Abstract We analyse the robust performance of the German labour market in the Great Recession, and investigate to what extent cyclical reductions in productivity and working time cushioned employment losses. We present stylized facts and apply time-series techniques to estimate counterfactual developments. Our results show that the magnitude of temporary working-time reductions was extraordinarily pronounced, whereas cyclical reductions in hourly productivity were in line with historical evidence. Using detailed information on instruments for the adjustment of working time, we uncover the institutional mechanisms behind this strong reduction. While short-time work played a significant role, even more important were working-time accounts and discretionary variations in regular working time, two new instruments which gained widespread use in the decade before the Great Recession. 1. Introduction In 2008 and 2009 Germany experienced the deepest recession in its post-war history. Real GDP dropped by more than 6% from its cyclical peak in the first quarter of 2008 to its trough in the third quarter of 2009 (Fig. 1). This output shock was also large compared to the experiences in other countries in the Great Recession; in the USA output declined by 4.2% from peak to trough, in the Euro Area as a whole by 5.5%.1 But unlike in previous recessions or other countries, this sharp decline in German output triggered no significant job losses. In fact, the total number of persons in employment was even higher in the trough of the recession than at the pre-recession peak (Fig. 1), while unemployment was lower. This remarkable stability of the German labour market has been termed ‘labour market miracle’ (e.g. Krugman, 2009; Möller, 2010). Fig. 1. View largeDownload slide Real GDP and employment in the Great Recession, 2008q1 = 100 Notes: Seasonally adjusted quarterly values, indexed to 2008q1 = 100. Employment is the total number of persons in employment. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations Fig. 1. View largeDownload slide Real GDP and employment in the Great Recession, 2008q1 = 100 Notes: Seasonally adjusted quarterly values, indexed to 2008q1 = 100. Employment is the total number of persons in employment. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations This paper contributes to the existing literature by presenting stylized facts, econometric evidence, and a discussion of the institutional instruments responsible for this development. We compare the Great Recession to other major recessions in Germany since 1970 and analyse to what extent reductions in hourly productivity and working time cushioned the recessions’ impact on employment. We find that cyclical reductions in hourly productivity played a significant role in safeguarding employment in all recessions, while temporary working-time reductions were historically less pronounced and occur irregularly. In the Great Recession, however, firms amply used them to hoard labour. We further conduct a time-series analysis and find that the development of cyclical hourly productivity in the upswing before, during, and after the Great Recession is well predictable using historical data, while the increase in working time in the upswing before the recession and its reduction at the beginning of the recession are much stronger than forecasted. Our findings suggest that both pro-cyclical variations in hourly productivity and working time strongly contributed to safeguarding jobs in the Great Recession. However, only the latter was extraordinarily pronounced compared to earlier business cycles, preventing an additional 660,000 job losses from peak to trough, which amounts to 1.6% of total employment. Using detailed information on instruments for the adjustment of working time, we uncover the institutional mechanisms behind this strong reduction. While short-time work played a significant role, even more important were working-time accounts and discretionary variations in regular working time, two new instruments which have become broadly available in the decade before the Great Recession. This paper is structured as follows: Section 2 compares the development of GDP, employment, productivity per hour, and hours worked per person in employment in post-war downturn-periods in Germany. Section 3 provides econometric evidence of the importance of cyclical reductions in productivity and working time to safeguard employment in the Great Recession. Section 4 discusses in more detail the institutional foundations of this labour market miracle. Section 5 relates our findings to the previous literature. Section 6 concludes. 2. Safeguarding employment in downturns: a historical comparison In his classic article, Arthur Okun (1962) established what subsequently became known as ‘Okun’s law’: the relation between changes in GDP and changes in unemployment. Usually there is no one-to-one relation between output and unemployment (for a recent assessment see Ball et al., 2013), thus other factors must buffer output losses. These buffers are average hours worked per person in employment and hourly labour productivity, as can be shown using a national accounting identity (e.g. Gordon, 1993): Real GDP (Y) is the product of the number of persons in employment (E), multiplied by their working time (WT), i.e. the number of effective hours worked per person in employment, and real productivity per hour worked by persons in employment (LP). Expressed in growth rates (g) and solved for employment, this gives:2 gE≈gY−gWT−gLF (1) In a mechanical sense, therefore, a strong drop in output with employment remaining constant, as in Germany during the Great Recession (see Fig. 1), implies that either working-time or labour productivity, or both, decrease significantly. In a first assessment of the extent of labour hoarding in the Great Recession, we apply this accounting identity to compare the development of GDP, employment, hourly productivity, and working hours per person in employment in major German recessions for which quarterly data are available.3 The comparison with earlier periods sheds light on the question as to whether employment, working time, and/or hourly productivity reacted in an exceptional way in the Great Recession compared to previous downturns. It is crucial to account for trend growth rates in such an assessment (see e.g. Möller, 2010; Merkl and Wesselbaum, 2011; Ohanian and Raffo, 2012; Ball et al., 2013), since otherwise recessions in a period of high trend growth would seem much less severe than recessions in an environment of low potential output growth. Similarly, the stabilizing impact of productivity or working time would be overstated for recessions during times of weak trend productivity increases, or strong trend declines in working time.4 Equation (1) also holds for trend growth rates ( g¯). Hence the cyclical deviations in employment growth, that is the difference between actual employment growth and its trend growth, denoted by g^E, can be expressed in terms of trend-deviations of GDP growth, working-time growth, and hourly labour productivity growth: g^E≡(gE−g¯E)≈g^Y−g^WT−g^LP=(gY−g¯Y)−(gWT−g¯WT)−(gLP−g¯LP) (2) For our assessment of German recessions, we apply seasonally adjusted quarterly data available from the national accounts from 1970q1 onwards (see Online Appendix A1 for details).5 Throughout this paper we calculate the trend of all variables by applying the Hodrick-Prescott filter with the standard smoothing parameter of λ = 1,600 for quarterly data (e.g. Ohanian and Raffo, 2012).6 In order to allow for breaks in the data because of German reunification, we compute different trends before and after 1991q1.7 To determine economic downturns, we apply the business cycle dating procedure developed by the German Council of Economic Experts.8 Using this procedure, we identify six downturns since 1970. However, we do not consider the downturn beginning in 1985, since the German Council of Economic Experts did not recognize it as a ‘pronounced economic downturn’.9 We omit the downturn that began in 1991 because of data problems due to German reunification. Our sample therefore contains four major recessions (Table 1), among them those due to the oil price shocks in the 1970s, as well as the long economic downswing of the first half of the 2000s, and the Great Recession. The recession of the early 1970s is of special interest for comparison as it was the most severe economic decline in Germany’s post-war history before 2008. Overall, two West German slumps and two slumps of the unified German economy are included in the analysis. Table 1 Economic downturns Peak Trough Rate of change of GDP in % Cyclical rate of change of GDP in % Downturn I 1973q2 1975q2 −0.5 −5.6 Downturn II 1979q4 1982q4 −0.9 −5.0 Downturn III 2001q1 2005q2 0.8 −4.0 Downturn IV 2008q1 2009q3 −6.1 −7.4 Peak Trough Rate of change of GDP in % Cyclical rate of change of GDP in % Downturn I 1973q2 1975q2 −0.5 −5.6 Downturn II 1979q4 1982q4 −0.9 −5.0 Downturn III 2001q1 2005q2 0.8 −4.0 Downturn IV 2008q1 2009q3 −6.1 −7.4 Notes: Business cycle dates are determined with the business cycle dating procedure developed by the German Council of Economic Experts. All variables are seasonally adjusted quarterly values. ‘Cyclical’ refers to the difference of actual and trend GDP growth, where trend GDP is constructed applying the Hodrick-Prescott filter with λ = 1,600. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations Table 2 applies the decomposition of employment growth according to eq. (2) to the four downturn episodes. The table shows the actual, trend, and cyclical change of every component in eq. (2) from peak to trough in each of the four downturns. The change in these variables is reported both in growth rates and in persons employed. While we are interested in the cyclical change of these variables, for completeness we also report actual and trend changes. Table 2 Contributions to safeguarding employment in downturns Downturn I Downturn II Downturn III Downturn IV 1973q2–1975q2 1979q4–1982q4 2001q1–2005q2 2008q1–2009q3 Rate of Persons Rate of Persons Rate of Persons Rate of Persons change in 1000 change in 1000 change in 1000 change in 1000 E Actual −3.3% −901 −0.3% −78 −1.4% −567 0.3% 113 Trend −0.8% −230 1.2% 323 0.5% 192 1.3% 508 Cycle −2.5% −671 −1.5% −401 −1.9% −759 −1.0% −395 Y Actual −0.5% −131 −0.9% −240 0.8% 304 −6.1% −2,451 Trend 5.2% 1,401 4.2% 1,132 4.7% 1,875 1.3% 534 Cycle −5.6% −1,532 −5.0% −1,372 −4.0% −1,570 −7.4% −2,984 LP Actual 7.0% 1,911 3.3% 895 4.6% 1,822 −2.6% −1,047 Trend 8.9% 2,417 5.8% 1,579 6.8% 2,687 0.8% 303 Cycle −1.9% −507 −2.5% −685 −2.2% −865 −3.4% −1,350 WT Actual −3.8% −1,041 −3.8% −1,024 −2.3% −897 −3.9% −1,549 Trend −2.7% −722 −2.6% −715 −2.4% −959 −0.7% −274 Cycle −1.2% −319 −1.1% −308 0.2% 62 −3.2% −1,275 Downturn I Downturn II Downturn III Downturn IV 1973q2–1975q2 1979q4–1982q4 2001q1–2005q2 2008q1–2009q3 Rate of Persons Rate of Persons Rate of Persons Rate of Persons change in 1000 change in 1000 change in 1000 change in 1000 E Actual −3.3% −901 −0.3% −78 −1.4% −567 0.3% 113 Trend −0.8% −230 1.2% 323 0.5% 192 1.3% 508 Cycle −2.5% −671 −1.5% −401 −1.9% −759 −1.0% −395 Y Actual −0.5% −131 −0.9% −240 0.8% 304 −6.1% −2,451 Trend 5.2% 1,401 4.2% 1,132 4.7% 1,875 1.3% 534 Cycle −5.6% −1,532 −5.0% −1,372 −4.0% −1,570 −7.4% −2,984 LP Actual 7.0% 1,911 3.3% 895 4.6% 1,822 −2.6% −1,047 Trend 8.9% 2,417 5.8% 1,579 6.8% 2,687 0.8% 303 Cycle −1.9% −507 −2.5% −685 −2.2% −865 −3.4% −1,350 WT Actual −3.8% −1,041 −3.8% −1,024 −2.3% −897 −3.9% −1,549 Trend −2.7% −722 −2.6% −715 −2.4% −959 −0.7% −274 Cycle −1.2% −319 −1.1% −308 0.2% 62 −3.2% −1,275 Notes: E are persons in employment, Y is real GDP, LP is real hourly labour productivity, WT is average working time. ‘Trend’ is constructed applying the Hodrick-Prescott filter with λ = 1,600. ‘Cycle’ is the difference between ‘Actual’ and ‘Trend’. Persons in 1,000 was calculated by multiplying the respective rate of change with the actual employment level at the beginning of the recession. Deviations from the accounting identity in eq. (2) are due to the following points: individual trends of each time series are calculated without taking into account eq. (2), each time series in the German national accounts is individually seasonally adjusted, and rounding differences. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. Table 2 shows that cyclical GDP strongly decreased in all economic downturns. In the first downturn, the cyclical decline of GDP from peak to trough is 5.6%, in the second 5%, in the third 4%, and in the fourth—the Great Recession—7.4%. In all recessions, the cyclical decline in hourly productivity strongly mitigated the effects of the downturn on employment. In the first downturn it saved 507,000 jobs or 1.9% of total employment. In Downturn II it saved 685,000 jobs or 2.5% of employment, in Downturn III 865,000 jobs or 2.2% of employment were saved, and in the Great Recession 1.35 million jobs or 3.4% of employment were saved. The contributions of working-time reductions to safeguarding employment were typically weaker and more infrequent. They saved 319,000 jobs or 1.2% of employment in the first downturn, and 308,000 or 1.1% in the second downturn. In the third downturn, cyclical working time even marginally increased and did not contribute to safeguarding jobs. In Downturn IV, however, cyclical working-time reductions safeguarded 1.28 million jobs or 3.2% of employment. In Table 3 we examine the relative contribution of changes in working time and hourly productivity to safeguarding jobs. We calculate the percentage shares of jobs destroyed due to cyclical output reductions as well as those saved by temporary reductions in working time and productivity. The percentage shares of jobs saved are obtained by dividing the cyclical change in persons employed due to LP and WT by the cyclical changes in employed people due to changes in GDP (as reported in Table 2). The share of jobs destroyed is 100% minus the sum of the LP and WT shares. Table 3 Relative contributions of cyclical productivity and working time to safeguarding employment in downturns Downturn I Downturn II Downturn III Downturn IV Share of cyclical reduction in output translating into employment loss: 46.1% 27.6% 48.9% 12.0% Share of cyclical reduction in output not translating into employment loss: Total 53.9% 72.4% 51.1% 88.0% of which due to cyclical reductions in: LP 33.1% 49.9% 55.1% 45.2% WT 20.8% 22.5% −3.9% 42.7% Downturn I Downturn II Downturn III Downturn IV Share of cyclical reduction in output translating into employment loss: 46.1% 27.6% 48.9% 12.0% Share of cyclical reduction in output not translating into employment loss: Total 53.9% 72.4% 51.1% 88.0% of which due to cyclical reductions in: LP 33.1% 49.9% 55.1% 45.2% WT 20.8% 22.5% −3.9% 42.7% Notes: LP is real hourly labour productivity, WT is average working time. ‘Cyclical’ refers to the difference of actual and trend values, where the trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Source: own calculations. About 30% to 50% of cyclical reductions in output were translated into job-losses in Downturns I to III, while in Downturn IV the corresponding value was only 12%. In all four recessions, between one third and more than half of the output shock was buffered by pro-cyclical productivity. Its relative contribution to safeguarding employment was highest in Downturns II and III and not exceptionally strong in Downturn IV. Cyclical reductions in hours worked safeguarded a much smaller share of employment than cyclical reductions in productivity in Downturns I to III. In the Great Recession, however, cyclical working-time reductions cushioned more than two fifth of the output shock, comparable to the contribution of cyclical reductions in productivity. Our preliminary conclusion from these stylized facts is that the cyclical reduction in working time played a significant role in safeguarding employment during the Great Recession. The same is true for labour productivity, but with the important difference that the relative magnitude of its cyclical decrease is in line with its development in earlier downturns. 3. Econometric evidence on labour hoarding in the Great Recession In this section, we pursue the analysis of Section 2 more systematically by presenting time-series evidence of the contributing factors to labour hoarding in the Great Recession. We estimate two models that explain relative trend-deviations of hourly productivity and working time by the output gap. We use these estimates to construct forecasts for the period 2005q3 to 2012q4, which captures the upswing before the Great Recession, the recession itself, and the first three years of the subsequent economic expansion. We then compare the forecasted and actual development of the relative trend-deviations of hourly productivity and working time in order to test if and to what extent they reacted unusually pronounced to the output shock in the crisis. For our regressions, we use time series information on hourly productivity, working time, and output to construct relative deviations from their respective trend, which we call ‘gaps’: xtgap=xt−x¯tx¯t, (3) where xt is the respective variable in time period t, and x¯t is its trend value (see Online Appendix A1 for details). Figure 2 shows the output gap, the productivity gap and the working-time gap over the whole sample period (see Online Appendix A3 for summary statistics). The output gap is strongly correlated with both the productivity gap and the working-time gap with correlation coefficients of 0.78 and 0.74 respectively. Fig. 2. View largeDownload slide Productivity per hour gap (A), working time gap (B), and output gap, 1970q1–2012q4 Notes: ‘Gap’ is the % deviation of the original time series from the trend component. The latter is constructed using the HP filter with λ = 1,600. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. Fig. 2. View largeDownload slide Productivity per hour gap (A), working time gap (B), and output gap, 1970q1–2012q4 Notes: ‘Gap’ is the % deviation of the original time series from the trend component. The latter is constructed using the HP filter with λ = 1,600. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. Next, we estimate two Autoregressive Distributed Lag (ADL) models. In the first model, the hourly productivity gap, LPgap, is explained by current and lagged values of the output gap, Ygap, and lagged values of the dependent variable (eq. 4). In the second model, a corresponding specification explains the working-time gap, WTgap, with current and lagged values of the output gap and lags of the dependent variable (eq. 5).10 LPtgap=∑k=1nα1,kLPt−kgap+∑j=0nα2,jYt−jgap+utLP (4) and WTtgap=∑k=1nβ1,kWTt−kgap+∑j=0nβ2,jYt−jgap+utWT (5) The αs and βs are coefficients and uLP and uWT are error terms for the productivity and the working-time estimations respectively. To address potential endogeneity of the output gap we apply a two-stage least squares approach and instrument the contemporaneous German output gap with the world output gap. Due to its strong export orientation, German economic performance heavily depends on global economic activity, while changes in German working time or hourly productivity can be expected to have no relevant impact on the global business cycle. The world output gap and the German output gap are highly correlated with a correlation coefficient of 0.72.11 We follow a general-to-specific approach to determine the lag length, starting with eight lags. Lags with the highest p-values are dropped until only those significant at the 10% level or less remain in the regression (the results are presented in Online Appendix A4). According to the R-squared metric, both models explain a high share of the variation in the dependent variable. There is no evidence for serial correlation in the residuals at conventional levels of significance.12 We use the estimation results to generate forecasts of the productivity and working-time gap based on the actual output gap for the period 2005q3 to 2012q4. The start of the forecast period is the beginning of the upswing before the Great Recession. Figure 3 shows the actual and forecasted gaps. Fig. 3 View largeDownload slide Actual and forecasted productivity (A) and working time gap (B), 2005q3–2012q4 Notes: Actual and forecasted productivity and working time gap from 2005q3 to 2012q4. ‘Gap’ is the % deviation of the original time series from the trend component. The latter is constructed using the HP filter with λ = 1,600. Forecasts are based on estimates of two-stage least squares ADL models. +/- 2 S.E. provide confidence bands of the forecasts at the 95% level. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. Fig. 3 View largeDownload slide Actual and forecasted productivity (A) and working time gap (B), 2005q3–2012q4 Notes: Actual and forecasted productivity and working time gap from 2005q3 to 2012q4. ‘Gap’ is the % deviation of the original time series from the trend component. The latter is constructed using the HP filter with λ = 1,600. Forecasts are based on estimates of two-stage least squares ADL models. +/- 2 S.E. provide confidence bands of the forecasts at the 95% level. Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. The forecasted productivity per hour gap tracks the actual development very closely over the whole forecast period (Fig. 3a) and is always within the confidence band of two standard errors. Hence, the cyclical reaction of productivity in the Great Recession was not significantly different from the average reaction in the 1970 to 2005 period. Thus, cyclical reductions in hourly productivity contributed strongly to safeguarding jobs in the Great Recession, but its magnitude was in line with historical evidence and is therefore unlikely to account for the employment ‘miracle’. Actual working time, on the contrary, departed significantly from its forecasted development (Fig. 3b). First, in the upswing before the Great Recession and especially since 2007, actual cyclical working time was much higher than forecasted. Second, in the Great Recession it decreased much stronger than predicted.13 Cyclical working-time reductions in 2008 and 2009 contributed exceptionally strongly to safeguarding employment and are therefore the central factor behind the employment ‘miracle’. After the recession, in 2011 and 2012, actual and forecasted working-time gaps are almost identical. How many jobs were saved in the recession? We use forecasts of the working-time and productivity gap, WTtgap,fand LPtgap,fto calculate the counterfactual level of employment, Etcf: Etcf=E¯t(1+Ytgap−WTtgap,f−LPtgap,f) (6) The results of this calculation are shown in Fig. 4. If hourly productivity and working time had reacted as they did on average in 1970q1 to 2005q2, employment would have been considerably higher before and lower after the recession. From the peak in the first quarter of 2008 to the trough in the third quarter of 2009, employment would have decreased by 546,000 persons instead of actually increasing by 113,000 persons. This difference amounts to about 659,000 persons or 1.63% of total employment. We conclude that the econometric results in Section 3 confirm the initial assessment of Section 2: While both working-time and labour productivity were strongly pro-cyclical, and thus contributed to safeguarding jobs in the recession, only the reduction in cyclical working time was exceptionally large compared to past experiences and additionally safeguarded jobs to a significant extent. Fig. 4. View largeDownload slide Actual and forecasted employment in 1,000 persons Notes: Actual and forecasted employment constructed according to eq. (6). Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. Fig. 4. View largeDownload slide Actual and forecasted employment in 1,000 persons Notes: Actual and forecasted employment constructed according to eq. (6). Source: Federal Statistical Office (Statistisches Bundesamt), authors’ calculations. 4. Instruments of working-time flexibility In this section, we identify the specific institutional mechanisms behind the strong reduction in working time in 2008 and 2009 and compare them to previous recessions. To do this, we analyse detailed information on the development and composition of working hours by working time instruments.14 In Germany, the following instruments of working time adjustment are available for work sharing: short-time work, overtime work, temporary reductions in collectively agreed/regular working hours, and working time accounts. We discuss each instrument in turn. The instrument that has gained most attention in the literature on the Great Recession is short-time work (Boeri and Bruecker, 2011; Will, 2011; Brenke et al., 2013; Balleer et al., 2016). Short-time work exists since the 1920s and is a well-established element in the toolkit of German active labour market policy. Firms flexibly used this instrument in the past, and its legal basis has been regularly changed in and between economic downturn periods.15 Currently it is a publicly subsidized form of work sharing where employees receive 60% (67% if they have children) of their net-wage from the government for the difference between their regular and actual working hours. The program is available to firms with cyclical economic problems. The remaining working-time instruments were not established by law, but negotiated between employers and employees and their representatives in a framework of corporatist industrial relations and implemented within collective and company agreements (Groß et al., 2000; Bispinck and WSI-Tarifarchiv, 2009).16 Paid overtime hours, i.e. the possibility to work more hours than contractually agreed, is the most common instrument of working-time flexibility.17 In contrast to short-time work and paid overtime hours, temporary reductions in collectively agreed/regular working hours and working-time accounts are relatively new instruments that have received less attention in the economic literature. The origin of both can be traced back to a collective agreement as a compromise of an industrial conflict about shorter weekly working hours in the metal and electrical industry in the mid-1980s which marked the beginning of working-time flexibilization in Germany (Müller-Jentsch, 2011, p. 126; see also Berg, 2008). Opening clauses that allow for temporary deviation from collectively agreed regular working hours at the firm level were traditionally especially important in the core industrial sectors. The first time this instrument was used was in 1993 with the introduction of the four-day week at the Volkswagen Company. In the 1990s, opening clauses were still largely limited to the manufacturing sector, but their importance increased over time. In 2005, already about 60% of firms and 75% of all employees in the manufacturing sector were covered by wage- and working-time-related opening clauses (Garloff and Gürtzgen, 2011). Meanwhile, many sectoral agreements beyond the manufacturing sector allow to reduce the agreed working time within given limits, or allow it to be in- or decreased in line with the economic situation within the framework of so-called working-time corridor arrangements (Heinbach and Schröper, 2007; Bispinck and WSI-Tarifarchiv, 2009, 2016). Finally, working-time accounts are a ‘modern company instrument of time management’ (Zapf, 2016, p. 3) to organize and regulate working hours flexibly over a certain period of time in an establishment (Bauer et al., 2004; Gerner, 2010). Individual deviations from regular or collectively agreed working hours lead to surpluses or deficits on these accounts, which typically have to be rebalanced within a certain predefined time period. Working-time accounts are implemented within the framework of collective and company agreements (Groß et al., 2000). In 2009 around 50% of all workers used them, compared to just 35% ten years earlier. In the industrial sectors and in large firms this share is even higher (Zapf and Brehmer, 2010). Additionally, over time the average upper and lower limits for the number of hours saved on working-time accounts grew and their average compensation period became longer (Groß, 2010). Hence, the possibility to use working-time accounts to adjust working hours in response to changing circumstances strongly increased in the decade before the Great Recession.18 To show how strongly these different working-time instruments were used in the four recessions since 1970, Fig. 5 presents the changes in cyclical hours worked per person in employment for these four instruments of work sharing.19 Fig. 5. View largeDownload slide Components of cyclical working time reductions in recessions from peak to trough, working hours per employee per quarter Notes: Downturn I: 1973q2 to 1975q2, Downturn II: 1979q4 to 1982q4, Downturn III: 2001q1 to 2005q2, Downturn IV: 2008q1 to 2009q3. ‘Cyclical’ refers to the difference of actual and trend changes, where the trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Trend and cycle have been calculated for overtime work and regular working time. Source: Institute for Employment Research (IAB) working time calculations; authors’ calculations. Fig. 5. View largeDownload slide Components of cyclical working time reductions in recessions from peak to trough, working hours per employee per quarter Notes: Downturn I: 1973q2 to 1975q2, Downturn II: 1979q4 to 1982q4, Downturn III: 2001q1 to 2005q2, Downturn IV: 2008q1 to 2009q3. ‘Cyclical’ refers to the difference of actual and trend changes, where the trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Trend and cycle have been calculated for overtime work and regular working time. Source: Institute for Employment Research (IAB) working time calculations; authors’ calculations. Different instruments were employed in the various recessions to temporarily decrease working time. Overtime reductions and short-time work were amply used in all recessions, with the exception of Downturn III, where aggregate working-time reductions hardly occurred. Working-time accounts had an important impact only in the Great Recession. Our results in Fig. 5 show that regular working time was strongly reduced in Downturn I, even though this instrument has only recently become available at a larger scale. This finding might be mainly the result of a coincidence, however, because a significant reduction in general working time was implemented in 1974 which was unrelated to the recession (Herzog-Stein and Seifert, 2010) and is partially picked up in our data as a cyclical reduction. Thus, the Great Recession was the first time when collectively agreed/regular working time was deliberately reduced to safeguard employment. Why was aggregate working time hardly reduced in Downturn III? Bosch (2011) argues that in this downturn firms were not interested in safeguarding employment because they expected that after the end of the reunification boom economic growth and labour demand would not return to previous levels. Even the government subsidized short-time work was hardly used. Bogedan (2010) notes that the government only allows companies to use short-time work after they have exhausted their other available working-time reduction instruments, which strengthens the point of Bosch (2011). In the Great Recession, however, all instruments were strongly used to reduce working time. As to answer why firms were interested in hoarding labour, Möller (2010) shows that the crisis mainly affected export-oriented manufacturing firms. Their employees often possess extensive firm-specific know-how, which makes it difficult and expensive to replace them. Export-oriented firms further strongly profited from the global upswing before the crisis, and suffered from a shortage of qualified workers, which might has made them reluctant to fire workers. Consistent with this interpretation, Bosch (2011, 2015) emphasizes that firms continued hiring apprentices and employing them after the completion of training because they expected skill shortages in the near future.20 Finally, the government also increased incentives to take up short-time work (e.g. Boeri and Bruecker, 2011). For example, it extended the maximum entitlement period from six to 24 months, and starting with January 2009, employers were granted to pay only half of the standard social security contribution and even none if the employee participated in certain vocational training programs. These economic incentives were combined with governmental media-campaigns motivating employers to use short-time work instead of laying off workers. Figure 6 shows the development of these cyclical working-time components for the Great Recession and its preceding upswing. It presents the same data as in Fig. 5, i.e. the four trend-adjusted working-time instruments. During the upswing, from late 2005 to early 2008, regular working hours were expanded, working-time accounts were filled, and overtime work was increased relative to the trend development of these working-time components. Given that flexible adjustments from collectively agreed or regular weekly working hours and working-time accounts are rather new instruments, this might also explain the unexpectedly strong increase in working time during the upswing, going along with a relatively weak increase in employment measured in persons. Fig. 6. View largeDownload slide Components of cyclical changes in working time, working hours per employee per quarter, 2005q1–2012q4 Notes: ‘Cyclical’ refers to the difference of actual and trend changes for each working time instrument (if the series shows a trend), where the trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Source: Institute for Employment Research (IAB) working time calculations; authors’ calculations. Fig. 6. View largeDownload slide Components of cyclical changes in working time, working hours per employee per quarter, 2005q1–2012q4 Notes: ‘Cyclical’ refers to the difference of actual and trend changes for each working time instrument (if the series shows a trend), where the trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Source: Institute for Employment Research (IAB) working time calculations; authors’ calculations. In the recession, starting with the fourth quarter of 2008, working-time account balances and overtime work were reduced strongly, while regular working hours were still increasing. However, in 2009, all instruments of working-time adjustment strongly contributed to the cyclical decline in working time. The contributions of short-time work and reductions of regular working hours increased over time, while working-time accounts and overtime work played a significant role especially at the beginning of the crisis, in line with Bosch (2011, 2015) who stresses the sequential use of different working-time instruments by companies. While short-time work played an important role, working-time accounts and regular working-time reductions contributed at least equally strongly. 5. How do our findings relate to previous research? Several other studies assess the robustness of the German labour market in the Great Recession. In our discussion, we focus on core differences in the results and methodological approaches to the most pertinent ones. Burda and Hunt (2011) have published the probably most influential article on the German labour market miracle that warrants a detailed discussion and comparison with our results. They argue based on time-series evidence of the aggregate economy that employment has hardly fallen in the Great Recession because employment growth was very low in the upswing before due to employers’ lack of confidence in its durability. Burda and Hunt (2011) further suggest that working-time reductions did not contribute strongly to saving jobs. According to them ‘hours per worker fell rapidly in the Great Recession’, but ‘their path is roughly comparable to that in the shallower 1973–75 recession’, while ‘the 4 percent reduction in productivity in the 2008–09 recession contrasts with strong increases in productivity in the four previous recessions’ (p. 280). They confirm this assessment when applying a time-series forecasting approach explaining working time, where for the 2008–9 recession their ‘models predict a fall in hours per worker similar to the actual fall’ (p. 287). Thus, their ‘analysis suggests that the decline in hours per worker in the 2008–09 recession was not surprising given the depth of the recession’ (p. 297). Our results indicate that Burda and Hunt’s (2011) conclusions only hold because they neglect the role of trend growth of working time and productivity. This is made explicit in our Table 2 (Section 2), where the actual development and cyclical effects of both variables are presented. While the actual developments are consistent with Burda and Hunt’s findings, the relevant cyclical changes support a different view. We illustrate this point in more detail by plotting actual and trend working time for the period 1970 to 2012 in Fig. 7. Trend working time is strongly decreasing over the whole sample period. This trend-decline in working time is unrelated to temporary labour hoarding in response to an output shock. It rather reflects structural factors, e.g. an increase in part-time employment often in the services sector and a decrease in full-time employment in the manufacturing sector, general working-time reductions, or shifts in preferences and the composition of the work force. Fig. 7. View largeDownload slide Actual and trend working time per quarter, 1970q1–2012q4 Notes: 1970q1-1990q4 West Germany, 1991q1-2012q4 Germany. The working time trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Source: Federal Statistical Office (Statistisches Bundesamt), Institute for Employment Research (IAB). Fig. 7. View largeDownload slide Actual and trend working time per quarter, 1970q1–2012q4 Notes: 1970q1-1990q4 West Germany, 1991q1-2012q4 Germany. The working time trend is constructed applying the Hodrick-Prescott filter with λ = 1,600. Source: Federal Statistical Office (Statistisches Bundesamt), Institute for Employment Research (IAB). Since Burda and Hunt (2011) inadequately account for trend developments in their analysis, they attribute the trend-decline in working time to labour hoarding by firms in recessions. Because trend working time decreased much more strongly in the 1970s and 1980s than in the 2000s, Burda and Hunt (2011) overstate the effect of labour hoarding due to working-time reductions for the earlier decades, and understate it for the more recent years. Contrary to what they suggest, but in line with our empirical findings presented in Sections 2 and 3, Fig. 7 underlines that the magnitude of cyclical reductions in working time in the Great Recession of 2008 to 2009 is unprecedented. A different interpretation of the evidence is provided by Boysen-Hogrefe and Groll (2010), who suggest that the wage moderation in the upswing before the crisis is the key mechanism behind the stability in employment. They argue that ‘the wage moderation … was one of the most, if not the most, important necessary condition for firms to be able to hoard labour, which led to the unprecedented decline in productivity [in the recession]’ (p. 45). Their conclusions, however, are not based on direct empirical evidence, whereas we test for the relevance of these mechanisms. Our findings refute this argument since the reduction in productivity is properly forecasted in our econometric model (see Section 3).21 Möller (2010) provides an important early study, discussing various factors behind the employment miracle. Some of them are already mentioned in the previous section, including the role of working-time accounts and the favourable behaviour of social partners, as well as short-time work, which allowed firms to buffer the demand shock by reducing working time. He also highlights labour hoarding in the form of reductions in productivity. However, Möller (2010) does not estimate counterfactual developments for working time or productivity, and is therefore not quantifying the relative importance of these mechanisms. Merkl and Wesselbaum (2011) examine cyclical changes in average working time more generally. They compare the importance of cyclical working-time adjustments between the USA and Germany for the time period 1970 to mid-2009 and find no difference between these two countries. Their results suggest that in the Great Recession German labour market institutions did not play a greater role in adjusting working time to safeguard employment than those of the more deregulated US labour market. However, applying the same methodology to longer samples, Wesselbaum (2011),22 and Herzog-Stein and Nüß (2015) show—consistent with our findings—that the relative importance of cyclical working-time adjustments increased markedly in Germany in the Great Recession but did not in the USA. Finally, our findings complement the results of several studies demonstrating the importance of short-time work in the crisis to stabilize employment (e.g. Boeri and Bruecker, 2011; Brenke et al., 2013; Balleer et al., 2016). 6. Conclusion We show that in all major German recessions the reduction in cyclical productivity per hour played a significant role in buffering the effects of output shocks on employment, while cyclical working-time reductions were more infrequent and quantitatively much less important. However, in the Great Recession temporary working-time reductions played a substantial role. To quantify the extent to which reductions in productivity and working time were unexpected by historical standards, we estimate time-series models explaining these variables from 1970 to 2005, and create forecasts until 2012. This allows us to compare the actual development of cyclical productivity per hour, as well as cyclical working time, with their respective counterfactual development. The results show that the German employment miracle in the Great Recession is mainly explained by strong temporary working-time reductions. These were enabled by new work-sharing instruments, i.e. working-time accounts and discretionary reductions of regular working hours. Both instruments were not established by law but within the context of corporatist negotiations between employers and employees. Further, the government funded short-time work scheme played a central role. Labour hoarding in the form of lower productivity is relatively costly for firms. Reducing working time is generally a cheaper tool to achieve the same outcome and might therefore be expected to be the preferred mechanism from the perspective of the employer. However, while labour hoarding in the form of lower productivity is relatively common across countries and time (e.g. Biddle, 2014), working-time reductions in downturns do not seem to play a significant role in most countries (van Rens, 2012). One likely reason for this is the lack of institutions allowing for flexible working-time arrangements. The German social partners have established such novel and effective institutions. Along with the already established short-time work scheme, these institutional features within the German labour market allow for high internal flexibility, even though external labour market flexibility is low. Our findings might also be helpful in understanding the European pattern of national labour market responses to the crisis. For instance, Amable and Mayhew (2011) find that, after correcting for the severity of the drop in output and the strength of macro policy responses, unemployment in the Great Recession rose by less in countries with strict employment protection legislation (EPL) and high collective-bargaining coverage. Boeri and Bruecker (2011) present cross-country evidence that strict EPL correlates with the use of short-time work in the Great Recession and that contractual arrangements which facilitate adjustments at the extensive margin reduce the use of short-time work at the company level, while a larger share of workers with vocational training increase it. Strict EPL, centralized collective bargaining, and workers with high firm-specific knowledge might thus be institutional prerequisites for adjusting labour demand along the internal margin to a significant extent (see also Estevez-Abe et al., 2001; Sturn, 2013; Llosa et al., 2014). Our analysis focuses on short-term stabilization of internal flexibility through working-time instruments. This short-term stabilization, however, might also have a positive long-run impact on the labour market. Recent empirical evidence by Klinger and Weber (2016) shows that the decades-long upward trend in German unemployment is largely explained by hysteresis, i.e. increases in cyclical unemployment becoming subsequently structural, which highlights the importance of counter-cyclical policies to prevent an increase in unemployment in a recession in the first place. Temporary working-time reductions are a novel and potentially powerful form of countercyclical stabilization ‘policies’. Because of these stabilizing cyclical effects of internal flexibility and its potential positive long-run outcomes, we conclude that there are good reasons for academic researchers to focus more on the quantification, causes, and consequences of high internal labour market flexibility beyond the German experience in the Great Recession. While the impact of external flexibility on unemployment has been heavily researched since the 1990s, the topic of internal flexibility has received little attention to date. Supplementary material Supplementary material—the Appendix and the Data files—are available online at the OUP website. Funding This work was supported by the Oesterreichische Nationalbank Anniversary Fund [Grant No. 16773 to S.S.]. Footnotes 1 Own calculations with real GDP data from the OECD and business cycle dates from the procedure developed by the German Council of Economic Experts for Germany, NBER’s Business Cycle Dating Committee, and the CEPR Euro Area Business Cycle Dating Committee. 2 For continuous growth rates, the relation presented in the following equation holds with equality. However, for discrete growth rates, it only holds approximately. The approximate case is chosen because we use quarterly growth rates. 3 We focus on the aggregate economy since in a sectoral analysis classification problems associated with the boom in temporary agency work since 2003 potentially lead to biased findings. Specifically, temporary agency workers are officially working in the services sector, but are predominantly contracted to manufacturing firms, which might result in an exaggeration of productivity increases in the manufacturing sector in a boom, as well as its reduction in a recession, while the opposite is true for working time in the non-manufacturing sector. 4 For instance, not accounting for trend developments would result in wrongly attributing structural change in the form of an increase in part-time employment in low-productivity services jobs and a reduction in full-time employment in high-productivity manufacturing jobs, resulting in lower productivity growth and a reduction in overall working time, to cyclical labour hoarding in a recession. 5 Information on working hours (as part of the German national accounts) and its components is provided by the Institut für Arbeitsmarkt- und Berufsforschung (IAB) based on various sources (for a detailed account see Wanger et al., 2016). The data used here do not include unpaid overtime hours. Evidence suggests that employees work a significant number of unpaid overtime hours per year (e.g. Brautzsch and Will, 2010; Zapf, 2012). To the extent that unpaid overtime work is pro-cyclical, i.e. increases in booms and decreases in recessions, like paid overtime work (see Section 4), this overstates the pro-cyclicality of hourly labour productivity, and understates the pro-cyclicality of working time. 6 The results remain reasonably robust when higher values for the smoothing parameter are used. 7 For pre-reunification Germany, the sample lasts from 1970q1 to 1991q4, while for post-reunification Germany it starts with 1990q1. This reduces the end-value problem of univariate filter methods for the reunification period. 8 The method is described in Online Appendix A2. 9 For details see Herzog-Stein and Seifert (2010) and the references mentioned there. 10 Reunification dummies for the year 1991 are found to be insignificant and are therefore not included. 11 The world output gap is based on quarterly world GDP provided by the IMF, seasonally adjusted with the BV4.1 procedure of the German Federal Statistical Office. The results are robust to other estimation approaches, such as ordinary least squares (OLS) and generalised method of moments (GMM) without instrumenting the output gap. 12 For both the productivity gap and the working-time gap model Wald tests reject that the coefficients of the output gap and its lags sum to zero at the 5% significance level. To assess the predictive power of these specifications, we estimated them until 1982q4, the beginning of a trough, and forecasted them until 1990q4. Further, we estimate them until 1999q1, the beginning of an upswing, and forecasted them until 2005q1. In both cases our models perform reasonably well in explaining actual developments. 13 This conclusion is also supported by the results of structural break tests. We estimate the working-time specification over the full sample, and perform Andrew-Quandt structural break tests, and find a break in 2006q1. 14 Data source is the IAB working-time computation (see footnote 5). The instruments are sub-categories of aggregate working time. For details see Bach and Koch (2002) and Wanger et al. (2016) as well as Online Appendix A1. 15 For detailed information on institutional reforms in the short-time working scheme over the last decades see Bogedan (2010) and Will (2011). 16 German industrial relations legislation distinguishes between collective agreements (Tarifverträge) by employers and unions usually signed at the industry level, and company or works agreements (Betriebsvereinbarungen), negotiated between a single employer and the works council that are based on opening clauses in collective agreements. 17 In practice remuneration of overtime hours can vary a lot from unpaid overtime to significant overtime premiums. It is even possible that some overtime hours are compensated by leisure time. The data used here do not include unpaid overtime hours. 18 For more details on working-time accounts and the determinants of its use to safeguard employment at the establishment level in the Great Recession see Herzog-Stein and Zapf (2014). 19 We seasonally adjust the working-time instruments data and subtract individually computed trends from overtime and regular working time to construct cyclical changes in working time. The change in regular working time is the sum of cyclical changes in full- and part-time jobs. Both are individually de-trended because there is a trend decline in average full-time working hours and a trend increase in part-time work. Since each instrument is de-trended individually, the numbers do not exactly add up to the cyclical component of aggregate working time discussed in Section 2. Short-time work and working-time accounts do not show a trend over time and are thus not de-trended. 20 This is an important reminder not to interpret the labour market miracle in a purely static sense that existing worker-job matches were simply preserved with the help of working-time flexibility. The successful crisis management ensured that on aggregate a certain employment level in the economy was preserved while the dynamic job and worker flows in the labour market continued, e.g. with marked changes in employment in different sectors of the economy and continued recruitment in hard hit sectors such as manufacturing. 21 Also Schaz and Spitznagel (2010) show that Germany’s hourly productivity was strongly pro-cyclical well before the Great Recession. 22 This result is only presented in the working paper version but not in the journal article published in 2016 under the same title. Acknowledgements We thank Michael Ash, Deepankar Basu, Peter Hohlfeld, Camille Logeay, Robert Pollin, Christian Proaño, Katja Rietzler, Sven Schreiber, Sabine Stephan, Thomas Theobald, Till van Treeck, and two anonymous referees for helpful comments and discussions. Any remaining errors are ours. References Amable B., Mayhew K. ( 2011) Unemployment in the OECD, Oxford Review of Economic Policy , 27, 207– 20. Google Scholar CrossRef Search ADS Bach H.-U., Koch S. ( 2002) Working time and the volume of work in Germany: the IAB concept of measurement, IAB Labour Market Research Topics , 53, 1– 27. Ball L.M., Leigh D., Loungani P. ( 2013) Okun’s Law: fit at fifty?, NBER Working Papers, 18668 , Cambridge, MA. Balleer A., Gehrke B., Lechthaler W., Merkl C. ( 2016) Does short-time work save jobs? A business cycle analysis, European Economic Review , 84, 99– 122. Google Scholar CrossRef Search ADS Bauer F., Groß H., Lehmann C., Munz E. ( 2004) Arbeitszeit 2003. Arbeitszeitgestaltung, Arbeitsorganisation und Tätigkeitsprofile . Technical Report, Institut zur Erforschung sozialer Chancen, Köln. Berg P. ( 2008) Working time flexibility in the German employment relations system: implications for Germany and lessons for the United States, Industrielle Beziehungen , 15, 133– 50. Biddle J.E. ( 2014) Retrospectives: the cyclical behavior of labor productivity and the emergence of the labor hoarding concept, Journal of Economic Perspectives , 28, 197– 212. Google Scholar CrossRef Search ADS Bispinck R. WSI-Tarifarchiv. ( 2009) Tarifliche Regelungen zur befristeten Arbeitszeitverkürzung, Elemente qualitativer Tarifpolitik , 67, Düsseldorf. Bispinck R. WSI-Tarifarchiv. ( 2016) Arbeitszeit – Was bietet der tarifvertragliche Instrumentenkoffer? - Eine Analyse von 23 Branchen und Tarifbereichen, Elemente qualitativer Tarifpolitik , 82, Düsseldorf. Boeri T., Bruecker H. ( 2011) Short-time work benefits revisited: some lessons from the Great Recession, Economic Policy , 26, 697– 765. Google Scholar CrossRef Search ADS Bogedan C. ( 2010) Arbeitsmarktpolitik aus der ‘Mottenkiste’? Kurzarbeitergeld im Lichte politischer Interessen, WSI Mitteilungen , 11, 577– 83. Bosch G. ( 2011) The German labour market after the financial crisis: miracle or just a good policy mix?, in D. Vaughan-Whitehead (ed.) Work Inequalities in the Crisis? Evidence from Europe , Edward Elgar, Cheltenham, 243–77. Bosch G. ( 2015) The German welfare state: From an inclusive to an exclusive Bismarckian model . in D. Vaughan-Whitehead (ed.) The European Social Model in Crisis - Is Europe Losing its Soul? Edward Elgar, Cheltenham, 175– 229. Google Scholar CrossRef Search ADS Boysen-Hogrefe J., Groll D. ( 2010) The German labour market miracle. National Institute Economic Review , 214, 38– 50. Google Scholar CrossRef Search ADS Brautzsch H.-U., Will H. ( 2010) Kurzarbeit: Wichtige Stütze des Arbeitsmarktes in der Krise, aber keine Dauerlösung. Wirtschaft Im Wandel , 16, 376– 83. Brenke K., Rinne U., Zimmermann K.F. ( 2013) Short-time work: the German answer to the Great Recession, International Labour Review , 152, 287– 305. Google Scholar CrossRef Search ADS Burda M.C., Hunt J. ( 2011) What explains the German labor market miracle in the Great Recession?, Brookings Papers on Economic Activity , 42, 273– 335. Google Scholar CrossRef Search ADS Estevez-Abe M., Iversen T., Soskice D. ( 2001) Social protection and the formation of skills: a reinterpretation of the welfare state , in P.A. Hall and D. Soskice (eds) Varieties of Capitalism – The Institutional Foundations of Comparative Advantage , Oxford University Press, Oxford. Google Scholar CrossRef Search ADS Garloff A., Gürtzgen N. ( 2011) Flexibilität durch Öffnungsklauseln – Wenn weniger mehr ist, IAB-Forum 1/2011 , 26– 29. Gerner H.-D. ( 2010) Arbeitszeitverlängerung, Arbeitszeitkonten und Teilzeitbeschäftigung: Ökonometrische Analysen . IAB Bibliothek , 322, Nürnberg. Gordon R.J. ( 1993) The jobless recovery: does it signal a new era of productivity-led growth?, Brookings Papers on Economic Activity , 24, 271– 316. Google Scholar CrossRef Search ADS Groß H. ( 2010) Vergleichende Analyse der Arbeits- und Betriebszeitentwicklung im Zeitraum von 1987 bis 2007. Beiträge aus der Forschung , 176. Sozialforschungsstelle Dortmund (sfs), Dortmund. Groß H., Munz E., Seifert H. ( 2000) Verbreitung und Struktur von Arbeitszeitkonten, Arbeit. Zeitschrift Für Arbeitszeitforschung, Arbeitsgestaltung und Arbeitspolitik , 9, 217– 99. Heinbach W.D., Schröper S. ( 2007) Typisierung der Tarifvertragslandschaft – Eine Clusteranalyse der tarifvertraglichen Öffnungsklauseln. Jahrbücher für Nationalökonomie und Statistik , 227, 219– 35. Google Scholar CrossRef Search ADS Herzog-Stein A., Seifert H. ( 2010) Der Arbeitsmarkt in der Großen Rezession – Bewährte Strategien in neuen Formen, WSI Mitteilungen , 11, 551– 59. Herzog-Stein A. and Zapf I. ( 2014) Navigating the Great Recession, ILR Review , 67, 891– 925. Google Scholar CrossRef Search ADS Herzog-Stein, A. and Nüß, P. (2015) Extensive versus intensive margin over the business cycle: new evidence for Germany and the United States, IMK Working Paper, 163, Düsseldorf. Klinger S., Weber E. ( 2016) Detecting unemployment hysteresis: a simultaneous unobserved components model with Markov switching, Economics Letters , 144, 115– 18. Google Scholar CrossRef Search ADS Krugman P. ( 2009) Free to lose, New York Times , November 1, A31. Llosa G., Ohanian L., Raffo A., Rogerson R. ( 2014) Firing costs and labor market fluctuations: a cross-country analysis, Society for Economic Dynamics 2014 Meeting Papers , 533. Merkl C., Wesselbaum D. ( 2011) Extensive versus intensive margin in Germany and the United States: any differences?, Applied Economics Letters , 18, 805– 8. Google Scholar CrossRef Search ADS Möller J. ( 2010) The German labor market response in the world recession: de-mystifying a miracle, Journal for Labour Market Research , 42, 325– 36. Google Scholar CrossRef Search ADS Müller-Jentsch W. ( 2011) Gewerkschaften und Soziale Marktwirtschaft seit 1945 , Philipp Reclam jun, Stuttgart. Ohanian L.E., Raffo A. ( 2012) Aggregate hours worked in OECD countries: new measurement and implications for business cycles, Journal of Monetary Economics , 59, 40– 56. Google Scholar CrossRef Search ADS Okun, A.M. (1962) Potential GNP: its measurement and significance. Proceedings of the Business and Economic Statistics Section of the American Statistical Association, 98–103. Schaz P., Spitznagel E. ( 2010) Makroökonomische Dynamik von Arbeitsmärkten – Ein Vergleich interner und externer Flexibilität in den USA und Deutschland, WSI Mitteilungen , 12, 626– 35. Sturn S. ( 2013) Are corporatist labour markets different? Labour market regimes and unemployment in OECD countries, International Labour Review , 152, 237– 54. Google Scholar CrossRef Search ADS Van Rens T. ( 2012) How important is the intensive margin of labor adjustment? Discussion of ‘Aggregate hours worked in OECD countries: New measurement and implications for business cycles’ by Lee Ohanian and Andrea Raffo, Journal of Monetary Economics , 59, 57– 63. Google Scholar CrossRef Search ADS Wanger S., Weigand R., Zapf I. ( 2016) Measuring hours worked in Germany – contents, data and methodological essentials of the IAB working time measurement, Journal for Labour Market Research , 49, 213– 38. Google Scholar CrossRef Search ADS Wesselbaum D. ( 2011) The intensive margin puzzle and labor market adjustment costs , Kiel Working Papers, 1701, Kiel. Will H. ( 2011) Germany’s short time compensation program: macroeconom(etr)ic insight , IMK Working Paper, 1, Düsseldorf. Zapf I. ( 2012) Flexibilität am Arbeitsmarkt durch Überstunden und Arbeitszeitkonten – Messkonzepte, Datenquelle und Ergebnisse im Kontext der IAB-Arbeitszeitrechnung , IAB-Forschungsbericht, 3, Nürnberg. Zapf I. ( 2016) Bestimmungsgründe von Arbeitszeitkonten bei Betrieben und Beschäftigten, Sozialpolitik.ch , 2/2016, 1– 35. Available at: https://www.sozialpolitik.ch/fileadmin/user_upload/2_2016_Zapf_final.pdf (last accessed 12 July 2017). Zapf I., Brehmer W. (2010) Flexibilität in der Wirtschaftskrise – Arbeitszeitkonten haben sich bewährt, IAB-Kurzbericht , 22, Nürnberg. © Oxford University Press 2017 All rights reserved
Oxford Economic Papers – Oxford University Press
Published: Jan 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera