The Speed of Leaving the Old Job

Wolff, Joachim; Trübswetter, Parvati Article The Speed of Leaving the Old Job Schmollers Jahrbuch – Zeitschrift für Wirtschaftsund Sozialwissenschaften. Journal of Applied Social Science Studies Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Wolff, Joachim; Trübswetter, Parvati (2003) : The Speed of Leaving the Old Job, Schmollers Jahrbuch – Zeitschrift für Wirtschaftsund Sozialwissenschaften. Journal of Applied Social Science Studies, ISSN 1865-5742, Duncker & Humblot, Berlin, Vol. 123, Iss. 1, pp. 55-69, https://doi.org/10.3790/schm.123.1.55 This Version is available at: https://hdl.handle.net/10419/292038 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ The Speed of Leaving the Old Job A Study on Job Changes and Exits into Unemployment During the East German Transition Process* By Joachim Wolff and Parvati Tru ¨bswetter Abstract This paper analyses how quickly old job matches ceased to exist during the East German transition process from July 1990 to December 1993. Using job spells from the German Socio-Economic Panel-East, we estimate the effects of important covariates on the transition rates from the old job to a new one and into unemployment by a competing-risks duration model. Our results suggest that the speed of exit into new jobs rises with skills. The special short-time allowance slowed down the exit rate into unemployment. Workers increased their rate of job change temporarily just after the end of this program. JEL Classification: C41, J63, P2 1. Introduction Since production in socialist economies was relatively labour intensive, the destruction of a large number of jobs in the old firms was inevitable in the transition process. This paper studies how quickly old job matches ceased to exist in East Germany and to what extent they terminated by a job change or by exit into unemployment. The reallocation of labour was expected to lead to considerable flows between employment and unemployment and vice versa. But in many transition economies these flows remained low by Western standards (Boeri, 1994). This also applies to East Germany (Wolff, 1998). To a large extent, the reallocation of labour from old less productive jobs to new ones took place by direct job changes. According to data from the Labour Market Monitor, East German Schmollers Jahrbuch 123 (2003) 1 Schmollers Jahrbuch 123 (2003), 55–70 Duncker & Humblot, Berlin * We are extremely grateful to Stephan Klasen, Rene ´Bo ¨heim, Jennifer Hunt, the participants of the XVI annual conference of the European Society for Population Economics and the participants of the GSOEP user conference 2002 for very helpful comments on earlier versions of this paper. Support from the Deutsche Forschungsgemeinschaft under the project Sfb 386 is gratefully acknowledged. OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 56 Joachim Wolff and Parvati Tru ¨bswetter job changes made up for roughly 55% of total hiring from November to November of the years 1991 to 1993 (German Federal Labour Office, 1994). This figure is considerably higher than in market economies. 1 East Germany chose a strategy of a rapid transition. With the introduction of the German Economic, Monetary, and Social Union in July 1990 and unification in October 1990, the West German institutional framework was adopted. A consequence was that wages would adjust rapidly to the much higher West German levels. Already in 1991 some collective agreements aimed at reaching this goal quickly. 2 Due to the wage rise and high subsidies for new capital investment in East Germany, a large proportion of the new jobs were good jobs with a high productivity and wage level. Hence, new firms chose more capital and human-capital intensive production technologies than prior to the transition process. Thus, they needed highly talented applicants to fill their vacancies, while old firms needed to retain talented workers to turn their jobs into profitable ones. The speed of exit of workers from old jobs into new jobs should hence increase in skills, past occupational position and past wages, which are proxies for their talents. The reverse should be true for exits into unemployment. Protection against dismissal may be another important determinant of the speed of exit from the old job. Protection may be more complete the older the worker and the higher his or her tenure. It may also be more complete for workers with children. Therefore, the transition rates both into new jobs and into unemployment should decline with age, children and tenure. As many old jobs were likely to become unprofitable, an important temporary program of special allowances for short-time work was introduced to reduce job losses. Prior to its end in January 1992, the old firms had a means of retaining their most talented workers, even if their jobs were not yet profitable. This should have kept the exit rates from the old jobs into unemployment initially relatively low compared with the period after December 1991. However, it could also explain variation in the speed of exit into new jobs over time. Before the end of the short-time allowance program, new employers had difficulties finding talented unemployed applicants. Therefore, they may have hired workers directly from their old firm, which would have caused transition rates into new jobs to be relatively high. When the short-time allowance program ended, the old firms had to shed quickly a large proportion of their staff, Schmollers Jahrbuch 123 (2003) 1 1Pissarides (1994) summarized evidence on job changes in several countries: Blanchard and Diamond (1989) found that between 1968 to 1986 about 20% of new hires in the U.S. result from job changes. According to estimates of Burda and Wyplosz (1994) for West-Germany in the year 1987 at 17% this proportion is even lower. 2A contract in the metal industry agreed to reach West German gross wages by 1994. The median real monthly consumer wage rose by more than 80% in the first six years of transition (Hunt, 2001). OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 The Speed of Leaving the Old Job 57 so that workers’ incentives to search on-the-job and to accept a job offer were relatively high. Consequently, we expect the job-to-job transition rate to peak just after December 1991. In the course of 1992 and 1993 it should be relatively low, as it became easier to fill vacancies with talented unemployed applicants. Previous empirical studies did not address the speed of exit from old to new jobs. Licht and Steiner (1994) used the German Socio-economic Panel East (GSOEP-East) to study employment duration for the first nine months since July 1990. They analysed the determinants of exit from employment into short-time work, unemployment and non-participation by a logistic duration model. One important result was that the likelihood of remaining employed rises with a worker’s wage in May 1990. Another empirical study on employment duration in East Germany of Hunt (2002) regards a much longer time horizon and distinguished between voluntary and involuntary separations from employment as competing risks. Her analysis also relies on the GSOEP-East and confirms that the wage in May 1990 had a positive effect on the probability of remaining employed of people who were employed in mid 1990. If wages under socialism increased with talents, this may reflect attempts by old firms to keep their most talented workers and attempts by firms creating jobs to hire them. Some of our hypotheses contradict implications of a model developed by Boeri (1997) to explain high job-to-job movements in transition economies. His starting point is that under socialism the most educated workers had little incentive to develop their skills on the job, due to wage compression. For this reason, and due to the fact that dismissing low-productivity workers was almost impossible, managers had little incentive to monitor their workers’ productivity. Under some further assumptions his model explains why a high proportion of new hires resulted from job changes. He assumes that past occupation, past wage, or vocational attainment during socialism do not signal the productivity of a worker to the employers who create new jobs. His model predicts high job-to-job movements, but according to this assumption and in contrast to our views, one would not expect that the speed at which people leave their old job for a new one varies considerably with differences in these attributes. Our analysis is based on retrospectively collected individual job duration data from the GSOEP-East. We study the determinants of the duration of old jobs over the period July 1990 to December 1993 and apply standard econometric duration models for investigating the above hypotheses. Schmollers Jahrbuch 123 (2003) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 58 Joachim Wolff and Parvati Tru ¨bswetter 2. The East German labour market in the first years of transition In East Germany the impact of the transition to a market economy on the labour market was remarkable. Employment fell from more than 9.7 million people in the first quarter of 1989 to 6.2 million people at its (first) trough five years later. More than 90% of this net job loss had occurred by the first quarter of 1992. By that time the unemployment rate had reached more than 15%. Table 1 displays some differences between the East and West German labour market with respect to the unemployment rate, the inflow rate into unemployment, as well as the outflow rate from unemployment from 1991 to 1993. These statistics are computed using data from the Federal Labour Office and the Microcensus. At monetary union in July 1990, the East German unemployment rate was less than four percent. By 1991 it was already above ten percent. In the following two years it went up to roughly 16% and was about twice as high as in West Germany. Even though there was a large employment reduction, by Western standards East German flows into and out of unemployment do not seem extremely high. Between 1991 and 1993, per month on average 1.5% of the people in the labour force became unemployed. For West Germany this average monthly inflow rate into unemployment was between one and 1.2%. Table 1 Unemployment rate and unemployment dynamics from 1991 to 1993 (in percent) East Germany West Germany 1991 1992 1993 1991 1992 1993 Unemployment Rate 10.3 14.8 15.8 6.3 6.6 8.2 Gross Inflow Rate into Unemployment a) 1.4 1.6 1.5 1.0 1.0 1.2 Gross Outflow Rate from Unemployment b) 9.6 10.8 10.3 18.3 16.9 14.9 a) Average monthly inflow into registered unemployment divided by the labour force. b) Average monthly outflow from registered unemployment divided by the annual average unemployment stock. Sources: Federal Labour Office and Microcensus One reason for the low inflow rates may have been the special short-time allowance that was in force until the end of 1991. It gave workers at risk of losing their job time to find a new one without entering unemployment. In West Germany this program only applies to firms affected by temporary economic shocks, while in East Germany it applied to any firm with a need to Schmollers Jahrbuch 123 (2003) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 The Speed of Leaving the Old Job 59 restructure. 3 In 1991 1.8 million workers – more than 25% of the East German employment – were on this program. Collective agreements on temporary protection against dismissal that were in force until June 1991 also partly explain the moderate inflow rates. Table 1 also shows that the outflow rates were low by Western standards. From 1991 to 1993 the monthly average outflow rate from unemployment in East Germany was close to 10% while in West Germany it ranged from 14.9 to 18.3%. The low East German outflow rates may reflect the fact that employers hired workers directly from their old jobs rather than from the pool of unemployed. 3. Methods and data 3.1 Econometric methods We analyse the discrete random variable duration of the old job, T. The hazard rate measures the probability of exiting from the old job in month tto any destination state, provided that the individual has been in the old job up to the beginning of that month. A hazard rate may additionally be conditioned on a vector of explanatory variables x  t  . The conditional probability of exiting to a specific destination state, J  j, is called transition rate and is denoted as: P  T  t ; J  j j T  t ; x  t    j  t j x  t  ; j  1 ; ... ; M  1  The unconditional probability that a spell does not end with an exit up to the month T  tis the survival probability and is defined as: S  t j x  t   Y t   1 1 ÿ X M j  1  j  t j x  t  "#  2  We assume the logistic specification for modelling the discrete random variable T (Allison, 1982; Jenkins, 1995). The transition rates are then specified as  j  t j x  t   exp   tj  x  t 01  j  1  P M m  1 exp   tm  x  t 01  m  ; j  1 ; ... ; M ;  3  where  tj represents a parameter for the time effect, which may vary from month to month. The parameter vector,  j, of the covariates differs for each Schmollers Jahrbuch 123 (2003) 1 3Short-time workers with children received 68% of their net wage loss, while those without children received 63%. OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 60 Joachim Wolff and Parvati Tru ¨bswetter exit state. We consider transitions to a new job and to unemployment. Withinfirm job changes are not considered to be a transition. Spells that end with a transition to some destination state jcontribute to the likelihood by the product of the corresponding transition rate with the survival probability. Spells that end without a transition during our observation period (right-censored spells) contribute to the likelihood function by the survival probability alone. Since transitions to non-participation are rare, we treat spells that end by such an exit as right-censored. Assuming the logistic specification, the likelihood function can be transformed into the likelihood of a multinomial logit model. 3.2 Data The GSOEP-East, a representative sample of the East German population on a household basis, is the data source for our analysis. 4 This panel study collects information on the start and end of jobs retrospectively, which enabled us to construct job spells of workers who held an ongoing job in July 1990. Their jobs are old jobs in the sense that they were created prior to monetary union. We defined July 1990 as the job start in order to analyse how long old job matches survived during the transition process. Completed spells end with an exit prior to 1994. Spells that continued in 1994 are regarded as right-censored. We used the calendar information to determine the exit state. Our sample was limited to workers who are younger than 53 years in 1990. Older workers had the additional option of exiting into special early retirement programs. Hence, their behaviour with respect to new job or unemployment exits would differ much from that of workers who were not eligible for early retirement. We also dropped job spells of women on maternity leave at their first interview in 1990. They did not provide information on a large number of important job-related covariates. We dropped spells if a covariate value is missing from the start of the spell. If a time-varying covariate is not missing at the start but at some other point in time during a spell, we treat the spell as right-censored in the month before the missing value occurs. This left us with a sample size of more than 2,400 observations. 3.3 Descriptive Statistics Table 2 displays the composition of our spell sample by destination state and gender, before spells are dropped owing to missing covariate information. Slightly more than 50% of the spells represent men. More than 22% of the spells of men but only 13.9% of the spells of women end in a job change (rows four and six). The modal female exit state is unemployment. We estiSchmollers Jahrbuch 123 (2003) 1 4For a general description of the GSOEP-East see Wagner et al. (1993). The data were provided by the Deutsches Institut fu ¨r Wirtschaftsforschung. OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 The Speed of Leaving the Old Job 61 Table 2 Composition of the sample of old job spells by destination state Men Women Number Proportion (percent) Number Proportion (percent) Sample size 1381 100 1264 100 Right-censored 805 58.3 730 57.8 Exits into new jobs 312 22.6 176 13.9 Exits into unemployment 228 16.5 334 26.4 Exits into non-participation 36 2.6 24 1.9 mated Kaplan-Meier survival probabilities in the old job. By the end of 1993 about 49.5% of men and 46% of women survived in their old job. We also estimated these probabilities for a sample of workers from the GSOEP-West, who held a job in July 1990. About 80% of men and 75% of women in West Germany still held their old job at the end of 1993. These proportions are substantially higher than for East Germany, and demonstrate that East German workers left their old jobs rapidly during the transition process. Table 3 shows the means of the main variables (and the standard deviations of continuous variables). The samples of 1,286 male and 1,193 female spells used in the regression analysis are smaller than those used to calculate the Kaplan-Meier estimates. The first set of covariates are the time intervals of the baseline hazard, which are not of equal length for various reasons. For example, we chose a very short interval of three months length for the period December 1991 to February 1992 to quantify the temporary rise of the exit rates after the short-time allowance ended. The last interval instead exceeds one year, as the number of exits in this interval is very small. The next set of covariates, age and the number of children, represent personal and household attributes. Table 3 shows that the share of young men exceeds that of young women. This is not surprising as young mothers are more likely to be non-participants than young fathers. As far as children are concerned the distribution is quite similar for both sexes. Table 3 also presents descriptive statistics on occupational qualification. 71% of men and 67.6% of women have no degree or an apprenticeship. 19% of men and 24.5% of women are either master craftsmen or got an engineering or technical degree. We refer to these categories as technical school. More than 10% of men and about 8% of women hold a degree from a university or a technical college. The last set of covariates characterize the old job in 1990. We distinguish between four occupational categories: The first – missing, self-employed or trainee – represents a very small share of both samples. The second, blue collar workers, is the modal group for men, while the third – foremen, un-or Schmollers Jahrbuch 123 (2003) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 62 Joachim Wolff and Parvati Tru ¨bswetter Table 3 Characteristics of the male and female spell samples Covariates Male sample Female sample Mean Std. Dev. Mean Std. Dev. Calendar Time July 90 – Aug. 90 0.090 0.093 Sept. 90 – Nov. 90 0.120 0.127 Dec. 90 – Feb. 91 0.108 0.114 Mar. 91 – May 91 0.095 0.099 June 91 – July 91 0.055 0.056 Aug. 91 – Nov. 91 0.103 0.103 Dec. 91 – Feb. 92 0.069 0.068 Mar. 92 – July 92 0.096 0.092 Aug. 92 – Dec. 93 0.264 0.247 Personal and household characteristics: Age < 26 0.183 0.158 26 to 35 0.334 0.346 36 to 45 0.278 0.298 > 45 0.205 0.199 Own children No children < 17 years 0.443 0.411 1 child 0.260 0.289 more than 1 child 0.297 0.300 Occupational qualification No degree or apprenticeship 0.707 0.676 Technical school 0.190 0.245 University or technical college 0.103 0.079 Characteristics of old job in 1990: Occupation Missing, self-employed, trainee 0.082 0.057 Blue-collar worker 0.624 0.295 Foremen, unor semi-skilled white collar workers 0.130 0.467 Professional or managerial position 0.165 0.181 Tenure in June 1990  2 years 0.163 0.159 3–4 years 0.126 0.135 5–7 years 0.121 0.135 8–10 years 0.103 0.111 11–15 years 0.164 0.175 > 15 years 0.323 0.285 Other ln (hourly wage) 1.781 0.734 1.726 0.647 Sample size (expanded sample) 27044 24402 Number of spells 1286 1193 Schmollers Jahrbuch 123 (2003) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38 The Speed of Leaving the Old Job 69 References Allison, P. D.(1982), Discete-time methods for the analysis of event histories, in: S. Leinhardt (ed.), Sociological Methodology, San Francisco, 61–98. Blanchard, O./Diamond, P. (1989), The beveridge curve, Brookings Papers on Economic Activity, 0, 1–76. Boeri, T. (1994), Labour market flows and the persistence of unemployment in Central and Eastern Europe, in: T. Boeri (ed.), Unemployment in Transition Countries: Transient or Persistent?, Paris, 13–57. – (1997), Heterogeneous workers, economic transformation and the stagnancy of transitional unemployment, European Economic Review, 41, 905–914. Burda, M./Wyplosz, C. (1994), Gross worker and job flows in Europe, European Economic Review, 38 (6), 1287–1315. German Federal Labour Office (1994), Amtliche Nachrichten der Bundesanstalt fu ¨rArbeit: Arbeitsmarkt 1993, 43, Nu ¨rnberg. Hunt, J. (2001), Post-unification wage growth in East Germany, Review of Economics and Statistics, 83 (1): 190–195. – (2002), The transition in East Germany: When is a ten-point fall in the gender wage gap bad news? Journal of Labor Economics, 20 (1), 148–169. Jenkins, S. P. (1995), Easy estimation methods for discrete-time duration models, Oxford Bulletin of Economics and Statistics, 57 (1), 129–137. Licht, G./Steiner, V. (1994), Where have all the workers gone? Employment termination in East Germany after unification, in: F. Buttler/G.-G. Wagner (eds.), Labour market dynamics in present day Germany, Frankfurt, New York, Boulder (Colorado), 40–66. Pissarides, C. (1994), Search unemployment with on-the-job search, Review of Economic Studies, 61, 457–475. Wagner, G. G./Burkhauser, R. V./Behringer, F. (1993), The english language public use file of the German Socio-Economic Panel, Journal of Human Resources, 28 (2), 429–433. Wolff, J. (1998), Essays in Unemployment Duration in two Economies in Transition: East Germany and Hungary. PhD thesis, European University Institute, Florence. Schmollers Jahrbuch 123 (2003) 1 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/schm.123.1.55 | Generated on 2023-04-04 12:31:38