally, economists have used observed purchase behavior and tradeoffs that consumers make in the marketplace as a basis for inferences about consumers' preferences for and (iinplicit) values of certain food products. However, health risks and nutritional content are only part of bundles of attributes characterizing food products. That is, health risks and nutrition are nonmarket goods without directly observable price components. Therefore, the conventional consumer demand analysis may not be directly applicable in measuring the values that consumers place on changes in food safety or nutrition components. In these circumstances, a natural tendency of economic modeling strategy is to establish some linkage between the nonmarket goods to be valued and observable market goods or private actions. Indeed, currently available empirical evidence suggests that consumers' concerns about food-related risks seem to motivate them to undertake some types of protective actions to reduce health risks (Swartz and Strand 1981, Smith et al. 1988, Foster and Just 1989, van Ravenswaay and Hoehn 1991a, Brown and Schrader 1990, Putter and Frazao 1991, IppolitoandMathios 1990, Hammitt 1986, Rae 1987, Zellner and Degner 1989, Ott et al. 1991, van Ravenswaay and Hoehn 1991b, Conklin et al. 1992, Born 1994). While these studies provide evidence on consumers' revealed or stated preferences for safer and healthier food, most of the valuation research on food safety and nutrition has failed to incorporate consumers' risk perception processes into the behavioral framework. In studies investigating consumers' aggregate responses, risk information was measured using crude proxies such 27

28

Young Sook Eom

as number of news accounts appearing or dummy variables. On the other hand, the studies of individual responses assumed that consumers were aware of the technical estimates of risks with the exception of van Ravenswaay and Hoehn (199lb) and Eom (1994). Unfortunately, the objective measure of food-related risk is not likely to be known exactly, even to scientific experts. Indeed, prominent psychologists such as Slovic et al. (1985) argue that there is no objective risk. In their view, all risks are subjective, whether judged by experts or lay people. In the formation of subjective risks, however, consumers often have imperfect and incorrect information about the event at risk and seem to be influenced by factors that are different from those influencing experts' risk assessments (Arrow 1982, Viscusi and Magat 1987, van Ravenswaay 1991). Recognizing this potential source of market failure, public information programs-prod uct labeling or hazard warning-are alternative policy options for government regulations concerning food safety and nutrition. Growing empirical evidence suggests that consumers can learn about risks and update their risk perceptions after receiving new information (Viscusi and O'Connor 1984, Smith and Johnson 1988). Therefore, a more plausible model describing consumers' behavior in the presence of food-related risks should begin with a formal analysis of the risk perception process. This chapter proposes to develop such a framework when consumption decisions must be made with incomplete information about food product attributes but self-protection actions to reduce adverse health effects are available. In terms of the three classes of household health production activities van Ravenswaay identified in the previous chapter (health maintenance, protection, and rehabilitation), the theoretical development in this chapter mainly focuses on household health protection activities to avoid harmful or hazardous exposure to contaminants, residues, or nutrients. When we interpret consumers' protective responses to information, it is important to distinguish between two different perspectives of valuation measures-ex ante and ex post evaluation. Consumers' protective decisions focus on the probability distribution of health effects, not on the reali:zation of health outcomes, arising from the consumption of certain food attributes. On the other hand, consumers' behavioral adjustments are undertaken after receiving (or acquiring) new information about food attributes. Hence, the valuation measures that will be derived in this chapter are ex ante measures with respect to health risks but ex post with respect to information about food attributes such as food safety or nutrients. The next section begins by reviewing the conventional expected utility theory as a description of consumer purchase behavior with uncertain product attributes. However, our framework will modify the expected utility theory to address criticisms of the conventional framework (Arrow 1982, Slovic et al. 1985) and will extend it to a self-protection model, allowing adjustments to risks through averting behavior. Subsequently, we introduce a subjective self-

Self Protection, Risk Information, and F.x Ante Values

29

protection model by treating the probability function as a subjective risk perception function based on available information. This formulation is static in the sense that learning is not taken into account. Finally, we introduce multiple time horizons into the subjective self-protection model to develop an integrated conceptual framework of risk perceptions, learning, and self-protection actions. This section also illustrates how information and self-protection can be incorporated into a Bayesian learning model. The last section provides some concluding remarks.

Expected Utility Theory and Ex Ante Value for Risk Reductiom This section outlines a model that describes a household's preferences in the presence of uncertain food quality, and uses it to derive monetary measures of willingness to pay for reductions in health risks before the quality uncertainty is resolved. A household's food purchasing decisions are made for all members of the household, so health risks from food consumption affect the entire household. However, following Becker's (1974) argument, we assume that a household decision process is the same as that of the household head, so that there exists a household preference that reflects all members' tastes. Suppose a representative household allocates its given income over the primary food item (X) and other composite goods (Y). While the household observes some food attributes (such as color, si:ze, shape, or freshness) prior to purchase, it does not know which particular "state of health" will actually occur from the consumption of X. For simplicity, the household is assumed to face only two states of the world-either the occurrence or nonoccurrence of adverse health outcomes. If it consumes the food item (X) suspected to contain harmful substances (residues or nutrients) over its lifetime, there is a probability 'Ir of the adverse health outcome. Because heath effects identified in the previous chapter (for example, getting cancer or having a heart attack) often involve unique and irreplaceable losses, the extreme characterization of health outcome seems reasonable. In these cases, the household is assumed to evaluate consumed food differently depending on the health outcome, implying state dependent preferences (Cook and Graham 1977). Thus, the household's state-dependent utility function can be defined as Ub(X, Y) if the health event occurs, and as Ug(X,Y) if it does not occur. Given the above assumptions, household consumption behavior can be described as maximizing the following expected utility: (1)

Max EU = (1 - T) Ug(X,Y) + 7rUb(X,Y). X,Y

In the expected utility framework, the probability of the adverse health outcome,

Young Sook Eom

30

11" (i.e., the risk) is known or objectively given to the household (or the household acts as if it were able to attach the exact probability to various possible states of the world). Thus, the risk, 71", is considered exogenous information over which the household has no control. Following Becker (1965), we assume that the household faces a "full income" budget constraint, M. The full income constraint is defined as: (2)

M

= wT + A = (mx

+

wtx)X

+

Y

= p)(

Y

+

where:

Illx: money price of goods X relative to price of Y,

w: wage rate, 1x: time spent for X relative to time spent on Y and assumed to be constant, Px: "full" price of X relative to "full" price of Y, T: time endowment, and A: non-wage asset income. Since a household is assumed to make consumption decisions consistent with its objective-maximizing expected utility given the budget constraint-solving the constrained maximization problem, equations 1 and 2, yields the following state dependent indirect utility functions: (3)

where Vb(M•Px) = Ub(X*(M,px), M - X*(M,px)); Vg(M,px) = Ug(X*(M,px), M - X*(M,px)); and X* denotes an optimally chosen level of the commodity X. The state-dependent indirect utility functions in equation 3 are assumed to be well behaved and to satisfy the usual properties such as nondecreasingness in Y and convexity in Px (()V/()M > 0 and ()V/()px < 0, i = b,g). It is important to note that risk, 11", is treated as a parameter, like price (px) and income (M) in the expected utility function. Hence, marginal willingness to pay for changes of the exogenous risk can be derived by taking the total differential of the expected utility function, equation 3. By setting dEU equal to zero and setting dpx equal to zero, we can solve for the income change that would need to be taken away from the household in response to an exogenous risk reduction if the two changes are to keep expected utility constant: (4)

aM ()7["

vb - vg

11"avb1aM + (1 - 11")1aV/iJM

=MRS

TM

.

The left-band side of equation 42 represents a gradient of the willingness to pay (WTP) risk schedule, which is equal to the marginal value of risk reduction.

Self Protection, Risk Information, and Ex Ante Values

31

The right-hand side of equation 4 is the marginal rate of substitution between 7r and M (MRS..-M), which is the difference between state-dependent utility functions divided by the expected marginal utility of income. However, it represents an a ante tradeoff between income and risk because the household must reveal its value for risk changes before it experiences the adverse health outcomes. Equation 4 implies that the a ante marginal willingness to pay (MWTP) for the risk reduction is equal to the a ante marginal rate of substitution between income and risk. Note that equation 4 was commonly measured in job risk cases to represent the risk-dollar tradeoff selected by a worker (Viscusi 1979). For the case of the risk of premature death, this ex ante MRS..-M represents the implicit value of one's statistical life. For the case of nonfatal injury, the rate of tradeoff represents the implicit value of per unit risk of injury.

Self-Protection Model and Ex Ante Value of Risk Reductions The situations considered in our analysis-consuming chemically contaminated food or intake of fat or cholesterol-can be viewed as a case in which a household can take actions to reduce the food risks. For example, households could change their food preparation methods (i.e., spend more time in cleanup or cooking), or could decrease the consumption of the suspected food items and eventually shift to food items that are viewed to be safer and healthier. 3 In addition, households could engage in some preventative health behavior such as having a cholesterol test done or visiting doctors regularly. However, given the uncertainty of ultimate health effects, any actions undertaken by a household cannot yield a certainty of protection, but can only provide the reduction of the probability of the adverse health outcomes (i.e., the risk). The recognition that risks can be affected by a household's action stimulated Ehrlich and Becker (1972) to develop the self-protection model. They argued that self-protection activities undertaken by a household would shift the whole probability distribution to the left to reduce the probability of adverse health outcomes and raise the probability of favorable health outcomes. Therefore, the household risk assessment was described as a function such as 7r = l/t(v), where v is a vector of self-protection actions undertaken. Our framework analyzing food purchase decisions follows Ehrlich and Becker's view and treats self-protection as a type of averting behavior. Thus, the expected utility framework discussed above is modified to describe the selection of self-protection actions undertaken (v) and how the self-protection actions would affect the probabilities (l/t(v)). Incorporating the opportunity of self-protection, the household aims to select the level of self-protection, v, as well as X and Y to maximiz.e its expected utility as shown earlier: (S)

EU

= (1 - l/t(v))

Ug(X,Y) + ~(v)

Ub(X,Y).

32

Young Sook Eom

The health risk, 11', is assumed to decrease as the household increases selfprotective activities (aTlitv < 0). In addition, the household is assumed to know how its self-protection actions would affect the risk function, 1/t(v), as Ehrlich and Becker (1972) described. With the above assumptions, the health risk became endogenous in the household decision process but still remains "objective• information in that the functional relationship of the risk, 1/t(v), is known to the household. The above description of the risk function allows the self-protection model to adopt the household production framework in the same way as averting behavior models (Gerking and Stanley 1986, Dickie et al. 1987, Berger et al. 1987). Risk, T, is equivalent to a final service flow produced using a vector of self-protection, v. Self-protection (v) may be considered to be household activities combining the householder's time and other resources (for example, a change in cooking preparation practices), and the purchase of nondurable products or services as essential inputs of the production process. Selfprotection is not a direct source of utility. It only serves to reduce the health risk (i.e., nonjointness in the household production). 4 Including the "full" price ofv, Pv• the household's "full income• budget constraint can be modified from equation 2 to equation 6:

(6)

M

= wT + A = (mx

+ wtx)X + Y + (mv + wtv)v

= p){ + Y + PvV.

In equation 6, the opportunity cost of a household's time spent on v, X, and Y is assumed to be equal to the market wage rate. The household production framework enables us to look at a household's choice problem in a two-stage decision process (Dickie et al. 1987, Deaton and Muellbauer 1980). In the first stage, the household is minimizing self-protection expenditures to obtain a given level of health risk such as:

(7)

C(pv,11"") s.t.

= C = min PvV T0 =

t/t(v).

This expenditure function has properties similar to a firm's cost function. It is positive, homogeneous of degree one, and concave in Pv· By Shephard's Lemma, the conditional demand function for self-protection is: (8)

- ac(pv,T°) apv

V -

=v(p , aO\I• v

33

Self Protection, Risk Information, and Ex Ante Values

In the second stage, expected utility is maximized subject to the budget constraint that induces self-protection expenditures from the first stage. Substituting equation 7 into equation 5, the Lagrangian is defined as:

(9)

L = [1

U1(X,Y)

- ~(C)]

Ub(X,Y)

+ ~q

+

µ[M - p){ - Y - CJ.

The optimally chosen self-protection expenditure, C, is to reduce risk but also to reduce income left over for consumption. Maximizing the expected utility for C, X, and Y subject to the budget constraint bas the first order conditions:

a'trtac cub - u,> - "

(10)

=

o

(11)

cc1 - 'tr) au,tar + 'trau~YJ

(12) (13)

-"

=

o

M - p){ - Y - C = 0.

Combining these equations gives:

(14)

ac a'tr

= =

1

ca'tr1aq

=

(Ub - U1) µ

(Ub - U1 ) cc1 - 1r)(au,taY) + 1r(au,1aY>1

= MRS.,;y·

The first term in equation 14 is the marginal self-protection expenditure which is observable. The last term is the ex ante marginal rate of substitution between 'tr and Y (MRS.,;y) which is unobservable. Because of the assumption of the linear budget constraint and the normalized price of Y, the marginal utility of a composite good Y is equivalent to the marginal utility of income (i.e., aU/aY = aU/aM) in each state of the world (i = g,b). This leads MRS-rv to be equal to MRS-rM· Hence, equation 14 can be rewritten as equation 15:

(15)

ac a'tr

=

cub - u,>

. , . .cc,_. 1-_-'tr...... )(---au.....,..... ,a...... M)-+-1r..... (a---u,-1a...-M)-J

The right-band side of equation 15 is exactly the same as that of equation 4. That means we can obtain equation 16 from equations 14 and 15:

34

Young Sook Eom

aM aT

(16)

=

ac ac(pv, T") aT = __a_T_

Equation 16 implies that the ex ante marginal willingness to pay for risk reduction is equal to the marginal self-protection expenditure of risk reduction. In other words, the marginal benefit of the reduction in risks is equal to the marginal cost of producing the same level of T, by increasing the use of v. For example, the willingness to pay for reduced risks from Salmonella contamination can be measured by the cost of additional time that consumers are willing to spend to prepare chicken more safely (Zellner and Degner 1989). In this framework, the estimation of marginal values of risk reductions requires only knowledge of household production technology and the price of protective behavior, Pv• not of unobservable households' preferences (see Smith 1990 for a more formal discussion of the household production framework). The second order condition to assure a maximum of expected utility requires that: ~(U !12

b

ac2

-U)2~

!I

ac

g

[au _b aM

au ] aM

__ g

+

(17)

[ (1

-

a2u aM2

'Ir) _ _ g + 'Ir

~ a2u]

aM

< 0.

In addition to the three assumptions made about the state-dependent utility functions, the only restriction required to guarantee inequality in equation 17 would be (a2TlaC2) > 0. It is noteworthy that derivation of the second order condition does not necessarily require risk averse preferences. Even though the marginal utility of consumption is increasing (i.e., a2U/aM2 > 0, i = b,g}, there may be cases in which inequality still holds. The desire to undertake self-protection might occur for risk lovers as well as for risk averters.

Subjective Self-Protection Model and Value of Information In the conventional expected utility framework and self-protection model, the probability or probability function of the adverse health outcome was treated as "objective" information in that any household facing the same problem will assign the same probability. However, the uncertain situation causing the adverse health outcomes is often unique and nonrepetitive. So there is little opportunity to gain the experience that is usually associated with learning.

Self Protection, Risk Information, and F.x Ante Values

35

Although the risk of short-term acute health problems due to pesticide poisoning, for example, is relatively well understood, the risk of health effects posed by long-term and low-level exposures to food contaminants (such as pesticide residues) or harmful nutrients (such as fat) are not as well known. We cannot assume that any individual, whether an informed consumer or a professional toxicologist, knows the technical risk or how it will respond to averting decisions. Nonetheless, it is reasonable to expect that a household will have subjective probabilistic beliefs. Any assignment of subjective probability is permissible, in principle, provided there is coherence in a household's judgement about the relative likelihood of various values of unknown states of the world (Winkler and Hays 1975). Each household may perceive a different degree of subjective risks according to its demographic background, knowledge about the event at risk, and past experiences with similar situations. These factors will serve as a set of information, I, to the household in the process of forming risk perceptions at a point in time. Thus, incorporating information, a household's subjective risk perception can be defined as: (18)

11"

= ¥-(v,I)

where v denotes a vector of self-protection actions, and I represents information available to the household. It is assumed that information, I in equation 18, is exogenously provided and thus is not subject to the household's choices and does not explicitly enter the budget constraint. Therefore, while risk perceptions become endogenous outcomes, information is still considered an exogenous factor in the household decision process. At the beginning of each period, a household is assumed to make selfprotection decisions and consumption plans X and Y, given a set of available information about the uncertain event. A household's objective function can be written as: (19)

EU

= (1

- ~v,I)

U8(X,"Y) + ¥-(v,I) Ub(X,Y).

Following the analysis discussed earlier, the household's constrained expected utility maximi7.ation problem can be stated equivalently as equation 20 in terms of state-dependent indirect utility functions to yield:

where Vb(M,px) = Ub(X*(M,px), M-X*(M,px}); V1(M,px) = U1(X*(M,px), = ~v v*(pv,l,r>)); and X*, Y*, M - X*(M,px)); 1/l(J>y,I) = ~C*(pv,lr>) and v* represent the optimally chosen levels of X, Y, and v.

36

Young Sook Eom

Because of the exogeneity of information at each period of consumption choice, marginal willingness to pay for additional information again can be derived by taking the total differential of the expected utility function, equation 20. The change of information that we consider is not complete but partial in the sense that the information affects households' risk perceptions while still leaving some uncertainty present. By setting dEU = 0 and holding dpx = dpv = d'lr = 0, we can solve for the income change that would be required in response to exogenous additional information to keep expected utility constant: (21) Compared with the valuation measure of risk reductions derived in equation 16, the expressions in equation 21 measure the marginal willingness to pay for information about health risk, reflecting individuals' incomplete knowledge about the risk. Consumers' risk perceptions are endogenously determined through the household health production activities in response to an exogenous change in information about the risk. In this subjective self-protection model, what consumers are evaluating is not food attributes (such as health risks) but information about the food attributes. Thus, the valuation measure of additional information in equation 21 captures both the direct effect of information on risk perceptions (a'lrlal) and indirect effects through marginal values of changes in risk (MRSrM).

to marginal Using the results of equations 14 and 15, which equate MRS~' self-protection expenditure, the specification of equation 21 can be reduced to equation 22: (22)

=

ac

a1 ·

The first and last terms in equation 22 state that the marginal value of additional information equals the marginal cost of information in terms of the reduction in self-protection expenditures. Since in the household production framework, whatever level of risk perception chosen in the expected utility maximization process must be produced at minimum cost, self-protection expenditure consists of C* = Pv v(pv,I, 'Ir). Taking the total differential of the self-protection expenditure function C*, the third term of equation 22 can be re-stated as equation 23: (23)

a'lr1a1

= a'lrlav

Pv

ac

= al .

Self Protection, Risk Information, and Ex Ante Values

37

Substituting equation 23 into equation 22 gives the expression for the marginal value of information as: (24)

aM

M "'

a1r1a1 ac a1rlav Pv "' al .

Equation 24 shows that the ex ante marginal willingness to pay for additional information is equal to the marginal cost of achieving the same level of risk perception, 'Ir", by increasing the self-protection expenditure C. 5 Again, this ex ante MWTP expression does not require that we observe the ex ante MRS but can be derived with knowledge of the technical relationship between information and self-protection actions in the risk perception functions (see Gerking and Stanley 1986 for parallel results from averting behavior models with certainty). Marginal WTP for new information in equation 24 deserves further explanation. First, a household's ex ante marginal willingness to pay will be higher as its •full• price ofv, Pv (mainly the opportunity cost of its time) is higher and its marginal productivity of v in risk perceptions (a1rliN) is lower. Equally important, ex ante MWTP will be higher for those households in which new information has greater impact on risk perceptions (a1rlal). Based on the result of equation 24, a more educated and/or more experienced household may express lower WTP in response to new information about health risks. This possible connection between individuals' demographic profiles and the acquisition of and use of information has been recognized by economists for some time. For example, Grossman (1972) hypothesized that schooling increases the efficiency of household health production, and therefore that better educated individuals may react to risk information differently from less educated people. Kenkel (1991) empirically found that education levels reflected in the number of years of schooling helped individuals to undertake more preventive actions by improving their knowledge of the relationship between protective behavior and health outcomes.

Self-Protection Model with Learning and Value of Information The analysis developed above incorporated self-protection and available information into the risk perception process. But the process still is •static"; it gives an account of effects of information on risk perceptions at a given point in time but does not describe how a household acquires and uses the information over time. Jn practice, the household takes self-protection actions while it acquires more information (through product labeling or new media reports) and learns about the risk. Jn this situation, the household's objective function at any time period t = i is equal to equation 25:

Young Sook Eom

38

The difference between equations 19 and 25 is that the acquisition of information in equation 25 is a part of the household's optimizing choice at a particular point in time, whereas information available in equation 19 was exogenously given. Since risk perceptions are endogenously determined, observed outcomes at time t = i-subjective risk perceptions and household behavioral decisions (Xi*, Yi*• and vi*)-reflect influences of both the acquired information and the feasibility of self-protection. If so, it will be a difficult, if not impossible, task to sort out the effects of information and self-protection on risk perceptions and behavioral decisions in a timeless expected utility framework. Hence, some restrictions on this integrated framework are required to separate the relative influence of acquired information and self-protection on risk perceptions from that on averting behavior decisions. Before proposing an integrated framework, consider first a simple Bayesian learning model in which households update risk perceptions by observing additional information in the form of new labeling over time (Viscusi 1989). For the sake of simplicity, we formulate the risk perception process only before receiving new information about food-related risks (i.e., t = 1) and after receiving new information (i.e., t = 2). After receiving new labeling information at t = 2, the household will update its risk perception, which can be described as a reduced form: (26) where

?r2: a household's perceived risk after receiving information, 11"1 : the perceived risk before receiving information, 11"s: sample risk inferred from risk message, ai (i = 1 ands): the weights for the risk perceptions.

The posterior risk perception in equation 26 is a weighted average of prior risk, ?r 1, and "sample" risk, 11"s. 6 The weights capture the household's assessment of the relative precision of the underlying true distribution of the risk. The Bayesian updating rule implies that the a 1 would be positive and the as would be 0 < as< 1. To develop an integrated modeling of averting behavior and learning over time, consider one way self-protective actions that can be incorporated with the Bayesian learning framework described as equation 26. Incorporating selfprotection into the Bayesian learning framework, the household's prior risk assessment at t = 1 would become ift(v 1), where v 1 is a vector of self-protection undertaken at t = 1.

Self Protection, Risk Information, and Ex Ante Values

39

Assume that the functional relationship 1/t(v) is known and processed recursively. As Crawford (1973) has shown, the recursive notation of the information set at time t = 2, 12, can be written as follows: (27) where 11'1 again denotes the "implicit" sample risk obtained through the product label, 11 designates a set of information available at time t = 1, and v 1 denotes the level of self-protection chosen at t = 1. The h(.) function in equation 27 can be interpreted as an updating rule. With new labeling information similar to that hypothesized to underlie equation 27, posterior risk perception is determined to be: (28) As we see in equation 28, the household's new information would alter its risk perception. It does not affect its perception of the effect of self-protection on the parameters of risk assessment. In this specification, learning becomes a part of a household's decision-making with risk. However, it is still separately processed from the household's behavioral decisions (i.e., "exogenous" learning). In other words, a household's decision-making with learning becomes a sequential process; the amount and framing of information lead to revisions in a household's risk perception to 11'2• Then, the household makes self-protection decisions, v2, using 12• With this background, we now attempt to develop an integrated framework describing the interaction between risk perceptions, learning, and behavioral decisions. To link the information acquisition and leaning processes, the selfprotection model developed earlier is extended to a two-period context in this section. With the extended time horizon for decision making, a more explicit consideration can be given to the way posterior risk perceptions are influenced by information as well as self-protection actions, and how learning takes place over time. Households' preferences are still represented by the von NeumannMorgenstem utility function, U(X 1, Xv, where Xi = (Xil•Xa• .. ·Xin>• t = 1,2 is a vector of disaggregated consumption goods at time t = 1,2. Thus, statedependent utility functions, U,(X 1,Xv and Ub(X 1,Xv, in this final model combine Cook and Graham's (1977) single-period state-dependent utility function with Epstein's (1975) two-period specifications. To implement the model, several assumptions must be made: first, risk perception, 11'1, and optimal averting expenditure, C1, at t = 1 are assumed to be known. So, when behavioral response to uncertainty takes place in the second period, 11'1 = 1/l(C1) is used as a prior risk perception in the posterior risk

40

Young Sook Eom

perception, T 2• Second, second-period price and income are known with certainty. Household savings, S, result in certain yields (1 + r)*S at t = 2, where r is an interest rate. Third, households still are assumed to be engaged in a twostage decision process according to the household production framework. The objective of the two-period model is to investigate how uncertainty about risk accompanied by the opportunity for learning influences the optimal consumption plans and self-protection decisions. Households in this framework select optimal levels of averting expenditure, Ci. levels of savings, S, and consumption levels for X1 and X2. Following Epstein (1975) and Chavas et al. (1986), the household's expected utility maximi7.ation problem can be written as: (29)

where E1 denotes the expectation operator conditional on information available at time t = 1. The household's budget constraint at each period would be:

at t = 1

(31)

at t = 2

(32)

where M1 and M2 are the household's full income at t = 1 and t = 2, respectively, and t = 1,2 is a vector of prices of goods including time costs as

Px/ ,

well as money prices. In the first period, the household has imperfect knowledge about the risk but has subjective prior beliefs. If the household has an opportunity to undertake self-protection, c1, and acquires new information through product labeling or public provision, the household will form the central tendency of the distribution of posterior perceived risk based on a Bayesian framework. A backward induction method is used to solve this sequential problem (DeGroot 1970). If a household receives new information about food-related risks during the first period, then its risk perception would be updated according to equation 28. Because of the "exogenous" learning process structured in equation 28, the set of information available when the household makes choices over X2 and Ci at t = 2, 12, can be treated as an exogenous factor. So, the second-period choice problem becomes:

Self Protection, Risk Information, and Ex Ante Values

(33)

41

EU2 = Max (1 - Ti) UgCX1,Xi) + T2 Ub(X1,Xi) X2,C2

(34)

(35) Note that the choice of v2 in equation 32 is converted to the choice of Ci in equation 34, because our framework is still based on the household production framework. To take advantage of the interrelationship between periods, first order conditions are solved using the Lagrangian: (36)

(37)

(38) Manipulating first order conditions will yield marginal conditions similar to those in equation 14. However, the expected utility function at t = 2 is maximiz.ed given X1• Thus, the expected value of the marginal utility of income, "1.• is also a function of X1• The solution of equations 36-38 will be:

Substituting X2* and Ci* into equation 33 yields an ex ante variable indirect utility function conditional on X1, S, and C1:

Young Sook Eom

42

The variable indirect utility function was defined by Epstein (1975) and was applied by Chavas et al. (1986) to derive the option price. In the first period, the household chooses its optimal level of consumption X 1, and savings, S, given averting expenditure, C 1, provided that X2 and Ci are determined in the optimal manner in the second period. However, the current consumption decisions at t = 1 must be made subject to uncertainty about future risks. The first-period choice problem is to:

EU1

(42)

= Max

E1[EUiJ

Xt,S

(31) Using the envelope theorem, first order conditions are:

(43)

(44) (45) The first order conditions, equations 43-45, can be solved for the optimal level of current consumption, X 1* and savings, S*, where X1* = X1(M1,Px/ ,r,11"2) and S* = S(M 1,px ',r,11"2)· 1

Substituting X 1* and S* into equation 33' yields an ex ante unconditional indirect expected utility function such as:

where

x1• 0)=

Picz ' r, I

?r2 •

X 1(M 1, Px , r, ?ri}; and 1

x2• =

X2(M1, M1

+ (1 +

r) * S, Px1 ,

Substituting equation 46 into equation 42 will lead the first period maximii.ation problem to equation 47 in terms of the ex ante indirect expected utility function:

43

Self Protection, Risk Information, a11d Ex Ante Values

(47)

EU1

=

E1{[l - 1/t

+

1/1(

0 .

Therefore, a relationship between the MWTP without learning and with learning can be derived from equations 24 and 53:

Self Protection, Risk Information, and Ex Ante Values

45

Equation 54 concisely summarizes the difference between values of information with and without the learning opportunities. In response to new information, the marginal changes of self-protection expenditure without the learning opportunity (the second term in equation 54) is greater than those when households have opportunities to learn about the uncertain event (the third term in equation 54). AB households learn more about health risks arising from food contaminants or nutrients, they may recognii.e that they do not need to spend as much on protective expenditures as before to achieve the same level of risk. The consequent reductions in protective expenditure would allow households to have more disposable income that can be reallocated to other consumption activities. This interpretation of the adjustment term may provide an explanation of consumers' strong reactions to extremely low but unfamiliar food risks (e.g., the Alar scare). If a household has to take self-protection actions in a single period context, while the process underlying food-related risks is involved in multiple time periods, then the household may show alarmist reactions and overestimate its tradeoffs between risk and income. This result in equation 54 is also consistent with the empirical findings of market experiments (for example Camerer 1987), where individuals' learning opportunities, gained through experience and better information, reduced biases in market prices with regard to the predictions of a Bayesian model. Concluding Remarks This chapter was motivated by observations suggesting that food-related risks (either arising from food contamination or dietary habits) are not well understood by consumers. Nonetheless, consumers seemed to take self-protection actions to reduce the risks while learning more about risks in response to new information. To describe such situations, I developed a conceptual framework investigating the effects of information and learning on consumers' protective behavior. The framework incorporating the learning process did not change the basic structure of the expected utility theory: self-protection and consumption decisions that affect utility directly are separated from the processes of risk perception and learning. AB long as we can identify protective behavior undertaken specifically to reduce health risks, the values of risk information could be measured from the knowledge of the technical relationship between risk and self-protection action, which is observable in principle. This three-way connection-perceptions, learning, and behavior-clearly has important implications

46

Young Sook Eom

for governmental efforts to address market failure associated with the provision of food safety and nutrition. Greater understanding on the part of consumers, based on learning opportunities, may reduce the degree of overestimation of small risks, as reflected in smaller self-protection expenditures. Unfortunately, the empirical implementation of the conceptual framework developed in this chapter requires intensive data collection efforts. Required are household-level primary data on patterns of food consumption and expenditures, time allocations including different cooking and shopping activities, wage rates, and the prices and quantities of protective behaviors, along with measures of levels of food attributes (such as food safety and nutrition) consumed by the same households. In addition, to understand households' learning processes, we need to accurately elicit consumer perceptions about food attributes before and after receiving new information and the effect of changes in perceptions on selfprotection activities. To date, the valuation research related to food safety and nutrition largely utilized three different data sources in various empirical applications: (1) consumers' revealed responses to food attributes or information about food attributes in actual market-based situations (hedonic price analysis, qualityaugmented product demand analysis), (2) consumers' stated purchase intentions or expressed willingness to pay for safer products in hypothetical market situations (contingent valuation method, conjoint analysis), or (3) laboratory experiments involving purchase decisions on attribute-differentiated products (e.g., Vickery sealed-bid auctions). Moreover, most of this empirical research has considered the three aspects of food safety economics- perceptions, behavior, and valuation-as separate alternatives. One line of research focuses on how best to elicit consumers' perceptions of food risks and how to examine the influences of sociodemographic characteristics on consumers' attitudes toward certain food attributes. Other researchers emphasize identifying the existence of a linkage between nonmarket food attributes and observed self-protection actions by analyzing consumer demand for attribute-differentiated food products. Others attempt to estimate values of certain food attributes or willingness to pay for safer products. One direction that future valuation research can take to meet intensive data needs is to examine the possibility of combining different sources of behavioral responses in a utility-theoretic consumer choice model such as the one developed in this chapter. For example, consumer preferences for food attributes can be jointly estimated by using both actual market demand responses for food products and contingent behavior responses to information about food attributes. This composite research strategy will exploit individuals' behavioral "windows" more completely and thus provide more reliable measures of the value of information about food safety and nutrition.

Self Protection, Risk Information, and Ex Ante Values

47

Notes 1. This chapter was taken from research done for my dissertation at North Carolina State University. I am indebted to V. Kerry Smith for insightful and constructive suggestions on earlier versions of this paper. 2. Since M and r changed while Px remained constant, the total differential reduced to a partial differential. Thus the partial derivative in equation 4, (aMtar) is equal to (dM/dr) given rx constant. This interpretation is applied to the discussions throughout this chapter. 3. This type of switching behavior entails discrete choices, which result in comer solutions for food consumption decisions. Eom (1994) describes a discrete choice model in a situation involving risks from pesticide residues on fresh produce. 4. Perhaps this nonjointness assumption in the household production framework: may be equivalent to the separability assumption between exposure to pollutants and unobservable randomness affecting health, which permitted Quiggin (1992) to derive some positive results in applying self-protection models. As a result of the nonjointness assumption, our analysis excludes certain preventive health behaviors (such as regular physical exercise) that improve households' general health conditions as well as mitigating the adverse health outcomes associated with particular food contaminants or nutrients. 5. When changes in expenditure on marketed protective behavior are used to measure individuals' values on nonmarket goods such as risk information, we have to recognize that there are three possible measures: (1) the change in expenditure on protective behavior, v, given a constant income, M, (2) the change in expenditure on v to hold the final service flow, r, constant, and (3) the change in expenditure on v to hold expected utility, EU, constant. The third measure is a correct measure of individuals' willingness to pay for the change in an exogenous factor, I. In the case in which a protective behavior, v, is a perfect substitute for the exogenous factor, I, the second measure will be equal to the third measure. However, the first measure is not the same as the third measure because of the income reallocation associated with the change in I (see Smith 1990 for details). 6. More specifically, Viscusi and O'Connor (1984) assumed the random event (the occurrence of the adverse health outcome) follows the sequence of Bernoulli trials and prior risk perceptions follow beta distributions. The beta distribution is quite flexible and can reflect a variety of skewed and symmetric shapes by varying the parameters of the distributions (Winkler and Hays 1975). These properties are useful to explain the selfprotection model. 7. Chavas et al. (1986) proved that the covariance (COV(µ2(12), C2i(li)) is positive if [(aµ2lalv * caeii1a12)] is positive.

References Arrow, Kenneth. 1982. Risk Perceptions in Psychology and Economics. Economic Inquiry 20: 1-9. Becker, G. S. 1965. A Theory of the Allocation of Time. Economic Journal 75:493517.

48

Young Sook Eom

Becker, G. S. 1974. A Theory of Social Interactions. Journal of Political Economy 82:1063-1093. Berger, M. C., B. L. Dillman, D. Kenkel, and G. S. Tolly. 1987. Valuing Changes in Health Risks: A Comparison of Alternative Measures. Southern Economic Journal 53:967-984. Brown, D. J. and L. F. Schrader. 1990. Cholesterol Information and Shell Egg Consumption. American Journal of Agricultural Economics 72(3):548-555. Camerer, C. F. 1987. Do Biases in Probability Judgement Matter in Markets? Experimental Evidence. American Economic Review 77:981-997. Chavas, Jean-Paul, R. C. Bishop, and K. Segerson. 1986. Ex-Ante Consumer Welfare Evaluation in Cost-Benefit Analysis. Journal of Environmental Economics and Management 13:255-268. Conklin, N. C., G. Thompson, and L. Riggs. 1992. Price and Quality Differentials in Organic and Conventional Produce Markets. A Report to Agricultural Marketing Service of USDA, June. Cook, P. J. and D. A. Graham. 19n. The Demand for Insurance and Protection: The Case of Irreplaceable Commodities. Quarterly Journal of Economics 91:143-156. Crawford, R. G. 1973. Implications for Leaming for Economic Models of Uncertainty. International Economic Review 14:587-600. Deaton, A. S. and J. Muellbauer. 1980. Economics and Consumer Behavior. Cambridge, MA: Cambridge University Press. DeGroot, M. 1970. Optimal Statistical Decisions. New York, NY: McGraw-Hill. Dickie, M., S. Gerking, W. Schulze, A. Coulson, and D. Tashkin. 1987. Value of Symptoms of Ozone Exposure: An Application of the Averting Behavior Methods. Volume II, Improving Accuracy and Reducing Cost of Environmental Benefit Assessments. U.S. Environmental Protection Agency, December. Ehrlich, I. and G. S. Becker. 1972. Market Insurance, Self-Insurance, and SelfProtection. Journal of Political Economy 80:623-648. Eom, Y. S. In Press. 1994. Pesticide Residues and Consumer Valuation of Food Safety. American Journal of Agricultural Economics. Epstein, L. G. 1975. A Disaggregated Analysis of Consumer Choice under Uncertainty. Econometrica 43:877-891. Foster, W. and R. E. Just. 1989. Consumer Valuation of Health Risk: The Case of Heptachlor Contamination of Milk in Hawaii. Journal of Environmental Economics and Management 17:266-283. Gerking, S. and L. Stanley. 1986. An Economic Analysis of Air Pollution and Health: The Case of St. Louis. Review of Economics and Statistics 71:228-234. Grossman, M. 1972. On the Concept of Health Capital and the Demand for Health. Journal of Political Economy 80:223-255. Hammitt, J. K. 1986. Organic Carrots: Consumer Willingness to Pay to Reduce Foodborne Risks. Prepared for U.S. Environmental Protection Agency, May. Ippolito, P. M. and A. D. Mathios. 1990. Information, Advertising, and Health Choices: A Study of the Cereal Market. Rand Journal ofEconomics 21(3):459-480. Kenkel, D. S. 1991. Health Behavior, Health Knowledge, and Schooling. Journal of Political Economy 99:287-305. Ott, S. L., C. L. Huang, and S. K. Misra. 1991. Consumers' Perceptions of Risks

Self Protection, Risk /'!formation, and Ex Ante Values

49

from Pesticide Residues and Demand for Certification of Residue-Free Produce. In

Economics ofFood Safety, ed. Julie A. Caswell, 175-188. New York, NY: Elsevier

Science Publishing Company. Putler, D.S. and E. Frazao. 1991. Assessing Effects of Diet/Health Awareness on the Consumption and Composition of Fat Intake. In Economics ofFood Safety, ed. Julie A. Caswell, 247-272. New York, NY: Elsevier Science Publishing Company. Quiggin, J. 1992. Risk, Self-Protection, and Ex Ante Economic Value-Some Positive Results. Journal of Environmental Economics and Management 23(1):40-53. Rae, D. 1987. Risks of Consuming Pesticide and Fungicide Additive: Perceptions and Behavior of Organic Food Consumers. Final Report to the Environmental Protection Agency. Slovic, P., B. Fischhoff, and S. Lichtenstein. 1985. Regulation of Risk: A Psychological Perspective. In Regulatory Policy and the Social Sciences, ed. R. Noll. Berkeley, CA: University of California Press. Smith, M. E., E. 0. van Ravenswaay, and S. R. Thompson. 1988. Sales Loss Determination in Food Contamination Incidents: An Application to Milk Bans in Hawaii. American Journal of Agricultural Economics 70(3):513-520. Smith, V. K. 1990. Household Production Function and Environmental Benefit Estimation. In Measuring the Demand for Environmental Improvement, ed. John Braden and Charles Kolstad. New York, NY: North-Holland. Smith, V. K. and F. R. Johnson. 1988. How Do Risk Perceptions Respond to Information? Review of Economics and Statistics 73:1-8. Swartz, D. G. and I. V. Strand, Jr. 1981. Avoidance Costs Associated with Imperfect Information: The Case of Kepon. Land Economics 57:139-150. van Ravenswaay, Eileen 0. 1991. Consumer Perception of Health Risks in Food. In Increasing Understanding of Public Problems and Policies-1990. Oak Brook, IL: Farm Foundation. van Ravenswaay, E. 0. and J.P. Hoehn. 1991a. The Impact of Health Risk on Food Demand: A Case Study of Alar and Apples. In Economics of Food Safety, ed. Julie A. Caswell, 155-174. New York, NY: Elsevier Science Publishing Company. van Ravenswaay, E. 0. and J. P. Hoehn. 1991b. Contingent Valuation and Food Safety: The Case of Pesticide Residues in Food. Department of Agricultural Economics, Michigan State University, Staff Paper No. 91-13. Viscusi, W. K. 1979. Job Hazards and Worker Quit Rates: An Analysis of Adaptive Behavior. International Economic Review 20:29-58. Viscusi, W. K. 1989. Prospective Reference Theory: Toward an Explanation of the Paradoxes. Journal of Risk and Uncertainty 2:235-264. Viscusi, W. K. and C. J. O'Connor. 1984. Adaptive Response to Chemical Labeling: Are Workers Bayesian Decision Makers? American Economic Review 74:942-956. Viscusi, W. K. and W. A. Magat. 1987. Learning About Risk. Cambridge, MA: Harvard University Press. Winkler, R. L. and W. L. Hays. 1975. Statistics: Probability, Inference, and Decision. New York, NY: Holt, Rinehart and Winston, Inc. Zellner, J. A. and R. L. Degner. 1989. Consumer Willingness to Pay for Food Safety. Paper presented at the 1989 Annual Meeting of the Southern Agricultural Economics Association, February.

3 Mitigation, Product Substitution, and Consumer Valuation of Undesirable Foodbome Effects Robert D. Weaver A subtle and often overlooked revolution commenced in the U.S. during the 1960s and has spread globally while simultaneously penetrating deeper and deeper layers of our social fabric. This revolution was one of consciousness of externalities. It began with recognition of the most obvious: smoke belching factory stacks and bus exhausts. Within this rubric, consciousness of the health implications of diet and food characteristics has steadily evolved targeting both general dietary effects on health as well as specific impacts of known and unknown, expected and unexpected characteristics and effects of food. Concern has included all things "artificial," for example additives such as colorants, flavorings, and other chemical residues; as well as undesirable "natural" things, for example foodborne pathogens such as aflatoxins, bacteria, molds, fungus; and other unwanted attributes which have direct health implications. Importantly, cause for consumer concern has evolved such that this list of undesirable things "natural" has been extended to include a wide variety of nutrients such as fat, sodium, calories, absence of fiber or vitamins, cholesterol, etc. The existence and value of the market for artificial sweeteners as well as egg and dairy substitutes attest to the significance of this trend in consumer concern and its manifestation in the marketplace. An underlying feature of these consumer concerns is their focus on what we will define as undesirable foodbome effects or UFBEs. UFBEs can be classified into three categories: (1) nutrients and nutritive characteristics, (2) food additives and residues, and (3) foodborne pathogens. This classification will prove useful in the consideration of the microeconomics of consumer valuation of changes in UFBEs. Cook (1989) provided a general chronicle of consumer concern for various types of UFBEs. More specific evidence has evolved over the decade through 51

52

Robert D. Weaver

a series of consumer surveys. Most recently, surveys have provided evidence that consumers have high levels of concern for possible health risks of chemical residues on fresh produce (Zind 1990, Ott and Maligaya 1989, Food Marketing Institute 1989, Sachs et al. 1987, Weaver et al. 1992). These apparently high levels of consumer concern have offered economists an attractive challenge to consider consumer behavior with respect to UFBEs. Of equal interest to economists has been measurement of consumer valuation of safety in food associated with the absence of UFBEs, whether consumer valuation is sufficiently stable to finance private sector innovation, and whether there is an opportunity for Pareto improvement from some type of government intervention. The objective of this chapter is to consider what guidance microeconomics has to offer for valuation of UFBEs. In particular, the chapter both reviews and extends existing literature by drawing on applied microeconomic theory to consider means of measuring consumer valuation of UFBEs. In the process, the chapter will lay a microeconomic foundation for valuation of UFBEs which may be of some usefulness for further empirical studies. The plan of the chapter is as follows. The salient features of the consumer choice problem are reviewed and three cases are identified for consideration: (1) an exogenous change in health risk information affects all food and the consumer has no opportunity to avoid exposure to the risk (no mitigation), (2) an exogenous change in health risk of one food product or type is announced for a temporary period, however, risk can be managed through defensive action (partial mitigation), and (3) alternative food products are characteri7.ed by different levels of perceived health risk allowing the consumer to completely avoid exposure (full mitigation). A brief review of literature suggests that past literature has focused on Cases 1 and 2. The microeconomics of the three cases are considered and a generali7.ed approach is presented that incorporates salient features of the consumer choice problem. Based on this generali7.ed model, implications for measurement of consumer valuation of UFBEs are derived. Results indicate that cost-of-illness, averting expenditure, and disutility of illness are each components of willingness to pay (WTP). Importantly, the chapter demonstrates that none of these individual components will, in general, suffice to measure WTP. Availability of alternative goods free of UFBEs renders willingness to pay discontinuous in the UFBE parameter. Limits on the usefulness of premia for uncontaminated food as a measure of WTP are also established. To clarify the focus of this chapter, two types of consumer response to UFBEs have been recogni7.ed in past literature. These include (1) consumer response to new information or announcements concerning UFBEs present in particular foods and the effects of the UFBEs and (2) demand response to and valuation of known levels of UFBEs. The first type of study considers the impact on consumer behavior of a change in information that results in an exogenous change in the perceived level of UFBE and foodbome risk associated with the food. For this reason, we label this type of study an announcement

Mitigation, Product Substitution, and Consumer Valuation

53

study. Announcements have affected all types of UFBEs and by definition involve UFBEs for which exposure after announcement can be in some sense managed by the consumer. For example, announcement of the presence of or health impacts of nutritional characteristics, additives, or residues may lead to shifts in demand for affected products. Similarly, announcements concerning the presence of foodbome pathogens may influence future consumption decisions to substitute other foods for which perceived risk is lower. The second type of study, demand response to and valuation of known levels of UFBEs, is of particular interest in this chapter and presumes that available information concerning UFBEs has not led to termination of consumption of the food. Of interest in this case is how available information influences the consumer's decisions.

Microeconomics of Cons1D11er Valuation of UFBEs In order to construct the microeconomic basis for measuring consumer valuation of changes in UFBEs, three cases will be considered which emerge from different interpretations of externalities and their effects that are associated with UFBEs. Each of the cases is relevant for particular types of UFBEs. For the first case, the UFBE is specified as an externality that is purely exogenous in origin. As in the classic Pigouvian case, no mitigating reaction by the affected individual is allowed. An unexpected exposure to a foodbome pathogen is a UFBE that falls in this case. Under such conditions, the microeconomics of consumer valuation of a change in the externality is straightforward. For the second case, we differentiate exposure to an extemality from the effect of an externality, allowing for partial mitigation by those exposed. Clearly, in many cases of extemality, the decision-maker has an opportunity to react by adapting decisions to reduce the impact of the externality. This type of partial mitigation is not possible in Case 1. As Shibata and Winrich (1983) noted, the polluted must be expected to defend themselves. UFBE examples that would fall into this case would include presence of chemical pesticide residues on the surface of produce. Here, washing or peeling could reduce exposure dramatically while allowing the consumer to continue to consume the product. The third case follows when full mitigation or avoidance is feasible, as is often possible with respect to exposure to UFBEs. In such a case, the consumer can choose alternative products and avoid exposure. The microeconomics of Case 1 are well-known (see, e.g., Swartz and Strand 1981) and will be considered only briefly here. In each of the second and third cases, a microeconomic perspective will be presented on the question of valuation of changes in the level of exposure to a UFBE, which is viewed as an extemality associated with food consumption. In particular, alternative indicators of value will be considered: cost-of-illness, defensive or averting expenditures, and willingness to pay. In a final model, the features of Cases 2 and 3

Robert D. Weaver

54

are synthesized to produce a general microeconomic model capable of analyzing observed consumer choice. Before proceeding, one further bit of housekeeping must be completed to clarify the focus of the chapter. In particular, any extemality, including the UFBEs to be considered here, may be either an expected characteristic of a situation or it may be a randomly occurring characteristic. For the case of UFBEs, expected characteristics might include the known, expected, or perceived health effects of sugar, fat, or a chemical residue. Randomly occurring characteristics generated by known probability density functions would include health effects resulting from known pathogens and are properly labeled foodborne risks. Where the pathogen's presence is unknown, uncertainty-not risk-would exist. This chapter will consider only the first type of effect. Generaliution of the theory to allow for random effects is straightforward.

Case 1: Exogenous Extemallty with No Mitigation In this case, the extent of exposure to the extemality is presumed known and exogenous to the individual. Where the effect of the exposure is also known, a variation of the Swartz and Strand (1981) model can be used to consider the implications of this case. In their model, two goods are assumed X 1 and X2 , where the quality of X 1 is indicated by Z 1 = Z 1(N) where N indicates the state of information concerning quality. X 1, X2, and N are assumed available in perfectly elastic supply to the individual at prices P 1, P 2 , and c. The individual is assumed to hold preferences over X 1 and X2 and make choices by maximizing utility U() representing those preferences subject to a constraint on income I. Swartz and Strand assume the quantity X 1 is a function of z1 and that no direct preferences exist for Z 1• Generalizing this specification, allow utility to be a function of Z 1, and suppose the consumer's problem is: (1)

max

U(X1,X2,Zi(N))

subject to P1X1

+ PzX2 + cN

= I.

X1,X2,N

Interior solutions are required suggesting that while some indirect mitigation might occur by substitution of X2 for X 1, full mitigation is not feasible. If X 1 is interpreted as food and X2 as all other goods and no product substitution is allowed, no indirect mitigation through substitution is allowed. In this model, if we assume the state of safety N is exogenous, then its price can be set to zero and the willingness to pay (WTP) for an improvement in food safety can be evaluated as resulting from a change in N. Swartz and Strand determine WTP as the integral of the inverse demand function for X 1• We define WTP as the compensation necessary to equate the indirect utility functions evaluated before and after the change in the level of food safety. For example, where V(l,N,P 1,Pi) defines the indirect utility function, WTP for a change from

Mitigation, Product Substitution, and Consumer Valuation

55

N' to N" could be defined as: (2)

Alternatively, by total differentiation of the indirect utility function and constraining the variation in utility to be z.ero, WTP may be derived as: (3)

W7P = a11aN = -(aVlaN)/{aVlal).

This model can be generaliml with no change in the approach to measuring WTP to allow for risk of illness by specifying the probability of illness as a function of the food safety level N, and perhaps personal characteristics of the consumer indicating state of health. The usefulness and appropriateness of this case for valuation of changes in food safety can now be assessed. By design, the usefulness of the model is limited to cases where no mitigation occurs. In the case of UFBEs, past literature has considered this specification within the context of oysters and shellfish (Swartz and Strand 1981), milk (Foster and Just 1989, Smith et al. 1988), and apples (van Ravenswaay and Hoehn 1991a, 1991b). However, for these cases, the two product specification of the above model is clearly uninteresting since substitution and complementarity among foods is both likely and of interest as a possible form of mitigating behavior. Further, for many types of UFBEs, defensive actions are clearly feasible and the consumer choice model must be generalized to accommodate that possibility. Fat can be trimmed, fruit and vegetables can be washed, and meat can be stored properly. Where mitigation is feasible, either through product substitution or through defensive action, the level of exposure (Nin the model above) becomes endogenous, rendering the Case 1 model misspecified for the problem under study.

Case 2: Exogenous Extemality with Partial Mitigation Coase (1960) recogniml as early as 1960 that the effects of virtually any externality could be controlled by the affected individuals at least in part through defensive, mitigating, or averting actions. Coase (1960), Baumol (1972), Baumol and Oates (1971), Mishan (1977), and Shibata and Winrich (1983) presented reconsiderations of the implications of mitigating behavior for tax or subsidy solutions to externalities. Applied literature has considered three cases of behavioral reaction to externalities that are of interest to the problem of valuation of UFBEs. These cases include behavioral reactions (e.g., cleaning) to pollution, defensive actions taken to improve health, and the value of public goods provision when there exist private good substitutes or complements. To limit the scope of this discussion, focus will be placed on the first two areas of literature.

Robert D. Weaver

56

Courant and Porter (1981) explored the usefulness of averting expenditure as a measure of the benefits of reducing exposure to an externality where opportunity for averting action is feasible. Harford (1984) extended the Courant and Porter model to directly incorporate interaction between averting action and the level of exposure to the externality in determining the unit cost of averting action. The Harford model serves as a good starting point for motivation of models suitable for the analysis of consumer valuation of UFBEs effects. Harford defines the consumer's choice problem (equation 4 below) as a simple generalization of utility maximi7Jltion subject to a budget constraint. Utility is affected by consumption of a numeraire good (X) and the level of cleanliness (C): U = U(X,C). Cleanliness is produced by the frequency of cleaning (F) and the extent of exposure to an externality (pollution, W): C = C(F,W). Cleaning and exposure also affect the unit cost of cleaning (q): q = q(F,W). Harford's model links exposure to the externality (W) and defensive action (cleaning, F) through a recursive, structural relation that generates an outcome (cleanliness) that affects utility. Harford' s model also makes the cost of mitigation dependent on the exposure level. In the absence of such dependence, Harford's model would be interpretable as simply a multiple good variation of Case 1 described above. Harford's model may be summariz.ed as follows: (4)

max U

= U(X,C) C = C(F,W) q = q(F,W)

Y=X+qF

subject to Cleanliness Unit Cost of Cleaning Budget Constraint.

Harford derives the marginal willingness to pay for a change in the externality (W) by requiring that utility remains constant and choice remains optimal. Mathematically, the total differentials of utility and the budget constraint are set to z:ero simultaneously with the condition that the first order conditions are met. Harford finds that the marginal willingness to pay is equal to the change in total cost of cleanliness with respect to the change in pollution (as indicated by the second term in equation 5 below) and is determined by the marginal cost of cleaning, the rate of substitution between cleaning and pollution, and the cleaning frequency: (S)

(dY/dW) Iu = d(qF)ldW = (q

+ q,F)(dF/dW) Ic + qv,F.

Harford's primary concern was to determine the usefulness of observed behavior such as the frequency and expense of cleaning for inferring the consumer's valuation of increased pollution. Courant and Porter (1981) showed that the benefits of pollution reduction could be measured by the average unit cost of cleaning times the decrease in cleaning induced by the decline in pollution. Harford' s extension shows that marginal WTP can be measured by the change in

Mitigation, Product Substitution, and Consumer Valuation

57

defensive expenditure, though this is not, in general, simply (qdF/dW) as Courant and Porter found. Instead, marginal WTP in his model will vary with the level of pollution (since qp varies with W). The conclusion is that averting expenditure cannot, in general, be measured by the change in averting behavior (e.g., cleaning) times the unit price of cleaning. Under simplified conditions, Harford shows that if the unit cost of cleaning depends on the level of cleanliness, rather than the level of pollution, then the Courant and Porter result emerges. The significance of these results is that change in total expenditure must be observed to measure WTP for a change in pollution, not simply change in mitigating behavior (e.g., F). Clearly, the nature of this result is of interest for the case of food safety where premia may exist for higher levels of safety. For example, where premia exist for certified pesticide-free produce, it is of interest to question whether such premia may be interpreted as a measure of WTP? Harford's results indicate that such premia would measure only one element ofWTP. Where averting or mitigating action is possible, the Harford model would seem to be of interest as a starting point for the measurement of consumer valuation of changes in UFBEs. Consider the case of pollution of drinking water. The question of consumer valuation of clean, safe water is immediately raised. Since bottled water is nearly universally available, and where it is not, boiling or filtration typically results in safety, consumers have defensive options. An attractive approach would be to use as an estimate of averting expenditure the quantity of boiling, filtration, or bottles of water times respective prices as a measure of consumer WTP. Harford's results would rule out such an approach as a measure of consumer WTP or true benefits when the unit cost of averting action is dependent on the extent of pollution and the extent of "cleaning" action. Since it is likely that unit costs would vary with pollution, Harford's proposed general approach would be necessary and use of averting expenditure would at best provide a lower bound estimate of true estimates. Harrington and Portney (1987) reconsidered an extension of the question analyzed by Harford asking how data on direct and indirect costs of illness as well as on averting expenditures could be used to estimate benefits of a reduction in pollution. Harrington and Portney find that the true benefits of a reduction in a health threat posed by pollution exceeds the sum of the cost of illness and the change in defensive expenditures. This result contrasts with Harford's result which indicated that defensive or averting expenditures could either underor over-estimate true benefits depending on the shape of the dose-response function. The Harrington and Portney model can be summarized as follows: (6)

max U(X,L,S) subject to S = S(D,P) M = mS(D,P) T-L-S=O I+ wT- wL - wS(D,P) -X-D - mS(D,P)

=0

Time Spent ill Cost of Illness Time Constraint Budget Constraint

58

Robert D. Weaver

where U is utility, X is the private consumption good, L is leisure, S is time spent ill, D is defensive expenditure, P is pollution, M is the cost of medical treatment, and w is the wage rate. In extensions, they add P to the utility function and allow w to vary with S. Using the indirect utility function V = V(I,P,w), Harrington and Portney (1987) derive a compensating variation measure of marginal WTP for a reduction in P. Use of the indirect utility function assures that WTP is evaluated at optimal choices. By total differentiation ofV( ), they find WTP (setting dV/dP = 0 and using the first order conditions): (7) a11ap

= -VplVI =-Sp/SD = (w + m) dS/dP - (U/X) dS/dP + Dp.

The final expression on the right represents Harrington and Portney's decomposition of the benefit measure into three portions: (1) the money cost of illness, (2) the disutility cost of illness, and (3) defensive expenditures. This result clearly illustrates that cost of illness or defensive expenditure would underestimate true benefits. As in the Harford model, the exposure to the externality (P) is specified by Harrington and Portney as strictly exogenous and independent of any consumption decision. This places the Harrington and Portney model within the context of Case 2. While consumer reaction is allowed through defensive action (D), as in Courant and Porter (1981), the unit cost of that action is fixed leaving the Harrington and Portney approach subject to Harford's criticisms. The Harrington and Portney model does take one step further than the Harford model in that income loss is allowed as a result of illness. Importantly, in an extension of their basic model, Harrington and Portney allow for the rate of loss (the wage rate) to vary with time spent ill. As do the Courant and Porter and the Harford models, the Harrington and Portney model establishes that benefit measurement is necessarily based on careful specification of the microeconomic decision problem faced by the consumer. Of great import is the specification of relations among defensive action and its cost, exposure, and its effect on utility. The usefulness of the Harford or Harrington and Portney models for deriving measures of consumer valuation of UFBEs effects can now be assessed by nesting them in a common notation appropriate for the case under study. In particular, define: max U = U(X,l,S)

(8)

L+Sl~T

I= l(D,E) C = C(D,E) Y

subject to

= X + qD - wL + C(D,E)

Time Spent Ill Cost of Illness Time Constraint Income Constraint

where X is the private consumption good, I is time spent ill, S is leisure, D is

Mitigation, Product Substitution, and Consumer Valuation

59

defensive action, E is exposure to a health debilitating extemality, C is the cost of illness, L is labor time, T is total time, Y is income, and q and w are fixed unit prices. Harford's problem follows from equation 8 when leisure, labor, and illness time are dropped from the problem. Harrington and Portney's problem follows from equation 8 when the cost of illness C results from fixed unit cost m, i.e., C = ml(D,E) where m = i +wand i indicates the unit price of medical treatment of illness. The essential feature of the Case 2 problem is that the level of the extemality (E) is independent of consumer choice. For the UFBE problem, several further features are important to incorporate. First, it would seem reasonable to postulate that illness directly enters the utility function as in equation 8. Similarly, it would seem reasonable to specify an illness function as in equation 8 which is dependent on exposure to the foodbome effect and any defensive actions that are taken. As for the cost of illness, both the opportunity cost of time lost and the cost of treatment should be recognized. Harford's variable average cost of defensive action is appealing. In sum, the Case 2 model for UFBE analysis could be specified as an extension of equation 8:

(9)

subject to max U = U(X,l,S) I= I(D,E) C = (m(D,E) + w) I(D,E) L+Sl~T

Y = X - wL + qD + m(D,E)l(D,E).

The differences between this model and those of Harford and Harrington and Portney are best appreciated from a consideration of its implications for the measurement of WTP. Using the indirect utility function to derive WTP:

dY!dEI v = -V_iVy

(10)

where V is evaluated at its initial value. Equation 10 provides the compensating variation, whereas when V is evaluated at the final value, equivalent variation results. Evaluating VE and Vy using the Lagrangian, the first-order conditions for optimal choice, and the total change in averting expenditures, equation 10 can be rewritten and decomposed to allow interpretation as follows: (11)

dY!dEI v = (m

+ w)IE + mEI

Marginal Cost of Illness Money Value of Disutility

wher~

Marginal Defensive Expenditure is the Lagrangian multiplier.

60

Robert D. Weaver

The elements of the decomposition of WTP are directly interpretable as (1) the marginal cost of illness, (2) the marginal money value of disutility, and (3) the marginal defensive expenditure, each with respect to an exogenous change in exposure to the extemality (E) which we interpret as UFBE exposure. Importantly, under general conditions, reliance on any one of the elements would result in an underestimation of the marginal benefit of the reduction in the extemality. Further, following Harford, the marginal cost of defensive activity is not constant, but varies with the level of defensive action and the level of exposure. Equation 11 extends past results by clarifying that the disutility cost involves both a direct and an indirect effect that results from the defensive expenditure induced by an increase in exposure (Dw, a resulting decrease in illness (Jo), and increase in utility (U1). The consumer choice problem considered here for Case 2 and the derived WTP measure provide a solid microeconomic foundation for measurement of consumer valuation when consumer choices do not affect the level of exposure directly. For the case ofUFBEs, it is difficult to identify examples which might fit this case. Consider the case where drinking water is polluted. If no other source of water were available, then water boiling or filtering could be appropriately thought of as mitigating behavior and the level of exposure could be viewed as exogenous and independent of consumption choices, although the extent of the effect is rendered endogenous through mitigation. However, such conditions seem unrealistic given widespread availability of bottled water. Purchases of bottled water would clearly affect the level of exposure implying the level of exposure is dependent on consumption decisions to substitute products. This condition would seem to be an important aspect of the choice problem relevant for valuation of food safety.

Case 3: Endogenous Externality with Full Mitigation The measurement of consumer valuation of externalities when the consumer choice problem satisfies Case 3 can now be considered by simplifying and generalizing the model presented in equation 9. The key feature of the Case 3 problem is that the level of exposure to the extemality is dependent on consumption decisions. For food consumption, this possibility would occur whenever the consumer is faced with an alternative product or action which can fully mitigate exposure. Two possibilities occur which deserve examination. In the first, as in Case 2, exposure results from or is produced by an exogenous stimulus, say e, though mitigation is allowed with an endogenous, continuous defensive action, say D. In this case, the effect of the exposure (I) might be specified as I = I(E(e,D)). While e is exogenous, the individual can continuously vary the impact of the exposure by varying the mitigating action defined by D. The implication is that the exposure realized (E) and, therefore, the impact (I) are endogenous. This

Mitigation, Product Substitution, and Consumer Valuation

61

case might be viewed as an example of the situation faced by a consumer faced with produce with surface residues of a pesticide. The exposure level e is exogenously determined, though the realized exposure is dependent on defensive action D such as proper washing. These conditions are consistent with Case 2. In Case 3, exposure may be fully mitigated through a binary decision to be exposed, e.g., as might result from the consumer's choice not to substitute an alternative product. Clearly, this is a case of great relevance to food safety since many food products are highly substitutable. In this case, multiple substitution possibilities imply mitigation could occur through substitution of other goods. Where a close substitute exists, complete avoidance of exposure or full mitigation may be feasible. In the UFBE application, these features necessarily take more specific form. For most situations, it could be expected that near perfect substitutes might exist, as in the case of organic produce. In this case, the consumer has a choice between two product types which are nearly perfect substitutes with respect to the consumption effects of the product which carry with it associated exposure to the •extemality. • Despite such near perfect substitution, the product alternative would be expected to have different prices since they would not be perfect substitutes when the extemality is considered. Past literature has not considered this case. Rather than extend the Case 2 specification to incorporate the features of Case 3, the following simplified Case 3 model for UFBE analysis, using a new notation, will be employed in order to retain focus on the implications of the unique features of the case: (12)

max U = U(X,Z,E)

subject to

x,;xe = 0 E

= E(Xe1· e)

where E(O,e)

=0

Again, U( ) is viewed as twice differentiable and strictly quasi-concave in X, Z, and E. In this specification, Z is defined as a typical private consumption good and its price is used as the numeraire to normalize all other prices. X.. and Xe are viewed as mutually exclusive consumption forms of a product. Importantly, however, X.. and Xe are considered perfect substitutes. In the form Xg, no extemality exposure (E) is associated. In the form Xe, exposure is associated as described by the function E( ). Notably, we assume the level of exposure E is functionally related to the volume of consumption of Xe. The parameter e is defined as the efficiency of the transformation of characteristics of Xe into exposure E. We interpret e to measure the rate of exposure and to be

62

Robert D. Weaver

exogenous. At one extreme this could be interpreted as simply an exogenous dosage rate of exposure. We purposely exclude any opportunity for direct defensive action from this case to sharpen the focus on product substitution as a means of full mitigation. The first-order conditions for this problem immediately indicate the complication added by the roles of Xe and X.. in affecting utility. In brief, the mutual exclusivity of these two goods results in a discontinuity of both choice and the resulting dual functions since the first-order conditions require: (13)

These conditions can be interpreted more conveniently by rewriting them in terms of the premium for Xu: Pu - Pc. By using the first-order condition inequality associated with each good when its optimal level is zero, the following rules may be written: (14)

Ux- Uzl'u

= 0.

From this perspective, it is clear that choice between product types depends on the magnitude of the marginal rate of substitution (mez) between exogenous exposure (e) and other products (Z) relative to the magnitude of the relative premium (Pu - P£). That is, the choice depends on the consumer's willingness to exchange units of E for units of Z as compared to the market's equilibrium rate of exchange of units of e for units of Z, as defined by the relative price Pu Pc• As such, equation 14 provides an important basis for specifying an empirical model of the choice to switch between Xe and X..· Further, it clarifies the role of premia for Xu vs. Xe. Substitution of the optimal choices derived from equation 14 into the choice problem defined by equation 12 provides a definition of the indirect utility function. As in the case of choice, the indirect utility function is discontinuous and may be written:

Mitigation, Product Substitution, and Consumer Valuation

63

V is defined as the supremum (or greatest) utility achievable and is conditional on the choice of Xu or Xe, as determined by the magnitude of mEZ relative to that of Pu -Pc. Within this notation, the total differential of the indirect utility function V can be taken to derive conditional WTP measures as follows: (16)

dY/de

= - (OV(Pe Y,elmEZ 1

:S Pu - PeJ!fJe)/

(OV(Pe 1 Y,elmEZ :S Pu - PeJ!aY) - (a\!'(Pu,Y,elmEZ ~ (a\!'(Pu,Y,elmEZ ~

if Xu= 0, or

Pu - PeJ!ae)/ Pu - PeJ!aY)

if Xe= 0.

These results indicate that for those consumers who consume the product Xe which has associated UFBEs, WTP will be positive and depend on the price of the good with UFBEs (PJ, income {Y), and the efficiency parameter e. For consumers that consume only Xu, the good free of UFBEs, WTP for a change in e is also positive, though in this case it is dependent on Pu, Y, and e. In either case, the WTP is conditional on the magnitude of mEZ relative to Pu - Pc· For consumers who switch from Xu to Xe, at current levels of prices and income, their WTP is determined by the first line in equation 16, in the same way as it is for consumers who were already consuming Xe. Before proceeding, it should be noted that the inequality conditions for choice or switching between Xe and Xu can be rewritten in terms of WTP. This follows from recognizing that: (17)

;JV/(Je

= (OU/fJE)(OE!fJXeJ(OX/fJe)

which implies that mEZ = - (;JVC/fJe)/(;JVc/aY) = WTP I Xe > 0 allows equation 14 to be restated in terms of WTPc: (14')

= WTPc and

Robert D. Weaver

64

A Microeconomi.c Model of Consumer Choice and Valuation of UFBEs Cases 2 and 3 presented above illustrate the implications of introducing each of two important features of the consumer choice problem when UFBEs are present in the marketplace. In this final section, a new model based on a synthesis of these features is introduced and a measure of WTP is derived and its decomposition analyzed. Of particular interest is to determine the implications of this model for the measurement of WTP for changes in exposure to UFBEs through consumption of contaminated food. The model presented abstracts from risk, the extension of the model to consider binary risk of exposure or uncertainty in exposure when consuming a potentially contaminated food product is trivial and would have no impact on the general nature of the results of this section. For example, introduction of a state preference approach for the binary risk case would be straightforward. In the previous two cases, the consumer was allowed to endogenize the effect of exposure through one of two strategies: (1) partial mitigation through defensive action as in Case 2, or (2) full mitigation through substitution of alternative products as in Case 3. Clearly, in the case of food safety, the consumer can be expected to pursue both approaches whenever they are both feasible. Consumer choice in this case is defined by the following problem:

max U = U(X,Z,l,S)

(18)

subject to

I= l(D,E) E = E(Xe'· e)

Y = P,;xu

where E(O,e)

=0

+ P,J < 1 >OPEN END -- MULTI-PUNCH --- ANSWER REQUIRED --

*** QUESTION # 2 *** Some of the questions I will be asking are your opinions about food safety. Others are specific questions about you. Please remember that if you feel any question is too personal and you do not wish to answer it, let me know and we can move on to the next section. (ENTER(.) AND HIT ENTER TWICE) GO TO Q. # 3 = = = = > < 1 > OPEN END -- MULTI-PUNCH --- ANSWER REQUIRED --

237

Using Contingent Valuation to Value Food Safety

***

***

QUESTION # 3 If I have your permission, let me begin by asking how many people, including yourself, do you usually shop for when you buy groceries? (ENTER NUMERIC VALUE ONLY. 98 = DK 99 = REF) GO TO Q. # 4 ====> < 1 >OPEN END 1. THRU -- NUMERIC OPEN END - RANGE IS -- ANSWER REQUIRED --

99.--

** QUESTION # 4 *** These next statements ask your opinion about various food safety issues. For each statement, please tell me whether you strongly agree, agree, feel neutral, disagree, or strongly disagree. (ENTER(.) AND HIT ENTER TWICE) GO TO Q. # 5 ====> < 1 >OPEN END -- ANSWER REQUIRED --

*** QUESTION # 5 *** I think that the current levels of pesticides in fresh fruits and vegetables are safe. GO TO Q. # 6 ====> < 1 > Strongly agree GO TO Q. # 6 ====> < 2 > Agree GO TO Q. # 6 ====> < 3 > Neutral GO TO Q. # 6 ====> < 4 > Disagree GO TO Q. # 6 ====> < 5 > Strongly disagree GO TO Q. # 6 ====> < 6 > # GO TO Q. # 6 = = = = > < 7 > # GO TO Q. # 6 = = = = > < 8 > DK GO TO Q. # 6 ====> < 9 > REF -- SPECIAL FEATURE * SHUFFLING QUESTIONS (3) -BEGINNING WITH QUESTION SAND ENDING WITH QUESTION 10 --

*** QUESTION # 6 *** I think that the government should ban all pesticides. GO GO GO GO GO GO

TO TO TO TO TO TO

Q. Q. Q. Q. Q. Q.

# # # # # #

7 7 7 7 7 7

====> ====> ====> ====> ====> ====>

< 1 < 2 < 3 < 4 < 5 < 6

> > > > > >

Strongly agree Agree Neutral Disagree Strongly disagree #

238

Jean Buzby, Jerry Skees, and Richard Ready

==== > < ==== > < ====>
> >

#

DK REF

*** QUESTION # 7 *** I prefer to buy organically grown fresh fruits and vegetables. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q.

# # # # # # # # #

8 8 8 8 8 8 8 8 8

====> ====> ====> ====> ==== > ==== > ==== > ==== > ==== >


> > > > >

Strongly agree Agree Neutral Disagree Strongly disagree #

#

DK REF

*** QUESTION # 8 *** I don't like to buy imported fresh produce. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q.

# 9 # 9 # 9 # 9 # 9 # 9 # 9 # 9 # 9

==== > ====> ==== > ==== > ==== > ==== > ====> ====> ==== >

< < < < < < < <
> > > > > > > >

Strongly agree Agree Neutral Disagree Strongly disagree

# #

DK REF

*** QUESTION # 9 *** I prefer to buy produce that is shiny. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTO Q. GOTOQ. GOTOQ. GOTOQ.

# # # # # # # # #

10 10 10 10 10 10 10 10 10

====> ====> ====> ====> ====> ====> ====> ====> ====>

> > > 4 > < s > < 6 > < 7 > < 8 > < 9 > < < <
====> ====> ====> ====> ====> ====> ====> ====>

< < < < < < < <
> > > > > > > >

Strongly agree Agree Neutral Disagree Strongly disagree # # DK REF

*** QUESTION # 11 *** I would like to finish with a few questions about you. If there is a certain question that you don't want to answer, just let me know. In what year were you born? (ENTER FULL FOUR DIGITS 9998 = DK 9999 = REF) GO TO Q. # 12 ====> < 1 > OPEN END -- NUMERIC OPEN END - RANGE IS 1880. IBRU -- ANSWER REQUIRED --

9999.--

*** QUESTION # 12 *** How many years of school have you completed? (ENTER NUMERIC VALUE ONLY 98 = DK 99 = REF) GO TO Q. # 13 = = = = > < 1 > OPEN END -- NUMERIC OPEN END - RANGE IS 0. THRU -- ANSWER REQUIRED --

99.--

*** QUESTION # 13 *** Which of the following best describes your racial or ethnic identification: GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ.

# # # # # # #

15 15 15 15 14 14 14

====> ====> ====> ====> ====> ====> ====>

< < < < < <
> > > 5 > 6 > 7 > 1 2 3 4

Afro-American (Black) White (Caucasian) Hispanic Asian (Oriental) or some other racial/ethnic group # #

240

Jean Buzby, Jerry Skees, and Richard Ready

GO TO Q. # 15 = = = = > GO TO Q. # 15 = = = = >

< 8 > DK < 9 > REF

*** QUESTION # 14 *** What other racial or ethnic group would that be? (ENTER VERBATIM RESPONSE) GO TO Q. # 15 ====> < 1 > OPEN END -- MULTI-PUNCH --- ANSWER REQUIRED --

*** QUESTION # 15 *** Could you please tell me approximately what was your annual household income before taxes in 1991? Was it: GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ. GOTOQ.

# # # # # # # # #

16 16 16 16 16 16 16 16 16

====> ====> ====> ====> ====> ====> ====> ====> ====>

< < < < < < < <
< 1 GO TO Q. # 17 = = = = > < 2 GO TO Q. # 17 = = = = > < 3 GO TO Q. # 17 = = = = > < 4 GO TO Q. # 17 ====> < 5 GO TO Q. # 17 = = = = > < 6 GO TO Q. # 17 = = = = > < 7 GO TO Q. # 18 = = = = > < 8 GO TO Q. # 17 = = = = > < 9

> > > > > > > > >

Under $10,000 $10,000 to $14,999 $15,000 to $24,999 $25,000 to $34,999 $35,000 to $49,999 $50,000 to $74,999 $15,000 or above DK REF

grapefruit for yourself or for anyone in > > > > > > > > >

Yes No # # # # # DK REF

*** QUESTION # 17 *** Those are all the questions I have. Thank you very much for your cooperation. (ENTER(.) AND HIT ENTER TWICE) GO TO Q. # 22 ====> < 1 >OPEN END -- MULTI-PUNCH --

Using Contingent Valuation to Value Food Safety

241

-- ANSWER REQUIRED -***QUESTION# 18 ***

As I mentioned earlier, one of the goals of this study is to learn more about consumer's attitudes regarding food safety. As a follow-up to this phone interview, we are mailing short questionnaires to individuals to get more details on their

food buying habits and food safety concerns. We would provide you with a self-addressed, stamped envelope to return your questionnaire and all your answers will be held in strictest confidence. Would you be willing to participate in this follow-up phase? GO TO Q. # 20 = = = = > < 1 > Yes GO TO Q. # 19 = = = = > < 2 > No *** QUESTION # 19 *** Thank you again for your help. (ENTER(.) AND HIT ENTER TWICE) GO TO Q. # 22 = = = = > < 1 > OPEN END -- MULTI-PUNCH -- ANSWER REQUIRED -*** QUESTION # 20 *** May I please have your name and mailing address? INCLUDE COMPLETE NAME, MAILING ADDRESS AND ZIP CODE. USE A SEPARATE LINE FOR EACH LINE OF ADDRESS. --VERIFY ALL INFORMATION-NAME STREET CITY, STATE ZIP CODE GO TO Q. # 21 ====> < 1 >OPEN END -- MULTI-PUNCH --- ANSWER REQUIRED -*** QUESTION # 21 *** Those are all the questions I have. Thank you for your cooperation. You should be receiving the follow-up mail questionnaire in about a week. (ENTER(.) AND HIT ENTER TWICE) GO TO Q. # 22 ====> < 1 >OPEN END -- MULTI-PUNCH -- ANSWER REQUIRED --

242

Jean Buzby, Jerry Skees, and Richard Ready

*** QUESTION # 22 *** Respondent's gender GO TO Q. # 23 = = = = > GO TO Q. # 23 ==== >

<
>

Male Female

*** QUESTION # 23 *** In general, the respondent's understanding of the questions was: GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q. GO TO Q.

# # # # # # # # #

24 24 24 24 24 24 24 24 24

==== > ====> ====> ====> ====> ==== > ==== > ====> ====>

< 1 > < 2 > < 3 > < 4 > < 5 > < 6 > < 7 > < 8 > < 9 >

Excellent Good Fair Poor # # # DK REF

243

Appendix 12.B MAIL SURVEY

CONSUMER FOOD SAFETY CONCERNS: YOUR VIEWS

Please return questionnaire to: J. C. Buzby Department of Agricultural &onomics Room 330 Agr. Engineering Building University of Kentucky Lexington, KY 20546-0276

244

Jean Buzby, Jerry Skees, and Richard Ready

Thankyoufor agreeing to participate in this study. Please.fill out this survey and return

it in the enclosed postage-paid envelope to the Department ofAgricultural Economics at

the University of Kentucky. Do not put your name anywhere on the survey. Your answers will be strictly confidential This study is for a university research project and is not an anempt to sell vou anything. Your help is critical to the success of this study and is appreciated very much.

245

Using Contingent Valuation to Value Food Safety UNIVERSITY OF KENTUCKY

This study is designed to help us karn how cons1llners think about food sqfety issues. Your input can help us achieve this goal Policy makers can make more informed decisions on food sqfety issues if they have better information. Pkase fill out your answers as compktely and accurately as you can. Feel free to add comments in the margins.

SECTION 1: Your Feelings About Food Safety Q-1

What are your most important food safety concerns? Please rank your three most important concerns with "1" for your most important concern and "2" for your second most important concern and "3" for your third important concern. The rest should remain blank. _ _HIGH SATURATED FATS AND CHOLESTEROL HIGH SUGAR CONTENT --_ _HIGH SALT CONTENT _ _ _PESTICIDE RESIDUES ---'PRESERVATIVES AND ADDITIVES (example: coloring) HORMONE AND ANTIBIOTIC RESIDUES --_ _FOOD POISONING OTHER THAN FROM PESTICIDES (examples: Botulism, Salmonella)

Q-2

Please indicate which of the following factors are important to you when deciding which fresh fruits and vegetables you will buy (circle one number for each row). Very lmoortant

PRICE FRESHNESS/QUALITY PRODUCT SELECTION PRODUCT APPEARANCE NUTRITIONAL VALUE IN SEASON ORGANICALLY GROWN Labeled "CERTIFIED PESTICIDE RESIDUE FREE" OTHER (Please Specify)

Moderately Imoortant

Not lmoortant

No Opinion

5 5

1 1 1 1 1 1 1 1

2 2 2 2 2 2

3 3 3 3 3 3

2 2

3 3

4 4

5 5

D D D D D D D D

1

2

3

4

5

D

4 4 4 4 4 4

5 5 5 5

246

Jean Buzby, Jerry Skees, and Richard Ready

Q-3

For each of the following statements, please indicate whether you: (1) strongly agree, (2) agree, (3) disagree, or (4) strongly disagree (circle one number for each row).

Your Opinion on Food Safety Strongly ~

~

Disagree

Strongly Disagree

No Opinion

I am concerned about food safety.

1

2

3

4

D

I prefer to buy organically grown fresh fruits and vegetables.

1

2

3

4

D

I would not pay more money to buy "certified pesticide residuefree" fresh produce.

1

2

3

4

D

In the past, I didn't buy certain fresh fruits and vegetables because of information presented on T. V. and in newspapers regarding harmful pesticide residues.

1

2

3

4

D

Q-4

Which, if any, of the following things do you do regularly to avoid pesticide residues in the fresh produce you buy? (Check all that apply.) DO NOTHING ---~RINSE FRESH PRODUCE WITH WATER

WASH PRODUCE WITH SOAP AND WATER ---~BUY ORGANIC PRODUCE

_ _.BUY FRESH PRODUCE TESTED FOR PESTICIDE RESIDUE AVOID IMPORTED PRODUCE --_ _GROW MY OWN FRESH PRODUCE _ _ _OTHER, Specify~-

Using Contingent Valuation to Value Food Safety

247

SECTION 2: Infonnation on Your Household This section asks questions about yourfamily s consumption of.fresh grapefruit (not juice, canned or ja"ed grapefruit). Fresh grapefruit was chosen for this study as a representative type ofproduce. Q-S

How many people including yourself do you buy groceries for? (Fill in number of adults and children.) ADULTS CHILDREN

Q-6

Of those people for whom you buy groceries (your answer in question S), how many eat fresh grapefruit? (Fill in number of adults and children.)

I

Q-7

When buying grapefruit, do you normally select individual grapefruit from a store display or do you buy them in prepackaged bags or other containers? (Circle one number.) 1. 2. 3. 4.

Q-8

INDIVIDUAL GRAPEFRUIT PACKAGED GRAPEFRUIT BOTH INDMDUAL AND PACKAGED GRAPEFRUIT NO OPINION

If you buy packaged grapefruit, how many individual grapefruit are in one package on the average? (Fill in number.)

II

II

Jean Buzby, Jerry Skees, and Richard Ready

248

The next few questions discuss information on the number of individual grapefruit that you purchase. Therefore, ifyou buy packaged grapefruit, please consider the number of grapefruit you said were in each package for the next few questions (your answer to question Q-8). Q-9

About how often does your household buy fresh grapefruit in the Fall, Winter, Spring, and Summer? (For each column, check one row with an "X. ") How often do you buy grapefruit in the: FALL '! •x• one box

WINTER '! •x• one box

SPRING '! •x• one box

SUMMER '! •x• one box

Every day Three days a week Two days a week Once a week Once a month Once every two months Once a season Never Other (Please specify)

Q-10 When you buy fresh grapefruit, how many individual grapefruit do you usually get each time? Write in a number for FALL, WINTER, SPRING, and SUMMER. Number bought each time in FALL (If none, write in a "O. ") Number bought each time in WINTER (If none, write in a "O. ") Number bought each time in SPRING (If none, write in a "O. ") Number bought each time in SUMMER (If none, write in a "O. ")

Using Contingent Valuation to Value Food Safety

249

Q-11 What kind of fresh grapefruit do you buy? (Circle all numbers that apply.)

1. 2. 3. 4.

WHITE SEEDY GRAPEFRUIT WHITE SEEDLESS GRAPEFRUIT RED/PINK SEEDY GRAPEFRUIT RED/PINK SEEDLESS GRAPEFRUIT 5. DOES NOT MATTER (any kind) 6. OTHER (Please s p e c i f y ) - - - - - - - - - - - - Q-12 Where do you usually buy fresh grapefruit? (Circle all numbers that apply.)

1. 2. 3. 4.

5.

6. 7. 8.

GROCERY STORE OR SUPERMARKET ROADSIDE STAND OR FARM FARMER'S MARKET MAIL ORDER ORGANIC FOOD STORE WHOLESALE OR COOPERATIVE FOOD MARKET I DON'T BUY GRAPEFRUIT, IT IS A GIFT OTHER (Please s p e c i f y ) - - - - - - - - - - - - -

Q-13 During which meal do you and your family normally eat fresh grapefruit? (Circle all numbers that apply.)

1. 2. 3. 4.

BREAKFAST LUNCH SNACK DINNER 5. OTHER, please specify _ _ _ _ _ _ _ _ _ _ _ _ __

Q-14 Are you male or female? (Circle one number.)

1. FEMALE 2. MALE

Jean Buzby, Jerry Skees, and Richard Ready

250

VERSION 1: 50% RISK REDUCTION SECTION 3: Reducing Pesticide Residues

PURPOSE: To learn what you think about health risks. The following example has been made up for this study and does not represent a real life situation. But, your help is really important. Please read all of this section before answering any questions. This section has only 3 quick and easy steps. STEP 1: UNDERSTAND THE SCENARIO

*

You walk into a store and want to buy grapefruit. There are two types of grapefruit: "A" and "B." Both types look and taste the same and have the same nutritional value.

*

In order to preserve quality, suppose that all grapefruit must be treated with either "Pesticide A" or "Pesticide B." Most of these pesticides stay on the peel and do not affect the appearance or taste of the grapefruit. "GRAPEFRUIT A"

.• t

"GRAPEFRUIT B"

~.- ,· · ._

,,,,.•.~- . ~ ·:.·

::\

.

.'

---·__ ......

*

"Grapefruit A" is treated with "Pesticide A," which has a very low chance of causing an early death of someone in your family.

*"Grapefruit B" is treated with "Pesticide B," which is twice as safe as the pesticide used on "Grapefruit A."

*

"Grapefruit A" normally costs about 50 cents each .

*"Grapefruit B" costs more than "Grapefruit A."

STEP 2: LOOK AT RISK LADDER (SEE NEXT PAGE)

*

Level 1 is the risk exposure over a lifetime by eating "Grapefruit A," the one treated with "Pesticide A." It is estimated that Pesticide A causes 5 deaths in every 100,000 people who consume "Grapefruit A" over a lifetime.

*

Level 2 is the risk exposure over a lifetime by·eating "Grapefruit B," the one treated with "Pesticide B." It is estimated that Pesticide B causes only 2.5 deaths in every 100,000 people who consume "Grapefruit B" over a lifetime.

*

If you choose to buy Grapefruit B instead of Grapefruit A, it would reduce your risk from Level 1 to Level 2 on the risk ladder.

Using Contingent Valuation to Value Food Safety

251

RISK LADDER COMPARING RISKS OF DEATH

This picture is a "risk ladder" which indicates the relative chance of dying from different causes. Please take a few moments to carefully consider the seriousness of the different risks. After you are familiar with the risk associated with eating Grapefruit A and Grapefruit B over a lifetime, please answer the final question Q-15.

Death Rates Per 100,000 Persons

166.3 132.7

car

ent

HIGH RISK

19.7

10.1

LOW RISK

9

stomach cancer

LEVEL 1 ....

5.5 5

4.9

3.8 LEVEL 2 ....

2.5 fire

2 1.4

.02 VERY LOW RISK 1he ladder is not drawn exactly to scale. These are the best available estimates.

252

Jean Buzby, Jerry Skees, and Richard Ready

VERSIONS 2 and 3: 99+% RISK REDUCTION SECTION 3: Reducing Pesticide Residues PURPOSE: To learn what you think about health risks. The foll.owing example has been made up for this study and does not represent a real life situation. But, your help is really important. Please read all of this section before answering any questions. This section has only 3 quick and easy steps.

STEP 1: UNDERSTAND THE SCENARIO

*

You walk into a store and want to buy grapefruit. There are two types of grapefruit: "A" and "B." Both types look and taste the same and have the same nutritional value.

*

In order to preserve quality, suppose that all grapefruit must be treated with either "Pesticide A" or "Pesticide B." Most of these pesticides stay on the peel and do not affect the appearance or taste of the grapefruit. "GRAPEFRUIT A"

.

"GRAPEFRUIT B"

-

.

'

. .

. .

*

"Grapefruit A" is treated with "Pesticide A," which has a very low chance of causing an early death of someone in your family.

*"Grapefruit B" is treated with "Pesticide B," which is relatively safer yet more expensive than the pesticide used on "Grapefruit A."

*

"Grapefruit A" normally costs about 50 cents each.

*"Grapefruit B" costs more than "Grapefruit A."

STEP 2: LOOK AT RISK LADDER (SEE NEXT PAGE) *

Level 1 is the risk exposure over a lifetime by eating "Grapefruit A," the one treated with "Pesticide A." It is estimated that Pesticide A causes 5 deaths in every 100,000 people who consume "Grapefruit A" over a lifetime.

*

Level 2 is the risk exposure over a lifetime by eating "Grapefruit B," the one treated with "Pesticide B." It is estimated that'Pesticide B causes only .0005 deaths in every 100,000 people (almost a 100% reduction) who consume "Grapefruit B" over a lifetime.

*

If you choose to buy Grapefruit B instead of Grapefruit A, it would reduce your risk from Level 1 to Level 2 on the risk ladder.

Using Contingent Valuation to Value Food Safety

253

RISK LADDER COMPARING RISKS OF DEATH

This picture is a "risk ladder" which indicates the relative chance of dying from different causes. Please take a few moments to carefully consider the seriousness of the different risks. After you are familiar with the risk associated with eating Grapefruit A and Grapefruit B over a lifetime, please answer the final question Q-15.

LEVEL 1 -.

Death Rates Per 100,000 Persons 168.3

car

HIGH RISK

132.7 19.7

10.1

LOW RISK

9

cancer 5.5 5 4.9 3.8

2 1.4

.02 LEVEL 2-.

.0005 VERY LOW RISK

1he 'ladder is not drawn exactly to scale. These are the best available

estimates.

254

Jean Buzby, Jerry Skees, and Richard Ready VERSIONS 1 and 2: PAYMENT CARD

Q-15

STEP 3: DETERMINE HOW MUCH YOU WOULD PAY TO AVOID THIS PESTICIDE

Please circle the one amount that indicates the most that you would pay above the purchase price of one "Grapefruit A" to buy each "Grapefruit 8."

* If you circle $0.00 this indicates that you would not pay more to buy the safer yet more expensive "Grapefruit B" and that you would buy "Grapefruit A" instead. Cents $0.00 $0.01 $0.02 $0.03 $0.04 $0.05 $0.06 $0.o7 $0.08 $0.09 $0.10 $0.12 $0.14 $0.16 $0.18 $0.20 $0.22 $0.24 $0.26 $0.28 $0.30 $0.32 $0.34 $0.36 $0.38 $0.40 $0.42 $0.44 $0.46 $0.48 $0.50

If greater than $.50 or if you would pay some number not shown above, please fill in number in box.

Using Contingent Valuation to Value Food Safety

255

VERSIONS 3 and 4: DICHOTOMOUS CHOICE STEP 3: DETERMINE HOW MUCH YOU WOULD PAY TO AVOID "PESTICIDE A" Suppose you walk into the store and you want to buy grapefruit. Both "Grapefruit A" and "Grapefruit B" are available.

*

"Grapefruit A" costs $0.50 per grapefruit.

*

"Grapefruit B" is treated with the relatively safer pesticide and costs $_ _ __ per grapefruit.

Q-15 Which kind of grapefruit would you buy? (Circle one number.) 1. Grapefruit A 2. Grapefruit B

If we assume that "Grapefruit A" is still available for $0.SO, what is the mQ!! that you would pay to buy each "Grapefruit B"'l (Please fill in amount in box.)

256

Jean Buzby, Jerry Skees, and Richard Ready

Q-16 If you would not pay more to buy "Grapefruit B" instead of "Grapefruit A," please use this space to tell us why you chose this answer.

Please use this page to write in any additional comments that you have.

THANK YOU FOR YOUR TIME. YOUR OPINIONS AND CONCERNS ARE VERY IMPORTANT TO US.

Qyou would like to receive a copy of the results of this survey, please write your name and address on the retu'?' envelope. Do not write your name on this survey.

The College of.Agriculture is an "Equal Opportunity Organization.

13 Anglers' Willingness to Pay for Information About Chemical Residues in Sport Fish: Design of a CV Questionnaire Douglas J. Krieger and John P. Hoehn 1 The contingent valuation (CV) approach asks individuals their willingness to pay (WTP) for a good in a hypothetical market setting. The resulting WTP estimates are conditioned on the assumption that respondents' stated choices correspond to their behavior in actual markets. Consumer behavior in actual markets depends on a variety of factors. These include characteristics of the good, characteristics of the payment, and the market itself (Fischhoff and Furby 1988). These factors exist explicitly in a hypothetical market only if they are described by the researcher. If a CV survey does not adequately describe the factors relevant to a decision, stated behavior will depend on what the respondent assumes. These assumptions may not correspond to the environment envisioned by the researcher. If respondents perceive a different market setting than the researcher intends, stated WTP may not reflect the values desired. The challenge of successful CV research is to clearly communicate a market , setting that most respondents interpret as the researcher intends. A clear description of all relevant aspects of the good, the payment, and a credible market, and clearly worded questions will reduce ambiguity and facilitate successful communication (Mitchell and Carson 1989). Thus, the design of the CV survey-the language, descriptions, and questions used-is crucial to obtaining good data. The implementation of a survey may also affect the quality and quantity of usable responses. Important considerations for a mail survey include the physical appearance of the survey instrument; the wording, order, and format of questions; the timing of mailings; and the process of following up on those who are slow to respond (Dillman 1978). 257

258

Douglas J. Krieger and John P. Hoehn

This chapter focuses on the design and implementation of a CV mail survey to assess anglers' WTP for information about chemical residues in sport fish. It illustrates the importance of pretesting to determine what aspects of the market description are important and to assess potential respondents' interpretation of specific questions and response choices. The first two sections outline the conceptual and policy setting for the survey design. The first reviews Michigan's current advisory program and two proposed alternatives. The second briefly outlines the conceptual definition of information value. The core of the chapter focuses on the process of designing the questionnaire. A summary section reviews our primary conclusions with respect to CV survey design.

Contaminants in Michigan's Sport Fish Tests of sport fish in the Great Lakes region have detected traces of chemical residues. In sufficient doses some of these chemicals are suspected of causing adverse human health effects. In response to this problem, state and provincial governments bordering the Great Lakes provide anglers with information about chemical residues in fish. These warnings typically take the form of public health advisories. These advisories list sites and species known to be contaminated and offer advice on how to reduce the risk of exposure (Hesse 1990). Michigan's advisory is issued by the Department of Public Health (MDPH) and printed in the annual fishing guide distributed by the Michigan Department of Natural Resources (MDNR). The current advisory contains a brief description of some chemicals found in Michigan's fish, their sources, and possible health effects. It also warns that concentrations of contaminants are likely to be higher in larger or older fish, predator species, fatty fish, and carp or catfish. It suggests that anglers who intend to eat their catch should trim and cook them to remove fat. A pictorial presentation illustrates the suggested method of trimming fish. The advisory also contains a table listing sites that have been found to be contaminated. Each entry cites the species and sizes that are contaminated, the specific chemical responsible for the advisory, and specific consumption advice. Depending on the extent of contamination, the advisory suggests either (1) no consumption, or (2) restricting consumption to no more than one meal per week. It also implies that no consumption restrictions are necessary for sites or species not mentioned in the advisory. It warns nursing mothers, pregnant women, women who intend to have children, and children under the age of 15 against eating any fish from either of the restricted categories (Hesse 1990). Since 1989 the advisory has also included a special warning about mercury in inland lakes. Because of widespread mercury contamination, the advisory warns anglers

Anglers' Willingness to Pay for Information About Chemical Residues

259

against eating more than one meal per week of specific species and sii.es of fish from any inland lake.

Adrisory Alternati.ves The advisories issued by various jurisdictions in the Great Lakes region differ in a number of dimensions (Foran and Vanderploeg 1989, Hesse 1990). These differences suggest a feasible set of alternatives to Michigan's current advisories. Probably the greatest difference between existing advisories is the treatment of sites where tested fish are found not to contain dangerous levels of contaminants. The advisories from Ontario and Minnesota include a list of such sites. However, most advisories (Michigan's included) tell anglers only about the sites and species that contain chemical concentrations exceeding state standards. Alternatives to Michigan's current advisory range from eliminating the advisory program altogether to an expanded program similar to Ontario's. Not all of these alternatives are politically or fiscally feasible. Discussions with state officials revealed two informational changes that could be incorporated in future advisory programs. 2 First, the current advisory program only partially discloses test results. It does not tell anglers about tested sites where chemical residues are not found or are below state standards. Alternatively, the advisory could fully disclose test results-provide a list of contaminated sites and a list of sites that posed little or no health risk. Since the current testing program already generates a list of relatively safe sites, printing costs are the primary constraint in implementing a full disclosure program. A second alternative is to test a greater number of sites with either full or partial disclosure of test results. Michigan contains over S,800 publicly accessible fishing sites. The current budget permits testing about 30 of these sites each year. 3 Information about a greater number of sites would provide anglers a broader base of knowledge about the risks associated with their fishing choices. With this information they could make choices more consistent with their preferences for risk bearing. Survey results suggest that anglers are interested in information about a greater number of sites (Ontario Ministries of the Environment and Natural Resources 1990).

The Value of Infonnation About Chemical Residues in Fish Information about potential risks to health has value because it helps people prevent mistakes. A mistake is an action taken in ignorance that a person would not have chosen if informed. Consider an angler who chooses among attributes of fishing so as to maximii.e utility. Characteristics that might affect anglers' utility include the species, number, and sii.e of fish caught (Vaughan and Russell 1982), aesthetic and physical characteristics of a fishing site, or the presence or

260

Douglas J. Krieger and John P. Hoehn

absence of chemical residues in the fish (Kikuchi 1986). Suppose the angler also cares about his or her state of health. The angler understands that health depends, in part, on exposure to chemical residues in fish. However, exposure is a random variable that depends on the unknown level of contamination in fish consumed. With uncertainty about the state of contamination, s, the angler chooses attributes of fishing, q = {q1,q2, ... ,C1J}, to maximi7.e expected utility over the perceived distribution of s, P(s). Define q• as the optimal choice given beliefs P(s). With q•, the angler achieves an expected utility of: (1)

Ju(q •, h(q •, s))P(s)ds = PrPr

where u( • ) is the utility function and h( • ) relates behavior and perceived contamination to health. Define this utility level as the prior utility of the prior act, PrPr· Now suppose the angler receives a message, y (new information in the advisory), that changes beliefs about the distribution of s. Represent these new beliefs by the Bayesian posterior distribution, P(s Iy), contingent on message y. Given new perceptions of contamination, the angler may believe his or her behavior prior to receipt of the message to be a mistake. The utility associated with the mistake is: (2)

The prior optimal act is a mistake because it is evaluated relative to posterior perceptions of contamination. Define this utility as the posterior utility of the prior act, P0 PrGiven new beliefs about contamination, the angler may wish to change behavior. Define optimal behavior after receipt of message y by qY. The utility associated with the posterior optimal behavior is: (3)

Define this as the posterior utility of the posterior act, P 0 P 0 • The value of information is the difference in utility it makes possible relative to a state of ignorance. However, the definition of prior utility must be considered with care. Realii.ed prior utility depends on the actual distribution of s. Define this as P(s'). For illustration, suppose P(s') represents a greater risk of contamination than P(s). The following analysis applies as well to the case where P(s') represents a lower risk than P(s). Ignorance of actual risks

Anglers' Willingness to Pay for Information About Chemical Residues

261

P(s') does not shield uninformed anglers from actual conditions. Both informed and uninformed anglers face possible future health effects determined by P(s'). The advantage of being informed is the opportunity to adjust behavior. To avoid issues of risk perception, assume that informed anglers perceive risks accurately (i.e., P(s Iy) = P(s')). The value of message y is the difference between the expected utility when informed, P0 P0 , and the expected utility of a mistake, P 0 P r: (4)

vy =

Ju(qY, h(qY, s))P(s Iy)tb - Ju(q • , h(q •, s))P(s Iy)tb.

By definition, qY represents optimal behavior given beliefs P(s Iy). Therefore, the first term on the right-band side of equation 4 is at least as large as the second and the value of message y must be nonnegative. The utility values derived above can be expressed in terms ofWTP using the expenditure function. Suppose improvements in the advisory alter perceptions of risk from chemical residues in fish. WTP for the additional information is the change in income required to maintain utility under posterior risk perceptions when behavioral adjustment is not permitted (Foster and Just 1989). Thus WTP for the current advisory, testing more sites, or full disclosure is the compensation anglers would require to continue their prior behavior after they are aware of the new information.

Contingent Valuation Survey Design and Implementation A contingent valuation survey designed to elicit WTP for advisory alternatives must first clearly communicate ideas to potential respondents. This means that questions must use language and concepts that are familiar and meaningful to anglers. We used three focus groups to explore bow anglers think about chemical residues in fish, the language they use to talk about it, and bow they think about and respond to the current advisory. Insights from focus groups aided in designing a draft questionnaire to assess anglers' WTP for advisory information. We pretested the draft questionnaire in personal interviews with twelve anglers. Participants were asked to talk through their response to the questionnaire. The interviews identified questions that were unclear or ambiguous and helped assess the adequacy of response choices. They were particularly useful in identifying assumptions respondents made about factors not explicitly mentioned in the questionnaire. They also helped identify irrelevant and redundant questions that could be removed or combined with other questions, thus reducing the length of the questionnaire. The process reduced a wordy and visually intimidating questionnaire of 46 questions to a clear and

262

Douglas J. Krieger and John P. Hoehn

interesting survey containing 15 questions. The remainder of this section highlights some of the more important lessons learned in focus groups and interviews and illustrates how they influenced the form of specific questions.

Eliciting WTP We used a referendum format to elicit WTP. The referendum approach presents respondents with the dichotomous choice of voting for or against a program offered at a specified cost. The approach produces binary yes/no responses instead of the direct dollar amounts generated by methods such as bidding games, payment cards, or open-ended formats. We used the referendum format for two reasons. First, it is probably easier for respondents than open-ended questions (McConnell 1990, Hanemann 1985). The decision to accept or reject a good at a given price is the most common type of market transaction people make. Second, people are familiar with the idea of voting on public programs. Ballots often contain measures offering public goods such as schools, water systems, sewer systems, or roads, for a given increase in taxes or fees (Mitchell and Carson 1989). Also, Hoehn and Randall (1987) conclude that the referendum format reduces the incentives for strategic over- or understatement of WTP. The questionnaire described an alternative to the current advisory and specified a cost in terms of a permanent annual increase in the cost of a fishing license. The license fee payment vehicle is appealing primarily because fees are collected by the state and people link them directly to spending on state programs. We tested an alternative payment vehicle in focus groups and interviews-a separate advisory booklet offered for sale where fishing licenses are sold. The booklet did not adequately capture WTP for information because it is difficult to restrict access. Many people stated they would read the booklet without buying it or they would share one with friends. The following comments are typical reactions to the booklet: "We couldn't [buy the booklet] or we'd have to go in with some friends or go borrow somebody's book." "And I think that's what would happen. I think a lot of people would stand at the counter and look up their section and then go on." "I'd probably stand there and look through it, look up my site, and set it back down."

By contrast, the license fee payment is more difficult to avoid if respondents wish to continue fishing. The license fee payment vehicle also had drawbacks. Several focus group participants thought it unlikely that the money collected would actually be used to improve advisories. As a basis for their beliefs, they cited several past

Anglers' Willingness to Pay for lnformalion About Chemical Residues

263

examples where the state reallocated collected funds. One participant phrased his concerns as follows: "I wouldn't mind paying a little bit more in a fishing license if they earmarked that's what it's going for. . .. if they're not taking money away from the other part of the license to do something else with. n

For the final questionnaire, we chose the license fee payment vehicle over the booklet. To control for objections to using a license fee increase, the questionnaire explicitly asked whether respondents objected to the license fee as a method of raising money for advisories. WTP Question Fonnat. A challenge in this particular survey was to clearly explain subtle changes in a complex public program. To obtain valid measures of WTP for changes in the advisory, respondents must first clearly understand the content of the current advisory. Focus groups revealed that many anglers believed that most sites had been tested. In reality, the state has tested fewer than three percent of all sites. Thus, anglers generally have greater confidence in the scope of advisory information than is warranted by actual testing. To provide anglers with an accurate and common point of reference, the questionnaire described the information content of the current advisory program. One dimension of information content is the proportion of sites tested. The initial WTP questions (reproduced in Figure 13.1) told respondents how many sites had actually been tested. However, the questions did not adequately identify the total number of sites. In the absence of an explicit denominator, respondents made different assumptions about the proportion of sites tested under a proposed program. Discussion with MDNR officials produced a more explicit definition of sites used in the final questionnaire. In addition to a clearer definition of sites, the interviews also prompted substantial revisions in question wording and format to improve clarity. The initial format of the three WTP questions did not adequately emphasize the information content of the current advisory, how the current advisory differed from proposed alternatives, or differences between alternatives. To remedy these problems, we further revised the question format. The final questionnaire explained details of the current program and some possible improvements on a separate page (see Figure 13.2). The suggested improvements corresponded to the alternatives of full disclosure and increased testing. Interview results were particularly useful in improving the brevity and clarity of the description. Interview participants quickly and accurately picked up the important features of the current advisory program from the written description when it was presented separately from the WTP questions. The page facing the description of the current advisory contained a description of alternative programs and the WTP question. Focus group results suggested that listing advisory programs in tabular form provided a clearer

264

Douglas J. Krieger and John P. Hoehn

FIGURE 13.1 Initial Wording of WTP Questions

1.

Fish from all of Michigan's Great Lakes have been tested for chemical residues. Michigan also contains 11,000 or so inland lakes and over 36,000 miles of rivers and streams. Only 331 tests have been conducted on these inland waters. The current advisory contains a list of water bodies where fish have been found to be unsafe. Water bodies that have been tested and found to be safe are not listed in the advisory. Suppose the state could no longer afford to test fish and print advisories unless fishing license fees were increased. Would you rather, (1) eliminate the current advisory program and keep fishing license costs the same, or (2) keep the advisory if it meant increasing license costs by $3.00? (Circle one number)

1.

2. 2.

Eliminate the advisory Keep the advisory

Suppose that an up-to-date list of 600 sites that had been tested and found to be safe could be printed in the advisory if money for printing was available. Would you rather, (1) keep advisories as they are and keep fishing license costs the same, or (2) include a list of 600 safe sites if it meant increasing license costs by $2.00? (Circle one number) 1.

2. 3.

Keep advisories the same Include list of safe sites

Only 331 tests for chemical residues have been conducted in Michigan's inland waters. About 30 new sites are tested each year. Suppose more sites could be tested if license fees were increased. Would you rather, (1) continue to test about 30 sites per year and keep fishing license costs the same, or (2) increase testing to 100 new sites per year if it meant increasing license costs by $4.00? (Circle one number)

1. 2.

Continue current testing Increase testing

comparison for respondents. Many participants said they were more likely to read and easily understand information presented in tables. The resulting tabular comparison of an intermediate form of the WTP questions depicted in Figure

Anglers' Willingness to Pay for Information About Chemical Residues

265

FIGURE 13.2 Final Description of Current Advisory

Michigan's Public Health Admory There are more than 5,800 public fishing sites in Michigan-2,200 sites on rivers and streams, 3,600 inland lakes, and the Great Lakes. The state has tested 350 of these sites for chemical residues in fish. About 30 new sites are tested each year. The current public health advisory tells you: •

that you should not eat too much fish from any inland lake because of widespread mercury contamination, and

•

it lists 50 sites where fish contain chemical residues above state limits.

The advisory does not tell you about: •

the 300 tested sites where chemical residues do not exist or are below state limits.

The advisory program could be changed: •

The advisory could list tested sites where chemical residues do not exist or are below state limits.

•

More than 30 new sites could be tested each year.

These changes would increase the amount of information in the advisory but they would also cost more money.

13.3 increased the ease with which respondents were able to compare programs. However, some respondents still failed to view the programs as independent. When asked for WTP for testing more sites they might say, "Well, if I'm already paying X dollars to list safe sites. . . • The final form of the WTP questions (reproduced in Figure 13.4) asked respondents to value only one program alternative. The proposed alternative tested more sites and either partially or fully disclosed test results. The final wording of the WTP questions also clarified the decision context and placed more emphasis on voting as a choice mechanism. A number of respondents throughout the pretest process voiced strong feelings about the integrity of the MDNR. Typical of these responses were:

266

Douglas J. Krieger and John P. Hoehn

FIGURE 13.3 Intermediate Form of WfP Questions

Table 1-Advisory Programs and their Cost to You Current Advisory

Program A

Program B

No

No

Yes

Number of new sites tested each year.

0

30

0

Cost to you in higher license fees.

$0

$5

$3

Program Options Lists tested sites where nonmercury residues pose little or no health risk.

In the next two questions suppose anglers could vote on Programs A and B. Vote for the program if it is worth the additional cost to you. Vote against the program if it is not worth the additional cost. Think about how important these programs are compared to other ways you could spend your limited budget.

1.

Would you vote for Program A if it permanently increased your yearly license cost by $5.00, or vote against it and keep the Current Advisory at no additional cost?

1. 2. 2.

Vote for Program A Vote against Program A and keep Current Advisory

Would you vote for Program B if it permanently increased your license cost by $3.00, or vote against it and keep the Current Advisory at no additional cost?

1. 2.

Vote for Program B Vote against Program Band keep Current Advisory

"People come from all over the United States to fish here and ifthe state DNR tells people just how contaminated some of these fish are, that would scare a lot of them away." "I think the DNR is more politicized than the Department of Health. They listen to too many special interests ... "

In general, respondents seemed to view the MDNR as the source of the advisory even though the advisories are actually issued by the MDPH. When questioned,

267

Anglers' Willingness to Pay for Infonnation About Chemical Residues

FIGURE 13.4 Final Form of WfP Question

Your Vote on Advisory Programs The table below shows two advisory programs. The "Current Advisory" is Michigan's current advisory program. "Program A" is a different program that could be put in place.

Program Options Lists tested sites where chemical residues are above state limits?

Current Advisory Yes

Program A Yes

Lists tested sites where chemical residues do not exist or are below state limits?

No

Yes

Number of new sites tested each year.

30

400

$0.00

$4.00

Cost to you in higher fishing license fees.

Suppose the Michigan Department of Natural Resources (DNR) sent you a ballot to vote "for" or "against" Program A. If a majority vote "for" Program A, it will replace the Current Advisory. If a majority vote "against" Program A, the Current Advisory will be continued. 9.

Would you vote for Program A if it permanently increased your yearly license cost by $4.00, or vote against it and keep the Current Advisory at no additional license cost?

1. 2. 3.

Vote for Program A Vote against Program A and keep Current Advisory Don't know or no opinion

respondents seemed to trust the MDPH more than the MDNR to provide unbiased information. To facilitate analysis of the importance of the information source, the final questionnaire was designed as a split sample with both the MDNR and MDPH as the stated source of the advisory. Finally, to make the questionnaire less intimidating and more easily read, the format was increased from a booklet measuring 51h by 81h inches to one

Douglas J. Krieger and John P. Hoehn

268

measuring 8 1h by 11 inches and the type siz.e increased. Appendix 13.A reproduces the complete text of the final questionnaire. A comparison of WTP estimates from the pretest and final surveys emphasi7.es the importance of clear descriptions of the contingent market. The pretest survey used the questions of Figure 13.1 to ask for respondent's WTP for (1) continuation of the current advisory program, (2) a partial disclosure program that tested either 100 or 300 sites per year, and (3) a full disclosure program that listed either 200 or 600 safe sites. An open-ended question that asked for respondents' maximum WTP followed each of the questions. Table 13.1 lists mean WTP values for each program offered in the pretest. Statistical analysis reveals no significant difference in WTP between mean bids for any of the programs described at a 1 percent level of significance. The similarity of the mean bids suggests that respondents may not have understood or responded to the differences between programs. The uniformity of bids within surveys further supports the notion that respondents did not interpret the programs as intended. A majority of respondents (72 percent) stated the same WTP for full as for partial disclosure advisories and 56 percent stated identical WTP for the current advisory, full disclosure, and partial disclosure with increased testing. This result seems remarkable given the significant quantitative differences between proposed programs. However, the result could arise if respondents (1) understand the differences but do not perceive them to be large, (2) understand the differences but do not view them as essential to the good or program to be valued, or (3) do not clearly perceive the differences due to poor question design. The pretest process improved respondents' comprehension of the differences between program alternatives. Final estimates exhibited a significant difference in WTP between full and partial disclosure programs. They also revealed a significant positive marginal WTP for testing more sites with full disclosure. In this case study, meaningful WTP responses depended crucially on a clear description of the contingent market. Pretesting, in turn, was a necessary step TABLE 13.1 Pretest WI'P Means

Program Partial Disclosure Partial Disclosure Partial Disclosure Full Disclosure Full Disclosure

Number of Sites

Mean WI'P

30 100 300 200

$3.49 $3.57 $3.06 $2.82 $2.95

600

Anglers' Willingness to Pay for Information About Chemical Residues

269

in the evolution of a clear survey. In retrospect, had the research proceeded from the initial draft of the survey without pretesting, many questions would not have measured intended concepts and the empirical results would likely have been meaningless. Experimental Desi.gn. The questionnaire described alternative programs in terms of (1) the number of new sites tested each year, (2) whether safe sites were listed, and (3) the cost in higher license fees. The experimental design consisted of combinations of three different testing levels, ten levels of program costs, whether safe sites were listed or not, and the state agency responsible for the advisory. A complete factorial design over these factors defined 120 unique combinations that were used for the final questionnaires. The questionnaire offered testing levels of 110, 620, and 1,240 new sites per year. These levels represented a range that was physically feasible for the state. Program costs were derived from the results of a mail pretest of 200 licensed anglers in the Lansing, Michigan area. The pretest asked WTP for program alternatives in an open-ended format. The final referendum format set bid levels at decile boundaries from the cumulative distribution of open-ended pretest bids. The final survey used bid amounts of $.40, $.95, $1.45, $1.90, $2.85, $4.10, $5.55, $8.75, $14.50, and $41.00. We chose the high bid in hopes of eliminating any positive response, thus avoiding the problem of arbitrarily truncating the empirical distribution. The distribution of bids and the number of questionnaires prepared with each bid amount can influence the statistical properties of the WTP estimator (Duffield and Patterson 1991, Boyle et al. 1988, Kanninen 1993, Cooper 1993). Strategies to minimize the variance of the WTP estimator use prior knowledge about the distribution of WTP to attempt to cluster bids around the true mean WTP. This minimizes the number of surveys •wasted• on bids far from the mean. This study employed a mail pretest to gain some prior knowledge about the distribution of WTP. However, it does not attempt an optimal allocation of bids. The pretest asked valuation questions that were somewhat different than those used in the final survey. The questionable quality of the pretest data relative to the final WTP questions seemed to dictate a conservative approach that reduced the risk of clustering many offered bids around a point far from the true mean. Focus group results also raised an important consideration in the choice of bid amounts. When asked their WTP for a particular program, participants' responses were often conditioned on their perceptions of a reasonable cost for providing the program. They seemed concerned that they were getting their money's worth and not paying more than their fair share. The following responses illustrate the nature of this concern. "Well, I have no idea how much this [publishing a list of safe sites] costs so it's really kind of hard to sit here and hem and haw over how many dimes I would ... actually give towards it."

270

Douglas J. Krieger and John P. Hoehn

" ... for a 30 page pamphlet or somethin', five bucks, well, they're just making money off it." "What's it cost to test a site? ... So, how many fishing licenses do they sell in a year? I mean, if they tack on SO cents a fishing license, how much money ... "

Fortunately, the bids determined from the pretest corresponded reasonably well to estimates of actual program costs. State officials estimated that about 100 additional sites could be tested for each $1.00 increase in license fees. This implies actual costs ranging from about $1.10 to $12.40 for the testing levels used in the questionnaire.

Explanatory Variables In addition to the advisory program characteristics discussed above, three classes of variables influence anglers' WTP for advisory alternatives. These are (1) the possibility of behavioral change to avert risk, (2) the perceived accuracy of advisories, and (3) the perceived severity of health consequences resulting from consumption of contaminated fish. This section reviews the development of selected questions to measure these explanatory variables. The section emphasizes the role of focus groups and pretest interviews in creating meaningful questions. Behavioral Change. The questionnaire assessed anticipated behavioral change resulting from listing safe sites or testing more sites. The advisory suggests a number of changes in behavior that can reduce risk. We used focus groups to explore which of these actions anglers were aware of and which they were likely to use. These focus group discussions influenced the form of pretest questions dealing with behavioral change. Pretest questions asked if respondents would make a specific behavioral change in response to a given change in the advisory. Interviews revealed that response choices were not rich enough. Participants often mentioned behaviors that were not included among the response categories. As an example of the issues that arose when designing these questions, consider the question used to assess behavioral response to full disclosure of test results. Figure 13.5 lists both the pretest and final versions of this question. The final version improves on the pretest version in two ways. First, the answer to the question will likely depend on whether a respondent believes the sites they use will be listed as safe or not. The pretest version provides no means to determine what assumptions respondents make regarding the sites they use. Second, the pretest version seems to lead the respondent by asking for response to a specific behavioral change we believed to be important. The final version lets respondents choose from a more comprehensive list of behavioral changes that focus groups and interviews suggested were relevant to anglers.

Anglers' Willingness to Pay for !'!formation About Chemical Residues

271

FIGURE 13.5 Behavioral Change Questions

Initial Question 1.

If sites that were tested and found to be safe were listed in the "Public Health

Advisory" would you try to fish at those sites rather than sites that were not listed? (Circle one number)

1. 2.

Yes No

Final Question

1.

In addition to the list of unsafe sites, suppose the advisory listed all tested sites where chemical residues in fish did not exist or were below state limits. If your favorite sites had not been tested would you ...1 (Circle all that apply)

1. 2. 3. 4.

Continue to rash at your favorite sites Stop eating rash from your favorite sites Fish only at sites where chemical residues are below state limits When choosing a new site, be more likely to go to a site where chemical residues were below state limits

Risk. We made few changes in the format of questions aimed at assessing risk perceptions between the pretest and final versions of the questionnaire. The questionnaire asked respondents for their guess about the probability of having health problems someday because of chemical residues in fish. Response categories covered a roughly logarithmic scale. These were: 1. no chance 2. 1 in a million 3. 1 in 100,000 4. 1 in 10,000 5. 1in1,000

6. 7. 8. 9. 10.

1in100 1in10 1in5 1in2 certain to happen

Focus group and interview participants generally perceived very small risks associated with chemical residues in fish. The logarithmic scale concentrates responses around small risks and follows the approach of other studies designed to measure small perceived risks (van Ravenswaay and Hoehn 1991). People generally found this question difficult. Subsequent interviews focused on the source of difficulty. In general, respondents had little difficulty inter-

272

Douglas J. Krieger and John P. Hoehn

preting risks as stated. One participant interpreted the logarithmic scale as linear but soon realiz.ed his mistake. The hesitation in responding to the question arose primarily from uncertainty about the accuracy of the guess. A typical reaction when asked to guess about the probability of a health problem was: "I don't know. . .. really, really slim I think. I don't think it's no chance. I'll guess one in a million. I'm really not sure."

The final version of the questionnaire followed this question by asking how sure people were of their guess. lnfoTllUltion Accuracy. The conceptual model identifies three factors that may affect the perceived accuracy of the advisories. These are (1) the perceived accuracy of tests to identify chemical residues in fish, (2) the adequacy of scientific knowledge linking exposure to health effects, and (3) trust in the state to impartially report test results. Interviews and the mail pretest revealed that few people questioned the accuracy of the tests themselves. Consequently, this question was eliminated from the final questionnaire. Pretesting resulted in relatively minor changes in the wording and format of questions addressing the other two sources of perceived accuracy. The questionnaire obtained poor measures of perceived accuracy. Perhaps the most severe problem is that it measured separate dimensions of accuracy-beliefs about the accuracy of scientific knowledge and the perceived accuracy of reporting. Measures of only some of the individual dimensions of perceived accuracy identify some of the anglers who believe the advisory to be inaccurate but not those who believe it to be accurate. For instance, a belief that any single dimension is inaccurate implies a belief that the advisory as a whole is inaccurate. Therefore, a respondent who believes any measured dimension to be inaccurate believes the advisory to be inaccurate. However, a respondent who believes all measured dimensions to be accurate may still believe the advisory to be inaccurate if they question the accuracy of an unmeasured dimension. Because the questionnaire does not measure all dimensions of perceived accuracy, it does not identify respondents who believe the advisory to be accurate.

Sampling Frame and Survey Implementation Chemical residues in Michigan's fish potentially affect three groups of individuals. These include current licensed anglers, those who do not fish but would if residues were not present, and those who do not fish but eat fish caught by others. For practical and conceptual reasons, this research focused on currently licensed anglers. From a practical perspective, licensed anglers are an easy group to identify-names and addresses are obtained by the MDNR when a license is purchased. Those who do not fish because of chemical

Anglers' Willingness to Pay for Information About Chemical Residues

273

residues and those who eat fish caught by others are relatively difficult and expensive to identify. The group who would fish if contaminants were not present is likely small. A casual examination of fishing license sales records revealed no noticeable decrease associated with the appearance of the advisories in 1970. The sampling frame does not include this group because of the difficulty and expense of identifying and reaching them. The number of people who eat fish caught by others is difficult to assess. This includes fish purchased in restaurants and stores and fish received from acquaintances who fish. Twenty percent of New York anglers reported giving away some of their catch (Knuth and Velicer 1990). Furthermore, commercial fishing operations in Michigan landed 15.7 million pounds of fish in the Great Lakes in 1988 (Michigan Department of Natural Resources 1990). This research examines the value of public health advisories designed to influence sport angling behavior. Thus, while the group of people who consume fish they do not catch is potentially large, they are not as likely to be directly influenced by the advisory. The sampling frame for this research consisted of individuals who purchased a Michigan fishing license for the 1991 fishing season. 4 The Fisheries Division of the MDNR provided a random sample of 1,578 anglers licensed to fish in 1991. For each angler we obtained name, address, birth date, and type of license purchased. Rodabaugh (1987) reports that 12.5-14.5 percent of surveyed anglers fished without purchasing a license. This figure reflects the actions of anglers who reside within one mile of the Shiawassee River and may represent the more casual anglers. Whatever the composition of this group, this survey will not include them. Design and implementation of the survey followed Dillman's (1978) total design method (TDM). The TDM stresses the many small details which, when taken together, have a potentially large impact on response rates and the quality of data from mail and phone surveys. Following the TDM, we sent the first mailing of 1,578 questionnaires on Tuesday, February 9, 1993. One week after the initial mailing we sent a reminder postcard to prompt response and to thank those who may already have responded. Three weeks after the initial mailing, we sent a second copy of the survey to the 1,012 members of the sample population who had not yet returned a completed questionnaire. Finally, seven weeks after the first mailing, we sent a third copy of the questionnaire by certified mail to the 576 people who had not yet responded. Of the 1,578 questionnaires originally sent, 230 were returned as undeliverable yielding a final sample of 1,348 anglers. The survey achieved an overall response rate of 73.4 percent. Figure 13.6 graphs the pattern of returns and illustrates the effect of each contact with respondents. Each contact prompted an increase in overall response. However, the magnitude of the response decreased with each contact as the remaining individuals

Douglas J. Krieger and John P. Hoehn

274

70

Su

Returned

60 50

40 30

20 10 0

/"/

/"'

Days

,.~

FIGURE 13.6 Pattern of Survey Response

were less inclined to complete the questionnaire. The first mailing obtained a response rate of 30 percent; the second, 25 percent; and the third, 24 percent. It is surprising that the final mailing elicited almost the same rate of response as the second. The following comment from a returned questionnaire suggests that the use of certified mail may have increased response rates relative to regular mail: "In-as-much that you have gone to the expense of sending this via certified mail, I felt that it would behoove me to fill out and return. . . . I do admire your efforts in safeguarding the health and safety of Michigan sportsmen. 11

The third contact also ran the risk of angering some people. The following comment expresses a common theme a bit more creatively than usual. This was written on the cover letter and returned without a completed questionnaire: "You can take this as my answer, if I were interested in answering this questionnaire I would have sent the first one back. . .. Now that you've wasted enough of the taxpayer's money to mail me three envelopes@ $1.52 each, you can save us all some money and use this paper constructively the next time you visit your favorite john. Thank you very much for your time. 11

Anglers' Willingness to Pay for Information About Chemical Residues

275

In spite of some negative responses, the certified mailing resulted in 150 additional completed questionnaires. This represents 15 percent of the overall response.

Conclusion More than anything else, this study emphasized the importance of interaction with potential respondents prior to and during the process of writing a questionnaire. Without such interaction, researchers must trust that respondents interpret questions and responses as they themselves do. They must also take for granted that questions are clear and unambiguous and do not elicit strong emotions or beliefs that may influence responses. Such assumptions in this study would have been a mistake. Pretest participants routinely interpreted questions, made assumptions, and reached conclusions that were unanticipated and inconsistent with the intended focus of a question. Focus groups and repeated one-on-one interviews refined the draft questionnaires. Revisions generated a set of questions that were interpreted in a like manner by most respondents. They also ensured that questions evoked interpretations consistent with our intentions. For example, revisions to the WTP question corrected an erroneous interpretation of the current advisory, clarified the differences between advisory alternatives, and eliminated the problem of interdependent valuation of several programs. Also, revisions of questions about anticipated behavioral change included a richer set of response choices that corresponded to those actually perceived as options by anglers. Revisions based on pretesting improved the quality of data obtained from the CV survey. Mean WTP estimates from the mail pretest exhibited no significant variation across programs or levels of testing. However, final estimates of WTP were consistent with theoretical expectations: WTP for a partial disclosure program was not significantly different from 7.ero, WTP for full disclosure was positive and significant, and the marginal value of testing an additional site with full disclosure was positive.

Notes 1. The authors would like to thank the Michigan Sea Grant College Program for support of this project 2. Personal communication with John Hesse, Chief, Environmental Health Assessment Division, Michigan Department of Public Health. 3. The size of the monitoring program is constrained primarily by the budget for chemical analysis which in 1992 totaled $320,000. 4. Michigan fishing licenses are valid through the end of March of the year following their issue.

276

Douglas J. Krieger and John P. Hoehn

References Boyle, Kevin J., Michael P. Welsh, and Richard C. Bishop. 1988. Validation of Empirical Measures of Welfare Change: Comment. Land Economics 64(February): 94-98. Cooper, Joseph C. 1993. Optimal Bid Selection for Dichotomous Choice Contingent Valuation Surveys. Journal of Environmental Economics and Management 24(January):2S-40. Dillman, Don A. 1978. Mail and Telephone Surveys: 'lhe Total Design Method. New York, NY: John Wtley & Sons, Inc. Duffield, John W. and David A. Patterson. 1991. Inference and Optimal Design for a Welfare Measure in Dichotomous Choice Contingent Valuation. Land Economics 67(May):22S-239. Fischhoff, Baruch and Lita Furby. 1988. Measuring Values: A Conceptual Framework for Interpreting Transactions with Special Reference to Contingent Valuation of Visibility. Journal of Risk and Uncertainty l(June):147-184. Foran, Jeffery A. and David Vanderploeg. 1989. Consumption Advisories for Sport Fish in the Great Lakes Basin: Jurisdictional Inconsistencies. Journal of Great Lakes Research 15(3):476-485. Foster, Wtlliam and Richard E. Just. 1989. Measuring Welfare Effects of Product Contamination with Consumer Uncertainty. Journal of Environmental Economics and Management 17:266-283. Hanemann, W. Michael. 1985. Some Issues in Continuous- and Disc~Repon Contingent Valuation Studies. Northeastern Journal of Agricullural and Resource Economics 14(Apri1):5-13. Hesse, John L. 1990. Summary and Analysis of Existing Sportfish Consumption Advisory Programs in the Great Lakes Basin. Unpublished Paper, Michigan Department of Public Health. Hoehn, John P. and Alan Randall. 1987. A Satisfactory Benefit Cost Indicator from Contingent Valuation. Journal of Environmental Economics and Management 14(June):226-247. Kanninen, Barbara J. 1993. Design of Sequential Experiments for Contingent Valuation Studies. Journal of Environmental Economics and Management 25(July):Sl-11. Kikuchi, Hideo. 1986. Segmenting Michigan's Sport Fishing Market: Evaluation of Two Approaches. Ph.D. Dissertation, Department of Parks and Recreation Resources, Michigan State University, East Lansing, MI. Knuth, Barbara A. and Christine M. Velicer. 1990. Receiver-Centered Risk Communication for Sportfisheries: Lessons from New York Licensed Anglers. Paper Presented at the American Fisheries Society Annual Meeting, Pittsburgh, PA. McConnell, Kenneth E. 1990. Models for Referendum Data: The Structure of Discrete Choice Models for Contingent Valuation. Journal of Environmental Economics and Management 18(March): 19-34. Michigan Department of Natural Resources. 1990. 1980-1990 Michigan Fisheries: A Foundation for the Future. Fisheries Special Report No. 13, ed. W. C. Latta. Lansing, Ml.

Anglers' Willingness to Pay for Information About Chemical Residues

277

Mitchell, Robert Cameron and Richard T. Carson. 1989. Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington, D.C.: Resources for the Future. Ontario Ministries of the Environment and Natural Resources. 1990. Guide to Eating Ontario Sport Fish. Toronto, Ontario. Rodabaugh, Gary L. 1987. An Investigation into Processes Contnbuting to Voluntary Exposure of Michigan Anglers to Contaminated Waterways and Contaminated Fish. Ph.D. Dissertation, Department of Fisheries, Michigan State University, East Lansing, MI. van Ravenswaay, Eileen 0. and John P. Hoehn. 1991. Consumer Willingness to Pay for Reducing Pesticide Residues in Food: Results of a Nationwide Survey. Department of Agricultural Economics Staff Paper #91-18, Michigan State University, East Lansing, MI. Vaughan, William J. and Clifford S. Russell. 1982. Freshwater Recreational Fishing: The National Benefits of Water Pollution Control. Research Paper. Washington, D.C.: Resources for the Future.

278

Appendix 13.A

MICHIGAN'S SPORT FISH CONSUMPTION ADVISORIES

Copyright 1993 Michigan State University Reprinted by permission.

Anglers' Willingness to Pay for Information About Chemical Residues

1.

Do you regularly do any of the following? (Circle all that apply)

1. 2. 3. 4. 2.

279

Firearm or bow hunting Bird or wildlife viewing Camping None of the above

About how many times per year do you fish at the following types of sites? (Fill in numbers)

1. Great Lakes 2. _ _ Inland lakes or ponm 3. Rivers or streams Other 4. 3.

On average, throughout the year, about how often do you eat fish that you catch in Michigan? (Circle one number)

1. 2. 3. 4. 4.

I do not eat fish that I catch Less than one meal per week About one meal per week Two or more meals per week

What do you think is the chance that you will someday have health problems because of qhemical residues in Michigan's sport fish? (Circle one number)

1. No chance 2. 1 in a million 3. 1in100,000 4. 1in10,000 s. 1in1,000

6. 7. 8. 9.

1in100 1in10 linS lin2 10. Certain to happen

280

S.

Douglas J. Krieger and John P. Hoehn

How certain are you that your guess about the chance of a health effect is correct? (Circle one number)

1. Very uncertain 2. Somewhat uncertain 3. Somewhat certain 4. Very certain S. I have no idea The public health advisory from the 1992 Fishing Guide is included with this questionnaire. Questions that mention the •advisory• refer to this insert. 6.

Has the advisory helped you to avoid health problems from chemical residues in fish? (Circle one number)

1. Yes 2. No 3. I don't know 7.

As a result of the advice in the advisory, do you . . . (Circle all that apply)

1. I have not read the advisory 2. Eat fish less often 3. Fish at different places 4. Eat smaller rash s. Eat different kinds of rash 6. Prepare rash to eat differently 7. Do nothing differently 8.

Are you concerned about chemical residues or other contaminants in other foods that you eat? (Circle one number)

1. Not at all concerned 2. Somewhat unconcerned 3. Somewhat concerned 4. Very concerned

Anglers' Willingness to Pay for Information About Chemical Residues

281

Michigan's Public Health Advisory There are more than 5,800 public fishing sites in Michigan. These include 2,200 sites on rivers and streams, 3,600 inland lakes, and the Great Lakes. The state has tested 350 of these sites for chemical residues in fish. About 30 new sites are tested each year. The current public health advisory tells you: • that you should not eat too much fish from any inland lake because of widespread mercury contamination, and • it lists 50 sites where fish contain chemical residues above state limits. The advisory does not tell you about: • the 300 tested sites where chemical residues do not exist or are below state limits. The advisory program could be changed. • In addition to the list of sites where chemical residues are above state limits, the advisory could list tested sites where chemical residues do not

exist or are below state limits.

• More than 30 new sites could be tested each year. These changes would increase the amount of information in the advisory but they would also cost more money.

282

Douglas J. Krieger and John P. Hoehn

Your Vote on Advisory Programs The table below shows two advisory programs. The "Current Advisory" is Michigan's current advisory program. "Program A" is a different program that could be put in place. Current Advisory

Program A

Lists tested sites where chemical residues are above state limits?

Yes

Yes

Lists tested sites where chemical residues do not exist or are below state limits?

No

Yes

Number of new sites tested each year.

30

110

$0.00

$.40

Program Options

Cost to you in higher fishing license fees.

Suppose the Michigan Department of Natural Resources (DNR) sent you a ballot to vote "for" or "against" Program A. If a majority of anglers vote "for" Program A, it will replace the Current Advisory. If a majority vote "against" Program A, the Current Advisory will be continued. 9.

Would you vote for Program A if it permanently increased your yearly license cost by $.40, or vote against it and keep the Current Advisory at no additional license cost?

1. Vote for Program A 2. Vote against Program A and keep Current Advisory 3. Don't know or no opinion

Anglers' Willingness to Pay for Information About Chemical Residues 10.

Do you agree or disagree with the following state.. ments?

283

Circle best response

It is OK to increase fish-

Strongly Agree

Somewhat Agree

Somewhat Disagree

Strongly Disagree

No Opinion

The health risks from chemical residues in fish are well understood by scientists.

Strongly Agree

Somewhat Agree

Somewhat Disagree

Strongly Disagree

No Opinion

If chemical residues in

Strongly Agree

Somewhat Agree

Somewhat Disagree

Strongly Disagree

No Opinion

The advisory understates the health risks from chemical residues in Michigan's fish.

Strongly Agree

Somewhat Agree

Somewhat Disagree

Strongly Disagree

No Opinion

ing license fees to pay for better public health advisories.

fish made someone sick, the illness would probably be fatal.

11. In addition to the list of unsafe sites, suppose the advisory listed all tested sites where chemical residues in fish did not exist or were below state limits. If your favorite site had not been tested would you ... ? (Circle all that apply)

1. Continue to fish at your favorite sites 2. Stop eating fish from your favorite sites 3. Fish only at sites where chemical residues are below state limits 4. When choosing a new site, be more likely to go to a site where chemical residues were below state limits 12. What would you do if next year's advisory said you should not eat any of your favorite species of fish from your favorite site? (Circle all that apply)

1. I would still fish at the site 2. I would eat fewer fish from the site

Douglas J. Krieger and John P. Hoehn

284

3. I would not eat fish from the site 4. I would fish at a different site S. I would stop fishing 13. Which of the following groups do you trust to provide the best information about contaminants in Michigan's sport fish? (Circle one number)

1. Federal government 2. State government 3. A well known commner's group 4. An environmental group S. A university laboratory 6. Other ~

14. What is the highest grade of school you have finished? number)

(Circle one

1. Grade school only

2. Did not finish high school

3. High school or GED 4. Vocational or technical school S. Some college 6. College graduate (BS or BA) 7. Some graduate or professional school 8. Graduate degree (PhD, MD, MA, MBA) 15. What choice below best describes your household's expected before-tax income from all sources for 1993? (Circle one number)

1. 2. 3. 4.

$0 to $9,999 $10,000 to $19,999 $20,000 to $29,999 $.10,000 to $.19,999 s. $40,000 to $49,999 6. $50,000 to $59,999 7. $60,000 to $69,999

8. 9. 10. 11. 12. 13. 14.

$70,000 to $79,999 $80,000 to $89,999 $90,000 to $99,999 $100,000 to $109,999 $110,000 to $149,999 $150,000 to $199,999 $200,000 and above

Anglers' Willingness to Pay for Information A.bout Chemical Residues

285

Ifyou have any comments about this questionnaire please write them on this

page.

When you are .finished with the questionnaire please fold it in half, place it in the enclosed business reply envelope, and return to: Douglas Krieger Project Director Department of Agricultural Economics Michigan State University East lAnsing, Ml 48824-1039 Thank you very much for your help.

Chlorine bleach used in the paper industry contributes to dioxins in Michigan waters. This questionnaire is printed on recycled paper made from 100% post consumer stock and processed without chlorine bleach.

14 Using Contingent Valuation Methods to Value the Health Risks from Pesticide Residues When Risks Are Ambiguous Eileen 0. van Ravenswaay and Jennifer Wohl

The debate about the optimal design of policies regarding pesticide residues in food has recently intensified. The controversy centers on whether policy makers should base policy design on consumers' concerns about pesticide residues, or on what is currently known by scientists about the risks from pesticide residues. These two approaches yield very different policy designs since consumers make decisions without full information about the health risks from pesticide residues. Furthermore, their interpretation of the risks they face often diverges significantly from that of scientists. Nonetheless, it is important to consider how consumers weigh the costs and benefits of policy changes. One approach to advising policy makers is to understand how much consumers value changes in risks from pesticide residues given their current risk perceptions, and then apply that information to the evaluation of policies based on scientific estimates of the risks. If we know, for example, how much consumers value risk changes from various baseline risks (e.g., the value of changing the risk of a health problem from pesticide residues from 1 in 100,000 to 1 in a million), we can extrapolate from that information how much they would value the changes in risks that scientists assert would be forthcoming with reduced pesticide residues in food. This chapter describes an approach to measuring consumers' risk perceptions and explains the design of a constructed market for valuing risk reductions relevant to food-safety policy on pesticides. It also identifies the possible importance of consumer uncertainty (called "ambiguity") about the health risks from pesticide residues in determining the benefits of reduced pesticide residues. The value of two pesticide residue policies that reduce risk and/or ambiguity is 287

288

Eileen 0. van Ravenswaay and Jennifer Wohl

sought. Policy I keeps the current federal standard for pesticide residues on food, but permits product claims to be made that the federal standard is met. Policy II toughens the federal residue standard and allows product claims that the new standard is met. We use a constructed market approach since markets for reduced pesticide residues do not currently exist. 1 The constructed market was presented to consumers using a contingent valuation survey instead of an experimental setting in which participants are offered actual products. This approach permits the use of a large and representative sample at lower cost than the equivalent experimental setting. There are several approaches available for measuring how much consumers value changes in risks. One approach is to measure the savings of health care costs and lost wages that would result from risk reductions. However, this measure underestimates the true value of changes in risk since it values only the direct costs of changes in risks. It does not include the inherent value of a longer or more healthy life, for example. Another measure is the amount consumers are willing to give up to achieve risk reductions in labor markets or in markets for consumer products known to reduce risks (e.g., the market for smoke detectors). When using this approach, however, we do not know the exact risk reductions people think they are buying. This is particularly problematic if there is not a consensus or readily available information about risks. Furthermore, the risk reductions acquired in markets are often not comparable to risk reductions provided by a policy change. Reductions in cancer risks, for example, are not comparable to reductions in fatal accidents in the work place or to reductions of risks from fires. Contingent valuation survey methods (CV) can be used to overcome these problems since they offer methods for measuring people's risk perceptions. We can then determine the risks people are actually valuing in either real or simulated markets. CV methods also allow us to measure consumers' willingness to pay for the type of risk reductions that are relevant to policy change. The next section specifies the variables we were interested in measuring in this study. The third section describes the CV methods used to measure the variables (the text of the full survey appears in Appendix 14.A). Section four presents some results obtained from a random sample of Michigan households. The concluding section uses Carson's (1991) validity criteria to evaluate the major strengths and weaknesses of the research design.

Theoretical Framework To reveal how much consumers are willing to pay for reduced risks from pesticide residues in food when risks are ambiguous, we need to model how consumer demand for food varies with the ambiguous risk attribute. To do this,

Using Contingent Valuation Methods to Value Health Risks

289

we incorporated ambiguity into van Ravenswaay and Hoehn's (1991) model of consumer willingness to pay for reduced risks from pesticide residues. In their model, a consumer is hypothesized to maximiz.e utility from budgetconstrained expenditures on a bundle of products offering certain amounts of attributes such as food calories, nutrients, cosmetic quality, and pesticide residues (Lancaster 1971). The consumer combines these product attributes into services such as hunger satisfaction, health maintenance, and health protection via a household production function. The consumer's choice problem is to select product attributes so as to maximiz.e utility obtained from services. The attribute of interest in their study was pesticide residues, and, ultimately, the health risks consumers associate with different residue levels. Van Ravenswaay and Hoehn (1991) develop a model of the demand for a single product, xY, that has the vector of characteristics a1 = (a11 ... au)'. The product xY is offered at price PY. They show that if the true demand function for that product is linear or semi-logarithmic, willingness to pay for a change in the amount of one of its attributes from aY1 to al 1 is: (1)

where pf is the price of good xY when attribute af 1 is present and PY is the price of xY when attribute ad 1 is present. The price pf is such that the quantity demanded remains at x1 after the attribute change. Van Ravenswaay and Hoehn (1991) estimate the shift in the demand curve due to changes in the levels of the pesticide residue. They use this shift to estimate willingness to pay for a change in pesticide residues. Weak complementarity allows us to use the area between the demand curves as a measure of households' willingness to pay for residue reductions if we can assume that the benefits of residue reductions accrue only to persons who purchase reduced residue products. Weak complementarity is said to occur when •the quantity demanded of a private good x1 is 7.ero, the marginal utility or marginal demand price of Q [the environmental attribute] is zero (Freeman 1979: 72). • We assume that consumers benefit from pesticide residue reductions only if they buy the reduced residue product (i.e., they do not experience any benefits from others' purchases). To develop an estimate of the implicit value of risk reduction, it is necessary to know the health risks consumers perceive to be associated with different pesticide residue levels. The van Ravenswaay and Hoehn (1991) model incorporates perceived risk from total consumption of residue (r, where 0 :S r 1 :S 1) into the model by assuming that risk is a product of the total amount of residue consumed, a11 x1, and c1, the factor of proportionality that translates dose into risk (i.e., a linear dose-response function with no threshold). A reduction in per unit residues from aY 1 to al 1 results in a proportional change in risk from:

290

Eileen 0. van Ravenswaay and Jennifer Wohl

(2)

to (3)

Dividing WTP by the perceived change m risk reduction gives average willingness to pay for risk reduction: (4)

wtp = WTPl(rj - lj).

In the survey described below, we observe both r? and r{ (respondents' perceptions of the probability of an adverse health outcome associated with different levels of residues on food), as well as x? and x} (the quantities of each type of food purchased with different levels of residues). We systematically vary P? and p} (the prices of each type of food). All other food attributes ("a" V a ¢ a 11 ) are held constant.

Adding Risk Ambiguity to the Model Van Ravenswaay and Hoehn (1991) applied the model described in the previous section to a constructed market in which apples were labeled for different levels of pesticide residues (the questionnaire is described in the next section). They concluded that although the risk perceptions associated with each of these levels were statistically significant in determining purchases of labeled apples, risk perceptions could not account for the full impact of the labels on apple demand. Van Ravenswaay and Hoehn suggested that the value of the labels might stem from the reduction in the uncertainty consumers had about the risks they face from pesticide residues in food. Consumers might be uncertain about the risks from pesticide residues for several reasons. They do not have perfect information about the levels of residues present in unlabeled food (variable "a" in the model above), nor do they have perfect information about the health effects of residues in either labeled or unlabeled foods (variable "c"). In fact, consumers regularly receive conflicting information from the media about the adequacy of government enforcement of residue standards and scientists' understanding of the health effects of pesticide residues. If consumers do not know with certainty the probability of an adverse health outcome associated with pesticide residues, a point estimate of the probability may not capture all the information relevant to decision makers when they make choices under uncertainty. The term "ambiguity" refers to uncertainty about the probability of an outcome. Ambiguous probabilities are random variables; their probability distributions are called "second-order probability distributions

Using Contingent Valuation Methods to Value Heallh Risks

291

(SOPs)." If consumers perceive the level of risk to be uncertain, some measure of the spread of the SOP will be important in explaining consumer reactions to changes in pesticide residue policy. The main conclusion of the literature on ambiguity is that when a decision maker is presented with two action choices-one where the possible set of outcomes and their probability distribution is known with certainty (nonambiguous, or pure risk), and the other where the set of possible outcomes is known, but the probability distribution is not known with certainty (ambiguous)-most people prefer the nonambiguous situation (Ellsberg 1961, Girdenfors and Sablin 1982, 1983, Einhorn and Hogarth 1985, Segal 1987).2 For example, if in one situation it is known with certainty that the probability of an adverse health outcome from pesticide residues is 1 in a million, but in another situation the probability is known to be somewhere between 1 in a billion and 1 in a thousand with a mean of 1 in a million, then consumers would prefer the situation with the known risk, even though the choices offer the same mean risk. If the ambiguity hypothesis is correct, consumers who perceive risks as ambiguous should be willing to pay more to avoid risks than consumers who perceive the same level of mean risk, but do not view the risk as ambiguous. Ambiguity about risk may stem from either uncertainty about the level of residues present (a} 1) or uncertainty about their health effects (c 1). Pesticide residue policies may reduce ambiguity by addressing one or both of these sources. For example, a policy may reduce ambiguity about the level of pesticide residues by requiring products to be tested and labeled for residue levels. A different policy may eliminate ambiguity about standards by funding more research in the area of risk assessments or by banning the use of chemicals for which the health risks are not well known. These types of policies reduce ambiguity about risks without necessarily changing the mean risk consumers perceive they face.

Design of the Contingent Valuation Survey The questionnaire used in this study (see Appendix 14.A) built upon the questionnaire developed by van Ravenswaay and Hoehn (1991) to estimate consumer willingness to pay for reduced risks from pesticide residues in apples. Van Ravenswaay and Hoehn sought to simulate an actual shopping situation as closely as possible, while having results based on a large, representative sample of households. Since the majority of households purchase apples and associate apples with pesticide residues, apples offer a convenient commodity to study. Since we used many of the same questions developed by van Ravenswaay and Hoehn, we will briefly describe the development of their questionnaire. First, van Ravenswaay and Hoehn conducted focus groups with consumers to understand how they shop for apples, the varieties of apples people buy, where

292

Eileen 0. van Ravenswaay and Jennifer Wohl

they purchased them, whether they were purchased packaged or loose, and whether people commonly paid for apples by the pound or by the apple. Focus groups were also used to discern consumers' thoughts about pest damage, pesticide residues on apples, and their perceptions of the risks associated with pesticide residues. Several labels regarding pesticide residues in apples were also tested to determine consumer comprehension and to elicit respondents' perceptions of the perceived changes in risks associated with different labels. The questions constructed from the focus groups were used to develop pretest questionnaires. These were pretested in a series of in-person interviews in which respondents were asked to describe how they interpreted each of the questions and why they chose particular responses. When no further changes were found to be needed in the pretest interviews, a pretest questionnaire was mailed to a random sample of 200 Michigan households. The data were analy7.ed to determine whether any unexpected results had been obtained. When none were found, the questionnaire was mailed to 2,200 randomly selected households nationwide. The final questionnaire asked respondents about their current purchases of apples, their purchase intentions for regular apples at specified prices, and their purchase intentions for apples that were described as tested and certified to have "no pesticide residues," "no detectable pesticide residues," and "no pesticide residues above federal limits." Respondents were given a range of different prices and asked how many apples with and without the labels they would buy if they were planning to buy apples on a typical shopping occasion in the fall. The season was specified so that all respondents would be considering similar supply conditions. Questions were also asked to elicit respondents' perceptions of the health risks associated with pesticide residues and the perceived changes in those risks with different residue labels. Respondents were asked the likelihood that a member of their household would experience any kind of health problem someday because of pesticide residues in foods. They could choose from the response categories: 1 in a million, 1 in 100,000, 1 in 10,000, 1 in 1,000, 1 in 100, 1 in 10, 1 in 5, 1 in 2, and certain to happen. Respondents were then asked to estimate the reduction in risks that would result when all foods were tested and certified to have the different levels of pesticide residues indicated above. In the present research several modifications were made to the van Ravenswaay and Hoehn (1991) survey. First, our survey was conducted by telephone rather than by mail. The van Ravenswaay and Hoehn study required respondents to react to photographs of apples with pest damage; a mail survey was thus necessary. In the present research, however, no variations in visual aspects of apples were considered. Telephone surveys offer several advantages over mail surveys. They allow the use of Random Digit Dialing (ROD) which ensures representative samples.

Using Contingent Valuation Methods lo Value Health Risks

293

They also reduce nonresponse bias since interviewers are able to clarify questions and prompt respondents for further information. Complicated skip patterns can also be easily incorporated into the questionnaire design. As in the van Ravenswaay and Hoehn (1991) questionnaire, the purchase intention questions in our survey were developed to reveal the quantity of apples respondents would likely buy at different prices during a typical grocery shopping occasion in the fall. However, in this survey respondents were given a choice between labeled and unlabeled apples, whereas the van Ravenswaay and Hoehn (1991) questionnaire did not offer this choice. In the previous survey, all apples were either all labeled or all unlabeled. Our modification allowed substitution possibilities to be explicitly measured. In the van Ravenswaay and Hoehn (1991) questionnaire, the consumer could substitute another fruit for apples, but the price and type of that fruit was unknown. In our survey, we assumed that the closest substitute to an unlabeled apple was an apple labeled for the level of pesticide residues. We offered respondents both products and specified their prices (prices varied across households). As in the van Ravenswaay and Hoehn (1991) survey, respondents were told to assume that all fruits other than apples were not labeled for pesticide residue levels. The labels offered to respondents were also changed. We omitted the "no detectable residues" label since van Ravenswaay and Hoehn (1991) found it was statistically indistinguishable from the label "meets federal standards for pesticide residues." The label "no pesticide residues" was changed to "produced without pesticides" since it is impossible to guarantee "no pesticide residues" when pesticides may be airborne, waterborne, or soilborne at the farm where the apples were grown, the warehouse where they were stored, or the store where they are sold. In order to make the questions shorter and more tractable, half the sample was given the label "produced without pesticides"; the other half received the label "no pesticide residues above federal limits." This design, plus the use of telephone rather than mail questionnaires, allowed for more randomiution of the price combinations given to respondents. Each household received one of forty price combinations of certified and regular apples ranging from $0.49 to $1.19 for the unlabeled apples and $0.49 to $1.59 for the certified apples. Different subsamples of respondents were given different sets of prices and label combinations. Respondents were told that all apples would look the same as those they usually buy. They were then asked if they would buy all of one type of apple (labeled or unlabeled), some of both, or none at all, and the quantities of those apples they would likely buy at the stated prices. Several improvements were also made to the method used to elicit risk perceptions. The van Ravenswaay and Hoehn (1991) study found that there was considerable variation across households in the actions households undertook to avoid pesticide residues. One-third of their sample reported that they washed fresh produce with soap and water, bought organic produce, or grew their own

294

Eileen 0. van Ravenswaay and Jennifer Wohl

produce. A method needed to be developed to ensure that each household was comparing labeled apples to conventional apples rather than to apples that might have reduced pesticide residues because of its own actions. For example, a person who already bought organic would be likely to compare the labeled apples to organic apples unless we specified otherwise. This was accomplished by asking respondents to consider a person from a household like theirs that did nothing at all to avoid or reduce pesticide residues in food. Another improvement in the risk perception questions was the increased specificity in the meaning of quantitative risk. In the van Ravenswaay and Hoehn (1991) survey, respondents were asked what the chances were that someone from their household would have a health problem someday because of pesticide residues in food. To ensure that respondents to the present survey were considering the risk to a person in the whole population, and not just the risk to themselves, we asked respondents in this survey to imagine that there were a million people from households like theirs that did nothing to reduce or avoid pesticide residues in food. We then asked them what they thought the chances are that a person from one of these households would have a health problem someday because of pesticide residues in food. The response categories used in the van Ravenswaay and Hoehn (1991) survey were modified to specify numbers of people out of one million who would be expected to have a health problem someday because of pesticide residues in food. A third improvement to the risk perception questions was that respondents were asked for a qualitative statement of the risks from pesticide residues before being asked for a quantitative estimate. Respondents were asked whether they would say there is "no chance," "it is very unlikely," "somewhat unlikely," "somewhat likely," "very likely," or "certain to happen" that someone from a household like theirs would have a health problem someday because of pesticide residues in food. The interviewer used this qualitative answer to prompt the respondent for a quantitative estimate of risks. This approach enabled us to examine the correlation between a qualitative, but more easily understood estimate, and a quantitative, but more precise estimate. In fact, we found the two measures to be highly correlated. Respondents were also asked an open-ended question about the types of health problems they associated with pesticide residues. The van Ravenswaay and Hoehn (1991) survey asked close-ended questions which may have prompted respondents to choose health effects they would not otherwise have thought of. Finally, a method was developed to measure respondents' ambiguity about their qualitative and quantitative risk estimates. Since we generally do not know the exact specification of the respondents' second-order probability (SOP) distribution around the mean probability, measuring the variance, or spread, of the SOP is not generally feasible. Several alternative measures of ambiguity have been developed. For example, Becker and Brownson (1964) used the "range" of the SOP as a measure of ambiguity. They assume the SOP is a

Using Contingent Valuation Methods to Value Health Risks

295

uniform distribution, the distance between the lowest and highest points on this distribution is the "range." The longer the range, the more ambiguity. Ambiguity can also be measured by the "degree of confidence" one has in the point estimate of the probability of an outcome (Ellsberg 1961). This approach assumes there is a confidence interval around the mean probability of an outcome. Confidence, then, is an indicator of the "spread" of the probability distribution around the probability. The higher the confidence, the smaller the confidence interval. Confidence can be measured in terms of a probability ("I am 95 percent sure that the probability of outcome x is 0. 75") or in qualitative terms ("I am 'very sure' that the probability of outcome x is 0. 75"). Ambiguity in this research is measured by asking survey respondents how sure they are about their estimate of the probability of an adverse health outcome. A Likert scale (1 = very sure, 2 = somewhat sure, 3 = somewhat unsure, 4 = very unsure) was used to measure confidence in risk estimates. This measure is used as a proxy for the spread, or variance, of the second-order probability distribution. Questions were also developed to determine the sources of ambiguity. We hypothesized that ambiguity stemmed from two sources: (1) uncertainty about current residue levels in food and (2) uncertainty about the relationship between residues and their associated health effects. Ambiguity about residue levels should be high if respondents feel that current federal standards were not being adequately enforced. Ambiguity about the relationship between residue levels and health effects should be high if respondents feel that scientists did not understand the true health effects of pesticides or if respondents believed that scientists were not honest with the public about the true health effects. Questions about these hypotheses were incorporated into the survey. All questions that were added to or revised in the van Ravenswaay and Hoehn (1991) questionnaire were pretested using personal interviews. The entire questionnaire was pretested in telephone interviews conducted by trained interviewers. This allowed us to develop detailed instructions on how interviewers should handle unusual or difficult situations, thus improving the reliability of data collection.

Some Survey Results The target population for this study was Michigan households that purchase food. The sample consisted of the telephone numbers of 1,730 randomly selected households in Michigan purchased from Survey Sampling Incorporated (SSI). The sample purchased from SSI was drawn using Random Digit Dialing (RDD). The surveys were conducted with adults over the age of 18 who did most of the food shopping for their household. They were conducted by telephone during June and July 1992. Sixty-seven percent (1,003 households)

296

Eileen 0. van Ravenswaay and Jennifer Wohl

of the eligible households contacted completed the survey, 33 percent (484 households) of those contacted refused to be surveyed. The telephone interviews were conducted by the Institute for Public Policy and Social Research (IPPSR) at Michigan State University. The perceived risk level from pesticide residues in food based on survey responses is presented in Table 14.1. The average perceived risk is between 1 in 10,000 and 1 in 1,000. This is the same mean perceived risk found by van Ravenswaay and Hoehn (1991), although the wording of the questions differs between the two surveys. However, many fewer households in our study perceived no chance or a chance of 1 in a million. This result was expected because the earlier survey did not control for the possibility that consumers may have been taking different types and amounts of actions to reduce or avoid pesticide residues in food. The results shown in Table 14.1 indicate that people have very divergent subjective estimates of the risk from pesticide residues in food. Consequently, surveys that assume that risk perceptions among all consumers are the same would lead to incorrect estimates of willingness to pay for risk reduction. In addition to the questions about the mean perceived health risk, Michigan respondents were also asked to indicate their level of ambiguity about risk. Despite the difficult nature of quantitatively assessing risks, Table 14.2 shows that 23 percent of respondents were very sure about their risk estimates, 45 percent were somewhat sure, 20 percent were somewhat unsure, and only 5 percent were very unsure. The results of this question demonstrate that people have different levels of confidence in their risk estimates (i.e., the spread of the SOP varies by individual). The contingent valuation survey method used here allows us to gauge this sureness and then use it as a product attribute that affects willingness to pay for residue reduction. The risk levels presented in Table 14.1 are the perceived risks associated with the consumption of conventionally produced foods. If foods are labeled for different levels of pesticide residues, many people will perceive that those risks change. Table 14.3 shows the risk reduction, in percentage terms, people felt they would be getting if foods were labeled as indicated. 3 As expected, respondents felt they got more risk reduction when foods are "produced without pesticides" than when foods "meet federal standards." However, the difference in risk reduction between the two labels is not large, indicating that consumers may perceive current federal standards as substantially reducing risks. Figure 14.1 shows the types of heath effects people associate with pesticide residues in food. The question asked was, •Suppose someone from a household like yours had a health problem someday that resulted from the current levels of pesticide residues in food. In your opinion, what would the health problem most likely be?" (open-ended). The results indicate that while there are a variety of health problems associated with pesticide residues in food, more than 50 percent of respondents believe cancer is the most likely illness.

297 TABLE 14.1 Perceived Chance of a Health Problem Response (Prompted) Question:

"Suppose there were a million people from households like yours who did nothing to reduce or avoid pesticide residues in food. What do you think the chances are that a person from one of these households would have a health problem someday because of pesticide residues in food?"

No chance 1 in a million 1in100,000 1in10,000 1in1,000 1in100 1in10 Certain to happen Don't know/no opinion/refused to answer Note: N

Percent Respondents

= 1,003.

2.4 4.1 14.1 23.0 22.8 10.8 8.4 8.2

6.3

Figures may not add to 100 percent due to rounding.

TABLE 14.2 Sureness About Health Risk Percent Respondents

Response (Prompted) Question:

"How sure are you that the chance of a health problem is _ _?" (Blank filled in with respondent's estimate.)

Very sure Somewhat sure Somewhat unsure Very unsure Don't know/no opinion Note: N

= 1,003.

23.4 44.7 19.9

S.3

6.7

Figures may not add to 100 percent due to rounding.

Table 14.4 shows respondents' attitudes towards both the government and the scientific community. Respondents were asked to what extent they agreed with the statements given. This table attempts to capture the source of ambiguity in

298

Eileen 0. van Ravenswaay and Jennifer Wohl

TABLE 14.3 Perceived Reduction in the Risks from Pesticide Residues When Foods Meet Federal Standards for Pesticide Residues or Are Produced Without Pesticides Label: "Meets Federal Standards" N = 434

=

Percent Respondents

Percent Reduction Question:

Label: "Produced Without Pesticides" 418 N

"Now suppose that all foods met the federal standard for pesticide residues (split sample received: Now suppose that pesticides were not used in producing foods). What percent do you think that would reduce the chances of a health problem happening someday to people who currently do nothing to reduce or avoid pesticide residues?" (open-ended)

0 0 to 20 20 to 40 40 to 60 60 to 80 80 to 99 100 Refused/no answer

0.9 10.6 14.3 28.1 21.9 9.4 4.1 10.6

1.7 13.4 7.2 29.9 19.4 13.2 15.3 0.0

Note: N = 852. Respondents who answered "there was no chance of a health problem," or who answered "don't know/no opinion/refused" to the question in Table 14.1 were not asked this question. Figures may not add to 100 percent due to rounding.

people's risk estimates. It appears that ambiguity does not stem overwhelmingly from one source, although it does seem that ambiguity derives less from scientific uncertainty about risks than from the trustworthiness of scientists and government regulators. Finally, Table 14.S shows respondents' purchase intentions for labeled and unlabeled apples at different prices. The table gives the percentage of respondents indicating they would buy that type of apple if both labeled and unlabeled apples were available. The data do not indicate the quantities purchased. There was not a large difference in purchase intentions between the subsample that evaluated the "meets federal standards" label and the subsample that evaluated the "produced without pesticides" label. This is evidence that there may be substantial value in reducing ambiguity about whether foods meet current federal standards for pesticide residues.

Using Contingent Valuation Methods to Value Health Risks

299

Gastro·lntestlnal problems Allergies Respiratory problems Nothing Other Don't know/refused 0.0% 20.0% 40.0% 80.0% 80.0% 100.0%

FIGURE 14.1 Health Effects Perceived to Be Associated with Pesticide Residues in Food

It is apparent from Table 14.5, that the difference in price between labeled and unlabeled apples is an important factor in respondents' decisions to purchase labeled apples. Respondents were interested in labeled apples only if the price was low enough relative to unlabeled apples. Specification of substitution possibilities is clearly important. A Critique of Survey Methods The results from contingent valuation surveys are often criticized because of the large divergence between what people say they will do, and what they actually do. Carson (1991) has developed a series of criteria for designing constructed markets to increase the reliability of the results. His criteria focus on the theoretical accuracy and policy relevance of the scenario offered in the survey, as well as on the extent to which the scenario is understandable, plausible, and meaningful to respondents. We use these criteria to evaluate the strengths and weaknesses of the contingent valuation survey design used in this study.

22.0

The health risks associated with current levels of pesticide residues in food are well known and understood by the scientific community. 39.6

39.7

38.0

37.9

Somewhat Agree

Figures may not add to 100 percent due to rounding.

12.S

The scientific community can be trusted to be truthful about what they know about health risks from pesticide residues.

= 1,003.

13.6

I trust that once the federal standards are set, all the food I buy will meet those standards.

Note: N

14.2

I trust the federal government to set the same standards that I would set in limiting the amount of pesticide residues allowed in food.

Statement

Strongly Agree

TABLE 14.4 Attitudes Toward the Government and the Scientific Community

23.2

25.4

27.8

25.7

Percent Respondents

Somewhat Disagree

12.8

21.1

20.0

22.0

Strongly Disagree

2.4

1.3

0.6

0.2

Don't Know/ No Opinion/ Refused

s

1.5

0.5

Don't know/refused

1.6

0.5

Note: N = 1,003. Figures may not add to 100 percent due to rounding. Numbers in table are percentages selecting type of apple at given prices, not how much would be purchased.

4.4

3.2

None at all

8.4

14.1

9.1

15.2

4.8

50.3

68.3

48.5

16.1

30.4

8.4

71.1

Certified apples

26.7

Price Difference: $0.20 or less

Price Difference: $0.30 or more

Label: Produced Without Pesticides

Price Difference: $0.30 or more

Some of both

16.8

Price Difference: $0.20 or less

Regular apples

Type of Apple

Label: No Pesticide Residues Above Federal Standards

Percent Respondents Indicating Willingness to Buy Type of Apple

TABLE 14.5 Willingness to Buy Labeled and Un1abeled Apples

~

w

302

Eileen 0. van Ravenswaay and Jennifer Wohl

To guarantee theoretical accuracy, the researcher must ensure that features of the created scenario are compatible with economic theory. For example, property rights must be clearly specified, the respondent's budget constraint must be binding, substitution possibilities must be clearly indicated, and the payment mechanism for the good in question should result in accurate statements of value. The theoretical framework outlined above warrants the use of a private-goods market. We use weak complementarity to justify the use of demand analysis as an approach to measuring willingness to pay for risk reduction. Weak complementarity requires that the benefit of risk reduction accrue only if the good is consumed (people do not benefit from reduced risk to others). The private market for apples is thus an appropriate setting for measuring willingness to pay for risk reduction. The budget constraint of the respondent is likely to be binding since respondents were given the prices of goods in a market setting. It is doubtful that respondents would strategically exaggerate the number of apples they would buy at these prices. It is also unlikely that people would have difficulties making decisions about products. Most people buy apples; it is easy for them to accurately predict their purchasing behavior. It is important in survey design to incorporate substitutes for the good in question. Otherwise, people are responding to questions out of the context in which they would actually be required to make payments. Our constructed market consisted of both labeled and unlabeled apples, with the prices for both indicated. Since labeled apples are considered the closest substitute for regular apples, this approach guarantees that people are considering a realistic buying scenario. Respondents are given the choice to buy some of either type of apple, some of both, or none at all. Respondents do not then have incentives to over- or understate their purchase intentions. Since it is doubtful that market goods would succeed if demand for the product does not materialize, respondents are also unlikely to think that their responses to a survey will make a difference in the ultimate provision of the good. Respondents probably do not misrepresent their purchase intentions in the hopes of influencing a policy outcome. Since our investigation is based on a goods-characteristics model, we designed the questionnaire such that the results indicate the value of only the food safety attribute in question. To ensure this, we structured the scenario such that respondents would assume that other product attributes did not vary. This was accomplished by asking respondents to consider the variety of apples they normally bought and to assume that the quality of all apples was the same as they normally observed. It is important that the results of the survey be relevant to policy makers. Our survey considered not only how consumers value changes in the risks from pesticide residues but also the effect of food labeling on consumer choice.

Using Contingent Valuation Methods to Value Health Risks

303

These are both issues of interest to decision makers in the area of food safety policy. The methods we developed to elicit perceptions about risks from pesticide residues and ambiguity about those risks can be used in any application that requires an understanding of consumers' perceptions of risk. One could question the policy relevance of a scenario that considers a market where apples are the only product tested and certified for residue levels. If policy makers were to require apples to be tested and certified, they would probably also require that other produce be tested and certified. However, we believe that respondents are unable to accurately predict and describe their behavior in the more general scenario where all foods are described as being tested for pesticide residues. This scenario would require changing the prices of all substitutes for apples; creating an easily understandable scenario then becomes infeasible. As the name implies, the results of contingent valuation studies are contingent upon the scenario presented to respondents. The highly specific scenario given to respondents in this survey makes it difficult to generalize the results. However, the specific scenario offers the advantage of yielding more valid results from the point of view of economic theory and respondent comprehension. Respondents will be more able and motivated to accurately predict their behavior if the scenario presented to them is understandable, plausible, and meaningful. Although contingent valuation surveys commonly ask respondents to directly state their willingness to pay for specific goods or services, this research did not use this approach. Instead, we ask respondents how many apples they would likely buy under different scenarios. We then use the information to estimate willingness to pay for risk reduction. Since consumers are more likely to have to make market choices than to be asked to pay directly for reduced risks, the market scenario is more understandable, plausible, and meaningful to respondents. We studied respondents' buying patterns of apples because most households purchase apples, apples have been associated with pesticide residues in the media, and using apples allows people to predict their behavior under very familiar and likely circumstances. One of the goals of this research was to develop reliable methods of soliciting risk perceptions. Extensive pretesting of the survey instrument was therefore conducted to ensure that questions were thoroughly understood. Crosstabulations between qualitative and quantitative responses to risk questions suggest that respondents did, in fact, understand the nature of the quantitative risk assessments they were asked to make. Careful consideration was also given to the order of questions. The survey was designed to first get respondents thinking about their current apple purchasing behavior before asking them to make hypothetical purchasing decisions. Many studies involving decisions under risk present respondents with scientific or objective risks and then assume that respondents use those estimates

304

Eileen 0. van Ravenswaay and Jennifer Wohl

in their decision calculus. In some cases, this approach may be useful, but the presentation of the objective risks should be coupled with questions that examine how such information alters risk perceptions. We cannot assume that respondents accept objective estimates of risk without consideration of their own experience and other information. In this research, we ask respondents to make their own assessment of the risks involved in pesticide residue consumption. This is a more realistic scenario since consumers are generally forced to make their own risk assessments before making purchase decisions. Furthermore, scientific or objective risk assessments are not generally available to consumers. To ensure that respondents understood the questions, risk perception questions were worded to allow respondents to think in terms of numbers of people affected rather than just in probabilities. Our survey also allows examination of ambiguity about risk. Questions were asked about how "sure" respondents felt about their risk estimates. Because there is a lot of uncertainty about the risks from pesticide residues, ambiguity may be an important factor in explaining consumers' willingness to pay for risk reduction. We also believe that allowing respondents to express their uncertainty about the health risks from pesticide residues improves the validity of the risk perception measures. When respondents are given the opportunity to express their reservations, they feel less pressure to be right, and are then more likely to give their best estimate of the risk instead of giving worst- or best-case estimates. We are interested in respondents' judgements of the most likely estimate of risk. We believe that additional research is needed to improve the validity of the methods of eliciting both risk and ambiguity perceptions. The methods described here should be tested against other possible approaches. We also need to further explore the possibility that food safety and nutrition involve risks that are ambiguous to consumers. The better we understand how consumers manage risks, the better we may be able to bridge the gap between scientists' understanding of the risks from pesticide residues and respondents' perceptions of those risks. We could then design policies that protect consumer health and ease consumers' anxiety about the risks they face.

Notes 1. Focus groups with consumers conducted by van Ravenswaay and Hoehn (1991) found that organic labels were not interpreted in a consistent fashion by consumers. Furthermore, organic products may be treated with organic pesticides and are not necessarily pesticide-free. Consumers also buy organic products for reasons other than pesticides. For example, some consumers believe organic foods have higher nutrient content and are more tasty. 2. See Camerer and Weber (1992) for a review of models of decision making that incorporate ambiguity.

Using Contingent Valuation Methods to Value Health Risks

305

3. The perceived risk levels associated with the labels can be calculated by applying the perceived risk reduction with the labels to the risk levels in Table 14.1.

References Becker, Selwyn W. and Fred 0. Brownson. 1964. What Price Ambiguity? Or the Role of Ambiguity in Decision-Making. Journal of Political "Economy 72(1):62-73. Camerer, Colin and Martin Weber. 1992. Recent Developments in Modeling Preferences: Uncertainty and Ambiguity. Journal of Risk and Uncertainty 5:325-370. Carson, Richard T. 1991. Constructed Markets. In Measuring the Demand for Environmental Quality, ed. J. B. Braden and C. D. Kolstad. New York, NY: North-Holland. Einhorn, Hillel J. and Robin M. Hogarth. 1985. Ambiguity and Uncertainty in Probabilistic Inference. Psychological Review 92(4):433-460. Ellsberg, Daniel. 1961. Risk, Ambiguity, and the Savage Axioms. Quarterly Journal of "Economics 75(4):643-669. Freeman, A. Myrick III. 1979. The Benefits of Environmental Improvement: Theory and Practice. Baltimore, MD: Johns Hopkins University Press. Girdenfors, P. and N. E. Sahlin. 1982. Unreliable Probabilities, Risk Taking, and Decision Making. Synthese 53:361-386. Girdenfors, P. and N. E. Sahlin. 1983. Decision Making with Unreliable Probabilities. British Journal of Mathematical and Statistical Psychology 36:240-241. Lancaster, Kelvin. 1971. Consumer Demand: A New Approach. New York, NY: Columbia University Press. Segal, Uzi. 1987. The Ellsberg Paradox and Risk Aversion: An Anticipated Utility Approach. International "Economic Review 28(1):175-202. van Ravenswaay, Eileen 0. and John P. Hoehn. 1991. Contingent Valuation and Food Safety: The Case of Pesticide Residues in Food. Department of Agricultural Economics Staff Paper No. 91-13, Michigan State University.

Appendix 14.A TELEPHONE SURVEY INSTRUMENT [Please note: all skip patterns and split-sample variations have been removed for better readability of survey. The full survey instrument is available from the authors upon request.] Hello, is this (confirm phone number)? My name is and I am calling from the Center for Survey Research at Michigan State University. We are conducting a study on behalf of the Department of Agricultural Economics at Michigan State University regarding pesticide residues in food.

Eileen 0. van Ravenswaay and Jennifer Wohl

306

According to our sampling design, I need to speak to the person in the household, who is at least 18 years of age, who does the most grocery shopping. Would that be you? Before we begin, let me tell you that any information you give me will be kept strictly confidential. Let me also tell you that this interview is completely voluntary. Should we come to any question that you don't want to answer, just let me know and we'll go on to the next question. Throughout the study, we will be asking for your opinions in terms of the food you buy for your household. Your household includes yourself, your dependents, and persons with whom you share income and household living expenses. We will also be talking about pesticide residues in food. Pesticides are used to control insects, diseases, and other pests that spoil food. To protect consumers' health, the federal government sets standards that limit the amount of pesticide residues that may be in food sold in the U.S. Ql

In terms of pesticide residues, how confident are you that the food your household eats is safe?

Would you say you are completely confident, mostly confident, somewhat confident, or not confident at all? < 1 > COMPLETELY CONFIDENT MOSTLY CONFIDENT SOMEWHAT CONFIDENT NOT CONFIDENT AT ALL DON'T KNOW/NO OPINION < 9> REFUSED/NO ANSWER Q2

Suppose someone from a household like yours did nothing at all to reduce or avoid pesticide residues in food. What do you think the chances would be that someone from that household will have a health problem someday because of pesticide residues in their food? Would you say there is no chance, it is very unlikely, somewhat unlikely, somewhat likely, very likely, or certain to happen? < 1 > NO CHANCE VERY UNLIKELY < 3 > SOMEWHAT UNLIKELY SOMEWHAT LIKELY VERY LIKELY CERTAIN TO HAPPEN DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Using Contingent Valuation Methods to Value Health Risks

3a

307

How sure are you that there is of a health problem because of pesticide residues in food? (blank is filled with respondent's answer from Q2) Would you say you are very sure, somewhat sure, somewhat unsure, or very unsure? < 1> VERY SURE SOMEWHAT SURE < 3 > SOMEWHAT UNSURE VERY UNSURE < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q4

Now I would like to get a better idea about what you mean by _(blank is filled with respondent's answer from Q2). Suppose there were a million people from households like yours who did nothing to reduce or avoid pesticide residues in food. What do you think the chances are that a person from one of these households would have a health problem someday because of pesticide residues in food? Would you say 1 person in a million, 1 in 100,000, 1 in 10,000, 1 in 1,000, 1 in 100, or 1 in 10? < 1> 1 PERSON IN A MILLION 1 IN 100,000 1IN10,000 1 IN 1,000 1IN100 1IN10 DON'T KNOW/NO OPINION REFUSED/NO ANSWER

QS

How sure are you that the chancei; are _ _ _ (blank is filled with respondent's answer from Q4). Would you say you are very sure, somewhat sure, somewhat unsure, or very unsure? < 1 > VERY SURE SOMEWHAT SURE < 3 > SOMEWHAT UNSURE VERY UNSURE < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

308

Eileen 0. van Ravenswaay and Jennifer Wohl

Q6

Is there anything you usually do to reduce or avoid pesticide residues in your food? (open-ended, field coded) < 1 > YES: specify NO DON'T KNOW < 99 > REFUSED

Q7

Suppose someone did the same things you usually do to reduce or avoid pesticide residues in food. What percent do you think that would reduce the chances of a health problem happening some day? < 0-100 > ENTER EXACT PERCENT < 998 > DON'T KNOW/NO OPINION < 999 > REFUSED/NO ANSWER

Q8

Next, I am going to read you two statements, please tell me to what extent you agree or disagree with each of them. I trust the federal government to set the same standards that I would set in limiting the amount of pesticide residues allowed in food. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree? < 1 > STRONGLY AGREE SOMEWHAT AGREE < 3 > SOMEWHAT DISAGREE STRONGLY DISAGREE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q9

I trust that once the federal standards are set, all the food I buy will meet those standards. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree? < 1> STRONGLY AGREE SOMEWHAT AGREE < 3 > SOMEWHAT DISAGREE STRONGLY DISAGREE < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Using Contingent Valuation Methods to Value Health Risks

309

QlO

Now suppose that all foods met the federal standard for pesticide residues. What percent do you think that would reduce the chances of a health problem happening someday to people who currently do nothing to reduce or avoid pesticide residues in food? ENTER EXACT PERCENT DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Qll

Suppose foods were tested and certified to meet federal standards for pesticide residues. Which of the following organiz.ations do you feel would be the most effective in conducting the tests and issuing certificates? Would you say the federal government, the state government, a well known consumer's group, or some other organiz.ation?

< 1 > FEDERAL GOVERNMENT STATE GOVERNMENT

< 3 > A WELL KNOWN CONSUMER'S GROUP

SOME OTHER ORGANIZATION: SPECIFY < 98 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Ql2

Suppose the federal government did the testing and certifying. How effective do you think such a program would be in ensuring that foods had no pesticide residues above federal standards? Would you say very effective, somewhat effective, somewhat ineffective, or totally ineffective?

< 1 > VERY EFFECTIVE

SOMEWHAT EFFECTIVE

< 8>

SOMEWHAT INEFFECTIVE TOTALLY INEFFECTIVE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

[Note: Half the sample received questions 10, 11, and 12 with "foods were produced without pesticides" replacing "foods met federal standards."] Q13

Suppose someone from a household like yours had a health problem someday that resulted from the current levels of pesticide residues in food. In your opinion, what would the health problem most likely be? (open-ended, field coded)

310

Q14a

Eileen 0. van Ravenswaay and Jennifer Wohl

Next I would like to ask a few questions about where you get your information about the health risks of pesticide residues. In the past 6 months have you gotten information about the health risks of pesticide residues from a television program? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q14b

In the past 6 months have you gotten information about the health risks of pesticide residues from your doctor or health specialist? YES NO DON'T KNOW/NO OPINION < 9 > REFUSED/NO ANSWER

Ql 4c

In the past 6 months have you gotten information about the health risks of pesticide residues from an article in a magazine? YES NO < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q14d

In the past 6 months have you gotten information about the health risks of pesticide residues from a newspaper? YES NO < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q14e

In the past 6 months have you gotten information about the health risks of pesticide residues from a health newsletter? YES NO < 8 > DON'T KNOW/NO OPINION < 9> REFUSED/NO ANSWER

Q14f

In the past 6 months have you gotten information about the health risks of pesticide residues from a radio program? YES NO < 8 > DON'T KNOW/NO OPINION

Using Contingent Valuation Methods to Value Health Risks

311

REFUSED/NO ANSWER Q14g

In the past 6 months have you gotten information about the health risks of pesticide residues from family, relatives, or friends? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q14h

In the past 6 months have you gotten information about the health risks of pesticide residues from any other sources? (open-ended, field coded) < 1 > YES: SPECIFY

QlS

Next, I am going to read you several statements. In these statements, the term 'plants and animals' refers to plants and animals produced for food. Please tell me to what extent you agree or disagree with each of them. If plants and animals were not protected in any way from insects, diseases, or other pests, the supply of food available to me would decrease.

Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree? < 1 > STRONGLY AGREE SOMEWHAT AGREE SOMEWHAT DISAGREE STRONGLY DISAGREE DON'T KNOW/NO OPINION REFUSED/NO ANSWER Q16

If plants and animals were not protected in any way from insects, diseases, or other pests, the food available to me would not look as good as it does now.

Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree? < 1 > STRONGLY AGREE SOMEWHAT AGREE < 3 > SOMEWHAT DISAGREE STRONGLY DISAGREE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

312

Q17

Eileen 0. van Ravenswaay and Jennifer Wohl If plants and animals were not protected in any way from insects, diseases, or other pests, the price of food available to me would increase.

Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree?

< 1> STRONGLY AGREE

SOMEWHAT AGREE

< 3 > SOMEWHAT DISAGREE STRONGLY DISAGREE

< 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER Q18

There are many equally effective ways other than using pesticides to protect plants and animals from insects, diseases, or other pests. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree?

< 1> STRONGLY AGREE



Q19

SOMEWHAT AGREE SOMEWHAT DISAGREE STRONGLY DISAGREE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

It is more expensive to use other ways of protecting plants and animals from pests than it is to use pesticides. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree?

< 1>

< 8>

Q20

STRONGLY AGREE SOMEWHAT AGREE SOMEWHAT DISAGREE STRONGLY DISAGREE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

The scientific community can be trusted to be truthful about what they know about health risks from pesticide residues. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree?

Using Contingent Valuation Methods to Value Health Risks

313

< 1 > STRONGLY AGREE

< 3>

< 8>

Q21

SOMEWHAT AGREE SOMEWHAT DISAGREE STRONGLY DISAGREE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

The health risb associated with current levels of pesticide residues in food are well known and understood by the scientific community. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree? < 1 > STRONGLY AGREE SOMEWHAT AGREE < 3 > SOMEWHAT DISAGREE STRONGLY DISAGREE < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q21a

Food labeled as organic means the food is grown without pesticides. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree?) STRONGLY AGREE SOMEWHAT AGREE < 3 > SOMEWHAT DISAGREE STRONGLY DISAGREE < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q21b

All food that is labeled as organic has been certified by a reputable laboratory to have been organically grown. Would you say you strongly agree, somewhat agree, somewhat disagree, or strongly disagree? < 1 > STRONGLY AGREE SOMEWHAT AGREE SOMEWHAT DISAGREE STRONGLY DISAGREE < 8> DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q22

The next few questions are about your food shopping routine.

314

Eileen 0. van Ravenswaay and Jennifer Wohl

How often is the grocery shopping done in your household? (openended, field coded) Q23

In the past year, bas your household bought any fresh apples? YES NO < 8 > DON'T KNOW /NO OPINION < 9 > REFUSED/NO ANSWER

Q23a

When you buy fresh apples, do you usually buy them individually, by the pound, by the bag, by the peck, or by the bushel? (open-ended, field coded) < 1> INDIVIDUAL POUNDS BAGS PECK BUSHEL < 0 > OTHER (SPECIFY) DON'T KNOW/NO OPINION < 99 > REFUSED/NO ANSWER

Q23b

How many individual apples or pounds are usually in a bag? < 1-997> DON'T KNOW/NO OPINION < 999 > REFUSED/NO ANSWER

Q24

About bow often does your household buy fresh apples in the fall? (open-ended, field coded)

Q24a

When you buy fresh apples in the fall, on average, bow many apples do you buy each time?

DON'T KNOW/NO OPINION < 999 > REFUSED/NO ANSWER

[Note: Q24 and Q24a were asked with "winter," "spring," and "summer" replacing fall.] Q28

Now, suppose it is next fall and you are planning to buy some fresh apples. The quality of all fresh apples is what you normally expect. Apples sold loose and prepackaged are all the same price per pound. The prices of all fresh fruits other than apples are what you normally expect.

Using Contingent Valuation Methods to Value Health Risks

315

How many apples of your usual variety would you buy if all fresh (blank filled with one of several prices) per apples were pound? Q29

·Now suppose you could also buy apples of your usual variety that have been tested and certified by the federal government to have no pesticide residues above federal standards. Fresh fruits other than apples are not per pound compared to _ certified. The certified apples are _ (blanks filled with one of several price combinations) per pound for the regular apples. Would you buy certified apples, regular apples, some of both, or none at all? < 1 > REGULAR APPLES < 2 > CERTIFIED APPLES < 3 > SOME OF BOTH NONE AT ALL < 8 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q29a

How many of the certified apples would you buy _ _ (blank filled with one of many prices) per pound?

< 998 > DON'T KNOW/NO OPINION < 999 > REFUSED/NO

Q29c

How many of the regular apples would you buy _ _ (blank filled with one of several prices) a pound?

[Note: Split sample variation of Q29, Q29a, and Q29c: "Tested and certified to have been produced without pesticides."] Q31

The last few questions are for statistical purposes only. We need the information to compare your opinions with the other households we are interviewing across Michigan. How many people in your household are under 5 years of age?

DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q3 la

How many people in your household are between 5 and 18 years of age?

316

Eileen 0. van Ravenswaay a71d Jennifer Wohl

< 98 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER Q31b

(Including yourself), how many people in your household are between 19 and 64 years of age?

DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q3 lc

(Including yourself), how many people in your household are over 64 years of age?

DON'T KNOW/NO OPINION < 99 > REFUSED/NO ANSWER

Q32

Respondent's gender: MALE FEMALE DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q33

What is your age?

DON'T KNOW/NO OPINION < 999 > REFUSED/NO ANSWER

Q34

What is the highest grade of school you have completed? GRADE SCHOOL ONLY < 1 > DID NOT FINISH HIGH SCHOOL < 2 > HIGH SCHOOL OR GED < 3 > VOCATIONAL OR TECHNICAL SCHOOL SOME COLLEGE COLLEGE GRADUATE (BA, BS) SOME GRADUATE OR PROFESSIONAL SCHOOL GRADUATE DEGREE (PHD, MD, MA, MBA) DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q35

To get a picture of people's financial situations, we need to know the general range of incomes of all respondents we interview. Now, thinking about your household's total annual income before taxes from all sources (including your job) in 1991, did your household receive $45,000 or more in 1991?

Using Contingent Valuation Methods to Value Health Risks

NO YES DON'T KNOW/NO OPINION REFUSED/NO ANSWER Q35a

Was it $30,000 or more? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q35b

Was it $20,000 or more? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q35c

Was it $10,000 or more? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q35d

Was it $50,000 or more? YES NO < 98 > DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q35e

Was it $60,000 or more? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

Q35f

Was it $70,000 or more? YES NO DON'T KNOW/NO OPINION REFUSED/NO ANSWER

317

15 Contingent Valuation of Consumers' Willingness to Purchase Pork with Lower Saturated Fat Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro Since the 1950s, research has shown that the saturated fatty acid content of pork can be reduced by feeding swine diets higher in unsaturated fat (Brooks 1971). This is usually accomplished by the addition of unsaturated fat or oil to the animals' diet. Advances in genetic engineering provide opportunities to improve quality traits, such as increased levels of unsaturated fatty acids or lower levels of saturated fatty acid, in commercial feed grains. The incorporation of these new varieties of grains into swine diets may make it possible to produce pork lower in saturated fat without the addition of oil to the diet (Sterling et al. 1994). This chapter examines the economic impacts of lower saturated fat (LSF) pork on market demand and price. Specifically, we address two objectives: (1) to estimate willingness to consume and (2) to estimate willingness to pay. To achieve these objectives, a nationwide telephone survey was conducted on consumers' current purchasing and consuming patterns, awareness and practices in regard to diet and health issues, and willingness to purchase and consume pork products with LSF. The analytical method used to evaluate product acceptance was effects-coding (EC) regression analysis, where willingness to pay and consume more were expressed as a function of socio-demographic variables.

Model Framework The model framework for consumer acceptance of LSF pork is based on Lancaster's attribute model of consumer choice which builds on the traditional 319

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

320

model of consumer demand (Ratchford 1975). In Lancaster's model, utility is derived from the attributes or characteristics a good possesses which influence the quantity of the good consumed. In other words, a consumer maximi:zes utility from the consumption of a bundle of products with certain attributes. Thus, the consumer's choice problem is to select attributes that maximi:ze utility under a budget constraint. Van Ravenswaay et al. (1992) developed a willingness to pay model for a single product X 1• With initial attributes r 0 offered at equilibrium price P?, the authors showed that if the demand function is linear or semialgorithmic for that product, willingness to pay (WTP) for a change in the level of one of its attributes from r0 to r 1 can be expressed as: (1)

Pl

is the willingness to pay price of X 1 after the attribute changes. where Studies have shown that households' preferences for goods are a function of socio-demographic characteristics in addition to price and attributes (Raunikar and Huang 1987). In these studies, the impact of socioeconomic and demographic variables on consumption of a good is often analy:zed from crosssectional data. Cross-sectional data usually exhibit minimal price variations. Typically, a traditional demand function with cross-sectional data is specified as a function of nonprice variables such as income, etc. Thus, in the absence of price variations and incorporating socio-demographic information, the demand · function for any commodity for cross-sectional data can be specified as:

(2) where s = a set of socioeconomic and demographic variables. Combining equations 1 and 2, WTP can be expressed as: (3) which implies that willingness to pay is a function of attributes of X 1, and the socioeconomic and demographic factors. In the same manner, the willingness to consume (WTC) X 1 for a change in the level of one of its attributes from r 0 to r 1 can be expressed as: (4)

Contingent Valuation of Consumers' Willingness to Purchase Pork

321

where X} is the quantity that the respondent is willing to consume with the equilibrium price remaining at PY after the attribute changes.

Data Collection and Survey Administration A national consumer survey was designed to collect data on willingness to pay and consume more fresh pork products with a 2S and SO percent reduction in saturated fat content. Earlier studies have shown that the technology does exist to reduce the quantity of fat in pork carcasses by S9 percent (Hollis 1989). The survey was administered over the telephone by a staff of professional telephone interviewers. A total of 1,213 potential respondents were contacted from a nationwide random commercial telephone list. A total of 417 were successfully interviewed, resulting in a response rate of 3S percent. Of the nonrespondents, 397 refused to be interviewed, 249 could not be reached, and 150 were screened for various reasons. The survey consisted of four parts (for the text of the survey see Appendix IS.A). The first part of the questionnaire concerned diet and health issues which may influence consumption patterns. The second part collected data on consumers' current consumption patterns. Part three explored consumers' potential willingness to pay for and consume pork products with reduced levels of saturated fat. This part began with an explanation of "Good Grain," a new variety of grain developed to be lower in saturated fat. Specifically, the respondents were asked: Knowing that for a serving of cooked pork (not ham) there are about 8 grams of saturated fat, if "Good Grain" can reduce the amount of saturated fat in pork from 8 grams to 6 grams, which is a 2S percent reduction, how much more per pound would you be willing to pay, and how much more would you be willing to consume? Finally, part four collected socio-demographic data on the respondents. The survey was subjected to focus group analysis for clarity and understanding. It was also pretested by phone on a group of respondents before data collection began.

Proflle of Respondents Data were collected to represent the U.S. population. Table 15.1 shows the profile of the respondents, who are primary food shoppers. The greatest number of responses came from the Midwest and South. Forty-three percent of the responding households were comprised of only two members. Over 56 percent of those responding had household incomes less than $45,000. An overwhelming majority (79 .1 percent) of the respondents were female indicating that females do the majority of the primary food shopping. Of particular interest is the ethnic background of the respondents, where the sample skewed in favor of a majority of white Americans (91.6 percent). A large majority (71.4

322

TABLE 15.1 Profile of Respondents Percent

Region•

Midwest Northeast South West

Household Size 1 2 3 4

5-6

7 or more

Income (000) < 15 15-24 25-34 35-44 45-54 55-64

> 65

Refused

Gender

Female Male

Ethnic

Black

White Asian

American Indian Hispanic Other

Number of Respondents

Survey

U.S.

131 89 123 74

31.4 21.3 29.5 17.8

24.0 21.0 34.0 21.0

63 181 63 70 33 7

15.1 43.4 15.1 16.8 7.9 1.7

25.0 32.0 17.0 15.0 9.0 1.7

53 73 51 41 22 33 86

58

13.91 12.72 17.51 12.23 9.83 5.27 7.91 20.62

24.3 17.5 15.8 11.9 11.0 5.9 9.5

330 87

79.14 20.86

53.0 47.0

25

5.99 91.61 0.72 0.24 0.72 0.72

12.0 75.0 3.4 0.5 9.1

382 3 1 3 3

(continues)

Contingent Valuation of Consumers' Willingness to Purchase Pork

323

TABLE 15.1 (continued) Percent Number of Respondents

Marital Status Married Unmarried

Age 18-24 25-34 35-44 45-54 55-64 65 and over Refused

Education

Below 12th Grade High School Graduate Some College College Graduate Postgraduate

Survey

U.S.

297 119

71.39 28.61

71.0 29.0

19 73 82 63 75 101 4

4.56 17.51 19.66 15.10 17.99 24.22 0.96

26.0 21.2 18.3 12.1 9.8 12.3

38 155 104 56 32

9.87 40.26 27.01 14.55 8.31

24.8 30.0 24.9 13.1 7.2

•Regions defined according to U.S. Department of Labor regional classification for their Consumer Expenditure Diary Surveys.

percent) of the respondents indicated they were married. The age group of 65 and over had the greatest representation (24.2 percent). Slightly less than half (49. 9 percent) of those responding had attained an educational level beyond that of high school, with less than 23 percent completing college or higher. In general, the survey respondents' profiles fitted the U.S. Census population profile quite well (see Table 15.1). The discrepancies in some areas are due to the focus of the survey which only targets pork consumers and primary shoppers.

Results In the telephone survey, respondents were asked a number of questions concerning diet and health awareness, which could influence their consumption patterns (for a complete text of the questionnaire, see Appendix 15.A). A number

324

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

of consumer surveys provide evidence that nutrition and wholesomeness are major concerns of the meat-buying public (Breidenstein and Carpenter 1983, Putter and Frauo 1988). Respondents were asked to identify their primary source of diet and health information. The media served as the principal source of information for a large number of respondents with 17.9 percent indicating their information came from a magazine, 17.6 percent referred to the newspaper, and 10.2 percent considered TV/radio as their primary source of health and diet information. Respondents also relied on the medical profession for health-related information with 40.6 percent reporting their information came from a doctor. When asked if they knew their serum cholesterol level, the majority (55.S percent) of the respondents said they did not. Of the 44.5 percent who did, 154 respondents gave an actual level which ranged from 66-347 mg/di. These cholesterol levels were divided into categories associated with the degree of risk for heart disease (National Institutes of Health 1987). Thirty-nine percent of the respondents were in the low risk category ( < 200 mg/di), 38.3 percent were in the borderline-high risk category (200-239 mg/di), and 22. 7 percent were in the high risk category(> 240 mg/di). Respondents were asked if they had heard of saturated and unsaturated fat. Less than one percent of the respondents indicated they had not heard of saturated and unsaturated fat. Regardless of their answers, respondents were given a standard definition (see Appendix 15.A) of saturated and unsaturated fat before WTC and WTP questions were asked. Respondents were also asked questions on how often they read information provided on food labels. Seventy-eight percent of the respondents said they usually read the label for information on fat, and 10 percent said they sometimes read for fat information. When asked about reading labels for information on cholesterol, 59 .1 percent said they usually did, and 12.S percent did sometimes. Respondents were asked if they read the labels of meat products for information on fat content. Nearly 51 percent indicated they usually did, 13. 7 percent said they sometimes did, and 35.S percent said they never read the labels of meat products for information on fat content. Respondents were asked if they had reduced their consumption of fresh pork or processed pork products in the last five years. Fifty-nine percent of the respondents said they had reduced their consumption of fresh pork and 56 percent had reduced their consumption of processed pork. When asked if concern about fat intake was the reason for the reduction in consumption of pork products, 68 and 77 percent said it was the reason for their reduced consumption of fresh and processed pork, respectively. Fifty-four percent of the respondents indicated they were willing to increase their consumption of fresh pork if the saturated fat content was reduced. Respondents were also asked how many more times per month they would eat pork if saturated fat levels were decreased by 25 percent and SO percent. Direct results from the survey indicate that for fresh pork products with 25 percent less

Contingent Valuation of Consumers' Willingness to Purchase Pork

325

saturated fat, 27 percent qf the respondents would pay 10 cents more, 14 percent would pay 25 cents more, and 11 percent of the respondents said they would pay 50 cents more per pound (see Table 15.2). Thirty-six percent said they would not pay any more for pork with a 25 percent fat reduction. If the saturated fat level of fresh pork products was reduced by 50 percent, 15 percent of the respondents said they would pay 10 cents more, 18 percent would pay 25 cents more, 10 percent would pay 50 cents more, and 8 percent indicated they would be willing to pay 75 cents more per pound. Twenty-seven percent of respondents said they would not pay any more for pork with a 50 percent saturated fat reduction.

Impact of Socio-Economic and Demographic Factors on WTC and WTP for LSF Porlc The impact of socio-economic and demographic factors on willingness to consume and pay more for LSF fresh pork was analy7.ed using effects coding (EC) regression. The EC dummy variable technique was used to code the qualitative socio-demographic variables of the regression model (Cohen and Cohen 1983). The models for willingness to consume and pay more were expressed as follows: TABLE 15.2 Wtllingness to Pay for LSF Pork Reduction in Saturated Fat Price Increase in Cents

25%

50%

% of Respondents 0

10 15 20 25 30

1 4

27 15

s

4

14

18

so

11

100 100

2

10 8 3

40

45

75

>

36 27

Note: There were 417 total responses.

1 1 0

3

1

3 1 1

4

326

(5)

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

WTC, WTP = 130 + fJ 1;(Reg) + fJu(Size) + {J3;(Edu) + /3 4;(Eth) + fJs;(Age) + {J6;(Inc) +

131;(Rdl)

+

fJs;( Gend)

where i represents the number of categories within the variable, Reg = region, Size= household size, Edu = education level, Eth =ethnic background, Age = age, Inc = income, Rdl = those who read labels, and Gend = gender (see Table 15.1 for categories). The intercept {J 0 represents WTC and WTP without the effects of demographic variables. The value of using effects coding instead of traditional dummy variable coding is the ability to easily obtain coefficients for all (k) levels of all attributes. In effects coding, the kth base level is represented as -1 instead of 0. This coding technique constrains the levels of each feature to sum to 0. The coefficient for the base level is easily calculated as the negative sum of the (k 1) level coefficients. The intercept becomes the mean willingness to pay/ consume, and socio-demographic coefficients measure deviation from the mean (Wirth 1989). Results of the EC regression models are summarized in Table 15.3. The empirical model specifies increase in consumption and price as a function of various socio-demographic factors. The results are shown for two levels of fat reduction, 25 percent and 50 percent. Table 15.3 shows that if the saturated fat content of fresh pork is reduced by 25 and 50 percent, the average consumption of this product without socio-demographic effects will increase an average of 3. 9 and 4.9 times per month. Moreover, depending on the level of saturated fat reduction, there are regional and socio-demographic differences that result in significantly greater increases ( +) or decreases (-) at at least the 10 percent significance level from the average increase in consumption per month. For example, for 25 percent saturated fat reduction, West ( + ), household size of three(-), household size greater than 4 ( + ), education level less than 12th grade (+), black (+), age between 25-34 (+), and age between 45-54 (-) were significantly different from the average increase in consumption of 3.9 times more per month. For the 50 percent saturated fat reduction, the Northeast(-), West (+),household size of three(-), age between 25-34 ( + ), age greater than 64 (-), and income greater than $64,000 ( + ), were different from the average increase of 4.9 times per month. Table 15.3 also shows that if the saturated fat content of fresh pork is reduced by 25 and 50 percent, the average willingness to pay without sociodemographic effects will increase an average of 20 and 29.9 cents per pound, respectively. Moreover, depending on the level of saturated fat reduction, there are regional and socio-demographic differences. For 25 percent saturated fat reduction, the Midwest (-), household size of two ( + ), income level between $55,000 to 64,000 ( + ), and male (-) were significantly different from the average willingness to pay more. For the 50 percent saturated fat reduction, the

327

TABLE 15.3 Estimated Influence of Socio-Demographic Characteristics on WfC and WTP for LSF Fresh Pork

SocioDemographic Characteristics

Average Increase

Consumption (times per month) 25 percent LSF

50 percent LSF

25 percent LSF

50 percent LSF

3.9

4.9

20.0

29.9

-S.1

-S.9

4.S

7.7

Region

Midwest Northeast South West Household Size One Two Three Four >Four

Education

Below 12 High-School Graduate Some College College Graduate Post Graduate

Ethnicity

Black White

Price (cents per pound)

-1.0 0.8

1.0

-1.4

-1.S

-10.3

1.1 1.1

0.9

Age

18-24 25-34 35-44

4S-S4 SS-64

> 64

1.1

1.3

-1.1

-1.8

(continues)

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

328

TABLE 15 .3 (continued)

SocioDemographic Characteristics

Consumption (times per month)

25 percent LSF

50 percent LSF

Price (cents per pound)

25 percent LSF

50 percent LSF

Income (000 $) Below 15 15-24

-10.4

25-34 35-44 45-54

55-64

2.0

> 64

10.6

21.0

-3.5

-5.3

Read Label Yes Sometimes No Gender Female Male No. of Respondents

200

211

255

292

Note: Numbers in table are significant at least at the 10 percent level.

Midwest(-), household sii.e of two ( +), household sii.e greater than four(-), income level between $15,000 and 24,000 (-) or $55,000 and 64,000 ( +), and male(-) were significantly different from the average willingness to pay more.

Conclusions Animal products contribute significantly to the total nutrients in the food supply, with meat accounting for the largest proportion of the calories. Advances in genetic engineering have made it possible to develop feed grains which are tailored to the specific needs of the livestock producer, and the desires of the health-conscious meat consumer. The ability to manipulate the fatty acid

Contingent Valuation of Consumers' Willingness to Purchase Pork

329

composition of pork by feeding tailor-made grain lower in saturated fat, may allow pork producers to create higher-quality, healthier pork products. A national consumer survey was designed to collect data on consumers' willingness to pay and consume pork products with a 25 and SO percent reduction in saturated fat. The survey also asked how much more consumers would be willing to pay for the improved pork products. The manner in which the survey questions were asked minimii.ed unrealistic responses by first establishing current level of price and consumption, and then asking the willingness to consume and pay more. The effects coding regression technique was used to evaluate how each level of each socio-demographic variable impacts WTC and WTP models. Results confirmed the hypothesis that if the level of saturated fat is reduced, pork consumption will increase. The model results showed that the largest potential for increase in consumption will be household sim of three or greater. Respondents from the West appeared to be the most willing to increase consumption of pork if its nutritional quality were improved. Respondents age 45 and older were the least likely to increase consumption of the improved or value-added pork products. Overall, survey respondents were willing to pay (on average 16 to 23 cents) more per pound for fresh pork with reduced levels of saturated fat. Respondents with larger incomes were more willing to pay a higher price for the improved products. Respondents from the Midwest region were generally less willing to pay more. This could be due to the fact that the consumers in the Midwest region are already consuming more pork than those in other regions. Household sims of two are more willing to pay higher prices for LSF pork. Respondents with income above $54,000 are more willing to pay higher prices. Finally, males are less likely to pay more. It was surprising that the demographic variables of education, ethnicity, and respondents who read labels, did not have much effect on increased consumption of pork or willingness to pay higher prices for healthier pork products. Respondents also said that the new pork products should be appropriately labeled to identify the added value. For example, pork from "Good Grain" should be labeled as lower in saturated fat content. Past purchasing patterns suggest this is essential to the successful marketing of new and improved pork products.

References Breidenstein, B. C. and Z. L. Carpenter. 1983. The Red Meat Industry: Product and Consumerism. Journal of Animal Science 57(sup. 2):119-132. Brooks, C. C. 1971. Fatty Acid Composition of Pork Lipids as Affected by Basal Diet, Fat Source, and Fat Level. Journal of Animal Science 33:1224-1231. Cohen, Jacob and Patricia Cohen. 1983. Applied Mulliple Regression-Correlation Analysis for Behavioral Sciences. 2nd Edition. New Jersey: Lawrence Erlbaum.

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

330

Hollis, G. R. 1989. Porcine Somatotropin: A Tool for Producing Leaner Hogs More Efficiently. Veterinary Medicine 1989(December):1188-1194. National Institutes of Health. 1987. Eating tQLower Your High Blood Cholesterol. No. 87-2920. Washington, D.C.: National Institutes of Health. Putler, D. and E. Fru.ao. 1988. Food Review. United States Department of Agriculture, Economic Research Service:16-22. Ratchford, Brian T. 1975. The New Economic Theory of Consumer Behavior: An Interpretive Essay. Journal of Consumer Research 2:65-75. Raunikar, Robert and Chung-Liang Huang. 1987. Food Demand Analysis. Ames, IA: Iowa State University Press. Sterling, Lesa G., Gary M. Fader, Beth H. Gutowski, and Catherine K. Halbrendt. 1994. The Effect of Source and Level of Dietary Fat on the Fatty Acid Composition of Muscle and Adipose Tissue in Swine. The ProfessionalAnimal Scientist 10: 11-17. van Ravenswaay, Eileen 0., Jennifer B. Wohl, and John P. Hoehn. 1992. Michigan Consumers' Perceptions of Pesticide Residues in Food. Department of Agricultural Economics Staff Paper No. 92-56, Michigan State University. Wuth, Ferdinand F. 1989. Mid-Atlantic Market Preferences Toward Farm-Raised Hybrid Striped Bass: A Conjoint Analysis. M.S. Thesis, University of Delaware.

Appendix 15.A TELEPHONE SURVEY City: _ _ _ _ _ _ _ __

-----------------

State: Tel.:

LOW SATURATED FAT-PORK PRODUCTS SURVEY Introduction: Hello, I am . I am calling for the University of Delaware. We are conducting a national survey to look at Food Purchasing Decisions and we are interested in interviewing you to get your opinion. Like to talk to the primary food shopper. If busy, call back later.

Time: 1.

---------

---------

Date:

Have you ever eaten or do you currently eat m of the following: Pork, bacon, sausage, or ham. Yes ( ) No ( ) (IF 'NO' THEN TERMINATE THE INTERVIEW)

331

Contingent Valuation of Consumers' Willingness to Purchase Pork

2.

Including yourself, how many people are there in your household?

(CHECK ONE)

Three Four

One ( ) Two ( )

Five or six Seven or more

( ) ( )

( ) ( )

3.

Are you or any of your family on a fat or cholesterol-restricted diet? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO S)

4.

For which of the following reasons?

(CHECK ALL THAT APPLY)

Medical Reasons ( ) Hereditary Reasons ( )

5.

Weight Reduction ( ) Other (Please Identify)_ _ _ __

What is your primary source of healthy eating tips?

(CHECK ONE)

Doctor Newspaper TV/Radio

Magazine ( ) Newsletter ( ) ( ) Books

( ) ( ) ( )

0th~-

6.

Do you know your blood cholesterol level? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 8)

7.

What is your level?

8.

On average how often, at home and away from home, do you eat the following?

a. b. c. d. e. f. g. h.

9. 10.

~-

Beef Chicken Turkey (luncheon meat) Fish or seafood Fresh pork/pork chops Bacon Sausages Luncheon ham

Times a Week

( ( ( ( ( ( ( (

) ) ) ) ) ) ) )

Times a Month

( ( ( ( ( ( ( (

) ) ) ) ) ) ) )

Times a Year

( ( ( ( ( ( ( (

) ) ) ) ) ) ) )

Do you read labels on the food you buy? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 13) Which of the following do you read for and how frequently?

Other

( ( ( ( ( ( ( (

) ) ) ) ) ) ) )

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

332

a. b. c. d. 11.

Usually

Calories Cholesterol level Fat content Other

( ) ( )

Sometimes ( (

) ) ( ) ( )

( )

( )

-----

Never ( ) ( ) ( ) ( )

If you are shopping and you see a product label that says either low-fat, cholesterol-free, sugar-free, low-sodium, how accurate do you think they generally are: (CIRCLE ONE)

VERY ACCURATE

1

2

3

4

NOT ACCURATE

5

12.

Do you read labels on meat and meat products such as luncheon meats to see what the fat content is? (CHECK ONE) Sometimes ( ) Never ( ) Usually ( )

13.

Have you reduced the amount of fresh pork (not ham) you ate in the last five years? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 15)

14.

Do you eat Im fresh pork-because in your opinion, it contains more fat than other meats? Yes ( ) No ( )

15.

pork (sausages, bacon, and Have you reduced the amount of p~ luncheon ham) you ate in the last five years? Yes ( ) No ( )

16.

Do you eat Im processed pork like sausages, bacon and ham-because you are concerned with the amount of fat? Yes ( ) No ( )

17.

Ifno,why? _ _ _ _ _ _ _ _ _ _ _ _ __

18.

Have you heard of the following kinds of fats? A. saturated Yes ( ) No ( ) B. unsaturated Yes ( ) No ( )

READ TO RESPONDENTS

To help you understand the next few questions I am going to read some facts about fat and grain.

Contingent Valuation of Consumers' Willingness to Purchase Pork

333

Fats are made up of saturated and unsaturated fat. A diet high in saturated fat tenm to raise the total blood cholesterol level. High blood cholesterol levels are associated with an increased risk of heart disease. STATEMENTS FOR REFERENCE A grain bas been developed to feed hogs that reduces the amount of saturated fat in pork products but maintains the same flavor and texture. We call this grain "Good Grain" for later reference. To give you some reference pointsA serving of cooked (i.e. roasted) beef contains about 9 grams of saturated fat. A serving of cooked (i.e. roasted) chicken contains about 3 grams of saturated fat. PROCEEDING TO THE NEXT QUESTION 19.

For a serving of cooked pork (not ham) there are about 8 grams of saturated fat. If "Good Grain" can reduce the amount of saturated fat in pork from 8 grams to 6 grams which is a 25 % reduction, how much more per pound would you be willing to pay? (CIRCLE THE

ANSWER, ASK FROM THE IDGHEST PRICE AND STOP ONCE ANSWER IS GIVEN) Price In-

7S cts/lb

20.

SO cts/lb

crease/Lb. 2S cts/lb

No

10 cts/lb

Other__

Change ( )

How about if 'Good Grain' can reduce the saturated fat from 8 grams to 4 grams which is a SO% reduction, how much more would you be willing to pay? Price In-

7S cts/lb

SO cts/lb

crease/Lb. 2S cts/lb

No

10 cts/lb

Other

21.

How much per pound do you pay for fresh pork?

22.

Average U.S. price is $3.03/lb. Are you paying this amount?

YOU GET PRICE SKIP TO 23)

Change ( )

- - - - -?

(IF

334

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

A. Yes ( ) (IF 'YES' SKIP TO 23) (IF 'NO' THEN PROCEED TO ASK) B. Higher? About $3.SO C. Lower? About $3.00

D. Don't care about price E. Don't know 23.

( ) ( )

( ) ( )

No ( ) Other (Specify) _ _ Other (Specify) _ _

If 'Good Grain' can reduce the amount of saturated fat, do you think you would eat more fresh pork? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 26)

(REPEAT TO RESPONDENTS) 24.

OK, if reduction is from 8 to 6 grams or a 25 % reduction then how many more times a week/month would you eat 'Good Grain' pork? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

25.

If reduction is 8 to 4 grams which is a SO% reduction. How many

more times a week/month do you think you would eat 'Good Grain' pork? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

AGAIN FOR YOUR REFERENCE, A SERVING OF COOKED BEEF CONTAINS 9 GRAMS OF SATURATED FAT A SERVING OF COOKED cmCKEN CONTAINS 3 GRAMS OF SATURATED FAT 26.

For a serving of cooked bacon there are about 18 grams of saturated fat. If 'Good Grain' can reduce the amount of saturated fat in bacon from 18 grams to 13 grams which is a 25 % reduction, how much more per pound would you be willing to pay? (CIRCLE THE ANSWER, ASK

FROM THE filGHEST PRICE AND STOP ONCE ANSWER IS GIVEN)

75 cts/lb

27.

50 cts/lb

Price Increase/Lb. 25 cts/lb

No

10 cts/lb

Other__

Change ( )

How about if 'Good Grain' can reduce the saturated fat from 18 grams to 9 grams which is a SO% reduction, how much more would you be willing to pay?

Contingent Valuation of Consumers' Willingness to Purchase Pork

75 cts/lb

50 cts/lb

Price Increase/Lb. 25 cts/lb

10 cts/lb

Other

335

No Change ( )

28.

How much per pound do you pay for bacon? _ _ _ (IF YOU GET A PRICE SKIP TO 30)

29.

Average U.S. price is $2.30/lb. Are you paying about this amount? A. Yes ( ) (IF 'YES' SKIP TO 30) No ( ) (IF 'NO' THEN PROCEED TO ASK)

B. Higher? About $2.SO

C. Lower? About $2.00 D. Don't care about price E. Don't know

( ( ( (

) ) ) )

Other (Specify) _ __ Other (Specify) _ __

30.

If 'Good Grain' can reduce the amount of saturated fat, do you think you would eat more bacon? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 33)

31.

OK, if reduction is from 18 grams to 13 grams or a 25 % reduction then how many more times a week/month would you eat 'Good Grain' bacon? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

32.

If reduction is 18 grams to 9 grams which is a 50 % reduction. How many more times a week/month do you think you would eat 'Good Grain' bacon? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

33.

For a serving of cooked sausages there are about lS grams of saturated fat. If 'Good Grain' can reduce the amount of saturated fat in sausages from lS grams to 11 grams which is a 25 % reduction, how much more per pound would you be willing to pay? (CIRCLE THE ANSWER, ASK FROM THE IDGHEST PRICE AND STOP ONCE ANSWER IS GIVEN)

75 cts/lb

34.

50 cts/lb

Price Increase/Lb. 25 cts/lb

10 ctsnb

Other

No Change ( )

How about if 'Good Grain' can reduce the saturated fat from lS grams to 8 grams which is a 50 % reduction, how much more would you be willing to pay?

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

336

15 cts/lb

50 cts/lb

Price Increase/Lb. 25 cts/lb

10 cts/lb

Other

----

No Change ( )

(IF YOU

35.

How much per pound do you pay for sausages?

36.

Average U.S. price is $2.30/lb. Are you paying about this amount? No ( ) A. Yes ( ) (IF 'YES' SKIP TO 37)

GET A PRICE SKIP TO 37)

(IF 'NO' THEN PROCEED TO ASK)

B. Higher? About $2.50 C. Lower? About $2.00 D. Don't care about price E. Don't know

( ) ( )

( ) ( )

Other (Specify) _ __ Other (Specify) _ __

37.

If 'Good Grain' can reduce the amount of saturated fat, do you think you would eat more sausages? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 40)

38.

OK, if reduction from 15 grams to 11 grams or a 25 % reduction then how many more times a week/month would you eat 'Good Grain' sausages? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

39.

If reduction is 15 grams to 8 grams which is a 50% reduction. How many more times a week/month do you think you would eat 'Good Grain' sausages? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

40.

For a serving of cooked luncheon ham there are about 8 grams of saturated fat. If 'Good Grain' can reduce the amount of saturated fat in ham from 8 grams to 6 grams which is a 25 % reduction, how much more per pound would you be willing to pay? (CIRCLE THE

ANSWER, ASK FROM THE WGllEST PRICE AND STOP ONCE ANSWER IS GIVEN)

15 cts/lb

41.

50 cts/lb

Price Increase/Lb. 25 cts/lb

10 cts/lb

Other__

No Change ( )

How about if 'Good Grain' can reduce the saturated fat from 8 grams to 4 grams which is a 50% reduction, how much more would you be willing to pay?

Contingent Valuation of Consumers' Willingness to Purchase Pork

75 cts/lb

50 cts/lb

337

No

Price Increase/Lb. 25 cts/lb

10 cts/lb

Change

Other

( )

42.

How much per pound do you pay for ham'? _ _ _ (IF YOU GET A PRICE SKIP TO 44)

43.

Average U.S. price is $2.70/lb. Are you paying about this amount? A. Yes ( ) (IF 'YES' SKIP TO 44) No ( ) (IF 'NO' THEN PROCEED TO ASK) B. C. D. E.

Higher? About $3.00 Lower? About $2.50 Don't care about price Don't know

( ( ( (

) ) ) )

Other (Specify) _ __ Other (Specify) _ __

44.

If 'Good Grain' can reduce the amount of saturated fat, do you think you would eat more ham'? Yes ( ) No ( ) (IF 'NO' THEN SKIP TO 47)

45.

OK, if reduction is from 8 grams to 6 grams or a 25 % reduction then how many more times a week/month would you eat 'Good Grain' ham? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

46.

If reduction is 8 grams to 4 grams which is a 50% reduction. How many more times a week/month do you think you would eat 'Good Grain' ham? TIMES a WEEK ( ) TIMES a MONTH ( ) TIMES a YEAR ( )

(ASK THE NEXT QUESTION IF ANY OF THE FOLLOWING 23, 30, 37 OR 44 WERE 'YES' ANSWERS) 47.

Since you have indicated that you will eat more pork products, will you eat less of the following: a. b. c. d.

48.

Eat less beef Eat less chicken Eat less turkey Eat less fish or seafood'?

Yes ( ) ( ) ( ) ( )

No ( ) ( ) ( ) ( )

if yes if yes if yes if yes

Times a Month

( ( ( (

) ) ) )

What meat products do you generally buy according to brand name not grocery store brand?

Catherine Halbrendt, Lesa Sterling, Sue Snider, and Gail Santoro

338

Food Group a. b. c. d. e. f. g. h.

Usually

Chicken Beef Porlc Turlcey Fish or seafood Bacon Sausages Luncheon ham

( ( ( ( ( ( ( (

Sometimes ( ( ( ( ( (

) ) ) ) ) ) ) )

Never

) )

( ( ( ( ( ( ( (

) ) )

) ( ) ( )

) ) ) ) ) ) ) )

49.

If 'Good Grain' pork products are developed would you prefer the products to carry a brand name? Yes ( ) No ( )

SO.

What is the average time (in minutes) you use to prepare a main course meal during: Weekdays ( ) Weekends ( )

51.

On average, how many times a month do you eat out? For breakfast? For dinner?

times times

For lunch?

times

52.

Of that how many times included porlc or pork products (such as pork chops, bacon, luncheon ham, and sausages?) times

53.

What is your highest level of education?

(READ ALL CATEGORIES)

Below grade 12 High School Graduate Some College or Vocationalffechnical School College Graduate Postgraduate

54.

) ) ) ) )

Which age category do you fall into? 18-24 ( )

25-34 ( ) 35-44 ( )

SS.

( ( ( ( (

45-54 55-64 65 and over

( ) ( ) ( )

Please indicate your marital status. Married ( )

Refused ( )

Unmarried ( )

Contingent Valuation of Consumers' Willingness to Purchase Pork

339

56.

Which Household Income category do you fall into? Below $15,000 ( ) $45,000-54,000 $15,000-24,000 ( ) $55,000-64,000 $25,000-34,000 ( ) $65,000 and over $35,000-44,000 ( ) Refused

51.

What is your ethnic background? Black (nonHispanic) ( ) American Indian/Native Alaskan ( ) White (nonHispanic) ( ) Hispanic ( ) Asian or Pacific Islander ( ) Other (Please specify)_ _ _ __

58.

What is the occupation of head of household?

59.

Sex of respondent? Male ( ) Female ( )

~

( ( ( (

) ) ) )

16 Criteria for Evaluating Results Obtained from Contingent Valuation Methods Richard B. Belzer and Richard P. Theroux 1 Recent years have seen rapidly expanding interest in a variety of nonobservational methods of collecting data on economic phenomena. Foremost among these methods is contingent valuation (CV), a methodology that has become prominent through its application to the problem of valuing environmental amenities and similar commodities which are not directly traded in marlcets (e.g., Cummings et al. 1986, Mitchell and Carson 1989, Environmental Pro~tin Agency 1993). The goal of CV methods is to simulate the same kind of ordered preferences which economic theory argues would be revealed through market behavior if such markets existed (Freeman 1979: 97). Thus, the application of CV methods has been largely limited to public-good commodities that are not traded in markets and for which there are few if any alternative methodologies. New interest in CV can be attributed primarily to natural resource damage litigation spawned by the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA), the Clean Water Act (CWA), the Oil Pollution Act of 1990 (OPA), and intensified oversight of federal regulatory activity conducted by the Executive Office of the President. Section 31 l(f) of CWA and section 107 of CERCLA authori7.e natural resource trustees to recover compensatory damages for injury to or destruction of natural resources resulting from a discharge of oil into navigable waters or a release of hazardous substances. Section 1006(e) of OPA authori7.ed the National Oceanic and Atmospheric Administration to develop rules for assessing natural resource damages for discharges of oil into navigable waters. Regulatory oversight is performed by the Office of Information and Regulatory Affairs (OIRA), a statutory office within the Office of Management and Budget (OMB), as directed by Executive Order No. 12866. 2 Among other things, this Order subjects significant regulatory actions of federal departments 341

342

Richard B. Belzer and Richard P. Theroux

and agencies to a rudimentary benefit-cost test. Included within the realm of regulatory actions are those involving the mandated provision or disclosure of information and requirements for the collection, retention, and submission of data (Office of Management and Budget 1990b). Regulatory actions that fail to offer social benefits in excess of social costs bear a special burden of policy justification, inasmuch as they violate traditional normative welfare economics standards. Much of the interest in applying CV methods to food safety issues has arisen because of these high-level demands for regulatory analysis.

OMB Guidance Concerning the Use of CV in Regulatory Impact Analysis With respect to federal regulation, OMB has published guidance for agencies to use in performing Regulatory Impact Analyses (RIAs) of major rules (Office of Management and Budget 1990c, Office of Management and Budget 1990d). 3 OMB guidance establishes a preference for observational or behavioral data in the development of benefit and cost estimates. Because such data reflect voluntary exchanges they can be presumed to reflect the economic preferences of individuals in the absence of a demonstrated market failure. Where benefits derive from risk assessments and an agency chooses to represent benefits with point estimates, the guidance calls for the use of expected value estimates of risk. Conventional risk assessment methods are not appropriate for use in benefit-cost analyses; embedded conservatism results in highly exaggerated point estimates of both risks and benefits, thus inserting a bias in favor of government intervention to remedy problems whose scope and severity are similarly overstated (Office of Management and Budget 1990a). OMB's guidance acknowledges the difficulty of estimating individuals' willingness to pay (WTP) for commodities that are not traded in markets and thus are impossible to value using conventional observational methods. However, the guidance also recognii.es that nonobservational methods such as CV warrant an additional burden of analytic rigor: Contingent valuation methods provide the only analytical approaches currently available for estimating the benefits of such untraded goods. The absence of observable and replicable behavior with respect to the benefit in question, combined with the difficulties of avoiding bias in contingent valuation studies, argues for great care and circumspection in the use of such methods. This means, for example, that estimates of willingness to pay must incorporate the variety of alternative means individuals have of expressing value for untraded goods. Moreover, analyses must faithfully capture individuals' budget constraints, which restrict their willingness to pay for untraded as well as traded goods and services. Benefit analyses derived from contingent valuation and similar methods thus require considerable analytic rigor in design and careful

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

343

execution. Absent such efforts, analyses based heavily on the benefits of untraded goods and services nonnally would fail the test of a satisfactory RIA (Office of Management and Budget 1990d: 661).

This flexible performance standard for RIAs was crafted after considering a number of comments from federal agencies. In responding to these comments, OMB acknowledged the need for CV and similar methods for estimating the value of untraded goods and services. Nevertheless, OMB insisted on a very high standard of care in the use of CV, including an extensive effort to test alternative explanatory hypotheses: Survey estimates may be necessary to estimate certain physical and psychological uses, because relevant behavior is unobservable. However, the problems that arise in the estimation of use value through survey methods are considerably more serious. Great care needs to be taken to ensure that survey designs do not introduce systematic biases by departing from market-based valuation principles. For example, slight changes in the way questions are presented can sometimes result in dramatic changes in responses, because of the hypothetical nature of data derived from survey instruments. This hypothetical character means that survey methods offer considerable opportunities for abuse. Analyses relying on survey instruments to estimate benefits should devote considerable efforts to quality control, data verification, and real-world hypothesis testing. Major departures from market-based principles can lead to serious distortions in the allocation of our Nation's scarce resources (Office of Management and Budget 1990c: 37-38).

Finally, the OMB guidance sounds a warning to federal agencies inclined to employ CV methods in support of federal regulation: CV-based analysis will be judged in accordance with extraordinary standards, particularly if it is relied upon for a large share of the estimated benefits: Because of the potential for misuse of survey methods, RIAs generally should avoid relying exclusively on value estimates derived from survey approaches ... [D]epartments and agencies that develop benefit estimates which rely heavily on the results of survey instruments bear an extraordinary burden to show that estimates obtained are reasonably consistent with observable market behavior and common sense (Office of Management and Budget 1990c: 38).

Interest in CV methods among federal regulatory agencies is a product of these institutional demands for rigorous policy analysis. It is important to keep foremost in mind that while CV research may serve to advance knowledge and improve economic methodologies generally, agencies' primary interest in CV research lies in its potential capacity to support or expand programmatic responsibilities and achieve organizational objectives. This creates an unavoidable tension between CV researchers and the agencies that fund their work. Continued funding depends to some extent on the ability

344

Richard B. Belzer and Richard P. Theroux

of CV methods to justify regulations and programs-if not now, then in the foreseeable future. Researchers thus face potentially perverse incentives to "see" promise where it is ephemeral and to overinterpret the results of the studies they perform. The purpose of this chapter is to provide a look at CV methods and applications from the perspective of a "consumer" of benefit-cost analyses prepared in support of governmental programs, policies, and regulations. While some practitioners have celebrated CV's "arrival" (Cummings et al. 1986), the controversy surrounding CV only continues to intensify (Cambridge Economics 1992). Our intent is neither to disparage nor promote the CV method, for we are agnostic as to its ultimate capacity to answer some very difficult empirical questions which the economics profession bas largely ignored. The issues and concerns set forth in this chapter should not be interpreted as comprehensive, nor are the suggested hypothesis tests intended to be either exhaustive or appropriate in every context. Rather, they represent relatively simple issues which have become evident in the course of reviewing government agency CV survey instruments and regulatory analyses relying upon CV methods and results. Further, agencies need encouragement to set high standards for the CV research they sponsor, both to stimulate high quality research and to resist the unavoidable countervailing pressure to utilim flawed results which might nevertheless appear to be attractive. Throughout this chapter we have made several important simplifying assumptions. First, we presume that a significant market failure has been demonstrated to exist with respect to a specific food safety risk. This means that carefully devised government intervention may increase social welfare in accordance with the Kaldor-Hicks criterion. We take no issue at all with private parties applying CV (or any other methods) as long as they alone bear the costs and capture the benefits of acting based on the results of such research. Our concern is solely with efforts to use government outlays or regulations to achieve public purposes. Second, we assume that researchers are pursuing CV methods because the relevant portion of the demand curve for risk reduction is not observable and there are no market-based data to rely upon for estimating consumers' WTP. Where the relevant segment of the demand curve is observable or can be approximated from other commodity markets in which food safety risk is a significant attribute, we expect that traditional valuation methods based on observations of behavior would be used instead of CV. Third, it is our firm conviction that governments bear a special burden whenever they propose to intervene in private transactions. Regardless of whether they spend monies collected through taxes or command others to make expenditures through regulation, government officials have a fiduciary duty when it comes to other people's money. Whether one fritters away one's own assets or inheritance is not generally a public concern. However, the power to consume the public purse or command private parties to reallocate the expenditure of

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

345

private funds carries with it a special ethical responsibility to be careful and deliberate. Thus, we presume that it is both fair and appropriate that those who would use the financial resources or coercive powers of the government also bear a disproportionately greater burden to demonstrate the wisdom and analytical soundness of their proposals. Finally, we are not experts in CV, nor are we advocates or critics. Our interest in the CV method and its potential applications arises because we are "consumers" of benefits analyses and regulatory policy initiatives in which CV methods are playing an increasingly important role. Many of the applications we have seen have made us acutely uncomfortable. We see a clear need for comprehensive criteria acceptable to all sides which could guide future public policy applications of CV methods. This chapter represents an effort to begin the dialogue necessary to develop such criteria.

Methodological Concerns Raised by Contingent Valuation Methods Practitioners of CV continue to struggle with a variety of serious methodological issues. At the risk of oversimplification, these issues may be captured by the following three questions: 1.

2. 3.

Do the expressed WTPs provided by CV survey respondents under hypothetical conditions adequately simulate behavior under real-world conditions? Do CV respondents have a clear understanding of the identity and character of the commodity which they have been asked to value? Are the expressed WTPs elicited by CV surveys reliable and valid estimates of respondents' economic preferences?

Each of these questions implicitly suggests the existence of necessary and sufficient conditions for CV estimates to be both valid and reliable. Because the CV method is still relatively new and its capable practitioners few, we seem to be a long way from developing sufficient conditions. Nevertheless, it is clearly necessary that the answers to these three questions must be "yes" before it is appropriate to use CV-based estimates in applied work having important public policy implications.

Hypothetical vs. Real-World Data Endemic to CV methods is the fact that the transactions examined are hypothetical, and answers to such questions are only hypothetically accurate. The problem, of course, is that absent actual behavior there is no way to ensure

346

Richard B. Belzer and Richard P. Theroux

that respondents give "real" answers. Experienced CV practitioners thus devote considerable efforts to establishing realism in an attempt to elicit "real" answers and to develop hypothesis tests which would identify possible biases (e.g., Cummings et al. 1986, Mitchell and Carson 1989). Others have performed experiments in which serious biases have been detected, such as discrepancies between cash and hypothetical payments (Bishop and Heberlein 1986, Neill et al. 1994), noncommitment biases related to budget constraints (Kemp and Maxwell 1992), and the presence of embedded goods and moral satisfaction (Kahneman and Knetsch 1992). Similar concerns were raised recently by a panel of distinguished economists asked by the federal government to evaluate the use of CV methods for valuing natural resource damages (National Oceanic and Atmospheric Administration 1993). In our view the general lack of effective budget constraints poses the most serious problem with hypothetical data. Public opinion polls routinely ask people to value certain things, but rarely do they ever focus, explicitly or implicitly, on the opportunity costs associated with obtaining these new assets. Absent any clear evidence that respondents understand the concept of opportunity cost and properly incorporate it into their responses, polls offer no useful economic information. Further, CV studies routinely suffer from an absence of effective budget constraints and little has been done in recent years to solve this problem. Kahneman and Knetsch (1992: 59) quote experienced CV practitioners who wrote in a 1983 survey of the state-of-the-art that "the summation of average CV values for public goods thus far available in the literature would exhaust the budget of the average individual." One can only imagine how much worse this comparison must be given the large number of CV studies performed since this statement was made. CV studies of environmental amenities have shown that the imposition of a budget constraint can dramatically alter expressed WTPs. In a study of the value of preventing oil spills in Alaska's Prince William Sound, Kemp and Maxwell (1992) obtained WTP estimates about 300 times smaller when they utilized a top-down disaggregation procedure than when they sought direct estimates of value. They concluded that the lower value was more plausible precisely because it required respondents to adapt to a binding budget constraint, but they averred from endorsing it as the "right" value because expressed WTPs tend to decline as the number of disaggregation levels increases and there is no obvious stopping point. Kahneman and Knetsch (1992: 61) observed a SO-fold reduction in median WTP in an experiment involving just two disaggregation steps. Where CV methods are applied to food safety problems, we believe that similarly rigorous efforts will be needed to develop realistic scenarios and effective budget constraints. It is the responsibility of CV practitioners to devise and test alternative hypotheses concerning possible biases resulting from the hypothetical nature of the transaction in question.

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

347

Clear Commodity Definition Economists asked to value commodities traded in markets look directly at such markets for data. Consumers reveal their preferences in their voluntary transactions. Thus, if the market price for doughnuts is 42 cents each, and there is no particular reason to believe that doughnut markets are rife with either externalities or market power, then the value of a doughnut to the marginal consumer is precisely 42 cents. Most consumers will value doughnuts at more than 42 cents, reflecting the consumers' surplus they reap from not having to pay a price greater than 42 cents despite a willingness to do so, but market price will be set by the marginal consumer. Most commodities traded in markets have attributes substantially more complex than doughnuts, but consumers still reveal their preferences through voluntary exchange. Consumers need not be able to quantify or even observe every attribute of a product to place an appropriate value on it. Information concerning product attributes is itself a valuable commodity, and one that suffers from certain public good problems, but even where information is available freely it cannot be costlessly internalized and processed. This means that lessthan-perfect information is in all cases optimal for individual decision making .. Privately owned environmental amenities also are traded in markets. An ocean view is a valuable real estate attribute, one that raises both its market clearing sales price and the rents which it can command. But the difficulty of quantifying an ocean view does not diminish consumers' capacity to monetize it or the ability of willing buyers and sellers to establish market-clearing prices. Sellers generate and disseminate information to seek out buyers, attempting to persuade them of the unique character, attractiveness, and perhaps exclusivity of these amenities. Potential buyers process this information, collect their own data and comparison data on possible substitutes, and evaluate market prices in the context of their own WTP. Prices rise or fall depending on the usual supply and demand conditions. Publicly owned amenities do not benefit from these market prices. Indeed, an obvious inefficiency arises insofar as publicly owned environments often have near zero prices. The National Parks suffer congestion externalities, for example, reducing both the quality of the environmental amenity and consumers' WTP to visit. Other valuation approaches, such as the travel cost and hedonic methods, must be used to estimate the value of publicly owned environments due to government's decision not to price these assets efficiently. Risks to health and safety represent another class of commodities which are clearly traded in markets but for which indirect methods must be used to estimate value. The economics literature on risk valuation is extensive and clearly demonstrates that individuals are willing to pay substantial sums to avoid risk despite the difficulty of identifying analytically rigorous measures for it and quantifying it (Viscusi 1983, Fisher et al. 1989). The task of the researcher is

348

Richard B. Beker and Richard P. Theroux

to statistically disentangle from the gamut of attributes inherent in the product or employment contract the attributes relating to risk. Alternative methods thus are available for analyzing actual behavioral data and ascertaining the reliability and validity of CV results. Given the range of methodological problems associated with CV and the depth of concern raised by many distinguished economists, researchers should seriously consider whether CV is the most appropriate tool for estimating consumers' WTP to reduce foodbome risks. 4 For any CV approach to generate potentially useful results, the survey must establish clearly the precise character of the commodity respondents are being asked to value. Thus, if the question is bow to value consumers' WTP to avoid E. coli infection, then the commodity must be defined as a specified likelihood (e.g., a 1 in 1,000 annual chance) of a well-defined adverse health outcome (e.g., hemorrhagic colitis). Clarity is necessary to minimize the extent to which respondents mistakenly provide values for unspecified alternative commodities (e.g., a 1in10,000 chance of debilitating kidney damage from hemolytic uremic syndrome). CV studies which purport to estimate consumers' WTP to reduce ambiguous risks by an unspecified amount do not yield information that is relevant for policy making where government intervention involves restricting consumer choice, mandating the expenditure of private resources, or expending public funds collected through taxation. The CV literature in the food safety area reveals uncertainty as to whether food safety risks should be portrayed in absolute or relative terms (e.g., see Lin and Milon, Chapter S). In theory, however, respondents should be indifferent to such framing questions if they have a clear understanding of the commodity they are expected to value and are sufficiently familiar with the commodity of interest. To the extent that framing matters, CV may be an inappropriate method for value elicitation inasmuch as there is no analytic basis for asserting that one particular framing approach is superior to others. In a similar vein, both the nature of the underlying risk and the commodity to be valued must be believable to survey respondents. If the problem appears to be hypothetical, then respondents may be inclined to give hypothetical answers. CV researchers often encounter zero bids from respondents who believe that the underlying problem does not actually exist or reject the proposed action as a valid response to such a problem. Because CV surveys represent hypothetical market transactions, respondents often fail to recognize their own budget constraints. High bids entail no additional opportunity cost. CV researchers must develop and implement methods to simulate real-world budget constraints. Finally, CV studies must use payment vehicles which would force respondents to bear the full costs of the benefits they receive. Increases in general taxation and similarly broad payment formulae may be acceptable in certain circumstances, but they create powerful incentives for survey respondents to express WTPs that are not founded on economic preferences and hence of limited utility in benefit-cost analysis.

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

349

Reliability and Validity in Expressed WTP Values Because there may be no market transactions providing sure evidence of voluntary exchange, the results of CV studies may be subject to a variety of alternative interpretations. It is not enough to assume that survey respondents interpret valuation questions in precisely the way they were intended. Rather, CV researchers must ensure that plausible alternative explanatory hypotheses can be safely ruled out. This process entails several critical elements. For example, researchers must identify all plausible explanations for the results they obtain and develop tests suitable for testing these alternative hypotheses; they cannot simply assume that the values respondents provide represent economic preferences. Verbal protocols must be employed to ensure that the decision processes used by respondents correspond to the economic model of behavior. Researchers must design CV surveys such that data needed to test alternative hypotheses are collected. Survey instruments must be designed, pre-tested, and refined based on the knowledge gleaned; it is not enough to assert that a particular design has been used effectively elsewhere and thus does not require similar validation. Samples must be large enough to execute the necessary hypothesis tests with sufficient statistical power; multiple alternative scenarios and payment vehicles are probably necessary. Finally, CV results must be compared with other expectations and economic data to ensure consistency between CV results and revealed preferences. Where CV results are intended to be used for policy purposes, they should be replicated by independent researchers. Suggested Checks for Comistency with the Comumer Choice Axioms of Elementary Economic Theory

Little effort has been devoted to the development and propagation of methods for testing whether CV results are consistent with revealed preferences. In this section we identify a variety of alternative hypotheses which CV researchers ought to be able to reject using generally accepted statistical methods. Each of these alternative hypotheses has been identified in the economics literature as a plausible explanation for observed CV results and as a rationale for rejecting their use in policy contexts.

Calibmti.on of Hypothetical and Actual WTP The primary use of survey methods involving economic phenomena is in market research, and survey results are routinely calibrated based on prior experience to adjust for the hypothetical nature of the exercise. Nevertheless, many products fail in real-world markets despite successful market research

350

Richard B. Belzer and Richard P. Theroux

studies, suggesting that calibration is more art than science. Marketing researchers routinely observe actual (ex post) purchase behavior that falls well short of the quantities consumers state (ex ante) they intend to purchase and have focused their efforts on developing models to explain and predict these shortfalls (Jamieson and Bass 1989). Recent work by Neill et al. (1994) begins to address this issue in the context of CV. They examined WTP responses in three settings-a CV survey, a hypothetical Vickrey auction, and a real Vickrey auction-to ascertain the extent to which CV methods could replicate a known incentive-compatible instrument and whether either hypothetical exercise could replicate actual purchase behavior. The authors found that the distribution of bids in the two hypothetical markets were not significantly different. However, median WTP in the hypothetical markets exceeded median WTP in the real auction by factors ranging from 6 to 15. These results have three important implications. First, hypothetical WTP generally exceeds actual WTP-often by substantial amounts-and cannot be assumed to represent actual WTP. Second, the difference between hypothetical and actual WTP may be independent of the nature of the simulated market used to elicit responses. Third, and perhaps most important, the use of an incentivecompatible payment instrument in a hypothetical setting does not assure results consistent with real purchase behavior. More research into the calibration of CV results clearly would be useful. However, this may conflict with the prevailing view among CV advocates that respondents' expressed WTPs properly reflect true economic values and thus do not require calibration at all. This represents an important divergence that deserves further study, because it is difficult to construct a theory under which both CV advocates and market researchers are correct. We believe that this conflict must be resolved before the results of CV studies can be treated as reliable indicators of willingness to pay suitable for public policy applications.

Law of Demand Subject as always to binding budget constraints, individuals should give consistent answers when they value single or multiple units of a commodity. That is, marginal WTP should decline in response to movement down the demand curve and increase in response to upward movement. As elementary as this notion may seem to be, CV studies do not routinely demonstrate conformity with the Law of Demand. Suppose that food safety measured as residual risk is the commodity in question. The consumer begins with q0 units of safety and would pay a price of Po for an additional unit. Once this first incremental unit has been acquired, however, the consumer's WTP declines to p' per unit. If entitled only to purchase multiple units of increased safety-that is, travel down the demand

Criteria/or Evaluating Results Obtained from Contingent Valuation Methods

351

curve-the consumer's total WTP must be less than twice the value of a single unit unless her price elasticity of demand is infinite. The opposite result should be observed for reductions in safety from the baseline. That is, after paying Po to avoid losing one unit of safety the consumer places the higher value p" on preserving the next unit. The consumer's total WTP to avoid a multiple unit loss of safety must exceed twice the value of losing a single unit. Hypothesis Tests. A proper test of conformity with the Law of Demand requires two demonstrations. First, CV respondents must show that the value they place on a subsequent unit of additional safety is no greater than (and probably less than) the value they place on the first unit. Second, CV respondents must show that the value they place on avoiding a subsequent unit loss of safety is no less than (and probably greater than) the value they place on avoiding the initial unit loss of safety. This "marginal value test" makes sense as long as the commodity can be characterized as unidimensional. Food safety risks appear to qualify because they can be readily characterized in quantitative terms based on their probability of occurrence. Goods which clearly have multiple attributes cannot be so easily subjected to this test, but nevertheless should be subjected to an analogous "total value test" in which it is demonstrated that respondents have higher WTPs for an unambiguously superior commodity. 5 Altemative Hypotheses. Some economists believe that the common absence of CV results consistent with the Law of Demand indicates the presence of "warm glow" effects, which are characterized as the moral satisfaction associated with responding affirmatively when asked to value "good" things (e.g., Diamond et al. 1992, Kahneman and Knetsch 1992). For example, a respondent may feel good about "making a contribution" or doing his "fair share" to acquire or protect a particular environmental amenity. Similarly, respondents may care deeply in an abstract sense about certain general food safety risks (e.g., pesticide residues) and embed this concern within their expressed WTP for avoiding specific food safety risks (e.g., Alar residue8 in red apples) or even unrelated risks (e.g., microbiological contamination in ground beet). Effective tests of upward and downward movement along the demand curve are essential to distinguish between these alternative hypotheses. It is the CV researcher's responsibility to test for the possible presence of "warm glow" effects and similar phenomena unrelated to individual economic preferences, for CV responses must reflect individual economic preferences for the commodity in question before they can be used in benefit-cost analysis (BCA). CV researchers can use split samples to directly test the relevant hypotheses and demonstrate conformity to the Law of Demand.

Casual Benefit-Cost Analysis Certain payment vehicles such as referenda and generalized tax increases are popular among CV practitioners because they comport with respondents'

352

Richard B. Belzer and Richard P. Theroux

personal experiences. Referenda have become so commonplace that voters are now quite familiar with this form of dichotomous choice. Similarly, CV respondents also readily understand how the provision of a certain public good may be paid for through higher taxes. Unfortunately, the advantage of familiarity is offset by the fact that respondents may be inclined to base their responses on casual BCA rather than personal preferences. Despite clear instructions to report personal valuations, some respondents apparently interpret their task as one of making a broad societal judgment about the desirability of a program or policy. This behavior may reflect widespread familiarity with voting and public opinion polls in which some element of choosing for a group seems to be implied. When such situations arise, respondents are unlikely to report the same preferences they would if the transaction actually arose in a conventional private-goods market (Buchanan 1954). A common practice among experienced survey researchers is to test survey instruments using verbal protocols. Respondents are asked to think aloud as they fill in the questionnaire. Researchers then review audio or video tapes of these sessions to learn more about the mental processes people use in deriving answers to critical valuation questions. Learning about these mental processes often is more important for effective market research than ascertaining precise value estimates. Casual BCA can be readily observed in these verbal protocols. Sensitivity to the payments of others reflects concerns about fairness and other values which are perfectly legitimate but cannot be incorporated into the benefit-cost paradigm. For example, excerpts of verbal responses in a recent CV study of ground water valuation clearly reveal that respondents provided values based on casual BCA rather than their own economic preferences (McClelland et al. 1992). Some respondents indicated WTP values which were either conditional upon others paying their "fair share" or mentally derived from considering how much revenue would be generated if everyone contributed a similar amount. The fundamental problem with casual BCA is that lay estimates of societal welfare are not equivalent to the underlying economic preferences of individuals. Results from CV studies tainted by casual BCA thus offer little insight into the underlying valuation question. Further, if BCA is to be used for societal decision making, expert analysis performed by trained economists is more likely to pass tests of validity and reliability than casual analysis performed on the spot by untrained survey respondents with limited information. Hypothesis Tests. Perhaps the best way to detect casual BCA is through a verbal protocol study executed as part of or parallel with a CV valuation study. Considerable effort must be devoted to develop hypothesis tests that are compatible with each particular CV survey, for no off-the-shelf test exists. One possible test of the hypothesis that survey responses reflect casual BCA rather than individual preferences would be to employ otherwise identical survey

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

353

instruments with radically different payment vehicles. Suppose that one payment vehicle involves a general tax increase that would be paid by all. The alternative payment vehicle could be a targeted tax or user fee aimed directly and solely at the respondent. A necessary condition for respondents to be providing values based on economic preferences is that the median WTP be the same across the two surveys. As a final note, careful efforts to define the commodity of interest or to provide extensive contextual information are generally worthwhile because they reduce sample variance resulting from uncertainty. They cannot overcome the problem of casual BCA, however. At the same time that additional information reduces respondents' uncertainty and gives them greater capacity to articulate meaningful answers, variance across answers may widen if respondents' preferences truly differ. Thus, reduced variance resulting from enhanced contextual information may reflect improved casual BCA rather than better articulation of individual economic preferences.

Altruism and Vengeance Standard welfare economics, the foundation for BCA, treats value as a personal matter independent of the gains and payments of others or the process by which decisions are made. Altruism, the expression of WTP derived from the enhanced utility of others, is a perfectly acceptable and economically rational expression of individual economic preferences with respect to private goods where the benefactor bears the full cost. However, it is not a legitimate quantity for BCA of public goods because it results in the double-counting of the benefits which are enjoyed by certain (but not all) members of the community (Milgrom 1992). That is, the incorporation of altruistic values in BCA would explicitly provide for the welfare of some individuals to be weighted more highly than others. Most economists are uncomfortable with such an approach because there are an infinite number of outcomes associated with weighting individual utilities, and economics offers no special expertise or legitimacy in the assignment of such weights. Another reason for leaving altruism out of BCA is that there is no ethical way to distinguish it from its antithesis, a construct we call "vengeance value. " This is the utility an individual obtains from imposing costs on (or denying benefits to) others. Vengeance clearly motivates behavior in private markets, and while it may be subject to opprobrium or other social sanctions when discovered it cannot be readily detected or deterred, especially when individuals quite sensibly keep their ulterior motives to themselves. Like altruism, vengeance value cannot be a legitimate element of BCA because it would implicitly reduce and could even eliminate the worth of some individuals merely due to the enmity of others. The possibility of vengeance value is particularly troublesome in CV surveys where the payment vehicle encourages respondents to think about the payments

354

Richard B. Belzer and Richard P. Theroux

which would be made by others. An individual who actually places a low value on a specified commodity may overstate his true economic preference if he believes that everyone else in the community would be compelled to pay as well and he places a high value on imposing sacrificial expenditures on others. Hypothesis Tests. A variety of tests may be devised to identify altruism (or vengeance) as an individual's underlying motive, but it is difficult to construct devices for filtering it out. One way to identify altruism may be to test otherwise identical surveys having different representation of the benefit. In one survey, only the respondent could enjoy the benefits of the program or policy of interest; in the other survey the same benefits would be distributed broadly. In the absence of altruism the median WTP of each survey sample should be the same. However, if the median WTP of the sample valuing a broadly distributed good exceeds the median WTP of the sample valuing an individually targeted good, then the former WTP estimate is tainted by altruism and should not be used for BCA. Unfortunately, we know of no way to tell whether altruism is embedded in the latter WTP estimate. 6 As before, verbal protocols may be extremely useful for detecting altruism (or vengeance). CV respondents often reveal quite legitimate and salutary concerns for their friends and neighbors. (They are unlikely to reveal enmity even if they are highly motivated by it.) Further refinement of a survey instrument may decrease the prevalence of this phenomenon, but we do not know of any way to eliminate it. CV practitioners must devote extraordinary efforts to control the problem in cases where altruism is particularly likely to arise. 7

"Absolution Value" Except for problems such as altruism (and now vengeance), economists generally do not dwell on the motives behind individuals' preferences. One such motive, which to our knowledge has received no attention in the benefits valuation literature, involves an individual's WTP to be absolved of sin. The sin in question may be personal (e.g., alcohol or drug abuse) or societal (e.g., rain forest destruction). Environmental restoration projects, such as the cleanup of the federal government's nuclear weapons testing and manufacturing facilities, appear to be strongly motivated by such "absolution values." A verbal protocol CV survey might well reveal that many respondents place a relatively high value on cleaning up these facilities because of the shame they feel for the nation's nuclear heritage rather than the demonstrable human health or environmental risks posed by such sites. Absolution value is a perfectly legitimate component of BCA under certain restrictive conditions. In particular, each individual's WTP must reflect only her own absolution. Once the absolution of others or society at large enters into the analysis, the same problems associated with altruism (and vengeance) arise.

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

355

An analogous situation could be observed in the food safety arena. Suppose that there are both naturally occurring and anthropogenic sources of a health risk of specified probability and magnitude. Suppose further that the median expressed WTP among survey respondents is significantly greater when the risk is anthropogenic than when it occurs naturally. Clearly, other factors besides risk per se have entered into the valuation exercise. It may be that respondents truly are willing to pay more to avoid risks if they are anthropogenic. However, it may also be the case that some respondents have embedded a societal absolution value into their expressed WTPs that is unrelated to their true economic preferences. It is the researchers' responsibility to ascertain these factors to ensure that if absolution value is present only the respondent and no others enjoy absolution. 8 Hypothesi.s Tests. Two otherwise identical surveys could be administered in which the specific food safety risk in question has either a natural or anthropogenic source. If CV respondents are expressing WTPs related to the probability and magnitude of the underlying risk without concern for its origin, then the median expressed WTP will not be significantly different between these two surveys. If they are significantly different, however, then the origin of the risk matters and further work is necessary to explore the respondents' valuation processes. Other evidence of absolution value may be obtained from survey questions or verbal protocols. Special ethical or moral concerns about past individual or collective actions may be revealed.

Strategi.c Behavior Such as Free-Ridi.ng A long-standing concern among economists is the possibility that survey respondents will give strategic answers to CV surveys. In particular, concern has focused on the effect of free-riding on expressed WTPs. Respondents may underbid if, for example, they fear that they will actually have to pay more than their true WTP. Alternatively, they may overbid if they expect the community average WTP will be substantially less than their own WTP. The problem of strategic behavior in general (and free-riding in particular) has been addressed at length in the CV literature. Mitchell and Carson (1989: 133), for example, reject the traditional Samuelsonian formulation that free. riding is highly rational and inevitable in favor of an alternative view that it is both irrational and avoidable. They find support for this view in the experimental economics literature and in the development of a variety of incentive-compatible demand revelation devices, such as voting schemes and auctions. 9 Hypothesi.s Tests. A broad review of the problems strategic behavior poses for CV analyses is beyond the scope of this chapter. However, Mitchell and Carson (1989: 165-168) suggest three types of hypothesis tests which represent minimum efforts to detect strategic behavior in CV surveys. Further, they

356

Richard B. Belzer and Richard P. Theroux

recommend procedures to be applied to deal with outliers symptomatic of strategic overbidding, such as the use of "trimmed means" (p. 369). The problem with all such devices, of course, is that researchers do not know a priori the precise level of expressed WTP that is "too high" to be plausible. Nor is there any acceptable methodology for upwardly adjusting zero (or just very low) values to account for possibly strategic underbids. Any technique used to remove implausible data and thus reduce bias runs the risk of introducing new biases by mistakenly altering strategic bids incorrectly, as well as altering bids which appeared suspicious but in fact reflected true WTP.

Payment Vehicles and Value Elicitation Fonnats CV surveys use a variety of payment vehicles to obtain respondents' expressed WTP, including bidding games, payment cards, open-ended elicitation, and dichotomous choice referenda. Bidding games use an iterative process like an auction to ascertain respondents' highest WTP. In payment card surveys, respondents are asked to select the value on the card which most closely approximates their maximum WTP. Open-ended elicitation formats ask respondents to provide their maximum WTPs without such prior cues. Finally, referenda approaches ask respondents to vote for or against a single proposal based on an explicit dollar value price. Many experienced CV practitioners have clear preferences for one or another of these formats, but there is no consensus in the CV literature in favor of any of them. Bidding games require the provision of a starting point which tends to impart a downward (upward) bias to the valuation exercise among respondents whose true WTP is above (below) the starting point. Open-ended value elicitation suffers from high nonresponse rates believed to arise from respondents' inability to specify prices without prior cues. The payment card approach helps solve this problem by providing such cues, but it also may cause starting-point bias similar to that observed in bidding games. The dichotomous choice referendum approach has the advantage of better approximating markets that consumers are familiar with in which prices appear to be set by the seller and not generally negotiable. It also reduces the likelihood of strategic behavior because respondents have less capacity to exaggerate their expressed WTPs. Unfortunately, dichotomous choice referenda also suffer from starting-point bias, they require a substantial increase in the number of surveys which must be administered, and assumptions must be made to parameterize the data to obtain estimates of the median WTP. CV researchers also must confront the possibility that the expressions of WTP provided by survey respondents represent wild guesses or random values. This problem is exacerbated in CV designs in which respondents are pressured to provide answers. When a substantial percentage of respondents prefers not to assign a value-perhaps because the commodity in question is

Criteria/or Evaluating Results Obtained from Contingent Valuation Methods

357

unfamiliar-then further efforts to coax value elicitation increases the chance that the values provided will be random numbers, guesses, or otherwise poorly related to their true WTP. Hypothesis Tests. At a minimum, CV researchers should test for the types of biases which can be expected to arise based on the value elicitation method they have employed. For example, if payment cards are used then it is necessary to show that the values included on these cards have not unwittingly created anchors. Two otherwise identical payment-card instruments could be administered where the value ranges on alternative payment cards overlap asymmetrically. Where both medians fall in the overlapping range, the median expressed WTPs should not be different when subjected to an appropriate statistical test. 10 Further, CV researchers should search for the presence of guesses and random numbers. A relatively weak test would involve comparing the median expressed WTP from a payment card instrument lacking a "don't know" option with another instrument in which this alternative is not available. A significant difference in expressed WTP suggests a systematic difference in the way these different groups of respondents processed survey information and formulated their answers.

Framing Effects Many psychologists and decision scientists have extensively criticiz.ed economists' reliance on theories of strict rationality. The fundamental dispute has been expressed recently as a debate between the perspectives of "articulated" versus "basic" values (Fischhoff 1991). The articulated values perspective assumes that individual responses to survey questions are rational and it attributes apparent contradictions to (sometimes imperceptible or unknown) differences in problem or question formulation. In contrast, the basic values perspective holds that individuals lack well-differentiated values except for the most familiar and noncontroversial evaluation efforts. CV methods are based on the articulated values paradigm; unsurprisingly, many of its critics subscribe to the basic values perspective. This debate is well beyond the scope of this chapter. Nevertheless, the debate has revealed important issues surrounding survey methods such as CV which deserve to be addressed before CV estimates make the giant leap toward inclusion in BCA. For example, if the values respondents provide to early survey questions influence how they interpret and respond to later questions, individual respondents may force their answers to be internally consistent but demonstrate incoherence across respondents (Diamond 1992). More importantly, it is then unclear which values are the appropriate ones to rely upon.

358

Richard B. Belzer and Richard P. Theroux

Problems Using CV Results in Federal Policy Making It is important to reiterate here that, given the breadth of methodological problems involved, CV surveys should not be used where there are market transactions which could provide the data necessary to perform traditional economic analysis. Proper CV survey design involves multiple alternative scenarios, multiple payment vehicles, verbal protocols, and a host of additional niggles likely to make the enterprise very expensive. Even if CV researchers can successfully solve problems such as those described briefly above, another set of issues arises regarding how CV results can be transferred to public policy applications.

Replicati.on Because of the range of methodological problems involved, CV studies that pass the necessary tests should be replicated, preferably by different researchers. The purpose of replication is to increase the likelihood that the initial study was not merely serendipitous. Government guidelines concerning quantitative risk assessment methodologies assert similar replication requirements as a means of deterring reliance on chance results (Office of Science and Technology Policy 1985, Environmental Protection Agency 1986). At a minimum, results obtained in the replication should be consistent with those obtained initially. If the median expressed WTP across essentially identical studies is significantly different, then it is unclear which study should be deemed authoritative. Researchers may be tempted to use the study yielding the lowest median WTP as a "conservative" estimate of potential benefits. This temptation should be avoided, for it creates a variety of perverse incentives among both researchers and government officials. The objective always should be to develop unbiased estimates of baseline conditions and the costs and benefits of various intervention alternatives (Office of Management and Budget 1990a, 1990d). CV results must be exceptionally robust where they could influence important public policy interventions. They must be able to withstand significant design perturbations without sybstantial deviation in median expressed WTPs. Again, if CV results are highly sensitive to survey design there is no clear basis for distinguishing among alternative designs.

I>iscrepancies Between Commodities in CV Studies and Proposed Government lnterventi.on A carefully constructed CV study may involve important simplifications and abstractions which are inconsistent with the manner in which government would

Criteria for Evaluating Results Obtained from Contingent Valuation Methods

359

actually implement a food safety program, policy, or regulation. However, it is not obvious how benefit estimates derived from a successful but narrowly constructed CV study can be properly extrapolated to a government program or regulation whose characteristics are substantially different from those analyzed. For example, a CV study in which respondents are asked to value reduced risks from eating raw oysters may not be applicable to a government program of general seafood inspection.

Perceived Versus Expert Estimates of Risk Where individuals make their own decisions concerning risk-taking and riskavoidance, it is perfectly acceptable to rely on risks that are perceived even if they are substantially different from expert risk estimates. Individuals can expend resources to obtain better information, update prior risk estimates, and make different decisions. Ultimately, however, they alone bear the cost of their own errors. Where governments make these decisions, however, they bear an obligation to use expert estimates of risk consistent with their fiduciary responsibility to properly care for other people's money. Failing to do so will waste resources and result in a suboptimal risk management portfolio: fewer health risks will be prevented for any given expenditure of societal resources. Further, once the government signals its intent to base policy decisions on perceived risks, it rewards those who strategically exaggerate their risk perceptions. Ignorance and intransigence become potential assets rather than liabilities.

Conclusion Contingent valuation represents an important technique for attempting to estimate the value of commodities which are not traded in markets. Generally, there are market transactions with risk attributes available to study. Thus, CV may be useful where segments of the market demand curve do not exist and cannot be approximated by examining substitutes. A decision to use CV opens up a large set of new methodological problems. Experienced CV practitioners believe that these problems can be overcome. CV critics argue that the problems are insurmountable and the method should be abandoned. Food safety researchers interested in CV should be wary of trying to finesse this conflict. OMB's guidance concerning the use of CV-based estimates for benefit estimation is largely performance oriented. While it avoids taking any position on CV, it does impose an extraordinary burden on any methodology that does not rely on real-world behavioral data. Food safety researchers and government agencies interested in CV thus are well advised to explore this method with considerable care.

360

Richard B. Belzer and Richard P. Theroux

We do not believe that we are imposing an unfair or discriminatory burden on CV studies relative to traditional methods based on observed behavior. In particular, we expect that economic studies based on observed market behavior also demonstrate that consumers clearly understand the commodities they are purchasing; that their preferences conform to the Law of Demand; that they do not display casual benefit-cost analysis apart from that which is consistent with individual utility maximization; that if they display altruism or vengeance they bear as individual consumers the full cost of these motives; that they cannot free-ride; and that no calibration is necessary to render observed transactions consistent with elementary axioms of economic theory. Unless their express purpose is to demonstrate the existence of a significant market failure, we would reject for purposes of benefit-cost analysis any study based on market observations that failed any one of these requirements. By elucidating criteria and suggested hypothesis tests, it is our hope that practitioners of the new art of contingent valuation will be able to achieve the same level of legitimacy and respect commonly accorded to economic analysis derived from market observations.

Notes 1. The views expressed in this paper are those of the authors and do not necessarily represent the views of the Office of Management and Budget. 2. Executive Order No. 12866 was signed by President Clinton on September 30, 1993 (58 Federal Register 51735-51744). It superseded Executive Order No. 12291 signed by President Reagan on February 17, 1981 (46 Federal Register 13193-13198). 3. This document has been reprinted in each annual Regulatory Program published since 1990. 4. In the remainder of this chapter we assume that a decision to proceed with a CV approach has already been made despite these warnings. 5. CV practitioners may be tempted to purposefully characterize the food safety commodity in question as multidimensional because of the difficulty of satisfying both elements of the marginal value test. This temptation should be resisted. The presence of multiple dimensions increases the likelihood that respondents will fail to adequately understand the commodity, thus rendering the valuation exercise meaningless. In addition, reviewers and "consumers" of CV studies are increasingly likely to question a survey that has been designed in such a fashion that consistency checks are infeasible. 6. A similar experiment could be performed to test for the presence of vengeance value by altering the payment vehicle in analogous ways. 7. Examples may include situations where programs or policies are intended to benefit children, the elderly, the infirm, or the disabled. 8. Note that if absolution value is present the nature of the commodity is radically changed. Absent both private markets for absolution and governmental mechanisms to provide it, estimates of WfP infected by absolution value are unlikely to have any practical use.

Criteriafor Evaluating Results Obtained from Contingent Valuation Methods

361

9. AB indicated earlier in the discussion of calibration, the use of an incentivecompatible voting scheme or auction is not sufficient to overcome the hypothetical nature of the CV exercise. 10. If one of the medians is not in the overlapping range, then the median WfP for the other survey should be close to the relevant extreme value in its range.

References Bishop, Richard C. and Thomas A. Heberlein. 1986. Does Contingent Valuation Work? In Valuing Environmental Goods: An Assessmentofthe Contingent Valuation Method, ed. Cummings et al., Chapter 9. Savage, MD: Rowman & Littlefield. Buchanan, James. 1954. Individual Choice in Voting and the Market. Journal of Political Economy 62:334-343. Cambridge Economics, Inc. 1992. Contingent Valuation: A Critical Assessment. Cambridge, MA: Cambridge Economics. Cummings, Ronald G., David S. Brookshire, and William D. Schultze, eds. 1986. Valuing Environmental Goods: An Assessment of the Contingent Valuation Method. Savage, MD: Rowman & Littlefield. Diamond, Peter. 1992. Comment on "Methods of Measuring Non-Use Values: A Contingent Valuation Study of Groundwater Cleanup," by Gary H. McClelland, William D. Schultze, Jeffrey K. Lazo, Donald M. Waldman, James K. Doyle, Steven R. Elliot, and Julie R. Irwin. Unpublished comment submitted to Environmental Economics Advisory Committee of the Science Advisory Board, U.S. Environmental Protection Agency (December 14). Diamond, Peter A., Jerry A. Hausman, Gregory K. Leonard, and Mike A. Denning. 1992. Does Contingent Valuation Measure Preferences? Experimental Evidence. In Contingent Valuation: A Critical Assessment, Tab 3. Cambridge, MA: Cambridge Economics. Environmental Protection Agency. 1986. Guidelines for Carcinogen Risk Assessment. Federal Register 51(46):33991ff (September 24). Environmental Protection Agency. 1993. Clean Water and the American Economy, October 19-21, 1992,· Proceedings: Ground Water (Volume 2). Washington, D.C.: U.S. EPA, Office of Water (EPA 800-R-93-00lb). Fischhoff, Baruch. 1991. Value Elicitation: Is There Anything in There? American Psychologist 46(8):835-847. Fisher, Ann, Lauraine G. Chestnut, and Daniel M. Violette. 1989. The Value of Reducing Risks of Death: A Note on New Evidence. Journal ofPolicy Analysis and Management 8(1):88-100. Freeman, A. Myrick. 1979. The Benefits of Environmental Improvement. Baltimore, MD: Johns Hopkins Press. Jamieson, Linda F. and Frank M. Bass. 1989. Adjusting Stated Intention Measures to Predict Trial Purchase of New Products: A Comparison of Models and Methods. Journal of Marketing Research 26:336-345.

362

Richard B. Belzer and Richard P. Theroux

.Kahneman, Daniel and Jack L. Knetsch. 1992. Valuing Public Goods: The Purchase of Moral Satisfaction. Journal ofEnvironmental Economics and Management 22:5770. Kemp, Michael A. and Christopher Maxwell. 1992. Exploring a Budget Constraint for Contingent Valuation Estimates. In Contingent Valuation: A Critical Assessment, Tab 4. Cambridge, MA: Cambridge Economics. McClelland, Gary H., William D. Schultze, Jeffrey K. Lazo, Donald M. Waldman, James K. Doyle, Steven R. Elliott, and Julie R. Irwin. 1992. Methods for

Measuring Non-Use Values: A Contingent Valuation Study of Groundwater Cleanup.

Boulder, CO: Center for Economic Analysis. Milgrom, Paul. 1992. Is Sympathy an Economic Value? Philosophy, Economics, and the Contingent Value Method. In Contingent Valuation: A Critical Assessment, Tab 11. Cambridge, MA: Cambridge Economics. Mitchell, Robert C. and Richard T. Carson. 1989. Using Surveys to Value Public Goods. Washington, D.C.: Resources for the Future. National Oceanic and Atmospheric Administration. 1993. Natural Resource Damage Assessments Under the Oil Pollution Act of 1990. Federal Register 58(10):46014614 (January 15). Neill, Helen R., Ronald G. Cummings, Philip T Ganderton, Glenn W. Harrison, and Thomas McGuckin. 1994. Hypothetical Surveys and Real Economic Commitments. Land Economics 70(2):145-154. Office of Management and Budget. 1990a. Current Regulatory Issues in Risk Assessment and Risk Management. In Regulatory Program of the United States Government: April 1, 1990-March 31, 1991, 13-26. Washington D.C: U.S. Government Printing Office. Office of Management and Budget. 1990b. Information as an Alternative Regulatory Strategy. In Regulatory Program of the United States Government: Aprill, 1990March 31, 1991, 26-32. Washington D.C.: U.S. Government Printing Office. Office of Management and Budget. 1990c. Regulatory Impact Analysis Guidance: Discussion of Comments. In Regulatory Program ofthe United States Government: April 1, 1990-March 31, 1991, 32-41. Washington D.C.: U.S. Government Printing Office. Office of Management and Budget. 1990d. Regulatory Impact Analysis Guidance. In

Regulatory Program of the United States Government: Aprill, 1990-March 31, 1991, 653-666 (AppendixV). Washington, D.C.: U.S. Government Printing Office.

Office of Science and Technology Policy. 1985. Chemical Carcinogens: A Review of the Science and Its Associated Principles. Federal Register 50(50):10374ff (March 14). Viscusi, W. Kip. 1983. Risk by Choice. Cambridge, MA: Harvard University Press.

PART FOUR

Inputs to Valuation Studies

17 Determining Foodbome Illness in the United States: A Step Toward Valuation Thomas E. Steahr1

Underlying the attempt to value foodbome illness in the United States is the belief there is consensus on what is included in this category. In fact, such a consensus does not exist. Sources of problems include the assumption that foodbome illness is distinguishable from other types of illness, e.g., that pathogens using food or water as a transmission vehicle to human populations may be identified in a separate category. A second source is the assumption that a list of foodbome illness exists and provides the basis for a count of persons with these diseases. A third source of the lack of consensus is that no single data base contains all cases of foodbome illness. The intent of this chapter is to identify illness which is most likely caused by foodbome pathogens, to construct a list of these diseases using the International Classification of Disease, 9th Revision codes, and to examine several national data sets for the frequency of foodbome illness and the demographic characteristics of persons with these diseases. This approach will result in a more comprehensive view of foodbome illness from which valuation research may proceed. One measurement of the prevalence of foodbome illness is provided by the Centers for Disease Control and Prevention. The surveillance system is based on reported outbreaks of foodbome illness. An outbreak is defined as two or more persons experiencing a similar illness after ingestion of a common food. The exceptions are botulism and chemical poisoning where one case is defined as an outbreak. While this surveillance system provides valuable information on the incidence of foodbome disease, it is limited in several ways. With the two exceptions above, all single cases of persons becoming ill due to foodbome pathogens do not qualify as outbreaks and are not reported as such. In addition, if two or more persons are involved, the original report must come from state and local health departments on standard forms. This means that reporting 365

366

Thomas E. Steahr

outbreaks depends on patient and physician recognition, motivation to contact the local health department, the resources of laboratories to identify the pathogen, and other factors. It has been suggested that only a small fraction of foodbome disease outbreaks are reported to the Centers for Disease Control and Prevention (Bean and Griffin 1990: 85). For example, small outbreaks of infections involving mild illness from ingesting food at family gatherings, workplace picnics, and similar social events are less likely to be reported than outbreaks involving hospitaliz.ations. Outbreaks caused by pathogens difficult to detect are also underreported (Bean and Griffin 1990). Outbreak data also underrepresent the prevalence of foodbome illness in the United States because the types of conditions receiving the most research attention from bacteriology during the past decades have been acute conditions resulting from gastroenteritis-causing pathogens. Specific human pathogens, such as Salmonella spp, Campylobacter spp, and Shigella spp, have been clearly linked to foodbome illness outbreaks (Bean et al. 1990, Bean and Griffin 1990). More recently, chronic conditions which have a longer period of debility and have more complex etiology are becoming directly or indirectly connected with foodbome pathogens. For example, reactive arthritis may be activated by Salmonella spp (among other pathogens); diarrheal experiences during foodbome illness may increase overall morbidity by reducing nutritional and immunity levels; and the heart diseases of endocarditis and myocarditis have been associated with foodbome pathogens (Archer and Young 1988: 387-390). These and other chronic health conditions suggest that the consequences and health costs of foodbome illness extend well beyond the immediate hospital costs to treat acute events. Microbiologists have established that foodbome disease encompasses a complex pattern ranging from a passive mode in which pathogens are transmitted to people via the eating of contaminated food acting as a simple vehicle, to an active mode in which the food supports growth of the pathogens so they replicate on food prior to eating. Moreover, foodbome disease is included within enteric or diarrheal disease and this ranks second only to respiratory diseases in prevalence in the United States. One estimate is that of all enteric disease, foodbome disease constitutes one-third of the total cases annually, or over 80 million cases (Archer and Young 1988: 378). This level of disease may be due primarily to unsanitary methods of producing food, improper methods of handling and preserving food, and unhygienic customs which allow persons to transmit their diseases via food. The complex etiology of food contamination has long been recognized (Bryan 1982). While research at the microbial level is required to understand the etiology of specific diseases, a broader approach is necessary if the incidence and prevalence of foodbome illness in human populations is to be understood. It is recognized that the reported data on foodbome illness are only a small part of the actual number of events that take place. There are a number of cultural,

Determining Foodborne Illness in the United States

367

social, economic, and administrative "filters" through which illness events must pass before becoming part of the official surveillance system. Cultural factors impacting on illness events include preference for uncooked meat or fish, or for large pieces of meat that prevent cooking from killing all bacteria, or for aging food without refrigeration. Social factors include attaching a meaning to the physical experience of pain which links it to food ingestion and then seeking medical care for this condition, rather than doing nothing, or self-care only, or care from relatives or friends. Research on these factors reveals a great deal of individual and cultural variation (Foster and Kaferstein 1985, Vargas 1990, Strain 1989). Characteristics such as age, gender, education, marital status, and health status are related to attaching meaning to pain and seeking medical health care (Kaplan et al. 1987, Swedlund and Armelagos 1990, Wolinsky et al. 1989). After such health care is sought, administrative conditions must be met to include the event in the official statistics. For example, after a physician is visited, a specimen must be taken and the organism correctly identified. This finding should be reported to the health department which forwards the data to the Centers for Disease Control and Prevention (Chalker and Blaser 1988). Since some of the illness events fall out of the system at various stages, the official statistics on foodbome illness are an underestimate of actual incidence or prevalence. Moreover, relatively little is known about the demographic and geographic patterns of persons with specific types of foodbome illness. A basic thesis of this chapter is that scientific knowledge of foodbome illness should be expanded to include the analysis of ill persons, as well as the analysis of the pathogen and how to eliminate its transmittal. Valuation research concerned with the economic impact of foodbome illness should recognize which high risk groups suffer most and what corrective action is needed to reduce this type of illness among young children, the elderly, persons with impaired immunity systems, etc. This information is required to establish trends and to document improvement following policy changes intended to ameliorate suffering from foodbome illness.

Available Data for Determining Foodbome Illness The preceding discussion suggests that surveillance statistics on foodbome illness in the United States are only the tip of the iceberg in terms of actual numbers of illness events. In developing countries of the world foodbome illness is recogni7.ed as a widespread problem and estimates of unreported cases approach 90 percent in nonindustriali7.ed countries (Abdussalam and Grossklaus 1991). While levels of unreported illness are lower in developed nations, there are still many cases that are not in the surveillance system. At the broadest level, persons become ill from eating a foodbome pathogen, do not seek medical treatment (or are not taken to a physician if they are children), and then return

368

Thomas E. Steahr

to normal functions after symptoms disappear. Measuring the frequency of these illness events requires special purpose surveys with carefully designed questionnaires. While some survey research of this type has been accomplished, significantly more is necessary to estimate the prevalence of nonreported foodbome infection. Beyond the initial level of unreported cases, there are official statistics which provide three sources of data on foodbome illness. One important indicator of the extent of these events is the number of persons who see a physician and are diagnosed as suffering from a foodbome pathogen. Fortunately, the National Center for Health Statistics, Public Health Service, Centers for Disease Control and Prevention, publishes such a data set. The National Ambulatory Medical Care Survey of 1990 provides data from samples of patient records selected from a national sample of office-based physicians. This sample describes the use of ambulatory medical care in the United States. In 1990 there were approximately 43,469 patient records provided by 1,684 physicians participating in the survey. Only office visits of nonfederally employed physicians classified by the American Medical Association or the American Osteopathic Association as "office-based patient care" were included. Specialties of anesthesiology, pathology and radiology were excluded. The random sample of 43,469 patient records represents the total of over 704 million physician office visits by all patients in the United States in 1990. The use of this data set to estimate the frequency of foodbome illness and determine selected demographic characteristics of the patient are discussed below. A second indicator of the prevalence of foodbome illness is the number of patients treated in hospitals. This may be an indicator of a more serious threat to health than physician office visits. These data exist in the National Hospital Discharge Survey of 1990, also collected by the National Center for Health Statistics, Public Health Service, Centers for Disease Control and Prevention (Graves 1992). This survey began in 1988 and is taken annually. In 1990, the survey covered discharges fromnoninstitutional hospitals located in all states and the District of Columbia (excluded are federal, military, and Veterans Administration hospitals). Only short-stay hospitals (average length of stay less than 30 days) or children's hospitals responded to the survey. The hospital discharge certificates for sampled patients are examined for foodbome illness as diagnosed by the hospital. The socio-demographic characteristics of those patients are described below. This random sample of hospital discharge certificates includes a weighting value for each certificate in terms of characteristics of the patient, such as age, gender, race, type of illness, etc. These weights are summed for the estimate of the actual number of discharge certificates that would have been counted if a complete enumeration of all patients was taken. In other words, the data presented here are the summations of weighted sample values provided by the Centers for Disease Control and Prevention from the original computer tapes.

Determining Foodborne Illness in the United States

369

A third and most serious indicator of the prevalence of foodborne disease is the level of mortality caused by specific foodborne pathogens. The most recent published data are for 1986 and are contained in the National Mortality Followback Survey. This was also collected by the National Center for Health Statistics, Public Health Service, Centers for Disease Control and Prevention. The data are a national sample of approximately 1 percent of the United States resident deaths of persons 25 years and older. Each state is represented in the sample (except Oregon) and information on 18,733 deaths was taken from four sets of data: (1) information from the death certificate, (2) an informant questionnaire, (3) a nursing home, hospital, or hospice questionnaire, and (4) a facility abstract record. The final record on each death contained all of this information and socio-demographic characteristics of persons deceased with mention of foodborne pathogens. Data for these are presented below.

Determination of Foodborne Din~ As noted earlier, a list of International Classification of Disease codes identifying only foodborne illness does not exist. Various diseases may be thought of as forming a continuum from primarily foodborne to rarely foodborne. The problem is to place each code on this continuum so that accurate counts of foodborne illness may be derived. Without this step, valuation of the economic costs and impacts of those diseases are hindered by the two problems of not knowing basic frequencies and not knowing basic demographic characteristics of patients suffering from foodborne illness. The classification used here is the International Classification of Disease, 4th Edition, Clinical Modification (ICD-9-CM) (Karaffa 1992). These codes are used to provide the link between each of the three data sets described above. An attempt to construct such a list was made based on careful reading of texts on foodborne bacteria (Doyle 1989, Cliver 1990, Riemann and Bryan 1979), by cross-checking the Merck Manual of Diagnosis and Therapy (Berkow and Fletcher 1992) for a given disease, and by reading the relevant articles in the health literature. It quickly became apparent that persons diagnosed with a given ICD-9 coded disease may or may not have contracted that illness via a foodborne or waterborne vehicle; only case by case analysis could determine a source. Direct contamination through open skin lesion is a frequent method of infection. It is not possible to determine the number of cases of a given ICD-9 disease that are caused by a foodborne vehicle without a case-specific analysis. An attempt to solve this problem was made by creating two lists of ICD-9 codes: List A and List B. List A is intended to include only those ICD-9 codes for diseases most likely the result of food- or waterborne pathogens. List B is intended to include those ICD-9 codes for diseases which may result from nonfoodborne causes but which also include an unknown number of cases caused

370

Thomas E. Steahr

by foodbome infection. Because the relative proportions are unknown, analysis is not performed on List B. It should be noted that the ICD-9-CM codes included on List A and List B are approximations and not a final division without potential change. Indeed, it is likely some conditions might be dropped from List A while other conditions added to List A from List B. New conditions not presently on List B may also be included on List A.

Foodborne Illness, Ust A The National Hospital Discharge Survey for 1990 is used to establish List A. Table 17 .1 presents data on patients discharged from short-stay hospitals by category of foodbome illness (List A) for the United States in 1990. The hospital discharge certificate for each inpatient contained seven lines on which a diagnosis might be recorded. The first line contained the principal diagnosis for each case and lines 2-7 contained related conditions for each patient. Table 17 .1 presents these separately so that the frequency of foodbome illness as the principal diagnosis may be determined. Based on numerical order of the ICD-9CM codes, cholera is the first category. While cholera is endemic in many nations, there were only 92 patients discharged from hospitals in the United States in 1990 with it. The etiological agent is Vibrio cholera and may be ingested by eating improperly cooked crabs, raw oysters, mussels, or clams (Cliver 1990: 243). Cholera is also spread by drinking water or eating food contaminated by the excrement of persons with the infection (Berkow and Fletcher 1992: 110). Typhoid and paratyphoid fevers (codes 002.0-002.9) are caused by Salmonella typhi (Cliver 1990: 186) which is also spread by either direct human fecal contamination of food or water or by indirect contamination by a food handler who is infected and spreads S. typhi to food. Foods that have been involved in the transmission of this disease include raw milk, shellfish, and raw salads (Cliver 1990: 187). In the United States, 1,882 hospital patients were diagnosed as having typhoid fever, with an additional 174 patients with paratyphoid fever A. Third on Table 17.1 is Other Salmonella Infection(ICD-9-CM code 003.0003.9), which accounted for 17,984 patients during 1990. The major source of infection is the ingestion of contaminated food and water (Cliver 1990: 189). Foods such as poultry, eggs, beef, and pork have been carriers. Direct personto-person contact with individuals recovering from salmonellosis may result in infection since they shed the organism in their feces for weeks to months (Cliver 1990: 190). Unsanitary behavior by these persons can spread the disease. In addition, contact with infected animals may spread Salmonella, such as was the case with pet turtles. Of the total number of patients with Salmonella infection, the majority, 11,490 or 63.9 percent, are diagnosed with gastroenteritis.

371

TABLE 17 .1 Patients Discharged from Hospitals by Category of Foodbome Disease, List A, United States, 1990 Final Diagnosis Entered on Certificate as: Disease Category Cholera Vibrio cholera Vibrio cholera el tor Cholera, unspecified

ICD-9 Code

Principal1

I

Additionalb

Total Mentions

001 001.0 001.1 001.9

92 0 0 92

0 0 0 0

92 0 0 92

002 002.0 002.1 002.2 002.3

2,0S6 1,882 174 0 0

0 0 0 0 0

2,0S6 1,882 174 0 0

002.9

0

0

0

Other Salmonella Infection Gastroenteritis Salmonella septicemia Localized salmonella Other specified salmonella Salmonella, unspecified

003 003.0 003.1 003.2

12,237 9,404 1,909

S,141

so

2,086 1,S70 0

17,984 11,490 3,479

003.8 003.9

293 S81

1,S17

Shigellosis Shigella dysenteriae Shigella tlexneri Shigella boydii Shigella sonnei Other specified shigella Shigellosis, unspecified

004 004.0 004.1 004.2 004.3 004.8 004.9

Other Food Poisoning Staphylococcal Botulism Clostridium perfringens

oos oos.o

Typhoid and ParaTyphoid Fevers Typhoid fever Paratyphoid fever A Paratyphoid fever B Paratyphoid fever C Paratyphoid fever, unspecified

OOS.1 OOS.2

so

S14

867 2,098

1,440 220 327 0 268 232 393

673 lS 49 0 76 0

2,113

S,121 308 494 0

993 0

6,114 308 729 0

S33

23S

0

23S

376 0 344 232 926

(continues)

372 TABLE 17 .1 (continued)

Disease Category Other clostridia Vibrio parahaemolyticus Other due to Bacillus cereus Food poisoning, unspecified Amebiasis Acute amebiasis, with abscess Chronic amebiasis, no mention of abscess Amebic nondysenteric colitis Amebic liver abscess Amebic lung abscess Amebic brain abscess Amebic skin ulceration Amebic infection elsewhere Amebiasis, unspecified Other Protozoal Intestinal Diseases Balantidiasis (B. coli) Giardiasis (Lambliasis) Coccidiosis (Isosporiasis) Intestinal Trichomoniasis Other protozoal intestinal diseases Unspecified protozoal intestinal diseases Intestinal Infections Due to Other Organisms (excludes food poisoning by these organisms (005.0-.9))

ICD-9 Code

Final Diagnosis Entered on Certificate as: Principal1

I

Additionalb

Total Mentions

005.3 005.4

0 0

0 0

0 0

005.8

52

0

52

005.9

4,267

758

5,025

006

1,011

330

1,341

006.0

89

231

320

006.1

0

0

0

006.2 006.3 006.4 006.5 006.6

0 922 0 0 0

0 0 0 0 0

0 922 0 0 0

006.8 006.9

0 0

46 53

46 53

007 007.0 007.1 007.2 007.3

1,927 0 1,671 256 0

2,937 0 1,296 1,416 225

4,864 0 2,967 1,672 225

007.8

0

0

0

007.9

0

0

0

101,978

74,304

176,282

008

(continues)

373 TABLE 17 .1 (continued)

Final Diagnosis Entered on Certificate as: Disease Category Escherichia coli Paracolon bacilli Aerobacter aerogenous Proteus Other specified bacteria Bacterial enteritis, unspecified Bacterial enteritis, specified Other organisms, NEC Ill-Defined Intestinal Infections Infectious colitis, etc. Presumed infectious colitis, etc. Infectious diarrhea Diarrhea presumed infectious origin Listeriosis Other Diseases Due to Viruses Viral Hepatitis Viral Hepatitis A, with hepatic coma Viral Hepatitis A, without hepatic coma Unspecified viral hepatitis without mention of hepatic coma Cysticercosis

Total Mentions

ICD-9 Code

Principal'

Additionalb

008.0 008.1 008.2 008.3 008.4

1,686 0 0 0 20,955

572 0 14 205 31,073

2,258 0 14 205 52,028

008.5

2,800

1,251

4,051

008.6 008.8

14,494 62,043

8,533 32,656

23,027 94,699

009 009.0

15,982 6,850

12,321 5,164

28,303 12,014

009.1 009.2

6,191 2,460

2,174 4,892

8,365 7,352

009.3

481

91

572

027.0

796

452

1,248

7,523

5,287

12,810

070.0

0

0

0

070.1

4,407

2,236

6,643

070.9

3,116

3,051

6,167

123.1

612

830

1,442

070.

(continues)

Thomas E. Steahr

374

TABLE 17.1 (continued) Final Diagnosis Entered on Certificate as: Disease Category Trichinosis

ICD-9 Code 124

Principal8

I

Additionalb

Total Mentions

0

42

42

Other Noninfectious Gastroenteritis and Colitis Unspecified Gastroenteritis and Colitis

558.9

294,818

267,229

562,047

Toxic Effect of Noxious Substances Eaten as food Fish and shellfish Mushrooms Berries and other plants Other specified Other unspecified

988 988.0 988.1 988.2 988.8 988.9

308 42 109 142 15 0

44 44 0 0 0 0

352 86 109 142 15 0

0

0

0

445,901

371,189

817,090

Toxic Effect of Other Substances Aflatoxin and other mycotoxin (food contaminate)

Totals

558

989 989.7

Source: National Hospital Discharge Survey, 1990, United States National Center for Health Statistics, Hyattsville, MD. Note: Total mentions in this table may not agree exactly with total mentions of foodbome illness on subsequent tables due to some discharge certificates reporting more than one ICD-9 code for foodbome disease. While some ICD-9 codes have no cases recorded in the 1990 data, other years may reveal foodbome illness from those diseases. 8 Principal diagnosis on hospital discharge certificates appears on the first line. bAdditional diagnosis on hospital discharge certificates appears on lines 2-7.

Shigellosis (ICD-9-CM codes 004.0-004.9) is an acute infection of the bowel caused by Shigella organisms resulting in dysentery. A direct source of infection is feces of infected persons via the fecal-oral route (Berkow and

Determining Foodborne Illness in the United States

375

Fletcher 1992: 106). Of course, the amount of feces necessary to contaminate food is very small, as microbes may reside under fingernails, on body hairs, or on the surface of unwashed skin in visually undetectable amounts. Food handlers who are infected and do not wash their hands after defecation and then handle food may pass the organism on (Doyle 1989: 207). Foods often involved are salads with potatoes, chicken, tuna, raw oysters, beans, and hamburgers (Doyle 1989: 208). During 1990 in the United States, 2,113 patients were discharged from hospitals after treatment for shigellosis. The next category is a collection of diseases placed into Other Food Poisoning (ICD-9-CM codes 005.0-005.9). There were 6,114 patients discharged from hospitals in the United States treated for these diseases. Staphylococcal food poisoning (code 005.0) is caused by toxins produced by enterotoxins, the most important of which is the Staphylococcus aureus species (Berkow and Fletcher 1992: 464). Common foods involved include raw milk, cooked meat, pork, packaged sliced bacon, and pastry products (Berkow and Fletcher 1992: 474-476). In 1990 there were 308 patients discharged from hospitals treated for staphylococcal infections. Foodbome Botulism (ICD-9-CM code 005.1) is caused by Clostridium botulinum toxin eaten in contaminated food (Berkow and Fletcher 1992: 817). Home-canned food is a common source but commercial foods may also be involved, such as vegetables, fish, beef, pork, and poultry (Berkow and Fletcher 1992: 817). There were 729 patients discharged from hospitals in the United States after treatment for botulism. The category with the most cases is unspecified food poisoning, with 5,025 or 82.2 percent of all discharged patients in the category. Amebiasis (ICD-9-CM codes 006.0-006. 9) is an infection of the colon caused by Entamoeba hisolytica and, while usually asymptomatic, patients may exhibit symptoms ranging from mild diarrhea to dysentery (Berkow and Fletcher 1992: 225). Direct spread is common when personal hygiene is poor or the disease may be spread indirectly via food or water. Fruits and vegetables may be contaminated when grown in soil fertilized with human feces, worked with contaminated water, or prepared by a contaminated food handler. There were 1,341 patients discharged from hospitals who were treated for this in 1990. Other proto:wal intestinal diseases (ICD-9-CM codes 007.0-007.9) included more patients, 4,864, during the time period and most of these were treated for Giardiasis (Lambliasis), with 2,967 or 61.0 percent with this disease. Giardiasis is an infection of the small intestine caused by Giardia lamblia and is usually asymptomatic (Berkow and Fletcher 1992: 228). The organism is passed directly between an infected person to another (between children or between sexual partners) or indirectly via contaminated food or water. The category on List A containing the second largest number of patients in 1990 is ICD-9-CM codes 008.0-008.9, Intestinal Infections Due to Other Organisms. The subcategory contained 94,699 cases, or 53.7 percent, of all 176,282 patients treated for these infections. According to discussion with the

376

Thomas E. Steahr

Centers for Disease Control and Prevention, many foodbome diseases where the causative organism is not identified are placed in this category and into codes 009 and 558.9. Further research is needed to refine these estimates. Ill-defined intestinal infections (ICD-9-CM codes 009.0-009.3) contains 28,303 patients treated for that disease. The largest percent, 42.4, were diagnosed with infectious colitis, enteritis, and gastroenteritis (ICD-9-CM code 009.0). Next on List A is Listeriosis (ICD-9-CM code 027.0) which is caused by an infection from Listeria monocytogenes. Outbreaks have been caused by cabbage; hot dogs, and undercooked chicken. Direct contact may be achieved by veterinarians or butchers handling infected animals (Berkow and Fletcher 1992: 98). There were 1,248 patients discharged from hospitals treated for this disease. Other Disease Due to Viruses (ICD-9-CM codes 070.0-070.9, excluding 070.2 and 070.3) are largely Viral Hepatitis A without hepatic coma. Viral Hepatitis A spreads primarily via the fecal-oral route, with fecal shedding of the virus contaminating hands which are not washed prior to food handling (Berkow and Fletcher 1992: 900). Eating raw shellfish from directly contaminated water also may be responsible. Viral Hepatitis A without sepatic coma was diagnosed in 6,643 patients discharged from hospitals in the United States in 1990. In addition, 6,167 patients were treated for unspecified viral hepatitis without mention of sepatic coma. Cysticercosis (ICD-9-CM code 123.1) is an intestinal infection caused by the cestode Taenia solium. The infection of the tapeworm is the result of eating improperly cooked pork which is a carrier of the adult worm (Berkow and Fletcher 1992: 252). There were 1,442 patients released from hospitals in the United States who received treatment for this infection. Trichinosis (ICD-9-CM code 124) which is a parasitic disease caused by Trichinella spiralis, a roundworm also found in raw or inadequately cooked pork (Berkow and Fletcher 1992: 245). Table 17 .1 shows 42 patients discharged from hospitals after treatment for Trichinosis. By far the largest category of patients on List A is ICD-9-CM code 558.9, Unspecified Gastroenteritis and colitis, with 562,047 patients discharged from hospitals treated for this disease. Toxic effect of noxious substances eaten as food (ICD-9-CM codes 988.0-988.9) included poisoning from toxins produced by microtoxin in fish, mushrooms, berries, and other plants. In 1990, there were 352 patients in hospitals treated for these illnesses. Based on hospital discharge certificate data for 1990 shown in Table 17.1, there were a total of 817 ,090 patients treated for specific foodbome illness. This figure is derived from the counting of the listed ICD-9-CM codes anywhere on the discharge certificate. The first line on the seven line list of diagnosis made for each patient is reserved for the principal illness treated during the hospital stay. Of the total of 817,090 mentions, 445,901 or 54.6 percent appeared as the principal diagnosis.

Determining Foodborne Illness in the United States

377

Foodborne Illness, Ust B As explained for List A, List B presents the weighted sample estimates based on the random sample of hospital discharge certificates drawn by the National Center for Health Statistics survey. They represent the estimate of the actual number of discharge certificates by ICD-9-CM codes that would have been counted if a complete enumeration of all patient records were taken. Disease categories which may contain infections due to food or waterborne vehicles are shown in List B (Table 17.2). Of the 692,789 patients treated for bacterial infection in conditions classified elsewhere and of unspecified site (ICD-9-CM codes 041.0-041.9), there were 349,674 diagnosed with E. coli bacteria. This infection can be very common and may be passed directly from person to person or by eating undercooked beef or unpasteurized milk (Berkow and Fletcher 1992: 815). More research is needed on this and other categories on List B to determine the proportion of each disease which is foodbome. Potentially, a substantial number of the 1,005,709 cases might be transferred to List A.

Patient Hospital Costs Table 17.3 summari7.es the amount offoodbome illness present in the United States in 1990 in terms of discharged patients, hospital care, and average annual hospital costs. The total number of patients treated during the period for particular diseases from List A is 815,721. There were a total of 1,369 fewer cases reported than on Table 17 .1 because some sampled discharge certificates reported more than one foodbome illness and were included in this first disease count. The total number of days of care for each diagnostic category is also shown. The 815,721 patients accumulated over 5 million days of hospital care, an average stay of 6.2 days. The average length of hospital stay ranged from a low of 1.5 days for treatment of toxic effect of noxious substances eaten as food to a high of 12.8 days for treatment for Listeriosis. This is consistent with the average length of say of 6.4 days for inpatients discharged from hospitals for all causes combined in 1990 (Graves 1992). As an estimate of patient direct costs for this hospital treatment, the 1990 average cost per day of $687 for the community hospitals in the United States is used (Statistical Abstract of the United States 1992). This was multiplied by days of care to estimate the annual hospital costs which are presented in Table 17.3. The estimated range of annual costs is large, depending on the number of patients involved and the days of care. The low is $57, 700 for patients treated for trichinosis and the high is over $2 billion for unspecified gastroenteritis and colitis. The total average annual hospital costs to patients for all foodbome diseases contained on List A is estimated to be $3.5 billion.

378 TABLE 17.2 Patients Discharged from Hospitals by Category ofFoodborne Disease, List B, United States, 1990

Final Diagnosis Entered on Certificate as: Additionalb

Total Mentions

218 0 0 0 0 218

0 0 0 0 0 0

218 0 0 0 0 218

026 026.1

0

0

0

027 027.1 027.2

491 0 491

1,771 116 1,655

2,262 116 2,146

027.8

0

0

0

027.9

0

0

0

031 031.0 031.1

3,642 2,295 51

6,751 3,710 186

10,393 6,005 237

031.8

961

1,850

2,811

031.9

335

1,005

1,340

032 032.0

0 0

278 0

278 0

032.1

0

0

0

Disease Category

ICD-9 Code

Brucellosis B. melitensis B. abortus B. suis B. canis Other brucellosis Brucellosis, unspecified

023 023.0 023.1 023.2 023.3 023.8 023.9

Rat-Bite Fever Streptobacillary fever Other Zoonotic Bacterial Diseases Erysipelothrix infection Pasteurellosis Other specified zoonotic bacterial diseases Unspecified zoonotic bacterial diseases Diseases Due to Other Mycobacteria Pulmonary Cutaneous Other specified mycobacteria diseases Unspecified mycobacteria Diphtheria Faucial diphtheria Masopharyngeal diphtheria

Principal•

(continues)

379 TABLE 17.2 (continued) Final Diagnosis Entered on Certificate

as:

Disease Category Anterior nasal diphtheria Other specified diphtheria Unspecified diphtheria

ICD-9 Code

Principal•

I

Additionalb

Total Mentions

032.2

0

0

0

032.8 032.9

0 0

0 278

0 278

Septicemia Streptococcal Staphylococcal Pneumococcal Septicemia due to anaerobes Septicemia, gram negative Other specified septicemia Unspecified septicemia

038 038.0 038.1 038.2

145,395 17,872 39,467 4,875

138,572 29,513 0 2,371

283,967 47,385 39,467 7,246

038.3

2,681

2,890

5,571

038.4

0

0

0

038.8 038.9

10,099 70,401

16,772 87,026

26,871 157,427

Other Bacterial Diseases Gas gangrene

040 040.0

1,014

2,447

3,461

Bacterial Infection in Conditions Classified Elsewhere and of Unspecified Site Streptococcus Staphylococcus E.coli Unspecified

041 041.0 041.1 041.4 041.9

5,300 2,418 2,032 692 158

687,489 136,101 180,514 348,982 21,892

692,789 138,519 182,546 349,674 22,050

Other Rickettsioses Q fever

083 083.0

0 0

94 94

94 94

Other Cestode Infection Taenia solium

123 123.0

0 0

51 0

51 0

(continues)

380 TABLE 17.2 (continued) Final Diagnosis Entered on Certificate as: Disease Category Taenia saginata Taeniasis, unspecified Diphyllobothriasis, intestinal Sparganosis Hymenolepiasis Other specified Cestode, unspecified Other Intestinal Helminthiases Ascariasis Anisakiasis Strongyloidiasis Trichuriasis Enterobiasis Capillariasis Trichostrongyliasis Other specified helminthiasis Mixed intestinal helminthiasis Intestinal helminthiasis, unspecified Other and Unspecified Helminthiases Toxocariasis Gnathostomiasis Other specified helminthiasis Helminth, unspecified Toxoplasmosis Encephalitis

ICD-9 Code

Principal•

I

Additionalb

Total Mentions

123.2 123.3

0 0

0 51

0 51

123.4 123.5 123.6 123.8 123.9

0 0 0 0 0

0 0 0 0 0

0 0 0 0 0

127 127.0 127.1 127.2 127.3 127.4 127.5 127.6

494 0 0 0 0 203 0 0

1,411 270 0 40 322 779 0 0

1,905 270 0 40 322 982 0 0

127.7

291

0

291

127.8

0

0

0

127.9

0

0

0

128 128.0 128.1

0 0 0

895 895 0

895 895 0

128.8 128.9

0 0

0 0

0 0

130 130.0

3,297 1,241

4,882 1,459

8,179 2,700

(continues)

381 TABLE 17.2 (continued) Final Diagnosis Entered on Certificate as: Disease Category

ICD-9 Code

Conjunctivitis Chorioretinitis Myocarditis Pneumonitis Hepatitis Other specified sites Multisystemic Unspecified

130.1 130.2 130.3 130.4 130.5 130.7 130.8 130.9

985 985.0 985.1 985.2 985.3 985.4 985.5 985.6 985.8 985.9

Toxic Effect of Ober Metals Mercury Arsenic Manganese Beryllium Antimony Cadmium Chromium Other, specified Unspecified metal

Tota1s

I

Additionalb

Total Mentions

0 0 0 19 0 1,026 78 933

251 150 0 207 0 1,282 215 1,318

251 150 0 226 0 2,308 293 2,251

431 0 333 0 0 0 0 0 98 0

786 0 0 0 0 0 0 0 786 0

1,217 0 333 0 0 0 0 0 884 0

160,282

845,427

1,005,709

Principal•

Source: National Hospital Discharge Survey, 1990, United States National Center for Health Statistics, Hyattsville, MD. •Principal diagnosis on hospital discharge certificates appears on the first line. bAdditional diagnosis on hospital discharge certificates appears on lines 2-7.

The total hospital costs to the patient in 1990 of $3.S billion may be viewed as an upper level estimate of the direct costs of foodbome illness because it includes patients whose principal diagnosis was not foodbome illness. Table 17 .4 presents patients discharged from hospitals with principal diagnosis as foodbome illness, List A, in the United States in 1990. There were a total of 445,901 patients in this category and they accumulated a total of 1,794,734 days of hospital care, which is an average of 4.0 days per patient with principal

382 TABLE 17.3 Patients Discharged from Hospitals by Category of Foodbome Disease, List A, by Days of Care, Average Length of Stay, and Average Annual Hospital Costs, United States, 1990

Total Hospital Time

Diagnostic Category and ICD-9-CM Code Cholera 001 Typhoid 002 Salmonella 003 Shigellosis 004 Other Food Poisoning 005 Amebiasis 006 Other Protozoa! Intestinal Disease 007 Intestinal Infections Due to Other Organisms 008 Ill-Defined Intestinal Infection 009 Listeriosis 027 Viral Hepatitis Ac 070 Cysticercosis 123 .1 Trichinosis 124 Unspecified Gastroenteritis and Colitis 558.9 Noxious Substance Eaten as Food 988 All Conditions Above

199()& Discharged Patients

Days of Care

Average Length of Stay

Annualb Hospital Costs ($1,000)

92 2,056 17,239 2,112

368 15,448 139,241 7,348

4.0 7.5 8.1 3.5

252.8 10,612.8 95,658.6 5,048.1

6,114 1,341

31,200 8,035

5.1 6.0

21,434.4 5,520.0

4,864

51,959

10.7

35,695.8

175,779

1,361,630

7.7

935,439.8

28,183 1,248 12,810 1,442 42

226,772 15,934 87,036 12,629 84

8.1 12.8 6.8 8.8 2.0

155,792.4 10,946.7 59,793.7 8,676.1 57.7

562,047

3,097,865

5.5

2,128,233.3

352 815,721

528 5,056,077

1.5 6.2

362.7 3,473,524.9

Source: National Hospital Discharge Survey, United States National Center for Health Statistics, Hyattsville, MD, annual data for 1990. 1lncludes patients with mention of disease on lines 1-7 of the hospital discharge certificate. bBased on 1990 national average cost per day of$687, from U.S. Bureau of Census, Statistical Abstracts of the United States, 1992, United States Government Printing Office, Washington, D.C., Table 170. c Excludes viral Hepatitis B with or without mention of hepatic coma.

Determining Foodborne Illness in the United States

383

TABLE 17.4 Patients Discharged from Hospitals with Principal Diagnosis as Foodborne Disease, List A, United States, 1990

Total Hospital Time

Diagnostic Category and ICD-9-CM Code Cholera 001 Typhoid 002 Salmonella 003 Shigellosis 004 Other Food Poisoning 005 Amebiasis 006 Other Protozoa! Intestinal Disease 007 Intestinal Infections Due to Other Organisms 008 Ill-Defined Intestinal Infection 009 Listeriosis 027 Viral Hepatitis A 070 Cysticercosis 123.1 Trichinosis 124 Unspecified Gastroenteritis and Colitis 558.9 Noxious Substance Eaten as Food 988 All Conditions Above

Discharged Patients 1990

Days of Care

Average Length of Stay

Annual Hospital Costs ($1,000)

92 2,056 12,237 1,440

368 15,448 61,631

5,045

4.0 7.5 5.0 3.5

252.8 10,612.8 42,340.5 3,465.9

5,121 1,011

26,185 5,428

5.1 5.4

17,989.1 3,729.0

1,927

12,542

6.5

8,616.3

101,978

496,443

4.9

341,056.3

15,982 796 7,523 612 0

54,241 6,512 34,360 6,809 0

3.4 8.2 4.6 11.1 0

37,263.6 4,473.7 23,605.3 4,677.8 0

294,818

1,069,414

3.6

734,687.4

308 445,901

308 1,794,734

1.0 4.0

211.6 1,232,982.3

Source: National Hospital Discharge Survey, United States National Center for Health Statistics, Hyattsville, MD.

diagnosis of foodbome illness. The national cost of $687 per day in a hospital yields $1.2 billion of cost to patients. This may be viewed as a minimal direct cost of foodbome illness in 1990. Omitted from both cost estimates are the costs of physician charges and drugs (which may double the estimates). In addition, the productivity losses for persons ill from microbial foodbome disease are not included and are generally much greater than medical costs (Roberts 1992, Roberts and Foegeding 1991).

Thomas E. Steahr

384

Foodbome Din~:

Additional Perspectives

The National Ambulatory Medical Care Survey of 1990 provides data on patient records selected in a national sample of office-based physicians. These records contained a weighting variable based on patient characteristics and other sampling considerations. The summation of these weights, as determined by the National Center for Health Statistics, represents the estimate of the number of cases of a given disease that would have been counted if all physician office visits were enumerated. Table 17.S presents basic data on the frequency of foodbome illness as indicated by physician office visits in 1990, hospital discharge certificates in 1990, and as appears on death certificates for 1986 (the most recent year available). Examination of the total number of cases reporting ICD-9-CM codes on List A shows the pattern one might expect; that the largest number of cases is for physician office visits (6,242,353 visits were diagnosed as involving foodbome illness), followed by hospital discharge certificates (815,721 patients) and death certificate mentions (208,384 appearances on death certificates). These figures yield ratios of7.65 physician's office visits for every hospital treatment involving foodbome illness and 3.91 hospital discharges for every death certificate listing foodbome disease. This pattern tends to support the image of the frequency of foodbome illness as a pyramid in which the largest number of cases are unreported to physicians. Indeed, if the ratio of office visits to hospital discharge is applied to estimate the number of foodbome illness cases that are unreported, the estimate would exceed 47 million cases in the United States in 1990 (7 .65 x 6.2 million = 47 million unreported cases). There are two major exceptions to the pyramid image of disease frequency: Listeriosis and Viral Hepatitis A. In the case of Listeriosis, there were 1,248 hospital discharge certificates in 1990 with mention of this disease but 42, 157 death certificates mentioned the disease. This disease is caused by Listeria spp, which are bacilli found in the environment. Infection is most common in children and the elderly and is usually the result of eating contaminated dairy products or raw vegetables (Berkow and Fletcher 1992: 98). Infection may also occur by direct contact of veterinarians with infected livestock and with butchers or workers who slaughter infected animals. A similar pattern exists for Viral Hepatitis A. There were 12,810 cases with mention of this disease on hospital discharge certificates but 32,203 death certificates mentioned this disease. Food and water contaminated with the virus is a common source of infection and eating raw shellfish is sometimes involved. It is not clear why the frequency of mentions on death certificates is much higher than for hospital discharge certificates.

385

Determining Foodborne Illness in the United States

TABLE 17.5 Frequency ofFoodbome Disease, Physician's Office Visits and Hospital Discharge Certificates for 1990 and Appearance on Death Certificate for 1986, United States

Disease Category and ICD-9-CM Code Cholera 001 Typhoid 002 Salmonella 003 Shigellosis 004 Other Food Poisoning 005 Amebiasis 006 Other Protozoa! Intestinal Disease 007 Intestinal Infections Due to Other Organisms 008 ID-Defined Intestinal Infection 009 Listeriosis 027 Viral Hepatitis A• 070 Cysticercosis 123.1 Trichinosis 124 Unspecified Gastroenteritis and Colitis 558.9 Noxious Substance Eaten as Food 988 All Conditions Above

1990 Physician's Office Visits

1990 Hospital Discharge

1986 Appears on Death Certificate

0 0 0 38,521 162,173 0 76,451

92 2,056 17,239 2,112 6,114 1,341 4,864

0 0 79,428 3,666 0 0 12,376

1,238,689 246,063 0 127,298 0 0

175,779 28,183 1,248 12,810 1,442 42

0 14,136 42,157 32,203 0 0

4,353,158

562,047 352 815,721

0 24,418 208,384

0

6,242,353

Source: National Ambulatory Medical Care Survey, 1990; National Hospital Discharge Survey, 1990; and National Mortality Followback Survey, 1986, National Center for Health Statistics, Division of Vital Statistics, Hyattsville, MD. Note: Table includes List A diseases only. •Excludes viral Hepatitis B with or without mention of hepatic coma.

Foodbome Illness: Patient Demographics Characteristics of persons suffering from foodbome illness are important to know if valuation research is to reflect the impact on youth, elderly, minority groups, etc. Identification of groups with an elevated risk of foodbome illness by geographic area where they live is necessary for effective policy to reduce

386

Thomas E. Steahr

suffering from these diseases. For example, if rates of foodbome illness for specific diseases show a systematic variation by age, sex, and race, then future levels might be predicted by application of these rates to projections of population by those characteristics. Demographic analysis also provides a basis for continued epidemiological investigation. If a certain foodbome illness occurs at elevated rates within an identifiable geographic area, continued investigation within that area is necessary to explain and correct that event. Moreover, if demographic analysis reveals a concentration of a given foodbome illness among children under 1 year of age or the elderly 75 years of age or over, a different type of educational program is necessary to reduce those rates than a generalized program appealing to all age groups. In brief, disaggregation of persons suffering from foodbome illness is required to establish meaningful trends over time in selected diseases, to formulate an effective policy to deal with those trends, to document the impact of those actions on disease rates for different types of persons, and to anticipate future levels of foodbome illness. Selected demographic characteristics of patients who visited a physician's office and were diagnosed with a foodbome illness from List A in 1990 are ' percent were presented in Table 17 .6. Of the more than 6 million patients, 54.6 female and only 45.5 percent were male. The geographic distribution revealed the highest concentration in the South and Midwest (63.3 percent of all cases}, with the Northeast region accounting for only 15.6 percent. The racial composition of this patient population was primarily white (82. 7 percent) and, in terms of residence, located in metropolitan areas (71.2 percent). The age structure reveals a concentration in the younger ages, with the largest group in the 1-4 year old category, with 1.2 million or 20.1 percent of all patients. In fact, children under the age of 14 years accounted for 43.5 percent of all patient visits to physician offices. This may be the result of parents being more likely to take their children to a physician when illness occurs than they are in the event of their own illness. In brief, a typical patient would be a white, female child living in the metropolitan areas of the South or Midwest. Demographic characteristics of patients discharged from hospitals with mention of foodbome illness from List A in 1990 are shown in Table 17. 7. For the over 815 thousand discharged patients, the largest percent were women (58.8 percent). The geographic distribution was clearly concentrated in the South with 37.5 percent of all cases living there. The West had the lowest proportion with 14.2 percent. The racial composition was largely white (70.4 percent) but at lower levels than office visit data. Information on marital status revealed that single (37.8 percent) was the most frequently mentioned category. Unlike data for office visits, the age distribution of patients discharged from hospitals is not concentrated at the younger ages. In fact, the largest number and percent of all patients were in the 75 and over age group (16.9 percent). Persons 65-74 years of age accounted for 10.9 percent of all patients. Days of stay in the hospital were normally less than seven days (80.1 percent of all stays). In brief, the

387 TABLE 17.6 Selected Demographic Characteristics, Physician Office Visits with Foodbome Illness, List A, United States, 1990

1990 Office Visits Number

Percent

Sex Male Female

6,242,353 2,836,449 3,405,904

100.0 45.4 54.6

Region Northeast Midwest South West

6,242,353 972,011 1,942,877 2,012,658 1,314,807

100.0 15.6 31.1 32.2 21.1

Race White Black Other Not stated

6,242,353 5,159,064 556,414 283,450 243,425

100.0 82.7 8.9 4.5 3.9

Residence Metropolitan Nonmetropolitan

6,242,353 4,447,185 1,795,168

100.0 71.2 28.8

Age Under 1 yr. 1-4 yrs. 5-14 15-24 25-34 35-44 45-54

6,242,353 745,338 1,258,090 720,428 434,728 637,656 893,805 366,670 383,259 440,625 361,754

100.0 11.9 20.1 11.5 7.0 10.2 14.3

Demographic Characteristics

55-64 65-74 75+ yrs.

5.9 6.1 7.1 5.8

Source: National Ambulatory Medical Care Survey, 1990, Department of Health and Human Services, National Center for Health Statistics, Hyattsville, MD.

388 TABLE 17.7 Selected Demographic Characteristics, Patients Discharged from Hospitals with Foodbome Illness,• List A, United States, 1990 1990 Hospital Discharge Certificate Demographic Characteristics Sex Male Female Region Northeast Midwest South West Race White Black Other Not stated Marital Status Married Single Widow/Divorced Not stated Age Under 1 yr. 1-4 yrs. 5-14 15-24 25-34 35-44 45-54

55-64

65-74 75+ yrs. Days in Hospital 0-7 days 8-14 15-21 22-28 29+ days

Number

Percent

815,721 336,254 479,468 815,721 163,392 230,869 305,976 115,484 815,721 574,063 87,818 23,132 130,708 815,721 185,030 308,917 88,210 233,564 815,721 98,971 87,066 50,636 64,561 82,629 72,490 62,051 70,519 89,097 137,692 815,721 653,513 96,870 30,028 14,356 20,954

100.0 41.2 58.8 100.0 20.0 28.3 37.5 14.2 100.0 70.4 10.8 2.8 16.0 100.0 22.7 37.8 10.8 28.6 100.0 12.1 10.7 6.2 7.9 10.1 8.9 7.6 8.6 10.9 16.9 100.0 80.1 11.8 3.7 1.8 2.6

Source: National Hospital Discharge Survey, United States National Center for Health Statistics, 1990 Survey. •includes patients with mention of foodbome disease on lines 1-7 of the hospital discharge certificate.

Determining Foodborne Illness in the United States

389

hospital discharge data revealed the typical patient to be a white, single female in the older age categories who likely lives in the South region or the Midwest. Demographic characteristics of persons who died with mention of foodbome illness on the death certificate from List A in the United States in 1986 are presented in Table 17.8. In terms of death data, males account for 65.6 percent

TABLE 17.8 Selected Demographic Characteristics, Deaths from Foodbome Illness•, List A, United States, 1986 1986 Deaths Demographic Characteristics

Number

Percent

Sex Male Female

208,384 136,695 71,689

100.0 65.6 34.4

Region Northeast Midwest South West

208,384 56,421 68,788 67,753 15,422

100.0 27.1 33.0 32.5 7.4

Race

208,384 172,791 32,273 3,320

100.0 82.9 15.S 1.6

Age Under 1 yr.

208,384

100.0

12,849 25,854 18,413 12,111 84,779 54,378

6.2 12.4 8.8 5.8 40.7 26.1

White Black Other

1-4 yrs.

5-14 15-24 25-34 35-44 45-54 55-64 65-74 15+ yrs.

Source: National Mortality Followback Survey, 1986, National Center for Health Statistics, Division of Vital Statistics, Hyattsville, MD. •Death data are for persons 25 years ofage and over.

390

Thomas E. Steahr

of all the 208,384 certificates in this year with mention of foodbome disease, a reversal of the patterns described for the previous data sets. The regional concentration is similar with 65.5 percent of all deaths occurring in either the South or the Midwest. Again the white population accounts for the majority of deaths, with 82.9 percent of the diseased classified in this category. Since this survey is restricted to persons who died after 25 years of age, the age structure is considerably older. Persons 65 years of age and over accounted for 66.8 percent of all death certificates in the sample. In brief, the death certificate data suggest the typical person who died with mention of foodbome disease to be an older, white male who resided in either the Midwest or South region.

Conclusions and Discussion At the level of microbiology, knowledge about the extent of foodbome illness is being extended from the earlier connection with short-term, acute conditions to the recognition that foodbome pathogens may also be involved in long-term, chronic health problems. This represents a significant expansion of the health arena within which foodbome disease must be assessed. Food is becoming a major sector of acute enteric disease and of chronic diseases of unknown etiology. One may anticipate a continued expansion of the impact of microbial foodbome illness as the national population ages during the next several decades, as the number of HIV-related diseases increases, and as the ethnic composition becomes more diverse with a wide range of food-related customs. An important conclusion from the recent recognition by microbiology that foodbome pathogens may be involved in chronic diseases as well as acute conditions, and from the recognition that social/cultural practices also contribute to foodbome illness, is that there is no single source in food production and consumption which may be held responsible. Commercially processed foods purchased in retail stores may be free of pathogens but become contaminated by people with unhygienic food handling habits. Therefore, attempts by policy makers to reduce the levels of foodbome illness in the United States must recognize the multidimensional nature of the problem and not rely on a single method to increase public protection. In terms of data needs, it is clear that the national sample surveys taken by the National Center for Health Statistics are of critical importance in the surveillance of foodbome disease. They should be continued on an annual basis and an increase in the sample size should even be considered. Beyond these data, there is a significant need for more special purpose sample surveys asking persons about food purchases, food preparation, illness events, etc. The actual frequency of unreported cases of foodbome illness needs to be established. In addition, characteristics of persons becoming ill need documentation so that high risk groups may be determined and corrective actions suggested.

Determining Foodborne Illness in the United States

39I

Note 1. The author wishes to thank the Food Marketing Policy Center, University of Connecticut, for research support of this project.

References Abdussalam, M. and D. Grossklaus. 1991. Foodbome Illness: A Growing Problem. World Health 18:9. Archer, Douglas L. and Frank E. Young. 1988. Contemporary Issues: Diseases With a Food Vector. Clinical Microbiological Review 1(4):377-398. Bean, Nancy H. and Patricia M. Griffin. 1990. Foodbome Disease Outbreaks in the United States, 1973-1987: Pathogens, Vehicles, and Trends. Journal of Food Protection 53(9):804-817. Bean, N. H., P. M. Griffin, J. S. Goulding, and C. B. Inez. 1990. Foodbome Disease Outbreaks, 5-Year Summary, 1983-1987. In Center for Disease Control, Morbidily and Mortality Weekly Report, 39 March, No. 55-1, CDC Surveillance Summaries. Berkow, R. and A. J. Fletcher, ed. 1992. Sixteenth Edition of the Merck Manual of Diagnosis and Therapy. Rahway, NJ: Merck Research Laboratories, Merck and Company, Inc. Bryan, F. L. 1982. Diseases Transmilted by Food: A Classification and Summary, 2'4 Edition. U.S. Department of Health and Human Services, Public Health Services, Center for Disease Control, No. 00-1845, 101. Chalker, R. B. and M. J. Blaser. 1988. A Review of Human Salmonellosis: III. Magnitude of Salmonella Infection in the United States. Review of Infectious Diseases 10(1):111-124. Cliver, Dean 0., ed. 1990. Foodbome Diseases. New York, NY: Academic Press. Doyle, Michael P., ed. 1989. Foodbome Bacterial Pathogens. New York, NY: Marcel Dekker, Inc. Foster, George M. and F. K. Kaferstein. 1985. Food Safety and the Behavioral Sciences. Social Science and Medicine 21(11):1273-1277. Graves, Edmund J. 1992. 1990 Summary: National Hospital Discharge Survey. Advance Data. National Center for Health Statistics, No. 210 (February 18). Kaplan, G. A., M. N. Haan, S. L. Syme, M. Minkler, and M. Wmldeby. 1987. Socioeconomic Status and Health. In Closing the Gap: The Burden of Unnecessary Illness, eds. Robert W. Amber and H. B. Dull, 125-129. New York, NY: Oxford University Press. Karaffa, N. C., ed. 1992. International Classification of Diseases, !J'I Revision, If" Edition, Clinical Modification, Volumes 1-3, 1992. Los Angeles, CA: Prentice Management Information Corporation. Riemann, Hans and Frank L. Bryan, eds. 1979. Foodborne Infections and Intoxications, 2nd Edition. New York, NY: Academic Press. Roberts, T. 1992. Estimated Foodbome Disease Risks and Costs in the United States. Paper Presented at the ASSA Meetings, New Orleans.

392

Thomas E. Steahr

Roberts, Tanya and Peggy M. Foegeding. 1991. Risk Assessment for Estimating the Economic Costs of Foodbome Disease Caused by Microorganisms. In Economics of Food Safety, ed. Julie A. Caswell, 103-129. New York, NY: Elsevier Science Publishing Company, Inc. Strain, Laurel A. 1989. Illness Behavior in Old Age: From Symptoms Awareness to Resolution. Journal of Aging Studies 3(4):325-340. Swedlund, Alan C. and George J. Armelagos. 1990. Disease in Populations in Transition. Westport, CT: Bergin and Garvey. Vargas, L. 1990. Old and New Transitions and Nutrition in Mexico. In Disease in Populations in Transition, ed. A.C. Swedlund and G. T. Armelagos. Westport, CT: Bergin and Garvey. Wolinsky, F. D., B. E. Aguirre, L. FaM, V. M. Keith, C. L. Arnold, J. C. Niederhauer, and K. Dietrich. 1989. Ethnic Differences in the Demand for Physician and Hospital Utilization Among Older Adults in Major American Cities: Conspicuous Evidence of Considerable Inequalities. The Milbank Quarterly 67(4):412-449.

18 Measuring the Food Safety Risk of Pesticides Kelly A. Day, Betsey A. Kuhn, and Ann M. Vandeman 1 Public concern about the use of pesticides and their possible effects on food safety has been increasing in recent years. Lawmakers are seeking to address these concerns, while assuring an abundant and affordable food supply. Whether or not new legislation is enacted, a tradeoff continues to exist between the reduction of pesticide residues and maintaining the current system of agricultural chemical use. The economic benefits of pesticide use are relatively easy to measure, given estimates of yield losses from pest damage. However, the economic costs of pesticide use, in terms of human health effects, are extremely difficult to measure. Even harder to assess are the benefits to human health of incremental reductions in pesticide residues. Were there a clearer measure of these benefits, policy makers might have better information about the food safety effects of pesticides and therefore more easily determine economically and socially rational pesticide policies. In this chapter, we have attempted a first step towards the valuation of pesticide reduction-the construction of a risk-weighted measure of food safety. While this work is in an infant stage, we hope it will focus attention on the issues underlying pesticide risk and human health concerns. Consumer surveys indicate that the public is concerned about the risks associated with consuming foods which contain pesticide residues. Horowitz and Carson (1990), for example, report that on a cost per death delayed basis, consumers would rather reduce the risks associated with pesticide residues on foods than those associated with PCBs in drinking water, radiation from the sun from use of CFCs, or automobile exhaust. Another survey (Sachs et al. 1987) compared consumers' concerns about pesticides in 1965 and 1984. The percent of respondents "with a great deal or some concern about the danger of eating fruits and vegetables that have been sprayed or dusted with pesticides" increased from 41.S percent in 1965 to 71.1 percent in 1984. 393

Kelly Day, Betsey Kuhn, and Ann Vandeman

394

In this chapter, we use the level of pesticide residues detected on a market basket of food prepared for home consumption as a proxy for the food safety risk associated with pesticides. We do not include other types of food safety risk, such as microbial contamination or natural toxins associated with food consumption (see, e.g., Roberts and van Ravenswaay 1989, Ames et al. 1987). We also do not offset the nutritional benefits of the food contained in the market basket against the pesticide residues detected on the food. Thus, the level of residues detected in a typical diet is not intended as a measure of general food safety. It is a measure of food safety risks associated with pesticide use and is appropriate for the purpose of investigating changes in these risks. If our purpose was to value the broad benefits of pesticide use, we might wish to include nutritional gains from supply changes attributable to the use of pesticides. The definition of benefits and costs, as well as the approach used to evaluate them, will greatly influence the valuation process.

A Method of Measuring Pesticide-Related Food Safety Risk The Total Diet Study A method for estimating food safety risk related to pesticides must be based on the most accurate measure of the chemical residues that humans consume in small quantities over time. The pesticide residue data we chose for this analysis were collected by the Federal Food and Drug Administration (FDA) in their Total Diet Study. 2 The Total Diet Study is the most appropriate source of data for this project because it measures actual residues in a typical diet drawn from a market basket of food. The study is conducted to determine levels of chemical contaminants, including pesticide residues, in foods prepared for consumption, and thus is the closest approximation to the average consumer's exposure to pesticides through the diet. The study has been conducted on a yearly basis since 1961, providing the only continuous time series of data on pesticide residues in the human diet. 3 In order to determine the residues present in human diets, FDA selects a market basket of food to represent typical consumption patterns. These market baskets are based on nationwide dietary surveys, and have been revised with changes in the typical diet throughout the history of the study. 4 Calorie counts also have been revised to reflect changing consumption patterns. The food is purchased at the retail level and is fully prepared for consumption. It is then tested for the presence of residues at FDA laboratories. The resulting estimates of residue concentrations are weighted according to the food consumption survey and summed to calculate daily intake figures for each chemical. We used the residue data reported for men 15 to 20 years-old5 to represent food safety risk. The Total Diet Study assessed other groups, including infants,

Measuring the Food Safety Risk of Pesticides

395

children, and the elderly. Infants and children were first included in the study in 1975, with the definition of infants being revised in 1982. Considerable attention has been focused on the effects of pesticide residues on this group. While residues for all age groups fall well below acceptable daily intake levels, 6 infants and children appear to have the greatest level of exposure (Archibald and Winter 1989). However, we did not attempt to assess their food safety risk for two reasons. First, data for this group were not collected during the first fourteen years of the Total Diet Study. Fifteen to 20 year-old males are the only age/sex group for whom residue data are reported consistently over the entire study period. Second, while research suggests that the risks of pesticide exposure may differ for children, how and to what degree risks are different is unclear (National Academy of the Sciences 1993a). Therefore, estimating the relative risks for children as compared to adults was not possible given the available information. In order to correct for the effect of changes in caloric intake on residue levels, we standardii.ed the data to 2,S20 calories, equal to the current estimate of total calories consumed by IS to 20 year-old males.

A Risk-Weighted Measure of Food Safety To construct a measure of food safety risk arising from pesticide residues, we needed to take account of a number of types of risk and differences in the level of risk for each of the 78 chemicals reported. Pesticides vary greatly in their toxic effects. For example, chlordane is a known carcinogen, while sulfur pesticides have no known carcinogenic properties. We reviewed the available toxicological data to evaluate the risks associated with each pesticide. Several measures have been developed to evaluate the potential for adverse physiological reactions from exposure to pesticides. The possible effects fall into two broad categories: acute toxicity, which generally refers to the damage resulting from a single dose or exposure, and chronic toxicological effects, which are health effects caused by repeated exposures to a substance over an extended period. The most commonly found acute toxicity measure is the W 50 (median lethal dose), or the amount of a substance necessary to kill SO percent of a sample population when exposed in a single dose. The LD50 is used to measure both oral and dermal acute toxicity. Expressed as milligrams of chemical per kilogram of body weight, the lower the LD50 , the more toxic the chemical. Hammitt used the reciprocal of the LD50 multiplied by 100 to form an acute index (Hammitt 1986). Acute toxicity is most relevant to populations who come in direct contact with a chemical, such as in the manufacture, mixing, loading, and application of pesticides. Because we sought a measure of the risks of pesticide residues for food consumers, we used indicators of chronic rather than acute risk.

396

Kelly Day, Betsey Kuhn, and Ann Vandeman

A comprehensive indicator of the chronic health risk associated with exposure to pesticides needs to take into account a number of possible health effects. These include:

1. Oncogenicity: equivocal tumor producing, unknown carcinogenicity, 2. Carcinogenicity: cancer producing, 3. Mutagenicity: causing unnatural changes in cells, in vitro, 4. Teratogenicity: causing birth defects, 5. Neurotoxicity: destructive to nerves or nerve tissue, 6. Immunotoxin: destructive to the immune system, 7. Damage to the reproductive system, and 8. Damage to other organ(s). There are several chronic risk measurements. However, the primary focus of chronic health risk assessment bas been limited to consideration of carcinogenicity and oncogenicity. The Q* Rating. The Q* is an estimate of the upper bound of extra incidents of tumor formation in humans that can be expected given a 70 year lifetime exposure to a certain dose of a chemical (both from dietary and other exposures). It is equal to the slope of the dose response curve from animal studies, and is expressed in terms of tumors/milligram of pesticide/kilogram of body weight/day. It does not account for the carcinogenicity or severity of the tumors produced. Also, it does not account for health effects other than oncogenicity. It is not peer reviewed, and is not available for all of the chemicals in the Total Diet Study. The Weight-of-the-Evidence Rankings. There is a chronic risk index created by the U.S. Environmental Protection Agency (EPA) for possible carcinogens, called the weight-of-the-evidence rankings. It is a system of classifications EPA uses to evaluate all potential human carcinogens. The ranking is based on oncogenicity (tumor production) tests, both positive and negative; mutagenicity studies; varieties of tumors induced; the rankings of structurally similar chemicals; and the replication of positive results. There are six possible rankings: A : Bl: B2: C : D : E :

Human carcinogen, Probable human carcinogen (some epidemiological or human data), Probable human carcinogen (no epidemiological data), Possible human carcinogen, Not classifiable as to carcinogenicity, and Evidence of non-carcinogenicity for humans.

EPA uses a three step process to assign a ranking to each chemical. First, human and animal study evidence is evaluated separately. Then these two

Measuring the Food Safety Risk of Pesticides

397

sources of evidence are combined to form an initial overall ranking. Finally, supporting evidence is used to arrive at a final ranking (U.S. EPA 1992). For our purpose of deriving an index of food safety risk, the weight-of-theevidence suffers from two major weaknesses. First, it exists for only 32 of the 78 pesticides found in the Total Diet Study. Second, it covers only oncological risks. Other health risks-neurotoxicity, immunotoxicity, nononcogenic reproductive effects, and teratogenicity-are not factored in (National Academy of Sciences 1987). Furthermore, the ranking is based solely on the available evidence, and lack of evidence does not correlate necessarily with low risk. Therefore, a possible human carcinogen may actually pose a greater risk to human health than a probable human carcinogen, despite the lower weight-ofthe-evidence ranking, simply because of a greater quantity of accumulated evidence (Harvard Center for Risk Analysis 1992).7 Because of these limitations, we chose to include additional information to create a more comprehensive risk index. A number of difficulties accompany any effort to create a risk index: inadequate understanding of pesticide risk; the lack of sufficient testing, particularly of chronic effects; and the problems associated with extrapolation of animal tests to humans (Hammitt 1986). Current toxicological testing generally uses "maximum tolerated doses" (MTD) to determine carcinogenicity, a much debated practice (National Academy of Sciences 1993b). A MTD is the largest quantity of a chemical that laboratory animals can ingest without causing serious health damage (other than cancer). These tests are used to detect carcinogenicity at relatively high levels of exposure. However, they provide limited information about the relationship between other exposure levels and adverse health effects. A linear relationship between dose levels is generally assumed, which may or may not be the case. A pesticide may have a high threshold of exposure that must be crossed in order for any adverse health effects to exist. On the other hand, a pesticide could be more huardous at lower levels than a linear curve would suggest. This uncertainty is exacerbated by the process of inferring human health risk from the results of animal tests. Generally, scientists assume that human bodies will respond in much the same way as laboratory animals, adjusting for weight differences. However, humans may respond differently than animals (e.g., metabolize a chemical differently). 8 Furthermore, chemicals which are assigned the same risk classification (e.g., probable human carcinogen) may have different probabilities of cancer causation. Therefore, even among experts in the toxicological field, there is extensive debate about methodology and considerable uncertainty surrounding the interpretation of toxicological test results. Though establishing the relative risks of pesticides is an extremely complex endeavor, it is very important when aggregating pesticide residues to avoid assigning equal weights to residues from substantially different chemicals. Existing toxicological data provide sufficient information to differentiate relative

398

Kelly Day, Betsey Kuhn, and Ann Vandeman

risks between pesticides. We are not the first to employ toxicological data in this manner. The starting point for developing our index was the work of Hammitt (1986), who constructed a series of food safety risk measures based on acute and chronic effects of pesticides. Hammitt (1986) used results from five types of toxicological tests-for carcinogenicity, neoplasticity, equivocal tumor agents, mutagenicity, and birth defects-to construct two alternative chronic risk indexes. The results were compiled in the Registry of Toxic Effects of Chemical Substances [RTECS]. The indexes considered only whether a test had been performed, not whether the test results were positive or negative. Therefore, Hammitt's indexes are reliable indicators of risk only if the toxicological tests have been performed and reported in RTECS. Three of the indexes are described below. Chronic risk index: Hammitt's chronic risk index is developed from the results of five types of toxicological tests contained in RTECS: Carcinogenicity, neoplasticity (non-spreading carcinogenic tumors), equivocal tumor promotion (oncogenicity), mutagenicity, and teragenicity. The first four types of chronic effects are weighted according to their relative severity. If a pesticide is considered carcinogenic, it is given a rating of 1.0. Pesticides which are neoplastic are rated at 112. Substances that cause equivocal tumor production and mutagenic reaction are rated 114 and 1/8, respectively. Teragenicity is considered separately. A positive test for teragenicity contributes 1.0 to the chronic risk index rating. With the exception of a positive teragenicity contribution, the effects are not additive and the chronic risk rating is simply the value of the highest ranking test. This index is of limited value if similar toxicological tests have not been performed and reported for all pesticide residues. The index also does not account for the severity of test reactions. Chronic sum index: For this index, Hammitt uses the same information considered in the chronic risk index, but sums the assigned weights of each toxic effect to derive a total risk weight for each chemical. In addition, carcinogenicity and mutagenicity are assigned equal weights. Again, missing tests reduce the reliability of the index, and potency is not considered. Combined risk index: This index combines Hammitt's acute and chronic risk indices (Hammitt 1986). The indices can be weighted according to the discretion of the researcher, although a simple average of the two indices is commonly used. A New Index. Because we wanted to create an index that accounted for a more comprehensive set of effects, we used three data bases to assemble toxicological information about the pesticides. The data bases were: Integrated Risk Information System (IRIS): IRIS is a data base maintained by the EPA containing toxicological and regulatory information for approximately 400 chemicals. The data base includes carcinogenic and noncarcinogenic risk information, such as Q* ratings and weight-of-the-evidence rankings, when available. The IRIS data base includes conclusions about chemical toxicity drawn by EPA scientists.

Measuring the Food Safety Risk of Pesticides

399

Registry of Toxic Effects of Chemical Substances (RTECS): RTECS is managed by the National Institute for Occupational Safety and Health. This data base contains acute and chronic health risk data on over 90,000 chemicals. We relied on RTECS primarily for information on the carcinogenic, mutagenic, and reproductive effects of chemicals. Chemical Carcinogenesis Research Information System (CCRIS): CCRIS is a data base created by the National Cancer Institute containing over 2,100 peerreviewed toxicological records. CCRIS reports the results for carcinogenic, mutagenic, and tumor promotion and inhibition studies. Our goal was to assign each pesticide to a risk category, either a higher, medium, or lower/unknown level of risk, based on data derived from these sources. We chose carcinogenicity as the first criteria to evaluate a pesticide's risk, based on regulatory precedent. Thus each pesticide in the higher risk category was a probable or declared carcinogen. Pesticides in the medium risk category were either suspected or possible human carcinogens, or they showed a greater number of positive mutagenic or developmental test results than other pesticides in the Total Diet Study. The lower risk category contains pesticides with negative or no carcinogenic test results or with no tests and limited evidence of mutagenic and reproductive effects. Unfortunately, those pesticides for which few toxicity studies have been conducted fall into this category. It is important also to emphasize that the rankings are relative risks, hence the terms higher, medium, or lower. Throughout the period of the Total Diet Study, with few exceptions, the quantities of pesticide residues detected have been well below the legal tolerance levels. 9 Our first step was to use a step-wise procedure to rank each pesticide according to evidence of carcinogenicity. The criteria we considered, in order, were the following: 1. We began with the weight-of-the-evidence ranking, when available. If

a pesticide was classified as a "probable human carcinogen• it automatically fell into the higher risk category. We placed "possible human carcinogens• initially in the medium risk category. 2. We considered other data on suspected carcinogenicity from EPA, such as information accompanying Special Review and cancellation reports. 3. Next we considered the results of carcinogenic studies from RTECS and CCRIS. We assigned greater weight to results from multiple sources. 4. Finally, we used the Basic Guide to Pesticides, published by the Rachel Carson Council (Briggs 1992), to verify the conclusions reached from other sources. While carcinogenicity was the first criteria for rating pesticides, we believed, following Hammitt (1986), that a complete index must incorporate other chronic health effects. Cropper et al. (1990) discussed the difficulties in quantifying

400

Kelly Day, Betsey Kuhn, and Ann Vandeman

noncarcinogenic effects and concluded that while quantification was infeasible, such effects should be noted and incorporated. Additional reports also support the inclusion of other chronic health effects (Harvard Center for Risk Analysis 1992). To evaluate noncarcinogenic health effects we examined the following sources: 1. First, EPA information from IRIS, press releases, special reviews, and cancellation reports. 2. Next we compiled the results of mutagenic and reproductive studies from RTECS and CCRIS, giving greater weight to positive results from multiple sources. 3. Again, we used the Basic Guide to Pesticides (Briggs 1992) as a source for confirmation of other studies. We deliberately avoided counting the number of toxicity studies because, for a variety of reasons, all pesticides have not been studied equally. Some of the older pesticides that are no longer registered have less toxicological data available. Other pesticides have received more public attention and, consequently, more testing. However, we did assign greater risk to those pesticides with a greater number of diverse, positive tests. Thus, our category assignments (higher, medium, lower) are admittedly and necessarily subjective, based on the risk a pesticide presented, subject to our interpretation of available evidence and relative to the risk of other pesticides in the Total Diet Study. The criteria included the nature and severity of effects, replicability of results, and regulatory actions taken by EPA. Table 18.1 shows the categorimtion of the 78 chemicals detected in the total diet study by relative risk category. While somewhat arbitrary, we feel this approach is defensible on two grounds. First, we believe that we could evaluate these data best using ordinal measures, because neither the data nor our level of expertise justified quantitative assessments of risk. Second, the valuation of relative risks is itself subjective. The use of carcinogenicity as a primary (and often the single) criteria for risk evaluation is both a function of the ease of data collection and the subjective value placed on cancer avoidance. Until both toxicological data and economic value assessments are more refined, assessment of relative risks involves subjectivity. For example, to value cancer avoided (either incidence or deaths) higher than the prevention of a birth defect may contradict the conclusion of an exclusively economic valuation process. We applied integer weights directly to each risk classification, with the higher risk category receiving a weight of 4, the medium risk category a weight of 2, and the lower risk category a weight of 1. We arrived at these weights through our review of the evidence, which, in our opinion, indicated a smaller difference of risk between the lower and medium risk categories than between the higher and medium categories. The weighted residues were then summed

401

Measuring the Food Safety Risk of Pesticides

TABLE 18.1 Chemicals Detected in the Total Diet Study, by Relative Risk Category

Higher Risk 2,4-D 2,4,5-T Aldrin Atrazine BHC Captan Chlordane Chlorobenzilate DDE DDT Dicofol {Kelthane) Dieldrin Dimethoate Endrin Folpet Gamma BHC (Lindane) Heptachlor Hexachlorobenzene Methoxychlor Trans-Nonachlor Nitro fen Parathion Pentachlorophenol (PCP) Permethrin Perthane o-Phenylphenol Quintozene Propargite TDE Toxophene

Medium Risk Acephate Aldicarb Azinphos-methyl Carbaryl Carbofuran Chlorpropham Demeton Demeton-S Deceton-S sulfone Diazinon Dicloran (DCNA) Endosulfan EPN Fenitrothion Fenvalerate Linuron Malathion Methidathion Methomyl Phorate Phosmet Phosphamidon Ronnel Thiabendazole

Lower or Unknown Risk Aldoxycarb Carbophenothion Chlorpyrifos Chlorpyrifos-methyl DCPA DEF Diphenylamine Disulfoton Disulfoton sulfone Ethion Fonofos lprodione Leptophos Methamidophos Methiocarb Mevinphos Omethoate Oxamyl Parathion-methyl Phosalone Pirimiphos-methyl Sulfur Tecnazene Vinclozolin

to create the risk index. The resulting index measures residues in "risk equivalents." This was an arbitrary weighting, based on our judgement. Other weighting schemes easily could be justified. However, we found that the results of the risk-weighted index were quite insensitive to the weights used.

Kelly Day, Betsey Kuhn, and Ann Vandeman

402

Results Figure 18.1 depicts both the risk-weighted and unweighted residue series. The risk-weighted series is equal to: (1)

where:

n

L

R

1 :::

i-1

W;P;1

Ri

= risk-weighted index in year t,

Pit

= residue for pesticide i in year t.

wi = category weight for pesticide i,

Adjusting for the relative risks of pesticides, Figure 18.1 suggests that food safety risk was greater in the late 1960s than the unweighted residue data indicate. The higher residue levels of several highly toxic pesticides used in the late 1960s account for the greater risk shown by the weighted index. The influence of these residues on the index declined in later years as several higher

6 Milligrams per kilogram of body wei ht (thousands) Unweighted residues Risk weighted residues 5

e

+

es

67 69 11 73 75

n

19 81 83 85 87 89 91 Year

FIGURE 18.1 Pesticide Residues Found in a Market Basket of Food

Measuring the Food Safety Risk of Pesticides

403

risk pesticides were cancelled. Among these were DDT and its derivatives, 2,4,5,-T, dieldrin, and endrin. Measured as a risk-weighted index of total food pesticide residues, our study shows that food safety risk, after declining from 1965 to 1972, remained relatively stable and then increased slightly in the late 1980s.

Factors Related to Changes in Food Safety Risk In an earlier paper (Day et al. 1993), we considered the effect of regulation on food safety, using the risk-weighted index of residues to measure changes in food safety risk from pesticide residues over time. Although the effects of regulation are not the focus of this chapter, it is useful to consider the determinants of the changes in food safety risk associated with pesticide residues. We briefly describe some of these determinants below.

Income Changes in real per capita disposable income can reflect changes in tastes and preferences. If food safety is a superior good, the demand for food safety should increase with income. One might expect consumers to be more concerned about food safety risk as their incomes increase. Efforts of producers, handlers, and retailers to reduce residues may in part be a response to this increased concern. At the same time, changes in consumption patterns are correlated with income. For example, fresh fruit and vegetable consumption increases with income (Lutz et al. 1992). Larger quantities of chemicals are used in the production of these commodities (USDA 1991 and 1992a). Furthermore, because these commodities are readily perishable, additional pesticides are applied to reduce spoilage. Both factors lead to higher residues on these foods. Therefore, the net effect of income on food safety risk, both increasing demand for food safety and increasing consumption of fruits and vegetables, is uncertain.

Pesticide Use Pesticide use also influences the quantity of resides found in food. We expect residues to increase with the intensity of pesticide use, not necessarily with total quantities applied. Pesticide use data derived from two sources reveal the trend in intensity of use. First, the total quantities of pesticides applied in agriculture, given in pounds of active ingredients, are reported by EPA and USDA (Aspelin et al. 1992, Osteen and S.zmedra 1989). This time series of pesticide quantities goes back to 1964, and is derived from a combination of industry sources and agricultural surveys. The second component, the number

Kelly Day, Betsey Kuhn, and Ann Vandeman

404

of acres of cropland, was obtained from Agricultural Statistics. By dividing the quantity of pesticides by the area of cropland we obtain a measure of the intensity of pesticide use. This series is depicted in Figure 18.2. Throughout the late 1960s and 1970s, there was a steady increase in pesticide use per acre. The lower prices of pesticides relative to other inputs encouraged the substitution of pesticides for labor (National Academy of the Sciences 1989). By the end of this period, the percentage of all cropland treated with pesticides, particularly herbicides, reached 95 to 98 percent (Osteen and Szmedra 1989). Thereafter, beginning in 1980, total and per acre pesticide use stabili7.ed. Because herbicides account for much of the increase and because often they are not applied directly to crops, we also show pesticide use excluding herbicides. This measure of use should be more closely related to changes in residues found on food. Without herbicides, total pesticide use per acre has changed little over the period examined. In terms of total pounds applied, insecticide use actually declined in the last decade (Osteen and Szmedra 1989), partly as a result of substituting more potent chemicals which are effective at lower rates of application.

~s-1.:;

3 Pounds of Active In redient Per Acre • Without Herbicides All Pesticides

0.5 -l~.1

+

es e1 ee

11 73 75

n

79 81 83 Year

FIGURE 18.2 Pesticide Use Per Acre, 1965-1991

es

87 89 91

Measuring the Food Safety Risk of Pesticides

405

Regulation Pesticide regulation did not originate in concerns for the safety of the products but because pesticide users were wary of manufacturers' unregulated claims of their effectiveness. The purpose of the first statute governing pesticides, the Insecticide Act of 1910, was to protect farmers and other pesticide users from fraudulent claims. Federal pesticide registration, first introduced in the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) of 1947, also was intended to ensure product efficacy. USDA was given responsibility for administering the registration program. Under FIFRA, tolerances for pesticides used on food crops were set by the Food and Drug Administration on the basis of toxicity information provided by the manufacturer. Manufacturers seeking to register a pesticide needed only to present data demonstrating that, when the product was used as directed on the label, residues left on food did not exceed the established tolerance level. It was not until 1962 when Rachel Carson, formerly a marine biologist in the U.S. Fish and Wildlife Service, published her account of the safety and environmental risks of pesticide use in Silent Spring that protecting the environment and human health gained legitimacy as possible objectives of pesticide regulation. Still, these objectives were not included in federal statutes until ten years later. In 1969, President Nixon asked the Council on Government Reorganiution to design a new environmental regulatory program. The Council recommended consolidating research, standards setting, and pollution control enforcement within a single department. In 1970, EPA was created and the environmental regulatory functions were transferred from USDA, FDA, and the Department of Interior to the new agency. The creation of EPA was a first step in shifting the focus from product efficacy to emphasizing human health in pesticide regulation. However, while the creation of EPA signalled this change in focus of regulation, it was not accompanied by changes in the law affecting pesticide registration, nor were restrictions imposed on pesticide use at that time. The 1972 amendments to FIFRA were the first major changes in federal regulation of pesticides since 1947, and the first instance of federal restrictions on pesticide use. Several of the changes included in the amendments were related to food safety; in particular, the requirement that the costs and benefits of pesticide use, including adverse human health effects, be an explicit part of the registration process.

Residue Detection Technology Residue testing technology has changed dramatically since the early 1960s. The sensitivity of detection methods has improved significantly. In 1965,

Kelly Day, Betsey Kuhn, and Ann Vandeman

406

pesticide residues could be detected at .003 parts per million (ppm), while in 1982, they were detectable at .001 ppm. Technological advancements also expanded the set of detectable pesticides substantially. When the Total Diet Study began in 1961, less than 24 pesticides were detectable. In 1984, over 200 pesticides and related chemicals could be detected (Pennington and Gunderson 1987). Consequently, later pesticide residue measurements included a greater number of detectable pesticides. lO The number of pesticides detected each year has increased substantially since the beginning of the study (Figure 18.3). Unfortunately, it is unclear how much is due to improved residue detection and how much is from increases in the set of available pesticides. We would expect improved residue detection to bias the food safety measure downward early in the series, and upward later in the series.

Conclusion Several points can be concluded from this study. First, the process of pesticide risk assessment is in its beginning stages for economists, and is preliminary to actual valuation work. No doubt a project such as this one would

es

67 69 11 73 75

n

19 01 83

Year

as

01 89 91

FIGURE 18.3 Number of Pesticides Detected in the Total Diet Study, 1965-1991

Measuring the Food Safety Risk of Pesticides

407

benefit from the input of toxicologists. While there is a need to communicate the potential risks associated with pesticide use, the information to assess these risks is limited. Second, if the object of a study is to determine human health risks posed by pesticides, it is important that the researcher assess more than carcinogenicity. If the risk index is limited to a Q* rating, or a weight-of-theevidence-ranking, then only oncogenic or carcinogenic risks are being measured. Finally, in order for valuation to be meaningful, there must be an accurate determination of risk. Without proper assessment of toxicological data, we will not produce reasonable measures of the cost of current use or the benefits of pesticide use reduction.

Notes 1. The views in this paper are those of the authors and do not necessarily reflect those of the Department of Agriculture. This research has benefitted greatly from the helpful comments and technical expertise of Harry Vroomen, Fred Joutz, and Cynthia Tyler. 2. The Total Diet Study results have been published in some years by the FDA, in other years in the Journal of the Association ofAnalytical Chemists, or published jointly by both. Sources: Corneliussen 1972, Duggan et al. 1971, 1983, U.S. Food and Drug Administration 1988, 1989, 1990, 1991, 1992, Gartrell et al. 1985a, 198Sb, Gunderson 1988, Johnson et al. 1984, Podrebarac 1984. 3. Although the Total Diet Study has been conducted annually, the results for some years during the early and mid-1980s were combined. Annual observations were interpolated using the results from these combined years. 4. The basis for the Total Diet Study is a series of food lists from consumer dietary surveys. These food lists have changed over time in accordance with changes in consumption patterns. Detailed below are the studies and number of foods analyzed: 1964-1970: The USDA's 1955 Household Food Consumption Survey was used to determine the typical diet consumed by 16-19 year old males in the U.S. The diet included 82 foods and assumed a large calorie intake (4,200 calories/day) to determine maximum exposure to various chemicals. 1971-1982: The diet and food lists were modified according to the USDA's 1965 Household Consumption Survey. A diet of 120 foods was developed for 15-20 year old males, and the caloric content was reduced to 3,900 calories/day. Four regional diets were constructed from the data, representing the East, Central, South, and Western U.S. In addition to these changes, FDA added an infant and toddler diet. 1982: The diet was revised according to the USDA 1977-1978 Nationwide Food Consumption Survey and the National Center for Health Statistics Second National Health and Nutrition Survey of 1976-1980. The Total Diet Study also returned to a national diet, including 234 foods. The caloric intake for 14-16 year-old males was reduced to 2,677 calories per day. 5. The age-sex group definition varies slightly through the history of the study. Until 1971, the group was 16-19 year-old males. From 1971-1982, the group was 15-20

408

Kelly Day, Betsey Kuhn, and Ann Vandeman

year-old males. From 1982 onward, the age-sex group was defined as 14-16 year-old males. 6. Acceptable daily intake levels are based on a reference dose. Animal tests are conducted to establish No-Observed-Adverse-Effect-Levels (NOAEL), the maximum dose at which no adverse health effects are observed. NOAELs are adjusted for uncertainty and other factors (by multiplying by a factor of 10 to 100 or more) to arrive at a reference dose that is considered safe for human consumption. 7. An extensive discussion of the drawbacks of qualitative carcinogenic measures can be found in Recommendation for Improving Cancer RiskAssessment(Harvard Center for Risk Analysis 1992). Defenses for the qualitative measures can be found in Regulating Pesticides in Food: The Delaney Paradox (National Academy of the Sciences 1987) and in "Risk Assessment: Scientists Find Federal Funds are Misguided" (Environmental Policy Alert 1992). 8. See note 6. 9. Tolerances are the maximum legal level of a residue permitted on a particular food product. Tolerances are determined by evaluating the toxicity of a chemical assuming the pesticide is applied at the highest allowable rate and maximum lifetime exposure for the food consumer. Because the Total Diet Study measures actual residues on food prepared for consumption, and aggregates these residues based on the amount of each food consumed in a typical diet, we argue that the Total Diet Study residues are superior to tolerances as a measure of food safety risk. 10. The method of reporting pesticide residues has changed over the period studied. In earlier years, individual chemical metabolites were reported as separate residues. Later, metabolites were combined. We corrected for changes in the number of residues reported by aggregating the individual metabolites to correspond with the later combined lists. However, other unknown reporting procedures still may have biased our measure of the number of residues.

References Ames, B. N., R. Magraw, and L. S. Gold. 1987. Ranking Possible Carcinogenic Hazards. Science 236:271-280. Archibald, Sandra 0. and Carl K. Wmter. 1989. Pesticides in Food: Assessing the Risks. Reprint of Chapter 1 in Chemicals in the Human Food Chain, ed. C. K. Wmter, J. N. Seiber, and C. F. Nuckton. Davis, CA: University of California Agricultural Issues Center. Aspelin, Arnold L., Arthur Grube, and Robert Torla. 1992. Pesticide Industry Sales and Usage: 1990 and 1991 Market Estimates. Washington, D.C.: Environmental Protection Agency. Briggs, Shirley A. and the Rachel Carson Council. 1992. Basic Guide to Pesticides. Washington D.C.: Hemisphere Publishing Corporation. Carson, Rachel. 1962. Silent Spring. Boston, MA: Houghton Mifflin Company. Comeliussen, P. E. 1972. Residues in Food and Feed: Pesticide Residues in Total Diet Samples (VI). Pesticide Monitoring Journal 5:313-341.

Measuring the Food Safety Risk of Pesticides

409

Cropper, Maureen L., Bill Evans, Steve Beradi, Maria Soares, and Paul Portney. 1990. An Analysis of EPA Regulation of Food-Use Pesticides. Discussion Paper CRM 9004. Washington, D.C.: Resources for the Future. Day, Kelly A., Betsey Kuhn, and Ann M. Vandeman. 1993. Pesticide Regulation and Food Safety Risk. Western Economic Association International Conference. Lake Tahoe, NV, June 21-25. Duggan, R. E., G. Q. Lipscomb, E. L. Cox, R. E. Heatwole, and P. C. Kling. 1971. Residues in Food and Feed: Pesticide Residue Levels in Foods in the United States from July 1, 1963 to June 30, 1969. Pesticide Monitoring Journal 5:331-341. Duggan, Reo E., Paul E. Corneliussen, Mary B. Duggan, Bernadette M. McMahon, and Robert J. Martin. 1983. Residue Monitoring Data: Pesticide Residue Levels in Foods in the United States from July l, 1969 to June 30, 1976. Washington, D.C.: Food and Drug Administration and Association of Official Analytical Chemists. Environmental Policy Alert. 1992. Risk Assessment: Scientists Find Federal Funds are Misguided. February 5:36. Gartrell, M. J., J.C. Craun, D.S. Podrebarac, and E. Gqnderson. 1985a. Pesticides, Selected Elements, and Other Chemicals in Adult Diet Samples, October 1978September 1979. Journal of the Association of Official Analytical Chemists 60: 146161. Gartrell, M. J., J. C. Craun, D. S. Podrebarac, and E. L. Gunderson. 1985b. Pesticides, Selected Elements, and Other Chemicals in Adult Diet Samples, October 1980March 1982. Journal of the Association of Official Analytical Chemists 60: 146-161. Gunderson, Ellis L. 1988. Chemical Contaminants Monitoring: FDA Total Diet Study, April 1982-April 1984, Dietary Intake of Pesticides, Selected Elements, and Other Chemicals. Journal ofthe Association of Official Analytical Chemists 71: 1200-1209. Hammitt, James K. 1986. Estimating Consumer Willingness to Pay for Food-Borne Risk. R-3347-EPA, October. Santa Monica, CA: The Rand Corporation. Harvard Center for Risk Analysis. 1992. Annual Report, 1992. Cambridge, MA: Harvard School of Public Health. Horowitz, John K. and Richard T. Carson. 1990. Do Consumers Really Care About the Characteristics of Health Risks? Paper Presented at the American Agricultural Economics Association, Vancouver, British Columbia, August 4-8. Johnson, Roger D., Dennis D. Manske, Dallas H. New, and David S. Podrebarac. 1984. Pesticide, Metal, and Other Chemical Residues in Adult Total Diet Samples (XIII), August 1976-September 1977. Journal of the Association of Official Analytical Chemists 67:154-166. Lutz, Steven M., David M. Smallwood, James R. Blaylock, and Mary Y. Hama. 1992.

Changes in Food Consumption and Expenditures in American Households During the

1980's. Statistical Bulletin No. 849. Washington, D.C.: Economic Research Service and Human Nutrition Information Service. National Academy of Sciences, National Research Council. 1987. Regulating Pesticides in Food: The Delaney Paradox. Washington, D.C.: National Academy Press. National Academy of Sciences, National Research Council. 1989. Alternative Agriculture. Washington D.C.: National Academy Press. National Academy of Sciences, National Research Council. 1993a. Pesticides in the Diets of Infants and Children. Washington, D.C.: National Academy Press.

410

Kelly Day, Betsey Kuhn, and Ann Vandeman

National Academy of Sciences, National Research Council. 1993b. News Release, January 28. Washington, D.C.: National Academy of Sciences. Osteen, Craig 0. and Philip I. Szmedra. 1989. Agricultural Pesticide Use Trends and Policy Issues. Agricultural Economic Report No. 622. Washington, D.C.: Economic Research Service, September. Pennington, Jean A. and Ellis Gunderson. 1987. History of the Food and Drug Administration's Total Diet Study-1961 to 1987. Journal of the Association of Official Analytical Chemists 70:772-782. Podrebarac, David S. 1984. Pesticide, Metal, and Other Chemical Residues in Adult Total Diet Samples (XIII), October 1977-September 1978. Journal ofthe Association of Official Analytical Chemists 67:176-185. Roberts, Tanya and Eileen van Ravenswaay. 1989. The &onomics ofSafeguarding the U.S. Food Supply. Agriculture Information Bulletin No. 566. Washington, D.C.: Economic Research Service, July. Sachs, Carolyn, Dorothy Blair, and Carolyn Richter. 1987. Consumer Pesticide Concerns: A 1965 and 1984 Comparison. Journal of Consumer Affairs 21:103. U.S. Department of Agriculture, National Agricultural Statistics Service and Economic Research Service. 1991. Agricultural Chemical Usage, 1990 Vegetables Summary. Report No. AgCh1(91). Washington, D.C.: U.S. Department of Agriculture, June. U.S. Department of Agriculture, National Agricultural Statistics Service and Economic Research Service. 1992a. Agricultural Chemical Usage, 1991 Fruits and Nuts Summary. Report No. AgCh1(92). Washington, D.C.: U.S. Department of Agriculture, June. U.S. Department of Agriculture. 1992b. Agricultural Statistics, various years. Washington, D.C.: Government Printing Office. U.S. Environmental Protection Agency. 1992. Background Document 2: EPA

Approach for Assessing the Risks Associated with Chronic Exposures to Carcinogens.

Washington, D.C.: EPA. U.S. Food and Drug Administration. 1988. Food and Drug Administration Pesticide Program-Residues in Foods-1987. Journal ofthe Association ofOfficial Analytical Chemists 71:156A-174A. U.S. Food and Drug Administration. 1989. Food and Drug Administration Pesticide Program-Residues in Foods-1988. Journal ofthe Association ofOfficial Analytical Chemists 72:133A-152A. U. S. Food and Drug Administration. 1990. Food and Drug Administration Pesticide Program-Residues in Foods-1989. Journal ofthe Association of Official Analytical Chemists 73:127A-146A. U. S. Food and Drug Administration. 1991. Food and Drug Administration Pesticide Program-Residues in Foods-1990. Journal ofthe Association ofOfficial Analytical Chemists 74:121A-152A. U.S. Food and Drug Administration. 1992. Food and Drug Administration Pesticide Program-Residues in Foods-1991. Journal ofthe Association of Official Analytical Chemists 75:20-25.

19 Adding Nutritional Quality to Analysis of Meat Demand Guijing Wang and Wen S. Chem1

Numerous meat demand studies have been undertaken in the past several decades, providing many insights into the effects of conventional factors such as price and income on meat demand. However, other factors such as nutritional quality have been overlooked. Price variations across consumers owing to quality differences have not been well understood. This chapter attempts to complement existing studies by evaluating meat demand from a nutritional quality perspective. In addition, from public health perspectives such as health promotion and disease prevention, it is important to understand the nutritional aspects of food quality. In this chapter, nutritional quality indexes of ten nutrients including seven essential nutrients (niacin, vitamin-B6, vitamin-B12, phosphorus, magnesium, iron, and zinc) and three undesirable nutrients (fat, saturated fat, and cholesterol) in five meats (beef, pork, lunch meat, poultry, and fish) are derived and discussed using data from the 1987-88 Nationwide Food Consumption Survey (NFCS). Unlike the conventional studies of Marsballian demand analyzing the quantity and price relationship, this study focuses on the relationship between price and quality attributes. This relationship is specified as a multiple regression model by combining the approach of Houthakker (1952) and hedonic methodology. Additionally, the heterogeneous nature of a food is detected by the diagnostics procedure developed by Belsley (1991). One important econometric issue, related to multicollinearity among nutrient attributes, is also appropriately addressed for model estimation. The proposed methodology is important both theoretically and empirically. Traditional consumer theory assumes that commodities are homogenous and prices are exogenous to consumers. These assumptions may be severely violated if broadly aggregated survey data are used. If price variations in cross-sectional 411

412

Guijing Wang and Wen S. Chern

data are mainly due to commodity heterogeneity (quality variations), the classical consumer demand theory may not be applied. Thus, a well defined nutritional quality index and a specification of the price-quality relationship are particularly useful to understand consumers' food demand behavior. Furthermore, future efforts toward nutritional improvement are striving for food quality rather than quantity. The information provided by this approach is of great interest to consumers since they know what they are paying for and what is the right food combination. Health care professionals can use the information to design appropriate diet plans for consumers and to initiate nutritional education programs. Food producers can utili7.e the information to improve nutritional quality of food. Government food agencies can propose appropriate food polices and assistance programs according to consumer nutritional intake and nutritional quality information.

Approaches to Studying Meat Demand Meat and poultry consumption constitutes the largest proportion of the food budget in the United States. According to the NFCS, for an average household in 1987-88, meat expenditures (excluding poultry and fish) accounted for 21 percent of the budget of total food consumed at home, the largest budget share among 19 food groups covered in the survey. The expenditure share of poultry and fish combined was 11 percent. The significant budget allocation to meat, poultry, and fish explains why the demand for these food groups has been studied extensively in the past several decades. During the early 1960s, meat demand became less price elastic in the United States. By examining the demand relations for beef, pork, and broilers during 1949-1956 (period 1) and 1956-1964 (period 2), Tomek (1965) found that price elasticities were less elastic in period 2 than in period 1. More importantly, he pointed out that the change in elasticity may be due to a quality change in the product, although he did not define and incorporate quality factors in his study. In recent years, meat demand studies have focused on the flexible demand system and the impacts of demographic variables. Examples of these studies include Christensen and Manser (1977), Chalfant (1987), and Heien and Pompelli (1988). Although many of these studies mentioned product quality characteristics to provide useful insights in explaining meat consumption behavior, none incorporated meat quality into the demand analysis. Chalfant and Alston (1988) studied structural changes in meat demand, accounting for consumer taste changes. They argued that the inadequacy of prices and expenditures in explaining observed patterns of meat consumption was due to specification errors in demand models. Other studies, e.g., Bales and Unnevehr (1993) have found significant structural changes. One major reason for these conflicting results may be that the price used in classical

Adding Nutritional Quality to Analysis of Meat Demand

413

demand models is incapable of accounting for quality variations across consumers and over time. Alston and Chalfant (1991) argued further the importance of correct model specification. However, data on and measurement of such information as the nutritional quality of meat are essential. One may argue that consumers are ignorant about nutrients contained in meats and that the nutritional quality of meat is irrelevant as a determinant of meat demand. However, nutrients are the basic physical quality attributes of food. The nutritional quality of meat is associated with its price. Understanding the nutritional quality of meat will enable us to distinguish consumption responses resulting from changes in price and quality. Moreover, the nutritional quality analysis of food will provide useful information for evaluating the American diet. Houthakker was among the first to discuss commodity quality in analyzing consumer behavior. Houthakker (1952) defined "quality elasticities" as the proportionate rise in the average price per unit associated with a small proportionate rise in income. Cox and Wohlgenant (1986) and Deaton (1990) have made further contributions to the attempt to incorporate quality effects into theoretical and empirical food demand models. However, they did not define and measure food quality explicitly. More recently, a number of studies have investigated the impacts of product attributes and nutritional information on the demand for food. Brown and Schrader (1990) studied cholesterol information and shell egg consumption. Their cholesterol information index was examined further by Capps and Schmitz (1991) in discussing health and nutrition factors in food demand analysis, and by Yen and Chern (1992) in their study of fats and oils consumption. Schmitz (1992) investigated the impact of nutritional information on demand for dairy products. Results from these studies indicated that consumer health and nutrition concerns had significant effects on food demand. Although these studies emphasized the importance of food attributes and nutritional information, the nutritional quality of food has never been quantified and incorporated in food demand analysis.

Methodology

Developing an lllllex of the Nutritional Quality of Meat Food quality can be measured by nutritional content. Hansen et al. (1979) developed an index of nutritional quality (INQ). The index expresses the nutritional quality of a food by comparing the nutrients in the food to the calories it contains. In order to compute the INQ, nutrient standards must be determined first. According to Hansen et al. (1979), a nutrient standard is the amount of nutrient needed to meet a human's daily need in order to maintain good health.

Guijing Wang and Wen S. Chern

414

The Recommended Dietary Allowances (RDA) can serve as the standards. For example, the standard for vitamin-B6 is 2 mg for males above 15 years old according to the 10th edition of the RDA (National Academy of Science 1989). Based on the nutrient standard, the percentage of the standard of nutrient j in meat i, Wij, is calculated for each meat by: 2 (1)

w,.1J

=

lit· x 100% ' ...J! s ~

where Nij is the amount of nutrient j contained per unit of meat i, and Sj is the standard for nutrient j. The INQ of nutrient j in meat i is expressed as the ratio of its percentage standard of nutrient j, Wij• relative to its percentage standard of energy (calories), Wie· It is expressed as:

(2)

w,.

INQ;· = -2. . !I

~e

For each nutrient in a meat, an INQ can be computed. Note that the INQijs vary from one household to another because of the different compositions of meat and/or different meals or different ways of cooking with the same meat. For example, beef includes steak, roast, ground beef, etc., and the composition affects the amount of various nutrients and yields different INQs for different households. In fact, if the same meat contained the same nutrients, there will be exact multicollinearity among nutrients over households. From this approach, the INQ for energy in any meat is unity. The INQs for other nutrients may be smaller or larger than unity. For a desirable nutrient, a meat having an INQ of 1.0 or greater for that nutrient is of good quality in the nutrient. Otherwise the meat is of low nutritional quality. For undesirable nutrients such as fat and cholesterol, the INQs have an opposite meaning. Here, greater INQs for fat and cholesterol imply the meat is of lower quality.

The Price-Quality Relationship In a cross-sectional demand analysis, prices are usually derived by dividing aggregated expenditures by quantities purchased. Thus, prices are actually the average unit values. They are determined by the composition of the subcommodities in the groups. For example, beef price is the average price of variety beef such as steak and roast. Therefore, a large part of the variation in prices may be due to the choice of different qualities of a commodity. This implies that price cannot be regarded as exogenous to consumers (Pudney 1989). Houthakker (1952) pointed out that:

Adding Nutritional Quality to Analysis of Meat Demand

415

Apparent differences in prices may occur if goods are available in a variety of qualities, .. ., if they [consumer unit] do pay different amounts per unit for the same good the qualities bought must be different (pp. 2, 6).

Deaton (1990) also noted: Commodities are considered as collections of heterogeneous goods within which consumers can choose more or less expensive items, so that the unit value of a commodity, the price paid per physical unit, is a matter of choice (p. 282).

Therefore, specifying and estimating the price-quality relationship is important in understanding consumption behavior. Houthakker modeled the pricequality relationship as:

(3) where Pi is the average price of the ith good defined by dividing expenditure by quantity, X is per capita income, a and (j are parameters to be estimated, and e is the error term. A positive estimate of (j is expected, which implies that higher income consumers purchase higher quality commodities. As shown by Deaton (1990) food quality varies with income, household characteristics, and other determinants of food consumption. Following his approach, a price-quality function for this study is specified as:

(4)

P; = a; + fi;log(FC) +

:E 'Yifj + e; , j

where p is the same as defined as Houthakker; FC is the per capita total food cost (including food cost at- and away-from-home); z includes indexes of nutritional quality (INQ) of meat and demographic variables; a, (j, and 'Y are parameters to be estimated; and e is the error term. The inclusion oflNQs in the price-quality equation for a meat depends on the results of the collinearity diagnostic testing described in the following section. If an equation includes a fat or cholesterol index, the expected sign of the estimated coefficient is negative. This implies that a higher fat or cholesterol index in a meat indicates lower quality of the meat. The signs of estimated coefficients of other included INQs are indeterminate. The demographic variables include the educational levels, race, age, and status of household head, and food stamp participation. Because it is reasonable to assume that a female head of household is more responsible for meal planning, the household head refers to the female head except when the household does not have a female head. The age is the actual years of the

416

Guijing Wang and Wen S. Chern

household head. The specification of other categorical (0,1) variables are: ED 1 = 1 if household head completed less than 9 years of school, ED2 = 1 if household head completed 1-4 years of college, ED3 = 1 if household head completed more than 4 years of college, Black = 1 if household head is black, Food Stamp = 1 if household is currently receiving food stamps, and Female Headed = 1 if household is only female headed. All three educational levels are used in the regression. Note that the singularity problem is avoided by leaving out the households whose heads completed 10-12 years of school. The educational level is hypothesized to be positively related to food quality, i.e., consumers with higher educational levels consume higher quality food than less educated consumers. The age of household head is expected to have a positive estimated coefficient. This means that older household heads are more knowledgeable and aware of food quality. Food stamp recipients should have a negative coefficient since they are low income consumers. The signs of the coefficients on other demographic variables are indeterminate.

Diagnostic Test of INQs There are ten nutrients under investigation in this study. In the price-quality relationship, all ten INQs should be included. The relationship between the ten INQs represents the heterogeneity characteristics of a food. The more heterogenous a meat, the more quality variation there will be in the meat. Moreover, the inclusion of all the INQs may result in serious estimation problems related to multicollinearity. Although the INQs in a meat vary between households as mentioned previously, a high collinearity between them is often encountered. It is obvious that for a more homogeneous commodity, a higher collinearity between the INQs should be expected. For regression, inefficient estimates based on least squares will occur if the collinearity is severe. Therefore, conducting a diagnostic test for the INQs is necessary to understanding the heterogeneous nature of food and selecting an appropriate subset of INQs for estimating the price-quality relationship accurately. There are many procedures proposed and commonly employed to detect collinearity, such as examination of the correlation matrix and the principal component method. The procedure followed here is the scaled condition index and variance-decomposition proportions method of Belsley (1991). 3 This method provides an effective way of diagnosing the presence of and variables involved in collinearity. It is more satisfactory in detecting multicollinearity, although it is less commonly used in the literature (Kennedy 1987).

Adding Nutritional Quality to Analysis of Meat Demand

417

The Belsley method is composed of two major steps. The first determines the number of collinear relations by computing condition indexes. The second is to determine variable involvement by variance-decomposition proportions. These two steps are delineated below. Let X be a m by n data matrix, it is always possible to decompose X as:

x = usv'

(S)

where U is am by n and Vis.an by n matrix, both U and V are orthogonal matrixes, and S is a n by n diagonal matrix:

S=

The si's are called singular values of X and are unique. Let Y be the dependent variable and b be the vector of the estimated coefficients from a least squares regression of Y on X. The variance-covariance matrix of b can be written as: (6)

where a'1- is the estimated variance of the disturbance in the regression. Apparently, the singular values are related to the precision of coefficient estimates from the least squares method. The condition index is defined as (7)

8rnax

flk = - - , k = 1, ... n, sk

where Srnax is the maximum singular value. Belsley pointed out that the condition index is meaningful only if the columns of the data matrix are scaled to have equal length. As a matter of practice, scaling each column of X to have a unity length is preferable and employed in this study. He suggested that a value of the condition index above 30 indicates a moderate to strong collinearity condition. The number of large condition

418

Guijing Wang and Wen S. Chern

indexes ( > 30) indicates the number of collinear relationships among the columns of the data matrix, X. The next step is to determine how many and which variables are involved in a collinear relation. For this purpose, the variance-decomposition proportions are derived. The proportion of the variance of a regression coefficient k associated with a condition index j are computed by:

(8)

- wlr,j 'trjk - , wk

k = 1, ... ,n,

where

and where Vkj is the element of the orthogonal matrix V at the kth row and jth column, and Sj is the jth singular value. A large proportion of the variance of a coefficient being associated with a large scaled condition index is an indicator that the variable is involved in a collinear relationship with other variables. It is apparent that at least two variables are involved if a collinear relationship is present. This procedure is applied to detect the collinearity problems among the INQs for each meat. The INQs involved in highly collinear relationships are eliminated accordingly. Belsley shows that it is not always possible to determine from the variance-decomposition proportions alone exactly which variables are involved in collinear relationships. For example, there may not be two or more high variancedecomposition proportions associated with a high scaled condition index. For this situation, auxiliary regressions can be formed to show the near dependencies in greater detail. For example, for a subset of variables which are suspected to be involved in dependencies, regressing one variable on other remaining variables can signal variable involvement. This also shows whether the variations of the deleted variables can be represented by the remaining variables. In this chapter, the auxiliary regression was only formed for fat and saturated fat in beef, pork, lunch meat, and poultry because fat and saturated fat in these foods are the most important attributes consumers are concerned about. Unfortunately, fat and saturated fat are eliminated from the analysis based on the collinearity diagnosis.

Data Data used in this study is obtained from the 1987-88 Nationwide Food Consumption Survey (NFCS) sponsored by the U.S. Department of Agriculture.

Adding Nutritional Quality to Analysis of Meat Demand

419

The survey was conducted between April 1987 and March 1988. This crosssectional data set provides the food consumption, cost, and nutrient contents of the food on a weekly basis for 4,273 housekeeping households in the U.S. These households are defined as those with at least one member having ten or more meals from the household food supply during the survey week. Among these households, only those who consumed a meat are used for this study. A group of five meats (beef, pork, lunch meat, poultry, and fish) is defined as a food bundle. Lunch meat includes frankfurters, hot dogs, and other lunch meat except boiled ham and roast beef. The five meats are selected for this study because they contain similar nutritive values. Close substitutional and complementary relationships are also expected among these food items. The sample means of per capita food consumption, prices, and sample si7.es for different household groups are highlighted in Table 19.1. The price is derived by dividing food cost (expenditure) by quantity consumed. Interestingly, comparing the quantity and price between different household groups, the larger quantities are almost always associated with lower prices. For example, the consumers with lower educational levels consumed more beef and paid much lower unit price. 4 This is a strong indicator of quality differences among foods consumed by different household groups. The ten nutrients investigated in this study are fat, saturated fat, cholesterol, niacin, vitamin-B6, vitamin-B12, phosphorus, magnesium, iron, and zinc. 5 The selection of these nutrients is based on the fact that they are commonly contained in and mainly provided by the meat-poultry-fish food group in the American diet. According to the Human Nutrition Information Service (1990), more than 44 percent ofvitamin-B12, zinc, and niacin are provided by this food group for women aged 19 to SO years. More than 25 percent of vitamin-B6, phosphorus, and iron consumed by the women are derived from this group, while the group provides 18 percent of the women's magnesium consumption. Fat and cholesterol in these meats are probably the major attributes which are well recogni7.ed and of concern to consumers. The nutrient standards for these ten nutrients are calculated based on the 10th edition of Recommended Dietary Allowance (RDA). Since other data are converted into the base of a standard person, the nutrient standards are derived by averaging the RDAs for male and female. The RDAs (standards) for energy, niacin, vitamin-B6, vitamin-B12, phosphorus, magnesium, iron, and zinc are 2,30S (kcal), 16.3 (mg), 1.74 (mg), 2 (mg), 1,040 (mg), 314 (mg), 12.4 (mg), and 13.S (mg), respectively. The standard for cholesterol is daily consumption of 300 (mg) or less for adults (NAS 1989: SO). For fat, the standard is computed as follows. A gram of fat provides 9 calories of energy (Drummond 1989: 39). The amount of energy that comes from fat and saturated fat should be less than 30 percent and 10 percent of total energy, respectively (NAS 1989: 49). Using the standard of energy (2,30S kcal), the standards of fat and saturated fat are 76.8 (g) and 25.6 (g), respectively.

2.30

1.47 (210)

ED3

1.36 (227)

1.16 (2748)

1.56

1.94

1.92 (266)

1.75 (3328)

No

2.10

1.80

2.46

0.98 (171)

1.81

2.08

p

2.18

I

1.02 (878)

1.44 (393)

1.18 (2975)

Q

Pork

Yes

Food Stamp Participant:

2.03

1.59 (1105)

ED2

1.91

pb

1.71

I

1.87 (471)

1.76 (3594)c

~

EDl

Education:

Total Sample

Item

Beef

0.58 (2524)

0.65 (231)

2.14

1.87

2.31

0.45 (153)

1.90

2.11

p

2.25

I

0.55 (788)

0.65 (378)

0.59 (2755)

Q

Lunch Meat

1.70 (2826)

1.62 (239)

1.60 (199)

1.64 (971)

1.82 (437)

1.70 (3108)

Q

I

1.43

1.10

1.60 (158)

1.55

1.11

1.40

p

(727)

Poultry

TABLE 19.1 Average Food Consumption Per Capita Per Week and Prices for Different Household Groups

0.88 (2079)

1.01 (159)

0.91

0.90

0.89 (262)

0.89 (2238)

Q

I

Fish

2.97

2.30

3.38

3.14

2.65

2.92

p

~

1.41 1.40

1.90 (770) 1.63 (2338)

2.05 2.13

0.64 (616) 0.57 (2139)

2.05 2.08

1.16 (2313)

cNumbers in parentheses are sample sizes.

bp

= quantity (lb). = price ($/lb).

1.92

1.75 (2775)

Other

11Q

1.86

1.82 (819)

Female Headed

1.24 (662)

2.77 2.96

0.97 (488) 0.87 (1750)

3.01

0.81

1.48 (1970)

1.62 (2675)

2.16

0.57 (2406)

2.12

1.09 (2598)

1.93

1.72 (3190)

Other

Household Head Status:

2.26

1.48

0.94 (268)

2.18 (433)

1.77

0.74 (349)

1.74

1.77 (377)

1.70

2.07 (404)

Black

Race:

~ ......

422

Guijing Wang and Wen S. Chern

Analysis of Nutritional Quality The sample means of the INQs of meats for the total sample are reported in Table 19.2. All five products are of good quality in niacin and vitamin-B12. Fish is of good quality in all nutrients except cholesterol. Except fish, all meats are not of good quality in magnesium. Pork is not a good source of iron. All the meats except fish provide a high level of fat relative to their energy content. Pork has the highest index of fat while poultry has the largest index of cholesterol. While the INQ of saturated fat is higher than the INQ of fat for beef, pork, and lunch meat, the opposite is true for poultry and fish. This is consistent with the fact that poultry and fish contain a more balanced fat composition6 (Drummond 1989). Table 19.3 presents the average INQs of meats for different household groups by educational levels. 7 In terms of fat and saturated fat, households with a well-educated household head consume higher quality meat because their INQs of fat and saturated fat are lower than the other households. This is particularly true in the case of pork and poultry. Fat and saturated fat intakes are lower for well-educated consumers since they consume less meat with the lower INQs of fat and saturated fat (Tables 19.1 and 19.2). However, in terms of cholesterol, the above intake pattern does not hold. For other major nutrients, households with a well-educated head have a tendency to consume meats with higher INQs. TABLE 19.2 Means of Nutritional Quality Indexes of Meat and Related Products

Nutrients Energy Fat Saturated Fat Cholesterol Niacin Vitamin-86 Vitamin-812 Phosphorus Magnesium Iron Zinc 1

Beef (3594)1

Pork (2975)

Lunch Meat (2755)

Poultry (3108)

Fish (2238)

1.00 2.36 2.85 2.26 1.98 0.93 9.32 1.32 0.51 1.41 2.70

1.00 2.65 2.88 1.54 1.58 1.02 2.64 1.17 0.32 0.45 0.94

1.00 2.57 2.84 1.71 1.54 0.89 6.15 0.86 0.31 0.99 1.19

1.00 1.81 1.55 3.17 4.40 2.23 2.49 1.67 0.70 1.07 1.40

1.00 0.66 0.43 3.12 6.56 1.92 26.22 3.77 1.99 2.03 2.23

Numbers in parentheses are sample sizes.

0.30 0.97 1.17

0.36 0.49 1.03

0.33

0.45

0.29

0.42

0.53

1.46

2.75

2.27

9.39

1.33

0.52

1.43

2.72

2.26

1.94

0.90

9.27

1.30

0.50

1.38

2.69

Niacin

Vitamin-B6

Vitamin-Bl2

Phosphorus

Magnesium

Iron

Zinc

8

2.43

9.50

Numbers in parentheses are sample sizes.

0.86

1.03

0.88

0.99

0.96

1.36

1.62

1.39

2.00

2.01

0.96

1.20

2.68

1.04

1.55

1.50

2.27

3.19

3.16

3.22

26.61 29.46 3.74

22.75

3.95

2.63

1.15

6.90

6.33 0.89

5.95 0.83

2.88 1.33

2.43

3.31 1.41

1.40 1.43

1.26 1.20

2.13 2.22 1.71 1.10 1.09 1.04

0.74

0.65 0.33 1.10

2.00 2.05

1.94 0.74

1.76

1.54

0.94

1.04

3.15

2.49 2.49

1.02

0.32

1.83 1.92 2.00 2.42

2.38

2.00

0.97

0.93

0.87

7.01

1.17

6.78

5.68 4.68 4.70

3.95

1.71

1.61

3.23

1.45

3.18

3.30

0.40

0.64

1.00

ED3 (158)

1.80

0.40

0.53

0.61

1.00

ED2 (727)

1.74

1.86

1.49

1.49

1.62

2.76

0.80

1.00

EDI (262)

FIBh

1.68

2.81

.74

1.00

1.00

1.15

ED3 (199)

ED2 (971)

Poultry

1.61

2.88

Cholesterol

2.87

2.96

2.80 2.78

2.81

2.65

1.90

2.52

2.55

2.59

2.58

2.89

2.72

Saturated Fat

2.33

2.33

2.39

1.00

1.00

1.00

1.00

1.00

Fat

1.00

1.00

1.00

1.00

1.00

EDI (437)

ED3 (153)

ED2 (788)

Lunch Meat

EDI (378)

ED3 (171)

Energy

ED2 (878)

EDI (393)

ED3 (210)

Nutrients

ED2 (1105)

Pork

EDI (471)8

Beef

TABLE 19.3 Means of Nutritional Quality Indexes of Meats for Different Consumer Groups by Educational Levels

e

424

Guijing Wang and Wen S. Chern

The sample means of INQs of meat (not reported here) for food stamp recipients versus nonrecipients, black versus nonblack, and only female headed versus nononly female headed households have patterns similar to those for educational levels. In terms of fat and saturated fat, households receiving food stamps, blacks, and those that are only female headed consume lower quality commodities than others. In terms of cholesterol, however, this is not true. The amount of nutrient intake can also be detected by combining the INQs with the meat consumption shown in Table 19.1. Nevertheless, the meat consumption patterns shown by the INQs for different household groups may better indicate the quality differences of food consumed.

Reg~ion

Results

As the first step in the price-quality analysis, the conditioning diagnostics of the ten INQs plus the intercept are carried out for each of the five meats. To avoid repetition, we only delineate the procedure for the conditioning diagnostics for beef. The same procedure is applied to other meat items. The condition diagnostic results for an intercept term (vl) and ten INQs (v2vll) are presented in Table 19.4. Because a scaled condition index above 30 indicates a moderate to a strong collinear relation, there are about seven collinear relations (number of scaled condition indexes larger than 30). Clearly v2 (fat) and v3 (saturated fat) are nearly perfectly collinear. V7 (vitamin-B12) and v4 (cholesterol) also have a relatively large proportion of variance associated with the larger scaled condition indexes. As a first attempt, the first three variables are eliminated. The diagnostic procedure continues for the remaining eight variables (Table 19.5). There are still four large condition indexes. The five variables of v4 (cholesterol), v5 (niacin), v6 (vitamin-B6), v8 (phosphorus), and v9 (magnesium) are strongly involved as shown by the proportions of variances associated with the largest condition index. Four out of the five are eliminated. Vitamin-B6 is kept because it has a relatively smaller variance proportion associated with the largest condition index. The condition diagnostic results for the remaining four variables are presented in Table 19.6. Now all condition indexes are less than 30 and no further diagnostics are needed. Including these four variables in the price-quality equation of beef should provide precise estimates of price-quality relationships. After the conditioning diagnostics, the number of INQs left for beef, pork, lunch meat, poultry, and fish are 4, 6, 7, 5, and 11. The numbers indicate that the matrix of INQs of beef is the most ill conditioned while that of fish is the best conditioned. The ill condition of the INQs matrix implies that the nutrients contained in beef are closely correlated. Therefore, we can say that beef is the most homogenous and fish is the most heterogenous commodity.

0.001

0.000

0.006

0.006

0.000

0.014

O.lSl

0.6S7

0.105

17

26

32

44

S4

76

S3

139

309

I

0.993

0.001

0.001

0.002

0.001

0.000

0.001

0.000

0.000

0.000

0.000

v2

I

0.9Sl

0.02S

0.004

O.OlS

0.000

0.000

0.001

0.000

0.000

0.000

0.000

v3

I

0.126

0.270

0.001

O.S3S

0.042

0.026

0.000

0.000

0.000

0.000

0.000

v4

I

0.039

0.042

0.2SS

0.141

0.116

0.036

0.13S

0.02S

0.17S

0.001

0.000

vs

I

0.006

0.230

0.109

0.217

0.155

0.053

O.OlS

0.1S9

0.013

0.010

0.000

v6

I

Proportions of

0.230

0.30S

0.046

0.014

0.319

0.001

O.OS9

0.009

0.016

0.000

0.000

v7

I

0.03S

0.739

0.131

o.oos

0.022

0.027

0.033

0.000

0.000

0.001

0.000

vs

I

0.04S

0.404

0.420

0.000

0.032

0.030

0.367

0.009

0.136

o.oss

0.000

0.003

0.001

0.002

0.000

vlO

0.421

I

0.019

0.045

0.001

0.006

0.001

0.000

v9

I

0.001

0.002

0.074

0.442

0.197

0.117

0.007

O.lSS

0.000

0.001

0.000

vll

8vl = constant, v2 = fat, v3 = saturated fat, v4 = cholesterol, vS = niacin, v6 = vitamin-B6, v7 = vitamin-B12, vs = phosphorus, v9 = magnesium, vlO = iron, and vll = zinc.

0.000

s

vl 8

0.000

I

I

1

Scaled Condition Index

TABLE 19.4 Scaled Condition Index.es and Variance-Decomposition Proportions (INQs of Beef)

~

""

426

Guijing Wang and Wen S. Chern

TABLE 19.S Scaled Condition Indexes and Variance-Decomposition Proportions (INQs of Beet)

Scaled Condition Index 1 10 19 22

Proportions of vl 1 0.000 0.016 0.081 0.012 0.003 0.100 0.619 0.168

3S

38 66 87

v4

vs

v6

v8

v9

vlO

vll

0.000 0.003 0.006 0.000 0.011 0.013

0.000 0.027 0.247 0.076

0.000 0.044 0.019 0.269 0.096 0.126 0.104 0.342

0.000 0.002 0.000 0.002 0.113 0.018 0.031 0.833

0.000 0.001 0.018 0.006 0.079 0.043 0.212 0.641

0.000 0.002 0.002

0.000 0.000

O.S04

0.464

O.OS6

0.104 0.002 0.488

o.oos

0.112 0.383 0.3Sl 0.14S

o.oos 0.228 0.238 0.282

0.23S 0.011

8vl =constant, v4 =cholesterol, vS =niacin, v6 = vitamin-86, v8 =phosphorus, v9 = magnesium, vlO = iron, and vll = zinc.

TABLE 19.6 Scaled Condition Indexes and Variance-Decomposition Proportions (INQs of Beet)

Scaled Condition Index 1 9 17 28

Proportions of vl 1

v6

0.002 0.388 0.609 0.001

0.001 0.190 0.376 0.433

vlO

vll

0.000

0.001 0.003 0.380 0.616

o.oos

0.03S

0.960

8vl = constant, v6 = vitamin-86, vlO = iron, and vll = zinc.

Table 19. 7 presents the regression results of price-quality functions for the five meats. The inclusions of INQs are based on the diagnostic results. All the five equations include the same set of economic and socio-demographic variables. Statistically, the majority of estimated parameters are significant at the 5 percent level in all the five equations. The R2's range from 0.20 to 0.33, which are reasonable for an analysis of cross-sectional data. By and large, the significant estimates have expected signs.

427

TABLE 19.7 Regression Results of Price-Quality Functions

Pork

Lunch Meat

Poultry

Parameter

3542)8

2925)

2711)

3062)

Fish (N = 2205)

Intercept

0.617* (5.70)b

1.200* (9.19)

0.818* (6.19)

-0.587* (-4.24)

0.240 (0.72)

Beef

(N =

(N =

(N =

(N =

Fat

0.713* (2.64)

Saturated Fat

-0.626 (-1.51) -0.602*

Cholesterol (~.4)

Niacin Vitamin-B6

-0.656* (-18.03) -0.076 (-1.41)

1.626* (19.56)

-0.206* (-2.60)

0.178* (7.44) 0.292* (24.21)

1.059* (12.64)

Vitamin-B12

-0.031** (-2.27) 0.068 (1.43)

-0.180* (-12.98)

Phosphorus

-0.009* (-4.41) -0.153* (-2.70)

Magnesium

4.219* (14.20)

1.338* (4.80)

0.158** (2.57)

Iron

-0.184** (-1.84)

0.613* (3.09)

0.214• (5.08)

0.851* (9.43)

0.188* (4.04)

Zinc

-0.363* (-8.68)

-0.460* (-3.70)

0.048 (0.64)

-0.151* (-2.64)

0.016* (5.14)

Log(FC)

0.248* (10.91)

0.200* (7.45)

0.213• (7.52)

0.157* (5.45)

0.542* (8.95)

(continues)

428

Guijing Wang and Wen S. Chern

TABLE 19.7 (continued)

Parameter EDl

Beef (N =

3542). -0.047 (-1.16)

Pork (N =

Lunch Meat (N = 2711)

(N = 3062)

-0.120.. (-2.52)

-0.150* (-3.14)

-0.038 (-0.76)

2925)

Poultry

Fish (N =

2205)

-0.0004 (-0.004)

ED2

0.094* (3.25)

0.085•• (2.49)

0.086.. (2.43)

0.001 (0.01)

0.187 .. (2.43)

ED3

0.295* (5.39)

0.278* (4.32)

0.133 .. (2.00)

0.066 (0.99)

0.387* (2.87)

Black

-0.118* (-2.86)

-0.186* (-3.90)

-0.159* (-3.34)

-0.240* (-4.89)

-0.794* (-7.06)

0.004* (3.60)

-0.002·· (-2.12)

0.003 (1.39)

Age

0.002· (2.75)

0.002·· (2.26)

-0.109•• (-2.14)

-0.080 (-1.34)

-0.033 (-0.56)

Female Headed

0.036 (1.12)

0.055 (1.45)

0.039 (1.00)

R2

0.25

0.20

0.27

Food Stamp

-0.043 (-0.68) 0.149* (3.77) 0.33

-0.190 (-1.37) 0.057 (0.66) 0.22

Note: The superscripts • and •• correspond to the levels of statistical significance of 1 percent and 5 percent, respectively. •N denotes the sample size. bNumbers in parentheses are t-statistics.

The coefficients on the INQ of cholesterol for pork and lunch meat are negative and statistically significant, as expected. The parameters of the fat and cholesterol indexes for fish have unexpected positive signs and are statistically significant. Because the fat index in fish is very low, consumers may not be concerned about the fat content as seriously as in other meat products. The low fat index may mislead consumers' perceptions of the cholesterol content in fish. The signs of the estimated coefficients of other INQs are mixed. If we look at

Adding Nutritional Quality to Analysis of Meat Demand

429

the specific major source of the respective nutrients, a meaningful explanation can be given. For example, if the expensive beef item bas a high INQ of vitamin-B6, the positive coefficient of the vitamin-B6 INQ in beef is expected. Similarly, if the cheap (low quality) beef item bas a high zinc INQ, the estimated coefficient of the zinc INQ in beef should be negative. Across the five meats examined in this study, the logged per capita total food cost bas a significant and positive coefficient estimate. These results imply that high income households do pay more for a unit of food consumed. Although the higher price may reflect the services purchased, such as packaging and shopping environment, it is more likely that quality, such as boneless, welltrimmed beef, is the major reason for higher prices. The regression results indicate that the higher the educational level achieved by the household head, the higher is the quality of food consumed. Poultry is an exception with no significant estimates associated with any of the three educational levels. By racial groups, blacks consume cheaper meat than nonblacks, ceteris paribus. Households headed by an elderly person consume higher quality meat except poultry and fish, which have negative and insignificant estimates. Food stamp program participation and household head status do not affect the quality of food consumed as much as other demographic variables. However, households receiving food stamps tend to consume cheaper meat. This pattern may reflect their behavior as low income consumers. Except for fish, all the price-quality equations do not include the INQ for fat or saturated fat due to collinearity problems. To verify whether the other nutrients included in the price-quality equations can represent the variation of fat or saturated fat, two auxiliary regressions were formed for each meat except fish. The auxiliary regressions are performed by regressing the INQ of fat or saturated fat on the selected INQs for each meat. The regression results are reported in Table 19.8. Almost all the coefficient estimates are statistically significant at the 5 percent level. The R2's are very high for pork, lunch meat, and poultry. The R2 for beef is 0.51 although only three INQs are included. These results suggest that the sets of selected INQs are representative of the INQs of fat and saturated fat in these meats. The explicit relationships between fat and prices of meat are estimated by replacing the set of selected INQs by the INQ of fat or saturated fat in pricequality functions. Because the INQs of fat and saturated fat are highly correlated, only one of them can be used in the regression. Selected regression results are reported in the bottom rows of Table 19.8. The parameters for other economic and demographic variables are consistent with the previous results and, therefore, are not reported. Both INQs of fat and saturated fat are all significantly and negatively related with prices for all four meats. This implies that a higher INQ of fat or saturated fat is associated with lower quality of these food items. Consumers are willing to pay a higher price for less fat and saturated fat content in beef, pork, lunch meat, and poultry.

-

-

-

Magnesium

-0.388 (-16.4)

-

-0.321 (-17.4)

Vitamin-B12

Vitamin-B6

-

-

Niacin

3.713 (204.4)

INQ2

-

I

-

2.991 (211.5)

INQ1

Cholesterol

Intercept

Auxiliary Regression:

Item

Beeth

-0.876 (-30.2)

-

-0.098 (-12.7)

-

-0.160 (-17.6)

3.495 (476.9)

INQ1

I

Pork

-1.445 (-35.5)

-

-0.087 (-7.97)

-

-0.102 (-8.05)

3.851 (374.7)

INQ2

-0.825 (-25.7)

-0.778 (-15.3)

-

-

-0.129 (-13.0)

-0.158 (-23.8)

3.578 (287.7)

INQ2

-0.437 (-28.6)

I

-0.194 (20.1)

-0.099 (-16.0)

0.015 (3.58)

3.067 (391.3)

INQ1

Lunch Meat

TABLE 19.8 Results of Auxiliary Regressions and Selected Regression Results of Price-Quality Functions•

0.006 (2.96)

-0.196 (-115.8)

3.270 (274.2)

INQ1

I

INQ2

0.006 (2.90)

-0.162 (-99.9)

2.72 (237.8)

Poultry

~

c

0.14

-0.802 (-14.0)

0.13

-0.582 (-13.3) 0.15

-0.649 (-16.8) 0.15

-0.535 (-17.3) 0.13

-1.000 (-15.0)

0.87

0.10

-0.484 (11.5)

0.87

0.30

-1.514 (-31.7)

0.87

0.31

-1.845 (-32.5)

0.82

a.rbe significance levels of parameter estimates are not indicated because all parameters except one in INQl of poultry are statistically significant at the 5 percent level. Fish is not included in this table because of its very low fat and saturated fat INQs. bINQl and INQ2 are the INQs of fat and saturated fat, respectively. For the auxiliary regression, they are the dependent variables while they are independent variables in the price-quality equations.

R2

Parameter

Price-Quality Function:

0.95

0.51

R2 0.96

-0.372 (-46.7)

-0.435 (-52.4)

0.220 (16.3)

0.016 (1.82)

0.148 (8.74)

-0.024 (-1.96)

-0.040 (-3.35)

-0.023 (-2.44)

Zinc

0.53

0.044 (3.49)

-0.002 (-0.18)

0.087 (11.3)

0.038 (7.93)

-0.903 (-33.4)

-0.425 (-22.1)

-0.283 (-9.90)

-0.193 (-8.65)

Iron

~~

432

Guijing Wang and Wen S. Chem

Conclusions This chapter proposed an alternative method to analyze consumer demand for meat. Incorporation of well-defined nutritional quality indexes provides useful insights into nutrition valuation. The results are useful to consumers in making food choices, to health professional and government agencies in initiating consumer education and food assistance programs, and to the meat industry in improving product quality to meet consumers' needs. The proposed methodology can be applied to other food items. The derived INQs of ten nutrients in five meats show that no meat is good in all nutritional aspects. This finding suggests that the optimal combination of nutrients can only be achieved through a combination of meats. Since "eat a variety of food• is usually the first dietary recommendation suggested by health professionals, a variety of meats may be a part of it. The varying patterns of INQs across different household groups and meat products suggest strong quality variations in meat consumed. The fat and saturated fat consumption of households with a well-educated, nonblack household head are lower than for other households because they consume less meat and the meat they consume has lower INQs of fat and saturated fat. The conditioning diagnostic results proved that some foods, such as fish, are more heterogenous in nutritional quality than other foods, such as beef. This finding has important implications for applied demand analysis. For crosssectional studies, the estimated demand parameters will be biased for a heterogeneous commodity if the quality effect is not accounted for. The regression results indicate that the price of a meat is associated with its nutrition attributes. Higher price may not mean less quantity demanded but higher demand for quality. This is somehow a controversy in traditional demand analysis. The significance and expected signs of the estimates also verify that the procedure used in this study for selecting INQs for price-quality functions is appropriate. Consumers have a negative valuation on fat, saturated fat, and cholesterol in all meat products except fish. The demand for a meat is affected by its fat and saturated fat content.

Notes 1. 2. labels, 3.

The comments of Julie A. Caswell and Ping Zhang are much appreciated. The Wij is similar to the percent daily value of nutrients used in the new food which is based on a 2,000 calorie diet. For the strengths and weaknesses of other approaches, the reader is referred to

Belsley (1991).

4. The sample statistics of meat consumption for households in the base educational

Adding Nutritional Quality to Analysis of Meat Demand

433

category are not reported due to space considerations. The patterns for these households are very consistent with the reported three categories. 5. Protein is not included for analysis based on the belief that the American diet is over-nutritious in protein. 6. A food is called a balanced source of saturated and unsaturated fat if it contains a balanced proportion of saturated and unsaturated fats. 7. The INQs for households in the base educational category are not reported due to space considerations. The patterns for these households are very consistent with the reported three categories.

References Alston, J.M. and J. A. Chalfant. 1991. Can We Take the Con Out of Meat Demand Studies? Western Journal of Agricultural Economics 16(1):36-48. Belsley, D. 1991. Conditioning Diagnostics. New York, NY: John Wtley and Sons, Inc. Brown, D. J. and L. F. Schrader. 1990. Cholesterol Information and Shell Egg Consumption. American Journal of Agricultural Economics 72:548-555. Capps, 0., Jr. and J. D. Schmitz. 1991. A Recognition of Health and Nutrition Factors in Food Demand Analysis. Western Journal ofAgricultural Economics 16(1):21-35. Chalfant, J. A. 1987. A Globally Flexible, Almost Ideal Demand System. Journal of Business and Economic Statistics 5(2):233-248. Chalfant, J. A. and J. M. Alston. 1988. Accounting for Changes in Tastes. Journal of Political Economy 96:391-410. Christensen, L. R. and M. E. Manser. 1977. Estimating U.S. Consumer Preferences for Meat with a Flexible Utility Function. Journal of Econometrics 5:31-53. Cox, T. L. and M. K. Wohlgenant. 1986. Price and Quality Effects in Cross-Sectional Demand Analysis. American Journal of Agricultural Economics 68:908-919. Deaton, A. 1990. Price Elasticities from Survey Data. Journal of Econometrics 44:281-309. Drummond, K. E. 1989. Nutrition/or the Food Service Professional. New York, NY: Van Nonstrand Reinhold. Bales, J. S. and L. J. Unnevehr. 1993. Simultaneity and Structural Change in U.S. Meat Demand. American Journal of Agricultural Economics 75:259-268. Hansen, G. R., B. W. Wyse, and A. W. Sorenson. 1979. Nutritional Quality Index of Food. Westport, CT: The AVI Publishing Company, Inc. Heien, D. and G. Pompelli. 1988. The Demand for Beef Products: Cross-Section Estimation of Demographic and Economic Effects. Western Journal ofAgricultural Economics 13(1):37-44. Houthakker, H. S. 1952. Compensated Changes in Quantities and Qualities Consumed. Review of Economic Studies 19:155-164. Human Nutrition Information Service. 1990. Good Sources ofNutrients. Washington, D.C.: U.S. Department of Agriculture, January. Kennedy, P. 1987. A Guide to Econometrics, Second Edition. Cambridge, MA: The MIT Press.

434

Guijing Wang and Wen S. Chem

National Academy of Sciences, National Research Council. 1989. Recommended Dietary Allowances, 10th ed. Washington, D.C.: National Academy Press. Nationwide Food Consumption Survey 1987-88. Washington, D.C.: Human Nutrition Infonnation Service, U.S. Department of Agriculture. Pudney, S. 1989. Modeling Individual Choice: The Econometrics of Comers, Kinks, and Holes. New York, NY: Basil Blackwell, Inc. Schmitz, J. 1992. The Impact of Nutritional Information on the Demand for Dairy Products. Paper Presented at the 1992 Annual Meeting of the Western Agricultural Economics Association. Colorado Springs, CO. Tomek, W. G. 1965. Changes in Price Elasticities of Demand for Beef, Pork, and Broilers. Journal of Farm Economics 47(3):793-802. Yen, S. T. and W. S. Chem. 1992. Flexible Demand Systems with Serially Correlated Errors: Fat and Oil Consumption in the United States. American Journal of Agricultural Economics 74(3):689-697.

20 Irradiation and Food Safety: Consumer Attitudes and Awareness Sukant K. Misra, Stanley M. Fletcher, and Chung L. Huang Food irradiation offers many advantages to producers, as well as consumers. Irradiation of food promises an improved sanitation level, extended food shelf life, safe transport of produce, replacement of chemical fumigants, and a reduction of mold (Schutz et al. 1989, Bruhn et al. 1986, Pszcmla 1990). Food irradiation technology, on a macro scale, has the potential to lower health care costs due to less foodbome illness (Roberts 1985), expand export business, and ease world hunger through the reduction of spoilage and waste (Diehl 1983). Opponents of irradiation claim that irradiation will make food radioactive, will reduce levels of essential nutrients, will help to conceal food contamination, will pose serious occupational and environmental ha7.ards, and in general will increase risks to public health (Pszcmla 1990). Conflicting claims by advocates and opponents of irradiation have generated intense controversy and have made irradiation a political and psychological issue. It is still uncertain whether the public will accept irradiated food despite comprehensive research and endorsement by major health organiz.ations, international expert committees, and scientific societies. The U.S. Food and Drug Administration (FDA) has approved irradiation at certain doses for fruits and vegetables, dry or dehydrated herbs, spices, seeds, teas, vegetable seasonings, for the Trichinae parasite in pork, for sprouting in white potatoes, and for disinfecting wheat and wheat flour (Pszcmla 1990). Recently, the FDA approved irradiation to control bacteria such as Salmonella, Listeria, and Campylobacter in fresh and frozen uncooked poultry products. The poultry industry has not been willing to incur the financial risk of investing in irradiation technology because of perceived consumer resistance to irradiated chicken (Feedstuffs 1992). Pork processors and the seafood industry have also been wary of the technology due to perceived consumer resistance. 435

436

Sukant Misra, Stanley Fletcher, and Chung Huang

Distrust of irradiation technology is allegedly generated by consumer misconceptions. Whether misconceptions exist or not, it is important that the public's attitudes and degree of awareness of the benefits of irradiation be studied thoroughly. The public's attitudes and awareness of the benefits of irradiation are crucial to consumer acceptance of irradiated food, and consumer acceptance is critical to the viable adoption of this technology and a realization of the advantages it offers. An understanding of the public's perception and judgement of irradiated food will aid food processors in deciding whether to incur the financial risk of investing in this technology. It will aid government agencies, such as FDA, in evaluating existing policies and in devising new regulations which will reconcile scientific and social forces. Further, it will assist in developing and implementing effective consumer education programs aimed at stressing the positive aspects of irradiation. The objectives of this chapter are to analym consumers' attitudes toward irradiated food, to examine the extent of consumers' knowledge and awareness of the process of irradiation, and to identify the several facets which should be included in an effective consumer education program.

Survey Design The consumer panel maintained by the Consumer Information Management System (Huang and Misra 1990) of the University of Georgia was used to administer a survey to accomplish the above objectives. Panel members represent a stratified sample (by income) of Georgia consumers. A mail survey among 500 panel households was conducted during the Fall of 1992 (contact the authors for a copy of the complete survey). The survey resulted in 236 returned questionnaires representing a response rate of 47 percent. Several questions on a variety of topics concerning irradiation were asked of the panel participants. Though no attempt was made to educate consumers about the irradiation technology, definitions of some of the terms frequently used in the questionnaire were provided. The questionnaire was carefully designed to minimim its effect on consumer responses and to collect information that reflects consumer's own attitudes. Questions related to consumer perceptions about irradiation and other food safety concerns were asked first to avoid any input (education) the survey instrument may have given. A pre-selected list of seven concerns was presented to determine the relative consumer risk perception about several food safety concerns. Respondents were asked to express their perception about these food safety concerns on a scale of "no problem" to "extremely serious problem." The seven food safety concerns were: pesticide residues, animal drug residues, growth hormones, food additives, bacteria, irradiation, and naturally-occurring toxins. The four most frequently cited disadvantages of irradiation were presented

Irradiation and Food Safety: Consumer Attitudes and Awareness

437

to focus more precisely on consumers' attitudes toward irradiation, and panelists were asked to express their concerns about these claims. The four statements were: a. b. c. d.

The possibility of food becoming radioactive due to irradiation. The possibility of reduced levels of nutrients due to irradiation. The risk of workers at irradiation facilities becoming ill. The risk of environmental pollution due to irradiation facilities.

Questions were asked directly on the necessity of irradiation for specific foods. These foods were fruits, vegetables, poultry products, pork products, beef products, and seafood. Panelists were also asked whether they would buy food that was treated with radiation. Those who expressed an unwillingness to buy irradiated food and those who were undecided were asked to indicate specific reasons for their response from a list of four pre-selected reasons. Another set of questions dealt with the extent of consumer knowledge and awareness of the process of irradiation. Panel members were initially asked to self-evaluate their knowledge of the irradiation process. In addition, ten factual statements were presented with "true," "false," and "don't know" answer options. A final set of questions was developed to assist in the formulation of an effective consumer education program. Panel members were presented with a number of arguments in favor of irradiation and the advantages of irradiation. Respondents were asked to evaluate each statement separately and to indicate how the evaluation affected their perception of the process of irradiation. They were also asked to identify the most persuasive argument and/or advantage statement in favor of the irradiation process. Finally, panel members were asked to identify the sources on which they depend for information and to express their confidence in comments on food safety by various interest groups.

Survey Results

Respondent Characteristics Table 20.1 presents the socioeconomic and demographic information on the sample used in this study. It was not swprising that the majority of the respondents were female, 61 percent, since the primary family food shopper completed the questionnaire. The racial composition of Georgia is approximately 74 percent white as compared with over 82 percent white households represented· in the sample. The mean age of the respondents was close to SO years and about 56 percent of the sample respondents had at least some college education. More than one-quarter (29 percent) of the respondents

438 TABLE 20.1 Sample Characteristics of Survey Respondents

Characteristics

White

Nonwhite

Total

---------- %----------GENDER Male Female

33.5 49.8

5.4 11.3

38.9 61.1

AGE Less than 35 years 36-45 years 46-65 years More than 65 years

14.7 21.7 29.0 18.0

4.1 4.1 6.5 1.8

18.9 25.8 35.5 19.8

EDUCATION Noncollege College

32.4 50.9

11.3 5.4

43.7 56.3

MARITAL STATUS Married OtherA

63.1 20.3

8.6 8.1

71.6 28.4

HOUSEHOLD INCOME Less than $20,000 $20,000-$29,000 $30,000-$39,000 $40,000-$49 ,000 $50,000 or more

20.2 14.4 13.5 10.6 25.0

9.1 2.4 2.4 0.5 1.9

29.3 16.8 15.9 11.1 26.9

HOUSEHOLD SIZE 1 person 2-4 persons 5 or more persons

13.1 62.6 7.7

3.6 9.0 4.1

16.7 71.6 11.7

PLACE OF RESIDENCE Urban Rural

42.1 41.2

11.3 5.4

53.4 46.6

8This group includes unmarried, divorced/separated, and widowed respondents.

I"adiation and Food Safety: Consumer Attitudes and Awareness

439

had annual household incomes of less than $20,000. Over a third (38 percent) had incomes of $40,000 or more. The sample tended to be demographically upscale with better educated and higher income consumers slightly overrepresented in comparison with census statistics.

Attitudes Toward Food Irradiation

Relative Risk Percepti.on. The list of seven food safety concerns served to focus respondent perceptions on the seriousness of problems associated with the listed concerns. The seven-point scale of •no problem• to "extremely serious problem• was converted into three groups: little or no problem, moderate problem, and serious or extremely serious problem. Table 20.2 presents the information on consumer perceptions toward these concerns. The results indicate that pesticide residues were perceived to be the greatest safety threat, followed by growth hormones, animal drug residues, bacteria, food additives, irradiation, and naturally occurring toxins. This finding is consistent with findings by the Food Marketing Institute (FMI 1991), where pesticide/herbicide residues, antibiotics, and hormones were perceived as more serious safety threats than irradiation. Bruhn et al. (1986) also found that consumers are more concerned about chemical sprays and preservatives than irradiation. Consumer responses, isolated for irradiation, indicate that approximately 40 percent of the respondents perceived irradiation to be either a serious or extremely serious problem. This result compares with the national results of Wiese Research Associates (1984) of 42 percent, FMI (1988) 36 percent, Brand Group (1986) 27 percent, and Bruhn et al. (1988) 29 percent for the Western region. TABLE 20.2 Consumer Perceptions of Suggested Food Safety Concerns

Safety Concern

Little or No Problem

Moderate Problem

Serious Problem

----------- %----------Pesticide residues Animal drug residues Growth hormones Food additives Bacteria Irradiation Naturally occurring toxins

19.1 21.9 20.9 25.0 25.6 34.6 54.9

26.7 27.4 27.4 32.4 24.6 26.1 22.8

54.2

50.1

51.6 42.6 49.8 39.3 22.3

Sukant Misra, Stanley Fletcher, and Chung Huang

440

Concem About Ramifications of Irradiation. Another measure of consumers' concerns about irradiated food is their perception of the ramifications of the technology. Opponents claim that irradiation will make food radioactive, will reduce levels of essential nutrients, and will pose serious occupational and environmental b.amrds. Respondents were asked to report their concern about these alleged ramifications of irradiation on a scale of "not concerned" to "extremely concerned." Results show that consumers are seriously concerned about the ramifications of the irradiation technology (Table 20.3). Are panelists inconsistent'? In Table 20.2, it was observed that consumers expressed a relatively low level of risk perception for irradiation. Yet, they expressed a rather high level of concern about specific irradiation ramifications. One possible explanation may be that though irradiation is not perceived as a more serious safety threat than chemicals and preservatives, its potential ramifications are still under close scrutiny. The risk of environmental pollution due to irradiation was listed as a matter of high or extremely high concern by over 65 percent of the respondents. Approximately 64 percent expressed high or extremely high concern for possible occupational b.amrds, 60 percent expressed similar concern on the possibility of food becoming radioactive, and about 53 percent were concerned about the possibility of reduced levels of nutrients due to irradiation. Interestingly, more consumers were concerned about possible occupational and environmental ha7.llrds due to food irradiation than to potential health risk. One possible explanation for this finding may be that consumers tend to link irradiation with the nuclear industry. Mention of irradiation technology, possibly, triggers thoughts of nuclear accidents and associated nuclear waste disposal in the minds of consumers. Is Irradiation Necesmry? Table 20.4 indicates that consumers perceive there is a necessity for irradiation of specific foods. About 60 percent of the respondents perceived that the irradiation of pork is very necessary. This was TABLE 20.3 Consumer Concern About Specific Irradiation Ramification

Ramifications

Little or No Concern

Moderate Concern

Extremely/ High Concern

------------ %--- ----------

Food becoming radioactive Reduced levels of nutrients Environmental risk Occupational hazard

23.8

20.5

13.1 14.8

16.0 26.2 21.8 21.4

60.2

53.3

65.1 63.8

441

Irradiation and Food Safety: Consumer Anitudes and Awareness TABLE 20.4 Consumer Perception About the Necessity of Irradiation

Food Product

Not Necessary

Somewhat Necessary

Very Necessary

-------------- %-------------Fruits Vegetables Poultry Pork

Beef

Seafood

34.0 31.1 12.9 12.9 16.2 14.7

40.0 42.0 30.0 27.5 38.4 27.2

26.0 26.9 57.1 59.6 45.4 58.1

followed by about 58 percent and 57 percent strongly recommending irradiation of seafood and poultry, respectively. Irradiation was also thought to be very necessary for beef products by about 45 percent of the respondents. The irradiation of fruits and vegetables was not perceived to be a high priority by the responding consumers. These results suggest that more consumers perceived irradiation as necessary for seafood and meat products, particularly in comparison to fruits and vegetables. Ironically, the poultry, pork, and seafood industries are hesitant to adopt irradiation technology, whereas the fruit and vegetable industries are proceeding with irradiation. The fruit and vegetable industries have been persistent in exploring the technology, despite opposition from consumer activists. The Vindicator food irradiation facility in Florida has continued to expand its operation to include citrus, fruits, spices, tomatoes, and onions (Turcsik 1992). Does a perception toward the necessity of irradiation translate into a willingness to purchase irradiated food? To assess this critical question, respondents were asked specifically whether they would buy food that was treated with radiation. Only 13 percent of these respondents expressed an unwillingness to buy irradiated food. About 31 percent of the respondents were willing to purchase irradiated food while a majority (56 percent) were undecided. This is consistent with the Brand Group (1986) study classification of "rejecters," "undecideds," and "acceptors" of irradiated food. This latter study estimated, on the basis of a national sample, that 5-10 percent can be classified as rejecters, 55-65 percent can be classified as undecideds, and 25-30 percent as acceptors. This result suggests that consumers have not changed their willingness to purchase irradiated food during the last 5 years. This may be simply because the issue of irradiation has not really been before the public.

442

Sukant Misra, Stanley Fletcher, and Chung Huang

Respondents who were either unwilling to purchase irradiated food or undecided were then asked to specify reasons for their response. The list of four pre-selected reasons was: a. b. c. d.

Is harmful and may lead to health complications. Poses occupational hazards for those involved. Poses serious environmental hazards. Not sure whether the process is safe.

In response, an overwhelming 93.5 percent of the respondents indicated that they are either unwilling to purchase irradiated food or are undecided primarily because they are not sure whether the process is safe. This result indicates that though the consumers perceive irradiation as necessary for certain foods, they demand more information about its safety.

Awareness About Food Irradiation Seif-Evaluation of Awareness. Respondents were asked whether they had heard of the irradiation process for preserving food prior to this survey. About 54.S percent of the respondents indicated "yes," with the remainder indicating "no." This result compares with the nationwide awareness percentages presented by Schutz et al. (1989) of 59.1 percent, Brand Group (1986) of 66 percent, and Wiese Research Associates (1984) of 23 percent, and the regional results of Bruhn et al. (1988) of 45 percent. Hearing about food irradiation does not necessarily provide a good measure of consumers' knowledge about the irradiation process. Respondents who had heard about irradiation were asked to rate their knowledge by choosing a statement from a pre-selected list of three statements. The three statements were: a. I am sufficiently informed about the irradiation process. b. I am somewhat informed about irradiation but do not feel comfortable to make an accurate assessment. c. I have heard about the irradiation process but do not know anything about it. Approximately 47 percent of the respondents who had heard about irradiation indicated that they do not know anything about it. The remainder of the respondents chose the second statement from the above list. None of the respondents indicated they were sufficiently informed. An estimate of the proportion of respondents having a meaningful level of knowledge about irradiation would seem to require that consideration be given only to that group of the respondents who have heard about irradiation and who feel that they are

Irradiation and Food Safety: Consumer Attitudes and Awareness

443

at least somewhat informed about the process. This group of respondents accounted for only 28 percent of the total responding panel members. Testing Consumer Knowledge. Another measure of consumer awareness of irradiation technology was developed based on the responses to ten factual statements with "true," "false," and "don't know" answer options (see Appendix 20.A, Table 20.Al). Three of these statements were true. Responses to these statements were translated to item scores and were coded so that 1 indicated a correct answer and 0 indicated a wrong answer or a don't know response. To construct an awareness variable, the item scores for each respondent were first summed to obtain a total score. The total awareness scores were then expressed as an index ranging from i.ero to 1. An index value of 1 corresponded to the highest possible total score of 10 points. The arithmetic mean of this constructed awareness index was 0.279. This corresponds to a very low level of consumer awareness of food irradiation since 1 corresponds to perfect awareness. No respondent received a perfect score of 1.

Influence of Demographic Variables Comlation of Demographics with Consumer Awareness. The data were

analy7.ed to determine if consumer awareness of irradiation differs significantly among demographic subgroups using Chi-square contingency tests. The awareness index was collapsed into two categories representing "low" and "moderate or high" awareness for this purpose. Respondents receiving a score of 0.5 or less were clustered together in the low awareness group and respondents with a score greater than 0.5 were categori7.ed as the moderate or high awareness group. The results suggest a correlation between awareness of food irradiation and a respondent's gender, education, and household income. Table 20.5 shows that a relatively higher proportion of females had lower awareness of the irradiation process than males, and that this difference was statistically different from i.ero at the 5 percent level of significance. Respondents with a college education and higher incomes appeared more likely to have a better awareness of irradiation than those with less than a college education and with lower incomes. These results are comparable with those of Schutz et al. (1989), who reported a higher level of awareness of irradiation for men, older people, and those with more education. Malone's (1990) study of consumer willingness to purchase irradiated fresh food also indicated a higher level of awareness for men, more highly educated people, and those with higher incomes. Comlation of Demographics with Consumer Perception. It was noted earlier from Table 20.2 that about 39.3 percent of the respondents perceived irradiation to be either a serious or extremely serious problem and that 26.1 percent and 34.6 percent of the respondents perceived irradiation to be a

444 TABLE 20.5 Cross-Tabulation of Consumer Awareness Index by Economic and Demographic Characteristics (Chi-Square Contingency Test) Consumer Awareness Characteristics

Low

High

Chi-Square Value

-------- %-------RACE White Nonwhite GENDER Male Female AGE Less than 35 3645 years 46-65 years More than 65 EDUCATION Noncollege College MARITAL STATUS Married Other HOUSEHOLD INCOME Less than $20,000 $20,000-$29,000 $30,000-$39 ,000 $40,000-$49 ,000 $50,000 or more HOUSEHOLD SIZE 1 person 2-4 persons 5 or more persons PLACE OF RESIDENCE Urban Rural

0.198 86.49 89.19

13.51 10.81

81.40 90.44

18.60 9.56

78.57 91.07 85.71 90.70

21.43 8.93 14.29 9.30

92.78 82.54

7.22 17.46

84.91 92.19

15.09 7.81

93.44 88.89 90.91 82.61 76.79

6.56 11.11 9.09 17.39 23.21

92.11 85.53 88.46

7.89 14.47 11.54

85.71 88.35

14.29 11.65

3.796**

3.995

5.084**

2.139

8.111***

1.226

0.338

Note: The superscripts ** and *** correspond to levels of statistical significance of 5 percent and 10 percent, respectively.

Irradiation and Food Safety: Consumer Attitudes and Awareness

445

moderate, or little or no problem, respectively. Cross-tabulation of consumer perceptions of irradiation by demographic variables and the awareness index revealed some interesting results. Consistent with the Chi-square contingency test for consumer awareness, results suggested a statistically significant correlation between consumer perceptions and the respondent's gender, education, and household income (Table 20.6). Females, respondents with less than a college education, and lower income respondents appeared more likely to perceive irradiation as a relatively more serious problem than their counterparts. This is consistent with the findings of Schutz et al. (1989) that there is a higher level of concern about irradiation among women and less educated individuals. A cross-tabulation of consumer perception and the awareness index showed statistical significance for the Chi-square statistic (Table 20.6). Consumers with low levels of knowledge about irradiation tended to perceive it as a more serious problem than those with moderate or high levels of awareness.

Disseminating lnfonnati.on Studies of consumer awareness and attitudes toward food irradiation, including this one, show that individuals are ambivalent toward irradiated food. The overall results of all existing studies, despite methodological differences, show a general lack of information about food irradiation and a persisting consumer uncertainty. Several studies (Urioste et al. 1990, Schutz et al. 1989, Bord and O'Conner 1989, Terry and Tabor 1988, Bruhn et al. 1988) have shown that consumer education campaigns have a positive influence on perceptions about the acceptability of irradiation. This study attempted to determine the importance of labeling and to identify various facets of an effective consumer education program from the consumer perspective. It identifies the advantages of irradiation and the arguments in favor of irradiation technology that the consumers believe to have a positive influence on their perceptions. Further, consumers' primary sources of information about irradiation and who they prefer to be relaying information were identified. Attitudes Toward Irradiation Labeling. Attitudes toward irradiation labeling were most certainly favorable. Over 74 percent of the respondents, responding to a question that displayed the international food irradiation symbol, indicated that it is very important for irradiated products to be clearly labeled. Another 24.3 percent of the respondents perceived it to be somewhat important. About 70 percent of the panelists felt that the international symbol of food irradiation, established to designate irradiated food products, is sufficient to inform consumers that the food is irradiated. However, the symbol was not perceived as a symbol of assurance for safety by over 80 percent of the respondents. This would seem to indicate that consumers want irradiated food to be labeled primarily because they want to know what they are buying.

446 TABLE 20.6 Cross-Tabulation of Consumer Perceptions oflrradiation by Economic and Demographic Characteristics, and Awareness Index (Chi-Square Contingency Test) Consumer Perception Characteristic

Little or No Problem

Moderate Problem

Serious Problem

ChiSquare Value

---------- %----------RACE White Nonwhite GENDER Male Female AGE Less than 35 36-45 years 46-65 years More than 65 EDUCATION Noncollege College MARITAL STATUS Married Other HOUSEHOLD INCOME Less than $20,000 $20,000-$29,000 $30,000-$39,000 $40,000-$49 ,000 $50,000 or more HOUSEHOLD SIZE 1 person 2-4 persons 5 or more persons PLACE OF RESIDENCE Urban Rural CONSUMER AWARENESS Low High

35.26 20.69

26.59 27.59

38.15 51.12

41.56 27.42

28.57 25.81

29.87 46.77

23.08 44.23 31.88 31.58

35.90 15.38 30.43 28.95

41.03 40.38 37.68 39.47

22.22 40.50

27.16 26.45

50.62 33.06

35.62 26.79

25.34 30.36

39.04 42.86

18.18 44.12 36.67 31.82 42.59

25.45 23.53 43.33 31.82 20.37

56.36 32.35 20.00 36.36 37.04

32.26 34.01 29.17

29.03 26.53 25.00

38.71 39.46 45.83

37.38 28.42

28.04 25.26

34.58 46.32

31.15 57.14

28.42 10.71

40.44 32.14

2.740 6.396 ..

7.303

8.625. 1.479 18.019••

0.473

3.092 8.087..

Note: The superscripts • and ••correspond to levels of statistical significance of 1 percent and 5 percent, respectively.

Irradiation and Food Safety: Consumer Attitudes and Awareness

447

Persuasi.ve Arguments in Favor of Irradiation. Seven statements representing some arguments in favor of irradiation were presented to the respondents (see Appendix 20.A, Table 20.A2). Respondents were asked to express how each of seven statements would affect their concern about irradiation. Only 20-22 percent of the respondents indicated that statements such as irradiation has government approval, irradiated food has been eaten by American and Soviet astronauts in space, and the enetgy used for irradiation is similar to ultraviolet light, would either totally eliminate or reduce their concern about irradiation to a great extent. Irradiation concerns of about a third of the respondents would be totally eliminated or reduced considerably if they were informed that irradiation has already been used in U.S. hospitals to steriliu medical and surgical products. Approximately the same proportion of respondents felt similarly about recommendations of various world organi:zations as to the safety of irradiation. Irradiation concerns for about 35 percent of the respondents would be totally eliminated or reduced considerably if they were informed that irradiation is used in U.S. hospitals to sterili7.e utensils and food for patients with critical immunity problems. The most persuasive argument in favor of irradiation (chosen by 41 percent) was that extensive research has shown irradiated food to be safe, wholesome, and nutritious. These results indicate that it is not sufficiently convincing to promote irradiation technology with seals of approval from the government and world organi:zations. Consumers appear to be asking for information based on extensive scientific research and real world tests. This should not be interpreted as contradicting earlier findings (Urioste et al. 1990, Bruhn and Noell 1987) that government authori:zation has a positive influence on the acceptability of irradiated food. Our finding only reinforces their judgement, while identifying additional persuasive arguments in favor of irradiation. Persuasi.ve Irradiation Benefits. Seven irradiation advantages were presented in another question, and panelists were asked to indicate how each of these statements was perceived as being an advantage for the process (Appendix 20.A, Table 20.A3). In general, all the advantage statements were perceived to indicate a major or somewhat major advantage of the irradiation process. The least persuasive advantage of the irradiation process was still perceived to be a major or somewhat major advantage by 60 percent of the respondents. The potential of irradiation technology in facilitating the production of Trichina-free pork and the transportation of processed food from specialil.ed production areas to other areas was perceived as a major or somewhat major advantage by about 60 percent of the respondents. Approximately 67 percent of the respondents considered the possibility of lowering the cost of foods due to irradiation as a major or somewhat major advantage of the process. The potential for reducing or eliminating the need for preservatives and extending the shelf life of fruits without using chemicals was considered as a

448

Sukant Misra, Stanley Fletcher, and Chung Huang

major or somewhat major advantage by 69 and 66 percent of the respondents, respectively. Irradiation as a solution to world hunger was perceived to be a major advantage by 72 percent of the respondents. The most persuasive advantage of the irradiation process was its potential for reducing or eliminating Salmonella contamination of poultry meat. Several previous studies have shown that information provided to the consumer regarding the benefits of irradiation results in a positive influence on their attitudes. The results in this section provide a measure of the persuasiveness of several specific benefits of the irradiation process. Whom Do Consumers Trust? Respondents were asked to express the extent of their confidence in comments on food safety issues by different interest groups. A five-point scale of "no confidence" to "high confidence" was used to record consumers' responses. The list of pre-selected groups included scientists from a university; representatives of a government agency, an independent laboratory, a chemical manufacturer, a consumer group, a grower association, or a supermarket; popular media personality or celebrity; and friends, family, or fellow workers. The top choice by the panelists was university scientists, chosen to be most trustworthy by about 58 percent of the respondents. Table 20. 7 shows that

TABLE 20. 7 Consumers' Confidence in Comments Made About Food Safety by Various Groups of Individuals•

Group

Little or No Confidence

Moderate Confidence

High Confidence

-------------- %-------------University scientists Government agency Independent laboratory Chemical manufacturer Consumer group Grower association Supermarket Media personality or celebrity Friends, family, fellow workers

13.9 38.8 13.8 61.6 16.9 41.0 49.3 52.1 30.4

27.8 38.3 29.9 25.9 29.8 36.1 37.0 21.3 30.8

58.3 22.9 56.3 12.5 53.3 22.9 13.6 8.6 38.8

•For purposes of analysis, the 5-point scale was collapsed into three groups. A ranking of 1 and 2 was taken to imply little or no confidence, a ranking of 3 was taken to imply moderate confidence, and a ranking of 4 and 5 was taken to imply high confidence.

Irradiation and Food Safety: Consumer Anitudes and Awareness

449

university scientists were closely followed by representatives of an independent laboratory (chosen by 56 percent of the panelists) and of a consumer group (chosen by 53 percent). That consumers expressed such a high level of confidence in representatives of a consumer group is noteworthy, especially since consumer groups are generally perceived as staunch opponents of food irradiation. This probably suggests that consumers want to be assured of the safety of the irradiation process with absolute certainty. They want the approval not only of scientists and independent entities, but also from the opponents of the process. Only about 23 percent of the respondents expressed a high degree of confidence in comments made by representatives of the government. Government agencies received a lower ranking than friends, family, and fellow workers, chosen by 39 percent of the respondents as highly trustworthy. It appears that the public's trust and confidence in the government's ability to guarantee the safety of food supplies has eroded. Representatives of supermarkets and chemical manufacturers and popular media personalities or celebrities, were the three lowest ranked groups in terms of consumers' confidence. Popular media personalities or celebrities are considered the least trustworthy as about 52 percent of the panelists had no confidence in their comments on food safety issues. Channels for IJi.sseminating lnformadon. Over 64 percent of the respondents identified radio and television as their major sources of information about irradiation. Approximately 54 percent indicated that they also depend on newspapers to receive information. Though about 38 percent of the respondents identified magazines as a major source, this finding is ambiguous since it was not possible to identify specific magazines. No other listed source was found to be effective as they were chosen by only a very small proportion of the panelists. These results clearly sug~t that radio, television, and newspapers are the most effective sources for disseminating information about irradiation to consumers.

Conclusions Consumers are ambivalent in their attitude toward irradiated foods and are concerned about perceived huards that may be associated with the technology. There is also some evidence that supports a view that public opinion is favorable to irradiation. One indication of a favorable public is that irradiation is perceived to be necessary for the preservation of several food products. Specifically, irradiation is perceived to be necessary for seafood and meat products. This finding should encourage the poultry, pork, beef, and seafood industries to reevaluate their perception of a public opinion which reflects only negative attitudes toward irradiation. Another indication of a more supportive

450

Sukant Misra, Stanley Fletcher, and Chung Huang

public is that only a very small proportion of those expressing opinions outrightly decline to buy irradiated food. It is true that irradiation technology has yet to receive an unconditional endorsement from the public. The public simply lacks sufficient information about the irradiation process and, therefore, is not convinced that it is safe. Survey results clearly suggest that consumers' knowledge and awareness about the process of irradiation is at a low level. This lack of awareness was found to be responsible for the low degree of enthusiasm which characterized consumers' responses relative to the potential for irradiated food in the marketplace. Low awareness was also found to be significantly important in explaining higher levels of concern on the safety of irradiation. The overall lack of awareness of the irradiation process makes the successful introduction of irradiation dependent upon the development of a consumer education program that targets the entire population. The survey results indicate that women and individuals with less income and education are less informed about irradiation suggesting a possible focus for the program. Several critical factors must be carefully considered in the development and implementation of an effective consumer education program. The survey suggests that the program should be equipped with extensive research findings on the effects of irradiation on food, individuals, and the environment. Environmental and occupational implications of the technology must be clearly and precisely addressed. Consumers must also be informed as to the current status of the technology in terms of where and how it is currently used and what have been the results of these real world tests. Irradiated products should also be clearly and precisely labeled and all the advantages and disadvantages of irradiation should be clearly and objectively spelled out. University scientists should be encouraged to communicate directly with the public in view of the eroding public confidence in the government's ability to guarantee the safety of food supplies. It appears critical that scientific research findings pertaining to irradiation technology be relayed directly to the public via popular media such as television, radio, and newspapers. Independent expert committees and representatives of consumer/environmental groups may also be involved to further strengthen creditability in the implementation of an effective consumer education program. In this chapter, a survey on consumer attitudes and evaluations of food safety technologies was analyzed. This study contributes to the broad spectrum of consumer valuation and attitudinal research, using survey instruments, that is designed specifically to explore consumers' knowledge and awareness about food safety and nutrition issues. Interpretation of the results of this survey on irradiation sheds additional insights on consumer attitudes. For example, we found that most consumers are unwilling to purchase irradiated food because they are uncertain about the safety of the process. The uncertainty can be attributed to the fact that the public has very little knowledge of the technology

Irradiation and Food Safety: Consumer Attitudes and Awareness

451

and is basically unaware of what it has to offer in improving the safety of food products. The results have important implications for food industry decision makers and government officials in their assessment of the market potential for irradiated foods. In particular, the identification of consumers' fears for the safety of the technology pinpoints the need, as the first priority, to develop and disseminate information concerning food irradiation to gain consumer acceptance in the marketplace. Thus, results of the study provide a useful means to identify the most important components required in the development of an effective consumer education program.

References Bord, R. J. and R. E. O'Connor. 1989. Who Wants Irradiation Food? Untangling Complex Public Opinion. Food Technology 43(10):87-90. Brand Group. 1986. Irradiated Seafood Product. A Position Paper for the Seafood Industry. Final Report. Chicago, IL: Brand Group. Bruhn, C. M., H. G. Schutz, and R. Sommer. 1986. Attitude Change Toward Food Irradiation Among Conventional and Alternative Consumers. Food Technology 40(1):86-91. Bruhn, C. M. and J. W. Noell. 1987. Consumer In-Store Response to Irradiated Papayas. Food Technology 41(1):83-85. Bruhn, C. M., H. G. Schutz, and R. Sommer. 1988. Food Irradiation and Consumer Values. "Ecology of Food and Nutrition 21:219. Diehl, J. F. 1983. Food Irradiation. In Developments in Food Preservation-2, ed. S. Thome. New York, NY: Applied Science Publishers. Feedst'flfs. 1992. Distrust of Technology Zaps Irradiated Food Market 64(24):1. Food Marketing Institute. 1988. Trends: Consumer Attitudes and the Supermarket. Washington, D.C.: Food Marketing Institute. Food Marketing Institute. 1991. Trends: Consumer Attitudes and the Supermarket. Washington, D.C.: Food Marketing Institute. Huang, C. L. and S. K. Misra. 1990. Designing a Data System for Consumer Research. Agribusiness 6(5):515-524. Malone, J. W., Jr. 1990. Consumer Willingness to Purchase and to Pay More for Potential Benefits of Irradiated Fresh Food Products. Agribusiness 6(2): 163-178. Pszczola, D. E. 1990. Food Irradiation: Countering the Tactics and Claims of Opponents. Food Technology 44(6):92-97. Roberts, T. 1985. Microbial Pathogens in Raw Pork, Chicken, and Beef: Benefits Estimates for Control Using Irradiation. American Journal ofAgricultural&onomics 67(5):957-965. Schutz, H. G., C. M. Bruhn, and K. V. Diaz-Knauf. 1989. Consumer Attitude Toward Irradiated Food: Effects of Labeling and Benefits Information. Food Technology 43(10):80-86. Terry, D. E. and R. L. Tabor. 1988. Consumer Acceptance of Irradiated Produce. Journal of Food Distribution Research 19(1):73-89.

452

Sukant Misra, Stanley Fletcher, and Chung Huang

Turcsik, R. 1992. Irradiation Plant Takes on Spices, Tomatoes, and Onions. Supermarket News 42(23):35. Urioste, A. M., C. A. Croci, and 0. A. Curzio. 1990. Consumer Acceptance of Irradiated Onions in Argentina. Food Technology 44(5):134-136. Wiese Research Associates. 1984. Consumer Reaction to the l"adiation Concept. Omaha, NE: Wiese Research Associates.

453

Appendix 20.A SURVEY QUESTIONS TABLE 20.Al Complete Text for Question on Factual Statements to Measure Consumer Awareness of Irradiation Please answer the following irradiation related questions by circling the number that you think represents the correct answer, where 1 = true, 2 = false, and 3 = don't know.

1~FW A. Food deliberately treated with radiation can at present be bought in some countries of the world.

1

2

3

irradiated foods available for retail sale.

1

2

3

Food contains natural radioactivity.

1

2

3

irradiated.

1

2

3

E.

Food that has been treated with irradiation cannot be recontaminated.

1

2

3

F.

Irradiated foods retain most of their appearance, taste, and quality, making them almost indistinguishable from fresh raw foods.

1

2

3

spoilage if it had gone bad.

1

2

3

H.

It is legal to irradiate food repeatedly.

1

2

3

I.

It is proved that consumption of irradiated food increases incidence of cancer in test animals.

1

2

3

J.

Food treated with radiation has a higher level of radioactivity than nonirradiated food.

1

2

3

B. The United States leads the world in the number of

c.

D. Scientists can easily detect food that has been

G. Irradiated food cannot be recognized for visible

454 TABLE 20.A2 Complete Text for Question on Arguments in Favor of Irradiation The following statements represent some arguments that have been presented in favor of irradiation. Please indicate, by circling the number, how it would affect your concern about irradiation. Where, 1 = would totally eliminate any concerns I may have, 2 = would reduce my concern a lot, 3 = would reduce my concern a little, 4 = wouldn't affect my concerns at all, and S = not sure.

111 A.

Not Sure

The U.S. government has already approved such processing for spices and other seasonings, fruits and vegetables, pork, and poultry.

1

2

3

4

s

The process has already been used in the U.S. to sterilize about 30% of all medical and surgical products used by American hospitals.

1

2

3

4

s

The United Nations Food and Agricultural Organization, the International Atomic Energy Agency, and the World Health Organization Joint Expert Committee have recommended the process as being safe.

1

2

3

4

s

Foods processed by this method have been eaten by the American and Soviet astronauts in space.

1

2

3

4

5

The process is used in U.S. hospitals for patients with critical immunity problems who cannot tolerate any disease-causing organisms in their food.

1

2

3

4

5

F.

The energy used for this purpose is similar to ultraviolet light, only more powerful.

1

2

3

4

5

G.

Extensive scientific research has shown that proper use of this process poses no health hazards and the irradiated food is safe, wholesome, and nutritious.

1

2

3

4

5

B.

C.

D.

E.

455

TABLE 20.A3 Complete Text for Question Pertaining to Advantages of Irradiation The following represent some of the advantages associated with the irradiation process. Please indicate, by circling the number, how do you see each of the following factors as being an advantage for the process. Where, 1 a major advantage, 2 a minor advantage, 4 no somewhat of a major advantage, 3 advantage, and 5 not sure.

=

A.

=

The process would reduce or eliminate the need for preservatives, such as nitrates which are carcinogenic, in the processed meat.

=

=

I:~

Major

=

III

1N~ Sure

1

2

3

4

5

B.

The process would reduce or eliminate Salmonella contamination of poultry meat.

1

2

3

4

5

c.

The process would reduce or eliminate the need to use chemicals on fruits and vegetables and extend their shelf life by several weeks to a month.

1

2

3

4

5

The process would allow the U.S. livestock market to produce Trichina-free pork and open up the world market in which entry of pork is currently denied because the meat cannot be guaranteed free of Trichina parasites.

1

2

3

4

5

The process would lower the cost of foods by greatly reducing the wastes due to spoilage or insect damage, and by eliminating the need for freezing and continued refrigeration.

1

2

3

4

5

The process would allow items processed in specialized areas of the country, such as fish, fruits, and vegetables, to be shipped chilled but not frozen to other areas.

1

2

3

4

5

1

2

3

4

5

D.

E.

F.

G. The process would help ease the world

hunger problem since today many countries lose 30% or more of their food production to spoilage and insects before it reaches the marketplace.

About the Book and Editor The delivery of safe food of appropriate nutritional quality is an increasingly important measure of the performance of the food production, manufacturing, and distribution system. Both consumers and society are demanding products with improved quality attributes-seeking, for example, lower contamination levels for foodbome pathogens and lower fat content. In this richly researched book, contributors present a thorough exploration of these issues, emphasizing especially the unresolved question of how much value consumers and society place on food safety and nutritional quality. Developing economic models and reporting the results of several original valuation studies by leading researchers, this volume makes a strong contribution to the newly developing field of valuing food safety and nutrition. Most important, the contributors critically evaluate the relative strengths, weaknesses, and requirements of the five major methodologies being employed in valuation research-contingent valuation, experimental markets, conjoint analysis, hedonic pricing, and cost of illness. They provide the detail on research design (including the text of surveys) necessary for thorough evaluation of results, making the book an extremely useful reference tool for researchers doing valuation work. The book also includes a detailed discussion of key factors such as illness data, risk analysis, and information on consumer attitudes. Julie A. Caswell is a professor, Department of Resource Economics, University of Massachusetts-Amherst.

457