The Economy, Crises, and the Labor Market: Can Institutions Serve as a Protective Shield for Employment? [1 ed.] 9783428534395, 9783428134397

Citation preview

Applied Economics Quarterly Supplement

2010

Issue 61 The Economy, Crises, and the Labor Market Can Institutions Serve as a Protective Shield for Employment?

Edited by Klaus F. Zimmermann and Christian Wey

asdfghjk Duncker & Humblot · Berlin

The Economy, Crises, and the Labor Market

Applied Economics Quarterly Supplement Issue 61

The Economy, Crises, and the Labor Market Can Institutions Serve as a Protective Shield for Employment?

Edited by

Klaus F. Zimmermann and Christian Wey

asdfghjk Duncker & Humblot · Berlin

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.

Technical editor: Deborah Anne Bowen All rights reserved. No part of this book may be reproduced, translated, or utilized in any form or by any means, electronic or mechanical, without the expressed written consent of the publisher. # 2010 Duncker & Humblot GmbH, Berlin Typesetting and printing: Berliner Buchdruckerei Union GmbH, Berlin Printed in Germany ISSN 1612-2127 ISBN 978-3-428-13439-7 (Print) ISBN 978-3-428-53439-5 (E-Book) ISBN 978-3-428-83439-6 (Print & E-Book) Printed on no aging resistant (non-acid) paper ∞ according to ISO 9706 *

Internet: http://www.duncker-humblot.de

Editorial This supplement to Applied Economics Quarterly reports on the 73rd Annual Meeting of the Association of German Economic Research Institutes (ARGE-Institute), which took place in Berlin on April 29, 2010. The topic was: “The Economy, Crises, and the Labor Market. Can Institutions Serve as a Protective Shield for Employment?” The year 2009 was marked by a deep global economic crisis triggered by turbulence on the financial markets. The crisis has affected different countries’ economies to differing degrees. The impact on national labor markets was even more severe, and wider in scope, than the resulting economic slump itself. It apppears likely that the different institutions are (at least partly) to blame. Against this backdrop, the 73rd ARGE meeting examined the relationships between institutions—labor market institutions but others as well—and labor market developments in times of crisis. The key question was whether and how institutions can serve as a “protective shield” for employment. The lectures focuses on the following aspects: European labor markets in international comparison, the role of labor market institutions in the crisis, labor market reforms and competition, labor force cohorts, meta-analysis of the minimum wage, and labor market regulation. We are grateful to the federal government—particularly to the Federal Ministry of Economics and Technology—for their generous support for the meeting. We would like to thank Heike Hofbauer-Geer (IAB Nürnberg), Jens SchmidtEhmcke (DIW Berlin), and Ralf Messer (ARGE-Institute) for the conceptual preparation of the conference, the organizer, Susanne Marcus (ARGE-Institute), and all the participants and attendees for their contributions. June 2010

Klaus F. Zimmermann Christian Wey

Contents Joachim Möller Germany’s Job Miracle in the World Recession—Shock-Absorbing Institutions in the Manufacturing Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

Werner Eichhorst, Michael Feil, and Paul Marx Crisis, What Crisis? Patterns of Adaptation in European Labor Markets . . . . . . . . . . . . .

29

Comment: Antje Mertens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

59

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl The Role of Labor Market Institutions in the Great Recession . . . . . . . . . . . . . . . . . . . . . . .

65

Comment: Claus Schnabel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

83

Alex Herzog-Stein and Camille Logeay Labor Market Reforms, Hysteresis, and Business Cycles in Germany: A SVAR Approach to Explain Unemployment Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

89

Comment: Jürgen Jerger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 Martin Dietz, Michael Stops, and Ulrich Walwei Safeguarding Jobs through Labor Hoarding in Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 Comment: Olaf Hübler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Bernhard Boockmann The Combined Employment Effects of Minimum Wages and Labor Market Regulation—a Meta-Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Comment: Alexandra Spitz-Oener . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

Germany’s Job Miracle in the World Recession—Shock-Absorbing Institutions in the Manufacturing Sector By Joachim Möller* Abstract This paper aims at analyzing the Germany’s so-called job miracle in the aftermath of the World Recession. The massive downturn mainly affected export-oriented manufacturing industries. These industries have shown an astonishingly moderate employment response. Hence there is strong evidence for substantial labor hoarding. By analyzing different branches of manufacturing one can shed some light on the buffering processes that took place at the firm level. It is shown that measures of within-firm flexibility—supported by labor market instruments like short-time work schemes—prevented Germany from major dismissals and a sharp increase in the unemployment rate. Keywords: World Recession, labor market adjustments, labor hoarding, labor market institutions, labor market reforms, business cycle, short-time work schemes JEL Classification: J2, J6

1. Introduction At the latest since mid-2009 it has been evident that the German labor market response to the World Recession was not following the usual pattern. In economic downturns of the past, the employment path mirrored changes in real GDP almost exactly, although typically with a delay of one or two quarters. While over the last two or three decades the amplitudes of de-trended growth rate of both time series showed very similar magnitudes, this is not the case in the current recession. For the year 2009, for instance, a 1-step-ahead forecast of a simple dynamic model relating relative employment changes to real GDP growth clearly exceeds the 95confidence band (see Möller 2010). Hence there is clear-cut indication for a structural break in the dynamic behavior of employment in the business cycle. Put differently, employment seems to be more or less decoupled from the enormous economic shock that hit the export-oriented branches of the German economy. This is what Paul Krugman in his New York Times column from November 11, 2009 de* IAB, IZA, and Department of Economics, University of Regensburg, Universitätsstraße 31, D-93053 Regensburg, Germany, e-mail: [email protected].

10

Joachim Möller

notes “Germany’s jobs miracle”. And the most recent data from the German labor market at the time of writing this article (June 2010) indicate that the miracle continues. Seasonally-adjusted German unemployment is falling although the real economy has by no means fully recovered from the biggest recession after World War II. At present, the German labor market seems to be an exceptional case among European economies. Harmonized unemployment rates in the European Union as a whole exceed 10 percent and are still rising, whereas German unemployment is down to 7.2 percent. This is remarkable because the economic burden of re-unification and the deep structural problems in eastern Germany are still not fully overcome. This paper aims at looking behind the phenomenon of job stability in the World Recession that was triggered by the financial crisis. The massive downturn mainly affected export-oriented manufacturing industries. In these industries there is strong evidence for substantial labor hoarding. By analyzing different branches of manufacturing one can shed some light on the buffering processes that took place at the firm level. It is argued that what might be called within-firm flexibility prevented Germany from major dismissals and a substantial increase in the unemployment rate. This flexibility profited from a well-functioning social partnership in the most affected industries and was greatly supported by labor market instruments like short-time work schemes.

2. Production and Productivity in Manufacturing Industries before and after the Crisis After Lehman Brothers filed for Chapter 11 bankruptcy protection on September 15, 2008, the subprime crisis quickly spilled over to the real economy. Throughout the world, firms were cutting down their plans to invest. In most countries consumer’s confidence and private spending on durable consumption goods declined sharply. In an export-oriented economy like Germany the first shock wave especially hit manufacturing industries with high export shares, typically producers of investment goods and consumer durables and their suppliers. Three points are important for the analysis of the shock wave: First, the firms in the most-affected industries are a positive selection with respect to productivity, technology and profitability (cf. e.g. Fryges and Wagner 2008). Second, in recent years unit labor costs were developing very favorable for German manufacturing firms, especially when compared to those of its competitors on the world market. Third, many of Germany’s export-oriented manufacturing firms have been experiencing an enormous upswing because their products meet the demand of capital goods and infrastructure modernization in the dynamically developing world regions. Taken together, these favorable economic framework conditions in German manufacturing led to an extraordinary strong boom after the slump in 2003 / 2004. This was even interpreted as a sign of re-industrialization in Germany strong enough to reverse the

Germany’s Job Miracle in the World Recession

11

long-run trends of manufacturing decline in the process of structural change. Although this might considered an exaggeration, it is incontestable that the growth rates of at least some parts of manufacturing industries—fostered by high demand and favorable cost conditions—were extraordinary even when compared to historical standards. Figure 1a depicts orders received by manufacturing and mining industries in a long-run perspective. The figure shows that industrial orders sharply increased from 2003 to 2008. According to Figure 1b the cyclical deviation from the trend was strikingly high before the crisis. The order of magnitude—20 log points above the trend—was comparable to industrial boom periods of the fifties, sixties and early seventies of the last century. Remarkably enough, the boost in demand clearly surpassed that during the re-unification boom of the early 1990s. Figure 2a shows that industrial production closely follows demand. Again, up to the first quarter of 2008 the positive deviation from trend was extraordinary although the cyclical component of production appears to be somewhat dampened when compared to the corresponding demand series. Because the crisis abruptly ended the boom period in German export-oriented manufacturing firms the height of fall was enormous. Table 1 reveals that the difference between peak and trough at the latest recession exceeded 30 log percentage points. This was the highest difference after World War II in Germany. Moreover, the duration between peak and trough was 16 months only. Hence industrial production between January 2008 and April 2009 fell at the fastest rate ever observed in the post-war period. In such an unusual situation, it is interesting to analyze productivity behavior. Figure 3a shows an index of production per employee in mining and manufacturing industries and the corresponding trend.1 It becomes obvious that the fall in production was accompanied by an unprecedented, massive downturn in industrial productivity. Figure 3b depicts the deviations from trend. From a situation of roughly 4 log percent above trend before the crisis, productivity fell to a level not less than 28 log percent below trend in April 2009. Although the series has somewhat recovered in the lasts months it is still around 15 log percent below its long-run expected level in March 2010. This situation has to be compared to maximum deviations from trend of about minus 5 log percent in previous recessions. Although a certain productivity slowdown is normal in economic downturns, the enormous size of the effect since mid-2008 calls for an explanation.

1 The trend was calculated using a polynomial of order 2 for the time period 1991:M1 to 2008:M8. The fitted trend was then extrapolated to the period 2009:M7 to 2010:M3.

12

Joachim Möller 5.0

4.5

4.0

3.5

3.0

2.5

2.0 50

55

60

65

70

75

80

85

90

95

00

05

10

Industrial Orders (Index 2005=100, in logs) Trend

Notes: Own calculations, source for original data; Deutsche Bundesbank; series are adjusted for seasonality and variations in calendar time; trend is calculated using a polynomial of order 4.

Figure 1a: Monthly Index of Industrial Orders and Trend (Index 2005 = 100, in logs, 1952:M1 to 2010:M3)

.3 .2 .1 .0 -.1 -.2 -.3 -.4 50

55

60

65

70

75

80

85

90

95

00

05

10

Notes: See Figure 1a.

Figure 1b: Monthly Index of Industrial Orders, Cyclical Deviation from Trend (in logs, 1952:M1 to 2010:M3)

Germany’s Job Miracle in the World Recession

13

4.8

4.4

4.0

3.6

3.2

2.8 50

55

60

65

70

75

80

85

90

95

00

05

10

Industrial Production (Index 2005=100, in logs) Trend

Notes: Own calculations, source for original data; Deutsche Bundesbank; series are adjusted for seasonality and variations in calendar time; trend is calculated using a polynomial of order 4.

Figure 2a: Monthly Index of Industrial Production and Trend (Index 2005 = 100, in logs, 1952:M1 to 2010:M3)

.15 .10 .05 .00 -.05 -.10 -.15 -.20 50

55

60

65

70

75

80

85

90

95

00

05

10

Notes: See Figure 2a.

Figure 2b: Monthly Index of Industrial Production, Cyclical Deviation from Trend (in logs, 1952:M1 to 2010:M3)

14

Joachim Möller Table 1 Major Cycles of Industrial Production peak

trough deviation from trend

period

deviation from trend

period

1965M01 1973M08 1979M12 1991M01 2001M02 2008M01

2.40 10.09 7.19 12.79 5.53 12.00

1967M05 1975M07 1982M11 1993M07 2003M09 2009M04

–13.62 –8.30 –10.12 –7.51 –4.68 –18.46

difference peak / trough

duration (in months)

–16.02 –18.39 –17.32 –20.29 –10.21 –30.46

29 24 36 31 32 16

Notes: Own calculations using data from the German Bundesbank; for calculation of trend see figure 2. An outlier in June 1984 was ignored.

4.8 4.7 4.6 4.5 4.4 4.3 4.2 4.1 4.0 3.9 92

94

96

98

00

02

04

06

08

Industrial Productivity (Index 2005=100, in logs) Trend (1991:1-2008:8) and Forecast (2008:9-2010:3)

Notes: Own calculations, source for original data; Deutsche Bundesbank; series are adjusted for seasonality and variations in calendar time; for calculation of trend see footnote 1.

Figure 3a: Monthly Productivity (Index 2005 = 100, in logs, 1991:M1 to 2010:M3)

Germany’s Job Miracle in the World Recession

15

.08 .04 .00 -.04 -.08 -.12 -.16 -.20 -.24 -.28 92

94

96

98

00

02

04

06

08

Notes: See Figure 3a.

Figure 3b: Monthly Productivity, Cyclical Deviation From Trend (in logs, 1991:M1 to 2010:M3)

3. Employment Dynamics and Labor Hoarding 3.1 Employment Changes in Different Branches of the Economy

Figure 4 gives the changes in employment across different branches of the economy in the first twelve months after September 2008. On the negative side the lion’s share of the employment decline is in manufacturing industries (–252,000) and temporary work and manpower services (–140,000), whereas—on the positive side—health and social services (+133,000) exhibit significant employment gains. Table 2 shows that in absolute terms employment changes in other industries are much lower and do not exceed 40,000. In relative terms the by far highest drop in employment was in temporary work and manpower services (–19.6%). Employment in manufacturing industries fell by 3.8 % and by 3.5 % in real estate activities, followed by mining and quarrying (–2.8 %), information and communication (–2.3 %) and transport and storage (–2.1%). Relative strong employment gains can be found in the energy sector (+2.0 %) and different branches of the service sector (hotels and restaurant (+2.4%), recreational, cultural, sporting activities (+3.2%), education (+3.7%) and health and social services (+4.1%). It can be seen that major employment losses in absolute and relative terms have mainly occured in manufacturing industries and related branches (with only real estate activities being an exception). It is remarkable that evidently the World Recession did not spillover to more consumer oriented services: On the contrary, these branches partly expanded their work force despite the crisis. Although there is an employment response in manufacturing industries, the extent is astonishingly moderate given the size of the demand shock. Here, one could ask whether the extraordinary fall of industrial productivity might be a statistical artifact.

16

Joachim Möller 200 150 100 50 0 -50 -100 -150 -200 -250

Health & social work

Education

Public Admin.

Busin. serv. excl. temps

Other services

Hotels/restaurants

Financial interm.

Cultural activ. etc.

Energy

Agriculture etc.

Scient., techn. Serv.

Construction

Priv. househ. etc.

Water supp., recycl.

Mining, quarrying

ICT

Real estate

Transport

Trade&Repair

Manufacturing

Temporary work

-300

Notes: Own calculations with data from the Statistical Department of the Federal Employment Agency. Employment numbers are end-of-the month values.

Figure 4: Change in Employment 2008:M9 to 2009:M9 (Absolute Numbers in Thousand)

Whereas the boom 2005 – 2008 was accompanied by an enormous increase in the employment of temporary workers, the crisis let to a sharp fall. Since around 2 / 3 of these workers were occupied in manufacturing industries—but counted as employees in service industries—the productivity boost before the crisis as well as the slowdown appears to be exaggerated in the statistical fugures. As shown in Figure, the decline in temporary workers in the 12 months after the crisis amounted to – 140,000 persons. If 2 / 3 of this decline was due to manufacturing, then the loss of employment in manufacturing industries in the first 12 months after September 2008 were roughly 345,000 instead of 252,000. Total employment in manufacturing industries fell from 6,638,000 in September 2008 to 6,385,000 in September 2009 and the share of temporary workers in those industries declined from 7.2 to 6.0 percent. Hence the loss of temporary employment accounts for 1.2 percentage points of the productivity slowdown in manufacturing industries. This is slightly less than 1 / 10 of the observed change. Therefore it has to be concluded that the massive fall in productivity cannot be explained by changes in the employment of temporary workers. The question remains why manufacturing firms did not adjust their workforce in the course of the crisis. As I argued elsewhere (see Möller 2010), the productivity slowdown is unlikely caused by non-adjustment due to employment protection.

Germany’s Job Miracle in the World Recession

17

Table 2 Changes in Employment by Industries

Total Economy Manufacturing Temporary work, manpower services Transport, storage Wholesale and retail trade, repair Information, communication Real estate activities Mining and quarrying Private household etc. Water supply, recycling Construction Agricult., hunt., forestry, fishing Scientific and technical services Energy Recreational, cultural, sporting activities Financial intermediation Other services Hotels and restaurants Business related services (excl. temp.work) Pub. Administration, defense, soc.sec. Education Health and social work

Employment on Sept. 30, Sept. 30, 2008 2009

Change Absolute Relative

27,995,661 6,637,452 718,274 1,439,919 4,107,861 850,834 218,700 95,776 62,625 220,872 1,621,561 219,998 1,557,642 234,646 224,950 1,010,096 822,038 828,388

27,799,973 6,385,318 577,681 1,409,598 4,078,282 831,091 211,043 93,078 63,398 222,227 1,623,788 224,247 1,561,945 239,480 232,085 1,017,978 833,905 848,630

–195,688 –252,134 –140,593 –30,321 –29,579 –19,743 –7,657 –2,698 773 1,355 2,227 4,249 4,303 4,834 7,135 7,882 11,867 20,242

–0.70 –3.80 –19.57 –2.11 –0.72 –2.32 –3.50 –2.82 1.23 0.61 0.14 1.93 0.28 2.06 3.17 0.78 1.44 2.44

1,099,453 1,666,028 1,059,845 3,294,820

1,121,081 1,693,971 1,099,274 3,428,129

21,628 27,943 39,429 133,309

1.97 1.68 3.72 4.05

Notes: Own calculations using data from the Statistical Department of the Federal Employment Agency; note that manpower services and temporary workers were excluded from business related services.

In former recessions, employment protection was not a major obstacle to reduce employment. There are several arguments that the restrictions for the firms’ behavior in the recent crisis are even less severe. For example, employment protection regulations for small firms have been abandoned and the share of temporary contracts has been substantially increasing. Rather than being forced to maintain their pre-crisis employment level, it is therefore more plausible that firms have voluntarily chosen not to downsize their labor force. This phenomenon is well known in labor economics as labor hoarding. Germany’s job miracle in the World Recession is closely related to this phenomenon.

18

Joachim Möller 3.2 The Theory of Labor Hoarding

The theoretical underpinnings of labor hoarding are given by literature starting with Bentolila and Bertola (1990) applying the real option theory of investment with uncertainty and irreversibility to the hiring / firing decision of firms (for an overview see Dixit and Pindyck 1994). If a firm is faced with a drop in the demand for its products that is perceived as temporary, it will typically choose some form of labor hoarding. The reason for this is that there are significant costs of adjusting the workforce. Downsizing the number of employees through lay-offs might imply severance payments, social compensation plans or simply indirect costs caused by a loss of reputation. In case of a recovery, finding new personal requires search activities and training measures which also involve some not negligible costs. Therefore, stabilizing employment might pay out for the firm although the under-utilization of the labor force is costly. More precisely, labor hoarding behavior is rational if the expected discounted present value of keeping a worker is lower than the sum of lay-off costs plus the expected discounted present value of future hiring and training costs. If the timing of the recovery is not known with certainty, the firm faces a nontrivial stochastic inter-temporal decision problem. Excluding a possible recall of workers, dismissals can be considered as an employer’s action that irreversibly destroys the former investment into the worker’s firm-specific human capital. The possibility of laying-off workers can be modeled as an option to the firm. Due to uncertainty there is a value of waiting before exerting the option. Hence a zone of inactivity exists, the extent of which depends on several parameters. It can be shown that typically the amount of sunk costs as well as the level of uncertainty raises the value of waiting. Theoretically, the probability for observing labor hoarding increases with optimism about future economic development, the amount of search and training costs or the loss of firm-specific human capital in case of dismissals. From an economic policy perspective, labor hoarding can be stimulated by paying wage subsidies during periods of slack demand.

3.3 Labor Hoarding in Manufacturing Industries

In principle, labor hoarding comes into play through two channels: (i) the reduction in working time and (ii) a decrease in labor density or in the output-orientation of labor. Both alternatives imply a reduction in workers’ productivity. Note that the second alternative additionally involves a slowdown of productivity per working hour. Using data for the branches of manufacturing, it is possible to disentangle both forms of labor hoarding. Table 3 first gives some descriptive statistics for aggregated two digit manufacturing industries. First one should stress that labor hoarding does not perfectly immunize employment in a sector from the shock. Major employment losses are visible in all big industries of the manufacturing sector: basic metals and metal

Germany’s Job Miracle in the World Recession

19

products (–73,000), transport equipment (–40,000), chemical and chemical products, non-metallic mineral products (–40,000), electrical, electronic and optical equipment (–37,000) as well as in machinery and equipment (–30,000). Only two industries experienced a (minor) employment gain: Food, beverages, tobacco (+1500) and repair (+7,500). Note that the job reductions in absolute terms do not necessarily indicate how strong the industry was affected by the crisis. The relative highest losses are found in textiles (–7.3%), basic metals (–6.5 %) and wood and wood products (–6.1). The size of the shock affecting the industries is closely related to their propensity to export.2 Figure 5 shows the correlation between the (negative) growth rates of turnover in the different manufacturing industries and their share of exports in 2008. Although there are two outliers—basic metals (see footnote 2) and repair—the negative relationship is statistically highly significant. Table 3 Changes in Employment by Manufacturing Industries Ind. Code

Employment on

Change

Sept. 30, 2008

Sept. 30, 2009

Absolute Relative

Total manufacturing

10 – 33

6,637,452

6,385,318

–252,134

Basic metals and metal products

24 – 25

1,136,783

1,063,477

–73,306

–3.80 –6.45

Transport equipment

29 – 30

955,501

915,468

–40,033

–4.19

Chemical, chemical products, non-metallic mineral products

19 – 23

1,050,060

1,011,003

–39,057

–3.72

Electrical and optical equipment, electronics

26 – 27

792,131

754,561

–37,570

–4.74

Machinery and equipment

28

996,746

966,801

–29,945

–3.00

Publishing, printing, reproduction

17 – 18

308,881

294,756

–14,125

–4.57

Textiles and textile products

13 – 15

144,129

133,682

–10,447

–7.25

Furniture and manufacturing n.e.c.

31 – 32

341,354

332,109

–9,245

–2.71

Wood and wood products

16

120,246

112,857

–7,389

–6.14

Food, beverages, tobacco

10 – 12

661,816

663,342

1,526

0.23

Repair and installation of equipment

33

129,805

137,262

7,457

5.74

Notes: Own calculations using data from the Statistical Department of the Federal Employment Agency.

2 Of course, the link to exports might also be indirect. This is the case, for example, if an industry is an important supplier to leading export industries as might be the case with basic metals as a supplier to transport industry or to machinery and equipment. Analyzing these input / output linkages is, however, beyond the scope of this article.

20

Joachim Möller

0

Repair

turnover growth 2008/09 -20 -10

Food

Print

Furniture

Wood Textile Transport

Chemical Electric

Machinery

-30

Metals

20

30

40 50 export share 2008

60

Notes: Own calculations with data from the Federal Statistical Office.

0

Figure 5: Relative Change in Turnover 2009 / 2008 and Export Share (2008) of Manufacturing Industries (in Percent)

Food Print Wood

Textile

Electric

Chemical

Furniture

Transport

Machinery

-25

productivity growth 2008/09 -20 -15 -10 -5

Repair

Metals

0

10 20 turnover growth 2005/2008

30

Notes: Own calculations with data from the Federal Statistical Office.

Figure 6: Turnover Growth 2005 / 2008 and Relative Change in Productivity (2008 / 2009) in Manufacturing Industries (in Percent)

40

Germany’s Job Miracle in the World Recession

21

As outlined above, a suitable (inverse) indicator for labor hoarding is the change in the productivity of the employees in the aftermath of a cyclical shock.3 If the growth of turnover during the economic upswing 2005 to 2008 could be taken as an indicator for optimism, then one would expect a positive correlation between labor hoarding and turnover growth in the past. Figure 6 shows that the turnover growth in the past is indeed negatively related to the relative change in productivity in response to the crisis. For example, basic metals and machinery and equipment were among the industries with the highest turnover growth in the upswing. Here the fall in productivity—which is at least partly due to under-utilization of the labor force—is especially strong. What corroborates this analysis is the fact that metals and machinery industries were among the industries reporting strong recruitment problems before the crisis (see Möller 2010). Recruitment problems in the past might be seen as a predictor of recruitment costs in the future. The theory of labor hoarding would therefore predict a positive relationship with labor hoarding. Production growth can be decomposed into relative change of three variables: employment, hours and productivity per hour. Note that the latter two reflect the two variants of labor hoarding. Figure 7 gives the results of the decomposition for the various branches of manufacturing in the time period 2005 to 2008. It can be seen that in the textiles sector turnover was shrinking. All other industries experienced positive turnover growth in the upswing period. Especially high turnover growth rates are found for machinery (+50 percent) and basic metals (+42 percent). These two branches also had the highest increase in employment together with the repair sector. The lion’s share of the turnover increase in the different industries, however, stems from the growth of productivity (as measured by the relative change of turnover per working hour). Variations in working hours are positive, but are generally of minor importance here. Figure 8 gives a corresponding decomposition for the time of the crisis (2008 / 2009). In this period only one industry (repair / installation of equipment) showed a growth of turnover, whereas the ten other industries under consideration here suffered from a more or less substantial decline. The shrinkage of turnover reaches from about 4 % in food, beverages, tobacco to almost 30% in basic metals. As can be seen from the figure, the employment responses are moderate when compared to the total size of the shock. Typically the sum of changes in productivity per working hour and working hours exceeds the employment change by far. For four out of the five industries with the biggest decline in turnover (basis metals, electrical products, machinery and transport) the relative change in productivity and working hours is of similar magnitude. Hence both components of labor hoarding play a similar role for buffering of the shock. For chemical products the reduction of productivity dominates the reduction in working hours. 3 Due to statistical reasons productivity growth here and in the following is measured as the relative change of turnover per employee or working hour.

22

Joachim Möller

60 50 40 30 20 10

r Re pa i

ro

M

El

ac hi ne ry Tr an sp or t Fu rn it u re

ec t

et al s M

Ch em

ic

al s

Pr in t in g

W oo d

il e s

-10

Te xt

Fo od

0

-20

Notes: Own calculations with data from the Federal Statistical Office; Relative changes of Employment (Empl.), Hours and Productivity (Prod.) as measured by production per working hour.

r ep ai R

Fu

rn

ns

itu

po

re

rt

ry Tr a

M

ac

hi

tro ec El

s et al M

ne

al C

he

m

ic

g tin in Pr

d W oo

es til Te x

Fo

od

s

Figure 7: Components of Relative GDP Change 2005 / 2008) in Manufacturing Industries (in Percent)

10 5 0 -5 -10 -15 -20 -25 -30 -35

Notes: Own calculations with data from the Federal Statistical Office; Relative changes of Employment (Empl.), Hours and Productivity (Prod.) as measured by production per working hour.

Figure 8: Components of Relative GDP Change 2008 / 2009) in Manufacturing Industries (in Percent)

Germany’s Job Miracle in the World Recession

23

The relative size of labor hoarding is comparatively low for the textiles and wood industries whereas it amounts to 80 percent and more for the industries that were strongly affected by the crisis. For example, the indicator of labor hoarding is 89 percent in machinery, 86 percent in basic metals, 84 percent in chemical products and 83 percent in transport.

4. Shocks and Institutions 4.1 Working-time Accounts and Alliances for Jobs

The importance of shock absorbing institutions has been increased over recent years. Together with a functioning social partnership in the export-oriented branches of the economy this has strengthened various channels for within-firm flexibility. An important instrument that has increasingly been used by firms is time flexibility through working-hours accounts. At the end of the nineties, 21 percent of firms were using working-hours accounts (Bellmann, Gewiese 2004). According to IAB establishment survey data this value increased to 32 percent in 2009. Among the exporting firms the share is even higher (40 percent). Moreover, the incidence of working time accounts increase with firm size. Therefore, the share of workers using working time accounts exceeds the share of firms.4 It seems plausible that especially exporting firms use the instrument as buffer stock against the volatility of demand at world markets. Altogether a sizeable surplus of working hours was accumulated on these accounts at the end of the boom period in the second quarter 2008 (see IAB 2009). These surpluses were increasingly reduced in the World Recession. According to calculations in Möller (2010), the quantitative effect amounts to more than 200,000 persons. It is highly plausible that the lion’s share took place in the exportoriented branches of manufacturing. A further phenomenon that has supported shock absorption is the increased cooperation between individual firms and their employees or worker councils. These so-called alliances for jobs usually imply that workers agree to wage restraint or even wage shortages in periods were the firm is under market pressure. In exchange the firm guarantees to refrain from dismissals. If job alliances were of greater importance, one would expect an effect on the total wage bill in case of a large negative shock to the economy. In a static world the change rate of the wage bill in such a situation should be less than the change rate of total hours. However, the wage bill also changes because of increases in the price level and trend productivity, for instance. I therefore subtracted the average change of negotiated wage in manufacturing industries from the change rate of the wage bill.5 Figure 9 compares 4 Already in 1999, 37 percent of workers were using working-hours accounts. Until 2003 the share increased to 41 percent (See Bauer, Munz 2005: 46, Footnote 10).

24

Joachim Möller

such an excess change rate of wages with the change rate of total hours across industries. It turns out that the so calculated change of the net wage rate is quite similar to the change in total hours. In four industries (textiles, wood / wooden products, printing / publishing and furniture) with declining total working hours the fall in the total wage bill exceeds that of total hours. For these industries there is some evidence that during the crisis wage restraint had some importance. For the big branches chemicals, metals, electro, machinery and transport at least a big effect is not visible at the aggregate industry level. Of course this result does not indicate that wage alliances were effective in some individual firms.

4 2 0 -2 -4 -6 -8 -10 -12 -14 -16

Total hours

Fo od Te xt ile s W oo Pr d in Ch ting em ica ls M et al s El ec tro M ac hi n Tr ery an sp o Fu rt rn itu re Re pa ir

wage bill

Notes: Own calculations with data from the Federal Statistical Office and the Deutsche Bundesbank; Net wage bill calculated as the growth of wage bill minus the average change rate of negotiated wages per hour in manufacturing industries including fringe benefits.

Figure 9: Change Rate of Total Hours and Net Wage Bill in Manufacturing Industries (2009 vs 2008 in Percent)

Firms exposed to the fluctuation of the world market typically possess some experience in dampening the effects of an external shock. Beside the instruments described above, the arsenal includes profit sharing schemes and the use of temporary workers. Furthermore, firms can react to demand fluctuations by transferring some of the risk to suppliers. In case of internationally active firms this might also be accomplished by a variation in the extent of cross-boarder outsourcing or offshoring. In addition, adjustment policies in periods of slack demand also include a higher effort in re-structuring and re-organizing production processes and intensifying the internal or external training of workers. Hence, during slumps the work5 The index of negotiated hourly wages in manufacturing industries (incl. mining and quarrying) was 132.5 in the last quarter 2008 and increased to 136.4 by the last quarter in 2009. This gives an average growth rate of 2.96 percent. The data source is Deutsche Bundesbank.

Germany’s Job Miracle in the World Recession

25

force will be partly employed in activities with no immediately measurable output. As a result, this contributes to the decline of productivity during recessions but increases the productivity reserves in the next upswing.

4.2 Short-time Work Schemes

Beside the other instruments also the usage of the short-time work subsidy scheme played a major role in dampening the labor market effects of the World Recession. This is especially true for manufacturing industries. The number of workers in short-time work reached the highest level since the early nineties when the instrument was used to dampen the labor market effects of re-unification in eastern Germany. Short-time work subsidies in the recent crisis have peaked in May 2009 when 1,516 million workers were affected. By December 2009, the number of short-time workers was reduced to 810,000 (see Figure 9). The lion’s share of short-time work can be found in manufacturing industries. Table 4 gives the numbers for December 2009. More than one fifth of all short-term workers in December 2009 were working in machinery and almost 14 percent in the automotive sector. Taken together, the five biggest industries in manufacturing accounted for more then half of all short-time workers. In these industries, between 12.1 and 17.3 percent of workers were using the subsidy. For firms the instrument has clearly reduced the cost of labor hoarding and thereby short-working time schemes have been an important stabilizing factor in the current recession. However, in fulltime equivalents the number of jobs protected by the instrument is by far not sufficient to fully account for Germany’s job miracle.6 Hence the other variants of within-firm flexibility also had to play an important role.

5. Conclusions The paper investigates the astonishing employment stability in response to the unparalleled real GDP shock of 2008 / 2009. The focus here is on manufacturing industries. We show that the basic channel of stabilizing employment is labor hoarding. Because of high export shares, manufacturing firms were heavily affected by the World Recession. When the crisis hit these firms they had just experienced an extraordinary boom with a sign of severe recruitment problems for qualified workers. In addition most of these firms are rather optimistic with respect to 6 One should bear in mind that the average shortage of working time for the affected workers is between 30 and 40 percent. Hence the calculated job protection is roughly one third of the total number of workers affected. With an average number of short-term workers in 2009 of slightly more than one million, the job protection effect can be calculated as between 300,000 and 400,000. As I have shown elsewhere (Möller 2010), the job losses of a 5 percent drop in GDP would be in the order of magnitude of at least 1.5 million in case of total adjustment of the work force.

26

Joachim Möller 1600 1400 1200 1000 800 600 400 200

20 08 :M 9 20 08 :M 10 20 08 :M 11 20 08 :M 12 20 09 :M 1 20 09 :M 2 20 09 :M 20 3 09 :M 4 20 09 :M 5 20 09 :M 6 20 09 :M 7 20 09 :M 8 20 09 :M 9 20 09 :M 10 20 09 :M 11 20 09 :M 12

0

Notes: Own calculations with data from the Federal Statistical Office and the Deutsche Bundesbank; Net wage bill calculated as the growth of wage bill minus the average change rate of negotiated wages per hour in manufacturing industries including fringe benefits.

Figure 10: Usage of Short-time Work Schemes (September 2008 to December 2009, in 1000 Persons)

future prospects on the world market. Hence labor hoarding appeared to be a rational strategy. Following this strategy was eased by several institutional conditions. Among these are working-hours accounts, collective agreements for flexible working time and—last but not least—the short-time working hours scheme. However, the latter is not sufficient for being the only explanatory factor in face of the enormous cushioning of employment in the World Recession. One can conclude that the labor market response in German manufacturing industries during the World Recession of 2008 / 2009 is an unprecedented phenomenon. It is mainly due to favorable initial conditions in the boom 2007 / 2008 and an optimal interplay between behavior of employers and employees or their representatives, shock absorbing institutions and the consequent use of labor market instruments. Provided that the current severe recession will not transform into a structural crisis and most of the affected firms will soon recover from the tremendous decline in the demand for their products, the Germany’s job miracle will become a textbook example of how to immunize the labor market from a major macroeconomic shock

Germany’s Job Miracle in the World Recession

27

Table 4 Usage of Short-Time Work Subsidies (Major Industries, December 2009) Industry Machinery Automotive Metal Prod. Basic Metals Electric Prod.

Total number of short-time workers

Share in total short-time work

Share of workers in short-time work

168,000 112,000 99,000 45,000 41,000

20.8 13.8 12.2 5.9 5.1

17.3 13.7 12.6 14.5 12.1

Notes: Data Source Statistical Department of the Federal Employment Agency.

References Bach, H.-U. / Hummel, M. / Klinger, S. / Spitznagel, E. / Zika, G. (2009): Arbeitsmarkt Projektion 2010, Die Krise wird deutliche Spuren hinterlassen, IAB-Kurzbericht 20. Bach, H.-U. / Spitznagel, E. (2009): Kurzarbeit: Betriebe zahlen mit – und haben was davon. IAB-Kurzbericht 17 / 2009. Bauer, F. / Munz, E. (2005): Arbeitszeiten in Deutschland: 40plus und hochflexibel, WSI-Mitteilungen, 1 – 2005, 40 – 48. Bellmann, L. / Gewiese, T. (2004): Die Dynamik der Nutzung von Arbeitszeitkonten – Theoretische Ansätze und Längsschnittanalysen mit dem IAB- Betriebspanel. In: Bellmann, L. / Schnabel, C. (Hrsg.): Betriebliche Arbeitszeitpolitik im Wandel. BeitrAB 288, Nürnberg, 139 – 184. Bentolila, S. / Bertola, G. (1990): Firing Costs and Labour Demand: How Bad Is Eurosclerosis?, Review of Economic Studies, 57(3), 381 – 402. Blanchard, O. / Wolfers, J. (2000): The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence, Economic Journal, 110(462), C1 – 33. Dixit, A. K. / Pindyck, R. S. (1994): Investment under Uncertainty, Princeton University Press, Princeton, N.J., 1994. Fryges, H. / Wagner, J. (2008): Exports and profitability first evidence for German manufacturing firms, Leuphana-Univ. Lüneburg, Working Paper Series in Economics 102, http:// hdl.handle.net/10419/28220. IAB (2009): Institut für Arbeitsmarkt- und Berufsforschung, Durchschnittliche Arbeitszeit und ihre Komponenten in Deutschland, http://doku.iab.de/grauepap/2009/tab-az09q3.pdf. Möller, J. (2010): The German Labor Market Response in the World Recession—De-mystifying a Miracle. In: Zeitschrift für ArbeitsmarktForschung 42(4), 325 – 336.

Crisis, What Crisis? Patterns of Adaptation in European Labor Markets By Werner Eichhorst,* Michael Feil,** and Paul Marx*** Abstract The current crisis, while of a global nature, has affected national labor markets to a varying extent. While some countries have experienced a steep increase in unemployment, employment in other developed economies has not fallen in parallel with a significant decline in GDP. Our analysis shows that labor market institutions frequently used to study employment performance can explain the development of unemployment in the situation of crisis in some clusters of countries much better than in others. One major factor to be incorporated in capturing national variations is the role of internal flexibility, in particular working time adjustment. This calls for a broader concept of labor market flexibility which takes into account different channels of adjustment. Keywords: Labor market institutions, internal flexibility, employment protection, economic crisis JEL Classification: J21, J23, J58

1. Introduction Although the current crisis is global, the effects felt by the national labor markets have varied to different degrees: some countries have witnessed a rapid rise in unemployment; others have not experienced the same fall in unemployment as in GDP. Particular countries with more regulated labor markets and developed instruments for internal flexibility have so far seemed to be better prepared to contain job losses, as the recent debate about the German “labor market miracle” shows (Möller 2010). Hence, among other factors such as the structure of the economy or macro-economic policies, labor market institutions are expected to help explain cross-country differences. This paper addresses the topic of institutional determinants of labor * IZA. Corresponding author: Werner Eichorst, IZA, Schaumburg-Lippe-Str. 5 – 9, 53113 Bonn, Germany. Phone: +49 (228) 3894 – 531, Fax: +49 (228) 3894 – 510, [email protected]. ** IAB. *** IZA. The authors thank Tim Schwarzmüller and Max Sties for their research assistance, as well as Antje Mertens for her insightful suggestions. Any remaining errors are our own.

30

Werner Eichhorst, Michael Feil, and Paul Marx

market adjustment from two different perspectives: first, we review existing research on the role of shocks and institutions and try to replicate previous econometric findings with new data. This is justified by a number of new indicators which have become available over recent years. Second, we suggest a new measure for labor market flexibility. This helps us to understand which role different institutional frameworks play in determining labor market performance in both the short and long run. The empirical part is based on both quantitative work covering the EU and OECD countries as well as on an in-depth analysis of selected national cases. The quantitative part uses standard indicators for external labor market flexibility. It shows that this set of factors can explain the development of unemployment in earlier crises and accounts for some variation of unemployment in the most recent period. However, since the explanatory power decreases for more recent data, a full understanding of current adjustment processes obviously requires a broader picture of labor market flexibility. In particular, enterprise-level strategies regarding working time adjustment and ad-hoc policy interventions “bias” the expected impact of institutions on short-term performance. This effect is arguable stronger in the current crisis, which led most policy makers in affluent economies to stabilize the labor market, e.g., by bailing-out vulnerable sectors. We therefore include case studies that account for hard-to-measure aspects of labor market flexibility, such as working time accounts and government sponsored adjustment.

2. Labor Market Institutions and Labor Market Adjustment Labor market institutions are an important variable in explaining divergent labor market developments across countries and over time. The European phenomenon of persistent mass unemployment following the “golden” postwar era shifted the focus of comparative labor market analysis. Instead of shocks, institutional rigidities were increasingly regarded as explanatory factors for employment outcomes (Blanchard 2006). Since the 1990s, institutions have been at the core of the debate about varying national labor market patterns. A large number of theoretical and empirical studies suggest, at least partly, a causal relation between institutional arrangements and labor market performance (Nickell 1997, Nickell et al. 2005, Bassanini and Duval 2006). It is argued that adverse institutions lead to a persistent deviation from labor market equilibria by distorting price- and wage-setting mechanisms, which take more time to adjust (Layard et al. 1991 and 2005). In this sense, four groups of labor market institutions are typically accused of creating or increasing unemployment: the wage-setting arrangement, unemployment benefits, taxation, and employment protection. Active labor market policies form a fifth group of institutions with significant (yet beneficial) influence on employment outcomes. Research into the interaction of institutions and shocks (Blanchard and Wolfers 2000, Bassanini and Duval 2006) aims at increasing the explanatory power

Crisis, What Crisis?

31

as institutions influence both the speed of labor market adjustment as well as the persistence of labor market imbalances. Shocks tend to produce persistent unemployment if there is wage rigidity, a generous benefit system, and few effective activation policies (see also Bassanini and Duval 2009). Yet another strand of literature emphasizes interactions between institutions and thereby paints a more complex picture of their macroeconomic and labor market effects. Institutional complementarities can explain why more regulated types of labor markets can produce favorable outcomes despite their rather “inflexible” institutions (Hall and Soskice 2001, Amable 2003). Starting from the observation that institutional factors do not work in isolation but form complex institutional arrangements, typologies of employment systems assume that diverging but coherent models could have potentially equal capacities to generate favorable employment outcomes. This work has so far been mainly theoretical, and empirical contributions often aim at identifying national patterns rather than providing econometric evidence for their labor market impact. Modeling complex institutional networks and understanding their impact on macroeconomic or labor market performance remains a challenge for comparative research. However, there are empirical findings that indeed suggest a multi- or at least dual-peaked landscape rather than one superior model outperforming the rest (Freeman 2000, Bassanini / Duval 2006). In this paper, we would like to complement the debate on labor market flexibility by pointing attention to some often neglected variables: in our understanding, focusing on “classical” institutions determining external flexibility risks neglecting important aspects of flexibility. To fully understand national institutional patterns, the analysis has to be extended to additional factors. The most important one is education (including early-childhood education, vocational training, and life-long learning). Another blind spot of most research is internal flexibility. This leads to biased results, as in many countries flexible working time models provide a compensation for rigid hiring-and-firing practices. Besides working time, organizational aspects make a major contribution to internal flexibility. Versatile workers provide an opportunity to respond to changing external conditions without drawing on resources from the external market. A note of caution, however, has to be made regarding the robustness of most findings. In general, one can argue that this type of research is less suitable for regression analysis since it most often relies on complex institutional analyses with strong qualitative evidence, which can be seen as less rigorous in empirical terms (Howell et al. 2007). Thus, we can assume that a sufficient degree of flexibility is needed for good employment performance—but this flexibility can be achieved in various ways. While there may be similar overall employment rates, different patterns of flexibility can be associated with diverging distributional outcomes—and different paths in coping with the crisis both in the short and long run.

32

Werner Eichhorst, Michael Feil, and Paul Marx

3. Mapping Patterns of Labor Market Flexibility In order to capture different patterns of labor market flexibility in OECD countries, we refer to a widely used typology developed by Atkinson (1985). It is based on two distinctions: first, it differentiates between internal and external flexibility, i.e., strategies which are applied inside a company or to the outside labor market, respectively; second, flexibility either derives from variation of workload (numerical) or from organizational adaptability (functional). A third way is to adjust wages and labor costs to the economic situation. Considering these criteria, five types of flexibility can be distinguished: (1) External numerical flexibility means the possibility to adapt the number of employees to the economic situation through layoffs or hiring (permanent or temporary). Determinants of external numerical flexibility are the extent of employment protection and the quantitative availability of manpower. The latter is influenced by features of the benefit system influencing labor supply. The same is true for taxation. High tax wedges and marginal effective tax rates can lead to several “traps” that provide incentives to remain inactive, unemployed or in a low-paid job. (2) Internal numerical flexibility is achieved without variations in the number of staff. It allows for adjustment of working time in order to meet a company’s current utilization rate. (3) External functional flexibility requires a skilled labor force adaptable to structural changes. Skill mismatch in the labor market can be avoided by promoting occupational mobility through active labor market policies, in particular job placement and training. A prerequisite for an adaptable labor force is the provision of appropriate primary, secondary, and tertiary education. (4) Internal functional flexibility means the ability to react to changing demand with a flexible organization of the production process. This requires broad and well-educated employees able to perform different tasks. Thus, investment in firm-specific human capital via continual (internal) qualification makes a major contribution to this type of flexibility. (5) Wage flexibility is when real wages can respond to changing macroeconomic conditions such as shocks. Wage rigidities result from features of a wage-setting regime such as statutory or collectively agreed minimum wages. As can be seen from the discussion above, Atkinson’s flexibility types are determined by various institutional factors. In our analysis we include eight groups of indicators than can be used as proxies for real-world institutional complexities (see Eichhorst et al. 2008 for a more detailed discussion). 1. Wage setting, 2. Unemployment benefits, 3. Taxes on labor,

Crisis, What Crisis?

33

4. Employment protection, 5. Active labor market policies, 6. Education and training, 7. Working time arrangements, 8. Human capital investment in a firm-specific setting. Based on a systematic aggregation of these institutional parameters, an overall indicator of labor market flexibility (see Annex 2) as well as a typology of countries can be established (for methodological details see also Eichhorst et al. 2009). Figure 1 provides the most recent institutional information on OECD countries regarding external and wage flexibility with the Anglo-Saxon countries and Switzerland, but also Denmark, at the top. USA Canada

23.9

17.5

12.9

20.0

15.5

10.5

Japan

22.1

15.4

8.7

United Kingdom Norway

19.2

9.6

15.0

23.2

9.0

7.6

Belgium

6.2

Portugal

5.9 0

16.3

11.5

9.4

11.3

7.3

Slovakia

17.7

7.5

13.3

Italy

16.1

7.1

16.4

Czech Republic

18.9

9.3

11.7

France

16.4

8.6

15.0

Hungary

19.8

9.4

12.4

Germany

17.7

8.4

16.1

Finland

17.5

13.4 16.2

8.5 5

15.6

9.4

18.7

Sweden Netherlands

15.0

9.8

19.3

Australia

18.3

13.6

13.5

Spain

15.6

17.0

14.5

Poland

19.0

18.0

11.4

Ireland

17.9

17.8

12.9

Austria

16.6

11.0

21.2

New Zealand

16.8

13.9

18.1

Denmark

20.6

14.6

14.5

Switzerland

24.2

20.2

15.1

Korea

10

15

20

25

30

35

40

45

50

55

60

Figure 1: Aggregate Flexibility Indicator (Version 1: External and Wage Flexibility, 2007)

Figure 2 also incorporates information on internal flexibility. Due to data limitations this is restricted to the EU and 2003. This version of the indicator shows the extent to which different flexibility options are available in European countries. Each country has several options, but both the overall aggregate level and the

34

Werner Eichhorst, Michael Feil, and Paul Marx

structure of flexibility varies across countries. Germany, for example, stands out as a champion in internal numerical flexibility (working time); whereas the UK shows much larger external numerical and wage flexibility. Denmark seems to provide a full range of flexibility channels; whereas in Southern Europe flexibility is more restricted (except for temporary employment).

Denmark Finland Ireland Germany United Kingdom Sweden Netherlands Czech Rep. Austria Hungary Belgium France Poland Italy Portugal Spain 0

10

20

30

40

50

60

70

80

Figure 2: Aggregate Flexibility Indicator (Version 2: Internal, External and Wage Flexibility, 2003)

Based on this indicator, clusters with similar patterns of flexibility can be formed. The groupings strongly resemble established typologies (Figure 3). The first cluster can be labeled “education-based” and comprises nine Scandinavian and Continental-European countries. It is characterized by low external numerical and wage flexibility, while external functional flexibility is high. The second cluster (Asian and the Anglo-Saxon countries) combines high external numerical and wage flexibility with intermediate external functional flexibility and can be called “market-oriented I.” The third group consists of the four CEE countries, which have high external functional and wage flexibility but very low external functional flexibility (“market-oriented II”). Finally, the Southern European countries find themselves in a cluster which is characterized by rather low flexibility in all dimensions (“low flexibility cluster”). Results of the second cluster analysis including internal flexibility are reported in Figure 4. As for external flexibility, Continental and Scandinavian countries are quite similar, although Denmark and Sweden have to be treated as a separate cluster. Both exhibit strong internal flexibility, while internal functional flexibility prevails.

35

Low

Market II

Market I

Education

Crisis, What Crisis?

Figure 3: Cluster Tree External / Wage Flexibility, 2003 (Hierarchical Cluster Analysis, Dendrogram Using Ward Method)

Figure 4: Cluster Tree overall Flexibility, 2003 (Hierarchical Cluster Analysis, Dendrogram Using Ward Method)

Overall, functional elements (internally and externally) are very important in Denmark and Sweden, and so their cluster is termed “functional model.” The remaining Scandinavian countries, together with the Continental ones, form the second

36

Werner Eichhorst, Michael Feil, and Paul Marx

cluster. They are characterized by an extensive use of working time arrangements (especially Finland, France, and Germany), but only by intermediate functional flexibility. In the third cluster, internal flexibility is clearly underdeveloped. It is formed by three Southern European countries (“low flexibility model”). The fourth cluster consists of cases that reveal very high external and wage flexibility and is therefore labeled “market model.” The heterogeneous group comprises two AngloSaxon and three CEE countries. Contrary to what one might expect, most of these countries do not completely eschew internal flexibility but, instead, show an average level for both sub-indicators.

4. Different Models of Flexibility and the Current Crisis OECD member states have been hit differently by the economic crisis. There are even a few countries which did not suffer a drop of GDP (Australia, Poland, and South Korea). Figure 5 relates changes in GDP to changes in employment. GDP growth rates per se, however, have to be analyzed with care, since they do not reflect the deviation from trend growth. Controlling for differences in trend growth rates, however, does not change the pattern very much. Comparisons with historic recessions after 1960 show that the drop in output has not been unprecedented in countries like the US, Japan, or Italy, where the first oil price shock resulted in a larger absolute deviation of actual from trend GDP than the latest financial turmoil. A look at the change of GDP to trend growth ratio, however, reveals that the economic and financial crisis is rather prominent.1 In any case, there is considerable divergence in the relation between GDP and employment in the current crisis, as Figure 5 shows. A two-tier pattern of employment-GDP changes emerges from Figure 5. A larger group below an imaginary 45°-degree line shows rather moderate employment losses. The champions in preserving employment in the crisis at this particular snapshot are Germany and the Netherlands. On the other hand, the big exceptions in this respect are Portugal, the US, Spain and Ireland. Obviously, we have to be aware of the fact that the impact of the crisis on labor markets may not yet be fully apparent. So what looks like a miracle in the short run could still turn out to be rather normal in the medium term, since it simply takes different periods of time until national labor markets react. However, there has so far been little evidence that many countries of the “small-reaction” group on the right in Figure 5 will follow the “outliers” on the left.

1 We use a standard Hodrick-Prescott filter to derive trend growth rates. Möller (2010: 328) applies the same method for Germany only.

Crisis, What Crisis? -10

37

-6

-2

Decline GDP (%)

-2

-7

D

Decline Employment (%)

-12

Source: OECD.

Figure 5: Decline in Employment and GDP between Country-specific Peaks before the Crisis and Third Quarter of 2009

A good starting point for our discussion of different adjustment patterns would be reliable figures on the volume of work throughout the OECD. The existing data is compiled in Table 1. To a large extent, changes in hours of work conditional on changes in GDP are small. Actually, the figures are quite puzzling. It is hard to believe that there are countries suffering from a large drop in GDP that have experienced constant or even slightly rising hours of worked. Whether the reasons for the differences between changes in GDP and changes in hours of worked are real or due to measurement error and timing problems cannot be solved here. It should be clear that a drop in GDP which is not matched by a similar reduction in total hours of work implies a large decrease in labor productivity. This is at odds with existing data showing only slow-downs in measures of productivity growth in economic downturns, as opposed to actual decreasing productivity figures, which are hard to find. This is not to say that there is no evidence of a fall in hourly labor productivity (Möller 2010: 334); but the gap appears simply to great to be explained by excessive labor hoarding. Table 1 also presents the adjustments during the recent recession by clusters. First and foremost, the country groupings based on more long-standing institutional patterns are only to a limited degree meaningful in the short-term perspective. For most variables, there is more intra than inter-cluster variation. This holds true for GDP (which is not surprising), unemployment rate and employment (which is more puzzling). Unemployment reacts more swiftly in market-oriented countries; however, this is largely driven by Ireland. The heterogeneity within clus-

38

Werner Eichhorst, Michael Feil, and Paul Marx

ters points once again to the relevance of crisis-specific interventions, which obviously distort the effect of institutions. Where the clustering is more obvious is in the adjustment of temporary employment, since here (as expected) only educationbased and southern “low-flex” cases show a notable reaction. Note that not only the usual suspects Spain and Italy rely on temporary workers as a flexibility source but that this path is also prevalent in the Scandinavian cases (except for Denmark). Regarding hours worked and wages, there is no clear pattern (although we clearly face a data problem here). Table 1 Adjustment by Country and Cluster. Percentage Change from Country-specific Beginning of Crisis (in Terms of GDP) to last Available Data Country

GDP

Unempl. rate

Empl.

Temp. empl.

Hours worked

Hourly earnings

Australia Canada Ireland Japan Korea New Zealand United Kingdom USA Market-oriented I

0.6 –3.2 –10.5 –7.2 0.7 –2.0 –6.0 –3.3 –3.9

36.3 39.1 175.2 39.9 17.2 84.3 51.9 81.7 65.7

–0.1 –1.8 –9.3 –2.2 0.0 –1.4 –1.9 –4.6 –2.7

0 2.7 –13.7 0 0 0 –0.8

0 0 –5.4 0 0 0 –1.5

–1.7

–1.0

0.2 –3.7 10.7 –6.7 –2.6 4.9 2.3 3.5 1.1

Austria Belgium Switzerland Denmark Finland Germany Netherlands Norway Sweden Education-based

–4.0 –3.4 –2.0 –6.4 –8.8 –5.6 –4.7 –1.3 –5.9 –4.7

45.0 18.3 31.7 101.1 33.2 -0.4 29.4 11.8 45.8 35.1

–0.7 0.0 0.5 –2.8 –3.5 –0.6 –0.7 –1.5 –2.5 –1.3

5.3 –0.2

–2.4 –0.7

0.7 3.1

0.6 –18.0 –1.7 –0.8 –14.9 –11.0 –7.6

–3.1 –1.6 –1.9 –1.7 –0.3

1.9 3.3 2.5 3.7 3.0 1.2 2.6

Czech Republic Hungary Poland Slovak Republic Market-oriented II

–4.1 –8.0 1.0 –5.5 –4.2

69.0 37.7 21.7 39.1 41.9

–1.8 –3.1 0.3 –4.1 –2.2

0.0 8.6 4.4 –4.0 2.3

–2.9 –2.7 0 –1.9

2.5 7.7 1.9 3.7 4.0

Spain France Italy Portugal Low flexibility

–4.5 –3.0 –5.9 –2.9 –4.1

103.3 25.8 19.4 30.1 44.6

–8.7 –0.8 –2.0 –3.8 –3.8

–20.4 –6.1 –12.7 –9.9 –12.3

1.0 0.3 –1.7 1.2 0.2

2.8 3.6 4.6 –5.6 1.4

Source: OECD.

–1.7

Crisis, What Crisis?

39

5. Quantitative Analysis 5.1 Assessing the Role of Labor Market Institutions

The basic idea for the kind of econometric research undertaken here stems from the “shocks-and-institutions” literature, where the focus is on the equilibrium or structural unemployment rate. We explore here the extension of the framework to changes in actual unemployment rates. The reason for this is rather obvious: any statement about the role of labor market institutions in the current crisis could at best hold for the actual unemployment rate, since the equilibrium rate will clearly depend on the effects of the crisis. In a sense we are shifting the yardstick in the literature from a decent explanation of the medium-term development to the short run. For part of the literature, this may seem inappropriate. It has been argued (see Blanchard and Wolfers 2000, among others) that the slow movements of institutions only justifies a model in which all variables are averaged over a longer period of time, most commonly five years. Nickell et al. (2005) on the other hand use annual data. Baccaro and Rei (2007) discuss these alternative ways of assessing the influence of shocks and institutions. They estimate both types of models (with annual and five-year-averaged data) and conclude that there is little evidence supporting the deregulatory view. An argument in favor of annual data and the short-run approach can be found in the interaction of shocks and institutions. If institutions shape labor market adjustments due to major changes in macroeconomic conditions, then their effects should show up in the short as well as the medium run. A typical example for this kind of reasoning is the role of employment protection. If existing jobs are well sheltered by strict labor laws, we will expect (un-)employment to react only slightly in response to adverse macroeconomic shocks. However, short- and medium-run effects could greatly differ if for instance strict employment protection prevents not only dismissals but also new hirings. Under these circumstances, one could expect deferred increases in unemployment and possibly similar medium-run effects more or less independent of differing labor market institutions. An obvious approach to the problem of lagged adjustments would be the estimation of an error-correction model. This method, however, requires the existence of a cointegration relationship between unemployment rates, shocks and institutions. Individual and joint properties of these series become quite tricky in the context of a panel model. Panel unit root tests typically indicate that at least some of the unemployment series contain a unit root. On the other hand, shock and institutional variables are mostly found to be stationary. It is thus not very surprising that the existence of a cointegration relationship can hardly be proved by available tests for panel cointegration. Another, although certainly simpler, way of dealing with the data is the estimation of models in first differences. This choice can be justified directly by nonstationarity and missing cointegration. Baccaro and Rei (2007) call the first-differ-

40

Werner Eichhorst, Michael Feil, and Paul Marx

encing strategy a “safety device.” This is also our understanding. However, there is also a very substantial argument to use first differences, since our interest is mostly on changes in key labor market variables and much less on the long run.2 The models that we run should be understood as first attempts to see how much explanatory power there is in shocks and institutions. We do not try to maximize the fit of the models, nor do we try to minimize the prediction errors. Loosely speaking, we want to see what institutions can explain and what they cannot. Our hypotheses about the role of institutions is that in economies with rigid labor markets, adverse macroeconomic shocks lead to a smaller increase in unemployment than in economies with flexible labor markets. Our starting point is the model, expressed in equation (1):3 …1†


urit ˆ 
uri;t

1

‡

X k

k xkit ‡

X l

l zlit ‡

X

m xkit
zlit ‡ ai ‡ "it :

m

We allow for country fixed effects …i † We also experimented with time effects, which are usually interpreted as common unobserved shocks. Since for most of the years they turned out insignificant, we decided not to include them. The model includes k macroeconomic (shock) variables …xit †, l labor market institutions …zit † and m ˆ k
l interaction terms of macro variables and institutions to pick up possible effects of institutions depending on macroeconomic conditions. We expect these interactions to be a function of the level of the institutions and not of their change.4 The data we use are closely related to those in Nickell et al. (2005). To some extent, we draw directly on their database. Most of the time series are taken from the OECD’s Source Database. This holds for all macroeconomic and for most of the institutional variables. Information on bargaining coordination is taken from the ICTWSS: Database on Institutional Characteristics of Trade Unions, assembled by the Amsterdam Institute for Advanced Labour Studies (AIAS). We constructed all the shock variables according to Nickell et al. (2005) except for the TFP shock, where we follow Bassanini and Duval (2006). The underlying idea of data gathering was the production of time series covering a maximum period of time. To do this, we also made direct use of Nickell et al. (2005) data. This refers to cases in which time series stored at the OECD are shorter than those in the Labour Market Institutions Database underlying the Nickell et al. (2005) paper. 2 Baccaro and Rei (2007: 550) also argue that the model in first differences is preferable because then there is no indication to include the lagged dependent variable, i.e., the first difference of the lagged unemployment rate. Our data does not completely confirm this argument, since we find for several specifications a significant coefficient for the lagged dependent variable. This is probably due to the fact that we use annual data as opposed to the fiveyear averages in Baccaro and Rei. 3 Tested by xtserial and abar in Stata. 4 We are aware of the possibility that interaction terms could be I…1†. Standard unit roots tests, however, do not support this hypothesis.

Crisis, What Crisis?

41

Our modeling strategy can be summarized as “general to specific.” The model outlined in (1) is a rather general way of addressing the data. We consider five macroeconomic variables: output gap, total factor productivity (TFP) shock, labor demand shock, real interest rates, and a terms-of-trade shock. The institutions block also consists of five variables: benefit reduction rates (BRR), trade union density (TUD), employment protection (EP), bargaining coordination (BC), and the tax wedge (TW). In the fully fledged version, this amounts to 35 coefficients to be estimated for the shocks, institutions, and interaction terms. Already from an ex-ante perspective this number appears fairly high. On the other hand, one should bear in mind that also interaction terms between institutions have been used in the literature (see, e.g., Belot and van Ours 2001). Adding this kind of interactions would obviously increase the number of parameters to be estimated even further. These early doubts about the number of independent variables were confirmed by running regressions with the full set of the complete variables or at least large subsets of them. Typical features of these models include large standard errors, wrongly signed coefficients and a general lack of robustness. We think that these problems can be traced back to a high degree of interdependence of the explanatory variables, or technically speaking, to some degree of multicollinearity. To cope with this issue, we resort to data-reducing methods. We use principal component analysis to reduce the number of shocks and institutions variables respectively to one. There is one exception on the institutions side. Since bargaining coordination is typically assumed to have a negative influence on unemployment, we do not include this variable in the institutions amalgam, consisting of the benefit replacement rate, trade union density and employment protection. Table 2 reports our baseline estimates for the full sample using alternative methods. Due to existence of significant autocorrelation, which persists even if we include the lagged endogenous variable as a regressor, we run different type of feasible generalized least squares estimator. We also employ the method proposed by Beck and Katz (1995), i.e., OLS estimates with panel-corrected standard errors (PCSE), correcting for serial correlation (AR(1)), heteroscedasticity and contemporaneously correlated errors across panels. The differences between the estimators turn out to be non-negligible, but qualitatively quite similar. The biggest difference is the estimated coefficient for the lagged unemployment rate, a finding that is perfectly in line with the different treatment of the data once an AR(1) process is explicitly modeled. Due to the interaction term, the coefficients on the macro variable and the labor market institutions index cannot be interpreted on their own. The marginal effect of macro shocks on changes in unemployment amounts to –.34, based on the estimates in the last column of Table 2. It is almost identical to the estimated coefficient of the “macro” variable itself. Since here, marginal effects are computed at sample means, we can conclude that for a hypothetical country with average labor market institutions, no considerable influence of institutions exist. Further calculations show that this finding holds almost uniformly across all countries. Put differ-

42

Werner Eichhorst, Michael Feil, and Paul Marx

ently, institutions do not seem to play any significant role in shaping the adjustments of the unemployment rate to macroeconomic shocks. As a corollary, we can note that the marginal effect of institutions on unemployment is not different from zero. Table 2 Fixed-effects Panel Models Using Different Estimators

Lagged Dependent Variable Macro Shocks Labor Market Institutions Interaction: Shocks-lntitutions Bargaining Coordination

FGLS, heterosked. Panels

PCSE

coef / se

coef / se

0.306*** (0.029) –0.255*** (0.017) 0.036 (0.039) –0.011 (0.015) 0.025 (0.023)

0.308*** (0.063) –0.289*** (0.035) 0.053 (0.086) –0.015 (0.019) 0.029 (0.032)

FGLS, PCSE, heterosked. AR1 Error Panels, AR1 Error coef / se coef / se 0.147*** (0.029) –0.307*** (0.016) 0.049 (0.049) –0.008 (0.015) 0.030 (0.026)

0.133** (0.058) –0.336*** (0.030) 0.066 (0.098) –0.021 (0.019) 0.027 (0.035)

note: ***p < 0.01, **p < 0.05, *p < 0.1.

These findings can be confirmed as we move on to separate models for three groups of countries based on the cluster analysis of section 3. The estimates for the three separate models—all estimated by FGLS with corrected standard errors—are reported in Table 3. The first group, which could be termed “Continental Europe,” comprises Austria, Belgium, Finland, Germany, the Netherlands, Norway, Sweden and Switzerland. In the second column of the table, the results for the “Anglo-Saxon” countries (Australia, Canada, Ireland, New Zealand, the United Kingdom and the United States) and Japan are given. Finally, the third column contains the regression output for the “Southern European” group (France, Italy, Portugal and Spain). As expected from the full model, including all 20 countries, the differences of the estimated coefficients on the macro variable are not that large. Turning directly to the marginal effects the impact of macro shocks on unemployment is largest in the Anglo-Saxon countries (–0.42), followed by “Continental Europe” (–.34). But even in “Southern Europe” the effect amounts to –.31. Moreover, we find large differences in the marginal effects of institutions evaluated at sample means. The effect of institutions is significant in the “Anglo-Saxon” and the “Southern Europe” group. While in the former institutions increase unemployment changes, they reduce them in the latter. “Continental European” coun-

Crisis, What Crisis?

43

tries range somehow in between. Since the sign of the “institutional effect” is undetermined from a theoretical point of view, this could be either due to offsetting partial effects of different labor market institutions as well as the result of a dominant role played by employment protection. The literature on labor market institutions and macroeconomic performance has stressed for a long time that employment protection has ambiguous effects. Finally, we cannot rule out the possibility that the insignificance result is due to an inappropriate clustering strategy. This can be seen from the different signs and significance levels the variable shows in the three separate regression models. Nevertheless, with respect to our basic research question these points are only of incidental interest, since we assume by and large that labor market institutions do not react to the crisis. For this reason, we have also chosen to include the institutions variable in levels and not in deviations. Surely, institutions can and to some extent were adjusted during the crisis, but these changes are typically confined to passive labor market policies such as unemployment benefits (and in particular benefit duration) and short-time working schedules. With respect to the central research question whether and to what extent labor market institutions cause unemployment our empirical approach is not suited to provide a substantial contribution. Comparing the three separate models with the panel including all 20 OECD countries, we note that there is heterogeneity that is not accounted for by allowing for country-specific effects. The estimated fixed effects in our models are almost always insignificant. So while there seem to be differences with respect to the slope parameters, there are no indications of country-specific trends.5 This is also a result of the rather long period of time covered by our data. The use of the two clustering strategies—principal component analysis and separate models for different groups of countries—makes the empirical models more robust. Robustness has been an issue in the literature for many years. There have always been doubts whether the models used also hold for slight variations in specifications and data (Baker et al. 2003, 2005; Baccaro and Rei 2007). We do not find such instabilities. Our main finding, that institutions do not matter much in shaping the short-run labor market responses, holds not only for different estimation methods but also for models that include lagged variables.

5 From a methodological point of view, the differences found by running separate models for the three country clusters raises the issue of “poolability.” Nickell et al. (2005) claim that the Roy-Zellner test (see Baltagi 1995) does not reject the assumption of common slopes. This finding is clearly at odds with our experiences. We do not find econometric support for the notion of “one model fits all.”

44

Werner Eichhorst, Michael Feil, and Paul Marx Table 3 Regressions by Country Groups Using PCSE, Correcting for Serially Correlated Error Terms

Lagged Dependent Variable Macro Shocks Labor Market Institutions Interaction: Shocks-lntitutions Bargaining Coordination

Continental Europe coef / se

Anglo-Saxon coef / se

Southern European coef / se

0.067 (0.063) –0.207*** (0.038) 0.134 (0.107) –0.149*** (0.032) 0.160** (0.076)

–0.047 (0.065) –0.286*** (0.048) 0.622** (0.273) 0.149*** (0.036) –0.060 (0.051)

0.366*** (0.097) –0.317*** (0.053) –0.460*** (0.157) 0.067 (0.078) 0.071 (0.072)

note: ***p < 0.01, **p < 0.05, *p < 0.1

5.2 Dynamic Simulations

An interesting way to assess our models is to plot the predicted changes in unemployment rates against the actual ones. The results of this exercise are depicted in Figures 6 to 8.

-4 -2 0 2 2005

1960 1975

1960

2005

2005

1960 1975

1990

2005

Sweden -2 0 2 4

-2-10 1 2

1975 1990

1960

2005

1975

1990

2005

Netherlands

-2-1 0 1 2 1975 1990

Norway

1960

1990

Germany

-20 2 4 6

Finland

-1 0 1 2 3

1975 1990

1960 1975

1990

1960

2005

1975

1990

2005

Switzerland

-.50 .5 11.5

1960

Denmark -4-2 0 2 4

Belgium

-1 0 1 2

Austria

1960

1975

1990

2005

year Actual change of unemployment rate

Dynamic simulation

Graphs by country

Figure 6: Actual and Simulated Changes of Unemployment Rates— “Continental Europe” Group

Crisis, What Crisis?

45

For a number of countries, the model is able to replicate changes in the unemployment rate quite well. Particularly for the Nordic countries, the model appears to produce roughly acceptable forecasts. For some other economies, the fit is rather poor. Germany, for instance, shows much more actual variation in the unemployment rate than the model predicts—this calls for a more detailed historical account. Concerning the latest figures, where our graphs plot OECD forecasts for 2010 and 2011 against predicted values, the model comes close to the actual and forecasted rates in Belgium, Finland, the Netherlands, and Norway for the first group. It also does a good job for Canada, Japan and the US for the second group.

4 2 1975 1990

2005

0 -2

1960 1975

1990

2005

1960

United Kingdom

1975

1990

2005

United States

1960

2 0

-2

-2

0

-1 -.5 0

2

.5

1

4

Japan

4

1960

-2

-1

0

0

1

2

2

6

Ireland

4

Canada

3

Australia

1975 1990

2005

1960 1975

1990

2005

1960

1975

1990

2005

year Actual change of unemployment rate

Dynamic simulation

Graphs by country

Figure 7: Actual and Simulated Changes of Unemployment Rates— “Anglo-Saxon” Group

Interestingly, our models do not predict increases in national unemployment rates due to the current crisis, beyond those experienced so far or forecasted to materialize in the near future. This finding also holds for the vast majority of peaks and troughs in the past. Up to a certain level, this feature is due to the fact that a panel is fitted to population means. However, since we model countryspecific shocks, which obviously differ between units of observation, there is at least the theoretical possibility—if countries are really similar in some sense— that our models come close to observed changes in unemployment. Denmark and the US are good examples that this could also hold in practice. The better fit for these two countries is likely due to the characteristics of the Danish and the American labor markets. Both countries have rather flexible labor markets—although different in nature—which react rather instantaneously to business cycle fluctuations.

46

Werner Eichhorst, Michael Feil, and Paul Marx

On the other hand, Germany and Switzerland, two economies that have grown at relatively small rates during the 1990s and early 2000s, do not follow the predicted slow movements in unemployment rates. This result points to other sources of variation which are not captured by our variables. In retrospect the German unemployment rate could be perceived as excessively high and volatile compared to other European labor markets. Except for changes in statistical definitions or related measures, such large deviations could also result from different economic structures or structural changes which are not captured in the macroeconomic data. Against the backdrop of the concentration of the current economic crisis on a few industries in several countries, this line of research could be taken up in future work. Admittedly, there are signs of misspecification in the sense of failure to describe the dynamic relationship between the independent and the dependent variables. This problem is even more exacerbated when we regressed changes in employment rates, instead of unemployment rates, on our set of explanatory variables. The modeling strategy pursued here, which is very similar to the one in Nickell et al. (2005), although including the lagged dependent variable and accounting for serially correlated error terms, is obviously not capable of explaining a significant part of the persistence mechanisms at work.

1960

1 0 -1 -2

-2

-1

0

1

2

Portugal

2

France

1970

1980

1990

2000

2010

1980

2000

2010

2000

2010

Spain

5

2 1

0

0

-5

-1 1960

1990

10

Italy

1970

1980

1990

2000

2010

1980

1990

year Actual change of unemployment rate

Dynamic simulation

Graphs by country

Figure 8: Actual and Simulated Changes of Unemployment Rates— “Southern Europe” Group

Crisis, What Crisis?

47

6. Case Studies Case studies can help deepen the analysis by going beyond standardized aggregate institutional indicators and taking into account the complexities of political and enterprise-level adjustment. We therefore give a more detailed account of four “ideal types” of European economies with largely different combinations of (a) active and passive labor market policies and (b) employment protection.

Labor market policies

Employment protection (core) Strong

Weak

Large

1 (Continental corporatist countries, DE)

3 (Flexicurity countries, DK)

Small

2 (Mediterranean countries, ES)

4 (Anglo-Saxon countries, UK)

Figure 9: Case Selection

6.1 Germany: Employment Stability despite Strong Export Orientation

Germany is a country which was heavily affected by the steep decline in international trade, which, in turn, led to a significant fall in orders and exports, particularly in core areas of the German production model, such as machinery and automobile manufacturing. However, despite its vulnerability due to the dependency on exports and the associated GDP decline of 5% in 2009, unemployment has basically remained stable, as has the total employment rate. Stable total employment is, on the one hand, due to the still dynamic growth in major parts of the service sector, which could offset limited losses in export-oriented sectors such as manufacturing and logistics. On the other hand, the relative success of the German story is explained by the fact that the core labor market of skilled workers in manufacturing is covered by strong legal dismissal protection—hence, short-term adjustment does not lead to quick layoffs but is dominated by an elaborate arrangement of internal flexibility (Möller 2010, Fuchs et al. 2010, OECD 2010). Three elements are crucial: 1. First, internal flexibility: this has already increased considerably over the past two decades, both within in the framework of collective bargaining and at the level of the enterprise. Hence, working time can be adjusted flexibly via working time accounts. In fact, surpluses in these accounts and overtime declined significantly: while employment was virtually unchanged from late 2008 to late

48

Werner Eichhorst, Michael Feil, and Paul Marx

2009, the total volume of hours worked declined by about 3 %. At the same time, the social partners were also allowed to adjust agreed wages or postpone wage increases in difficult times. 2. Second, the concentration of redundancies in the marginal workforce: over the past five years, manufacturing employers have increasingly relied on temporary agency staff to establish a flexible segment of the workforce which can be swiftly adjusted under uncertain economic prospects. Therefore, when the crisis began, employers started to reduce the number of agency workers active in their firms by about 300,000, so that employment decline could basically be limited to this category of workers. On a smaller scale, the same is true for the nonrenewal of fixed-term contracts. 3. Third, heavy reliance on a public short-time work allowance embodied in the institutional repertoire of the unemployment insurance and active labor market policy system. In 2009, 1.1 million workers (about 350,000 full-time equivalents) worked short-time. The short-time work or partial unemployment scheme was, of course, already in place before the crisis began, but it has been modified in three aspects: (i) the maximum duration for which hours not worked are reimbursed by the unemployment fund at the regular replace rate applicable in case of unemployment was increased from six to 24 months for inflow in 2009 (18 months in 2010); (ii) for cases of short-time work arising in 2009 and 2010, employers are exempt from social security contributions for hours not worked— regarding employee contributions from the first day of short-time work, regarding employer contributions from the seventh month of reduced working time (or earlier in case of employer-provided training); and (iii) administrative requirements for firms entering this scheme were simplified considerably. Given the abrupt character of the crisis and the uncertainty of its duration, employers have been reluctant to dismiss skilled staff as long as partial unemployment is feasible and a recovery is expected. The German fiscal stimulus package seems to have so far had a limited direct impact on the labor market outside expansion of partial unemployment schemes. Despite some measures to help stabilize consumer confidence, such as a marginal cut in income taxes and social security contributions, and a “cash for clunkers” scheme, the fiscal package has yet to show its major impact. Most recent figures on the development of GDP and exports show early signs of recovery. Hence, the extent of employment losses may be contained, and labor market performance in 2009 and 2010 could be better than in earlier forecasts. Hence, working time flexibility and complementary short-time work allowances could have helped bridge the slump in manufacturing without endangering the skilled core labor force. 6.2 Denmark

In the decade before the crisis, Denmark became a celebrated model country for the flexicurity approach. The combination of its traditionally flexible labor law,

Crisis, What Crisis?

49

generous unemployment benefits, and active labor market policies ensured a very smooth functioning of the employment model. Thanks to this “golden triangle,” the Danish labor market was characterized by high mobility and far above-average employment rates. Unemployment was virtually non-existent. However, the recent crisis had a comparatively strong impact on the economy and especially on the labor market. Given the excellent situation in the past, an unemployment rate of more than 6% meant a severe increase. While Denmark is still doing comparatively well, it is approaching the EU average concerning most labor market indicators. The straightforward interpretation of this trend is that the Danish employment system features fewer internal institutional buffers against an output shock. As one would expect, liberal dismissal regulation contributes to a large inflow into unemployment during an economic crisis. On the other hand, labor market recovery should be more dynamic once the economy returns to growth, as hiring barriers are lower and activation policies are in place. Furthermore, Danish policy makers have refrained from interfering with the economy by way of sectoral subsidization. Whether this institutional advantage of the Danish model will prove to be beneficial in the near future also depends on the second aspect of the flexicurity concept—social security and activation. The Danish welfare state provides high income replacement in the case of job loss. Even after five years of unemployment, net benefits of previous low-wage earners can amount to more than 80%. This is the most generous level in the OECD (average: 45%). To diminish disincentives to take up work, labor market reforms in previous years have strengthened the activation of the unemployed. While active labor market policies were sufficient to promote employment in normal economic times, it remains to be seen how the system will work against the background of severe unemployment. So far, Danish active labor market policies have been characterized by policy continuity—also with respect to activation strategies trying to increase labor supply. Yet, the crisis also revealed the macro-economic advantages of high level of social protection. It can be shown that unemployment insurance had a significant effect as an automatic stabilizer in those countries with sufficient levels of income protection (e.g., vis-à-vis the United States). This holds true for Denmark in particular. Together with Austria and Sweden, the income stabilization effect of benefits is the largest in the OECD (Dolls et al. 2009). In a similar fashion, the budget available for active measures is tied to the development of the unemployment rate. Therefore, the Danish regime adapts automatically and needs less discretionary intervention by policy makers. So although the crisis had a severe impact, the generous Danish welfare state helped prevent worse effects. This is an insight that should be reflected in the reform discussions of countries that traditionally refrain from implementing strong automatic stabilizers. On the other hand, these beneficial short-term effects of the

50

Werner Eichhorst, Michael Feil, and Paul Marx

welfare system imply risks in terms of labor market performance. The Danish experience has shown that the right mix of welfare state generosity and activation is crucial to reconcile both objectives. While the country has been very successful in the past, the current crisis could make further adjustments necessary.

6.3 Spain

Concerning the international economic crisis, the Spanish labor market is one of the hardest-hit in the European Union. In December 2009, unemployment rose to almost 20%—twice the euro zone average; within in the EU, only Latvia has comparable rates. The reason for this devastating labor market performance can be found in the features of the previous upswing as well as in the bifurcated employment model. Before the crisis, Spain had experienced a long-lasting boom period, with aboveaverage growth rates and strong job creation. However, both heavily relied on a housing bubble, which burst in the wake of the international crisis. Accordingly, employment losses were particularly severe in the oversized construction sector. In addition, Spain features relatively flexible job protection regulation at the margin of the labor market. As opposed to strictly regulated permanent contracts, temporary workers form a very flexible tier in the labor market. While this pattern of asymmetric flexibility is by now quite typical for highly regulated European labor markets, Spain in this regard is an extreme case. Before the crisis, approximately one-third of all employees worked on a fixed-term contract. Educational purposes or extended probationary periods, which play a large role in other countries, are almost non-existent. Around 85% of all employees in Spain working on a temporary contract do so involuntarily. Only Cyprus, Greece and Portugal exhibit similarly high values in the EU. In addition, approximately two-thirds hold a contract with a duration of only six months or less (twice the EU-15 average). It was mainly this segment of highly flexible workers who carried the burden of labor market adaptation in the recent recession. Between the beginning of 2007 and the end 2009, the share of temporary workers in total employment decreased from 32 to 26%. The total reduction amounted to more than 1.2 million workers. At the same time, employment among permanent workers was very stable: about 90% of the jobs lost were temporary ones. With a large fiscal stimulus package, the Spanish government contributed to keeping GDP decline moderate. However, policy measures designed to keep unemployment at bay were less successful. In an attempt to encourage German-style labor hoarding, social security payments were lowered for companies that reduced employees’ working hours instead of dismissing them. In addition, wage subsidies for workers on short hours and subsidies for companies hiring part-time workers were created. Hence, Spanish policy makers tried to stimulate policies in favor of working time and, therefore, internal flexibility, which is seen as better alternative

Crisis, What Crisis?

51

to excessive external flexibility stemming from fixed-term contracts. At the same time, however, policy makers also extended unemployment benefits. A possible reason for the failure of this program to prevent unemployment from rising continuously can be found in characteristics of the workers laid-off. The jobs that were created in the upswing preceding the crisis and now no longer exist are predominantly low-productivity jobs in construction; hence, there is no interaction with existing strategies of skill-oriented internal flexibility, as can be observed in German manufacturing. By and large, there are few incentives for Spanish employers to retain lowskilled and flexible workers (whose redundancies do not involve significant turnover costs for firms) during a phase of slump. And there is little ad-hoc policy measures can do about it. Hence, the seemingly crisis-specific problems point to more deeply rooted challenges for the Spanish employment model: to shift jobcreation to sectors with higher productivity and to overcome segmentation in the labor market. Concerning the latter, the lessons of the crisis do not seem to have increased chances for reform. While there has an expansion of unemployment benefits to counter social concerns, the liberalization of dismissal laws still ranks low on the agenda. 6.4 United Kingdom

The United Kingdom was also heavily hit by the current crisis—but in contrast to Spain and Germany, the British economy was not affected mainly due to an oversized construction sector or heavy reliance on manufacturing. The crisis rather hit the UK as Europe’s financial centre, which had benefited from a dynamic development of the banking sector and strong housing price growth for many years (Bell and Blanchflower 2009). This has been associated with a dynamic development of job creation in the private service sector. High exposure to global financial business and, at the same time, a rather flexible labor market and smaller welfare state—compared to most Continental European countries—imply high labor market vulnerability. Compared with, e.g., Germany, external flexibility is more dominant than internal adjustment of firms. In accordance with a pattern dominated by external flexibility, working time adjustment did not contribute significantly to the adjustment process following the recent shock. More flexibility could be observed with wages in the private sector. In good times, a less regulated labor market can contribute to strong employment growth; in bad times, this reduces the buffering effect of intra-organizational arrangements. At the same time, the tax / benefit system in the UK provides less automatic stabilization than in Continental European or Scandinavian countries. The increase of unemployment was stronger in the UK than it was in Germany, however, the UK performed better than Spain, Denmark, and the US. A more indepth analysis shows that, at least until recently, job losses in the private sector were compensated for by a stable and ever expanding public sector. Major job cuts

52

Werner Eichhorst, Michael Feil, and Paul Marx

could be observed in the vulnerable exposed sectors such as manufacturing, finance and construction as well as distributional services. Given the limited role of automatic stabilization and the size of the shock experienced by the British economy, it is interesting to see that the UK adopted a largerthan-average fiscal stimulus package, which helped mitigate the immediate labor market impact of the crisis and will also contribute to additional fiscal constraints in the future. At the same time, the UK is also one of the countries which implemented the most comprehensive crisis-related reforms in the realm of unemployment protection and active labor market policies. The existing institutional repertoire was expanded in order to cope with increasing unemployment and speed up reintegration into employment. In this perspective, discretionary policy making complemented active labor market instruments which were already in place at the outset of the current crisis, such as the JobCentre Plus and Rapid Response Services, to achieve a broader activation with more intensive support for all jobseekers out of work for more than six months, an activation guarantee for young people and a more general scheme implemented in fall 2009 (Flexible New Deal) replacing earlier, more selective ones. Furthermore, incentives for employers who recruit and train people unemployed for at least six months were introduced and extra funding given for training, not only for the unemployed but also for start-up support.

7. Conclusion Countries vary widely with respect to the labor market impact of the current global economic crisis. Macroeconomic shocks and policies, institutional factors and policy packages interact. While the GDP shock and the structure of the economy, i.e., the share of exposed and therefore vulnerable sectors, determine the problem pressure, fiscal packages, institutional arrangements, and labor market-related policies can help explain the actual impact the crisis has in a given national context. But classical (external) labor market institutions cannot explain everything; and this is certainly confirmed by our regression models. There seems to be, however, some support in the data that institutions matter. The importance of institutions differs between countries. There are also signs that they vary between different periods of time. Since our models are rather parsimonious, we consider our results encouraging. There are, however, some obvious shortcomings. Firstly, we have so far not been able to include all margins of flexibility. The variables we used for the econometric work can be linked to external numerical as well as wage flexibility. They do not contain information about other forms of flexibility. This is an obvious shortcoming. Due to the difficulties in gathering adequate data for a large set of countries, we remain skeptical whether this shortfall could be overcome in the near future. Another related issue is the question of universality. Are the effects of labor market institutions really the same across the OECD or some subsets of countries? Finally, measurement of institutions is certainly an issue, but this has long been discussed in the literature.

Crisis, What Crisis?

53

Economies which were able to rely on strong internal flexibility have hitherto been able to avoid a major increase of unemployment. Here, the core of the labor market is protected by relatively strict employment protection and benefits from working time adjustment that compensates for layoffs. This holds in particular for Germany, where working time accounts and complementary short-time work allowances helped stabilize the manufacturing sector. It comes as no surprise that many countries expanded or introduced short-time work schemes in order to stabilize existing jobs during a crisis perceived as a transitory one. Short-time work programs allow for a smooth short-term adjustment, but also mean a costly bet on a quick recovery. However, even in highly regulated labor markets with elaborate systems of internal flexibility, non-standard workers exhibit a much stronger risk of being made redundant during a crisis. Hence, the burden of adjustment in the current downturn is distributed rather unequally and tends to deepen dualization in the labor market. This is particularly apparent in the case of Spain, where the termination of temporary contracts is a major channel of adjustment. Non-standard workers not only face a higher risk of unemployment, but they also tend to be less well protected by social insurance. However, policy makers in some countries have taken the initiative to ease access to benefits for non-standard workers or to increase generosity. One lesson from the crisis could be to move towards more equal distributions of risks and appropriate social protection by making benefit systems more universal, so that non-standard workers are not disadvantaged twice: in terms of employment stability and in terms of social benefits. Welfare states not only provide income replacement for individuals, they also stabilize the whole economy. Automatic stabilizers, embodied in a progressive tax system and social benefit arrangements, have certainly contributed to the mitigation of the direct impact of the crisis on the economy. Social policies in place at the beginning of the crisis can set in immediately after the shock. Automatic stabilization policies are, therefore, much easier to implement than discretionary measures, which tend to create greater challenges in terms of appropriate implementation and often can only become effective after a certain time lag. Hence, it seems desirable for the future to have proper automatic stabilization policies (unemployment protection and active labor market policies) in place in order to limit the necessity for discretionary action. A review of policy initiatives during the crisis also shows that most discretionary changes in social and labor market policies were tied to existing programs and changed them in a parametric way (e.g., expansion of short-time work, more generous unemployment benefits or hiring subsidies, and more training or job search assistance) rather than introducing completely new elements. Keeping unemployment low by means of internal flexibility and complementary public support, as well as automatic or discretionary policy interventions, may be a good thing in the short run. At some point, however, a cyclical crisis may turn into a structural one which requires a different policy approach: in particular, policies to cope with structural change. While many policy instruments implemented in the current situation were designed to bridge a temporary loss of demand based on the

54

Werner Eichhorst, Michael Feil, and Paul Marx

assumption that we were facing a short, but harsh, business slump, a decline of employment in manufacturing, construction or the financial sector may become inevitable in many countries if productive employment in these sectors is not viable at the pre-crisis level in the foreseeable future. If structural change is accelerated the longer the crisis lasts, well-designed policies should refrain from trying to preserve non-sustainable jobs and postpone structural change by means of sectoral subsidies. This would rather call for a timely exit from crisis-related subsidies and stabilization measures and for stronger policy action to stimulate economic innovation and mobility in the labor market. External flexibility and job reallocation gain in importance. Encouraging internal functional flexibility through lifelong learning and workplace-related training is one thing. Mobility from shrinking or declining firms and sectors to other more dynamic sectors and employers is a second major challenge, as productive employment is essential not only for individual incomes but also for the public budget and the sustainability of social policies. Welfare states and social benefits, while helpful in providing income support and buffering the immediate shock, could create work disincentives and run the risk of persistent unemployment or inactivity. They have to be complemented by active labor market policies and activation approaches encouraging quick and sustainable reemployment in the early stage of unemployment. The country studies have so far shown that there is no major rollback of activation policies; rather, more emphasis is put on active labor market policies, job search assistance and training, as well as on hiring subsidies and start-up grants. In this respect, countries can benefit from earlier reforms directed towards activation and an effective delivery of labor market policies and public employment services. References Amable, B. (2003): The Diversity of Modern Capitalism. Oxford, Oxford University Press. Atkinson, J. (1985): Flexibility, Uncertainty and Manpower Management: IMS Report 89, Brighton, Institute of Manpower Studies. Baccaro, L. / Rei, D. (2007): “Institutional determinants of unemployment in OECD countries: Does the deregulatory view hold water?” International Organization 61 (3), 27 – 569. Baker, D. / Glyn, A. / Howell, D. R. / Schmitt, J. (2003): “Labor market institutions and unemployment: A critical assessment of the cross-country evidence,” Oxford University Department of Economics Discussion Paper 168. – (2005): “Labor market institutions and unemployment: A critical assessment of the crosscountry evidence,” in Fighting Unemployment: The Limits of Free Market Orthodoxy, edited by D. R. Howell, Oxford, Oxford University Press, 72 – 118. Baltagi, B. (1995): Econometric Analysis of Panel Data, Chichester, John Wiley and Sons. Bassanini, A. P. / Duval, R. (2006): “Employment patterns in OECD countries: Reassessing the role of policies and institutions,” OECD Economics Department Working Paper 486. – (2009): “Unemployment, institutions, and reform complementarities: Re-assessing the aggregate evidence for OECD countries,” Oxford Review of Economic Policy 25 (1), 40 – 59.

Crisis, What Crisis?

55

Beck, N. / Katz, J. N. (1995): “What to do (and not to do) with time-series cross-section data,” American Political Science Review 89: 634 – 647. Bell, D. N. F. / Blanchflower, D. G. (2009): “What should be done about rising unemployment in the UK,” IZA Discussion Paper 4040. Belot, M. / van Ours, J. C. (2001): “Unemployment and labor market institutions: An empirical analysis,” Journal of the Japanese and International Economies 15 (4), 403 – 418. Blanchard, O. (2006): “European unemployment: The evolution of facts and ideas,” Economic Policy 21 (45), 5 – 59. Blanchard, O. / Wolfers, J. (2000): “The role of shocks and institutions in the rise of European unemployment: The aggregate evidence,” The Economic Journal 110 (462), 1 – 33. Daveri, F. / Tabellini, G. E. (2000): “Unemployment, growth and taxation in industrial countries,” Economic Policy 15 (30), 47 – 104. Dolls, M. / Fuest, C. / Peichl, A. (2009): “Automatic stabilizers and economic crisis: US vs. Europe,” IZA Discussion Paper 4310. Eichhorst, W. / Feil, M. / Braun, C. (2008): “What have we learned? Assessing labor market institutions and indicators,” IZA Discussion Paper 3470. Eichhorst, W. / Marx, P. / Tobsch, V. (2009): “Institutional arrangements, employment performance and the quality of work,” IZA Discussion Paper 4595. Freeman, R. B. (2000): “Single peaked vs. diversified capitalism: The relation between economic institutions and outcomes,” NBER Working Paper 7556. Fuchs, J. / Hummel, M. / Klinger, S. / Spitznagel, E. / Wanger, S. / Zika, G. (2010): “Entwicklung des Arbeitsmarktes 2010: Die Spuren der Krise sind noch länger sichtbar”, IAB Kurzbericht 3 / 2010. Hall, P. A. / Soskice, D. (2001): “An introduction to varieties of capitalism,” in Varieties of Capitalism: The Institutional Foundations of Comparative Advantage, edited by P. A. Hall and D. Soskice, Oxford, Oxford University Press, 1 – 68. Hasselpflug, S. (2005): “Availability criteria in 25 countries,” Ministry of Finance Denmark Working Paper 12 / 2005. Howell, D. R. / Baker, D. / Glyn, A. / Schmitt, J. (2007): “Are protective labor market institutions at the root of unemployment? A critical review of the evidence,” Capitalism and Society 2 (1), 1 – 71. Layard, R. / Nickell, S. / Jackman, R. (1991): Unemployment: Macroeconomic Performance and the Labour Market, Oxford, Oxford University Press. – (2005): Unemployment: Macroeconomic Performance and the Labour Market, 2nd edition with new introduction, Oxford, Oxford University Press. Möller, J. (2010): “The German labor market response in the world recession—de-mystifying a miracle,” Zeitschrift für Arbeitsmarktforschung 42 (4), 325 – 336. Nickell, S. (1997): “Unemployment and labor market rigidities: Europe versus North America,” Journal of Economic Perspectives 11 (3), 55 – 74. Nickell, S. / Nunziata, L. / Ochel, W. (2005): “Unemployment in the OECD since the 1960s. What do we know?” The Economic Journal 115 (500), 1 – 27.

56

Werner Eichhorst, Michael Feil, and Paul Marx

OECD (2006): OECD Employment Outlook 2006, Paris. – (2008): OECD Education at a Glance 2008: OECD Indicators, Paris. – 2009a): OECD Employment Outlook 2009, Paris. – (2009b): OECD Education at a Glance 2009: OECD Indicators, Paris. – (2009c): OECD Benefits and Wages 2009: OECD Indicators, Paris. – (2009d): OECD Taxing Wages 2009: OECD Indicators, Paris. – (2010): OECD Economic Surveys Germany, Paris. Venn, D. (2009): “Legislation, collective bargaining and enforcement: Updating the OECD employment protection indicators,” OECD Working Paper 89. Visser, J. (2009): Database on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts in 34 Countries between 1960 and 2007. Amsterdam Institute for Advanced Labour Studies (AIAS).

Annex 1 Channels of Flexibility and Associated Indicators SubIndicators

Institution

Variables

Source (2003) Source (2008)

Empl. Protection

Summary Indicator of EPL Stringency

OECD

Venn (2009)

Net Replacement Rates (1st and 60th month) Strictness of Available Criteria

OECD

OECD (2009c)

Benefit System

Hasselpflug (2005)

Hasselpflug (2005)

Taxation

Tax Wedge

OECD

OECD (2009d)

ALMP

ALMP Spending

OECD

OECD.Stat

Education

Educational Attainment Expenditure per Student

OECD OECD

OECD (2009b) OECD (2009b)

Wage

Wagesetting

Coll. Bargaining Coverage Rate Coll. Bargaining Co-Ordination Wage-Dispersion

OECD OECD OECD

Visser (2009) Visser (2009) OECD (2009a)

InternalNumerical

Working Time

Dominant Response to Workload Rationale for Using Flexible Working Hours Working Time Accounts

ESWT ESWT

– –

ESWT

–

EWCS

–

Internal Qualification

Incidence of Training Provided by Employer Volume of Non-Formal JobRelated Training

Work Organization

Incidence of Task Rotation Necessity of Learning New Things

EWCS EWCS

ExternalNumerical

ExternalFunctional

InternalFunctional

OECD (2008) – – –

Structure of the Composite Indicator with Scores for Sub-indicators (Circled Numbers)

Annex 2

Crisis, What Crisis? 57

Crisis, What Crisis? Patterns of Adaptation in European Labor Markets Comment By Antje Mertens1 Introduction In the middle of the most severe economic downturn since the Second World War, Eichhorst, Feil and Marx deal with the question of how labor market institutions influence unemployment dynamics in a crisis. The main questions they ask are: Why did unemployment increase dramatically in some countries and less severely in others? What labor market institutions are driving the outcomes? Which aspects of flexibility are most important? These are important questions to ask, especially since structural unemployment seems to have declined lately as labor market conditions have changed and unemployment mobility has increased in Europe (Boeri and Garibaldi 2009). To answer these questions, Eichhorst, Feil, and Marx use cross-sectional and longitudinal data for a large set of OECD countries and apply a variety of approaches: (i) flexibility indicators, (ii) quantitative time series analysis, and (iii) case studies. Although at first glance, the diversity of methods may be somewhat confusing, the paper carefully collects and discusses available information from different angles. At the same time, the results point to the limitations of all three approaches, which will be discussed below. Their main findings can be summarized as follows: 1. Defining groups of countries according to institutional patterns does not seem meaningful from a short-term perspective. Even if long-term flexibility patterns coincide, short-term adjustment paths differ significantly. 2. Quantitative analyses in the tradition of Nickell (1997) and Nickell et al. (2005) help to predict unemployment dynamics in some countries—notably the Nordic countries—but the fit is by no means close for all countries, and particularly not for Germany. 3. Finally, internal flexibility—in particular working time adjustments—accounts for some of the variation. This is potentially the strongest argument.

1 Berlin School of Economics and Law, Faculty 2, Neue Nahnhofstr. 11 – 17, 10245 Berlin, e-mail: [email protected].

60

Antje Mertens

1. Flexibility Indicators and External Functional Flexibility The paper starts by constructing aggregate flexibility indicators as developed by Eichhorst, Marx and Tobsch (2009). These indicators allow the sorting of countries according to flexibility categories, but also facilitate the clustering of countries into groups. In the process, information is condensed and on the one hand made more accessible. On the other hand, the aggregation of information potentially masks important differences between and within countries. However, if indicators are unable to describe institutions and their changes perfectly, especially short-term variation in cross-country comparison, results become fuzzy and we have to be very careful in their interpretation. The major categories of flexibility incorporated here are external numerical, external functional, and wage flexibility. Of these indicators, external functional flexibility is especially hard to specify. The basic idea is to measure how easy it is for firms to exchange qualifications no longer needed with new qualifications via the labor market. Leschke, Schmid, and Griga (2006) as well as Seifert and Tangian (2006) use the term external functional flexibility for temporary work and other forms of outsourcing like Transfergesellschaften2. Eichhorst, Feil, and Marx measure external functional flexibility with spending on active labor market policies (ALMP) and education. Although these indicators are potentially part of external functional flexibility, we would need to include the efficiency of ALMP, potentially identified by Meta studies (Card at al. 2009, Kluve 2006, Steiner and Caliendo 2005). Getting even closer to measuring external functional flexibility, i.e., the possibility to substitute “old qualifications” for “new qualifications,” one would need to know how mobile workers in the different countries really are, as labor market flows out of unemployment and from job to- job show considerable inter-country differences. Moreover, the typical type of training—for instance, the dual system in Germany or school-based vocational education in Sweden—potentially influences external functional flexibility (Korpi and Mertens 2003).

2. Quantitative Analysis Considering additional information along the lines just described would almost certainly improve the clustering of countries performed on the basis of the flexibility indicators. The cluster analysis is used to sort countries and reduce heterogeneity in the quantitative analysis. However, clusters are still relatively heterogeneous, as the following example shows: Germany, Sweden, and Denmark are in the same cluster, although flexibility indicators differ markedly ranging from 39.9 for Germany to 43.7 for Sweden and all the way to 48.8 for Denmark. It is especially 2 Transfergesellschaften are interim employment associations created by downsized firms together with government employment offices. Their goal is to facilitate job search for workers who are no longer needed.

Crisis, What Crisis?—Comment

61

surprising to find Denmark (with its high external flexibility) and Germany (which is known to have relatively low external flexibility) in one cluster. Lowering the cutoff for forming clusters would reduce heterogeneity further, which in turn could help to make the quantitative analysis more robust and convincing. The quantitative analysis by Eichorst, Feil, and Marx is in the tradition of Blanchard and Wolfers (2000) and rests very much on the paper by Nickell et al. (2005). It primarily extends the period under analysis but estimates separate models for the clusters identified instead of models for the pooled sample like the original paper does. As major problems of these types of models have been intensively discussed in the literature before, the reader is referred to Freeman (2005) and Baccaro and Rei (2007) for further reference. Two points are noteworthy here. First, the institutional variables used (benefit reduction rate, trade union density, employment protection, bargaining coordination and the tax wedge) are reduced by factor analysis, so only one single institutional coefficient is estimated for each cluster. This intensifies the problem already discussed above that aggregate institutional indicators will never be able to perfectly map country differences in institutions. Even though some significant coefficients are found, it is impossible to disentangle the importance of the different institutions. Unfortunately, results even are “. . . at odds with the usual intuition . . .” as the authors state themselves. Having said this, the paper is still valuable as it offers an empirical means of clustering countries according to their labor market flexibility patterns and as it re-tests an empirical method that has received both praise and critique. Moreover, looking at actual and simulated unemployment dynamics, the models show remarkably good predictions for some countries. Then again, some predictions are very imprecise, especially for Germany. This rightly led the authors to examine some countries through case studies. 3. Detailed Country Analysis The idea that single institutions, often combined with certain sectoral structures, may explain major aspects of unemployment dynamics is dealt with in the last section of the paper. For the present discussion, let us focus on Spain and Germany, two countries with relatively high employment protection for regular employees. In Spain, the combination of a bursting housing bubble and high numbers of fixedterm employees led to dramatically rising unemployment rates affecting as much as 20 % of the labor force. In fact, “. . . temporary employment accounted for close to 90% of all job losses in the 12 months to June 2009” (OECD 2009). In Germany, on the other hand, where the manufacturing sector was hit especially severely, average unemployment increased little. Firms have been hording labor because of flexible work arrangements such as working time accounts and because they suffered from a shortage of qualified professional workers before the crisis (Möller 2010). Moreover, massive short-time work schemes helped to dampen unemployment response.

62

Antje Mertens

While we know that Spain’s unemployment rate soared partly due to the high number of fixed-term contract workers, it is not well understood why fixed-term contracts have increased so much more in Spain since the mid-1980s than in Germany. This is important to understand in order to explain how unemployment has responded. Both countries are always classified as having relatively high employment protection (see OECD, 1999, Tables 2.5 and 2.6 for an overview), even though some studies place both countries in a different rank order. Moreover, Germany and Spain are being ranked very closely in terms of employment protection for regular employees (OECD 2004), with an index of 2.7 for Germany and 2.6 for Spain. Still, the use of fixed-term contracts has been and is much higher in Spain than in Germany, with Spain at roughly one-third compared to less than 10 % in Germany (OECD 2010 and Statistisches Bundesamt 2010). Part of the explanation may be differences in severance payments between the two countries. While there are no official figures on the incidence and size of severance payments in Germany, evidence from micro-data suggests that between 1990 and 2002, roughly 30% of all West German employees who were dismissed actually received severance pay (Goerke and Pannenberg 2005). Average severance payments in (West) Germany vary substantially: assuming average monthly earnings of A 2000 in West Germany, they amount to between 5 and 17 months’ pay. Grund (2006) reports that between 1991 and 2004, displaced workers received severance payments of 8010,000 on average. In contrast to Germany, severance payments in Spain are mandatory and regulated by law. Still, 22% of all workers take the case to court. If the court decides that the dismissal was unfair—which happens in around 72% of cases—then not only does severance pay increase, but foregone wages have to be paid (Galdón-Sánchez and Güell 2000). In Germany, only 11 – 15% of all cases go to court and in 60% of these cases, employers lose the case in the court of first instance and 52 % in the court of second instance (Höland, Kahl, and Zeibig 2005). Whether and how mandatory severance payments and higher probabilities of losing the case determine the levels of fixed-term employment remains unresolved, but this example shows how aggregate indicators mask important country differences. Basically all of the factors mentioned above—subtle differences in employment protection, flexible work arrangements, industrial specificities, and even short-time work programs cannot be appropriately controlled for in the macro analysis. Since these factors are driving unemployment dynamics in the current crisis to a large extent, it is unsurprising that the quantitative macro analysis does not fit well. Still, the results from Eichhorst, Feil, and Marx show that we need to understand the influence of institutions better at the micro and country level. More microeconomic panel studies as well as case studies are needed on these issues.

Crisis, What Crisis?—Comment

63

References Baccaro, L. / Rei, D. (2007): “Institutional Determinants of Unemployment in OECD Countries: Does the Deregulatory View Hold Water?”, International Organization 61(3), 527 – 569. Blanchard, O. / Wolfers, J. (2000): “The Role of Shocks and Institutions in the Rise of European Unemployment: the Aggregate Evidence,” The Economic Journal 110(462), 1 – 33. Boeri, T. / Garibaldi, P. (2009): “Beyond Eurosclerosis,” Economic Policy 59 (July), 409 – 462. Card, D. / Kluve, J. / Weber, A. (2009): “Active Labor Market Policy Evaluations: A MetaAnalysis,” IZA Discussion Paper 4002. Eichhorst, W. / Marx, P. / Tobsch, V. (2009): “Institutional Arrangements, Employment Performance and the Quality of Work,” IZA Discussion Paper No. 4595. Freeman, R. B. (2005), “Labour Market Institutions Without Blinders: The Debate over Flexibility and Labour Market Performance,” NBER Working Paper 11286. Galdón-Sánchez, J. E. / Güell, M. (2000): “Let’s Go to Court! Firing Costs and Dismissal Conflicts,” Princeton University Working Paper 444, August 2000. Goerke, L. / Pannenberg, M. (2005): “Berufliche Weiterbildung ‘on-the-job’ und Auflösung von Beschäftigungsverhältnissen,” in L. Bellmann and D. Sadowski (eds.), Bildungsökonomische Analysen mit Mikrodaten, (Beiträge zur Arbeitsmarkt- und Berufsforschung, 295), Nürnberg, 151 – 179. Grund, Ch. (2006): “Abfindungszahlungen an entlassene Arbeitnehmer—eine empirische Untersuchung für Deutschland in den Jahren 1991 bis 2004,” in H. S. Buscher, C. Dreger, and U. Walwei (eds.), Institutionen und Arbeitsmarkt. Messkonzepte, Wirkungszusammenhänge und internationaler Vergleich, (Beiträge zur Arbeitsmarkt- und Berufsforschung, 304), Nürnberg, 117 – 129. Höland, A. / Kahl, U. / Zeibig, N. (2005): “Wirklichkeit und Wahrnehmung des Kündigungsschutzes in den Arbeitsgerichten,” in WSI Mitteilungen10 / 2005, 561 – 567. Kluve, J. (2006). “The Effectiveness of European Active Labor Market Policy,” IZA Discussion Paper 2018. Korpi, T. / Mertens, A. (2003): “Training Systems and Labor Mobility—A Comparison between Germany and Sweden,” Scandinavian Journal of Economics 105(4), 597 – 617. Leschke, J. / Schmid, G. / Griga, D. (2006): “On the Marriage of Flexibility and Security: Lessons from the Hartz-reforms in Germany,” WZB Discussion Paper April 2006 – 118. Möller, J. (2010): “The German Labor Market Response in the World Recession—De-mystifying a Miracle,” Zeitschrift für Arbeitsmarktforschung 42, 325 – 336. Nickell, S. (1997): “Unemployment and Labor Market Rigidities: Europe versus North America,” Journal of Economic Perspectives 11(3), 55 – 74. Nickell, S. / Nunziata, L. / Ochel, W. (2005): “Unemployment in the OECD Since the 1960s. What Do We Know?,” The Economic Journal, 115 (500), 1 – 27. OECD (1999): Employment Outlook, Paris: OECD. – (2004): Employment Outlook. Paris: OECD.

64

Antje Mertens

– (2009): Employment Outlook. Paris: OECD. – (2010): Labour Force Statistics, http:// stats.oecd.org/Index.aspx?DatasetCode=ALFS_ SUMTAB, accessed May, 19, 2010. Seifert, H. / Tangian, A. (2006): “Globalization and deregulation: Does flexicurity protect atypically employed?” WSI-Diskussionspapier 143, March 2006. Statistisches Bundesamt (2010): “Befristete Beschäftigung: Jeder elfte Vertrag hat ein Verfallsdatum,” http://www.destatis.de/jetspeed/portal/cms/Sites/destatis/Internet/DE/Content/ Publikationen/STATmagazin/Arbeitsmarkt/2010_032/2010_03Beschaeftigung,templateId= renderPrint.psml, accessed May, 19, 2010. Steiner, V. / Caliendo, M. (2005), “Aktive Arbeitsmarktpolitik in Deutschland—Bestandsaufnahme und Bewertung der mikroökonomischen Evaluationsergebnisse,” Zeitschrift für Arbeitsmarkt- und Berufsforschung 38 (2 / 3), 396 – 418.

The Role of Labor Market Institutions in the Great Recession By Jens Boysen-Hogrefe,* Dominik Groll,** Wolfgang Lechthaler,*** and Christian Merkl**** Abstract The recent Great Recession had very heterogeneous effects on the labor market outcomes in industrialized countries. We analyze the role of three labor market institutions in this context, namely the level of firing costs, the existence of short-time work and the wage formation process. This paper combines two different perspectives, a structural dynamic model perspective and an empirical cross-country perspective. Using the Lechthaler, Merkl, and Snower (2010) model, we first simulate the effects of the three labor market institutions during a recession. Using the panel of the EU-15 countries without Luxembourg, we then test the predictions of the model. Indeed, we find evidence that the three labor market institutions can partially explain the different labor market reactions across countries during the Great Recession. However, further empirical research is needed, as more data can be expected to become available, especially with respect to the use of short-time work in different countries. Keywords: employment, firing costs, short-time work, unit labor costs JEL Classification: E24, E32, J64

1. Introduction In 2009, the world economy faced the most severe recession since the Great Depression 80 years before. The depth and consequences of the so called Great Recession were quite heterogeneous across countries. Labor market institutions certainly played an important role for the actual economic outcomes in 2009. This paper combines two different perspectives to shed more light on the role of labor market institutions, namely a structural dynamic model perspective and an empiri* Kiel Institute for the World Economy. ** Corresponding author. Kiel Institute for the World Economy, Hindenburgufer 66, 24105 Kiel, Germany. E-mail: [email protected] *** Kiel Institute for the World Economy. **** Kiel Institute for the World Economy, University of Kiel, IZA. The authors thank the participants of the Macro Reading Group of the Kiel Institute and the participants of the annual ARGE conference on April 29, 2010 in Berlin, especially Claus Schnabel, for valuable comments. Any remaining errors are our own.

66

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

cal cross-country perspective. We combine both approaches, as either of them faces virtues and challenges in light of the current economic crisis. The former approach derives the results from the dynamic optimization of microeconomic agents. Therefore, changes in the behavior of economic agents, which are particularly severe in a great recession, can be expected to be taken into account. However, the dynamic model may be misspecified or unsuitable to quantify actual effects. Therefore, the latter approach allows us to obtain a better feeling for actual quantitative effects. However, the correlations of the empirical exercise cannot necessarily be interpreted in causal manner, particularly given the scarcity of data points during the Great Recession. This brings us back to the virtues of the theoretical perspective. Interestingly, our two employed approaches deliver similar results for the role of labor market institutions for output and employment outcomes. We consider three different labor market institutions: First, we analyze the effects of higher firing costs. In line with the theoretical predictions, our cross-country estimations show that industrial employment has fallen by more in those countries with lower firing costs. Thus, firing costs may act as a short-run shock absorber for the industrial sector. However, this shock absorber is associated with two types of costs. The empirical analysis shows that it creates a duality in the labor market. While the aggregate employment reductions may be reduced by higher firing costs, workers in the temporary employment segment are hit more severely in times of economic crisis, i.e., they have to bear the major costs of the crisis. Further, our theoretical model predicts a slower recovery in those countries with higher firing costs. Second, we analyze the effects of short-time work (Kurzarbeit) on output and employment. Our theoretical model predicts a more severe recession in terms of GDP in countries which make stronger use of this instrument. But at the same time, short-time work is a suitable instrument to stabilize employment in light of a negative aggregate shock. We compare different countries on a case study basis and find evidence that is in line with our model predictions. Third, we analyze the effects of wage formation (which may be influenced by reforms of the unemployment benefit system or by changes in the bargaining process) on employment. The theoretical model predicts a move to higher equilibrium employment when workers’ wage claims are reduced (e.g., under a less generous unemployment benefit system). If a negative aggregate shock occurs, while the economy is still in the transition to the higher equilibrium employment level, the adverse effect on employment is mitigated. The model prediction is backed by the fact that those countries with the lowest increase in unit labor costs during the years before the crisis performed best in terms of employment during the Great Recession. The rest of the paper is structured as follows. Section 2 discusses issues of methodology. Section 3 briefly illustrates the mechanisms of the underlying dynamic

The Role of Labor Market Institutions in the Great Recession

67

model. Section 4 shows dynamic model simulations and cross-country regressions for the three different labor market institutions. And Section 5 concludes.

2. Methodology An analysis of the role of labor market institutions in the current crisis is particularly challenging both from an empirical and a theoretical perspective. A pure data driven approach faces several shortcomings. First, the crisis is far from being over. Thus, analyzing the current data has the caveat that we are unable to know whether the worst is already over or whether some countries will still see the worst to come. Second, real time output data is known to be subject to severe revisions. This is already true in usual times (see, e.g., Orphanides and van Norden 2002). But we can expect this to be even more the case in windy times. Third, as the time series are very short, it is all the more dangerous to interpret correlations in a causal manner. Fourth, cross-country analyses generally suffer from lack of observations. Structural theoretical modelers may infer that this gives reason enough for not touching the data at all and for purely relying on shock and policy exercises in microfounded dynamic optimizing frameworks. But such an approach also faces several challenges. First, in order to analyze the role of labor market institutions in the current crisis, a rich enough model framework has to be identified. The workhorse medium-scale business cycle frameworks all assume a representative agent environment with frictionless labor markets (see Smets and Wouters 2003 for a European example). If modern business cycle frameworks contain a labor market, it is very often driven by a pure matching function approach (e.g., Faia 2009), giving only a limited role to labor market institutions. Second, even with an appropriate framework, which represents the data correlations in normal times fairly well, it is unclear whether this is also true in times of crisis or whether the underlying structure of the model has changed. Third, as the crisis is very recent, the time series are too short to fit a model exclusively to this data. Bearing all shortcomings in mind, we will combine the theoretical and empirical approaches, serving as a potential cross-check. Theoretically, we use the model framework by Lechthaler, Merkl and Snower (2010, LMS henceforth), which combines a standard New Keynesian model with a frictional labor market, with heterogeneous idiosyncratic productivity and linear hiring and firing costs. LMS show that this framework is able to replicate several important business cycle facts, such as strong amplification effects (i.e., strong reaction of labor market variables to macroeconomic shocks), persistence and appropriate correlations between aggregate variables. However, ex ante it is unclear how this model framework performs in light of the recent economic turmoil. We will start our analysis by using the predictions of this model framework on different labor market institutions, which have a clear-cut causal interpretation. Afterwards, we will perform an empirical cross-country analysis to check whether the correlations are in line with this interpretation. If this is the

68

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

case, we feel on fairly safe grounds to draw conclusions on the role of the respective labor market institution on macroeconomic outcomes. We will analyze the role of firing costs, short-time work and the wage formation process. The empirical cross-country analysis considers data from several European countries, namely the EU-15 without Luxembourg. We assume that these countries are sufficiently homogeneous with respect to structural conditions so that simple correlations are meaningful without the consideration of additional control variables. Expanding the sample to include all European countries or even countries beyond Europe would have the disadvantage that the heterogeneity across countries would increase by more than what could be controlled for by including additional control variables regarding the still very small sample size. Data is taken from EUROSTAT and OECD (see Appendix).

3. Model Economy 3.1 General Structure

The theoretical model we use is a variant of the framework introduced in LMS. It embeds microfounded labor market frictions into an otherwise standard dynamic stochastic general equilibrium model with sticky prices. The model structure is best illustrated by the following figure.

Figure 1: Stylized Model Structure

The Role of Labor Market Institutions in the Great Recession

69

Households provide labor to a sector producing intermediate goods. The employment decisions of intermediate goods producers are subject to linear hiring and firing costs and will be described in more detail further below. They sell their output to the wholesale sector under perfect competition. The firms of the wholesale sector use the intermediate goods as input to produce slightly differentiated consumption goods. The fact that these products are slightly differentiated gives them the power to set the price (monopolistic competition). It is assumed that they can change the price of their product each period but they face quadratic price adjustment costs. Finally, the retail sector bundles the differentiated consumption goods to one aggregate good and sells it to the consumers. Monetary policy is described by a Taylorrule which sets the interest rate at which households can save and borrow. As is standard, the Taylor-rule reacts to the output gap and inflation. Thus, apart from the labor market, we use a standard model as it can be found, e.g., in Gali (2008) or Woodford (2003). 3.2 The Labor Market

Since the labor market is the main innovation in the LMS model and since this is the part that is particularly relevant for our analysis, we will describe it in more detail, but restrict ourselves to the main intuition.1 To cover the structure of continental European labor markets, which are dominated by unions, we assume that wages are bargained collectively between the median worker and the firm. Wage negotiations are described by the Nash-product, which depends on the respective bargaining powers of firms and workers.2 To endogenize hiring and firing decisions, we assume that each worker is hit each period by a match-specific idiosyncratic shock. Thus, the output of a worker is given by y ˆ a ", where a is aggregate productivity and " is the idiosyncratic component, which is perfectly observable by both, worker and firm. Due to the match-specific component of output, productivity of workers can vary a great deal: Some workers will be very productive, generating high profits, while other workers generate losses for their employers, since the value of their output is lower than the wage. However, in assessing the value of a worker the firm does not only take account of present profits / losses but also of expected future profits. Workers are protected by firing costs and thus the firm only fires workers whose value is lower than the negative firing cost. A similar optimization determines the hiring decision of the firm: Only unemployed workers, which are sufficiently productive to generate profits larger than the hiring cost will be employed. Less productive workers will remain unemployed. 1 For technical details and more intuition see LMS. To make the model more realistic, we also included the tax and transfer system in our model. For details see Faia et al. (2010). 2 Although real wage rigidities may be present in European countries, they are not included in the model for simplicity. Including real wage rigidities would affect the results only quantitatively, not qualitatively.

70

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

One important implication of hiring and firing costs is that they drive a wedge between the retention rate (the probability that a worker stays in the firm) and the hiring rate. In a perfectly competitive labor market the two rates are the same, implying that the employment history of a worker is irrelevant. It doesn’t matter whether a worker had a job last period or if she was unemployed, the probability of having a job in the current period is the same. With hiring and / or firing costs3 the situation is different. A worker who had a job last period has a higher probability of having a job in the current period (which is obviously the case in real life). This result has important implications for the dynamic properties of the model. Employment (not only for the single worker but also for the whole economy) becomes path-dependent, i.e., employment this period depends on employment last period. As a consequence, the economy reacts much more sluggishly to changes in economic conditions and thus output becomes much more persistent, a feature that is well in line with empirical data.

4. The Effects of Different Labor Market Institutions 4.1 Firing Costs

To ensure that the model predictions are comparable to the data, we use the standard parameterization for a typical continental European economy, as in Faia et al. (2009). We keep all parameters constant, except for the firing costs, which are changed to 50 and 70 percent of the worker’s productivity (instead of 60 percent in the baseline scenario). Figure 2 shows the reaction of employment to a temporary negative aggregate productivity shock under different levels of firing costs. It is clearly visible from the figure that firing costs may act as a shock absorber, i.e., the economy slides into a more severe recession with lower firing costs. The reason lies in the fact that firing costs make employment adjustments (and thereby output adjustments) more costly. During an economic downturn some workers become unproductive and get fired by their employers. The higher firing costs are, the more reluctant employers will be to fire the workers. Thus, the decrease in employment is dampened. One important feature of this mechanism is that the effect works in both directions. Employment fluctuations are not only dampened during a downturn but also during a boom: With higher firing costs, firms are more reluctant to hire new workers, because they know it will be more costly to reduce the workforce to its old level once the boom is over. Thus, firing costs lead to smaller increases in output and employment during an economic boom. Further, it is noteworthy that economies with larger firing costs return less quickly to the steady state. Thus, firing costs act as a shock-absorber in the im3

In fact, only one of the two is necessary to generate the effect.

The Role of Labor Market Institutions in the Great Recession

71

mediate aftermath of the shock. However, they also delay the recovery for exactly the same reasons. EMPLOYMENT PATH (% DEVIATION) 0 fc = 0.5 fc = 0.6 fc = 0.7

-0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9

0

5

10

15 Quarters

20

25

30

Figure 2: Reaction to a Negative Productivity Shock under Different Firing Costs (FC)

Faia et al. (2009) and Merkl and Schmitz (2010) show that firing costs do not only reduce employment fluctuations for supply shocks, but also for realistic combinations of aggregate supply and demand shocks. Further, Merkl and Schmitz (2010) show that the output volatilities within the Euro area were in line with the model predictions before the Great Recession. Abbritti and Weber (2009) draw similar conclusions for employment volatility, based on a panel VAR for an OECD sample. Thus, there is evidence that the model predictions are in line with the data in usual times. At first sight, the development during the Great Recession does not seem to further confirm the relationship between employment and firing costs. As a proxy for firing costs, an often used measure is the OECD’s employment protection legislation (EPL) indicator. According to the Deutsche Bundesbank (2009), it is not possible to observe a close link between the EPL index and the reaction of employment during the crisis across countries. Similarly, Möller (2010) does not find a significant effect of the EPL index on the change of the unemployment rate during the recession. However, when looking at the evolution of the number of employees (i.e., excluding the self-employed) in the total industry sector, a clearer picture emerges. Figure 3 depicts the values for the EPL index for regular jobs and the change in the number of total industry employees between 2008Q1 and 2009Q1.

72

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

A negative relationship can be detected. Countries with higher firing costs experienced a less dramatic drop in total industry employment, as predicted by the model.4

Change in number of total industry employees 2008Q1-2009Q1 (percent)

2

-2

D

I

0 0.5

1

1.5

2A

2.5

3

F

3.5

NL

-6

FIN B

-8

DK

-12

-16

S

y = 3.2894x - 13.199 R² = 0.248

IRL

-10

-14

4.5

P

GR

-4

4

E

GB Employment protection legislation index

Figure 3: EPL and Total Industry Employment

Several reasons exist why the relationship between firing costs and the reaction of employment during the Great Recession cannot be detected using aggregate employment. First, the recession hit sectors in countries differently—e.g., in Spain construction industry was hit harder than in Germany—and employment protection often varies between sectors. Further, parts of the workforce are self-employed. Holland et al. (2010) point at severe effects of the crisis on self-employed and state substantial differences between countries in this regard. As the employment protection legislation does not affect self-employed, the changes in this part of the labor market varying between countries imply substantial noise for the regression. Finally, as firing costs can be avoided by reducing the number of temporary workers, the analysis may be affected by differing shares of temporary workers on the total workforce. Therefore, we also take into account changes in temporary work during the crisis. As has been reported by Booth et al. (2002), there is evidence that countries with higher firing costs are characterized by a higher degree of duality at the labor market, i.e., a larger share of temporary employment vis-à-vis regular employment. According to the authors, “the positive correlation can be interpreted as a clear sign that temporary contracts act as a way of providing employment flexibility in those countries with severe EPL for permanent jobs.” (pF184). In the course of the Great Recession, the data seems to suggest that, in line with intuition, temporary 4 The result is robust with respect to reasonable variations of the time span used in the calculation of the percentage change of the number of total industry employees.

The Role of Labor Market Institutions in the Great Recession

73

workers suffer more, the more protected are workers with regular contracts. Figure 4 shows that, in countries with higher firing costs for regular jobs as measured by the EPL index, the share of temporary workers in total industry declined more sharply.5 We cannot compare this empirical observation with predictions from the model, since the model only features one labor market in the sense that firing costs are the same for all workers. Labor market duality, however, is an interesting topic for future theoretical research.6

Change in the share of temporary workers in total industry 2008 Q1 - 2009 Q1 (percentage points)

2 1

DK GB

A

IRL

0 0.5

1

1.5

2

I

B

GR 2.5

-1 -2

D

3

3.5

4

4.5

NL FIN

S P

F -3 y = -1.0551x + 1.2148 R² = 0.2616

-4 -5

E

-6 Employment protection legislation index

Figure 4: EPL and Temporary Workers in Total Industry 4.2 Short-Time Work

We model short-time work by assuming that the government may allow firms to use this scheme for unproductive workers (see Faia et al. 2010). Whenever a worker does not generate any contemporaneous profits, the government allows that a certain specified share of this worker is co-financed by the government, while the working time is reduced by an equal share. It is assumed that the government uses short-time work as a discretionary instrument, i.e., it acts in countercyclical manner by allocating a lot of resources to it during the crisis and reducing the resources afterwards.7 Note that the R2 rises above 0.5 if Spain is left out. Spain in this sense is an outlier as it has a huge share of temporary workers compared to all other countries. 6 For a very recent attempt to tackle this issue, see Costain et al. (2009). 7 We assume that the short-time work intervention follows an AR(1) process with a quarterly autocorrelation of 0.9. 5

74

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl SHORT TIME WORK 0.5

0

-0.5

-1

Y: No Intervention Y: Intervention n: No Intervention n: Intervention

-1.5

-2

0

5

10

15

20

25

30

Figure 5: Output (Y) and Employment (n) Development for a Recession with and without Short-Time Work

To analyze the effects of short-time work, we compare a recession with and without short-time work. For analytical simplicity, in Figure 5 we use a productivity shock to model the recession.8 It is clearly visible that employment decreases less or even increases with the short-time work intervention (depending on the size of the short-time work intervention). At the same time, the fall in output is aggravated by the short-time work intervention. On the one side, financing part of the wage bill lowers the firing threshold, hence increases the number of employed workers. This explains the large employment effect. On the other side, short-time work endogenously reduces average productivity per worker for two reasons. First, workers with very low productivity are kept instead of being fired. Second, these workers reduce their weekly working time, i.e., overall production is reduced. An empirical assessment based on a cross country analysis is particularly hard to conduct in this case. Short-time regulations can be found in almost all countries of the Euro area, but they widely differ in terms of duration, replacement rates, or remaining costs for the employers. Furthermore, data on the incidence of short-time work is scarce. Figures provided by the OECD only cover the period up to 2008 and also lack comparability. For these reasons, we use a case study approach and compare the experiences of two countries with important short-time 8 We are well aware that the current crisis was not solely generated by a productivity shock. Aggregate demand shocks, however, would lead to similar conclusions.

The Role of Labor Market Institutions in the Great Recession

75

work schemes (Germany and Italy)9 with the Euro area as a whole and France, which has very little short-time work (CEEMET 2009, Banco d’Italia 2010, EMCO 2009). Figure 6 shows inverse unemployment as well as GDP for the Euro area and for Germany. The loss in terms of GDP is higher in Germany, while (un)employment is much less affected. If one assumes that all economies where hit by a huge shock in Q4 2008 and that Germany had a bigger intervention in terms of short-time work than the Euro area, the empirical counterparts do resemble the theoretical impulse response functions quite well. In addition, we compare the figures of Italy and France, see Figure 7. Both countries where hit by the severe downturn in international trade and had no serious internal problems like housing crises. While several similarities exist, Italy has a more generous short-time work scheme. Again in line with the theoretical model, the drop in GDP is higher in Italy, while employment is more damaged in France. Summarizing, both exercises confirm the predictions of the theoretical model. It is obvious, however, that the results of this case study are only suggestive. As more data concerning short-time work becomes available, a more systematic analysis is necessary.

Figure 6: GDP and Employment in Germany and the Euro Area

9 A further candidate with an important short-time work regulation is Austria; see Mahringer (2009).

76

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

Figure 7: GDP and Employment in France and Italy 4.3 Wage Formation

The wage formation mechanism can be influenced by a variety of labor market institutions, e.g., by the generosity of unemployment benefits, which affect the fall-back option in the bargaining process, or by the strength of unions. This section discusses the macroeconomic effects of wage moderation (i.e., lower wage claims by workers). To generate wage moderation in our theoretical model, we reduce the level of unemployment benefits permanently. However, similar effects could, for example, be generated by using a lower bargaining power for workers. Wage moderation makes it profitable for firms to hire and retain less productive workers because the average value of an entrant increases (see upper right panel in Figure 8). Thus, the steady state job-finding rate increases and the steady-state firing rate decreases. As a consequence, the economy moves to a higher level of employment and output, as can clearly be seen in Figure 8. Next, we perform a simulation exercise where the economy is hit by a negative aggregate productivity shock and policy makers change labor market institutions10 such that workers’ wage claims are reduced.11 Figure 9 shows the results for output 10 For simplicity, it is assumed that the labor market reform and the negative productivity shock occur at the same time. 11 This exercise may be in line with what happened in Germany and other countries, which reformed their unemployment benefits system before the Great Recession. For example, between 2003 and 2005, Germany has performed one of the major labor market reforms after the Second World War, the so called “Hartz reforms.” Among other things, these reforms de facto reduced unemployment benefits. Thus, according to our model the German economy may still have been in transition to the higher equilibrium employment level when foreign demand collapsed at the end of 2008.

The Role of Labor Market Institutions in the Great Recession

77

and employment for this type of exercise. Basically, the economy is hit by two shocks: A recession, decreasing output and employment, and a welfare reform, increasing output and employment. The net-effect lies in between and thus the welfare reform mitigates the negative effects of the recession. In the long-run, the economy moves to a higher steady state in terms of output and employment, similar to Figure 8. OUTPUT PATH (%DEVIATION)

VALUE OF AN ENTRANT (%DEVIATION)

0.35

0.4

0.3

0.3

0.25 0.2

0.2

0.15

0.1

0.1 0

0.05 0

0

10

20

30

40

-0.1

HIRING AND FIRING(%DEVIATION)

10

20

30

40

EMPLOYMENT (%DEVIATION)

3

0.35

2

0.3

1

0.25

0

0.2

-1

0.15

-2

0.1 Hiring Firing

-3 -4 0

0

10

20

30

40

0.05 0

0

10

20

30

40

Figure 8: The Effects of Wage Moderation

The link between labor market institutions and wage moderation is very difficult to establish,12 at least in a cross-country setting. Therefore, we focus on the link between wage moderation and employment, without taking a stance on what led to the wage moderation (in our model, various measures would have observationally equivalent effects). From a model perspective, the average values of entrants and incumbent workers (for a given macroeconomic stage) would be the best proxy for wage moderation, since they contain also the future development of wages that firms expect and take into account when hiring new workers and keeping or firing incumbent workers . But since this measure is not readily available in the data, we use unit labor costs as our proxy for wage moderation. Figure 10 relates the development of unit labor costs between 2002 and 2008 and the change of employment across European countries during the recession. In12 The biggest problem in this respect is the assessment of whether or not certain labor market reforms in different countries have had effects on work incentives and wages.

78

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

deed, those countries that experienced a less pronounced increase in unit labor costs in the past saw their employment deteriorate much less than countries with higher unit labor cost increases.13 UNEMPLOYMENT BENEFITS 0.5

0

-0.5

-1

Output - normal recession Output - recession with drop in b Employment - normal recession Employment - recession with drop in b

-1.5

-2

0

10

20

30

40

50

60

Change in nominal unit labor costs 20022008 (percent)

Figure 9: Wage Moderation in a Recession

-10

35

IRL

30 25

DK

E

P

y = -2.2248x + 8.5991 R² = 0.6478

GB

20

I GR

15

F

B

10

FIN

S

A

NL

5

D

0 -8

-6

-4

-2

0

2

Change in number of employees 2008Q2-2009Q2 (percent)

Figure 10: Unit Labor Costs and Employment 13 The result is robust with respect to reasonable variations of the time spans used in the calculation of the percentage changes of unit labor costs and employment.

The Role of Labor Market Institutions in the Great Recession

79

5. Conclusions In this paper, we have analyzed the role of labor market institutions in the Great Recession, both from a dynamic model perspective and from an empirical crosscountry perspective. First, the existing empirical literature shows that higher firing costs reduce macroeconomic volatilities (both employment and output), in line with our model prediction. In the Great Recession, this connection (i.e., firing costs as shock absorber) can be found for the industrial sector, where firing cost regulations are presumably most relevant. In addition, we can see that in countries with severe firing regulation, the pool of temporary employees in the industrial sector is affected more severely by employment reductions, leaving a more unclear aggregate picture. Second, our model predicts that short-time work may aggravate recessions with respect to GDP, but stabilizes employment. Comparing countries with important schemes for short-time work and countries without such schemes, we find evidence for this prediction. However, data is still very scarce with respect to comparing short-time work across countries. Finally, our model predicts that less generous unemployment benefits may help to moderate wages and increase equilibrium employment. Evidence for the connection between unit labor costs and the reaction of employment during the recession can clearly be found in the data. Thus, we conclude that although the current recession is very unusual in terms of its size, it has not abandoned usual economic mechanisms. However, it must be stressed that labor market institutions are not able to explain entirely the employment outcomes across countries during the Great Recession. Our exercise also shows that there are interesting questions for future research. From a theoretical perspective, the duality of labor markets and its welfare implications deserve more attention. From an empirical perspective, the current heavy short-time work interventions in many countries ask for further quantification and analysis. References Abbritti, M. / Weber, S. (2009): “Labor Market Institutions and the Business Cycle: Unemployment Rigidities vs. Real Wage Rigidities,” Mimeo, Graduate Institute of International Studies, Geneva. Banca d’Italia (2010): Economic Bulletin January 2010, 27 – 30. Booth, A. / Dolado, J. / Frank, J. (2002): “Symposium on Temporary Work – Introduction,” Economic Journal 112, F181 – F188. CEEMET (Council of European Employers of the Metal, Engineering and Technology-Based Industries) (2009): “Short-Time Working 2009,” via Internet (16 December 2009): . Costain, J. / Jimeno, J. F. / Thomas, C. (2009): “Employment Fluctuations in a Dual Labor Market,” Mimeo, Bank of Spain.

80

Jens Boysen-Hogrefe, Dominik Groll, Wolfgang Lechthaler, and Christian Merkl

Deutsche Bundesbank (2009): “Labor Markets in the Global Recession,” Monthly Report November 2009, 20 – 21. EMCO (Employment Committee) (2009): The 2009 Cambridge Review of the National Reform Programmes – Annual Report, via Internet (26 March 2010): . Faia, E. (2009): “Ramsey Monetary Policy with Labor Market Frictions,” Journal of Monetary Economics, 56(4), 570 – 581. Faia, E. / Lechthaler, W. / Merkl, C. (2009): “Labor Turnover Costs, Workers’ Heterogeneity, and Optimal Monetary Policy,” IZA Discussion Papers, No. 4322. Faia, E. / Lechthaler, W. / Merkl, C. (2010): “Fiscal Multipliers and the Labour Market in the Open Economy,” IZA Discussion Paper, No. 4849. Gali, J. (2008): Monetary Policy, Inflation, and the Business Cycle: An Introduction to the New Keynesian Framework, Princeton University Press, Princeton. Holland, D. / Kirby, S. / Whitworth, R. (2010): “A Comparison of Labour Market Responses to the Global Downturn,” National Institute Economic Review, 211, F38 – F42. Lechthaler, W. / Merkl, C. / Snower, D. (2010): “Monetary Persistence and the Labor Market: A New Perspective,” Journal of Economic Dynamics and Control,34(5), 968 – 983. Mahringer, H. (2009): “Der Arbeitsmarkt in der Finanzmarkt- und Wirtschaftskrise,” WiFo Monatsbericht Dezember, 967 – 978. Merkl, C. / Schmitz, T. (2010): “Macroeconomic Volatilities and the Labor Market: First Results from the Euro Experiment,” European Journal of Political Economy, forthcoming. Möller, J. (2010): “The German Labor Market Response in the World Recession – De-mystifying a Miracle,” Zeitschrift für ArbeitsmarktForschung, 42, 325 – 336. Orphanides, A. / van Norden, S. (2002): “The Unreliability of Output-Gap Estimates in Real Time,” The Review of Economics and Statistics, 84(4), 569 – 583. Smets, F. / Wouters, R. (2003): “An Estimated Dynamic Stochastic General Equilibrium Model of the Euro Area,” Journal of the European Economic Association, 1(5), 1123 – 1175. Woodford, M. (2003): Interest and Prices: Foundations of a Theory of Monetary Policy, Princeton University Press, Princeton.

Appendix: Data Sources EUROSTAT: Employees in total industry: Employment and unemployment (Labor Force Survey) ! LFS series—Detailed quarterly survey results (from 1998) ! Employees—LFS series ! Employees by sex, age groups and economic activity (from 2008, NACE rev.2) (lfsq_eegan2) ! Age: 15 years and over, NACE_R2: C, D, E. http:// epp.eurostat.ec.europa.eu/portal/ page/portal/employment_unemployment_lfs/data/database Employees total: Employment and unemployment (Labour Force Survey) ! LFS series— Detailed quarterly survey results (from 1998) ! Employees—LFS series ! Employees by sex, age groups and economic activity (from 2008, NACE rev.2) (lfsq_eegan2) ! Age: 15 years and over, NACE_R2: Total. http:// epp.eurostat.ec.europa.eu/portal/page/portal/ employment_unemployment_lfs/ data / database

The Role of Labor Market Institutions in the Great Recession

81

Temporary workers in total industry: Employment and unemployment (Labour Force Survey) ! LFS series—Detailed quarterly survey results (from 1998) ! Temporary employment– LFS series ! Temporary employees by sex, age groups and economic activity (from 2008, NACE rev. 2) (lfsq_etgan2) ! Age: 15 years and over, NACE_R2: C, D, E. http:// epp.eurostat.ec.europa.eu/portal/page/portal/employment_unemployment_lfs/data/database Temporary workers total: Employment and unemployment (Labour Force Survey) ! LFS series—Detailed quarterly survey results (from 1998) ! Temporary employment—LFS series ! Temporary employees by sex, age groups and economic activity (from 2008, NACE rev. 2) (lfsq_etgan2) ! Age: 15 years and over, NACE_R2: Total. http:// epp.eurostat. ec.europa.eu/portal/page/portal/employment_unemployment_lfs/data/database Unemployment: Employment and unemployment (Labour Force Survey) ! LFS main indicators ! Unemployment—LFS adjusted series ! Unemployment rate, quarterly average, by sex and age goups (une_rt_q) http:// epp.eurostat.ec.europa.eu/portal/page/portal/ employment_unemployment_lfs/data/database GDP: National accounts (including GDP) ! Quarterly national accounts ! GDP and main components ! GDP and main components—volumes (namq_gdp_k) ! Index, seasonally adjusted http:// epp.eurostat.ec.europa.eu/portal/page/portal/national_accounts/data/database Nominal unit labor costs: National accounts (including GDP) ! Annual national accounts ! Auxiliary indicators to national accounts—Annual data ! Unit labor cost—Annual data (nama_aux_ulc): http:// epp.eurostat.ec.europa.eu/portal/page/portal/national_accounts/ data/database OECD: Strictness of employment protection: Regular employment, Version 3 (2008): http://stats. oecd.org/Index.aspx?DataSetCode=EPL_R

The Role of Labor Market Institutions in the Great Recession Comment By Claus Schnabel*

Starting with the expansion of the welfare state and the oil crises in the 1970s, the importance of labor market institutions for macroeconomic performance has been widely researched in the last decades. Therefore it seems natural to investigate which role labor market institutions have played in the (ongoing) world-wide economic crisis that started in 2007 / 08.1 This is the aim of the stimulating paper by Boysen-Hogrefe et al. (2010), which specifically deals with three labor market institutions: higher firing costs due to employment protection laws, wage moderation due to reformed unemployment benefit systems, and short-term work (which in my opinion is an instrument of labor market policy rather than an institution). The authors analyze the effects of these institutions by combining two completely different approaches, namely a structural dynamic model perspective and an empirical cross-country perspective. Interestingly, both approaches deliver similar results concerning the role of labor market institutions for output and employment outcomes. The authors find that aggregate employment reductions are reduced by higher firing costs, but workers in the temporary employment segment are hit more severely in times of economic crisis, and recovery may be slower in those countries with higher firing costs. While short-time work is a suitable instrument to stabilize employment in light of a negative aggregate shock, the authors’ theoretical model predicts a more severe recession in terms of GDP in countries which make stronger use of this instrument. Finally, there are some indications that labor market reforms (such as a less generous unemployment benefit system) that result in wage moderation may dampen employment losses in the recession. * University of Erlangen-Nürnberg, Lange Gasse 20, 90403 Nürnberg, Germany, e-mail: [email protected] 1 Although this crisis has been referred to as “the great recession” by the authors of this paper and by others, we should be aware that the same phrase has been used to refer to almost every recession of the preceding decades; see the brief etymology of this term provided by Catherine Rampell in the New York Times Economix Blog, March 11, 2009 (http:// economix. blogs.nytimes.com/2009/03/11/great-recession-a-brief-etymology/).

84

Claus Schnabel

Even if the paper somehow oscillates between international comparisons and the specific situation in Germany, it delivers some plausible results and interesting insights. However, in the following I will question how robust these results are and I will argue that some important characteristics and institutions of (European) labor markets such as collective bargaining, wage rigidity and co-determination should also be taken into account when analyzing the output and employment effects of a severe recession. 1. Theoretical Modeling The paper makes use of the New Keynesian model framework by Lechthaler, Merkl and Snower (2010), which adds microfounded labor market frictions to an otherwise standard dynamic stochastic general equilibrium model with sticky prices. Hiring and firing decisions are endogenised, and employment becomes path-dependent. Since this model is only sketched in the paper, I will not discuss it in detail. In my opinion, the model is better suited than previous ones to simulate the output and employment effects of a severe recession in Europe. Somehow surprising, however, the authors use a productivity shock to model the recession whereas an aggregate demand shock would have been much more appropriate. They argue that this is done for analytical simplicity and that aggregate demand shocks would lead to similar conclusions. A second and more important objection concerns the modeling of wage determination in the paper. The model assumes that wages are bargained collectively between the median worker and the firm, and wage negotiations are described by the Nash-product, which depends on the respective bargaining powers of firms and workers. This might be seen as a rather crude (but somehow acceptable) way of taking into account trade unions, employers associations and sectoral bargaining which are characteristic for European labor markets. The crucial point, however, is that the authors largely neglect the phenomenon of (nominal and real) wage rigidity that characterizes labor markets in many European countries (see, e.g., Knoppik and Beissinger 2009, Babecky et al. 2009). While it is sometimes argued that the inclusion of real wage rigidity in New Keynesian models (as attempted, e.g., by Christoffel and Linzert 2005) is ad hoc and irrelevant (see Krause and Lubik 2007), in my opinion the simulations presented by Boysen-Hogreve et al. (2010) would be more convincing if they somehow took account of wage rigidities—no matter whether this was done ad hoc or based on microeconomic foundations.2

2 Lechthaler, Merkl and Snower (2010) admit that microfounding these real wage rigidities may entail other changes in the general equilibrium model that have significant effects on their predictions and policy implications. Interestingly, Boysen-Hogreve et al. (2010, footnote 2) argue that including real wage rigidities would affect their results only quantitatively, not qualitatively.

The Role of Labor Market Institutions in the Great Recession—Comment

85

2. Empirical Analysis In their empirical analysis, Boysen-Hogrefe et al. (2010) make use of simple comparisons between selected countries (in the case of short-term work) and of bivariate cross-country correlations for 14 countries. Without further justification, they restrict their sample to the EU 15 without Luxemburg, arguing that this panel is sufficiently homogeneous with respect to structural conditions so that simple correlations are meaningful without the consideration of additional control variables. An important and open question is whether the results would be similar if a different sample of countries was used, if countries and economic indicators were selected less arbitrarily, and if a more thorough econometric analysis was applied. All of these problems show up in the assessment of short-term work, where the authors base their empirical conclusions solely on two graphical comparisons: Germany is compared to the Euro area (not to the EU 15 without Luxemburg), and France is compared to Italy. While the authors claim that “both exercises confirm the predictions of the theoretical model”, I would be reluctant to draw firm conclusions from such selective comparisons. When analyzing the role of less generous welfare institutions and wage moderation, the authors argue that in Germany the “Hartz reforms” de facto reduced unemployment benefits and that the German economy may still have been in transition to the higher equilibrium employment level when foreign demand collapsed at the end of 2008 (thus neglecting the fact that the maximum duration of unemployment benefits was extended to 24 months in 2008 for older workers). Assuming that labor market reforms which reduce unemployment benefits lead to a slower increase in wages and thus a slower increase in unit labor costs, they try to provide empirical evidence by relating the development of nominal unit labor costs between 2002 and 2008 and the change of employment across European countries during the recession. Since it does not become clear why the time period 2002 to 2008 is chosen and since unit labor costs clearly reflect more factors than just labor market reforms (e.g. systems of wage determination not taken into account here), it is questionable what we really learn from the empirical analysis. Concerning the role of firing costs, the empirical analysis seems to be more meaningful at first sight. The authors present bivariate graphics and corresponding regressions relating the employment protection legislation index to the change in the “number of total industry employees” (not to the change in overall employment as used when investigating the role of wage moderation). While the authors are quite cautious in interpreting their results, it is highly questionable that simple correlations are meaningful when important control variables are missing (leading to an omitted variable bias). What would be interesting to know is whether and to which extent employment protection legislation is still associated with lower reductions in employment and with higher reductions in the share of temporary workers once the impact of other variables such as the reduction in output is taken into account—but this question cannot be answered by the empirical approach chosen.

86

Claus Schnabel

Furthermore, Figure 3 indicates that some countries seem to be outliers in the sense that they have extremely low or high values in the employment protection legislation index and / or they have experienced no or extremely strong reductions in employment. It would be interesting to know whether these countries are also outliers in a statistical sense—would excluding these countries from the sample substantially change the regression results?

3. Further Labor Market Institutions and their Impact on Employment While Boysen-Hogrefe et al. (2010) focus on two labor market institutions and one instrument of labor market policy which they are able to analyze both theoretically and empirically, there exist other institutions and characteristics of labor market that may also be relevant in the present crisis. These include wage rigidity (already mentioned above), minimum wages, collective bargaining, and co-determination. While a full discussion of these institutions and their likely effects is beyond the scope of this comment, it may be worthwhile to have a brief look at the empirical evidence concerning the (un)employment effects of these institutions, bearing in mind that this evidence stems from observation periods prior to the recent crisis:  Wage rigidity: There is ample empirical evidence for the existence of substantial downward nominal wage rigidity in European countries (see Knoppik and Beissinger 2009, Babecky et al. 2009), including Germany (see also Pfeiffer 2003). Downwardly rigid wages can be the result of statutory minimum wages, of union resistance to wage cuts, of employer considerations when paying efficiency wages and of other factors such as social norms, and they are usually associated with lower levels of employment.  Systems of wage determination: While there exists a large literature on the effects of corporatism and (de)centralized wage bargaining on macroeconomic performance started by Calmfors and Driffill (1988), this has not produced clear-cut and stable empirical results (see Aidt and Tzannatos 2002 for a survey). There is, however, some evidence that coordination in collective bargaining among and between trade unions and employers can help to offset adverse effects of unionism on employment (see, e.g., Nickell et al. 2005) and that wage, price and employment responses to economic shocks depend on collective wage bargaining (see Bertola et al. 2010).  Co-determination: Systems of co-determination by employees at company and / or establishment level can be found in many European countries (although the extent of employee participation rights differs substantially). It is not unlikely that firms’ reactions to the recent crisis have been influenced by ongoing consultation and co-determination processes. Even before the recent recession, works councils in Germany were found to reduce dismissals (see Frick 1996) and se-

The Role of Labor Market Institutions in the Great Recession—Comment

87

parations to non-employment (see Hirsch et al. 2009), whereas board-level representation of employees seems to be associated with longer payrolls (see Gorton and Schmid 2004). While minimum wages do not play an important role in Germany, the other three institutions sketched above seem to have been relevant for firms’ reactions to the recent crisis. The massive labor hoarding observed was not only made possible by generous short-term work provisions. It also reflected unions’ acceptance of low (or no) wage increases facilitated by the strong coordination and cooperation in the corporatist system of wage determination and labor market regulation in Germany (Sozialpartnerschaft). In addition, co-determination at company and establishment level may have played a role because it enabled unions, works councils and employees to “convince” management not to lay off employees (yet). Probably employment might have been stabilized more easily by higher wage flexibility, but in most sectors nominal (hourly) wages have been quite rigid downwards, and real wages also do not seem to have fallen much in the recession.

4. Conclusion While Boysen-Hogrefe et al. (2010) have written a stimulating paper on an important topic, the jury on the effects of labor market institutions in the recent recession is still out. Note that the authors were conducting their analyses while the recession was not over yet, and in some cases they were limited by the lack of (recent) data. Future empirical studies should make use of data for a larger number of countries (in particular when investigating short-term work), they should analyze longer time periods, and they should conduct multivariate rather than bivariate analyses. Furthermore, additional characteristics and institutions of European labor markets such as wage rigidity, collective bargaining and co-determination as well as the complex interdependencies of these and other institutions must be taken into account before we can claim to have understood the impact of labor market institutions on macroeconomic performance.

References Aidt, T. / Tzannatos, Z. (2002): Unions and Collective Bargaining: Economic Effects in a Global Environment, Washington, D.C. Babecky, J. / Caju, P. / Kosma, T. / Lawless, M. / Messina, J. / Room, T. (2009): Downward nominal and real wage rigidity: survey evidence from European firms, World Bank Policy Research Working Paper No. 5159, Washington, D.C. Bertola, G. / Dabusinskas, A. / Hoeberichts, M. / Izquierdo, M. / Kwapil, C. / Montornès, J. / Radowski, D. (2010): Price, Wage and Employment Response to Shocks: Evidence from the WDN Survey, European Central Bank Working Paper No. 1164, Frankfurt.

88

Claus Schnabel

Boysen-Hogrefe, J. / Groll, D. / Lechthaler, W. / Merkl, C. (2010): “The Role of Labor Market Institutions in the Great Recession”, this issue. Calmfors, L. / Driffill, J. (1988): Bargaining structure, corporatism and macroeconomic performance, Economic Policy 6, 13 – 61. Christoffel, K. P. / Linzert, T. (2005): The Role of Real Wage Rigidity and Labor Market Frictions for Unemployment and Inflation Dynamics, IZA Discussion Paper 1896, Bonn. Frick, B. (1996): “Co-determination and Personnel Turnover: The German Experience”, Labour 10, 407 – 430. Gorton, G. / Schmid, F. A. (2004): “Capital, labor, and the firm: A study of German codetermination”, Journal of the European Economic Association 2, 863 – 905. Hirsch, B. / Schank, T. / Schnabel, C. (2009): “Works Councils and Separations: Voice, Monopoly, and Insurance Effects”, IZA Discussion Paper No. 4126, Bonn, forthcoming in Industrial Relations. Knoppik, C. / Beissinger, T. (2009): “Downward Nominal Wage Rigidity in Europe – An Analysis of European Micro Data from the ECHP 1994 – 2001”, Empirical Economics 36, 321 – 338. Krause, M. U. / Lubik, T. A. (2007): “The (Ir)relevance of Real Wage Rigidity in the New Keynesian Model with Search Frictions”, Journal of Monetary Economics 54, 706 – 727. Lechthaler, W. / Merkl, C. / Snower, D. (2010): “Monetary Persistence and the Labor Market: A New Perspective”, Journal of Economic Dynamics and Control 34, 968 – 983. Nickell, S. / Nunziata, L. / Ochel, W. (2005): “Unemployment in the OECD since the 1960s”, Economic Journal 115, 1 – 27. Pfeiffer, F. (2003): Lohnrigiditäten im gemischten Lohnbildungssystem, Baden-Baden.

Labor Market Reforms, Hysteresis, and Business Cycles in Germany: A SVAR Approach to Explain Unemployment Developments By Alex Herzog-Stein* and Camille Logeay** Abstract Since the 1970s the German economy suffered from a pattern of step-wise rises in unemployment which was only broken during the last expansion in economic activity as the unemployment rate dropped to levels unseen since the 1990s. Against the backdrop of recent major labor market reforms this raises the question about the reasons for these shifts in unemployment over time. Therefore this study performs a time series analysis to examine the different impacts of institutional factors, cyclical shocks and hysteresis on unemployment in West Germany. This is done firstly via a descriptive comparative approach and secondly by estimating an autoregressive model in error-correction form for West Germany. The econometric results provide support for the hypothesis that cyclical shocks to the demand for labor and hysteresis played an important role in shaping labor market outcomes in Germany. Keywords: unemployment, vacancies JEL Classification: J60

1. Introduction In the last forty years the course of the German labor market was remarkable. In fact coinciding and partly due to the two oil crises, unemployment in Germany increased slowly but steadily from the 1970s onwards. After every economic slump unemployment remained at a higher level than before. In the course of the German reunification and the subsequent decade the unemployment situation in both parts of Germany, i.e., East and West Germany, deteriorated significantly further. During the same period in which Germany struggled some European countries like the * Corresponding Author. Institute of Economic and Social Research in the Hans Böckler Foundation (WSI), Düsseldorf, Germany. E-mail address: Alexander-Herzog-Stein@boeckler. de. ** Macroeconomic Policy Institute in the Hans Böckler Foundation (IMK), Düsseldorf, Germany. The authors thank Jürgen Jerger and Sven Schreiber for their insightful comments and suggestions and Christopher Lehmann for his research assistance. Any remaining errors are our own.

90

Alex Herzog-Stein and Camille Logeay

United Kingdom and the Netherlands successfully overcame their formerly even more severe unemployment problems. Voices were raised which lay the blame on labor market institutions especially the German system of wage replacement benefits for the unemployed and a lack of labor market flexibility, and asked for radical changes in labor market institutions and for labor market deregulation (see, e.g., German Council of Economic Experts 2000). With the recent labor market reforms many of these policy proposals were implemented. Marginal employment and temporary-agency work were deregulated and in 2005 unemployment assistance and social aid were replaced by a new more restrictive system of basic assistance called unemployment benefit II. In the following years German unemployment declined significantly and an economic debate began whether the improvement in labormarket outcomes was due to the implemented structural reforms or the consequence of the global economic boom. Against this backdrop it is of great importance to investigate the German unemployment performance since the 1970s. So far there is no general agreed answer among economists to the question about what the underlying causes of the observed development of the unemployment rate are. There are two contrasting views: The first one—favored by the economic mainstream—is the prevalent neoclassical explanation using labor market institutions to explain the observed unemployment time pattern. The second approach suggests that the observed step-wise rise in unemployment is the consequence of lacks in aggregate demand together with hysteresis mechanisms transforming cyclical into permanent unemployment over time. An attempt is made to discriminate empirically between these two alternative views. This is done with the help of the econometric approach suggested by Stegman and Stegman (2004), which is based on the Beveridge Curve interpreted as summary indicator of labor market efficiency as proposed by Solow (1998), which until now no one has tried to use to study German labor market performance since the 1970s: In a time series analysis a structural vector autoregressive model in error-correction form for West Germany is estimated and then it is investigated whether changes in the efficiency of labor market operations are associated with changes in labor market institutions and regulations, or instead correlated rather with cyclical factors. Furthermore, it is analyzed whether shifts in the Beveridge Curve are correlated with changes in search effectiveness which would lend support to the hypothesis that hysteresis is an important mechanism in explaining changes in labor market performance over time in Germany. In addition this econometric analysis is supplemented by a descriptive empirical approach comparing different aspects of the economic performance during the last two German business cycles. The paper is structured in the following way. In section two recent political efforts to reform the labor market in Germany are described. In section three based on the so called standard model of modern labor economics two contrasting views explaining the huge increases in unemployment—the institutional approach versus the hysteresis approach—are described and the empirical evidence with respect to

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

91

each explanation is presented. This is followed by a short discussion about labor market rigidities and the Beveridge Curve as a summary indicator of labor market efficiency in section four. In section 5 a time series analysis is performed to examine the different impacts of institutional factors, cyclical shocks and hysteresis on unemployment in Germany. This is done firstly via a descriptive comparative approach and secondly by estimating an autoregressive model in error-correction form for West Germany. Finally, section six concludes.

2. Labor Market Reforms in Germany As a consequence of the hotly debated so called Hartz Reforms of the German federal government under Chancellor Gerhard Schröder and the resulting societal split in advocates and opponents of these reforms the fact is easily overlooked that significant labor market reforms were already implemented in the 1990s. For example, with the “Acts on the Reform of Employment Promotion” (Arbeitsförderungs-Reformgesetz) in 1997 and 1998 Employment Promotion was reorganized, codified and then integrated into the new “Social Code Book 3” (Sozialgesetzbuch Drittes Buch). Furthermore the “criteria regarding suitable jobs” (Zumutbarkeitsregelungen) were tightened and unemployment benefit duration was reduced.1 However the aim of recalling these earlier labor market reforms is not to obscure the fact that with the Agenda 2010 and the Hartz Reforms the culmination point of these reform and deregulation efforts was reached. The “JobActiv Act”, the four “Laws for a modern provision of services on the labor market” (Gesetze für moderne Dienstleistungen am Arbeitsmarkt), and the “Act on Labor Market Reforms” (Gesetz zu Reformen am Arbeitsmarkt) had far reaching consequences for many labor market institutions. Some forms of non-standard employment like e.g. marginal employment and temporary-agency work were deregulated and actively promoted. Employment protection was reduced. The activation paradigm (Rights and Duties) became dominant in the field of employment promotion. Far reaching changes to the system of wage replacement benefits for the unemployed were introduced. Thus the duration of benefit entitlement was reduced and in 2005 unemployment assistance and social aid were replaced by a new more restrictive system of means-tested basic assistance called “unemployment benefit II” (Arbeitslosengeld II). At the same time in-work benefits as part of the “unemployment benefit II” were extended. The main aim of these latter measures was to reduce the level of the reservation wage and to increase work incentives. While the deregulation of non-standard employment was aiming to increase labor market flexibility, both the deregulation of non-standard forms of employment together with lower reservation wages and in1 For detailed overviews on German labour market institutions and labour market reforms see Bothfeld (2007) and Ebbinghaus and Eichhorst (2006).

92

Alex Herzog-Stein and Camille Logeay

creased work incentives were intended to open up potential employment in new areas especially in the service sector in general. More specifically the aim was to increase employment opportunities for the low-skilled and the long-term unemployed by increasing the low-wage sector in Germany.

3. Labor Market Institutions, Hysteresis, and Aggregate Demand: Theoretical Considerations and Empirical Evidence The dramatic shifts in unemployment in the OECD countries since the 1970s are a well documented fact. In many countries, especially in Europe, unemployment increased drastically and in general over time an upward trend in unemployment has been observable. This is also evident in those countries where unemployment declined significantly after reaching historically high levels, such as the Netherlands. In response to these developments since the late 1980s in modern labor economics a theoretical “standard model” has been established. This analytical framework takes the characteristics and the institutions of modern industrial economies explicitly into account. Based on imperfect competition in goods and labor markets this standard model is able to describe several important aspects like on the one hand high unemployment as an equilibrium phenomenon and on the other hand demand induced unemployment and persistency mechanisms describing how demand-deficient unemployment may be transformed into long-lasting equilibrium unemployment. Thus, in general this standard model presents an analytical framework that is broad enough to encompass competing explanations for the rise in unemployment over time and allows for a variety of supply-side as well as demand-side factors that determine the equilibrium level of unemployment.2 However, in recent decades among economists a narrow interpretation of the theoretical framework described above became the dominant explanation for the observed unemployment pattern in Germany and Europe. Labelled by Stockhammer (2008) as the “NAIRU story” it claims that labor market institutions are at the heart of the unemployment problem. On the one hand changes in country-specific labor market institutions resulted in overgenerous welfare states with, e.g., generous and durable unemployment benefits and pushed up wages, while on the other hand labor market institutions such as trade unions and strong employment protection legislation caused rigid labor markets. Overall, these institutional factors were 2 The most influential contributions are Layard, Nickell and Jackman (1991) and Phelps (1994). Interesting literature surveys and overview articles with different emphases are Bean (1994), Franz (1996), Beißinger and Möller (2000), Calmfors and Holmlund (2000), Hein (2004), and Kromphardt and Schneider (2007). Still by far the best economic textbook offering a detailed introduction to this approach and providing some interesting extensions is Carlin and Soskice (1990). For a comprehensive study analysing the German case within the framework of the standard model see Beißinger (1996).

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

93

responsible for a significant increase in equilibrium unemployment or NAIRU and hence the rise in overall unemployment rate since the 1970s. According to this view the blame for the German unemployment problem lay especially with the German system of wage replacement benefits for the unemployed, the institutional design of wage formation and a lack of labor market flexibility (see, e.g., Siebert 1997, 2003 and German Council of Economic Experts 2000). Alternative explanations stressing the importance of aggregate demand in fighting unemployment went largely unheard especially in Germany (see, e.g., Kromphardt 1999 and Solow 2000). In the next two sections therefore, firstly, the empirical evidence with respect to the influence of institutions in explaining unemployment is summarized. Secondly, the empirical literature about hysteresis, an important factor in explaining long-run unemployment effects of aggregate demand, is reviewed.

3.1 Institutions and Unemployment

With the advent of better and large panel datasets since the mid 1990s numerous empirical studies investigated the role and influence of labor market institutions on unemployment. The empirical approach is based on international panel-data sets and estimates multivariate equations with the unemployment rate as dependent variable. Apart from lagged values of the unemployment rate, explanatory variables normally included in this type of regressions are variables representing different types or aspects of labor market institutions, economic shocks3 and interactions between institutional variables and economic shocks. The institutional variables used are mostly indices based on classifications and ratings of the respective labor market institutions. Overall, the attempts to explain the rise and the magnitude of the country-specific unemployment rates with the help of labor market institutions have so far been disillusioning. The empirical results (see Table 7 in Appendix 3) cast serious doubts on the validity of this approach, or in other words: “The orthodox explanation of unemployment that argues that institutions matter has been subject to fairly extensive econometric testing, and in recent years, the validity of the empirical results supporting this view has been called into question. It has proved difficult to estimate a set of cross-country panel unemployment regressions that contain a lagged unemployment rate and a full set of year and country dummies and show that any of the labor market rigidity variables work. This is a crucial test. This is the first main similarity between European labor markets: labor market institutions do not tend to cause unemployment.” (Bell and Blanchflower 2009. p. 12) 3 In many studies the macroeconomic variables used are detrended. Hence by construction they can only explain cyclical movements of the unemployment rate around its trend. As a consequence there is the possibility that the effects of the macroeconomic variables are underestimated and the effects of the institutional variables overestimated (see Baccaro and Rei 2007, pp. 534 – 35 and 541; Bassanini and Duval 2006, p. 14).

94

Alex Herzog-Stein and Camille Logeay

Additionally some of the macroeconometric studies provide some interesting hints about the significance of labor market institutions for the observed increase in unemployment in Germany over time. Five of the studies presented in Table 7 (IMF 2003, Nickell et al. 2003, Nickell et al. 2005, Bassanini and Duval 2006, Bassanini and Duval 2009) do not only present the average explanatory power of the estimated equations, but also how good their estimations can explain the movement of unemployment over time in individual countries. In the IMF analysis the model is simulated keeping the institutional variables fixed on their level in 1970. In the case of Germany there is no improvement relative to the starting situation where the institutional variables are not fixed (see Figure 4.5). Surprisingly the IMF states that there was nearly no change in German institutions in the investigated period 1970 to 1998. This is a very interesting statement since the IMF used the dataset from Nickell and Nunziata (2001). However an investigation of the dataset from Nickell and Nunziata (2001) shows that in the time period 1960 to 1995 German labor market institutions changed quite a bit. Furthermore, in the time period investigated by the IMF (1970 to 1995) the statistically significant variables Employment Protection, Union Density and Tax Wedge decreased by 10.3 per cent respectively increased by 4.6 and 9.1 percentage points. Nickell et al. (2003, 2005) present results from comparable simulation exercises keeping the institutional variables fixed at their respective 1960s levels (see Figure 2). With respect to changing institutions they conclude that “They explain very little in Germany” (Nickell et al. 2003, p. 419). Furthermore if they exclude Germany from their analysis the explanatory power of their simulations of the increase in unemployment in the remainder of Europe in the period 1960 to 1990 – 1995 increases by eight percentage points (ibid., pp. 419 / 420). Bassanini and Duval (2006, Figure 1) quantify for their basic model without interactions between institutions and economic shocks how much each institution contributed to the actual change in the unemployment rate in each country. For Germany they conclude: “In particular, the gradual pick up in unemployment in Germany since unification is not properly explained by either policy or control variables included in the analysis” (ibid., p. 15). In Bassanini and Duval (2009) they simulate a non-linear model which seems to be able to explain the development of unemployment in all countries quite well (ibid., p. 9). Overall these results show that the conclusions valid for the average of the OECD countries do not necessarily hold for Germany. At least the labor market effects as a consequence of the German reunification until the end of the 1990s (the end of the time period investigated by most studies) seem to have played a significant role in explaining the unemployment level in Germany and hence dominate the results.

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

95

3.2 Hysteresis and Unemployment

The standard model in its general form described above does also allow for a condition of the economic system called hysteresis. The term hysteresis is used to describe the property of a dynamic economic system to be path dependent. In such a system the long-run equilibrium is not only determined by the long-run values of exogenous variables but also by the initial condition of each state variable (Franz 1990, p. 110). In reality hysteresis is often not clearly distinguished from the theoretically different property of a dynamic system called persistence. Persistence describes the property that a dynamic economic system exhibits a slow speed of adjustment towards its long-run equilibrium which is only determined by the long-run values of exogenous variables once the system has been disturbed by a shock. However in practice the distinction between a high degree of persistence and hysteresis seem to be of little relevance, because the medium-term NAIRU appears to be most policy relevant and given uncertainty in economic development the two concepts may be anyway (empirically) indistinguishable (Logeay and Tober 2006, p. 410). However, a high degree of persistence or hysteresis imply that cyclical economic shocks causing temporary changes in unemployment have longrun effects and insufficient aggregate demand might result in a higher equilibrium rate of unemployment. There are many causes and reasons why an economic system can be hysteretic.4 However its main aspect5 is “. . . that the number of long-term unemployed persons rises and that the latter exert less of an influence on labor market developments and on wages in particular than do the short-term unemployed.” (Logeay and Tober 2006, p. 411). Various studies investigated the existence of hysteresis for various countries.6 For example the influential work by Layard et al. (1991) tests for the significance of lagged unemployment in the NAIRU equation, and find strong evidence of hysteresis for Europe albeit not for the United States. But the most common approach is to use unit-root tests. However, it has the disadvantage that it cannot discriminate whether the unemployment rate is integrated of order one because of hysteresis or due to a change in the institutional structure of the labor market.7 Based on the panel unit root test developed by Im et al. (2003), León-Ledesma (2002) finds strong evidence for a unit root in Europe and weak evidence in the case of the United States. Furthermore, León-Ledesma and For details see the survey by Røed (1997) and Logeay (2006, pp. 90 ff.). Hysteresis in unemployment may be also caused by the adjustment of the capital stock over the business cycle or follow from multiply equilibria as a consequence of increasing returns to scale, high firing costs and the degree of labour market tightness (Røed 1997, pp. 401 ff.). However they are beyond the scope of this paper. 6 The aim of the following is to provide only a short overview over the large empirical literature testing the hysteresis hypothesis. For more information see Logeay (2006), pp. 118 ff. 7 For more details see, e.g., Logeay (2006, p. 118). 4 5

96

Alex Herzog-Stein and Camille Logeay

McAdam (2004) test for hysteresis in the ten new EU countries with the help of the Perron test in addition to the ADF test. They find that controlling for structural breaks leads to a rejection of the hysteresis hypothesis. These findings are supported by a corresponding panel analysis. However, when the authors interpret instead hysteresis in terms of multiple equilibria and apply the Markov-switching technique they find a low-unemployment regime, a high-unemployment regime and a high persistency of unemployment: thus supporting the hypothesis of hysteresis. A second group of studies uses the Kalman-filter technique. Tober and Logeay (2006) estimate the dependence of the NAIRU on lagged unemployment and longterm unemployment using the Kalman-filter technique. They find that the lagged unemployment rate (Euro Area) and the long-term unemployment rate (Germany) have a significant effect, which indicates the presence of hysteresis. For the United States Salemi (1999) simultaneously estimates the NAIRU using exogenous explanatory variables and the wage-setting and price-setting system using the Kalmanfilter technique. Testing for hysteresis by investigating whether the coefficient of the lagged natural rate of unemployment is one in the natural rate equation he does not find evidence for hysteresis. Jaeger and Parkinson (1994) use the Kalman-filter technique. They find significant hysteresis effects in Germany and the United Kingdom but not in Canada. A third group of studies tests for the existence of hysteresis using cointegration tests. For Norway Johansen (1995) tests for cointegration between the unemployment rate and the wage level, on the one hand, and between the unemployment rate, long-term unemployment and the wage level, on the other. He finds cointegration in both cases with long-term unemployment and the wage level being positively correlated. Hence for a given unemployment rate a larger share of long-term unemployed leads to a higher wage exactly as predicted by hysteresis. Overall these studies present ample evidence for hysteresis in various countries and regions. As a general pattern it seems that the studies find hysteresis at work in European countries, but there seems to be no evidence for hysteresis in the United States. With respect to Germany, several studies find evidence for hysteresis in the German labor market.

4. Labor Market Rigidity and the Beveridge Curve As described in the previous section although there is a general theoretical framework available to address the European and German unemployment puzzle, there is no general agreement about the factors responsible for the sharp increase in European unemployment and its persistence since the 1970s. Conceptually, there is also no general agreement among economists what is meant by labor market flexibility respectively labor market rigidity. Thus, in the Keynes Lecture in Economics 1997 Robert Solow remarked that while rigidities in the labor market are

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

97

blamed for the high unemployment in Europe in the last decades “. . . ‘labor market rigidity’ is never defined very precisely or directly in this context, but only by the enumeration of tell-tale symptoms.” (Solow 1998, p. 190) Instead, labor market flexibility should be thought of in terms of efficiency of the labor market and the “minimization of ‘structural’ unemployment” (Stegman and Stegman 2004, p. 751). Solow then goes on and suggests the Beveridge Curve or more precisely the location of the Beveridge Curve as his candidate for a useful summary indicator of labor market rigidity. If the Beveridge Curve is used as indicator of rigidities in the labor market then—as Stegman and Stegman (2004) point out—changes in labor market efficiency or in other words changes in the extent of “structural” problems in the labor market are reflected in a changing location of the Beveridge Curve. The more rigid the labor market the further away the Beveridge Curve is going to be from the origin. Furthermore, fluctuations over the business cycle cause anti-clockwise loops around the long-run relationship.8 There is a substantial empirical literature on the Beveridge Curve. Earlier examples are Budd et al. (1988), Jackman et al. (1989) and Borsch-Supan (1991). A more recent example is Nickell et al. (2003). They all use the standard approach of regressing unemployment on vacancies, and a stet of variables, such as the rate of employment separation, changes in labor supply and indicator variables approximating labor market institutions. However, as Stegman and Stegman (2004) points out they all do not address the question of stationarity. Furthermore, in recent German publications the Beveridge curve is used as an indicator about the state of the labor market and whether the labor market reforms initiated in the last decade have worked. For example, Gartner and Klinger (2008) use the Beveridge Curve in combination with the Philips Curve to assess whether there are differences between the most recent economic boom and the previous one (1999 / 2000). Furthermore the Deutsche Bundesbank interprets changes of a Beveridge Curve for Germany to the left in the years 2006 to 2008 as a hint that structural labor market problems have become less severe (Bundesbank 2008, p. 56). Next, in the following part this paper addresses the role of labor market institutions and the success of labor market reforms in improving labor market outcomes in Germany. Therefore firstly the economic performance and labor market developments over the last two business cycles are compared with each other. Then, the econometric approach suggested by Stegman and Stegman (2004) is used: based on the Beveridge Curve as summary indicator of labor market efficiency and by estimating a structural vector autoregressive model in error-correction form for West Germany it is investigated whether changes in the efficiency of labor market operations are associated with changes in labor market institutions and regulations, 8 For more details, see Stegman and Stegman (2004, p. 753) and Blanchard and Diamond (1989).

98

Alex Herzog-Stein and Camille Logeay

or instead correlated rather with cyclical factors. Furthermore, it is analyzed whether shifts in the Beveridge Curve are correlated with changes in search effectiveness and provide in this way empirical evidence for hysteresis.

5. Institutional Factors, Business Cycles and Hysteresis: A Time Series Analysis 5.1 Comparing Business Cycles: A Descriptive Comparative Approach

The following analysis compares the economic performance and labor market developments in Germany over the last two business cycles. Based on such a comparison it is attempted to find differences and similarities which might be attributed to recent labor market reforms. In this way indications can be collected, which either support or contradict the hypothesis that labor market reforms improved the economic and labor market performance in Germany. However, to be in a position to compare different business cycles first it is necessary to investigate the German business cycles and to date them. As in German Council of Economic Experts (2007, 2008) an approach based on the concept of relative output gaps is used to identify the different business cycles. For the determination of the relative output gaps different filter methods are used9. A through in economic activity is defined as a local minimum of the relative output gap series after which the relative output gap closes over time and stays positive for at least four quarters. In similar fashion a peak in economic activity is defined as a local maximum of the relative output gap series after which the relative output gap closes over time and stays negative for at least four quarters. The two most recent troughs occurred in the first quarter 1999 and the second quarter 2005. Peaks occurred in the first quarter 2001 and the first quarter 2008. Therefore Business Cycle I has a length of 25 quarters and its starting point is the trough which occurred in the first quarter 1999 and its end is the trough in the second quarter 2005. Business cycle II has so far a length of 18 quarters. Its starting point is the trough in the second quarter 2005 and its end is not determined yet. Looking at the economic performance in the two business cycles it is striking how similar GDP developed initially (see Figure 1). The major differences between the two business cycles are not the strength of GDP growth but the fact that the boom period in Business Cycle I ended after eight quarters whereas in Business Cycle II the expansion lasted eleven quarters.10 Furthermore the actual recession is much more severe. Whereas the recession in Business Cycle I lasted 17 quarters and looked very much like a long period of stagnation.

9 10

For a more detailed description of this method see Herzog-Stein (2010). In each graph the peak of the business cycle is indicated by a black triangular symbol.

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

99

Real Gross Domestic Product (seasonally and calendar adjusted)

Index ( Trough=100)

Cycle I (99q1 - 01q1 - 05q2)

Cycle II (05q2 - 08q1 - )

110 109 108 107 106 105 104 103 102 101 100 99 98 0

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

Quartale

Real Exports (seasonally and calendar adjusted) Cycle I (99q1 - 01q1 - 05q2)

Cycle II (05q2 - 08q1 - )

170.00 160.00

Index (Trou gh =100)

150.00 140.00 130.00 120.00 110.00 100.00 90.00 0

01

02

03 04

05 06

07

08 09

10 11 12

13 14

15

16

17

18

19 20

21 22 23

24 25

Quarters

Real Labor Costs per Employee (GDP-Deflator) (seasonally and calendar adjusted) Cycle II (05q2 - 08q1 - )

Cycle I (99q1 - 01q1 - 05q2) 105.00 104.00

In dex (Trou g h=100)

103.00 102.00 101.00 100.00 99.00 98.00 97.00 0

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

21

22

23

24

25

Quarters

Real Private Consumption (seasonally and calendar adjusted) Cycle I (99q1 - 01q1 - 05q2)

Cycle II (05q2 - 08q1 - )

106 105 In dex (Tro ug h=100)

104 103 102 101 100 99 98 0

01

02

03

04

05

06

07

08

09

10

11

12

13

Quarters

Sources: Federal Statistics Office of Germany; Own Calculations.

Figure 1

14

15

16

17

18

19

20

100

Alex Herzog-Stein and Camille Logeay

In both business cycles exports dominated. During both expansions real exports grew by around 29%. However during the recession in Business Cycle I real exports continued to grow significantly, whereas the actual recession is very much dominated by the slump in exports due to the global economic slump. Growth in domestic demand was weak in both expansions. After eight quarters real domestic demand had increased by 2.7% in Business Cycle I and by 3.1% in Business Cycle II. However at the end of the last expansion real domestic demand growth was equal to 5.7 %. This is an indication how important the length of a period of expansion is in terms of the economic performance. During the actual business cycle real wage costs per worker declined. However this pattern is also observable in Business Cycle I. From the seventh quarter onwards real wage cost per worker declined after initially real wage cost had grown by 4.2%. Correspondingly, private consumption stagnated more or less in the actual business cycle (peak: +1.8 %; after 18 quarters: +0.8 %). During Business Cycle I there was at least some moderate growth: during the expansion private consumption grew by 4.5% and then stagnated roughly during the recession. When the labor market performance in the two business cycles is compared the picture is mixed. The increase in total hours worked was larger and the reduction in unemployment was much more pronounced during the expansion in Business Cycle II than during the expansion in Business Cycle I. This is also true when the development in total hours worked and unemployment is compared after eight quarters (peak of the first business cycle) or after eighteen quarters (see Figure 2). However the labor market performance with respect to the development of employment subject to social security contributions and total employment is less impressive in Business Cycle II. After eight quarters total employment had increased a bit more in Business Cycle I (+2.9% versus +2.3%) and employment subject to social security contributions a bit less (+1.9 % versus 2.5) in Business Cycle II. Remarkable however is the performance during the actual recession. But it is hard to attribute this to the recent labor market reforms because their aim has been to increase external flexibility and this looks more like the result of increased internal flexibility.11 Furthermore, a comparison of the number of total hours worked when economic activity was at its peaks and at its troughs shows that there is no difference between the two business cycles (see Figure 3). Total employment has increased over time and at the last peak in economic activity was significantly higher than in the previous business cycle, but the number of employees subject to social security contributions was significantly lower than at the peak in economic activity in Business Cycle I. Finally, the potential labor force in Germany decreased since 2006. This negative labor supply effect is also a reason why unemployment declined so much during the actual business cycle. Furthermore it might be one of the reasons why employment was so far less affected by the economic recession. 11

For details see Herzog-Stein and Seifert 2010.

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

101

Total Hours Worked (Total Employees) (seasonally and calendar adjusted) Cycle I (99q1 - 01q1 - 05q2)

Cycle II (05q2 - 08q1 - )

105 104 Index (T ro u gh =100)

103 102 101 100 99 98 97 96 0

01

02 03

04 05

06 07

08

09

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

20 21

22 23

24

25

20

22

23

24

25

23

24

25

Quarters

Sources: Federal Statistics Office of Germany; Own Calculations. Total Employment (seasonally and calendar adjusted) Zyklus I (99q1 - 01q1 - 05q2)

>Zyklus II (05q2 - 08q1 - )

105

In dex ( Trough=100)

104 103 102 101 100 99 98 0

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

16 17

18 19

Quarters

Sources: Federal Statistics Office of Germany; Own Calculations. Registered Unemployment (seasonally adjusted) Zyklus I (99q1 - 01q1 - 05q2)

Zyklus II (05q2 - 08q1 - )

130

In dex (T ro ug h=100)

120 110 100 90 80 70 60 0

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

16

17

18

19

21

Quarters

Sources: German Bundesbank; Own Calculations. Employees Subject to Social Security Contributions (seasonally adjusted) Zyklus I (99q1 - 01q1 - 05q2)

Zyklus II (05q2 - 08q1 - )

108. 00 106. 00

Ind ex (T rou gh =100)

104. 00 102. 00 100. 00 98. 00 96. 00 94. 00 92. 00 0

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15

Quarters

Sources: Federal Employment Agency of Germany; Own Calculations.

Figure 2

16

17

18

19

20

21

22

102

Alex Herzog-Stein and Camille Logeay Total Employment, Total Hours Worked, and Employees Subject to Social Security Contributions (First Quarter 1999 to Fourth Quarter 2009) 108

Total Employment 1999q1=100t

Ind ex (1999q 1=100)

106 104 102 100 Total Hours Worked (Total Employees) 1999q1=100

98

Employees Subject to Social Security Contributions 1999q1=100t

96 94 1999

2002

2005

2008

Quarters

Sources: Federal Statistics Office of Germany; Federal Employment Agency of Germany; Own Calculations. Potential Labor Force (Differences in thousand, yo y) 500

Participation*

400 300 208

Total

226

200

Migration*

154 117

108

100

64

38

37 0

-29 -100

-65

-65 -98

Demographics -200 -300 98

99

0

1

2

3

4

5

6

7

8

9

Years

* Due to methodological changes there are not data avaible for 2009. Sources: Institut for Employment Research (IAB); Own Calculations.

Figure 3

Overall the results of the above analysis are mixed. While on the whole labor market performance during the actual business cycle seems to have been better, there are several factors which put this into perspective. The expansion lasted much longer and there was a very favorable global economic environment. Furthermore, in contrast to the previous business cycle the shrinking labor supply reduces labor market pressure. Thus, in conclusion this analysis does not provide conclusive hints that either support or contradict the hypothesis that labor market reforms improved the economic and labor market performance in Germany. 5.2 VAR AND VECM Analyses: An Econometric Approach

5.2.1 Description and Statistical Properties of the Data The data used for the econometric analysis are three time series: the unemployment rate (W_ALQ), the vacancy rate (W_OSQ), and the share of the long-term unemployment (W_LALQ) defined as the proportion of long-term unemployed in relation to the total number of unemployed. The analysis is restricted to West Ger-

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

103

many to avoid the evident statistical and definitional break in the years of the German reunification. Because monthly data for long-term unemployment are available only since December 1997 and before that date are only available annually, it was not possible to use the monthly or quarterly frequencies of the other variables. The used vacancy series provides only information about the vacancies notified to the Federal Employment Agency. From the national job vacancy statistics12 it is known that this is only a part of the supply of jobs available in the economy. However, data from the national job vacancy statistics are only available on a yearly basis and since 1989. The used sample consists therefore of annual data for the period 1969 to 2009. Despite the fact that only annual data are used the available time series are therefore longer than those used in the econometric studies summarized in Table 7 in Appendix 3. In the VAR / VEC-estimations an impulse dummy for 2005 (i05) is included to correct for the very sharp statistical break caused by the introduction of the Hartz IV legislation in that year. For detailed descriptions of the data see Appendix 1. Figure 4 presents the data used in the VAR / VEC-estimations. Two Beveridge curves for West Germany are presented in Figure 5 for the time periods 1962 to 2009 and 1985 to 2009 respectively. A break at the beginning of the 1980s in form of an outward shift of the Beveridge Curves is quite identifiable (left panel). There could be a second but less obvious break in the 1990s or 2000s in form of another outward shift (right panel). There is no clear indication for a third structural break in the form of an inward shift of the Beveridge curve after mid 2000 which then could be attributed to the Hartz Reforms especially the changes induced by the Hartz IV legislation. It looks more like short-run adjustments due to cyclical shocks in form of anti-clockwise loops around the long-run relationship as predicted for example by Blanchard and Diamond 1989. All variables are tested for unit roots using the Augmented Dickey-Fuller Test. These tests of the three time series for stationarity indicate that they all exhibit some non-stationary behavior and can be fairly considered as integrated of order one (I(1)). The results of the Augmented Dickey-Fuller Test are appended in Table 6 in Appendix 2. 6. Econometric Results In this section following the analysis of Stegman and Stegman (2004) West German data on the unemployment rate and vacancy rate is examined with a time series analysis in form of a structural vector autoregressive model in error-correction form (SVECM) in order to model shifts in the West German Beveridge Curve and their underlying causes econometrically.13 For details see Kettner 2008. For details on the econometric aspects of the estimation of VECM see for example Lütkepohl (1993), Stock and Watson (2001) or Lütkepohl and Kräzig (2004). 12 13

0

2

4

6

8

10

60

65

70

75

80

85

90

95

00

Unemployment Rate (W_ALQ) (in %)

05

0

1

2

3

4

60

65

70

75

85

90

Figure 4

80

95

Vacancy Rate (W_OSQ) (in %)

00

05

West Germany 1960 – 2009 (annual data)

0

10

20

30

40

60

65

70

75

80

85

90

95

00

05

Share of Long-Term Unemployed (W_LALQ) (in % of all unemployed)

104 Alex Herzog-Stein and Camille Logeay

Figure 5

Beveridge Curves for West Germany

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

105

106

Alex Herzog-Stein and Camille Logeay

Cointegration Analyses In the first step the aim of this econometric analysis is to establish whether there is a stable long-run Beveridge curve, i.e., a stable long-run relationship between the unemployment rate and vacancy rate. Therefore it is investigated whether there is a cointegration relationship between the variables of the Beveridge Curve. From a theoretical perspective it would make sense to perform a log-linear estimation (see Stegman and Stegman 2004, p. 755 and pp. 763 / 764). However, a closer investigation of the residuals gives no hints that a log-linear specification would be appropriate. Therefore no logarithmic transformation of the variables is performed. Initially, a cointegration analysis of the standard Beveridge variables unemployment rate and vacancy rate suggests the absence of cointegration between the unemployment rate and the vacancy rate. Hence the data do not imply a long-run stable relationship between the unemployment rate and vacancy rate over the sample period (Table 1; Bivariate Case). Following the interpretation of Stegman and Stegman (2004) this can be seen as evidence that the West German Beveridge Curve has shifted over time. Table 1 Engle-Granger Cointegration Tests (Sample: 1969 – 2009) Dependent Variable

t-stat

Crit. Val at:

1%

5%

10 %

–4.37 –4.12

–3.59 –3.46

–3.22 –3.13

–4.10 –3.92

–3.71 –3.58

Bivariate Case (T = 48 obs) W_ALQ W_OSQ

ADF ADF

–1.58 –2.11

W_ALQ W_OSQ

ADF ADF

–4.12 –2.86

T = 25 T = 50

Trivariate Case (T = 41 obs) T = 25 T = 50

–4.92 –4.59

Critical Values after McKinnon (1991).

Next as suggested by Stegman and Stegman (2004) and the literature on hysteresis the share of long-term unemployment (W_LALQ) is included as an additional variable in the cointegration model. In this way, the model is able to capture the effects of hysteresis, because the share of long-term unemployment as a proxy for the search effectiveness of the unemployed is treated as endogenous. The mechanism aimed to model is that there is an inverse relationship between search effectiveness and the share of long-term unemployment: “The persistence mechanisms of hysteresis hypotheses imply that, as the search effectiveness of the unemployed decreases, the proportion of unemployment which is long term increase.” (Stegman and Stegman 2004, p. 755) After the additional variable W_LALQ is added to the model, the Engle-Granger cointegration test suggests the existence of a long-run cointegrating relationship

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

107

between the unemployment rate, vacancy rate, and the share of long-term unemployment (Table 1, Trivariate Case). Additionally, the above result of the Engle-Granger procedure (one-equation) is supported by the Johansen cointegration tests (system estimation) shown in Table 2. Table 2 Johansen Cointegration Tests (Sample: 1980 – 2009) Hypothesized No. of CE(s)

Unrestricted Cointegration Rank Test (Trace) EigenTrace Small sample 1% 5% 10% value Statistic correction Critical Value (Trace Statistic)***

None At most 1 At most 2

0.563 0.333 0.147

Hypothesized No. of CE(s) None At most 1 At most 2

38.939 15.776 4.438

54.240 19.933 2.451

35.458 19.937 6.635

29.797 15.495 3.841

27.067 13.429 2.706

Unrestricted Cointegration Rank Test (Maximum Eigenvalue) EigenMaxSmall sample cor1% 5% 10% value Eigen rection (Max-Eigen Critical Value Statistic Statistic)*** 0.563 0.333 0.147

23.163 11.337 4.438

34.306 17.482 2.451

25.861 18.520 6.635

21.132 14.265 3.841

Prob.**

0.3% 4.5% 3.5% Prob.**

18.893 2.6% 12.297 13.8 % 2.706 3.5%

** MacKinnon-Haug-Michelis (1999) p-values. *** After Cheung / Lai (1993), test-stat* (T variables.

np) / T with T = number of observations, n = number of

Overall, for the sample period 1969 to 2009 the estimates for the cointegrating coefficients have the correct sign. They are equal to –2.37 for W_OSQ and 0.21 for W_LALQ, when the coefficient on W_ALQ is normalized to unity. They have the correct sign. The results of the cointegration analyses provide support for the hypothesis that the West German Beveridge Curve is unstable and that changes in the share of long-term unemployment as a proxy for changes in the search effectiveness of the unemployed have played a role in shifting the West German Beveridge Curve over the investigated sample period. A Structural Vector Error Correction Model After having identified a long-run cointegration relationship between the unemployment rate, vacancy rate, and the share of long-term unemployment for West Germany, next a structural vector autoregression analysis in form of a SVECM with one cointegration relationship is pursued. On the basis of these estimated SVECM an analysis of the short-run dynamics is performed that is consistent with the long-run relationship established above.

108

Alex Herzog-Stein and Camille Logeay

Based on a priori arguments about the way labor markets function Stegman and Stegman (2004) chose the Choleski decomposition proposed by Sims (1980) to impose restrictions on residual coefficients that enable the identification of the structural parameters.14 Following their approach the same Choleski ordering which restricts the residual coefficients triangularly is used to identify the structural parameters of the models. Firstly, a shock to the error term of the OSQ-equation contemporaneously affects all three variables. Secondly, a shock to the error term of the ALQ-equation has only a contemporaneous effect on unemployment itself and on the share of long-term unemployment. Finally, a shock to the error term of the LALQ-equation impacts contemporaneously only on the share of long-term unemployment. The structural shocks estimated along this identification scheme are interpreted as follows: the first shock (OSQ-shock) is a labor demand or business cycle shock, the second shock (ALQ-shock) is a labor-supply shock and the third shock (LALQ-shock) is a mismatch or ALMP shock. Given that, in contrast to Stegman and Stegman (2004) who use quarterly data, in this study annual data are utilized the applied Cholesky decomposition could be questioned. However, the importance of the ordering of the variables in the Cholesky decomposition for the impulse response functions and forecast error variance decompositions depends on the degree of correlation between the estimated residuals of the reduced form. Would they be uncorrelated, the ordering would not matter as they would subsume to the structural shocks (Enders 1995, pp. 276 – 277). An investigation of the estimated residuals of the reduced form shows that the W_OSQ residuals series is significantly correlated with the W_ALQ and W_LALQ residual series whereas these two are less correlated (only at a 5% but not at 1 % significance level). Changing the ordering of the variables will therefore affect the impulse response functions and forecast error variance decompositions presumably at the most for the business cycle shock. On the other hand another ordering would mean that a business cycle shock does not affect immediately either unemployment rate or long-term unemployment share or both. This is unconvincing and in contrast to the stylized facts. Exchanging ALQ with LALQ in the ordering changes almost nothing in the shape of the impulse response functions and forecast error variance decompositions. Therefore the ordering with OSQ in the pole position seems appropriate. Next the forecast error variance decompositions and the impulse response functions are presented. The forecast error variance decomposition (FEVD) gives the proportion of the variance of the endogenous variables—or more precisely of their stochastic part only—that is explained by each of the structural shocks. The impulse response functions (IRF) give the dynamic multipliers of the system, i.e., the response of the endogenous variables to each of the structural shocks over time. 14

For details see Stegman and Stegman (2004), p. 756.

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

109

Firstly, the FEVD is reported in Table 3. The variance decomposition of the W_OSQ-variance shows that it is essentially determined by itself, or in other words that the vacancies series is in the system almost an exogenous variable. At all horizons shocks to labor demand explain more than 85%. The explanatory power of the labor-supply shocks is negligible. Mismatch or ALMP shocks become more important in the long run but account only for approximately 12% of the forecast error variance in vacancies. Table 3 Variance Decomposition of W_OSQ: Period Std. Err. W_OSQ 1 0.234 100.00 2 0.384 99.46 3 0.487 98.33 4 0.563 96.76 5 0.625 94.95 6 0.680 93.15 7 0.729 91.49 8 0.775 90.00 9 0.818 88.70 10 0.858 87.56

W_ALQ 0.00 0.01 0.01 0.01 0.04 0.08 0.13 0.18 0.22 0.26

W_LALQ 0.00 0.53 1.66 3.23 5.01 6.77 8.38 9.82 11.08 12.18

Variance Decomposition of W_ALQ: Period Std. Err. W_OSQ 1 0.611 72.49 2 1.275 86.19 3 1.775 90.69 4 2.121 92.31 5 2.392 92.88 6 2.633 93.10 7 2.857 93.22 8 3.064 93.31 9 3.256 93.37 10 3.436 93.40

W_ALQ 27.51 13.78 9.28 7.66 7.06 6.80 6.63 6.51 6.42 6.35

W_LALQ 0.00 0.03 0.03 0.04 0.06 0.10 0.15 0.18 0.22 0.25

Variance Decomposition of W_LALQ: Period Std. Err. W_OSQ 1 1.456 20.61 2 2.394 19.30 3 3.851 48.62 4 5.074 61.03 5 5.936 65.33 6 6.613 67.14 7 7.218 68.28 8 7.780 69.20 9 8.301 69.91 10 8.785 70.46

W_ALQ 0.98 23.90 19.90 15.26 13.36 12.73 12.53 12.41 12.31 12.26

W_LALQ 78.41 56.80 31.48 23.71 21.32 20.13 19.19 18.40 17.77 17.29

110

Alex Herzog-Stein and Camille Logeay

The variance decomposition suggests that the W_ALQ-variance is more endogenous to the model and mostly determined by the business cycle shock. Over all horizons they explain in the range of 70% to 90% of the forecast error variance of the unemployment rate and they become more and more important over time. Overall this suggests a dominant role for cyclical shocks and hints at a very strong Okun’s relationship. It seems that past aggregate demand shocks are important in explaining the sharp increases in unemployment over time. The W_LALQ series shows evidence of being highly endogenous to the model. The variance of the share of long-term unemployment is determined in the longrun by business cycle shocks, around 70% of the forecast error variance is in the long run explained by shocks to labor demand. The mismatch or AMLP shocks have only a short-run effect that is however very strong (80 % versus a long-run effect of 20 %). The labor-supply shocks has a bell-shaped effect that is but small compared to the other two shocks, explaining 10% to 25 % of the forecast error variance in long-term unemployment. This points to hysteresis and persistence effects as business cycle shocks translate over time into long-term unemployment. Next, the impulse responses of each of the three variables with bootstrapped confidence intervals are reported. In the first column of Figure 6 the responses of each of the variables to a business cycle shock are presented; the second column presents the responses to a labor-supply shock and column three to a mismatch or ALMP shock. The IRF confirm that the business cycle shock is clearly dominant and the leading shock for all variables (first column). The responses of W_ALQ and W_LALQ to a business cycle shock are significant and consistent with the hypotheses presented here. A positive business cycle shock causes unemployment to fall permanently, with some slight undershooting initially. At the same time it reduces the share of long-term unemployment in the medium and long run. Initially however as expected, a positive business cycle shock increases the share of long-term unemployment as the short-term unemployed are hired first. The responses to a labor-supply shock in the labor market are presented in the second column of Figure 6. In response to a shock in unemployment the share of longterm unemployment increases over time as the labor market deteriorates and results in a significantly higher proportion of long-term unemployment as a consequence of persistence mechanisms of hysteresis. Initially, due to the restrictions imposed by the Cholesky ordering there is no contemporary response of vacancies to a W_ALQshock. In the following periods the response of vacancies is statistically insignificant. In column three of Figure 6 the responses to a mismatch or ALMP shock are shown. Due to the imposed restrictions there is no contemporaneous effect on either vacancies or unemployment rate. However the response of the unemployment rate is statistically insignificant over all horizons. Vacancies increase in response to the exogenous increase in mismatch as a consequence of the increase in the share of long-term unemployment. This response is statistically significant in the medium and long run.

111

Figure 6

VECM Orthogonal Impulse Responses

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

112

Alex Herzog-Stein and Camille Logeay

In conclusion the econometric results presented here for West Germany confirm the hypothesis that cyclical shocks together with hysteresis mechanisms played an important role in shaping labor market outcomes in Germany.

7. Conclusion The German economy suffered from a pattern of step-wise rises in unemployment since the 1970s which was only broken during the last expansion in economic activity as the unemployment rate dropped to levels unseen since the 1990s. This raises the question whether recent major labor market reforms, which changed important labor market institutions significantly, are responsible for this improved labor market performance or whether it is the result of a long global cyclical upswing which came to an end in 2008. Against this backdrop the analysis presented in this paper has addressed the role of labor market institutions and the success of labor market reforms in improving labor market outcomes versus the alternative explanation of cyclical effects of hysteresis in the German case. Therefore firstly the economic performance and labor market developments over the last two business cycles have been compared with each other. Then, as a novelty the econometric approach suggested by Stegman and Stegman (2004) has been used for a time series analysis of the German labor market. Based on the Beveridge Curve as summary indicator of labor market efficiency and by estimating a structural vector autoregressive model in error-correction form for West Germany the question has been investigated whether changes in the efficiency of labor market operations are associated with changes in labor market institutions and regulations, or instead correlated rather with cyclical factors. Furthermore, this paper analyzed whether shifts in the Beveridge Curve are correlated with changes in search effectiveness which would lend support to the hypothesis that with respect to the German labor market hysteresis is an important mechanism in explaining changes in labor market performance over time. The results of the comparison of the last two business cycles are mixed. While on the whole labor market performance during the actual business cycle seems to have been better, there are several factors which put this into perspective. The expansion lasted much longer and there was a very favorable global economic environment. Furthermore, in contrast to the previous business cycle the shrinking labor supply reduces labor market pressure. Thus, in conclusion the comparative analysis does not provide conclusive hints that either support or contradict the hypothesis that labor market reforms improved the economic and labor market performance in Germany. The estimation of a structural vector autoregressive model in error-correction form for West Germany provides support for the hypothesis that cyclical shocks to the demand for labor and hysteresis played a major role in shaping labor market outcomes in Germany. Furthermore the importance of the business cycle for the

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

113

development and structure of unemployment is corroborated. In conclusion the results of the econometric analysis presented here for West Germany so far suggest cyclical demand shocks in combination with hysteresis rather than changes of labor market institutions and deregulation are important in explaining changes in labor market efficiency. The research results presented here are only a first step and still of a preliminary nature. Future research will try to repeat the analysis presented here using quarterly data and closer examine the robustness of the presented results. Furthermore, attempts will be made to extend the analysis by estimating in addition to the Beveridge Curve a Phillips Curve using again the VECM approach. Additionally the influence of measures of active labor market policy will be closer examined.

References Baccaro, L. / Rei, D. (2007): “Institutional Determinants of Unemployment in OECD Countries: Does the Deregulatory View Hold Water?”, in: International Organization 61(3), 527 – 569. Bassanini, A. / Duval, R. (2006): “The Determinants of Unemployment Across OECD Countries: Reassessing the Role of Policies and Institutions,” in: OECD Economic Studies 42, 7 – 86. – (2009): “Unemployment, Institutions, and Reform Complementarities: Re-assessing the Aggregate Evidence for OECD countries,” in: Oxford Review of Economic Policy, Oxford University Press 25(1), 40 – 59. Bean, C.R. (1994): “European Unemployment: A Survey,” in: Journal of Economic Literature 32, 573 – 619. Beißinger, T. (1996): “Inflation und Arbeitslosigkeit in der Bundesrepublik Deutschland. Eine Analyse anhand von Modellen mit unvollkommenem Wettbewerb, Marburg, Metropolis Verlag. Beißinger, T. / Möller, J. (2000): “Unemployment: Theoretical Explanations”, in: “Globalization and Unemployment”, edited by H. Wagner, Berlin, Springer, 89 – 133. Bell, D. / Blanchflower, D. (2009): “What should be done about “Rising Unemployment in the UK?”, IZA Discussion Paper 4040. Blanchard, O. / Diamond, P. (1989): “The Beveridge Curve,” in: Brookings Papers on Economic Activity 1, 1 – 76. Borsch-Supan, A. (1991): “Panel Data Analysis of the Beveridge Curve: is there a Macroeconomic Relation between the Rate of Unemployment and the Vacancy Rate?”, in: Economica 58, 279 – 297. Bothfeld, S. (2007): “Labor Market Institutions in Germany: Current Status and Ongoing Reforms,” WSI Discussion Paper 152. Budd, A. / Levine, P. / Smith, P. (1988): “Unemployment, Vacancies and the Long-term Unemployed,” in: The Economic Journal 98, 1071 – 1091.

114

Alex Herzog-Stein and Camille Logeay

Calmfors, L. / Holmlund, B. (2000): “Unemployment and Economic Growth: A partial Survey”, in: Swedish Economic Policy Review 7, 107 – 153. Carlin, W. / Soskice, D. (1990): “Macroeconomics and the Wage Bargain: A Modern Approach to Employment, Inflation, and the Exchange Rate”, Oxford, Oxford University Press. Deutsche Bundesbank (2008): “Verbesserter Ausgleich am Arbeitsmarkt? – Evidenz von der Beveridge-Kurve,” Monatsbericht November 2008, Frankfurt / Main. Ebbinghaus, B. / Eichhorst, W. (2006): “Employment Regulation and Labor Market Policy in Germany 1991 – 2005”, IZA-Discussion Paper 2505. Enders, W. (1995): “Applied Econometric Times Series”, New York, John Wiley and Sons. Fahr, R. / Sunde, U. (2006): “Did the Hartz Reforms Speed-Up the Matching Process? A Macro-Evaluation Using Empirical Matching Functions”, in: German Economic Review 10(3), 284 – 316. Franz, W. (1990): “Hysteresis in Economic Relationships: An Overview,” in: Empirical Economics 15, 109 – 125. – (1996): “Theoretische Ansätze zur Erklärung der Arbeitslosigkeit. Wo stehen wir 1995?“,“ in: Arbeitslosigkeit und Möglichkeiten ihrer Überwindung, edited by B. Gahlen, H. Hesse, and H. J. Ramser, Tübingen, J.C.B. Mohr, 3 – 45. Gartner, H. / Klinger, S. (2008): “Ein janusköpfiger Aufschwung: Beschäftigungsgewinne und Polarisierung”, in: WSI-Mitteilungen 8, 439 – 446. German Council of Economic Experts (2000): “Chancen auf einen höheren Wachstumspfad”, Annual Report 2000 / 2001, Wiesbaden. – (2007): “Das Erreichte nicht verspielen”, Annual Report 2007 / 2008, Wiesbaden. – (2008): “Die Finanzkrise meistern – Wachstumskräfte stärken”, Annual Report 2008 / 2009, Wiesbaden. Hein, E. (2004): “Die NAIRU – eine Post-keynesianische Interpretation”, in: Intervention. Zeitschrift für Ökonomie 1(1), 43 – 66. Herzog-Stein, A. (2010): “Arbeitsmarktentwicklung und Bedeutung der Arbeitszeit im ersten Jahr der Krise: Ein Vergleich verschiedener Abschwungphasen seit 1970”, in: Zeitkonflikte. Renaissance der Arbeitszeitpolitik, edited by H. Groß, H. Seifert, im Erscheinen. Herzog-Stein, A. / Seifert, H. (2010): “Deutsches ‘Beschäftigungswunder’ und flexible Arbeitszeiten”, WSI Discussion Paper 169. Im, K. / Pesaran, M. H. / Shin, Y. (2003): “Testing for Unit Roots in Heterogeneous Panels,” in: Journal of Econometrics 115(1), 53 – 74. IMF (2003): World Economic Outlook: Unemployment and Labor Market Institutions: Why Reforms pay off, chapter 4, Washington, D. C. Jackman, R. / Layard, R. / Pissarides, C. (1989): “On Vacancies,” in: Oxford Bulletin of Economics and Statistics 51(4), 377 – 394. Jaeger, A. / Parkinson, M. (1994): “Some Evidence on Hysteresis in Unemployment Rates,” in: European Economic Review 38(2), 329 – 342. Johansen, K. (1995): “Norwegian Wage Curves,” in: Oxford Bulletin of Economics and Statistics 52(2), 229 – 247.

Labor Market Reforms, Hysteresis, and Business Cycles in Germany

115

Kettner, A. (2008): “Description of National Job Vacancy Statistics: Germany,” mimeo. Konle-Seidl, R. / Stephan, G. (2009): “Arbeitsmarktpolitische Maßnahmen für Beschäftigte und Arbeitslose: Möglichkeiten und Grenzen in der Krise”, in: Sozialer Fortschritt 11, 235 – 241. Kromphardt, J. (1999): “Ansatzpunkte der Beschäftigungspolitik aus keynesianischer Sicht”, in: Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 32, 499 – 513. Kromphardt, J. / Schneider, S. (2007): “Verknüpfung des Arbeitsmarktes mit dem Güter- und Geldmarkt bei Zinssteuerung und Inflation”, in: Volkswirtschaftslehre WISU 4, 476 – 581. Layard, R. / Nickell, S. / Jackman, R. (1991): “Unemployment: Macroeconomic Performance and the Labor Market,” Oxford, Oxford University Press. León-Ledesma, M. A. (2002): “Unemployment Hysteresis in the United States and the EU: A Panel Approach,” in: Bulletin of Economic Research 54(2), 95 – 103. León-Ledesma, M. A. / McAdam, P. (2004): “Unemployment, Hysteresis and Transition,” in: Scottish Journal of Political Economy 51(3), 377 – 401. Logeay, C. (2006): “Die Nairu: Erklärung oder Reflex der Arbeitslosenquote?”, Marburg: Metropolis Verlag. Logeay, C. / Tober, S. (2006): “Hysteresis and the Nairu in the Euro Area,” in: Scottish Journal of Political Economy 53(4), 409 – 429. Lütkepohl, H. (1993): Introduction to Multiple Time Series Analysis (2nd edition), Berlin, Springer Verlag. Lütkepohl, H. / Kräzig, M. (2004): Applied Time Series Econometrics, Cambridge, Cambridge University Press. Nickell, S. / Nunziata, L. (2001): Labor Market Institutions Database, London School of Economics, London, mimeo. Nickell, S. / Nunziata, L. / Ochel, W. (2005): “Unemployment in the OECD since the 1960s. What do we know?”, in: The Economic Journal 115(500), 1 – 27. Nickell, S. / Nunziata, L. / Ochel, W. / Quintini, G. (2003), “The Beveridge Curve, Unemployment, and Wages in the OECD from the 1960s to the 1990s,” in: Knowledge, Information, and Expectations in Modern Macroeconomics: In Honor of Edmund S. Phelps, edited by P. Aghion, R. Frydman, J. Stiglitz, M. Woodford, Princeton, Princeton University Press, 394 – 431. Phelps, E. S. (1994): “Structural Slumps – The Modern Equilibrium Theory of Unemployment, Interest, and Assets,” Cambridge (Massachusetts), Harvard University Press. Røed, K. (1997): “Hysteresis in Unemployment,” in: Journal of Economic Surveys 11(4), 389 – 418. Salemi, M. K. (1999): “Estimating the Natural Rate of Unemployment and Testing the Natural Rate Hypothesis,” in: Journal of Applied Econometrics 14(1), 1 – 25. Siebert, H. (1997): “Labor Market Rigidities: At the Root of Unemployment in Europe,” in: Journal of Economic Perspectives 11(3), 37 – 54. – (2003): “The Failure of the German Labor Market,” Kiel Working Paper 1169.

116

Alex Herzog-Stein and Camille Logeay

Sims, C. (1980): “Macroeconomics and Reality,” in: Econometrica 48, 1 – 49. Solow, R. M. (1998): “What is Labor-Market Flexibility? What is it Good for?”, Keynes Lecture in Economics, in: Proceedings of the British Academy 97, 189 – 211. – (2000): “Unemployment in the United States and in Europe: A Contrast and the Reasons,” CESifo Working Paper 231. Stegman, A. / Stegman, T. (2004): “Labor Market Reform and the Macroeconomic Efficiency of the Labor Market in Australia,” in: Cambridge Journal of Economics 28, 743 – 766. Stock, J. H. / Watson, M. W. (2001): “Vector Autoregressions,” in: Journal of Economic Perspectives 15(4), 101 – 116. Stockhammer, E. (2008): Is the NAIRU Theory a Monetarist, New Keynesian, Post Keynesian or a Marxist Theory?, in: Metroeconomica 59(3), 479 – 510.

Appendix 1 Table 4 W ALQ ALO OSQ OS LALO LALQ

West German data. Unemployment rate in % (registred unemployed / civilian labor force). Seasonally adjusted, Source: Deutsche Bundesbank. Registred unemployed persons in thousand. Seasonally adjusted, Source: Deutsche Bundesbank. Vacancy rate in % (vacancies / civilian labor force). Seasonally adjusted, own calculations. Vacancies in thousand. Seasonally adjusted, Source: Deutsche Bundesbank. Long-term unemployment (more than one year) in thousand persons. Value for september till 1997, yearly mean from 1998 onwards. Sources: IAB, Federal Employment Agency and German Council of Economic Experts. Share of long-term unemployed in relation to total number of unemployed in % (LALO / ALO). Own calculations.

Table 5

W_ALO* W_ALQ* W_LALO W_LALQ W_OS* W_OSQ*

Maximum

Minimum

Std. Dev.

Observations

Sample

3246.17 9.88 1121.39 39.60 796.83 3.11

146.00 0.57 7.76 4.15 76.00 0.27

1016.29 3.13 366.06 11.12 172.63 0.76

50 48 41 41 50 48

1960 – 2009 1962 – 2009 1969 – 2009 1969 – 2009 1960 – 2009 1962 – 2009

* Annual means calculated from the seasonally adjusted data.

+

2 (AIC) 2 (HQ) 1 (AIC) 1 (HQ) 1 (AIC) 1 (HQ) 1 (AIC) 1 (HQ) 2 (AIC) 1 (HQ) 2 (AIC) 1 (HQ) 1 (AIC) 1 (HQ) 1 (AIC) 1 (HQ) 0 (AIC) 0 (HQ) 0 (AIC) 0 (HQ) 0 (AIC) 0 (HQ) 0 (AIC) 0 (HQ)

Info-criterium and Lag-length Constant Constant Constant, Linear Trend Constant, Linear Trend Constant Constant Constant, Linear Trend Constant, Linear Trend Constant Constant Constant, Linear Trend Constant, Linear Trend Constant Constant Constant, Linear Trend Constant, Linear Trend Constant Constant Constant, Linear Trend Constant, Linear Trend Constant Constant Constant, Linear Trend Constant, Linear Trend

Constant & Trend –1.559 –1.559 –2.832 –2.832 –4.923 –4.923 –4.983 –4.983 –2.289 –2.313 –1.868 –2.369 –5.729 –5.729 –5.899 –5.899 –1.498 –1.498 –0.751 –0.751 –5.208 –5.208 –5.488 –5.488

t-statistic

MacKinnon (1996) one-sided p-values; Significance Levels: *** 1 % Level, ** 5 % Level, * 10 % Level.

W_ALQ, level W_ALQ, level W_ALQ, level W ALQ, level W ALQ, first difference W_ALQ, first difference W_ALQ, first difference W_ALQ, first difference W_OSQ, level W OSQ, level W_OSQ, level W OSQ, level W_OSQ, first difference W_OSQ, first difference W_OSQ, first difference W_OSQ, first difference W_LALQ, level W_LALQ, level W_LALQ, level W_LALQ, level W_LALQ, first difference W LALQ, first difference W_LALQ, first difference W_LALQ, first difference

Variable

Augmented Dickey-Fuller Tests West Germany: 1960 – 2009; maximum lag length = 3

Table 6

Appendix 2

49.5% 49.5% 19.4% 19.4% 0.0% 0.0% 0.1% 0.1% 18.0% 17.2% 65.5% 39.0% 0.0% 0.0% 0.0% 0.0% 52.5% 52.5% 96.2% 96.2% 0.0% 0.0% 0.0% 0.0%

one-sided p-value+

*** *** *** ***

*** *** *** ***

*** *** *** ***

H0 of unit root

45 45 46 46 45 45 45 45 45 46 45 46 45 45 45 45 40 40 40 40 39 39 39 39

nb obs

Labor Market Reforms, Hysteresis, and Business Cycles in Germany 117

+

–

Taxes

Coordination

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

Real interest rate

Labor demand

Money supply

Real import price / oil price

Share of young people (15 – 24 years)

insignificant n.a.

Productivity

Output gap

Shocks Change in inflation

ALMP

Employment protection

insignificant –

+

Union coverage

Union density

insignificant +

+

Nickell (1997)

Benefit duration

Institutions Benefit replacement ratio

Studies:

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

–

n.a.

+ partly significant – partly significant

–

+

insignificant n.a.

n.a.

+

Elmeskov et al. (1998)

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

–

n.a.

+ partly significant insignificant insignificant insignificant + partly significant – partly significant insignificant insignificant

OECD (1999)

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

–

insignificant –

–

+

+

+

+

+

Nickell & Layard (1999)

n.a.

n.a.

n.a.

–

+

–

n.a.

n.a.

– partly significant + partly significant insignificant

+ partly significant + partly significant + partly significant + partly significant +

Blanchard & Wolfers (2000)

insignificant / +

n.a.

insignificant n.a.

insignificant +

n.a.

–

insignificant

insignificant – partly significant + partly significant – partly significant +

+ partly significant n.a.

Bertola et al. (2001)

Table 7

Appendix 3

n.a.

insignificant +

–

+

–

n.a.

n.a.

insignificant n.a.

n.a.

n.a.

+

countryspecific insignificant –

n.a.

+

–

– insignificant n.a.

+

n.a.

+

+ partly significant n.a.

IWF (2003)

+

insignificant n.a.

+

+

Nickell et al. (2003)

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

–

insignificant insignificant insignificant n.a.

+ partly significant n.a.

+ partly significant n.a.

Belot & Van Ours (2004)

n.a.

insignificant +

+ partly significant -

–

n.a.

n.a.

insignificant n.a.

+ partly significant –

+ partly significant insignificant n.a.

+

Nickell et al. (2005)

n.a.

n.a.

n.a.

n.a.

– partly significant +

– partly significant n.a.

insignificant n.a.

+ / – partly significant – partly significant n.a.

+

insignificant n.a.

Baccaro & Rei (2007)

+ (terms of trade) n.a.

n.a.

+

+

–

–

n.a.

insignificant – partly significant

–

+

insignificant n.a.

n.a.

+

Bassanini & Duval (2006)

n.a.

n.a

n.a.

n.a.

n.a.

n.a.

–

n.a.

– partly significant + / – partly significant n.a.

+

+ partly significant n.a.

n.a.

+

Bassanini & Duval (2009)

118 Alex Herzog-Stein and Camille Logeay

no no no

yes no no yes no

no no yes no no

no yes yes yes no

no yes yes yes yes

yes no yes no yes

yes no yes yes no

yes no yes yes yes

yes yes yes yes no

yes no yes yes no

yes yes yes yes no

yes no

N.B.: If available, only those estimations are considered which include institutions and macroeconomic shocks with or without interactions. Only the most common explanatory Variables are presented. To ensure comparability only estimations with the unemployment rate as dependent variable are considered.

Sources: Nickell (1997): “Unemployment and Labor Market Rigidities: Europe Versus North America “Journal of Economic Perspektives VoI. 11(3), S. 55 – 74. Elmeskov et al. (1998): “Key Lessons for Labour Market Reforms: Evidence from OECD Countries’ Experiences”. Swedish Economic Policy Review Vol. 5(2), S. 205 – 252. OECD (1999): Ernployment Outlook, Chap. 2. Nickell / Layard (1999): “Labor Market Institution and Economic Performance”. O. Ashenfelter und D. Card (Hrsg.), Handbook of Labor Economics Vol.3 / Chap. 46, S. 3029 – 3084. Blanchard / Wolfers (2000): “The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence”. Economic Journal Vol. 110(March), 5. C1 – C33. Bertola et al. (2001): “Comparative Analysis of Labor Market Outcomes: Lessons for the US from International Long-Run Evidence”. NBER Working Paper Nr. 8526. Nickell et al. (2003): “The Beveridge Curve, Unemployment, and Wages in the OECD from the 1960s to the 1990s” P. Aghion, R. Frydman, J. Stiglitz und M. Woodford (Hrsg.), Knowledge, Information, and Expectations in Modem Macroeconomics: In Honor of Edmund S. Phelps, S. 394 – 431. IWF (2003): World Economic Outlook, Chapter 4. S. 129 – 150. Belot / van Ours (2004): “Does the Recent Success of Some OECD Countries in Lowering their Unemployment Rates Lie in the Clever Design of their Labour Market Reforms?”. Oxford Economic Papers, Vol. 2004(56), S. 621 – 642. Nickell et al. (2005): “Unemployment in the OECD since the 1960s. what do we know?”. The Economic Journal, Vol 115(500), S. 1 – 27. Baccaro / Rei (2007): “Institutional Determinants of Unemployment in OECD Countries: Does the Deregulatory View Hold Water?”. International Organization, Vol. 61(3), S. 527 – 569. Bassanini / Duval (2006): “The Determinants of Unemployment Across OECD Countries: Reassessing the Role of Policies and Institutions”. OECD Economic Studies, Vol. 2006(42-I), S. 7 – 86. Bassanini / Duval (2009): “Unemployment, institutions, and reform complementarities: re-assessing the aggregate evidence for OECD countries”. Oxford Review of Economic Policy, Vol. 25(1), S. 40 – 59.

Notes: The dependent variable is always the unemployment rate. + (–): An increase in this variable increases (decreases) unemployment n.a.: The corresponding variable was not used in this study.

Number of countries

yes no no

yes no

1983 – 1994 1983 – 1995 1985 – 1997 1983 – 1994 1960 – 1995 1070 – 1996 1961 – 1992 1960 – 1998 1960 – 1999 1961 – 1995 1960 – 1998 1982 – 2003 1982 – 2003 6 years annual 6 years 6 years 5 years 5 years annual annual 5 years annual annual / annual annual / 5 years 5 years 20 19 19 20 20 20 20 20 17 20 18 20 20

no yes no

Fixed Effects Country Time Country-specific trend

Dataset Time period Periodicity

no no

Interactions Institutions + Institutions Institutions + Shocks

Labor Market Reforms, Hysteresis, and Business Cycles in Germany 119

Labor Market Reforms, Hysteresis, and Business Cycles in Germany: A SVAR Approach to Explain Unemployment Developments Comment By Jürgen Jerger*

The paper by Herzog-Stein and Logeay provides an account of the more recent history of unemployment in Germany from a macroeconom(etr)ic point of view. This is–at least in principle—a welcome change from the now dominant literature on microeconometric evaluations of labor market reforms. The obvious disadvantage of using macroeconometric data is the limited amount of observations, especially on the relevant policy variables that are in the focus of reforms. Usually this problem is dealt with by using panel data, thereby exploring cross-country variation in the data in addition to the scant variation over time (see Layard / Nickell 1999, Nickell / Nunziata / Ochel 2005). In contrast to this literature, Herzog-Stein and Logeay confine themselves to German data, using a SVAR approach adopted from Stegman / Stegmann (2004). Moreover, they provide a useful account of the labor market reforms in Germany and a descriptive analysis of major features of the last two business cycles. In section 3 of the paper, the authors discuss alternative explanations of unemployment along the lines of the well-known wage-setting / price-setting model and review the empirical literature that builds on this framework. Institutions such as unemployment benefits, taxes and employment protection, to name just a few, can be easily linked to the key relationships of this model. It is well known that the empirical performance of this model is far from undisputable. This is especially so in the case of Germany which is partly due to the fact that the German reunification as a major driver of recent labor market developments, is not easily captured as an “institutional” feature. Nevertheless, the literature has certainly shown that institutions do matter and usually affect (un)employment in a way that is consistent with theory. This is even more so if one takes into account the interaction between shocks and institutions (e.g., Blanchard / Wolfers 2007). Furthermore, interactions between different aspects of labor market institutions—e.g., the level and duration of unemployment benefits—have been shown to matter. Herzog-Stein and Logeay * Osteuropa-Institut Regensburg and University of Regensburg, Department of Economics, E-mail address: [email protected].

122

Jürgen Jerger

choose to downplay these successes somewhat by pointing to the existence and relevance of hysteresis in the labor markets. This choice seems to be motivated by the attempt to separate the impacts of “institutions” and “hysteresis,” respectively. In my view, however, this distinction is not really helpful for two reasons. First, hysteresis (or persistence if one allows for less extreme forms of path-dependency) is a time series property reflecting the propagation of shocks that due to its symmetry at least in principle has no normative value attached—in contrast to institutions or shocks that affect unemployment. Secondly, and more importantly, institutions themselves affect the presence and degree of persistence. The relevance of the insider-outsider mechanism, for example, clearly depends on institutional choices that in turn affect the persistence of unemployment. Therefore, establishing the relevance of hysteresis or persistence is by no means equivalent to the statement that institutions are of lesser importance for the explanation of labor market developments. One might even argue to the contrary that institutions need to be designed such that adverse shocks are not allowed to translate into persistent unemployment—as it was empirically the case in Germany since the 1970s. A good deal of the recent labor market reforms in Germany was at least partly motivated by this idea. The empirical section 5 of the paper starts out with an informative comparison of the two most recent business cycles in Germany. The authors summarize this descriptive exercise in the conclusions of their paper with the statement that the “analysis does not provide conclusive hints that either support or contradict the hypothesis that labor market reforms improved the economic and labor market performance in Germany.” Whereas this is necessarily true simply since the purely descriptive analysis does not allow for any hypothesis testing, the differences between the two cycles seem to be quite noteworthy. More specifically, the relative robustness of the employment level in the recent downswing is very markedly different from previous experience indeed. Hence, just looking at the data strongly suggests that something has happened that changed the business cycle behaviour of the German labor market to the better. The recent labour market reforms and the resulting wage restraint do not seem to be too far-fetched candidates for this welcome change. Along the lines of Stegman / Stegman (2004) the econometric exercises start with a test for cointegration between vacancies and unemployment. The absence of such a stable long-run relationship is interpreted as evidence for shifts of the Beveridge curve. Whereas the difficulties of empirically identifying those shifts are well known, the absence of a cointegration relationship does not necessarily imply the absence of a stable Beveridge curve. More specifically, movements along a stable but non-linear Beveridge curve are not necessarily reflected in a linear cointegration relationship. Therefore, the absence of a linear relationship does not logically imply the absence of a stable Beveridge curve. Furthermore, there are severe measurement problems, especially concerning the vacancy rate. These problems also may affect the validity of the other results in the paper.

Labor Market Reforms—Comment

123

When long-term unemployment is added to the two variables mentioned before, there is evidence of a cointegration vector. Herzog-Stein and Logeay interpret “the share of long-term unemployment as a proxy for search effectiveness.” This is far from clear, however, because other potential channels that are consistent with excess supply on the labor market—as opposed to search frictions that entail both vacancies and unemployment—are at least as likely to be reflected in long-term unemployment. Based on the cointegration relationship, the authors estimate a structural vector error correction model (VECM) in order to look at the short-run dynamics of different shocks. These shocks, labeled as business cycle shock, hysteresis shock, and mismatch shock, respectively, are identified by a simple Choleski decomposition. Neither the order of the variables (and the implied restrictions) nor the nature of the structural shocks are explained in any detail, leaving doubts concerning the precise meaning of the exercise. In particular, it remains quite unclear exactly what a “hysteresis shock” really means. The use of the Choleski decomposition is especially problematic in the empirical context of annual data, since it is not plausible that any of the variables does not react to (some of) the shocks within one year. In my view, it would have been more straightforward to formulate appropriate longterm restrictions for the identification of the respective structural shocks. More crucially, however, the set-up is simply not able to identify anything about possible effects of labor market reforms or labor market institutions as claimed in the title of the paper. This is simply due to the fact that the model does not contain any variable that can be associated with labor market policies and is estimated on the entire sample without allowing for time-varying coefficients. Hence, the empirical model cannot produce any conclusion concerning the effects of the recent labor market reforms. What can be shown, however, is that unemployment and long-term unemployment react qualitatively similar to all shocks—which shouldn’t come as a big surprise—and that the business cycle shock is rather important. It is, however, more than courageous to jump from this result to the conclusion that “demand shocks in combination with hysteresis rather than the reform of labor market institutions and deregulation are important in explaining changes in labor market efficiency” as the authors claim. My reservations concerning this interpretation are due to both major critical remarks of my comments: First, there is no convincing rationale to separate institutions and hysteresis as alternative patterns of explanation; and second, the empirical model does not allow for testing the effects of institutions or the change of institutions. References Blanchard, O. / Wolfers, J. (2007): “The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence,” in Recent Developments in Labor Economics, Volume 3, edited by J. T. Addison, International Library of Critical Writings in Economics, vol. 207. Cheltenham (U.K.) and Northampton (Mass.): Elgar; 2007, 208 – 240.

124

Jürgen Jerger

Nickell, S. / Layard, R. (1999): “Labor Market Institutions and Economic Performance,” in Handbook of labor economics, Volume 3C, edited by O. Ashenfelter and David Card, Handbooks in Economics, vol. 5, 1999, 3029 – 3084. Nickell, S. / Nunziata, L. / Ochel, W. (2005): “Unemployment in the OECD Since the 1960s: What Do We Know?”, Economic Journal, 115(500), 1 – 27. Stegman, A. / Stegman, T. (2004): “Labour Market Reform and the Macroeconomic Efficiency of the Labour Market in Australia,” Cambridge Journal of Economics, 28(5), 743 – 766.

Safeguarding Jobs through Labor Hoarding in Germany By Martin Dietz,* Michael Stops,** and Ulrich Walwei***

Abstract As a consequence of the global financial crisis Germany has experienced the deepest slowdown of its economy since World War II. However, at least up to now the German labor market has not shown a strong reaction to the financial crisis. Given the sharp decrease in GDP the levels of employment and unemployment are still quite stable. One possible reason for the recent development is an increased level of labor hoarding, indicating that firms do not immediately adjust labor input in line with demand for their products. The paper uses both aggregate and firm-level data in order to examine the extent to which labor hoarding has contributed towards stabilizing the labor market during periods of recession. In addition, we examine the extent to which subsidized types of labor hoarding, such as short-time work, may have facilitated the retention of workers by employers. The paper shows that labor hoarding has been of certain relevance for the German labor market in times of economic slack. This is obviously true during the current crisis. Nevertheless, short-time work has also been used by firms which were not suffering significantly from an underutilization of their capacities. To avoid windfall gains the state should consider more effective targeting systems or advocate functional equivalents such as more flexibility in working time. Keywords: Labor hoarding, economic crisis, employment, public policy JEL Classification: J21, J23, J38

1. Introduction As a consequence of the global financial crisis Germany has experienced the deepest slowdown of its economy since World War II. In particular export-oriented firms in manufacturing are facing shrinking demand and therefore an underutilization of capital and labor. However, the German labor market is still quite robust. Up to now, the reduction of employment and the increase in unemployment remain * IAB-Institute for Employment Research. ** IAB-Institute for Employment Research. *** Corresponding author. IAB-Institute for Employment Research. E-mail address: ulrich. [email protected]. We thank Laura Dittrich for outstanding research assistance.

126

Martin Dietz, Michael Stops, and Ulrich Walwei

rather low. This is remarkable for two reasons. On the one hand, the labor market response in previous recessions was quicker and stronger. On the other hand, the labor market response has been weaker than in other countries. Macroeconomic figures are first indicators pointing to the possible relevance of labor hoarding at firm level. Besides the effort of firms to preserve jobs, the German government has made an important contribution which has helped to safeguard employment by making short-time work more attractive during the crisis. In our paper we shed light on different options for companies to adapt their employment level to economic shocks. In particular we focus on the phenomenon of labor hoarding, which means here that firms do not immediately alter labor input in line with a decline in their production. The paper starts by discussing companys’ reasons for this behavior, considering both the costs of labor hoarding as well as its opportunity costs. Also state intervention in the form of subsidies for short-time work can play an important role. In the third section we look for evidence of labor hoarding at the macro level of the German labor market. This part of the paper is based on aggregate data and looks at developments of labor productivity, which is used as a proxy for the utilization of labor. The data cover several business cycles as well as developments within different industries. At the firm level there is, unfortunately, a lack of data concerning the current situation. For this reason we take data from the IAB Establishment Panel for the years 2002 to 2004 in order to examine firms’ strategies of labor hoarding during that period of economic downswing. What are the characteristics of firms engaged in labor hoarding and to what extent did they use short-time work? By analyzing these data we try to gain some information concerning the portfolio of firms’ strategies for dealing with a negative shock, concerning windfall gains and the possible substitution of non-subsidized labor hoarding by subsidized short-time work. The results can be used to shape more effective short-time work schemes or functional equivalents in the future.

2. Labor Hoarding as a Response to Economic Shocks When looking at the transmission of shocks on goods markets to the labor market we typically observe some kind of a time lag. Obvious reasons are institutional features of the labor market providing workers with job security in times of uncertainty, e.g., employment protection legislation. Thus, to a certain degree labor hoarding is unavoidable on an institutionalized labor market such as we see in Germany. The length of the delayed labor market response to shocks on the goods markets depends, for example, on the rigidity of employment protection for standard work arrangements and the way more flexible (temporary) work arrangements are legislated. Indeed, we observe different labor market reactions in an international comparison. While countries with a weak employment protection system like the US and Ireland

Safeguarding Jobs through Labor Hoarding in Germany

127

or with a large proportion of fixed-term contracts like Spain show quick and strong reactions with respect to employment, in Germany the labor market response has remained fairly small with respect to the strongest decrease in GDP since World War II (Chart 1). This partly reflects the fact that German companies still face relatively strict employment protection legislation for their core staff. Germany comes third in a ranking of the OECD countries according to the strictness of their protection of permanent workers against (individual) dismissals (Venn 2009). 2

0.5 0

-0.2

-0.3 -2

-2.0

-1.6

-0.3

-1.6 -2.4

-4

-3.6 -4.2 -5.1

-6

-8

-5.0

-5.0

-4.7

-3.7

-3.6

-4.5

-7.0

-7.5 -8.9

-10 Ireland

Denmark

UK

Germany

Italy

EU

Netherlands

Spain

Austria

USA

Source: Eurostat.

Chart 1: Changes in Real GDP and in Employment in Selected Countries 2009 compared to 2008

When taking a closer look at the labor market impacts of the current crisis in Germany it is important to consider the options that companies face with respect to employment decisions. In general we can differentiate between external and internal measures for generating flexibility of labor input. With regard to external flexibility, lay-offs are an obvious response to decreases in demand. This reaction is restricted by employment protection legislation. Due to the strict legislation concerning standard work arrangements, atypical forms of employment such as marginal part-time work and agency work have gained in importance in Germany during recent years. Especially the hiring of agency workers serves as an instrument to react quickly to short-term fluctuations in labor demand.1 Indeed we observe that 1 There is also increasing use of fixed-term contracts, which often serve as a prolonged probationary period, especially for younger people. However, the use of fixed-term workers is not as flexible as hiring agency workers. Hohendanner (2010) shows for Germany that the number of fixed-term contracts decreased only moderately from 2008 to 2009 as they are often used in businesses not affected strongly by the crisis.

128

Martin Dietz, Michael Stops, and Ulrich Walwei

agency work has suffered heavily from the crisis. The numbers decreased sharply by almost 40 percent from a maximum of 820,000 in July 2008 to around 580,000 in April 2009 (Statistik der Bundesagentur für Arbeit 2010). So agency work may also be interpreted as a buffer protecting the core of a company’s workforce from short-run fluctuations. However, institutional rigidities and partial adjustments through temporary work cannot fully explain the robustness of the German labor market during the current recession. Firms have also increased internal flexibility to a large extent in order to keep their employees. So the question is: What are the possible reasons for firms preserving jobs despite facing a clear drop in labor demand? We need to take a closer look at the issue of labor hoarding, its definition, the rationale behind it and possible intervention by the state. A straightforward definition of labor hoarding is that current employment is higher than required employment (Blankart 1973). The problem with this definition, however, is that the amount of required employment is hard to evaluate from an outsider’s perspective. So the firm’s adjustment of labor input might be a better way to approach the phenomenon of labor hoarding: we observe labor hoarding when firms do not immediately adjust labor input in line with demand for their products and in particular allow their utilization of labor to vary over the business cycle (Taylor 1982). This may result in a reduction of hours worked and a reduction of work intensity, or to lower productivity per head and per working hour (Bosworth / Westaway 1990). In line with this, Hamermesh (1993) defines labor hoarding as a less than proportionate decrease in total hours worked in response to a negative demand shock. What is the rationale behind labor hoarding? From a perspective of profit maximization in a world with homogeneous labor and without transaction costs such behavior can hardly be explained. Under these circumstances, an underutilization of labor means that companies stand to the right of their labor demand curve, so they fail to minimize costs at a given production volume. Competition would cause these companies to disappear from the markets. However, real-life markets are much more complex and companies’ decisions on hiring and firing are made on a broader basis than just weighing up actual wage rates and marginal productivity (de Koning 1989). This is rational because a pure strategy of external flexibility implies various transaction costs (Oi 1962, Brechling 1965, Bowers / Deaton / Turk 1982, Horning 1994). In addition, side effects on the work attitudes of the staff have to be considered and can be regarded as indirect costs of labor hoarding (Okun 1981). All in all, the company has to examine the entire opportunity costs to reach a decision on labor hoarding. Let us briefly summarize some important factors that impose costs on lay-offs, thus making labor hoarding relatively attractive. We have already mentioned costs caused by formal regulations concerning employment protection or redundancy payments. But there are also several costs that are not related to state intervention.

Safeguarding Jobs through Labor Hoarding in Germany

129

Transaction costs play an important role on imperfect markets. Firms will face search costs when demand for their goods rises again and they need to hire new employees. In addition, costs will be incurred for setting up the labor contract and conducting job training (Bentolila / Bertola 1990). Of course the impact of transaction costs will vary according to several factors. First, the overall labor market situation will be important. When there has been a period of difficulties with respect to the supply of labor in specific sectors in the past, firms will expect long periods of vacancies and thus high recruitment costs. Under such circumstances labor hoarding becomes plausible up to a certain degree. In particular with regard to skilled workers there are incentives for labor hoarding because these are the workers that will be needed urgently during the (next) upswing (James / Thomas 1998). Besides these market-related aspects, there are also benefits of labor hoarding resulting from the employment relationship itself. If tasks are highly firm-specific the termination of an employment contract is accompanied by a loss of important firm-specific knowledge which will be time consuming and costly to regain (Oi 1962, Becker 1975, Williamson et al. 1975). Again, this will be more relevant for high-skilled workers with complex tasks than for low-skilled people engaged in routine jobs (Kruppe / Mosley 1996). Moreover, labor contracts are incomplete by nature and informal rules or social norms play an important role for the effectiveness of labor relations. This becomes more and more important in knowledge societies where tasks are complex and input-output relations are difficult to monitor. Here, principal-agent problems are difficult to handle by means of financial incentives or sanctions, and trust and reciprocity are important ways of enhancing work effort (Akerlof 1982, Fehr / Gächter 2000). These informal mechanisms need a long-term relationship to evolve and become stable (Buttler / Walwei 1993, Marsden 1995). If the employment contract becomes relational (Macneil 1974) which means that the labor relation has a value as such, the costs of lay-offs will increase further, resulting in an irreversible loss of firm-specific and relational capital. This not only affects the specific employment contract—the loss of trust may spill over and destroy reciprocal behavior on a broader basis among the remaining staff. In addition, the company’s reputation as a reliable employer is weakened, which might induce quitting and hamper future recruitment (Okun 1981). In contrast, labor hoarding is an approach that is complementary to norms of reciprocity and trust, because it signals that firms are willing to invest in safeguarding a stable employment relationship. In doing so, a high level of work motivation can be sustained. There are thus several reasons for labor hoarding that can be derived from an institutional perspective as well as from market reasons and the logic of modern employment relationships. These benefits of labor hoarding have to be weighed against its costs, which are associated with lower labor productivity and a corresponding increase in unit labor costs. So the price for labor hoarding is a decreas-

130

Martin Dietz, Michael Stops, and Ulrich Walwei

ing competitive position of the firm on the goods market in the short run. An important factor for this consideration is how long the firm expects the decline in demand to last. Labor hoarding will be more favorable when companies are facing only short-term declines in production, while expectations of a deep and long-lasting decline increases the likelihood of lay-offs and makes transaction cost arguments less important: labor hoarding tends to become inefficient. Output uncertainty is crucial in this respect and can not really be avoided (de Koning 1989). If firms adjust their labor input to the levels of sales volume that may be achieved under favorable conditions, they run the risk of being left with unneeded workers. If they decide to reduce staff, they may be short of workers afterwards. Finally, labor hoarding can also be encouraged by state intervention. One wellknown option is to subsidize working time reductions via short-time work. Typically the unemployment insurance scheme partly compensates for workers’ income losses and the costs of labor hoarding for firms. The benefits of short-time work schemes for firms are obviously similar to non-subsidized labor hoarding. The additional advantage is that labor costs are reduced immediately. The magnitude of savings on the part of the firm depends on specific features of the short-time work scheme. Important issues in this respect are the level and the duration of benefits to compensate for some of the wage losses and possible direct subsidies, e.g., concerning social security contributions. Germany is one of the countries with a long tradition of short-time work. At the moment there are three types:  Cyclical short-time work (“Konjunkturelle Kurzarbeit”) in case of a temporary, unavoidable loss of employment due to economic factors or to an unavoidable event.  Transitional short-time work (“Transfer-Kurzarbeit”) in case of a permanent loss of employment due to re-structuring measures on the establishment level.  Seasonal short-time work (“Saison-Kurzarbeit”) in case of non-productive times due to weather conditions.

In particular during the course of the current crisis, conditions for cyclical shorttime work have been made more attractive to firms by extending the maximum duration, giving financial support for training schemes and reducing social security contributions. Table 1 offers information about requirements concerning cyclical short-time work, which is by far the most important type of short-time work. However, the costs of subsidized working time reductions are not only borne by the state with respect to wage-related tax losses and the unemployment insurance system with respect to benefit payments. Employees accept a reduction of their net income. Firms have to deal with a lower capital utilization and they have to bear the remaining costs such as social security contributions for the reduced working hours or holiday pay. For the case of Germany, Bach et al. (2009) estimate that the remaining costs for firms amount to up to 35 percent of the usual labor costs.

Safeguarding Jobs through Labor Hoarding in Germany

131

Table 1 Requirements for the Use of Cyclical Short-time Work and Recent Changes Requirements for use Significant unavoidable loss of work

Pre-crisis regulations (since 1997) Temporary loss of work due to economic reasons. Other options of internal flexibility need to be utilized.

At least one third of the staff must be affected.

Estimated loss of income for the entire staff of at least 10 p.c. Requirements At least one regularly employed jobholder. to the establishment Good chance that firm returns to regular working hours.

Recent Changes No or minor changes January 2009: Under certain conditions working time accounts need not necessarily be reduced. February 2009 until March 2012: Compensation can be granted even if the firm is not able to provide full employment to at least one jobholder. No or minor changes No or minor changes No or minor changes

Individual requirements

Employment contract need to be No or minor changes maintained. No or minor changes Short-time worker is obliged to accept job offers from Federal Employment Agency. Excluded are recipients of unemNo or minor changes ployment benefits or subsistence allowances while participating in publicly financed training measures.

Notification

Employers or works councils are obliged to notify the estimated loss of work to the local employment agency.

No or minor changes

Contributions are reduced to 80 p.c. No or minor changes Social security for the loss of working hours. contributions Employers have to bear the full February 2009: amount. Federal Employment Agency covers 50 p.c. of the contributions for the loss of work. In case of training during the loss of work, the agency covers 100 p.c. of the contributions. Costs of training measures can be reimbursed. July 2009 until March 2012: Agency covers 100 p.c. of the contributions for the loss of work from the seventh month of short-time work. (Continued next page)

132

Martin Dietz, Michael Stops, and Ulrich Walwei

(Continued Table 1) Requirements for use

Pre-crisis regulations (since 1997)

Recent Changes

Duration of benefits

Maximum duration of six months.

January 2007 until June 2007: 15 months July 2007 until December 2008: 18 months January 2009 until December 2009: 24 months January 2010 until December 2010: 18 months

Level of benefit

60 p.c. of the net wage loss due to shorter hours. 67 p.c. with at least one dependent child.

No or minor changes No or minor changes

Source: Social Code III, as amended from time to time.

Given output uncertainty, the option of short-time work creates an incentive for firms to run the risk of labor hoarding. Nevertheless, a firm can only benefit from the costly retention of workers if the reduction in working hours is actually temporary and if the affected workers do not leave the firm voluntarily. We also need to assume that laid-off workers may not return to their employer even in the absence of a short-time work scheme. Such “recalls” are used for significant shares of workers in countries where temporary lay-offs are feasible (Kruppe / Mosley 1996). In the next chapter we take a closer look at macroeconomic figures in order to gain more evidence as to whether labor hoarding has played a significant role in times of recession in Germany and particularly during the current crisis.

3. The Impact of Labor Hoarding and Short-time Work from a Macro Perspective Given the sharp decrease in GDP in the fourth quarter of 2008 and the first quarter of 2009 the level of employment is still quite stable. Long-term time series show that employment and total hours worked are less volatile than output (see Chart 2), while employment measured in total hours worked is more volatile than employment measured in heads. This can also be observed by taking a closer look at the consequences of the financial crisis and is in line with our general considerations about labor hoarding. The current development is remarkable because the negative labor market responses to previous recessions were stronger and quicker. There is also evidence that labor hoarding in terms of a lower utilization of labor input has played a role in safeguarding jobs. The utilization of labor input can be measured either in terms of labor productivity per person or of labor productivity

Safeguarding Jobs through Labor Hoarding in Germany

133

2

1

0

-1

-2

-3

-4

-5 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Source: Federal Statistical Office, IAB, own calculations.

Chart 2: GDP Growth, Emploment and Total Hours Worked 1st Quarter 1991 – 4th Quarter 2009 percentage changes compared to the previous quarter, seasonally adjusted

per hour (see Chart 3). The pro-cyclical behavior of labor productivity per person reflects the observation that employment adjusts less than output over the cycle. One reason might be that workers produce less in times of recession and more in times of economic boom (de Koning 1989, Franz 2006). Labor productivity per hour has been higher than its per-person counterpart for most of the period. An important explanation for this is the persistent downward trend in total hours worked. As a consequence of the financial crisis both labor productivity per person and labor productivity per hour have decreased significantly to levels which have never been observed before in Germany. In order to assess the impact that the financial crisis is having on the labor market, and particularly the role of labor hoarding, we need to ask how far below its long-run level labor utilization is. In this respect labor productivity can be used as a proxy for labor utilization. However, a linear trend is not necessarily a good measure of the long-run utilization of the workforce (Felices 2003). A straight trend line might not properly represent the long-run behavior of labor productivity. We assume that a time-varying trend is a more suitable measure for changes in labor productivity over time. A Hodrick-Prescott filter provides an estimation of such a time-varying trend (Hodrick / Prescott 1997). Using this procedure, Charts 4 and 5 illustrate whether labor utilization is above or below its long-run level. In Charts

134

Martin Dietz, Michael Stops, and Ulrich Walwei

4a und 4b, labor productivity, measured as output per head or per hour, varies around an upward-sloping trend that could reflect its long-run equilibrium. If we assume this to be the case, then labor productivity below this line would indicate labor utilization below its long-run trend and would therefore reflect the possibility of labor hoarding. Conversely, labor productivity above the line reflects a more intensive labor utilization. 3

2

1

0

-1

-2

-3

-4

-5

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Source: Federal Statistical Office, IAB, own calculations.

Chart 3: GDP Growth and Labor Productivity 1st Quarter 1991 – 4th Quarter 2009 percentage changes compared to the previous quarter, seasonally adjusted

Charts 5a and 5b show the deviations between labor productivity and its fitted trend. Positive differences from the trend representing values above the zero line indicate an increasing labor intensity or labor hoarding below its long-run trend. Similarly, labor hoarding in terms of an underutilization of the workforce occurs when negative differences from the trend can be observed. Since reunification this has happened several times (e.g., quarters I – IV, 1993; quarters II – III, 1994; quarter I, 2004, to quarter IV, 2005) but there has never been a negative deviation of labor utilization from the long-term trend at the level which has been reached during the current crisis.

Safeguarding Jobs through Labor Hoarding in Germany

135

120 115 110 105 100 95 90 85 80 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Source: Federal Statistical Office, IAB, own calculations.

Chart 4a: Labor Productivity per Hour: Actual and Trend* 1st Quarter 1991 – 4th Quarter 2009 Index 2000 = 100, seasonally adjusted * Trend calculated using a Hodrick-Prescott filter with the smoothing parameter  = 1600. 120 115 110

105 100

95

90 85

80

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Source: Federal Statistical Office, IAB, own calculations.

Chart 4b: Labor Productivity per Head: Actual and Trend* 1st Quarter 1991 – 4th Quarter 2009 Index 2000 = 100, seasonally adjusted * Trend calculated using a Hodrick-Prescott filter with the smoothing parameter  = 1600.

136

Martin Dietz, Michael Stops, and Ulrich Walwei

6

4

2

0

-2

-4

-6 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Source: Federal Statistical Office, IAB, own calculations.

Chart 5a: Labor Utilization Based on Labor Productivity per Hour 1st Quarter 1991 – 4th Quarter 2009 cyclical component*, seasonally adjusted * Cyclical component calculated using a Hodrick-Prescott filter with the smoothing parameter  = 1600.

6

4

2

0

-2

-4

-6

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Source: Federal Statistical Office, IAB, own calculations.

Chart 5b: Labor Utilization based on Labor Productivity per Capita 1st Quarter 1991 – 4th Quarter 2009 cyclical component*, seasonally adjusted * Cyclical component calculated using a Hodrick-Prescott filter with the smoothing parameter  = 1600.

Safeguarding Jobs through Labor Hoarding in Germany

137

Labor utilization may differ significantly between industries. Using the same procedure as for the economy as a whole the analyses reveal that particularly during the current crisis we can identify industries with either a higher or a lower level of labor being hoarded. We distinguish between three sectors (primary, secondary and tertiary sector) and five industries: manufacturing; construction; trade, hotels, restaurants and transport; finance, housing, business-related services; public and private services. The results reveal that the secondary and tertiary sectors and within these sectors particularly industries like manufacturing as well as trade, hotels, restaurants and transport have recently shown large negative deviations of labor utilization from the long-term trend, also compared to the maximum negative deviations in the past (see Tables 2a and 2b). We can therefore assume that especially firms in these industries may have used labor hoarding in order to retain workers. Table 2a Cyclical Component of Labor Productivity per Capita* by Industry

Primary Sector Secondary Sector Manufacturing Construction Tertiary Sector Trade, Hotels, Restaurants and Transport Finance, Housing and Business-Related Services Public and Private Services All Sectors

2008 QI

2008 Q II

2008 Q III

2008 Q IV

2009 QI

2009 Q II

2009 Q III

2009 Q IV

4.1 2.8 5.8 –3.6 0.1

–0.1 16.4 10.5 5.2 6.3

–1.6 13.9 5.1 8.1 7.9

–2.4 –5.3 –1.5 –4.6 –2.1

4.6 –25.2 –15.7 –9.5 –8.8

–0.4 –11.8 –13.6 1.8 –3.4

–1.8 1.1 –7.7 8.8 1.0

–0.6 –5.0 –2.6 –2.4 –4.8

–1.1

4.8

4.9

–1.0

–7.3

–2.7

–0.4

–3.4

0.7

0.2

0.7

–1.7

–0.2

–0.2

0.8

–1.0

0.1 1.8

1.0 3.3

1.8 3.0

0.0 –0.8

–1.4 –5.0

–0.5 –3.8

0.5 –1.3

–0.5 –1.5

* Cyclical component calculated using a Hodrick-Prescott filter with the smoothing parameter  = 1600. Source: Federal Statistical Office, IAB, own calculations.

As stated earlier there are several explanations for labor hoarding. Amongst other factors the decision not to dismiss core workers may also reflect recent skill shortages in those regions or industries hit by the crisis. Indeed the German economy, particularly manufacturing companies, suffered from a lack of highly skilled labor during the last economic upswing. So the companies are aware of an upcoming recruitment problem, expect increasing hiring costs and thus may be willing to preserve jobs in these sectors. This is even more important because the crisis is especially affecting regions and sectors which have exhibited very good labor market performance during the past years: highly competitive and export-oriented

138

Martin Dietz, Michael Stops, and Ulrich Walwei

manufacturing industries (Möller 2009). These companies are often situated in prospering regions especially in the south-west of Germany, where one has observed almost full employment during recent years. There, firms had made a considerable profit in the last upswing and are thus able to bear the costs of labor hoarding at least for a limited period of time without danger of bankruptcy. Table 2b Maximum Negative Deviation of Labor Utilization from the Long-term Trend* (1st Quarter 1991 – 4th Quarter 2007) All Sectors Primary Sector Secondary Sector Manufacturing Construction Tertiary Sector Trade, Hotels, Restaurants and Transport Finance, Housing and Business-Related Services Public and Private Services All Sectors

2003 Q III: 1996 Q I: 2002 Q I: 1996 Q I: 1994 Q I: 2005 Q I: 2000 Q IV: 1997 Q I: 2005 Q I:

–14.6 –20.0 –5.4 –16.9 –8.1 –5.4 –3.5 –2.3 –2.8

0.8 –2.1 –2.1 –2.1 –1.3 –2.8 –0.2 –2.6

* Cyclical component calculated using a Hodrick-Prescott filter with the smoothing parameter  = 1600. Source: Federal Statistical Office, IAB, own calculations.

Changes in labor productivity used as a proxy for labor utilization cover variations of labor hoarding to a certain extent. But there are drawbacks which need to be mentioned. Interpreting changes in labor productivity per head or per hour over time as a more or less intensive use of labor input underestimates the changing influence of non-cyclical movements in labor utilization. Labor productivity is also affected by other factors of production, e.g., cyclical movements in total factor productivity, changes in the capital stock and its utilization, as well as varying returns associated with different inputs (Basu / Fernald 2000). Variations in labor productivity might also be affected by the measurement of GDP. A reliable measurement of labor productivity implies being able to measure labor input properly. Aggregate output includes only regular production or marketable output and more or less neglects other meaningful work which might be important for future production, such as painting the factory, machine maintenance or internal training. This kind of work which does not normally contribute to aggregate output can be of particular importance during recessions. In some industries more labor input may be needed to sell the same amount of services, for example in logistics, or it may take the same amount of labor to serve a smaller number of people, e.g., in restaurants, theatres or cinemas. Labor productivity could therefore also decrease due to the specificities of certain tasks as well as the problems in output measurement, and

Safeguarding Jobs through Labor Hoarding in Germany

139

not solely because of firms hoarding labor. In addition, there are problems in estimating the level of productivity per hour and per head properly. This is due to the fact that the measurement of working hours is to some extent imprecise. Because certain elements of total hours worked, such as unpaid overtime, are underreported in administrative data and surveys, cyclical variations in working hours might be underestimated in the available data. This would imply that labor productivity per hour is overestimated in recessions and underestimated in boom periods. Nevertheless, using labor productivity as a proxy for labor utilization reveals that labor hoarding played a not inconsiderable role during previous recessions and is doing so in the current crisis, too. This is particularly true in certain industries such as manufacturing. All in all, we can recapitulate that the specific situation concerning the last economic upswing and the incidence of the current crisis are favorable for labor hoarding in Germany. Because the percentage deviation of labor productivity per head from the long-term trend is much higher than its per-hour counterpart, the reduction of working hours has obviously been an additional significant factor for safeguarding jobs. Given the different costs of external flexibility, firms face incentives to increase internal work flexibility by varying working hours or the intensity of labor input over the business cycle. Indeed companies have increasingly created internal schemes in order to react more flexibly with respect to short-run fluctuations, e.g., the use of working-time accounts has become more and more important (Bellmann / Gerner 2010). Working-time accounts allow for increasing working hours during times of high demand—these hours saved by the employees act as a buffer stock which can be spent to reduce working time during periods of lower activity. In doing so, real labor costs are shifted from a period of high production activity to a period of low production activity. This guarantees a stable income for workers facing fluctuations in their working hours. Working-time accounts come to a natural end when they are empty. In addition, so-called alliances for jobs at the company level or collective agreements aim at safeguarding jobs by allowing a joint reduction of working time and pay (Bellmann / Gerner 2010). The decomposition of the total change in working hours of salaried workers from 2008 to 2009 by the Institute for Employment Research (IAB) indicates that four components were especially relevant. Shorter weekly working hours (minus 18.4 hours a year) were of the greatest importance. Collective agreements and firm-specific alliances for jobs allow longer working hours during peak periods of demand and shorter hours during recessions on condition that the contracted working time is maintained on average over a defined period. The second most important component was the strong use of short-time work, which amounts to minus 13.4 hours a year. A reduction of paid overtime (minus 7.9 hours a year) and a phasing down of working-time accounts (minus 7.9 hours a year) additionally contributed to the reduction of average working hours in 2009 (see Chart 6). The decomposition suggests that subsidized labor hoarding such as short-time work certainly helps to explain the strong reduction of working hours but it is not the major explanation.

140

Martin Dietz, Michael Stops, and Ulrich Walwei

10

4.9

0 -10

-7.9

-7.9

-13.4 -20

-18.4

-30 -40 -42.9 -50 Weekly Working Sickness, Short-time Work Hours Holidays, other Components

Overtime

Working Hour Accounts

Total

Source: IAB.

Chart 6: Changes in Working Hours per Year in Germany, 2009 yearly averages, changes in hours

The calculated impact of short-time work on working hours per year is based on the use of the scheme. In 2009 an average of about 1.1 million people worked short-time. This was the highest number since the early 1990s, when short-time work was used as an instrument to dampen the structural consequences of German unification. Short-time work2 is currently being used most intensely in manufacturing and in western Germany, in highly competitive sectors and thus serves more as an instrument to counter cyclical movements and not structural changes. So in the current crisis firms are using subsidized short-time work as a means to partly finance labor hoarding and bridge the crisis, making their workers return to their regular jobs. By contrast it was used as an additional instrument of the social security network following German unification in the 1990s. Based on an average loss of 36 percent of the working time, the amount of shorttime work corresponds to about 340,000 full-time employees in 2009 (own calculations based on IAB 2010). As some of the short-time workers are employed on a part-time basis, the calculated effect of short-time work on safeguarding employment is likely to be slightly higher. Full-time equivalents are often put on a level with their impact on the labor market. But this equation overestimates the incremental employment effects of the program. It fails to take windfall gains into consideration. A certain amount of displacement can take place, as both viable and non-viable firms are supported during the crisis. In addition, it also assumes that the only alternative to short-time work is non-employment and disregards the possibility that in some cases dismissals may not be avoided by means of short-time work but only postponed (Kruppe / Mosley 1996). 2 Please note: we are not able to distinguish between the three types of short-time work with the available data.

Safeguarding Jobs through Labor Hoarding in Germany

141

Two crucial issues for an assessment of short-time work are the extent to which it could have been substituted by non-subsidized types of labor hoarding and the relevance of displacement effects. These issues can, of course, not be answered using macro data. In the following section we therefore use micro data and take a closer look at subsidized and non-subsidized labor hoarding at the firm level.

4. Labor Hoarding and Short-time Work at the Firm Level To investigate the extent to which firms suffering from a recession hoard labor, we use data from the IAB Establishment Panel for the period from 2002 to 2004 (see also Box 1). The development of sales volume and the employment of firms during a period of economic downturn are of special interest in this respect. We took these data for two reasons. Firstly, we wanted to come as close as possible to the last recession, which took place from 2002 to 2003 (Räth 2009). For obvious reasons firm data until 2010 / 2011 are not available at present. Secondly, we will later add some further information about establishments using short-time work and those not doing so. This information is available for the 2003 survey period of the IAB Establishment Panel.

The IAB (Institute for Employment Research) Establishment Panel is an annual survey of establishments. It represents all industries and establishment sizes nationwide and can also be analyzed on a longitudinal basis. The survey was initiated in western Germany in 1993. Besides firms’ structure information, such as economic activity, it contains information about sales and intermediate inputs per year, employment levels as of June of each year and information about short-time work for the year 2003 (Fischer et al. 2009). Box 1: The IAB Establishment Panel

First we have to identify establishments with a considerable decrease in sales volume. We use the following definition: a considerable decrease in sales volume is given when the firm would have been able to keep the sales volume per employee constant in 2002 by laying off at least one employee.3 This group consists of around 17 percent of all establishments with about 34 percent of all employees. We then divided these establishments into two sub-groups: the first group maintained or even increased their employment levels, whereas the second group decreased the number of employees between June 2002 and June 20044 (see Table 3). A good 3 This definition is, of course, debatable. It obviously has different implications for large and small establishments. We thus conducted a sensitivity analysis using an alternative definition: a reduction was regarded as considerable if the decrease in sales volume exceeded 10 per cent. The results do not show significant differences from the analysis presented in the paper.

142

Martin Dietz, Michael Stops, and Ulrich Walwei

half of the establishments suffering from a considerable reduction in their sales volume increased their number of employees or kept it constant. Because of decreasing sales from 2002 to 2003 and the either constant or increasing level of employment during the period June 2002 to June 2004, these firms are most likely to have been hoarding labor because their output-labor ratio measured in output per employee decreased.5 For the other sub-group of establishments facing a considerable decline in sales volume the employment decisions point in the same direction—the number of employees fell. Table 3 Changes in Employment in Establishments with a Considerable Decrease in Sales Volume Establishments with a considerable decrease in sales volume 2002 – 2003 (17% + / –1.5% of all establishments with 34 % of all employees in 2002)* Share of the number of establishments (share of all employees) in this group Labor hoarding June 2002 – June 2004 53% + / –3.2 % (40 %)

Decline in employment June 2002 – June 2004 47% + / –3.2% (60%)

* Notes: (1) 42 % + / –1.8 % of the establishments in the sample had slightly decreasing, constant or increasing sales. For 41 % + / –1.8 % of all establishments in the sample there is no information about changes in sales volume, short-time work, and / or changes in employment. (2) Weighted shares with an estimated 95 % confidence interval for the number of establishments (taken from Fischer et al. 2008). Source: IAB Establishment Panel, own calculations.

If we look at different industries we see some variation. In the tertiary sector (services) the share of establishments with considerably decreasing sales which showed signs of labor hoarding is higher. In the secondary sector (manufacturing and construction) the opposite is the case. A larger number of establishments with decreases in sales reduced their employment.6 4 We took the level of employment in June 2002 as a proxy for the starting point of the observation at the beginning of 2002 and the level of employment in June 2004 for the final point of the observation at the end of 2003. 5 While we considered the volume of work in our previous definition of labor hoarding we have to refer to the number of employees here, because the data do not contain sufficient information about the hours worked. 6 Weighted shares and 95 % confidence intervals for the sectors are: 57 % + / –4.4% of all establishments in the tertiary sector tended to hoard labor. 57% + / –5.3% of all establishments in the secondary sector reduced the number of their employees. Due to the large confidence interval there is no clear result for the employment changes in firms in the primary sector which suffered from decreases in sales volume: 55 % + / –14% of the establishments tended to hoard labor.

Safeguarding Jobs through Labor Hoarding in Germany

143

Additionally, we can show that small establishments with a considerable drop in sales volume are more likely to maintain or increase the number of employees (Table 4). The data also allow us to consider the development of shares of part-time workers, temporary workers, skilled and unskilled workers as proportions of the total workforce. However, further analyses gave no clear evidence as to how establishments with decreasing sales make use of labor hoarding. The recession in 2002 / 2003 was not as pronounced as the current one. The results for the current recession would probably be more definite. Additionally, to gain more information about companies which hoard labor, future analyses could link data containing details on employees. Table 4 Changes in Employment in Establishments with a Considerable Decrease in Sales Volume by Size Class Establishments with a considerable decrease in sales volume 2002 – 2003 Labor hoarding and decline in employment by size class (Share of all establishments in the corresponding size class) Size class 1 – 9 employees 10 – 49 employees 50 – 249 employees 250 or more employees

Labor hoarding

Decline in employment

63% (+ / –8 %) 45% (+ / –5.7%) 36% (+ / –6.2%) 33% (+ / –6.7%)

37% (+ / –8 %) 55% (+ / –5.7%) 64% (+ / –6.2%) 67% (+ / –6.7%)

Note: Weighted shares with estimated 95 % confidence intervals for the number of establishments (taken from Fischer et al. (2008, p. 37). Source: IAB Establishment Panel, own calculations.

In a subsequent step we combined our results with information about the use of short-time work in the first six months of 2003. An important goal of the shorttime work scheme in Germany is to dampen undesirable effects of recessions on employment. But what firms do make use of short-time work? Can we assume that they are exclusively firms that are suffering from a recession or do we also observe firms with increasing sales? There is not much information on this concerning the period of recession from 2002 to 2003 (Deeke 2005). Crimmann / Wießner (2009) found a negative correlation between qualitative assessments by firms concerning their revenues in 2002 and the probability of using short-time work, and a positive correlation between negative business expectations for 2003 and the probability of using short-time work. We are able to extend this view by examining how the employ-

144

Martin Dietz, Michael Stops, and Ulrich Walwei

ment levels and the sales volume of firms using short-time work have developed.7 Nevertheless, the share of establishments using short-time work was quite small in the first six months of 2003, at only 2%. For this reason the following results should be interpreted with caution. With regard to the industries it can be shown that in the first six months of 2003 short-time work was mostly used by the secondary sector (Table 5), although the tertiary sector comprises the largest number of establishments and employees. Table 5 Use of Short-time Work in the First Six Months of 2003 by Sector All sectors

Primary sector

Secondary sector

Tertiary sector

Yes (2 %) No (98%)

1% (–)* 4 % (+ / –1.0%)

55% (+ / –9.9 %) 21% (+ / –1.9 %)

44% (+ / –9.9 %) 75% (+ / –1.9 %)

* Number of establishments observed is too small. Note: Weighted shares with estimated 95 % confidence intervals for the number of establishments (taken from Fischer et al. (2008, p. 37). Source: IAB Establishment Panel, own calculations.

An interesting question, however, is to what extent companies with varying sales volume trends have made use of short-term work. In this respect Table 6 displays two findings. On the one hand establishments which were not suffering from considerable losses of sales volume and used short-time work were more likely to have a steady or increasing level of employment afterwards than those which showed a decline in employment between June 2002 and June 2004. This does not rule out the possibility that some of these firms were suffering from economic difficulties, so the use of short-time work may be justified. But especially in the case of increasing sales one has to ask whether the firm itself should be responsible for the necessary adjustments and not the state. On the other hand most of the establishments using short-time work and suffering from a considerable decrease in sales volume had reduced their workforce. However, the latter finding may suggest, but does not necessarily mean, that short-time work has had no impact at all: without public aid there may have been even more lay-offs. To investigate this issue, however, a counterfactual situation would be needed implying that short-time work had not been used by a certain establishment. This situation is neither theoretically nor empirically constructible and therefore there is no clear evidence of whether shorttime work has dampened the decline in employment or has had no effects on its development. 7 We pay for our dynamic perspective with a (further) loss of observed establishments relative to the population. This is partly compensated by supporting panel extrapolation factors. These factors can be used for descriptive structural analysis under certain conditions, but the findings may be imprecise (Fischer et al. 2008). However, in the following they are presented for illustration purposes.

Safeguarding Jobs through Labor Hoarding in Germany

145

Table 6 Establishments with Short-time Work in the First Six Months of 2003 by Sub-group Establishments with slightly decreasing, constant, or increasing sales volume 2002 – 2003

Steady or increasing employment

33 %(+ / –9.5%)

Decline in employment

19 %(+ / –8.0%)

Establishments with a considerable decrease in sales volume 2002 – 2003

Labor hoarding Decline in employment

2% (–)* 45%(+ / –10%)

* Number of establishments observed is too small. Note: Weighted shares with estimated 95 % confidence intervals for the number of establishments (taken from Fischer et al. (2008, p. 37). Source: Differentiation on the basis of Frick (2005), data from the IAB Establishment Panel, own calculation.

We conclude that a not inconsiderable share of German establishments that were affected by the recession in 2002 / 2003 already tended to hoard labor. Unfortunately, it is not known whether short-time work stabilized employment or not. Furthermore, the use of short-time work obviously does not depend solely on the economic situation of the establishment. Therefore, also firms with increasing or at least constant sales as well as stable employment used short-time work. One possible reason for this might be that the scheme was not targeted properly. This could be avoided in two ways: (1) better conditions for firms suffering from a difficult economic situation and (2) restricted access to short-time work for firms in a more favorable economic situation. In the current recession the first aspect may already have been taken into account by subsidizing social security contributions in the case of downtime (Bach et al. 2009). But this does not exclude the possibility that recently firms in a more favorable situation have again used the instrument. For this reason, stricter regulations concerning the economic situation of a firm should be considered.

5. Conclusions As a consequence of the global financial crisis Germany has experienced the deepest slowdown of its economy since World War II. However, at least up to now the German labor market has not shown a strong reaction to the financial crises. Given the sharp decrease in GDP the levels of employment and unemployment are still quite stable. German companies continue to face stringent regulations for standard work arrangements, so atypical and more flexible work arrangements have gained importance during recent years. These work arrangements provide external flexibility at least to a certain degree and are accompanied by a set of measures aimed at internal flexibility within the core workforce. Both strategies facilitate adaptation to macroeconomic shocks and form a protective shield against job

146

Martin Dietz, Michael Stops, and Ulrich Walwei

losses among the core workforce, which is responsible for the stability of the German labor market until the beginning of 2010. One of the potential sources of internal flexibility is labor hoarding, which means that firms do not immediately adjust labor input in line with demand for their products. Our paper indicates that labor hoarding has been of certain relevance for the German labor market in times of economic slack. This is particularly true of the current crisis. But in general, labor hoarding is not a “silver bullet” at all. It has certain limitations, especially due to output uncertainty. When the crisis is deep and long, there are particular risks—a firm’s ability to bear the costs of labor hoarding declines and lay-offs become more reasonable. Other side effects of labor hoarding may include a lower level of labor turnover, a slowdown of structural change or the risk of jobless growth in the next economic upswing. Recent surveys indicate that companies which are affected by the crisis state that their most important response to a lack of demand is to reduce hiring (Heckmann et al. 2009). For outsiders the combination of labor hoarding and a reduction of recruitments may result in serious problems in entering the labor market. Indeed an analysis of labor market flows shows that there are only few entries to the labor market at the moment (Rothe 2010). This problem will probably continue to exist throughout the crisis and extend into the following economic upswing, as internal and external flexibility operate in both directions. Employment reactions are weaker both in the recession and also in the upswing: firms will first increase the intensity of labor input, engage staff in working overtime and restock the working-time accounts. In a second step they will recruit new staff, possibly on a temporary basis at the beginning. So a period of jobless growth is likely after leaving the current crisis, which bears higher risks of human capital deviation and long-term unemployment. There is also some evidence that short-time work has contributed to labor hoarding. Nevertheless, firm data covering an earlier recession show that short-time work was also used by firms which were not suffering significantly from an underutilization of their capacities. This cannot be ruled out at all in the case of the current crisis, especially with regard to the reforms leading to more generous short-time work schemes. Companies decide about labor hoarding according to a cost-benefit ratio which in general does not include the social costs of labor hoarding. Additional costs for the state may result, e.g., if short-time workers are dismissed later on anyway or if firms could have done without public aid. In addition, generous short-time work schemes bear the risk of conserving structures or rescuing companies which are no longer competitive. Because of these side effects, effective targeting measures for short-time work are important. The aim would then be to restrict access to the subsidies and spend the money on the “right” companies. In order to minimize windfall gains, schemes may therefore impose special conditions. Nevertheless, there is a problem at that point in time when the decision for or against short-time work has to be made: neither the firm nor the state know whether the firms’ negative business expectations will become true. One solution could be a combination of short-time work with a clause

Safeguarding Jobs through Labor Hoarding in Germany

147

for repayment in the case of an (unexpected) good economic situation from an expost perspective. An even more radical strategy would be to rely solely on functional equivalents such as an even more intensive use of flexible work schemes within firms. Our paper is only a starting point for further analyses. It will be necessary to take a more careful look at more recent firm data in the near future covering the period between 2009 and 2011 and examine whether there are still indications of windfall gains. The analyses could be deepened further by linking firm data with employee data in order to identify determinants of labor hoarding with a (micro-) econometric approach. References Akerlof, G. A. (1982): “Labor Contracts as Partial Gift Exchange,” Quarterly Journal of Economics, 97 (4), 543 – 569. Bach, H. U. / Spitznagel, E. (2009): “Kurzarbeit: Betriebe zahlen – und haben etwas davon”, IAB-Kurzbericht 17 / 2009. Basu, S. / Fernald, J. (2000): “Why is productivity procyclical? Why do we care?,” NBER Working Paper Series, No. 7040. Becker, G. (1975): Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. New York: Columbia University Press, NBER. Bellmann, L. / Gerner, H.-D. (2010): “Economic Crisis, Rising Unemployment and Policy Responses,” Nuremberg: Paper prepared for the IZA / OECD Workshop. Paris 8. / 9. February 2010. Bentolila, S. / Bertola, G. (1990): “Firing Costs and Labor Demand: How Bad Is Eurosclerosis?,” Review of Economic Studies, 57, 381 – 402. Blankart, B. (1973): “Arbeitskräftenachfrage im Konjunkturablauf – das Problem des temporären Hortens von Arbeitskräften”, Zeitschrift für Volkswirtschaft und Statistik, 109 (2), 171 – 185. Bosworth, D. / Westaway, T. (1990): “Labor Hoarding, and Recorded Unemployment in OECD Countries,” in: Unemployment in Europe, edited by J. Muysken and C. de Neubourg, London: Macmillan, 150 – 164. Bowers, J. / Deaton, D. / Turk, J. (1982): Labor Hoarding in British Industry, Oxford: Blackwell. Brechling, F. P. R. (1965): “The Relationship between Output and Employment in British Manufacturing Industries,” Review of Economic Studies, 32, 187 – 216. Buttler, F. / Walwei, U. (1993): “Employment Security and Efficiency: Assumptions in the Current Debate and Empirical Evidence for West Germany,” in: Employment Security and Labor Market Behavior. Interdisciplinary Approaches and International Evidence, edited by C. F. Buechtemann, Ithaca: ILR Press, 255 – 266. Crimmann, A. / Wießner, F. (2009): “Wirtschafts- und Finanzkrise: Verschnaufpause dank Kurzarbeit”, IAB-Kurzbericht, 14 / 2009.

148

Martin Dietz, Michael Stops, and Ulrich Walwei

De Koning, J. (1989): “Labor Hoarding in Dutch Manufacturing Industry,” De Economist, 137, 155 – 172. Deeke, A: (2005): “Kurzarbeit als Instrument betrieblicher Flexibilität. Ergebnisse aus dem IAB-Betriebspanel 2003”, IAB-Forschungsbericht, 12 / 2005. Fehr, E. / Gächter, S. (2000): “Fairness and Retaliation: The Economics of Reciprocity,” Journal of Economic Perspectives, 14 (3), 159 – 181. Felices, G. (2003): “Assessing the Extent of Labor Hoarding,” Bank of England Quarterly Bulletin. Fischer, G. / Janik, F. / Müller, D. / Schmucker, A. (2008): “The IAB establishment panel—from sample to survey to projection,” FDZ-Methodenreport, 01 / 2008. – (2009): “The IAB Establishment Panel— things users should know,” Schmollers Jahrbuch. Zeitschrift für Wirtschafts- und Sozialwissenschaften, 129 (1), 133 – 148. Franz, W. (2006): Arbeitsmarktökonomik, Berlin: Springer. Frick, A. / Wirz, A. (2005): “Wirksamkeit der Kurzarbeitsregelung in der Rezession 2001 – 2003”, seco Publikation – Arbeitsmarktpolitik Nr 13. Hamermesh, D. (1993): Labor Demand, Princeton: Princeton University Press. Heckmann, M. / Kettner, A. / Pausch, S. / Szameitat, J. / Vogler-Ludwig, K. (2009): “Unternehmensbefragung im II. Quartal 2009: Wie Betriebe in der Krise Beschäftigung stützen”, IAB-Kurzbericht, 18 / 2009. Hodrick, R. J. / Prescott, E. C. (1997): “Postwar US business cycles: An empirical investigation,” Journal of Money, Credit and Banking, 29(1), 1–16. Hohendanner, C. (2010): “Befristete Arbeitsverträge zwischen Auf- und Abschwung. Unsichere Zeiten, unsichere Verträge?”, IAB Kurzbericht 14 / 2010. Horning, B. (1994): “Labor Hoarding and the Business Cycle,” International Economic Review, 35, 87 – 100. IAB (2010): “Durchschnittliche Arbeitszeit und ihre Komponenten in Deutschland”, http:// doku.iab.de/grauepap/2010/tab-az09.pdf. 2010-04-02. James, J. A. / Thomas, M. (1998): “Labor Hoarding and Selective Retention of Skilled Workers in U.S. Industry in the Late Nineteenth-Century: The Workers’ Perspective,” in: The Microeconomic Analysis of the Household and the Labor Market, Proceedings of the Twelfth International Economic History Congress, edited by C. Nunez, Madrid: Fundacion Fomento de la Historia Economica. Macneil, I. (1974): “The Many Futures of Contracts,” Southern California Law Review, 47, 691 – 816. Marsden, D. (1995): “Deregulation or Cooperation? The Future of Europe’s Labor Markets,” Labor—Special Issue 1995, 67-S91. Möller, J. (2009): “The German Labor Market Response in the World Recession—De-mystifying a Miracle,” Zeitschrift für Arbeitsmarkt- und Berufsforschung, 325 – 336. Mosley, H. / Kruppe, T. (1996): “Employment Stabilisation through Short-time Work,” International Handbook of Labor Market Policy and Evaluation, 594 – 619. OECD (2009): Employment Outlook, Tackling the Job Crisis. Paris: OECD Publishing.

Safeguarding Jobs through Labor Hoarding in Germany

149

Oi, W. Y. (1962): “Labor as a Quasi-Fixed Factor of Production,” Journal of Political Economy 70, 538 – 555. Okun, A. M. (1981): Prices and Quantities: A Macroeconomic Analysis, Washington: The Brookings Institution. Räth, N. (2009): “Rezessionen in historischer Betrachtung”, Statistisches Bundesamt. Rothe, T. (2010): “Tägliche Dynamik auf dem deutschen Arbeitsmarkt”, Wirtschaftsdienst, 90, 64 – 66. Statistik der Bundesagentur für Arbeit (2010): “Beschäftigungsstatistik, Arbeitnehmerüberlassung, Berichtszeitraum. 1. Halbjahr 2009”. http://www.pub.arbeitsagentur.de/hst/services/ statistik/detail/b.html?call=l. 2010-04-07. Taylor, J. (1982): “The Theory and Measurement of Labor Hoarding,” Applied Economics, 14, 591 – 601. Venn, D. (2009): “Legislation, Collective Bargaining and Enforcement: Updating the OECD employment protection indicators,” OECD Working Paper No. 89. Williamson, O. E. / Wachter, M. / Harris, J. (1975): “Understanding the Employment Relation: The Analysis of Idiosyncratic Exchange,” The Bell Journal of Economics, 6, 250 – 278.

Safeguarding Jobs through Labor Hoarding in Germany Comment By Olaf Hübler* 1. Introduction Labor hoarding contributes to explain why during the current crisis Germany has experienced only a slight decline in employment levels. Employers may seek to minimize short-run costs by reducing the demand for labor at the intensive margin (Bell / Blanchflower 2010). However, there is little empirical evidence to support this hypothesis. One exception is the study undertaken by Dietz, Stops and Walwei (2010), who reached three conclusions: (1) many firms in Germany have hoarded labor during the last two years despite a sharp decline in demand; (2) short-time work has evidently contributed to labor hoarding; and (3) short-time work, generating as it does windfall gains, is an inefficient instrument. Therefore, they suggest a modified arrangement. Their empirical studies are based on several macroeconomic time series from 1991 to 2009 that permit them to gauge the possible effects of labor hoarding. Additionally, microeconomic analyses, in which labor hoarding is defined somewhat differently, are executed, based on firms’ level data from 2002 – 2003. The results from these two years cannot be applied to the current crisis, as Crimmann, Wießner and Bellmann (2010) show in their comparison. Labor hoarding is a more general category than short-time work (which has a legal definition), but there is no consensus on what exactly it is. As a theoretical construct it does not fit naturally into empirical investigations. Taylor’s construct (Taylor 1982) differs from the operationalizations used by Dietz, Stops and Walwei (2010). Thus a certain ambiguity persists in the term itself. In the next session I highlight some arguments for and against labor hoarding and short-time work. Then I move on to problems with the empirical determination of labor hoarding. Finally, Dietz, Stops and Walwei’s paper is supplemented and my attention is focused on short-time work. Empirical evidence from the literature and some new results are presented. * Institute of Empirical Economic Research, Leibniz University of Hannover, Hannover, Germany. E-mail address: [email protected].

152

Olaf Hübler

2. Arguments for and against Labor Hoarding and Short-time Work Why labor hoarding and especially short-time work are intensively used in Germany and a few other countries but not widely and to the same degree remains an open question. Assessments of these instruments are chiefly at odds regarding their possible disadvantages and the relative value of other instruments. Labor hoarding, used when triggered by a decrease in demand for products, is a very flexible instrument, permitting an instantaneous positive adjustment of labor if the demand subsequently increases. It involves no layoff or rehiring costs, and in the case of short-time work it means a reduction in hoarding costs. Furthermore, there is less loss of human capital then there would be in the case of a standard layoff. Since short-time workers foresee the risk of an imminent period of unemployment if they do not start to look for work before they are laid off, they are motivated to do so, whereas the unemployed have little chance of re-entering the work force. Since workers under the labor-hoarding system work relatively few hours, they have more time than others in which to train for another, perhaps better job. The German government supports the use of training programs in conjunction with short-time work. It is a win-win arrangement: the employers with their improved work force and the employees with their lowered risk of joining the longterm unemployed. There are two possible problems with labor hoarding: that it obliges a shift from variable to fixed labor costs, and that it threatens to hinder structural change. Because labor hoarding retards needed adjustments to the labor force, it impedes the smooth functioning of the market mechanism and thereby generates undesirable windfall gains and substitution processes. For example, qualified work is hoarded and unqualified work is displaced to an excessive degree. A firm’s assessment of costs and benefits of its own use of short-time workers cannot be considered objective and reliable. Many factors have to be considered; for instance, with a prolongation of the period in which short-time allowance is paid there may come the need for yet more layoffs, and ultimately bankruptcy. At that point, all hoarding costs are become sunk costs. If the government or the employer does not bear the full burden of these costs, then the employees will have to accept wage cuts and this may have, in turn, a negative impact on, for example, rents and unemployment benefits. Any assessment of the advantages and disadvantages of labor hoarding is a question of perspective. The various parties involved—the employer, the employees and the government—may be focusing on short-term or long-term issues, or weighting labor hoarding against other options. In addition, we must not equate the economic consequences for a single firm with those for the entire nation. The degree to which workers are qualified determines the degree to which labor hoarding affects them, since firms tend to hoard qualified workers, so as to avoid the need to try to rehire them come the next business boom. Whether a worker’s sex and age deter-

Safeguarding Jobs through Labor Hoarding in Germany—Comment

153

mine how he or she handles an increase in his or her free time needs to be determined. For example, many women may consider it an opportunity to better care of their children, while it may leave many men feeling at loose ends and demoralized; older workers may look upon shorter hours as a sort of pre-retirement, whereas those at the start of their careers may miss the more intense rhythm of permanent everyday work. Analyses have yet to be made in order to determine the comparative efficacy of options other than short-time work: for instance, in-plant offers of part-time work, a longer work-week, working-time accounts, training programs, and the use of a job-transfer service. Public-sector possibilities include tax cuts, wage subsidies, or car-scrap bonus programs, and extended early retirement benefits. Dietz, Stops and Walwei (2010) suggest a two-part modification of the shorttime work scheme: “(1) better conditions for firms suffering from a difficult economic situation and (2) restricted access to short time work for firms in a more favorable economic situation.” The problem with the first part is that there is no objective way to measure the degree to which a firm is suffering. As for the second part, which implies that badly-off firms would have greater access to short-time work, such expanded access, in many cases (especially that of firms whose troubles had begun before the crisis), would do no more than delay the inevitable: bankruptcy. On the other hand, a short-time allowance can provide short-term help to firms that are fundamentally sound, with a solid history and a promising future, but that have been hit hard by the current crisis.

3. Problems with the Empirical Determination of Labor Hoarding According to Taylor (1982) labor hoarding occurs when firms do not immediately bring labor input into line with the demand for their products. If we assume that such a delay, with hoarding as its consequence, is the outcome of a deliberate decision, the issue of unscheduled hoarding remains, since it is bound to occur, on account of incomplete information, a lag in reaction time, indivisibilities of input factors and a certain inflexibility on the labor-input side. Unfortunately, there is no empirical method by which one can distinguish the two types of hoarding. Similar problems arise when one tries to measure labor hoarding in terms of its possible effects: for instance, a reduction in working hours or in labor productivity. Such developments may be due, instead, to changes in the labor supply or to less unpaid work and an increase in absenteeism. These reactions can also contribute to less effective working time. Moreover, the motivation to react with a reduction in working time may distinctly different from that to accept implicitly a lower productivity. How would a business enterprise choose between instituting a system of shorter working hours and reducing its rate of productivity? The latter move might have negative long-run consequences, since once workers have grown accustomed

154

Olaf Hübler

to lower expectations on the part of management it is difficult to motivate them to return to their previous, higher performance level. In contrast, a reduction in hours of working signals to its workers that the enterprise is in trouble and that layoffs are likely, and this dreaded prospect is likely to spur them to be more productive. Nevertheless, the enterprise might opt instead for a reduction in productivity because it is less visible than the reduction in working hours and poses less of a threat to the enterprise’s stability. As Dietz, Stops and Walwei (2010) demonstrate in their charts, the productivity during 2006 – 2008 was very high in Germany. A reduced productivity during the current crisis without alarming the workforce may permit a recovery in the long run because it reduces the risk of negative side effects, for instance, negative health effects. The following simple model reveals the advantages of a reduction in productivity instead of in working hours. Consider the two equations …1†

r ˆ O=H

…2†

A ˆ H=h ;

where r is the level of labor productivity, O is the output (the demand for products), H is the total number, h is the individual number of working hours, and A is the number of employees. If (1) and (2) are combined, then the growth equation via the total differential O^ ˆ A^ ‡ h^ ‡ r^

…3†

follows, where ^ describes the growth rate of a variable. If O shrinks, O^ < 0, it is ^ obvious from (3) that the adjustment is possible on account of a combination of A; ^ ^ h and r^. This means that when it comes to labor hoarding, the firms avoid A < 0 but A^ ˆ 0 is assumed. Then they follow a strategy that has h^ < 0 or r^ < 0. However, if we incorporate two modifications,—namely, the existence of fixed labor costs F and only partial wage cuts due to a reduction in working hours—a more complex growth equation follows. Assume that the relative increase in wage rates, made to compensate for the hours lost (the cost of this lost being born by the firms) is a and f represents the relative fixed labor costs measured in variable labor costs …f ˆ F=…hw††; then …4†

 O^ ˆ A^ ‡ 1


 …a ‡ f †=…1 ‡ f † j" h^ ‡ r^

follows instead of (3), where w represents the wages per employee, and f and a are fixed (f = const., a = const). This means the negative growth rate h^ has to be larger than the negative growth rate r^ in order to have the same effects as r^. This is due to 0 < a < 1, 0 < f < 1 and j"j, where " is the wage elasticity of the labor demand which is absolutely smaller than one as empirical investigations show (Görres 1984, 24, Hamermesh 1993, 94 – 104).

Safeguarding Jobs through Labor Hoarding in Germany—Comment

155

A further problem may arise if firms prefer hoarding of qualified workers and lay off unqualified workers. Usually, it is argued that effective labor productivity decreases under a policy of labor hoarding. This is often but not always the case. Assume that in a recession, on account of a reduced demand for products, both less qualified work (q) and less unqualified (u) work are required. Firms might hesitate to lay off qualified workers, knowing how difficult it will be to rehire them once the recession is over, where is laying off unqualified workers poses no such problem. The result is that they lay off more unqualified workers than necessary but hoard the qualified workers, who therefore during the recession perform some of the tasks of the laid-off unqualified workers …O…u††. We assume that the performance of the qualified workers …r…qjO…q††† is higher than that of the unqualified workers …r…ujO…u††† if both perform their usual tasks. This means c…q† ˆ r…qjO…q††=r…ujO…u†† > 1. The average productivity in the firm before the recession …Av…r† b† is …5†

Av…r† b ˆ r…ujO…u††
…u† ‡ r…qjO…q††
…1

…u†† ;

where …u† is the ratio of unqualified workers in the firm. In the interest of simplicity we assume that the total number of working hours is the same for qualified and unqualified workers …H…q† ˆ H…u††. If the qualified workers perform some of the tasks normally performed by the unqualified workers …r…qjO…u†††, we define c…u† ˆ r…qjO…u††=r…ujO…u††. It seems obvious that r…qjO…q†† > r…qjO…u†† but we do not assume either r…qjO…u†† > r…ujO…u†† or vice versa. On the one hand, their being qualified lends support to the case for the former inequality: the qualified workers learn faster than unqualified to optimize the production process. On the other hand, one could argue that the unqualified workers are more experienced at performing their tasks. If the (negative) growth rate of qualified working hours in the firm is defined by …q† ˆ
H…q†=H…q† where that of unqualified working hours is …u† ˆ
H…u†=H…u†, then the average productivity increases after the reorganization of the work force …Av…r† a† if the following inequality is fulfilled: …6†



…u† c…q†

  1 ‡ 2 …q† c…u†

 c…q† > 0 :

This means Av…r† b < Av…r† a. For this inequality to be fulfilled, it is sufficient that c…u† > c…q† > 1, but this is implausible. However, another condition seems possible: …7†



…u†= …q† > 2 c…q†

  c…u† = c…q†

 1 ;

where …q† < …u† is in accord with our assumptions. If r…qjO…u†† > r…ujO…u††, then inequality (7) is even more likely. As these arguments and counter-arguments indicate, it is difficult to measure labor hoarding and its effects: hence the paucity of empirical studies devoted to this topic. Short-time work has fared somewhat better, but further analysis of its causes and effects is also needed.

156

Olaf Hübler

4. Existing Empirical Evidence to Studies of Short-time Work It was in the 1970s that the issue of short-time work, especially in the manufacturing sector, began to attract the attention of policy makers and researchers. In Germany in 1978 alone, 86 million hours were cancelled by short-time work. Flechsenhar (1979) has calculated that this is an equivalent of 33.000 unemployed people. There are only a handful of further empirical studies of short-time work in Germany and its impact on employment. A summary of their conclusions follows.  The macroeconomic development of short-time work varies widely from period to period, on account of cyclical developments and changes in labor market policy. In 1991 the rate of short-time work in eastern Germany was extremely high and during the next two years it did not decline significantly. It then decreased significantly, only to rise sharply in 2009 (Eichhorst / Marx 2009).  Short-time work in eastern Germany after reunification failed to have the positive effect that had been expected, but instead very often led to unemployment (Hübler 1997).  The positive effect of short-time work in 1993 – 1994 lasted for only three months and then induced a sharp long-run drop in wage levels (Speckesser 2010).  The reduction of standard working hours in enterprises that had established agreements with all their employees during the period 2001 – 2003 failed to improve the employment situation. Instead, extending working hours, adopting working-time accounts, and applying qualification measures to the workforce were found to be effective remedies (Hübler 2005).  An empirical investigation based on the 2003 IAB establishment survey shows that companies using mostly short-time work are distinguished by a large percentage of workers with an academic degree, few leased laborer and workers with temporary contracts, few part-time workers, extreme fluctuations in business activity, and scant profitability. This measure is relatively widespread in large firms, in the production sector and in western Germany (Crimmann / Wießner 2009).  Drawing on a survey in 2009 among of about 1000 Bavarian companies, Wiegmann (2009) found that 26 percent use and 4 percent plan to use the short-time instrument. In two thirds of these companies not more than the half of the staff works on a short-time basis.  Analyses, by means of probit estimates, to determinants of short-time work reveal several significant developments between 2003 and 2009. First, the chance that exceptionally qualified employees would work on a short-time basis went from high to low. Second, export businesses are more likely to use a work-sharing system during the current crisis than they were in 2003. Third, in 2003 there were no discernible differences between eastern and western Germany, but currently short-time work is more widespread in the latter (Crimmann / Wießner / Bellmann 2010).

Safeguarding Jobs through Labor Hoarding in Germany—Comment

157

 Drawing on three KOF surveys (Konjunkturforschungsstelle der Eidgenössischen Technischen Hochschule Zürich) of 514 Swiss firms in the industry Frick (2009) shows that firms using short-time work reduce their workforce during a crisis to a greater extend than do other firms; that, as a rule, they engage in less labor hoarding; and that in the long run a work-sharing system fails to stabilize the employment situation.  Individuals who have previously been promoted are less likely than others to turn to a job-transfer service as an alternative to traditional short-time work (Gerlach / Hübler 2009). The investigation is based on an extensive dataset of a large German manufacturing company from January 1999 to December 2005.  The degree to which a company that uses short-time work has been affected by the current crisis has little or no effect on the evolution of its average wage rate (Bellmann / Gerner 2010). The analysis is focussed on the data of the 1st half of the years 2008 and 2009 of the IAB establishment survey.

5. New Results from the IAB Establishment Panel 5.1 Database and Descriptive Statistics

In this section we analyse the 2003, 2006, and 2009 data on short-time work in Lower Saxony focussing our attention primarily on 2009. The database is the business-establishment file of Germany’s Federal Employment Agency, containing files of all such establishments employing at least one person covered by social security. The IAB Establishment Panel performs an annual survey of approximately 16,000 such establishments, including 1078 in Lower Saxony. Interviews with the firms’ high-ranking managers are conducted for the Institute for Employment Research (IAB) by TNS Infratest Munich. A questionnaire covers changes in the structure of the workforce as well as the business development, and earnings. The 2009 survey comprises ninety-four questions, five of which concern shorttime work:  Did your company use short-time work in the first half of this year?  If yes, how many employees were concerned?  Have you instituted a training program in connection with short-time work?  If yes, how many employees were involved?  Who has financed the program?

Short-time work is more widespread in larger firms than in smaller ones, and especially so during the current severe recession; in fact, the correlation is statistically significant in 2003 and 2009 (Table 1). A business that has already had difficulties in the past is more likely than others to resort to short-time work. This correlation suggests that short-time work may have negative long-term consequences.

158

Olaf Hübler

2 tests in the cross tables show that this is the case between 2003 and 2006 and, but to lesser degree, between 2003 and 2009 (Table 2). Table 1 Short-time Work and Firm Size 2009 1

FSIZEk STW 0 STW 1

2

2006 3

4

1

2

3

4

1

2

2003 3

4

97.09 89.70 78.57 76.06 99.02 98.80 98.21 98.90 98.23 94.36 92.79 100.00 2.91 10.30 21.43 23.94 0.98 1.20 1.79 1.10 1.77 5.64 7.21 0.00

Person 2 p-value

59.52 0.000

0.727 0.867

13.53 0.004

Note: STW = 1 if the company uses short-time work; FSIZEk = 1 if firm size 10 and < 100 employees; FSIZEk = 3 if firm size > 100 and < 500 employees; FSIZEk = 4 if firm size > 500 employees. Source: IAB Establishment Panel 2003, 2006 and 2009—Lower Saxony.

Table 2 Cross Tables of Firms’ Transitions of Short-time Work (STW) among 2003, 2006, and 2009 STW_06 STW_09 0 1 P 2 p-value STW_06 0 1 P

0

1

330 71

5 1

401

6 0.004 0.947

P

STW_03

P

0

1

335 72

323 65

12 7

335 72

407

388

19 5.020 0.025

407

386 2

15 4

401 6

388

19 52.600 0.000

407

2 p-value Source: IAB Establishment Panel 2003, 2006, and 2009—Lower Saxony.

The growth rates of firms using short-time work in 2009 are on average lower than those of other firms—see Table 3. We find also some establishments which have a positive growth rate, even though they do not expect to be able to increase the size of their workforce. The graph of the kernel density estimator shows that the growth of establishments that do not use short-time work concentrates around zero (Figure 1). The fact that the kurtosis is larger than that of a normal distribution …4;STW ˆ1 = 4.61; 4;STW ˆ0 = 6.62) means that the tails are thicker.

Safeguarding Jobs through Labor Hoarding in Germany—Comment

159

Table 3 Frequency Distribution of the Growth Rate Firm Size

–1.0 < = GR_FS < –0.5 –0.5 < = GR_FS < 0.0 0.0 = GR_FS 0.0 < GR_FS < = 0.5 0.5 < GR_FS < = 1 1.0 < GR_FS Total

All

STW = 1& ISE = 0

STW = 0 & ISE = 0

2.23 30.24 31.45 34.23 1.11 0.74

2.65 62.83 9.73 23.89 0.88 0.00

2.37 26.78 36.15 32.85 1.19 0.66

100.00

100.00

100.00

Note: STW—dummy (= 1, if short-time worker); ISE—dummy (= 0, if no increasing sales expected); GR_FS—growth rate of firm size measured by the number of employees.

0

2

4

Density 6

8

10

Source: IAB Establishment Panel 2009—Lower Saxony.

-.5

0 WRB short-time work

.5 no short-time work

Figure 1: Epanechnikov Density Estimator of the Growth Rate of Firm Size (WRB), Measured by the Number of Employees

5.2 Econometric Results

A fractional-polynomial estimator determines the two-variable nonlinear relation between the growth rate of the firm size, measured by the number of employees, and the rate of short-time workers in a given firm. It is interesting that, although the estimation is imprecise (i.e., the confidence interval is flat) those companies using short-time work at a medium rate (between 0.5 and 0.7) have the best chance of

160

Olaf Hübler

-.2

-.15

-.1

-.05

0

increasing the size of their workforce (Figure 2). A company with a low rate means, c.p., has to bear relatively high costs for its other workers; the learning effect of this dilemma is a cautious approach to hiring. The use of short-time-work at high rate is an indicator of serious trouble, for which there is no such short-term solution.

0

.2

.4

.6

.8

1

Short_time 95% CI

predicted Firm_size

Figure 2: Fractional-polynomial Estimation of the Growth Rate of Firm Size as a Function of the Ratio of Short-time Workers Presented as a Confidence Interval of the Mean

The effects of establishment characteristics on the ratio of short-time work are determined and presented in Table 4. The estimates show that firms with their own research and development (R&D) section have on average a higher ratio of shorttime workers than do other firms. During a crisis these firms try to reduce costs. One way is to restrict R&D activities, since they are effective only in the long run. The positive sign of the working time account coefficient indicates that short-time work and working time accounts are complementary. If the demand for products decreases, the employers apply both instruments. These estimates do not clarify whether these two instruments are associated with the same reduction in working hours or whether some sectors of the workforce find one instrument preferable to the other in this regard. Both early retirement and short-time work are used during the current crisis and seem to be effective, too, but their correlation is insignificant. The sign of the coefficients of hiring and expectation of increased sales are as expected. The positively significant coefficient of foreign sales demonstrates that export-oriented firms are being hit particularly hard by the current decline in product demand. That workers with a university degree get a negative sign is unexpected; one would expect firms to hoard this sector of the workforce more than any other. One possible explanation is that, while firms with a workforce featuring

Safeguarding Jobs through Labor Hoarding in Germany—Comment

161

these qualified workers are less vulnerable to the current crisis. This could mean that the conditions for short-time allowance are not fulfilled although the demand for qualified workers is reduced. The employer does not dismiss these workers but hoards them and employs them at tasks for which they are overqualified. In this case less qualified workers are dismissed or offered a short-time allowance, because there is not enough unqualified work to go around. Table 4 Tobit Estimates of the Ratio of Short-time Workers (RSTW) with Respect to all Workers as a Function of Establishment Characteristics in 2009 Number of obs LR 2 (12) Prob > 2 Pseudo R2 Log likelihood

= 892 = 160.77 = 0.0000 = 0.2199 = –285.21482 RSTW !

R&D (= 1, if yes) Working time account (= 1, if yes) No collective agreement (= 1, if yes) Hirings (= 1, if yes) Works council (= 1, if yes) Member of an employers’ association (= 1, if yes) Profitability last year (= 1, if very good, . . . ; = 5 if bad) Public ownership (= 1, if yes) Partial retirement (= 1, if yes) Increasing sales expected (= 1, if yes) Foreign sales (ratio of total sales) Workers with university degree (ratio of total employees) Constant 

Coef.

Std. Err.

t

P > jtj

0.4394

0.1148

3.83

0.000

0.3271

0.1045

3.13

0.002

–0.1335

0.1004

–1.33

0.184

–0.3539

0.0960

–3.68

0.000

0.2287

0.1147

1.99

0.046

0.3576

0.2334

1.53

0.126

0.1452

0.0630

2.30

0.021

–0.5113

0.4093

–1.25

0.212

0.1706

0.1198

1.43

0.154

–0.5040

0.1575

–3.20

0.001

0.0080

0.0022

3.61

0.000

–1.7124

0.5701

–3.00

0.003

–1.8633

0.3622

–5.14

0.000

0.7572

0.0611

Source: IAB Establishment Panel 2009—Lower Saxony.

162

Olaf Hübler

We present two estimates of the influence of short-time work on the firm’s growth rate: in one, the ratio of short-time workers is used as regressor; in the other, only those short-time workers who used the increase in their non-work time to join a training program are considered (Table 5). A Tobit estimator is applied. Whenever absolute rates are greater than 100 percent, they are censored. It is of interest that short-time work in combination with training has a positive effect on the employment rate, whereas short-time work alone has the opposite effect. It seems curious that few firms use this dual instrument, of short-time work coupled with a training program, since it has proved to be effective (Crimmann, Wießner and Bellmann 2010). Table 5 Tobit Estimates of the Growth Rate of Firm Size in 2009 with Respect to Short-time Work, Training and other Establishment Characteristics Number of obs LR 2 (15) Prob > 2 Log likelihood Pseudo R2 Growth rate of firm size ! Ratio of short-time workers Short-time work and training R&D Working time account No collective agreement Hirings Working council Member of an employers’ association Profitablity last year Public ownership Partial retirement Increasing sales expected Foreign sales Workers with university degree Ratio of temporary workers Training measures (= 1, if yes) Constant 

892 127.88 0.0000 –276.52951 0.1878 Coef.

Std. Err.

–0.1428

0.0716

–0.0295 –0.0160 –0.0119 0.2240 –0.0715 0.0486 –0.0073 –0.0111 0.0403 0.0425 –0.0001 0.1080 –0.0366 0.0106 –0.2398 0.2526

Source: IAB Establishment Panel 2009—Lower Saxony.

899 128.69 0.0000 –282.4425 0.1855 Coef.

Std.Err.

0.0322 0.0236 0.0235 0.0247 0.0298 0.0406 0.0123 0.0816 0.0337 0.0265 0.0007 0.0768 0.0847 0.0244 0.0617

0.0006 –0.0373 –0.0226 –0.0140 0.2348 –0.0798 0.0499 –0.0089 –0.0067 0.0465 0.0450 –0.0003 0.1252 –0.0381 0.0149 –0.2441

0.0011 0.0323 0.0237 0.0236 0.0248 0.0299 0.0411 0.0124 0.0822 0.0334 0.0266 0.0009 0.0773 0.0853 0.0245 0.0624

0.0120

0.2554

0.0120

Safeguarding Jobs through Labor Hoarding in Germany—Comment

163

However, the estimates presented in Table 5 do not prove that short-time work is the cause of the growth rate of a firm’s workforce during the crisis, because it is possible that the cause lies instead in characteristics that were also typical before the crisis began. We therefore apply a propensity score matching estimator (Table 6) and also take into consideration the nonlinear course between a firm’s growth rate and the ratio of short-time work presented in Figure 2; the effect is a reversal of the effect of short-time work in Table 5 when it alone is applied. At first, an increase in the percentage of short-time workers is positively correlated with an increase in the size of the firm’s workforce. In other words, the workforce decreases in a recession due to dismissals as long as no short-time work is adopted. Short-time allowance helps to stop this development. However, if the percentage of short-time workers exceeds a certain threshold in an establishment management interprets this situation as a sign that the economic situation is deteriorating, concludes that hoarding is an inappropriate policy, and begins to lay off workers in increasing numbers. Table 6 Propensity Score Matching Estimator of the Growth Rate of Firm Size Number of obs LR 2 (16) Prob > 2 Pseudo R2 Log likelihood

= 892 = 175.91 = 0.0000 = 0.1539 = –483.66081

Growth rate of firm size ! Ratio of short-time workers (RSTW) RSTW2 R&D Working time account No collective agreement Hirings Works council Member of an employers’ association Profitability last year Public ownership Partial retirement Increasing sales expected Foreign sales Workers with university degree Ratio of temporary workers Training measures Constant

Coef.

Std. Err.

z

P > jzj

2.4949 –2.0407 0.0002 0.2053 –0.0656 0.6036 0.2585 0.3523 0.1650 –0.0402 0.5341 0.0143 0.0034 0.7884 0.4381 0.1058 –1.0816

1.1235 1.4014 0.1499 0.1024 0.1012 0.1034 0.1359 0.1671 0.0528 0.3999 0.1823 0.1226 0.0033 0.3890 0.3837 0.1017 0.2553

2.22 –1.46 0.00 2.01 –0.65 5.83 1.90 2.11 3.13 –0.10 2.93 0.12 1.03 2.03 1.14 1.04 –4.24

0.026 0.145 0.999 0.045 0.517 0.000 0.057 0.035 0.002 0.920 0.003 0.907 0.301 0.043 0.254 0.298 0.000

Source: IAB Establishment Panel 2009—Lower Saxony.

164

Olaf Hübler

In Table 5 we have presented short-run effects, but the final estimation is a matter of ones. Using, once again, a propensity score matching estimator, we find that the effects switch from positive to negative when the time is extended (Table 7): further evidence that short-time work should be used only for a short period. Otherwise this instrument is counterproductive.

Table 7 Propensity Score Matching Estimator of the Growth Rate of Firm Size with Respect to Current and Lagged Short-time Work (STW) Number of obs LR 2 (3) Prob > 2 Pseudo R2 Log likelihood

= 407 = 19.35 = 0.0002 = 0.0399 = –232.63155

Growth rate of firm size ! STW_09 STW_06 STW_03 Constant

Coef.

Std. Err.

z

P > jzj

0.8773 0.6320 –0.3951 0.4667

0.2202 0.6353 0.3350 0.0723

3.98 0.99 –1.18 6.45

0.000 0.320 0.238 0.000

Source: IAB Establishment Panel 2003, 2006, and 2009—Lower Saxony.

6. Conclusions It is evident that labor hoarding and short-time work have contributed to Germany’s relatively robust labor market in the current crisis. We can state with confidence that if it is on account of cyclical, as opposed to structural, problems that firms resort to labor hoarding and short-time work, then short-time work is effective, but for a short period only. A far better remedy is a combination of short-time work and training programs, because it minimizes the problems that are sure to arise when the next business boom begins. The risk of layoffs is greater in a firm whose entire workforce is engaged in short-time work than in a firm where only a portion of the work is short-time. However, since the effects of short-time work vary greatly over time, firms would do well to limit short-time work to certain sectors of their workforce, and thereby avoid the windfall gains that excessive use of the short-time system generates. However, in the case of firms in serious difficulty, the use of short-time work is almost bound to lead, in the long run, to layoffs; such firms should hire a job-transfer service and institute training programs. Labor hoarding and short-time work offer the greatest chance of long-term employment in the case of well-off firms whose employment problems are limited to temporary bottlenecks.

Safeguarding Jobs through Labor Hoarding in Germany—Comment

165

With the passage of time, data in the current crisis will increase, and these data will permit us to generate more reliable robustness checks regarding the causes and effects of labor hoarding and short-time work. A regression discontinuity approach and DiD estimators can be applied, at the same time, as well. Different sectors of the workforce can then be analyzed, and the interactions among various factors affecting employment rates—most importantly, the cyclic patterns of shorttime work—can be determined with precision.

References Bell, D. N. F. / Blanchflower, D. G. (2010): “Recession and Unemployment in the OECD,” CESifo Forum 11, 14 – 22. Bellmann, L. / Gerner, H.-D. (2010): “Reversed Roles? Wage Effects of the Current Crisis,” mimeo. Crimmann, A. / Wießner, F. (2009): “Verschnaufpause dank Kurzarbeit”, IAB-Kurzbericht 14 / 2009. Crimmann, A. / Wießner, F. / Bellmann, L. (2010): “The German Work Sharing Scheme: An Instrument for the Crisis,” mimeo. Dietz, M. / Stops, M. / Walwei, U. (2010): “Securing Jobs through Labour Hoarding in Germany,” in this volume. Eichhorst, W. / Marx, P. (2009): “Kurzarbeit: Sinnvoller Konjunkturpuffer oder verlängertes Arbeitslosengeld?”, IZA Standpunkte Nr. 5. Flechsenhar, H.-R. (1979): “Kurzarbeit—Strukturen und Beschäftigungswirkung”, Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 12, 362 – 372. – (1978): “Kurzarbeit—Kosten und Finanzierung”, Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 11, 443 – 456. Frick, A. (2009): “Die Wirkung der Kurzarbeit,” ETH, mimeo. Gerlach, K. / Hübler, O. (2009): “Employment Adjustments on the Internal and External Labour Market—An Empirical Study with Personnel Records of a German Company,” Jahrbücher für Nationalökonomie und Statistik 229, 198 – 213. Görres, P. A. (1984): “Die Umverteilung der Arbeit,” Frankfurt / New York, Campus. Hamermesh, D. S. (1993):“Labor Demand,” Princeton (N.J.), Princeton University Press. Hübler, O. (1997): “Evaluation beschäftigungspolitischer Maßnahmen in Ostdeutschland,” Jahrbücher für Nationalökonomie und Statistik 216, 21 – 44. – (2005): “Sind betriebliche Bündnisse für Arbeit erfolgreich?”, Jahrbücher für Nationalökonomie und Statistik 225, 630 – 652. Speckesser, S. (2010): “Employment Retention in the Recession: Microeconomic Effects on the Short-Time Work Programme in Germany,” unpublished. Wiegmann, J. (2009): “Kurzarbeit—Ergebnisse des BIHK-Unternehmensbarometers,” BIHK, mimeo.

The Combined Employment Effects of Minimum Wages and Labor Market Regulation—a Meta-Analysis By Bernhard Boockmann* Abstract This paper provides a meta-analysis of 55 empirical studies estimating the employment effects of minimum wages in 15 industrial countries. It strongly confirms the notion that the effects of minimum wages are heterogeneous between countries. As possible sources of heterogeneity, it considers the benefit replacement ratio, employment protection and the collective bargaining system. While the results are in line with theoretical expectations, the degree to which they are robust differs across these institutions. Keywords: Minimum wage, regulation, employment, meta-analysis JEL Classification: J38, J20, C12

1. Introduction Following the influential study by Card and Krueger (1995a), a wealth of empirical studies on the employment effects of minimum wages has been produced for a large number of countries. The available evidence for the US has been summarized in meta-analyses by Card and Krueger (1995b) and, more recently, by Doucouliagos and Stanley (2009). Meta-analysis is a useful statistical tool for reviewing empirical results increasingly adopted in the economics profession. This study provides a meta-analysis for the effect of minimum wages on employment in major industrial countries since 1995. A central theme of the paper is that the effects of particular regulations such as mandates for minimum wages are likely to depend on a country’s wider institutional setting, as well as on norms, macroeconomic conditions and other circumstances. However, it is difficult to estimate empirically how important these inter* Institute for Applied Economic Research (Institut für Angewandte Wirtschaftsforschung, IAW), Tübingen; University of Tübingen; IZA Bonn. Ob dem Himmelreich 1, D-72074 Tübingen, Germany, Tel.: +49 7071 9896 – 20, Fax: +49 7071 9896 – 99, e-mail: bernhard. [email protected]. I wish to thank Alexandra Spitz-Oener and participants of the 2010 ARGE conference for helpful suggestions and Sarah Hofmann and Thomas Schneider for highly committed research assistance. Any remaining errors are my own.

168

Bernhard Boockmann

dependencies are. In modern empirical analysis, regulation effects are mostly estimated on the basis of micro-level data and using techniques such as natural experiments. This approach, however, takes the institutional and social framework as given and does not allow for comparisons between different settings. By contrast, studies based on country-level panel data offer the potential for cross-country comparisons but often suffer from low statistical power. In this trade-off, the use of metaanalysis may provide a bridge between the use of sophisticated methods of statistical evaluation and the objective to compare systematically between countries. The interdependence between different policy interventions is a question of high political importance and the object of much academic research. Following the important contributions by Coe and Snower (1997) and Orszag and Snower (1998), studies such as Amable and Gatti (2006), Bassanini and Duval (2006, 2009), Eichhorst and Konle-Seidl (2005) and Elmeskov et al. (1998) give support to the notion that policy recommendations and reform proposals for labor markets need to differ according to a country’s institutional framework. Empirically, interdependence between particular labor market regulations such as employment protection, replacement rates, collective bargaining structures and tax rates has been analyzed studies such as Belot and van Ours (2004) and Belot et al. (2007). These studies find significant evidence for the interaction of different policies. Concerning minimum wages, interaction effects with other regulations on youth unemployment have been empirically investigated by Neumark and Wascher (2004a) for a sample of 17 OECD countries. Using pooled cross-section time-series data, these authors find that the employment effects of minimum wages vary considerably across countries. Institutions such as employment protection, union coverage and active labor market policies are shown to explain part of these differences. In this paper, we focus on the level of benefit payments, employment protection and the collective bargaining system as possible sources of policy complementarities. First, benefits received by the long-term unemployed often set effective minimum wages. If market wages fall below this level, workers rationally prefer to remain out of work. Thus, minimum wages have employment effects only above a certain threshold. Conversely, reducing benefits may only increase employment if minimum wages are low (Orszag and Snower 1998). Second, costs of employment protection may be shifted to workers in the form of lower wages unless shifting is prevented by minimum wages (Lazear 1990, Cahuc and Zylberberg 1999). Thus, at least in the long run, adverse employment effects of employment protection and minimum wages may reinforce each other. In the short run, however, employment protection may also reduce employment losses that would otherwise result from an increase of the minimum wage (Neumark and Wascher 2004a). Third, collective bargaining also sets minimum wages that may be a substitute for legal minimum wages. If collective bargaining results in uniform nationwide standards that are generally applied, the additional effect of a statutory minimum

The Combined Employment Effects of Minimum Wages

169

wage is likely to be limited (Cahuc et al. 2001). The inverse relation between the use of statutory minimum wages and union bargaining has been given a theoretical underpinning by Aghion et al. (2008). In addition to these institutions, minimum wages may interact with several other country characteristics, including tax policies (Bassanini and Duval 2006), active labor market, training and housing policies. Product market regulation could be a close substitute to minimum wage regulation if minimum wages are used to increase competing firms’ costs of production (see, however, Amable and Gatti 2004 for a model with different conclusions). In addition, Alesina et al. (2010) stress interactions between labor market regulation and the role of the family. In this paper, however, we are explicitly concerned only with labour market regulation and welfare state characteristics. The high significance of country fixed effects found in our empirical results, however, makes it appear quite likely that other interaction effects are present.

2. Meta-Analysis and Minimum Wage Research Meta-analysis is a quantitative method to summarize the content of a sample of empirical studies.1 In a meta-regression approach, the dependent variable is a statistic such as an elasticity or a regression parameter obtained from each study in the sample. A study may contribute more than one estimate to the analysis. The dependent variable is statistically explained by characteristics of the study such as its empirical data base, the methods used, place and time of publication and others. If successful, meta-analysis may reconcile the diversity of findings often obtained by researchers working on the same empirical question. It may also hint at inadequacies in the literature such as publication bias. While meta-analysis originated in the sciences, it is frequently used in empirical economics. There are numerous examples for applications to empirical labor economics, such as Greenberg et al. (2005), Kluve (2006) and Card et al. (2009) for the empirical effects of welfare and active labor market programs, Jarrell and Stanley (1990) for the union influence on wages, Pereira and Martins (2004) for the returns to education, Longhi et al. (2005) for immigration and earnings and Weichselbaumer und Winter-Ebmer (2005, 2007) for the presence of gender discrimination in wage-setting. The employment effects of minimum wages have been meta-analyzed by Card und Krueger (1995b) and Doucouliagos and Stanley (2009). Card and Krueger used meta-analysis to examine publication bias in a sample of 14 time-series studies for the US. They start with the observation that in a meta-analysis using an unbiased sample of coefficients (where differences result from the usual sampling 1 More detailed information on the methods of meta-analysis and their use in economics is provided in Stanley (2001) and Stanley and Jarrell (2005).

170

Bernhard Boockmann

error and not from systematic selection), the relation between sample size and reported t-values should be inverse. Contrary to this expectation, they find that the reported employment effects of the minimum wage very often have a t-statistic of slightly above two, regardless of sample size. This suggests the presence of publication bias, such that published studies are systematically selected on the basis of the t-values of the estimated parameters.2 The more recent study by Doucouliagos and Stanley (2009), based on a sample of 64 studies for the US using different methodologies, similarly finds strong evidence for publication bias and little or no evidence of a negative association between minimum wages and employment. More often than examining publication bias, however, meta-analysis is used to synthesize the literature and derive conclusions with respect to the questions the original studies sought to answer. Thus, several independent variables typically used in meta-analyses contain information about the specific kind of intervention and the population groups to which the intervention is targeted. Variables relating to study design (e.g., data sources and sampling, econometric methods used) are often used as control variables. Finally, if the studies refer to different countries, one may also include measures defining the institutional context and other country characteristics. This is the feature most important for our study. An example for country characteristics identified by meta-analysis is WinterEbmer and Weichselbaumer (2007). The country-level policy variables used in their study relate, first, to the ratification of international conventions against discrimination and, second, to economic freedom (as measured by the Freedom House index) as a measure of competition. They find that, consistent with expectations, both international rules and competition reduce the wage differential estimated by the econometric studies in the sample.

3. Data Data collection is a first and crucial step in meta-analysis. Since statistical analysis requires a sufficient data base, only topics frequently investigated in empirical research can be analyzed. Given the large number of studies published after Card and Krueger (1995a), the minimum wage is a suitable topic for a meta-analysis. Since 1995 marks a turning point in research on minimum wages and in order to obtain a relatively homogenous set of studies, only studies published after 1995 are included in the sample. Many of these studies are not concerned with the effect of the minimum wage on employment but study other outcomes potentially influenced by minimum wages, such as labor earnings, the wage distribution, or product prices. By contrast, only estimates relating to employment are included in this study. The data set consists of all econometric studies published in journals or as discussion papers found in an extensive literature search. Standard economic data2

See, however, Neumark and Wascher (1998) for an alternative explanation.

The Combined Employment Effects of Minimum Wages

171

bases (ECONLIT, ECONIS and RePEc) were systematically searched using “minimum wage” and several flections as a keyword. In addition, internet searches using Google were made to detect new studies not yet published in journals or included in standard databases. The lists of references in the database were checked for further studies to ensure that our database is as complete as possible. A minimum requirement for inclusion of a study in our data is that the employment effect is estimated econometrically by an estimator more sophisticated than simply a bivariate correlation. Most of the studies in the sample use either regression techniques or quasi-experimental difference-in-differences (DiD) methods. Since the objective is to identify the effects of variables measured at country level, studies estimating effects for several countries (e.g., in a pooled cross-section timeseries study for a panel of countries) were excluded. Apart from these requirements, all available studies were used, leaving us with a data set of 55 studies relating to 15 industrial countries.3 Both studies based on micro and aggregate data are included. The list of papers is contained in the Appendix. Given that most studies contain more than a single estimate for the employment effects, a choice has to be made of how many to include. Two extreme ways to deal with this problem are the “best-set” and the “all-set” of estimates (Doucouliagos and Stanley 2009). In the first case, only the estimate preferred by the study author(s) is included in the data. The disadvantage is that an objective standard for the preferred estimate does not exist. In the second case, all estimates are included. The problem here is that a lot of redundant information is included if authors engage in extensive robustness analysis or present results for slightly different data versions. The best choice is, therefore, to determine what constitutes a separate entry in the data base on a case-by-case basis. To standardize the decision, a codebook was developed giving instructions how to distinguish substantial variations between estimates from redundant information. The rules of the codebook are contained in Table 1. While the application of the rules did not uniquely determine coding decisions, they provided guidelines that were sufficiently transparent and manageable. Checks revealed that the decisions made on the basis of the codebook were consistent. The data was coded partly by the present author and partly by a research assistant. As a result, the 55 studies yielded 304 separate estimates, with a minimum of one estimate and a maximum of 21. Since the likelihood of obtaining similar estimates is higher within studies than between studies, standard errors with clusters for studies are used in estimation. Not surprisingly, the distribution of estimates over countries is very uneven. Roughly one half of the studies relate to the US. Other countries with many studies are the UK, Canada and France (see Figure 1). 3 There are also a number of studies for developing countries and emerging economies. To keep the country sample homogenous, they were not included in this study.

172

Bernhard Boockmann Table 1 Codebook for Inclusion of Estimation Results as Separate Entries

160 140 120 100 80 60 40 20

Figure 1: Number of Data Points by Country, Estimation Sample

Spain

Portugal

Ireland

Germany

Finland

Sweden

Slovak Republic

0

Czech Republic

7

Austria

6

New Zealand

5

Australia

4

Canada

3

France

2

Include results for different groups of workers (sex, age, qualification etc.) separately. Otherwise, include only the specification with the most comprehensive data base (e.g., do not include sub-periods, use the specification with the largest set of controls unless rule 5 applies, etc.). If possible, choose the representation that is most comparable across studies (e.g., report elasticities rather than measures depending on the scale of the variables). Do not include small variants of estimations as separate entries. Small variations are specifications that differ in the set of covariates or the specific functional form (e.g. linear versus logarithmic). If no other decision rule applies, include the smallest and the largest estimate among a set of similar results. If specifications differ substantially (e.g., use of different estimation procedures, different data bases), include results as separate entries. Do not include results of generalizations not supported by the data (e.g., insignificant lags, insignificant groups of dummies). Exclude specifications explicitly marked as inferior to preferred specifications by the author(s).

UK

1

Definition

USA

Number

The Combined Employment Effects of Minimum Wages

173

4. Empirical Model A metric used in existing meta-studies is the elasticity of employment with respect to the minimum wage. Using this metric, however, would result in the loss of a large number of studies and countries, since elasticities are often not reported or cannot be derived from the results. This concerns, in particular, micro-level studies where the (change of) individual employment status is the dependent variable. Moreover, employment is defined differently across studies (employment numbers, full-time equivalents, hours etc.), and some of the studies use unemployment instead of employment as the concept of measurement. The measure of the minimum wages also varies across studies. Most studies use the absolute amount of the minimum wage per hour worked. Several macro-level studies, however, use the Kaitz index, i.e. the relation between minimum wage and average wage, or relate the minimum wage to other reference measures. In some cases, the proportion of workers directly affected by the minimum wage is taken as a measure of regulation intensity. Since a common metric for the size of the effect is lacking, we follow the example of other studies such as Card et al. (2009) and define the dependent variable on the basis of statistical significance and the direction of the influence. This results in an ordinal measure with four categories (statistically significant and positive, positive but insignificant, negative but insignificant, and significantly negative) where the cut-off level for significance is five percent. Alternative measures with fewer or more categories are used for robustness checks. To account for different concepts of measurement, the definition of employment and minimum wage is included as a control variable in the estimations. Table 2 contains the frequencies of the four categories of the dependent variable used in the analysis. Roughly one third of the estimates are significantly negative, insignificant and negative or positive (either significantly or insignificantly). About 8 per cent of the estimation results represent a significantly positive effect. Of course, no conclusions as to the “true” effect of minimum wages can be derived from these frequencies. To take account of the ordinal nature of the dependent variable, the ordered probit model is used for estimation. The independent variables are listed (with means and standard deviations) in Table 3. A first group concerns the population group for which the study has been made. The employment consequences of minimum wages are often believed to be particularly strong for young workers. This is reflected in the fact that a quarter of the estimation results refer to individuals below age 19. About one third of the data points represent separate estimation results for men and women while in two thirds of the estimates no distinction according to gender is made. In addition, a small number of studies restrict attention to lowskilled workers or to workers from low-paying industries.

174

Bernhard Boockmann Table 2 Dependent Variable, Estimation Sample

sigificantly negative (5 % level) insigificantly negative insigificantly positive sigificantly positive (5 % level)

Number

Percent

96 107 78 23

31.58 35.2 25.66 7.57

Table 3 Independent Variables, Estimation Sample Mean Age less than 19 Age 19 – 25 Gender: Men only Gender: Women only Low-skilled workers Low-wage sectors Small industry Data: Micro Data: Macro panel Observation period Outcome: unemployment Estimator: DiD Estimator: IV Peer reviewed journal Minimum wage level Basic assistance level Employment protection Bargaining coordination

0.25 0.06 0.18 0.18 0.02 0.07 0.08 0.56 0.34 1992.23 0.02 0.58 0.03 0.64 4.64 2.95 1.08 1.97

Std. Dev. 0.44 0.23 0.39 0.39 0.14 0.25 0.28 0.50 0.47 7.06 0.14 0.49 0.16 0.48 0.95 1.11 1.05 1.10

Studies using the difference-in-differences (DiD) methodology often restrict attention to one or several narrowly defined industries, such as the fast food sector in Card and Krueger (1995a). The employment consequences of a minimum wage may differ in a specific sector and the whole economy. As Neumark and Wascher (2007) remark, if the minimum wage affects both the sector under study and a competing sector, and the competing sector produces with higher labor intensity, the price response will be larger in the competing sector, increasing demand in the sector considered in the study. A positive employment effect of the minimum wage in a small industry may, thus, be fully consistent with the competitive model of the

The Combined Employment Effects of Minimum Wages

175

labor market. Therefore, we include a dummy indicating whether the study relates to workers in a small industry (such as fast food stores, supermarkets, hairdressing, or residential care homes). Concerning the level of aggregation of the data, we distinguish between studies based on individual-level and aggregate data, where the former either relate to individuals or firms and establishments. If regional averages are derived from individual-level data and analysis is performed at the regional level, the study is coded as using aggregate data. Among the studies based on aggregate data, we distinguish time-series and panel studies. For the period after 1995, the vast majority of aggregate data studies use panel data. As mentioned before, the sample studies differ in the measurement concept of the outcome variable. Originally, we defined a set of dummies for the different outcomes. Since most of them never turned out significant, we retain only one variable indicating if the outcome measure is the unemployment rate rather than a measure for employment. We group the estimators into three categories: OLS (including time-series estimators), IV and DiD. More than half of the estimates in the sample originate from applications of the DiD technique, attesting to the huge influence of the Card-Krueger approach in the minimum wage literature. A last study characteristic is whether the results were published in a peer-reviewed journal or appeared as a discussion paper or policy report. If journal referees are biased in their judgement in favor of statistically significant results and / or in favor of the economic “common sense,” authors will have difficulties in placing insignificant or positive findings in reviewed journals.4 An extensive analysis of the subtle methods used to detect publication bias is included in Doucouliagos and Stanley (2009). An important group of independent variable relate to labor market regulation in the country for which the study has been done. A first characteristic is the tightness of existing minimum wage regulation. An increase in the minimum wage from a level already quite advanced may have a different effect than introducing a minimum wage in a country in which it has not previously existed. Apart from a measure for minimum wage regulation, we also use indicators for the level of benefits received by the long-term unemployed, the level of employment protection and the collective bargaining system. These indicators are used to detect possible policy complementarities between different fields of labor regulation, their choice being motivated by the literature cited in introductory section. All the variables used are constant for countries and years and relate to the late 1990s, which corresponds to the average of the observation period of the studies included in the analysis. For the minimum wage level, we use Neumark and Wascher (2004b, Table 1) as the data source. For employment protection, we use 4 We checked whether the effect of this variable changes if we include the publication date as an additional regressor, to take account of not-yet-published as opposed to never-published papers. There were little changes in the coefficient of this variable from this robustness check.

176

Bernhard Boockmann

the OECD summary indicator of employment protection for the late 1990s, as published in the 2004 OECD Employment Outlook (OECD 2005). This broad measure captures employment protection of regular employment relationships as well as regulations for fixed-term and temporary agency work. The benefit replacement rates are taken from the 2007 edition of “Benefits and Wages,” a regular OECD publication. The specific measure used relates to the average for four different household types and over the first 60 months of unemployment and is measured in the year 2001. It does not distinguish between different branches of the social security system, such as unemployment insurance or social assistance. The collective bargaining system can be distinguished according to several characteristics, such as centralization, coordination and bargaining coverage (Kenworthy 1999). Most relevant in our context is the potential of unions and employer associations to act as standard-setters. Hence, we choose the degree of bargaining coordination as the variable to include. It is taken from the OECD Employment Outlook 2003 (OECD 2004, Table 3.5) and relates to 1995 – 2000. 5. Results The estimation results (ordered probit coefficients and t-values) are displayed for different specifications in Table 4. Table 5 provides additional robustness analysis. The dependent variable is coded such that a negative coefficient implies a higher likelihood of a negative estimated effect on employment. A general-tospecific-approach was used to search for the most appropriate specification. The first column (ALLVARS) contains all available independent variables except those for used to check for policy complementarities, which are added later. Only variables at least coming close to statistical significance (using a p-level of 0.15 as the threshold) are retained in the next specification (INSIG-OUT). This specification is used as the baseline for the next two estimations. In the third specification (DUMMIES), country fixed effects are added. Finally, the last specification (PREFERRED) adds the institutional variables to the baseline specification. Concerning the variables for different population groups, no significant results are found except that studies for low-skilled workers tend to find more negative minimum age effects on employment. The dummy variable for studies related to “small” industries compared to the economy has the expected positive sign, but is not significant. Both micro data and panel data generate more negative (and more significant) employment results than time-series data. Thus, in contrast to the meta-analysis by Card and Krueger (1995b), in our sample more statistical power seems to produce more significant results. The period of observation positively influences the dependent variable, which suggests that the effect of minimum wages has become less negative over time. Since this variable is highly collinear with publication date, an alternative explanation is that studies finding negative employment effect have become scarcer over time. Unfortunately, the data do not allow discriminating between these explanations.

The Combined Employment Effects of Minimum Wages

177

Table 4 Ordered Probit Estimation Results from Meta-Analysis Specification

ALLVARS coeff.

t-stat.

Age less than 19

–0.04

–0.16

Age 19 – 25

–0.22

–0.77

Gender: Men only

0.06

0.23

Gender: Women only

0.11

0.56

Low-skilled workers

–0.60*

–2.26

Low-wage sectors

–0.05

–0.20

Small industry

INSIG-OUT

DUMMIES

PREFERRED

coeff.

t-stat.

coeff.

t-stat.

coeff.

t-stat.

–0.64*

–3.47

–0.64*

–2.82

–0.56*

–3.01

0.60

1.45

0.56

1.44

0.58

1.08

0.70

1.49

Data: Micro

–0.73*

–2.10

–0.59*

–2.34

–0.66*

–2.22

–0.81*

–2.88

Data: Macro panel

–1.29*

–4.51

–1.19*

–4.78

–0.92*

–3.14

–1.16*

–5.18

Observation period

0.05*

2.60

0.05*

2.84

0.03

1.79

0.03*

2.22

Outcome: unemployment

1.23*

2.89

1.43*

4.69

1.24*

2.93

1.25*

3.35

0.30

1.44

0.26

1.32

0.27

1.48

Estimator: DiD

0.00

0.01

–0.22

–0.28

Peer reviewed journal

0.34

1.72

Minimum wage level

0.12

1.28

Estimator: IV

Benefit replacement ratio Employment protection Bargaining coordination Country dummies

NO

NO

YES

0.55*

2.95

–0.53*

–3.35

0.16*

2.37

NO

Studies

55

55

55

Countries

15

15

15

15

304

304

304

304

–354.27

–356.79

–339.23

–346.50

Observations Log-Likelihood Joint significance 2 value

2509.23*

55

15.88*

* denotes entries with a least five percent significance. Standard errors are in parentheses. The excluded groups for the dummy variable groups are: Age: all age groups; Gender: men and women; Data: macro cross-section; Estimator: OLS.

Unemployment as an outcome measure is less negatively affected5 by the minimum wage as compared to employment. The estimation technique (OLS, IV or DiD) does not statistically significantly affect the estimation outcomes. Finally, there is no indication of publication bias in the sense that negative and significant results are published with higher likelihood than insignificant or positive employment effects. Rather, the coefficient is positive, but it is only marginally significant in the baseline specification and loses its significance if further variables are added. 5 The coding ensures that a positive effect on unemployment is coded in the same way as a negative effect on employment.

178

Bernhard Boockmann

According to joint significance tests, country fixed effects are highly significant in statistically explaining the outcomes of minimum wage studies (DUMMIES). Taking the USA as the baseline, the coefficients of the individual dummies have positive signs for most European countries except France and Portugal. The high significance of the country dummies is not self-evident. For instance, in the metaanalysis of active labor market policies by Card et al. (2009), country dummies were found to be insignificant. Thus, the notion that minimum wage effects are heterogeneous across countries is supported by our analysis. In the final specification (PREFERRED), the dummies are replaced by the institutional variables on employment protection, net replacement rates and bargaining coordination. The results suggest that the three dummies used do not fully explain the heterogeneity found in the DUMMIES specification. However, the absolute increase in the log likelihood value is about sixty per cent of the increase achieved by the inclusion of the country dummies. All three variables are statistically significant. Higher benefit replacement rates and union-employer bargaining coordination decrease the negative effects of the minimum wage, while stricter employment protection reinforces the negative employment effects.6 The other variables are only little affected by the inclusion of country fixed effects or institutional variables. Before we turn to interpreting these results, we present some robustness analysis. This is important because several countries contribute only few data points to the sample. Hence, the results may not be stable. We have checked the robustness in several ways (see Table 5 for results). Exclusion of Institutional Variables The institutional variable taken together may describe the countries’ regulation system but it is unclear whether they have a separate impact on the estimated minimum wage effects. Including these variables one-by-one reveals that only the benefit replacement ratio remains significant (specifications BRR, EPL and COORD). However, even here the effect is only significant at the ten per cent level, and it is reduced by two thirds in magnitude as compared to specification PREFERRED. This suggests that the interaction between institutions is more complex than can be captured by bivariate interaction effects. As a consequence, one should be very careful with statements about the precise nature of the institutional interactions.

6 We can also compare the effects quantitatively by looking at the marginal effects. Focussing on the likelihood of a significantly negative employment effect, the marginal effects of a change in the institutional variables by one standard deviation are –0.17 for the benefit replacement ratio, 0.16 for the employment protection indicator and –0.05 for bargaining coordination. Thus, the effects of benefits and employment protection are quantitatively much more pronounced than the effect of coordination. Overall, marginal effects for the probability of a significantly negative outcomes do not differ from coefficient estimates in direction and statistical significance (results are available on request from the author).

BRR

0.25

0.19

Peer reviewed journal

Benefit replacement ratio

15

304

–352.76

Countries

Observations

Log-Likelihood

–356.46

304

15 –355.80

304

15

55

1.23

1.36

4.29

2.92

–5.09

–2.50

1.49

–3.22

–350.71

304

15

55

NO

0.17*

–0.62*

0.61*

0.14

1.36*

0.03*

–0.96*

–0.82*

0.69

–0.72*

2.30

–2.73

2.41

0.78

4.18

2.00

–3.39

–2.61

1.43

–3.22

t-stat.

WEIGHTED coeff.

–408.65

304

15

55

NO

0.14*

–0.52*

0.55*

0.26

1.22*

0.03*

–1.16*

–0.84*

0.78

–0.65*

2.17

–3.27

2.92

1.46

3.31

2.22

–4.80

–2.93

1.62

–3.68

t-stat.

6-CAT coeff.

–228.47

304

15

55

NO

0.15*

–0.44*

0.42*

0.22

1.16*

0.03

–1.33*

0.80*

0.88

–0.39*

2.49

–2.89

2.23

1.08

4.77

1.55

–5.11

–2.36

1.91

–2.07

t-stat.

3-CAT coeff.

2.72*

0.81

15

15

55

0.19*

–0.42*

–0.14

5.25

0.17*

–4.29*

–3.38*

2.35

–2.23

–0.49

0.93

3.07

–5.69

–4.22

4.45

t-stat.

MOULTON coeff.

* denotes entries with a least five percent significance. Standard errors are in parentheses. The excluded groups for the dummy variable groups are: Age: all age groups; Gender: men and women; Data: macro cross-section; Estimator: OLS.

Adjusted R2

55

Studies

55

0.29

1.40*

0.05*

–1.15*

–0.58*

0.56

–0.61*

0.08

0.52

1.35

4.19

2.82

–4.98

–2.35

1.50

–3.06

t-stat.

COORD coeff.

NO

NO

0.05

0.29

1.39*

0.05*

–1.18*

–0.61*

0.58

–0.60*

t-stat.

Country dummies

1.79

1.23

EPL

coeff.

Bargaining coordination

NO

1.25*

Outcome: unemployment

Employment protection

2.77

0.04*

3.25

–5.35

–1.19*

Data: Macro panel

Observation period

1.65

–2.63

t-stat.

–2.87

0.68

–0.51*

coeff.

–0.77*

Data: Micro

Small industry

Low-skilled workers

Specification

Table 5: Further Results from Meta-Analysis (Robustness Analysis)

The Combined Employment Effects of Minimum Wages 179

180

Bernhard Boockmann

Weighted Estimation To avoid giving studies with many different estimates disproportionate influence on the results, estimates may be weighted by the inverse of the number of estimates included, such that the impact of each study on the outcome is the same. Column WEIGHTED contains the results. Weighting does not affect the conclusions substantially. Different Definitions of the Dependent Variable The coding of the dependent variable in four categories is arbitrary. One could argue for a coarser measure in which all insignificant outcomes are joined into one category. Alternatively, a finer measure distinguishing between insignificant and marginally significant estimates may also appear sensible. In the specifications headed 3-CAT and 6-CAT, these two alternatives are tried. Again, the results are robust to these specification changes. Moulton Bias The main independent variables of interest are fixed for countries. This could introduce bias in the standard errors because the iid assumption is violated (Moulton 1986). To check for this possibility, meta-analysis is performed at the country level, using country averages of the dependent and independent variables. Given that the original dependent variable is ordinal, this imposes further structure on the data. Moreover, with only 15 observations the sample becomes very small. The results (headed MOULTON) are similar to the results with estimates as data points. Employment protection increases the negative impact of minimum wages on employment, union centralization reduces it. The benefit replacement ratio, however, becomes insignificant. The other variables, as far as collinearity problems do not require their exclusion, give similar results as found previously. 6. Interpretation and Conclusions A first conclusion from the results of the meta-analysis is that the notion that the employment effects of minimum wages are heterogeneous between countries (Neumark and Wascher 2004a 2004b; König and Möller 2008) receives strong support from the data. Even after controlling for study characteristics, country-specific institutions or, alternatively, country dummies remain strongly significant. This is not self-evident; for instance, in the meta-analysis by Card et al. (2009) on active labor market policy, country dummies are jointly insignificant. As a conclusion, differences in country characteristics such as the institutional framework should be considered before any conclusions from other countries’ experiences with minimum wages are drawn. The three institutional variables included to capture interactions with other regulations influence the estimated employment effects of the minimum wage in plau-

The Combined Employment Effects of Minimum Wages

181

sible ways. The benefit replacement ratio weakens any negative employment effects. This is consistent with strategic substitutability between these instruments: if one is in place, it limits the harmful effect of the other. Stricter employment protection enhances the negative employment effect of minimum wages. This result is in contrast to the finding by Neumark and Wascher (2004a). The discrepancy can perhaps be reconciled if the result of Neumark and Wascher is interpreted as a short-run effect. In the short run, employment losses due to minimum wages may be reduced by strict employment protection. This short-run effect will, however, disappear over time, while the additional costs to employers from having both minimum wages and employment protection will persist. Unfortunately, our data does not allow separating the estimations that represent short-run or long-run effects. Finally, bargaining coordination seems to reduce the employment effects of minimum wages. However, the effect is not particularly pronounced quantitatively. While the estimated effects appear plausible, their robustness is not beyond doubt. Although minimum wage research has expanded massively over the last 15 years, the number of studies currently available does not allow for more detailed analysis. Moreover, the complementarity of minimum wage effects with other country characteristics, such as the tax structure or family values, remains an open question. Thus, while the results of this study strongly warn against the unconsidered application of other countries’ experiences with minimum wages in the political discourse, more research is needed on the interaction between minimum wages and other institutions. References Aghion, P. / Algan, Y. / Cahuc, P. (2008): “Can policy interact with culture? Minimum wage and the quality of labor relations,” IZA Discussion Paper No. 3680. Alesina, A. / Algan, Y. / Cahuc, P. / Giuliano, P. (2010): “Family values and the regulation of Labor,” IZA Discussion Paper No. 4747. Amable, B. / Gatti, D. (2006): “Labor and product market reforms: questioning policy complementarity,” Industrial and Corporate Change 15, 101 – 122. Bassanini, A. / Duval, R. (2006): “The Determinants of unemployment across OECD countries,” OECD Economic Studies 42, 7 – 86. – (2009): “Unemployment, institutions, and reform complementarities: re-assessing the aggregate evidence for OECD countries,” Oxford Review of Economic Policy 25, 40 – 59. Belot, M. / Boone, J. / van Ours, J. C. (2007): “Welfare-improving employment protection,” Economica 74, 381 – 396. Belot, M. / van Ours, J. C. (2004): “Does the recent success of some OECD countries in lowering their unemployment rates lie in the clever design of their labor market reforms?,” Oxford Economic Papers 56, 621 – 642. Cahuc, P. / Zylberberg, A. (1999): “Job protection, minimum wage and unemployment,” IZA Discussion Paper No. 95.

182

Bernhard Boockmann

Cahuc, P. / Saint-Martin, A. / Zylberberg, A. (2001): “The consequences of the minimum wage when other wages are bargained over,” European Economic Review 45, 337 – 352. Card, D. / Kluve, J. / Weber, A. (2009): “Active labor market policy evaluations: a meta-analysis,” IZA Discussion Paper No. 4002. Card, D. / Krueger, A. B. (1995a): Myth and Measurement: The New Economics of the Minimum Wage, Princeton University Press. – (1995b): “Time-series minimum-wage studies: a meta-analysis,” American Economic Review 85, 238 – 43. Coe, D. T. / Snower, D. J. (1997): “Policy complementarities: the case for fundamental labor market reform,” CEPR Discussion Paper No. 1585. Eichhorst, W. / Konle-Seidl, R. (2005): “The interaction of labor market regulation and labor market policies in welfare state reform,” IZA Discussion Paper No. 1718. Elmeskov, J. / Martin, J. / Scarpetta, S. (1998): “Key lessons for labour market reforms: evidence from OECD countries’ experiences,” Swedish Economic Policy Review 5, 205 – 52. Greenberg, D. / Ashworth, K. / Cebulla, A. / Walker, R. (2005): “When welfare-to-work programs seem to work well: explaining why Riverside and Portland shine so brightly,” Industrial and Labor Relations Review 59, 34 – 50. Jarrell, S. B. / Stanley, T. D. (1990): “A meta-analysis of the union-nonunion wage gap,” Industrial and Labor Relations Review 44, 54 – 67. Kenworthy, L. (2001): “Wage-setting measures,” World Politics 54: 57 – 98. Kluve, J. (2006): “The effectiveness of European active labor market policy”, IZA Discussion Paper No. 2018. Lazear, E. (1990): “Job security provisions and employment,” Quarterly Journal of Economics 105, 699 – 726. Longhi, S. / Nijkamp, P. / Poot, J. (2005): “A meta-analytic assessment of the effect of immigration on wages,” Journal of Economic Surveys 19, 451 – 477. Moulton, B. (1986): “Random group effects and the precision of regression estimates,” Journal of Econometrics 32, 385 – 397. Neumark, D. / Wascher, W. (1998): “Is the time-series evidence on minimum wage effects contaminated by publication bias?”, Economic Inquiry 36, 458 – 470. – (2004a): “Minimum wages, labor market institutions, and youth employment: a cross-national analysis,” Industrial and Labor Relations Review 57, 223 – 248. – (2004b): “The influence of labour market institutions on the disemployment effect of the minimum wage,” CESifo DICE Report 2, 40 – 47. – (2007): “Minimum wages and employment,” IZA Discussion Paper No. 2570. – (2008): Minimum Wages, Cambridge (MA), MIT Press. OECD (2005): OECD Employment Outlook 2004. Paris, OECD. – Benefits and Wages. OECD Indicators Paris, OECD. Orszag, M. / Snower, D. (1998): “Anatomy of policy complementarities,” Swedish Economic Policy Review 5, 303 – 343.

The Combined Employment Effects of Minimum Wages

183

Pereira, P. T. / Martins, P. S. (2004): “Returns to education and wage equations,” Applied Economics 36, 525 – 31. Stanley, T. D. (2001): “Wheat from chaff: meta-analysis as quantitative literature review,” Journal of Economic Perspectives 15, 131 – 150. Stanley, T. D. / Jarrell, S. B. (2005): “Meta-regression analysis: a quantitative method of literature surveys,” Journal of Economic Surveys 19, 299 – 308. Weichselbaumer, D. / Winter-Ebmer, R. (2005): “A meta-analysis of the international gender wage gap,” Journal of Economic Surveys 19, 479 – 511. – (2007): “The effects of competition and equal treatment laws on gender wage differentials,” Economic Policy 22, 235 – 287.

Appendix: Studies Included in the Sample 1. Abowd, J. M. / Kramarz, F. / Margolis, D. N. (1999): “Minimum wages and employment in France and the United States,” NBER Working Paper 6996. 2. Abowd, J. M. / Kramarz, F. / Margolis, D. N. / Phillipon, T. (2000): “The tail of two countries: minimum wages and employment in France and the United States,” IZA Discussion Paper No. 203. 3. Abowd, J. M. / Kramarz, F. / Lemieux, T. / Margolis, D. N. (1997): “Minimum wages and youth employment in France and the United States,” NBER Working Paper 6111. 4. Addison, J. T. / Blackburn, M. L. / Cotti, C. D. (2008): “The effect of minimum wages on wages and employment–county-level estimates for the United States,” IZA Discussion Paper No. 3300. 5. Addison, J. T. / Blackburn, M. L. / Cotti, C. D. (2009): “Do minimum wages raise employment? Evidence from the U.S. retail-trade sector,” Labour Economics 16, 397 – 408. 6. Baker, M. / Benjamin, D. / Stanger, S. (1999): “The highs and lows of the minimum wage effect: a time-series cross-section study of the Canadian law,” Journal of Labor Economics 17, 318 – 350. 7. Bazen, S. / Marimoutou, V. (2002): “Looking for a needle in a haystack? A re-examination of the time series relationship between teenage employment and minimum wages in the United States,” Oxford Bulletin of Economics and Statistics, 64, 699 – 725. 8. Bernstein, J. / Schmitt, J. (2000): “The impact of the minimum wage: policy lifts wages, maintains floor for low-wage labor market,” Economic Policy Institute Briefing Paper. 9. Böckerman, P. / Uusitalo, R. (2007): “Minimum wages and youth employment—evidence from the Finnish retail trade sector,” Labour Institute for Economic Research Discussion Papers No. 238. 10. Brenner, M. D. / Wicks-Lim, J. / Polin R. (2002): “Measuring the impact of living wage laws: a critical appraisal of David Neumark’s How Living Wage Laws Affect Low-Wage Workers and Low-Income Families,” Working Paper 43, Political Economy Research Institute.

184

Bernhard Boockmann

11. Burkhauser, R. V. / Couch, K. V. / Wittenburg, D. C. (2000): “A reassessment of the new economics of the minimum wage literature with monthly data from the Current Population Survey,” Journal of Labor Economics 18, 653 – 680. 12. Campolieti, M. / Gunderson, M. / Riddell, C. (2006): “Minimum wage impacts from a prespecified research design: Canada 1981 – 1997,” Industrial Relations 45, 195 – 216. 13. Campolieti, M. / Fang, T. / Gunderson, M. (2005a): “How minimum wages affect schooling-employment outcomes in Canada, 1993 – 1999,” Journal of Labor Research 26, 169 – 182. 14. Campolieti, M. / Fang, T. / Gunderson, M. (2005b): “Minimum wage impacts on youth employment transitions, 1993 – 1999,” Canadian Journal of Economics 38, 81 – 104. 15. Card, D. / Krueger, A. B. (2000): “Minimum wages and employment: a case study of the fast-food industry in New Jersey and Pennsylvania: reply,” American Economic Review 90, 1397 – 1420. 16. Chapman, J. (2004): “Employment and the minimum wage: evidence from recent State labor market trends,” Economic Policy Institute Briefing Paper. 17. Currie, J. / Fallick, B. (1996): “The minimum wage and the employment of youth: evidence from the NLSY,” Journal of Human Resources 31, 404 – 428. 18. Deere, D. / Murphy, K. M. / Welch, F. (1995): “Employment and the 1990 – 1991 minimum-wage hike,” American Economic Review Papers and Proceedings 85, 232 – 237. 19. Dickens, R. / Draca, M. (2005): “The employment effects of the October 2003 increase in the national minimum wage” Report prepared for the Low Pay Commission. 20. Dickens, R. / Machin, S. / Manning, A. (1999): “The effects of minimum wages on employment: theory and evidence,” Journal of Labor Economics 17, 1 – 22. 21. Dolado, J. / Kramarz, F. / Machin, S. / Manning, A. / Margolis, D. / Teulings, C. / SaintPaul, G. / Keen, M. (1996): “The economic impact of minimum wages in Europe,” Economic Policy 11, 317 – 372. 22. Eriksson, T. / Pytlikova, M. (2004): “Firm-level consequences of large minimum-wage increases in the Czech and Slovak Republics,” Labour 18, 75 – 103. 23. Hyslop, D. / Stillman, S. (2004): “Youth minimum wage reform and the labour market.” IZA Discussion Paper No. 1091. 24. Hyslop, D. / Stillman, S. (2007): “Youth minimum wage reform and the labour market in New Zealand,” Labour Economics 14, 201 – 230. 25. Kennan, J. (1995): “The elusive effects of minimum wages,” Journal of Economic Literature 33, 1950 – 1965. 26. Leigh, A. (2003): “Employment effects of minimum wages: evidence from a quasi-experiment,” Australian Economic Review 36, 361 – 373. 27. Machin, S. / Manning, A. / Rahman, L. (2003). “Where the minimum wage bites hard: introduction of minimum wages to a low wage sector,” Journal of the European Economic Association 1, 154 – 180. 28. McDonald, J. T. / Myatt, A. E. (2004): “The minimum wage effect on youth employment in Canada: testing the robustness of cross-province panel studies”, mimeo, Department of Economics, University of New Brunswick.

The Combined Employment Effects of Minimum Wages

185

29. Möller, J. / König, M. (2008): “Mindestlohneffekte des Entsendegesetzes? Eine Mikrodatenanalyse für die deutsche Bauwirtschaft,” Zeitschrift für ArbeitsmarktForschung 41, 327 – 346. 30. Neumark, D. / Wascher, W. (2000): “The effect of New Jersey’s minimum wage increase on fast-food employment: a reevaluation using payroll records,” American Economic Review 90, 1362 – 1296. 31. Neumark, D. / Wascher, W. (2002): “State-level estimates of minimum wage effects: new evidence and interpretations from disequilibrium models,” Journal of Human Resources 37, 35 – 62. 32. Neumark, D. / Schweitzer, M. / Wascher, W. (2004): “Minimum wage effects throughout the wage distribution,” Journal of Human Resources 39, 425 – 450. 33. Neumark, D. / Nizalova, O. (2004): “Minimum wage effects in the longer run,” NBER Working Paper 10656. 34. Neumark, D. (2001): “The employment effects of minimum wages: evidence from a prespecified research design,” Industrial Relations 40, 121 – 144. 35. Neumark, D. (2002): How Living Wage Laws Affect Low-Wage Workers and Low-Income Families,” San Francisco, Public Policy Institute of California. 36. O’Neill, D. / Nolan, B. / Williams, J. (2006): “Evaluating the introduction of a national minimum wage—evidence from a new survey of firms in Ireland,” Labour 20, 63 – 90. 37. Orazem, P. F. / Mattila, J. P. (2002): “Minimum wage effects on hours, employment, and number of firms: the Iowa case,” Journal of Labor Research 23, 3 – 23. 38. Pabilonia, S. W. (2002): “The effects of federal and state minimum wages upon teen employment and earnings,” mimeo, Bureau of Labor Statistics. 39. Partridge, M. D. / Partridge, J. S. (1998): “Are teen unemployment rates influenced by State minimum wage laws?,” Growth and Change 29, 359 – 382. 40. Partridge, M. D. / Partridge, J. S. (1999a): “Do minimum wage hikes raise US long-term unemployment? Evidence using State minimum wage rates,” Regional Studies 38, 713 – 726. 41. Partridge, M. D. / Partridge, J. S. (1999b): “Do minimum wage hikes reduce employment? State-level evidence from the low-wage retail sector,” Journal of Labor Research 20, 393 – 413. 42. Pereira, S. C. (2003): “The impact of minimum wages on youth employment in Portugal,” European Economic Review 47, 229 – 244. 43. Powers, E. T. (2009): “The impact of minimum-wage increases—evidence from fast-food establishments in Illinois and Indiana,” Journal of Labor Research 30, 365 – 394. 44. Ragacs, C. (2002). “On the empirics of minimum wages: stylized facts for Austria,” Working Paper, Vienna University of Economics & B.A. 45. Sabia, J. J. (2006): “The effect of minimum wage increases on retail and small business employment,” mimeo, Employment Policies Institute. 46. Sabia, J. J. (2009): “The effects of minimum wage increases on retail employment and hours—new evidence from monthly CPS data,” Journal of Labor Research 30, 75 – 97.

186

Bernhard Boockmann

47. Skedinger, P. (2006): “Minimum wages and employment in Swedish hotels and restaurants,” Labour Economics 13, 259 – 90. 48. Stewart, M. B. (2002): “Estimating the impact of the minimum wage using geographical wage variation,” Oxford Bulletin of Economics & Statistics 64, 583 – 605. 49. Stewart, M. B. (2003): “Modelling the employment effects of the minimum wage,” Final Report to the Low Pay Commission. 50. Stewart, M. B. (2004a): “The impact of the introduction of the U.K. minimum wage on the employment probabilities of low-wage workers,” Journal of the European Economic Association 2, 67 – 97. 51. Stewart, M. B. (2004b): “The employment effects of the national minimum wage,” Economic Journal 114, C110 – 116. 52. Wessels, W. (2007): “A reexamination of Card and Krueger’s state-level study of the minimum wage,” Journal of Labor Research 28, 135 – 146. 53. Williams, N. / Mills, J. A. (2001): “The minimum wage and teenage employment: evidence from time series,” Applied Economics 33, 285 – 300. 54. Yuen, T. (2003): “The effect of minimum wages on youth employment in Canada: a panel study,” Journal of Human Resources 38, 647 – 672. 55. Zavodny, M. (2000). “The Effect of the minimum wage on employment and hours,” Labour Economics 7, 729 – 750.

The Combined Employment Effects of Minimum Wages and Labor Market Regulation—a Meta-Analysis Comment By Alexandra Spitz-Oener*

Boockmann contributes to an interesting area of labor market research that emphasizes the importance of the institutional background against which policy interventions occur. Specifically, he analyzes to what extent labor market institutions determine the employment consequences of minimum wages. He conducts a metaanalysis that combines the results of 55 studies on the employment effects of minimum wages in 15 industrialized countries. The main explanatory variables in this study are measures for a country’s benefit replacement ratio, its employment protection legislation and the extent of the collective bargaining system. While these measures are able to explain some of the heterogeneity in the employment consequences of minimum wages across countries that we observe, one conclusion to draw from this study is that a large part of the explanatory power comes from country fixed-effects, that is, country specific factors that are unobserved in the empirical analysis. This is a general criticism of cross-country studies and as such not very constructive. However, I use this comment to suggest the degree of product market regulation as an additional explanatory variable that should be included in the specification and that can be easily obtained by recent OECD publications. In addition, the specification should also include an interaction term between labor market regulation and product market regulation, as both types of regulation may influence each other when it comes to the employment consequences of minimum wages. Williamson (1968) was the first to argue that wage rates have an impact on the product market as they can be used as barriers to entry (more recent contributions along these lines are Haucap et al. 2001, Heitzler and Wey 2010). The economic model is motivated by the Supreme Court decision United Mine Workers vs. Pennington (UMW vs. P), that discussed whether the union had colluded with a group of employers in the bituminous coal industry to impose a uniform wage rate throughout the industry in order to establish a barrier to firm entry or force small rivals to abandon the industry. * Address: Humboldt University Berlin, Spandauer Str. 1, 10178 Berlin, Germany, e-mail: [email protected], IAB, CASE, and IZA.

188

Alexandra Spitz-Oener

Let me briefly go through the intuition of the model: Wage rates can act as barriers to firm entry if increases in wages produce differential shifts in average production cost across firms. If there are, for example, systematic differences in production techniques between small and large firms, so that increases in wages increase average cost to a larger extend in small then in large firms, an increase in wages will force small firms to leave the market or prevent potential new (small) firms from entering. Systematic differences in production techniques arise, for example, when small scale operations produce with a larger labor to capital ratio than large scale production entities. Owing to the larger labor intensity of small scale producers, average cost increases brought about by the wage increases are larger for the small producers than for the large producers. For this reason, large firms will find an increase in wages attractive. While it is beyond the scope of this study to determine the product market consequences of minimum wages, I think that its appeal would benefit from an extension of the set of institutions that are considered in the analysis along the lines suggested. Depending how regulated product markets already are in countries, minimum wage might be differentially effective in regulating the product market. Product market regulation, ceteris paribus, hampers competition and generates rents, and it then depends on the bargaining power of workers (represented by labor market institutions) to what extend firms share part of the rents with workers (Haucap et al. 2001, Blanchard and Giavazzi 2003). In this institutional setting, the employment consequences of minimum wages are a priori not clear.

References Haucap, J. / Pauly, U. / Wey, C. (2001): Collective Wage Setting when Wages are generally binding: An Antitrust Perspective, International Review of Law and Economics 21(3), 287 – 307. Heitzler, S. / Wey, C. (2010): Raising Rivals’ Fixed (Labor) Costs: The Deutsche Post Case, DIW DP 1008. Williamson, O. E. (1968): Wage Rates as Barriers to Entry: The Pennington Case in Perspective, Quarterly Journal of Economics 82(1), 85 – 116.

Participants Franziska Bremus

DIW Berlin

Rolf Bürkl

GfK SE

Martin Clemens

Institute for the Study of Labor, Bonn

Irwin Collier

Freie Universität Berlin

Tarnike Darjania Knut Degenhart Martin Dietz

Institute for Employment Research, Nuremberg

Werner Eichhorst

Institute for the Study of Labor, Bonn

Heinz Engelstädter

IWVWW, Berlin

Rainer Ertel

Niedersächsisches Institut für Wirtschaftsforschung

Michael Feil

Institute for Employment Research, Nuremberg

Joachim Fischer

HTW Berlin

Wolfgang Franz

Centre for European Economic Research, Mannheim

Luise Görges

Berlin School of Economics and Law

Dominik Groll

Kiel Institute for the World Economy

Hendrik Hagedorn

DIW Berlin

Bernhard Heitzer

Federal Ministry of Economics and Technology

Alexander Herzog-Stein

Hans Böckler Foundation, Düsseldorf

Volker Hofmann

Association of German Banks

Jürgen Jerger

University of Regensburg

Liliane Jodkowski Ulf Kadritzke

Berlin School of Economics and Law

Andrea Kargus

Institute for Employment Research, Nuremberg

Willi Koll

Federal Ministry of Economics and Technology

Stefan Klinger

Workers’ lobby of the CDU / CSU Bundestag group

Thomas Köhler

Federal Ministry of Finance

Bettina Kudla

MdB

Harald Kuhne

Federal Ministry of Economics and Technology

Stefanie Knoth

Federal Ministry of Finance

Camille Logeay

Hans Böckler Foundation, Düsseldorf

190

Participants

Susanne Marcus

DIW Berlin

Ralf Messer

ARGE

Antje Mertens

Berlin School of Economics and Law

Heinrich Milosiu

Fraunhofer IIS Erlangen

Joachim Möller

Institute for Employment Research, Nuremberg

Klaus Müller

Federal Ministry of Economics and Technology

Jürgen Müller

Berlin School of Economics and Law

Nadja Neder

Institute for Employment Research, Nuremberg

Guido Raddatz

Market Economy Foundation

Steffen Roth

Institute for Economic Policy, University of Cologne

Andreas Sachs

Centre for European Economic Research, Mannheim

Wolfgang Scheremet

Federal Ministry of Economics and Technology

Christof Schiller

Hertie School of Governance

Jens Schmidt-Ehmcke

DIW Berlin

Andreas Schmidt-Rögnitz

University of Applied Sciences, Berlin

Claus Schnabel

University of Erlangen-Nuremberg

Christian Schneider

Office of Uwe Kekeritz, Member of the Bundestag

Christine Maria Sentz

Workers’ lobby of the CDU / CSU Bundestag group

Benjamin Sommer

Federal Ministry of Economics and Technology

Dennis Snower

Kiel Institute for the World Economy

Alexandra Spitz-Oener

Humboldt-Universität zu Berlin

Michael Stops

Institute for Employment Research, Nuremberg

Ulrich Walwei

Institute for Employment Research, Nuremberg

Klaus F. Zimmermann

DIW Berlin

Florian Zinsmeister

Federal Ministry of Economics and Technology