Reversing Early Retirement in Germany A Longitudinal Analysis of the Effects of Recent Pension Reforms on the Timing of the Transition to Retirement and on Pension Incomes Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld Abstract: This article investigates the effects and risks of recent pension reforms in Germany. While German pension policy systematically supported early retirement for many years in order to relieve the regulated labour market in times of economic stagnation, there has been a substantial change of the pension policy paradigm in the more recent past. Latest reforms expect older people to prolong working life. Using data from the German Socio-Economic Panel (GSOEP) and applying micro- level longitudinal research methods, this contribution shows that the recent rever- sal of early retirement in Germany has been at the price of growing social inequali- ties in old age. Keywords: Early retirement · Reversal of early retirement · Social inequality · Pension income · Longitudinal research 1 Introduction For several decades, German pension policies strongly relied on pushing older em- ployees out of the labour market very early. The massive expansion of early re- tirement programmes since the 1970s had mainly an economic background. Early retirement was systematically used to relieve the comparatively regulated German labour market in times of economic stagnation and increasing unemployment rates (see, for example, Guillemard 1991; Kohli et al. 1991; Arnds/Bonin 2002; Gruber/ Wise 1999, 2004, 2005; Blossfeld e al. 2006). As a result, the transition out of em- ployment became strongly destandardised in Germany and the legal retirement age of 65 was (and still is) in practice meaningless for the timing of the employment exit of most older workers. In East Germany, after reunifi cation, the actual age of employment exit was even approximately at the age of55 (i.e. ten years before legal retirement age, see Buchholz 2008). This strong emphasis on early retirement in Germany also becomes clear by looking at the employment rates of older men in Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft Vol. 38, 4 (2013): 881-906 (Date of release: 19.12.2013) © Federal Institute for Population Research 2013 URL: www.comparativepopulationstudies.de DOI: 10.12765/CPoS-2013-23en URN: urn:nbn:de:bib-cpos-2013-23en6 • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld882 pre-retirement age, which is the age group of 60- to 64-year-olds. For many years, only less than one third of the older men in this age group were still working in Ger- many. Indeed, in comparison to several other European and Northern American so- cieties, the employment rates of older people have been below average in Germany for many years (Hofäcker/Pollnerová 2006; Ebbinghaus 2000, 2008; Börsch-Supan 1992, 2000). Yet, for several years now, we can observe a clear policy change. German policy makers increasingly expect older people to prolong their working life. For example, reforms have been introduced increasing the penalties in case of an early with- drawal from employment, various early retirement paths were gradually closed, the share of private pensions was increased,1 and since the beginning of 2012, the legal retirement age has been incrementally increased from the age of 65 to the age of 67. These various reforms are a reaction to the growing fi nancial burdening of the public pension system in times of demographic aging which also creates an increas- ing imbalance between those persons contributing to the public pension system and those claiming pensions in Germany. Further, these reforms aim at reducing the growth of non-wage labour costs caused by increasing needs of the social in- surance system. For several years now, about 20 percent of the gross wage of a regular dependent employee in Germany is spent on the public pension insurance (Deutsche Rentenversicherung 2012). This article aims at analysing the effects of the profound pension policy change in Germany. At the core of our empirical analyses will be the following questions: How successful is the new policy paradigm? Are people really able to prolong their working lives, as intended by recent pension reforms or do they still exit early from employment? And, are all older people able to meet the new policy expectation to prolong working life or do specifi c parts of the older population fail in main- taining longer employment careers? Finally, what are the risks of recent reforms especially with regard to fi nancial well-being during retirement?2 To answer these questions, we investigate the late working lives and pension incomes of older peo- ple of the three birth cohorts 1934-39, 1940-45 and 1946-51. Specifi cally, we ex- amine the timing of their transition to retirement, the timing of their transition to non-employment, their risks of old age unemployment as well as the level of their 1 It has to be noted, however, that the share of private pensions is still very low in Germany (Börsch-Supan et al. 2008; Börsch-Supan/Wilke 2003), especially compared to liberal welfare states, such as the United States of America, but also compared to various European societies. 2 This set of questions was also at the core of the international comparative research project fl exCAREER which was recently completed at the Universities of Bamberg and Göttingen in Germany and which was funded by the German Research Foundation (DFG). The central aim of the fl exCAREER project was to empirically investigate if, how, and to what extent, the employ- ment and income situation of late-career employees and retirees have developed in nine Euro- pean societies (that is Germany, the Netherlands, Italy, Spain, Denmark, Sweden, Great Britain, Hungary and Estonia) and the United States of America in times of demographic aging. For the detailed results of all country studies and the international comparison, please see Blossfeld et al. (2011). Reversing Early Retirement in Germany • 883 pension incomes. Our longitudinal analyses are based on data from the German Socio-Economic Panel (GSOEP). The structure of our article is as follows: In the next section, we present an over- view of the German pension system paying special attention to the massive expan- sion of early retirement regulations in the 1970s and 1980s and the profound change German pension policy has undergone since the 1990s. Afterwards, we outline our research question and research design. In the succeeding section, we present the results of our empirical micro-level analyses. We conclude with a short summary and discussion of our results. 2 “Breaking with Traditions” in Germany: From a system supporting early retirement to a system increasingly penalizing early exits In Germany, the transition to retirement was massively fl exibilised and destandard- ised since the 1970s. Mainly as a reaction to the stagnation of economic growth after the oil price crises in the 1970s, growing unemployment rates and increasing labour market problems, the German government introduced various reforms which allowed fi rms to lay off older workers in a “socially peaceful” manner by sending them into early retirement (Kohli et al. 1991; Wübbeke 1999; Buchholz 2008). Since this time, the legal retirement age of 65 became increasingly meaningless (e.g. Wüb- beke 1999; Buchholz 2008; Ebbinghaus 2008). The fi rst fl exibilisation of retirement was introduced with the pension reform of 1972 which allowed an early transition to retirement already at the age of 63 for large parts of the workforce (Arnds/Bonin 2002: 12). Although this fi rst reform was actually introduced for reasons of work hu- manisation, early retirement soon became a common instrument of fi rms to adapt to a more tense economic situation, and sending older employees into early retire- ment became increasingly popular to relieve the regulated national labour market in times of growing unemployment rates (Esping-Andersen 1990: 227; Gatter/Hart- mann 1995: 413; Wübbeke 1999: 105; Buchholz 2008: 105-110). Yet, the introduction of the fl exible retirement age of 63 was not the only reform. In the 1980s, various other reforms have been undertaken to allow an even earlier transition to retirement. For example, between 1984 and 1988 a special early retire- ment scheme (the so-called “Vorruhestandsregelung”) was introduced to face grow- ing labour market problems. This special scheme lead to fi rms already “laying off” older employees at the age of 58 while generously fi nancing the period until these former older employees became eligible for public pensions. Further, a special pro- gramme for partial retirement was introduced which allowed an earlier withdrawal from the labour market. However, one of the most prominent early retirement pro- grammes became the German unemployment insurance which allowed fi rms to “lay off” older workers already at the age of 57 years and 4 months. In the subsequent 32 months (i.e. until the age of 60), these people were eligible for generous unemploy- ment insurance payments replacing about 60 to 67 percent of the former net income. Often, these people received additional payments from their former employer. At the age of 60, persons were then able to retire via a special scheme of the Ger- • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld884 man pension system which was originally designed to allow long-term unemployed to retire earlier (Gatter/Hartmann 1995; Arnds/Bonin 2002; Schmid 2002; Buchholz 2008). It has to be noted that this unemployment path to retirement was fi nancially highly attractive for both employers and older employees. Older employees usually did not face a signifi cant worsening of their fi nancial situation due to subsidy pay- ments by their former employer. For employers, the unemployment scheme allowed to shed older workers already several years before both the legal and fl exible retire- ment ages, while the vast part of the former payroll costs for the older workers were paid by the German unemployment and pension insurance, and employers only paid about one third of these older workers’ former net wage (Buchholz 2008). Yet, also for the German government this solution was highly attractive because it allowed to lower the offi cial unemployment statistics3 and to introduce “hidden dismissals” within the highly regulated German labour market. For East Germany, the German government even extended this regulation in the fi rst years after reunifi cation al- lowing older workers to claim unemployment insurance payments not only for 32 months but even for 5 years (so-called “Altersübergangsregelung”). As a result of this regulation, older workers in East Germany “retired” already at the age of 55 (which is 10 years before the legal retirement age and 8 years before the fl exible retirement age) in the fi rst years after reunifi cation (Ernst 1996; Buchholz 2008). It has to be noted that these older unemployed in East Germany never entered the of- fi cial German unemployment statistics. Offi cially, the share of unemployed persons among those aged 55 or older was less than 5 percent in autumn 1992 although this special unemployment scheme was used by almost one million people in the fi rst two years after German reunifi cation (Ernst 1995; Buchholz 2008). Against the background of this massive expansion of early retirement regula- tions it is thus not surprising that old age employment rates strongly decreased in Germany since the 1970s. Figure 14 shows the development of employment rates of 60 to 64 year old men in Germany and various other countries. It becomes clear that employment rates of workers in pre-retirement age decreased in all coun- tries since the 1970s. Yet, the magnitude of this decline was very different, and Germany is one of the countries in which the decline was the strongest (we fi nd a decrease of about 43 percentage points between 1970 and 2000). More detailed micro-level analyses based on longitudinal data (Buchholz 2008) have shown that 3 For many years, many of the older unemployed were not presented in offi cial unemployment statistics (Engstler/Brussig 2006). However, in our empirical study we will be able to capture the share of the older unemployed more realistically because our analyses will be based on data from the German Socio-Economic Panel in which persons interviewed report the labour market status. Thus, our analyses will not be (as) biased as the offi cial unemployment statistics in Germany. 4 In this fi gure, which is based on cross-sectional data, we only present the employment rates of men in order to avoid confusing the development of early retirement with increasing female employment rates which occurred in the same period. The selection of the countries presented in this fi gure is based on the countries that participated in our international comparative re- search project fl exCAREER at the Universities of Bamberg and Göttingen in Germany (see foot- note 2). Reversing Early Retirement in Germany • 885 the use of early retirement was highly selective and took place particularly in fi rms and sectors which faced growing economic uncertainties and a high pressure of rationalisation and restructuring (i.e. huge fi rms of the classical industrial sector). This clearly proves that employers have also been main actors for the expanding use of early retirement in Germany. A comparable picture could be drawn for the Netherlands, which also belong to the conservative welfare regime, as well as for the Southern European5 countries in which early retirement was also massively used to relieve the economy and the regulated national labour markets. Fig. 1: Employment rates of 60–64 year old men in 1970, 1980, 1990 and 2000 0 10 20 30 40 50 60 70 80 90 DE NL ES IT SE DK US UK HU EE 1970 1980 1990 2000 Employment Rate Notes: NL 1971 instead of 1970, ES 1972 instead of 1970, DK 1983 instead of 1980, UK 1984 instead of 1980, HU 1992 instead of 1990; DK 1970 participation instead of employ- ment rate; EE 1980 based on ILO estimates. Source: Own calculations based on OECD Labour Force Statistics 5 As the retirement age in Italy is very low, one would have to extend the analysis to the age group 55 to 59 years to show a development which is comparable to Germany, the Netherlands and Spain (Hofäcker/Pollnerová 2006). • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld886 Still, while in the 1970s, 1980s and early 1990s supporting the massive use of early retirement seemed to be an appropriate and effective measure to tackle the increasing labour market problems and economic stagnation, this has strongly changed in the more recent past which is marked by debates about the fi nancial sustainability of the national pension system and efforts aiming at unburdening the national welfare state budget. For example, in 2000, public pension expendi- tures amounted to about 200 billion Euros in Germany, representing approximately 20 percent of public spending and 12 percent of the GDP (OECD 2001; Börsch-Su- pan/Wilke 2003). Against the background of these fi gures, it is hence not surprising that public debates and political decisions on pension schemes have altered dra- matically in Germany in the past ten to fi fteen years. Actually, the efforts of today’s pension policy aim at setting incentives for maintaining long(er) working lives in order to compensate the increasing fi nancing problems of the public pension sys- tem in times of demographic aging (Börsch-Supan 2003). There have been several reforms which expect people to work longer or, in case they are unable to do so, to “pay the price” for an early employment exit. With the pension reforms of 1992 and 1999, access to early retirement programmes has been increasingly restricted by gradually closing various paths to early retirement, raising the legal retirement age, and increasing pension reductions in cases of early exit from the labour market. The pension reform from 2001 aimed at reducing public pension benefi ts, and addition- ally strengthened the incentives for private pension savings. Further, the so-called “Hartz” labour market reforms led to a decrease in early retirement as the oppor- tunities to retire early after a period of unemployment were reduced. Although we are not yet able to comprehensively access the effects of this reform, it has to be mentioned that since the beginning of 2012, the legal retirement age is being pro- gressively increased to the age of 67. Although the German public pension system can still be described as relatively generous compared to those of other countries (especially those of liberal welfare states such as the U.S. and Great Britain; Börsch-Supan et al. 2008; Börsch-Supan/ Wilke 2003), it has to be noted that these reforms caused a fundamental shift in Germany’s general pension policy paradigm. In the 1970s, 1980s and early 1990s, the German pension system was also very generous with those who “failed” to work until legal retirement age. Actuarially non-neutral early pensions enabled a secure fi nancial situation for large parts of the older population. Consequently, Germany’s original policy aimed at reducing social inequalities among older peo- ple with unequal opportunities on the labour market. However, the most recent reforms fundamentally changed this situation. The German government strongly decreased its efforts to buffer and cushion employment risks in later life. Instead, with the latest reforms, labour market risks are increasingly privatised and indi- vidualised, and older people are increasingly expected to account for those risks by themselves. Generally, these recent reforms in Germany (and other European countries) start from the premise that the transition to retirement and the use of early retire- ment can be explained by micro-economic theory. It is argued that early retire- ment is a result of rationally acting and maximizing individuals (see, for example, Reversing Early Retirement in Germany • 887 Börsch-Supan 1998; Riphahn/Schmidt 1997; Siddiqui 1997). According to this ar- gument, older people opt for early retirement (and more leisure time) as long as the national pension system offers generous early pensions that are not or only slightly adjusted to the fact that these people leave employment earlier, no longer contribute to the social security system and claim pensions for a longer period. In this theoretical framework of modelling the transition to (early) retirement, the main (and only) actor is the older worker. According to this approach, the different rates of early retirement in various countries can be explained only by the differ- ent national (early) retirement systems. It is argued that, in Germany, the rate of early retirement is much higher than in the UK, for example, because pensions of early retirees are very generous (Börsch-Supan 1998; Riphahn/Schmidt 1997; Siddiqui 1997). Thus, the so-called “pull factor” of the pension system is much stronger in Germany than in the liberal welfare state model of the UK. However, the micro-economic approach is also heavily criticised, especially by sociologists (see, for example, Kohli 1991; Arnds/Bonin 2002; Blossfeld et al. 2006; Ebbing- haus 2008). These authors state that employers, governments as well as various country-specifi c institutional features also have to be modelled in order to ap- propriately understand early retirement. Further, there is not only a “pull factor” but also a “push factor” that determines early retirement. For example, employers and governments use early retirement to react to growing economic uncertainty, to realise economic restructuring and to relieve the national labour market. Thus, it has to be critically examined in our following analyses if reforms solely targeting and reducing the “pull factors” (by increasing pension reductions in case of early retirement) are likely to be successful. Further, the actual and potential risks of such reforms with regard to social inequalities among the retired population have to be identifi ed. Figure 2 presents the employment rates of 60- to 64 year old men in Germany and other modern societies for 2000 and 2009.6 As can be seen, we indeed fi nd a substantial growth of the employment rate among the older workforce in Germany. Within only nine years, the employment rate of 60 to 64 year old men increased by about 20 percentage points. Not surprisingly, German policy-makers celebrate this increase in old age employment as a huge success of their latest reforms. 6 Again, we will focus on the employment rates of men in order to avoid mixing up early retire- ment trends with increasing female employment rates across cohorts. The selection of coun- tries is again based on the countries which participated in our international comparative fl exCA- REER project (see footnote 2). • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld888 3 Research Approach and Design Necessity for a longitudinal research approach to assess the reversal of early retirement For a variety of reasons it is very risky to assess the success of recent pension reforms on the basis of broad cross-sectional fi gures as they are used in Figure 2: • First of all, this simple cross-sectional approach does not allow to control whether the recent increase in old age employment is caused by a chang- ing composition of the older workforce. It is well known that, as a result of educational expansion, the qualifi cation level of today’s cohorts of older workers is signifi cantly higher compared to that of past cohorts. As higher qualifi ed people tend to retire later, we would thus “automatically” fi nd a trend towards longer working lives across time. Yet, in that case, the increase of old age employment rates would be the result of the different educational composition of recent cohorts of older workers and not the result of the Fig. 2: Employment rates of 60-64 year old men in 2000 and 2009 0 10 20 30 40 50 60 70 80 90 DE NL ES IT SE DK US UK HU EE 2000 2009 Employment rate Source: Own calculations based on OECD Labour Force Statistics Reversing Early Retirement in Germany • 889 recent pension reforms. Simple cross-sectional estimations do not allow ap- propriately controlling for this fact. • Second, on the basis of these cross-sectional fi gures, we are not able to understand whether the whole strata of the older workforce benefi tted from increasing employment rates in later life or if only parts of the older work- force (namely, already privileged workers, such as the higher qualifi ed) are able to maintain longer working lives while already disadvantaged and lowly qualifi ed older persons still have to retire early and – compared to the 1970s, 1980s and 1990s – bear the price of signifi cant pension reductions. • Thirdly, with cross-sectional data, it is impossible to causally model and es- timate the consequences of the timing of the transition to retirement on indi- viduals’ pension incomes. However, this would be very important in order to understand if recent reforms have been really successful or whether instead, they rather tend to be a hidden pension reduction, maybe even at the price of increasing social inequalities in old age. In fact, a micro-level longitudinal research design is necessary to address all these important questions. Our following empirical analyses are thus based on data from the German Socio-Economic Panel (GSOEP) which provides longitudinal information on individuals and households in Germany since 1984. Using GSOEP data allows us to investigate if there remains a signifi cant effect for the reversal of retirement after controlling for potential changes in the composition of the older workforce across cohorts (especially the increasing level of education across co- horts). Additionally, by using a micro-level research design, we are able to assess whether all parts of the older workforce benefi tted from the recent increase of old age employment rates or not. Finally, we can carry out causal analyses that allow us to investigate the fi nancial consequences of early retirement and changes therein in the light of latest pension reforms in Germany. Research design, data used and sample description For our purpose, we use a sub-sample of the GSOEP by selecting 50 year old people born between 1934 and 1951 and by following their working lives and their transi- tion to retirement. Our empirical analyses are based on GSOEP data for the years 1984 till 2007. For the years 1984 to 1989, the GSOEP includes information for only Western Germany; since 1990, it also includes Eastern German households. Mi- grants are overrepresented in the GSOEP in order to allow detailed analyses of this sub-group of the population. With regards to our research design, we apply the following modelling: As the transition to non-employment and the transition to retirement are not necessarily co-occurring events in Germany, it is important to differentiate between direct and indirect transitions to retirement in order to assess the success, risks and effects of the latest pension reforms appropriately. Consequently, we will inves- tigate both the indirect and direct transition to retirement by, on the one hand, modelling the point in time of the transition to retirement and, on the other hand, • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld890 the point in time of the transition to non-employment. The transition to retirement captures the time when a person fi rst claims pension benefi ts and is no longer part of the labour market. In contrast, the transition to non-employment indicates the time when persons end working life. For those who make a direct transition from employment to retirement, the time of the transition to non-employment would obviously be identical to the time of the transition to retirement. However, for those making an indirect transition to retirement after a phase of unemployment, we would no longer use the time when persons claim pension benefi ts for the fi rst time (e.g. at the age of 60), but the time when these persons became unemployed (e.g. the age of 57).7 Additionally, we model transitions to unemployment.8 This enables us to assess whether the recent reversal of early retirement has also been able to bring the massive use of unemployment as a path to early retirement to a halt or if the reversal of old age unemployment has been less effective. Finally, we model the level of pension incomes of the persons in our sample in order to exam- ine if the recent attempts of policy-makers to reverse early retirement have been accompanied by signifi cant pension losses in younger birth cohorts of retirees. Our sample for analysing the transition to retirement consists of individuals who were employed or unemployed at the age of 50. Following this defi nition, our sam- ple for these analyses consists of 1993 men and 1422 women (55 percent West Germans, 21 percent East Germans, 24 percent with a migration background). To analyse the transition to non-employment and the risks of old age unemployment, of course, only individuals who were employed at the age of 50 were selected. Ac- cording to this defi nition, our sample for the unemployment analyses includes on 1853 men and 1255 women (58 percent West Germans, 19 percent East Germans and 23 percent with a migration background). To estimate developments over time, we compare three different birth cohorts, that is, people born in 1934-39, 1940-45 and 1946-51. While the two older birth co- horts benefi tted from the early generous retirement regulations, the youngest birth cohort 1946-51 is the one most affected by latest pension reforms in Germany. This cohort is expected to rather prolong their working life or, in case of failing to do so, to “pay the price” for an earlier employment exit.9 In order to study the causal relationship between older individuals’ success on the labour market, their path to retirement and the resulting effects on their pension incomes, we introduce various 7 Obviously, it could be critically argued that a transition to unemployment does not necessarily have to be followed by a transition to retirement but could instead also be followed by a transi- tion back to employment. Yet, as an additional sequence analysis with our data has shown, the transition to (fi rst) unemployment after the age of 50 is usually connected with a subsequent transition to retirement. Thus, our analyses for transitions to non-employment and unemploy- ment can de facto serve for modelling indirect transitions to retirement. 8 We use the self-reported labour market status of the respondents of the GSOEP which is avail- able for all years of the panel study. 9 Unfortunately, we are not able to differentiate between more birth cohorts due to the small sample sizes. However, for the reader it is important to know that the youngest birth cohort of our analyses is the one which is most affected by the recent pension policy change in Germany. Reversing Early Retirement in Germany • 891 covariates, namely the individuals’ level of education and the characteristics of their employer, that is, fi rm size and branch of industry, into our models. Methodology For the longitudinal analyses presented in this article, we apply descriptive and mul- tivariate methods of event history analysis. Event history analysis models the time- dependency and dynamics of the individuals’ transitions and life events (Blossfeld/ Rohwer 2002). As a descriptive method, we use product-limit estimations which calculate so-called survivor functions (Kaplan/Meier 1958) that allow us to under- stand how long it takes until the individuals of our sample experience the event of concern (in our case, for example, the transition to retirement). Additionally, we apply logistic regression models (Yamaguchi 1991) which permit to include vari- ous time-constant and time-varying covariates into our longitudinal analyses. This enables us to estimate if the rate and time of a specifi c event is infl uenced by the characteristics of the respondents (e.g. their educational level or the economic sec- tor they are working in). To estimate the absolute pension income, we fi nally use linear regression models. 4 Empirical evidence for the effects of recent reforms The transition to non-employment and to retirement – descriptive fi ndings Figure 3 presents the results of product-limit estimations for the transition to non- employment and the transition to retirement for the three birth cohorts under study (in a fi rst step, for both men and women; in a second step for men only to make our longitudinal results more comparable to the cross-sectional results presented in Figures 1 and 2). These product-limit estimations allow us to descriptively investi- gate the particular time of the transition to retirement and the transition to non-em- ployment as well as changes in these transitions across cohorts. For our youngest cohort, that is the cohort 1946-51, observations end at the age of 61, the time of their last interview in the GSOEP data we use. These longitudinal analyses indicate a clear delay of the transition to retirement in Germany in recent years. For example, at the age of 60, almost 25 percent of the oldest cohort have retired (among men: 24 percent). In the middle cohort 1940-45, this fi gure reduces to 17 percent (among men: 20 percent) and amounts to only 10 percent in the youngest cohort 1946-51. At the age of 63, which was introduced as the fl exible retirement age in Germany in 1972, the majority of the respondents born between 1934 and 1939, namely 68 percent, have retired (among men: 64 per- cent). In the cohort 1940-45, this was only the case for about 58 percent of the respondents (among men: 50 percent). Thus, we fi nd a signifi cant delay in the tran- sition to retirement for younger cohorts. Yet, these estimations also highlight that neither the legal retirement age of 65 nor the fl exible retirement age of 63 play an important role for structuring the timing of the retirement transitions in Germany. • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld892 Fig. 3: Transition to retirement vs. transition to non-employment by birth cohorts (survivor percent derived from product-limit-estimations) 0 10 20 30 40 50 60 70 80 90 100 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 Age Retirement 1934-39 Retirement 1940-45 Retirement 1946-51 Non-employment 1934-39 Non-employment 1940-45 Non-employment 1946-51 0 10 20 30 40 50 60 70 80 90 100 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 Age Retirement 1934-39 Retirement 1940-45 Retirement 1946-51 Non-employment 1934-39 Non-employment 1940-45 Non-employment 1946-51 Men and Women Men only percent percent Notes: The analyses for the transition to retirement are based on persons employed or unemployed at the age of 50. For the transition to non-employment, our estimations are based only on persons employed at the age of 50. Source: Own calculations based on the GSOEP (1984-2007). Reversing Early Retirement in Germany • 893 However, there is another important fact that can be derived from Figure 3. Comparing the estimates for the transition to retirement with the estimates for the transition to non-employment (which could be either a transition to retirement or a transition to unemployment) reveals that for a substantial number of older people the transition to retirement indeed takes place only indirectly after a phase of unem- ployment. Although the product-limit estimations for the transition to non-employ- ment also indicate a prolongation of working life across birth cohorts, it has to be stated that in all cohorts of our study people tend to exit employment signifi cantly earlier than they retire. The difference between the direct and indirect transitions to retirement amounts up to remarkable 20 percentage points. For example, in the cohort 1934-39, 24 percent entered retirement at the age of 60 while 42 percent have already left employment. In the cohort 1940-45, 17 percent retired at the age of 60 while 33 percent have already left employment at this age. In the youngest co- hort 1946-51, the share of those who retired at the age of 60 amounts to 10 percent while the share of the non-employed is about 25 percent. Consequently, there is a considerable gap between these two transitions. This clearly shows that indirect transitions to retirement after a phase of unemployment are widespread in Ger- many. This result already indicates that reforms aiming at reversing early retirement in Germany should not only be concerned about changing pension regulations but should also target the employability of older people. Transition to retirement In the following, we present the results of our multivariate models for the timing of the transition to retirement and changes therein across cohorts (Table 1). This multivariate approach enables us to examine if the delayed transition to retirement across cohorts remains signifi cant after controlling for additional factors, especially the educational level of the individuals which has risen across cohorts. In these more complex models we also fi nd that younger birth cohorts (that is people born between 1940-45 and 1946-51) retire signifi cantly later than the cohort of people born between 1934 and 1939. These effects remain highly signifi cant even if we additionally control for various factors, namely the level of education, branch of industry and fi rm size (models 2, 3 and 4). Hence, we fi nd a delay of retirement in Germany that is indeed independent of the changing composition of the population of the older workforce. Table 1 additionally reveals an infl uence of individuals’ educational level, although this infl uence is weak. Only very highly qualifi ed people with a college or university degree retire signifi cantly later. For the remaining people (which is the majority of the workforce), a signifi cant difference cannot be found. Yet, it has to be noted that in the same model we control if the respondents experience the transition to retirement after a phase of unemployment or not, and it is likely that especially lower qualifi ed people face a higher risk of retiring only indirectly after a phase of unemployment. Our following models on the risks of ending working life in unemployment (see Ta- ble 2) test if educational differences increase when we look specifi cally at indirect transitions to retirement which have become less favourable for individuals. • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld894 Tab. 1: Timing of the transition to retirement (logistic regression model) Notes: ** Effect signifi cant at p < 0.01; * effect signifi cant at p < 0.05; + effect signifi cant at p < 0.10. In our models, we also control for the cumulative experience in unemployment, self-em- ployment, marginal and part-time employment. Yet, as we focus on the interpretation of cohort, education, branch of industry and fi rm size and do not interpret the results of these covariates, we decided to not present the effects in this table. More detailed information are provided by Rinklake and Buchholz (2011). Source: Own calculations based on the GSOEP (1984-2007) 1 2 3 4 Constant –4.36** –4.30** –4.18** –4.27** Age 50–57 (ref.) –– –– –– –– 58–59 0.66** 0.68** 0.67** 0.65** 60–61 2.91** 2.95** 2.95** 2.96** 62–63 2.52** 2.59** 2.59** 2.61** 64 plus 3.47** 2.58** 3.61** 3.64** Population group West German (ref.) –– –– –– –– East German –0.14 –0.06 –0.10 –0.02 Migration background –0.28** –0.38** –0.38** –0.39** Sex Men (ref.) –– –– –– –– Women 0.17* 0.12 0.15+ 0.20* Birth cohorts 1934–39 (ref.) –– –– –– –– 1940–45 –0.51** –0.46** –0.45** –0.45** 1946–51 –1.09** –1.01** –1.00** –0.97** Late career characteristics Currently unemployed 0.80** 0.78** 0.78** 0.77** Employed at the age of 50 0.44** 0.46** 0.37** 0.28** Educational level Lower secondary degree without occupational qualification 0.06 0.05 0.02 Lower secondary degree with occupational qualification (ref.) –– –– –– Upper secondary degree without occupational qualification –0.08 –0.12 –0.16 Upper secondary degree with occupational qualification –0.06 –0.07 –0.09 College or university degree –0.65** –0.66** –0.72** Branch of industry Extractive industry 0.49* 0.55* Transformative industry (ref.) –– –– Private services –0.12 –0.11 Social services 0.03 –0.01 Firm size Up to 19 employees –0.14 20–199 employees (ref.) –– 200–1,999 employees 0.13 Over 2,000 employees 0.50** Events 1,131 1,131 1,131 1,131 Total persons 3,415 3,415 3,415 3,415 Censored persons 2,284 2,284 2,284 2,284 –2*diff (LogL) 2,518.92 2,555.92 2,566.48 2,605.23 Reversing Early Retirement in Germany • 895 Finally, the results in Table 1 again display a signifi cant infl uence of the employ- er’s characteristics on the timing of the transition to retirement. As other empirical studies have already revealed (Wübbeke 1999; Buchholz 2008), the option of early retirement is especially used in large fi rms. It is argued that especially in huge and tayloristically organised fi rms early retirement is used to realise restructuring and downsizing after the stagnation of the economic miracle in the 1970s. Regarding the branch of industry, we fi nd a signifi cant difference between the transformative sec- tor (that is the classical industrial sector) and the so-called extractive sector which includes agriculture, but especially mining and related industries. Risks of old age unemployment In the next step of our empirical analyses, we concentrate on the risks of becoming unemployed after the age of 50 and consequently on the risks of making an indirect transition to retirement. Our descriptive analyses presented in Figure 3 have shown that, as a result of the massive use of the unemployment insurance as a “path to early retirement”, there is a signifi cant gap between the timing of the transition to retirement and the timing of the transition out of employment. Therefore, it is important to also analyse the transition to old age unemployment in order to ap- propriately understand early retirement in Germany. The results of our analyses are presented in Table 2. Firstly, it has to be noted that the risks of unemployment are especially high at the age of 58 and 59. Before and after this age, the risks of becoming unemployed are much lower. The reason for this systematic age effect is that, for many years, at this age, the German pension system allowed for a fi nancially secure and early withdrawal from the labour market. Thus, the unemployment insurance was used as a welfare state subsystem of the pension system (for more details, see section 2). In the recent past, however, this path to retirement was increasingly penalized. Yet, in order to answer our research question the results of the cohort compari- son are most important. Indeed, our multivariate analyses for old age unemploy- ment display a decreasing signifi cance for younger cohorts. Similar to the results for the transition to retirement, this cohort-specifi c change remains highly signifi - cant even after controlling for education, fi rm size and branch of industry (models 2, 3 and 4). Yet, compared to the trend reversal of early retirement, the measures to reverse the trend of using unemployment as a path to early retirement apparently were far less successful. In our analyses of the transition to retirement (Table 1), we fi nd signifi cant effects for both the middle and the youngest cohort. In our models of old age unemployment, however, the middle cohort 1940-45 does not differ signifi - cantly from our oldest birth cohort. We only fi nd a signifi cant effect for the young- est cohort 1946-51 and, compared to the cohort-specifi c coeffi cients estimated in our models for retirement (Table 1), the size of this coeffi cient is far smaller in the models for unemployment. Indeed, additional product-limit estimations show that even in the youngest cohort 1946-51, old-age unemployment is very widespread: 19 percent of this cohort became unemployed before the age of 58 compared to 24 percent in the cohort 1934-39 (Rinklake/Buchholz 2011). It thus can be concluded • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld896 Tab. 2: Transition to unemployment (logistic regression model) Notes: ** Effect signifi cant at p < 0.01; * effect signifi cant at p < 0.05; + effect signifi cant at p < 0.10. In our models, we also control for the cumulative experience in unemployment, self-em- ployment, marginal and part-time employment. Yet, as we focus on the interpretation of cohort, education, branch of industry and fi rm size and do not interpret the results of these covariates, we decided not to present the effects in this table. Further details are provided by Rinklake and Buchholz (2011). Source: Own calculations based on the GSOEP (1984-2007) 1 2 3 4 Constant –3.51** –3.39** –3.22** –3.08** Age 50–57 (ref.) –– –– –– –– 58–59 0.36** 0.39** 0.30** 0.40** 60–61 –0.27 –0.20 –0.20 –0.19 62–63 –0.34 –0.24 –0.21 –0.21 64 plus –0.82 –0.68 –0.70 –0.70 Population group West German (ref.) –– –– –– –– East German 0.89** 1.01** 1.03** 0.98** Migration background 0.41** 0.31** 0.21* 0.23* Sex Men (ref.) –– –– –– –– Women –0.19* –0.21* –0.06 –0.09* Birth cohorts 1934–39 (ref.) –– –– –– –– 1940–45 0.01 0.05 0.07 0.07 1946–51 –0.40** –0.29** –0.29** –0.31** Educational level Lower secondary degree without occupational qualification 0.01 –0.02 –0.01 Lower secondary degree with occupational qualification (ref.) –– –– –– Upper secondary degree without occupational qualification –0.18 –0.15 –0.12+ Upper secondary degree with occupational qualification –0.28** –0.22** –0.20* College or university degree –0.83** –0.61** –0.59** Branch of industry Extractive industry –0.16 –0.16 Transformative industry (ref.) –– –– Private services –0.39** –0.36** Social services –0.88** –0.86** Firm size Up to 19 employees 0.78 20–199 employees (ref.) –– 200–1,999 employees –0.25* Over 2,000 employees –0.32** Events 670 670 670 670 Total persons 3,108 3,108 3,108 3,108 Censored persons 2,438 2,438 2,438 2,438 –2*diff (LogL) 173.38 228.18 290.27 311.03 Reversing Early Retirement in Germany • 897 that the signifi cant increase of retirement ages is not necessarily connected to the fact that all older people were also able to maintain longer working lives. On the basis of the results presented in Table 2, it also becomes evident that the transition to unemployment and, thus, the indirect path to retirement, is much more stratifi ed than the timing of the transition to retirement. This applies to both individuals’ characteristics (i.e. educational level) and the characteristics of their employer (i.e. branch of industry or fi rm size). In contrast to the results for the timing of retirement, education is far more important for the transition to unemployment. It is no longer only the highest qualifi ed (namely individuals with college or university education) who differ from older individuals with lower secondary degree and oc- cupational qualifi cation. For the risks of ending working life in unemployment, we fi nd a signifi cant effect for those with upper secondary degree and occupational qualifi cation (models 2, 3 and 4) and even a slight effect for those with upper sec- ondary degree without occupational qualifi cation (model 4). Further, with regard to the branch of the industry, we now fi nd a broader variety. The risks of unemployment are especially high for older workers in the transforma- tive and extractive sector.10 Both industries have faced a high need for economic restructuring and downsizing in Germany since the 1970s, and our results show that employers clearly made use of the unemployment path to retirement to tackle growing economic pressure (Buchholz 2008). It is important to highlight these sig- nifi cant branch effects because this clearly shows that the transition to retirement (especially the indirect transition to retirement via unemployment) can obviously not be modelled as a decision solely made by older workers as proposed by the micro-economic pull approach. On the contrary, employers signifi cantly determine the transition to retirement, and this hidden form of employment fl exibility was also publicly supported to relieve the regulated German labour market for a long time. Transition to non-employment This section investigates the transition to non-employment, which accounts for di- rect and indirect transitions at the same time (Table 3). As the fi ndings for fi rm size and branch of industry remain the same as in the models for the transition to retire- ment, we concentrate on the results for the cohorts and the interaction between the cohort and the educational level. In our multivariate analyses for the timing of the transition to non-employment, we fi nd a signifi cant cohort-specifi c change. Compared to the oldest cohort 1934– 39, members of the birth cohorts 1940-45 and 1946-51 remain in employment long- er and the effects remain signifi cant even after controlling for education, fi rm size and branch of industry. Yet, compared to the models for the transition to retirement, 10 These effects remain highly signifi cant even if we control for occupational class in order to ac- count for the fact that the share of manual workers is higher in this sector. Additional analyses, in which the extractive sector was used as the reference category, also revealed signifi cant differences to the public and private service sector. • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld898 Tab. 3: Transition to non-employment (logistic regression model) Notes: ** Effect signifi cant at p < 0.01; * effect signifi cant at p < 0.05; + effect signifi cant at p < 0.10. In our models, we also control for the cumulative experience in unemployment, self-em- ployment, marginal and part-time employment. Yet, as we focus on the interpretation of cohort, education, branch of industry and fi rm size and do not interpret the results of these covariates, we decided not to present the effects in this table. Further details are provided by Rinklake and Buchholz (2011). Source: Own calculations based on the GSOEP (1984-2007) 1 2 3 4 5 Constant –3.16** –3.08** –3.10** –3.30** –3.31** Age 50–57 (ref.) –– –– –– –– –– 58–59 0.70** 0.72** 0.72** 0.71** 0.72** 60–61 2.37** 2.41** 2.42** 2.42** 2.43** 62–63 2.13** 2.19** 2.19** 2.20** 2.20** 64 plus 2.92** 3.01** 3.01** 3.03** 3.04** Population group West German (ref.) –– –– –– –– –– East German –0.41** –0.33** –0.36** –0.29** –0.30** Migration background –0.21** –0.29** –0.29** –0.28** –0.27** Sex Men (ref.) –– –– –– –– –– Women 0.31** 0.31** 0.27** 0.31** 0.31** Birth cohorts 1934–39 (ref.) –– –– –– –– –– 1940–45 –0.35** –0.31** –0.30** –0.29** –0.24* 1946–51 –0.73** –0.66** –0.66** –0.63** –0.74** Educational level Lower secondary degree without occupational qualification 0.05 0.04 0.01 0.01 Lower secondary degree with occupational qualification (ref.) –– –– –– –– Upper secondary degree without occupational qualification –0.06 –0.07 –0.11 –0.18 Upper secondary degree with occupational qualification –0.11 –0.10 –0.13 –0.11 College or university degree –0.48** –0.46** –0.51** –0.52* Interaction educational level * birth cohort Birth cohort 1940–45 Lower secondary degree without occupational qualification –0.22 Upper secondary degree without occupational qualification –0.06 Upper secondary degree with occupational qualification –0.03 College or university degree 0.01 Birth cohort 1946–51 Lower secondary degree without occupational qualification 0.56** Upper secondary degree without occupational qualification 0.46 Upper secondary degree with occupational qualification 0.02 College or university degree 0.03 Branch of industry Extractive industry 0.50** 0.53** 0.54** Transformative industry (ref.) –– –– –– Private services 0.02 0.01 0.01 Social services –0.01 –0.04 –0.04 Firm size Up to 19 employees 0.02 0.03 20–199 employees (ref.) –– –– 200–1,999 employees 0.16* 0.16* Over 2,000 employees 0.48** 0.48** Events 1,617 1,617 1,617 1,617 1,617 Total persons 3,108 3,108 3,108 3,108 3,108 Censored persons 1,419 1,419 1,419 1,419 1,419 –2*diff (LogL) 2,128.37 2,161.48 2,169.2 2,215.2 2,229.4 Reversing Early Retirement in Germany • 899 the size of the estimated coeffi cients is smaller. This again supports the conclusion, which we have already drawn on the basis of our analyses for unemployment, that the delayed transition to retirement in younger cohorts cannot be fully explained by the fact that all older workers of younger cohorts are also able to maintain longer working lives. Indeed, the interaction effects between cohort and educational level indicate that the prolongation of working life has been selective and came at the price of growing social inequalities among the older workforce. Especially the lowest qualifi ed older workers, that is those with a lower secondary degree without occupational qualifi - cation, are not able to meet the new policy expectation to remain in employment longer. In the youngest cohort 1946-51, these low-qualifi ed workers now signifi cant- ly differ from those with a medium educational level. This was not the case in the two earlier birth cohorts, in which the length of the working life was more or less the same for the majority of the older people (with the exception of the highest qualifi ed who always maintained longer employment careers, see above). Pension income In the last step, we now focus on the respondents’ pension incomes. Table 4 pre- sents the results of our analyses on the individual level of pension income after the respondents of our original sample claimed pensions for the fi rst time. With these fi nal empirical analyses we are able to evaluate how the income situation of retirees has developed since the mid-1980s and to what extent recent reforms have led to decreasing pension incomes. At fi rst glance, our results indicate that younger cohorts’ fi nancial situation is signifi cantly better. However, it has to be noted that with our data, we are not able to control for the individuals’ employment income throughout their entire working life although the individual working income strongly determines the level of pen- sion benefi ts persons receive after retirement in Germany. However, other authors (e.g. Mayer/Huinink 1990) report that employment incomes strongly vary between the cohorts of our study as a consequence of the economic boom Germany experi- enced in the 1960s and early 1970s and the specifi c employment opportunities this boom created for the cohorts of our study. Mayer and Huinink (1990) report that it is especially the middle and the youngest cohort which benefi tted from the economic boom. They entered the labour market at the time of these economically favourable conditions and therefore had better chances to achieve higher incomes. In contrast, people born in the 1930s entered the labour market after World War II when the general economic situation was less favourable. As a consequence, they started their career at a lower income level, and this had a long-term effect on their entire employment career (Mayer/Huinink 1990). Hence, the fact that the individual work- ing incomes in our empirical study have increased for the younger cohorts, directly affects their level of pensions we observe in our analyses presented in Table 4. The absolute pension incomes of the retirees of our younger cohorts are higher com- pared to those born in the 1930s. Indeed, when we include an individual’s last labour • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld900 market income as a proxy measure for the different individual working incomes of our cohorts (model 2), the cohorts no longer differ signifi cantly from each other. Yet, the most important question for the purpose of this article is whether those who (have to) use the path to early retirement via unemployment receive lower pen- sions than those who do not (have to) use it. Indeed, in model 1 we fi nd a signifi cant Tab. 4: Absolute pension income (linear regression model) Notes: ** Effect signifi cant at p < 0.01; * effect signifi cant at p < 0.05; + effect signifi cant at p < 0.10. Public, occupational and private pensions are included; the pension income is adjusted for infl ation. Branch of industry and fi rm size did not have any effect on the level of pension income and their inclusion did not improve our estimations. This is not surprising, because pub- lic pensions are still the main income source of retirees in Germany. Due to their lack of signifi cance, the effects of fi rm size and branch of industry are not presented in Table 4. Source: Rinklake and Buchholz (2011: 69-70); own calculations based on the GSOEP (1984-2007) 1 2 3 4 Constant 7.42** 6.93** 7.64** 7.29** Age 50–57 (ref.) – – – – 58–59 1.98* 1.07 1.67+ 1.86* 60–61 1.28+ 0.01 1.40* 1.55* 62–63 3.69** 2.03** 3.25** 3.39** 64 plus 4.25** 2.44** 3.47** 3.67** Population group West German (ref.) – – – – East German –2.77** –1.42* –3.90** –3.79** Migration background –2.54** –2.22** –1.45* –1.47* Sex Men (ref.) – – – – Women –5.01** –4.03** –4.41** –4.48** Birth cohorts 1934–39 (ref.) – – – – 1940–45 1.35* 0.34 0.90+ 1.08+ 1946–51 2.95** 1.14 2.57** 3.34** Last working income 0.01** Educational level Lower secondary degree without occupational qualification –1.07+ –1.09+ Lower secondary degree with occupational qualification (ref.) – – Upper secondary degree without occupational qualification –0.30 –0.14 Upper secondary degree with occupational qualification 1.30+ 1.41* College or university degree 7.05** 7.03** Late career characteristics Employed at the age of 50 1.95* –0.08 0.97 1.05 Unemployment path to retirement –1.22+ –0.77 –0.69 0.39 Retirement path * Birth cohort Unemployment path to retirement * Cohort 1940–45 –1.26 Unemployment path to retirement * Cohort 1946–51 –6.65** Adjusted R-square 0.14 0.28 0.20 0.21 Number of cases 1,206 1,206 1,206 1,206 Reversing Early Retirement in Germany • 901 effect indicating that those individuals who have been unemployed receive lower pensions. However, this effect is signifi cant only on a low level in our general model. Indeed, the introduction of interaction terms for the path to retirement and cohort into our model (see model 4) reveals that retiring via unemployment has become signifi cantly unfavourable for younger cohorts. In contrast to our two older birth co- horts, the unemployment path to retirement is connected with signifi cant pension losses in our youngest cohort 1946-51. This was not the case in previously born co- horts (model 4) for whom the German pension system still offered strong incentives to make use of this scheme. The profound recent pension reforms, however, put an end to this generous policy. Today, older people are expected to either prolong their working lives or accept higher penalties for early retirement. Hence, our em- pirical results clearly indicate that recent pension reforms in Germany which aim at reducing early retirement led to growing social inequalities in old age. This is due to specifi c parts of the older workforce, namely the low-qualifi ed (see Table 3), are not able to meet the new expectation to remain employed longer. Yet, in contrast to the past, these workers have to pay the price for their “failure” in the form of signifi cant pension losses (see Table 4). 5 Conclusion and Outlook The aim of this contribution is to investigate the effects and risks of recent pension reforms in Germany. For many years, German pension policy systematically sup- ported early retirement in order to relieve the regulated labour market in times of economic stagnation and growing rates of unemployment. For this purpose, various generous early retirement programmes have been introduced in the 1970s, 1980s and early-1990s. These programmes were systematically used by fi rms and the gov- ernment to reduce the workforce in a “socially peaceful” manner. Not surprisingly, the employment rates in pre-retirement age have been very low in Germany for many years. However, in the last ten to fi fteen years, we can observe a clear change in the German pension policy paradigm. Latest pension reforms increasingly expect workers to delay the transition to retirement and, in case they fail to maintain longer working lives, individuals have to accept a signifi cant pension reduction. Cross-sectional fi gures indeed indicate that employment rates in the pre-retire- ment age have substantially risen since 2000. However, as we argue, it is very risky to estimate the success of recent pension reforms on the basis of such broad cross- sectional fi gures for a variety of reasons: First, these cross-sectional estimates do not account for the fact that the increasing old age employment rates might also be caused by the changing composition of today’s cohorts of older workers (espe- cially with regard to the educational level). Second, based on such cross-sectional data, it is impossible to understand whether the whole strata of the older workforce benefi tted from increasing employment rates in later life or if only specifi c parts of the older workforce are able to maintain longer working lives while already disad- vantaged and low-qualifi ed older persons still have to retire early – yet, compared to the 1970s, 1980s and 1990s at the price of signifi cant pension reductions. And • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld902 fi nally, a simple cross-sectional approach does not allow to causally model the con- sequences of the timing of the transition to retirement on individuals’ pension in- comes. However, such a causal modelling would be necessary in order to estimate if recent reforms have been really successful or instead rather tend to be a hidden pension reduction. To appropriately address these issues, a micro-level longitudinal research ap- proach, as applied in our empirical study, is necessary. This allows us to statistically investigate (1) if there remains a signifi cant effect for the reversal of retirement if it is controlled for potential changes in the composition of the older workforce across cohorts, (2) if all parts of the older workforce benefi tted from the recent increase of old age employment rates, and (3) if the latest pension reforms aiming at delaying the transition to retirement have been a “hidden” pension reduction for some parts of the older population. Regarding these matters, the central fi ndings of our longitudinal micro-level study are as follows: (1) There is a signifi cant delay of the transition to retirement and out of employ- ment across cohorts even if we control for the potentially changing composi- tion of the older workforce. Thus, our longitudinal results indicate that there has been a signifi cant postponement of the transition out of employment and into retirement across cohorts which is independent of the higher level of education in recent cohorts, for example. However, our results also show that the delayed transition to retirement in younger cohorts is not necessarily connected to the fact that all older people also succeed in maintaining longer employment lives. Especially, the trend reversal of old age unemployment was far less successful than the trend reversal of early retirement. (2) Our empirical analyses have shown that the changing pension policy para- digm in Germany led to growing social inequalities in old age because the prolongation of working life has been socially selective. Especially the low- est qualifi ed older workers are not able to meet the new policy expectation to remain in employment longer. In the youngest cohort 1946-51, there is a signifi cant gap between lowly qualifi ed workers and those with medium edu- cational levels for the timing of the employment exit. This is not the case in the two earlier birth cohorts in which the length of working life was more or less the same for the majority of the older people. (3) Finally, our results show that the new pension policy paradigm has been a pension reduction for some parts of the older population because they are not able to maintain longer working lives. While early retirement regulations have been fi nancially very generous throughout the 1970s, 1980s and 1990s, latest pension reforms in Germany aim at punishing early retirement with signifi cant fi nancial cutbacks. Indeed, our empirical analyses indicate that it is especially those who still have to use the unemployment path to retirement, which are mostly the low-qualifi ed workers, who are nowadays confronted with signifi cant pension losses while this was not the case for the two older birth cohorts of our study. Reversing Early Retirement in Germany • 903 What do these results tell us with regard to our key research interest of evaluat- ing the effects of the changing pension policy paradigm in Germany? Obviously, the systematic expansion and the massive use of early retirement have come to a halt since 2000. Both cross-sectional data as well as our multivariate longitudinal results display a signifi cant prolongation of working life and a substantial postponement of the transition to retirement. However, it is evident that, despite this recent trend reversal, early retirement is still predominant in Germany. Many older people expe- rience the transition to retirement before both, fl exible and legal retirement age. It also has to be noted that a substantial number of older people still exit from employ- ment early due to the indirect paths to retirement. Especially the unemployment path to retirement still play an important role. This clearly suggests that attempts to reverse early retirement should not only concentrate on changing pension regula- tions but should also address the employability of older workers. Additionally, our longitudinal results show that the recent reforms have not been able to target the entire older workforce. To the contrary, especially the already disadvantaged parts of the older workforce, that is the low-qualifi ed individuals of the latest cohort, have not been reached by the latest pension reforms and still (have to) exit employment very early. However, these workers now have to pay the price for their “failure” to meet the new policy expectation to prolong working life by accepting signifi cant pension losses. In the past, on the other hand, their market “failure” was cushioned by more generous early pension regulations. Thus, recent pension reforms have caused growing inequalities among the older population. What is the reason for this limited success of Germany’s latest (and still ongoing) pension reforms? As we have outlined at the beginning of our contribution, these reforms are based on the assumption that early retirement can be explained by mi- cro-economic theory. This approach suggests that early retirement is the (sole) re- sult of rational acting and maximizing individuals who opt for an early labour market exit as long as pension systems offer incentives for early retirement (so-called pull factors). However, it has been critically stated, that the micro-economic approach fails to model the complexity of early retirement in Western societies (see, for exam- ple, Kohli 1991; Wübbeke 1999; Arnds/Bonin 2002; Blossfeld et al. 2006; Buchholz 2008; Ebbinghaus 2008). According to these authors, it is important to also take “push factors” into account to explain early retirement. These factors contribute to pushing older workers out of the labour market, for example, due to the employers’ interests to carry out rationalisation and restructuring (Wübbeke 1999; Arnds/Bonin 2002; Buchholz 2008), due to the governments’ interests to unburden the labour market (Kohli 1991; Blossfeld et al. 2006; Ebbinghaus 2008; Blossfeld et al. 2011) or due to the countries’ institutional characteristics, which hinder the employability of older workers (e.g. a lacking infrastructure for lifelong learning and low support of active labour market policies) (Blossfeld et al. 2006; Ebbinghaus 2008; Blossfeld et al. 2011). Further, our results clearly show that early retirement is not only caused by “pull factors” but also by “push factors” as we fi nd strong empirical evidence that early retirement takes place especially in fi rms, sectors and jobs, which have been confronted with the high pressure of rationalisation and restructuring. This means that, as long as Germany exclusively relies on weakening the “pull factors” of early • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld904 retirement without aiming at minimizing the “push factors” at the same time, the reforms remain unlikely to successfully enable the prolongation of the working life of all older people. These one-sided reforms merely targeting the pull factors may be useful to lower the fi nancial burdening of the public pension insurance. However, as our longitudinal results clearly demonstrate, this policy leads to growing social inequalities in old age. Thus, for the weaker parts of the older workforce, these re- forms are actually hidden pension reductions. For inducing a more successful trend reversal of early retirement, the social-democratic welfare states of Scandinavia can serve as role models. These countries prove the importance and the success of an integrated institutional framework that does not only expect people to work longer but also allows them to remain employed – namely, by offering a well-established infrastructure for lifelong learning and by putting more efforts into active labour market policies (Ebbinghaus 2005; Blossfeld et al. 2006; Blossfeld et al. 2011). References Arnds, Pascal; Bonin, Holger 2002: Frühverrentung in Deutschland: Ökonomische An- reize und institutionelle Strukturen. IZA Discussion Paper No. 666. Bonn: Institut zur Zukunft der Arbeit. Blossfeld, Hans-Peter; Rohwer, Götz 2002: Techniques of Event History Modeling. Lon- don: Lawrence Erlbaum Associates. Blossfeld, Hans-Peter; Buchholz, Sandra; Hofäcker Dirk (Eds.) 2006: Globalization, Un- certainty and Late Careers in Society, London/New York: Routledge. Blossfeld, Hans-Peter; Buchholz, Sandra; Kurz, Karin (Eds.) 2011: Aging Populations, Globalization and the Labor Market. Comparing Late Working Life and Retirement in Modern Societies. Cheltenham, UK/Northampton, MA, USA: Edward Elgar. Börsch-Supan, Axel 1992: Population aging, social-security design, and early retire- ment. In: Journal of Institutional and Theoretical Economics – Zeitschrift für die ge- samte Staatswissenschaft 148: 533-557. Börsch-Supan, Axel 1998: Incentive Effects of Social Security on Labor Force Participa- tion. Evidence in Germany and across Europe. NBER Working Paper No. 6780. Cam- bridge: National Bureau of Economic Research. Börsch-Supan, Axel 2000: Incentive effects of social security on labor force participa- tion: Evidence in Germany and across Europe. Journal of Puplic Economics 78: 25-50. Börsch-Supan, Axel 2003: Labor market effects of population aging, Review of Labour Economics and Industrial Relations 17: 5-44. Börsch-Supan, Axel; Wilke, Christina 2003: The German Public Pension System: How it Was, how it Will Be. MEA Discussion Paper 34-03. Mannheim: Mannheim Research Institute for the Economics of Aging. Börsch-Supan, Axel; Bucher-Koenen, Tabea; Reil-Held, Annette; Wilke, Christina 2008: Zum zukünftigen Stellenwert der ersten Säule im Gesamtsystem der Alterssicherung. MEA Discussion Paper 08-45, Mannheim: Mannheim Research Institute for the Eco- nomics of Aging. Buchholz, Sandra 2008: Die Flexibilisierung des Erwerbsverlaufs. Eine Analyse von Ein- stiegs- und Ausstiegsprozessen in Ost- und Westdeutschland. Wiesbaden: VS Verlag für Sozialwissenschaften. Reversing Early Retirement in Germany • 905 Deutsche Rentenversicherung 2012: Entwicklung der Beitragssätze. Berlin: DRV. Ebbinghaus, Bernhard 2000: When Labour and Capital Collude: The Varieties of Welfare Capitalism and Early Retirement in Europe, Japan and the USA. Cambridge, MA, USA: Cambridge University Press. Ebbinghaus, Bernhard 2005: Vom Stilllegen von Arbeit zum lebenslangen Lernen. Das überfällige Ende der Frühverrentung in Europa, Japan und den USA. In: MPlfG Year- book 2003-04: 37-42. Ebbinghaus, Bernhard 2008: Comparative regime analysis: Early exit from work in Eu- rope, Japan, and the USA. In: Kenworthy, Lane; Hicks, Alexander (Eds.): Methods and Substance in Macro-Comparative Analysis. Basingstoke: Palgrave Macmillan: 260- 289. Engstler, Heribert; Brussig, Martin 2006: Arbeitslosigkeit am Ende des Erwerbslebens. In: Informationsdienst Altersfragen 4,33: 2-6. Ernst, Jochen 1995: Frühverrentung in Ostdeutschland: Ergebnisse einer empirischen Erhebung zu den Bedingungen und sozialen Folgen des vorzeitigen Ruhestandes. Frankfurt am Main: Europäischer Verlag der Wissenschaften. Ernst, Jochen 1996: Alterserwerbstätigkeit und Ruhestand in Ostdeutschland. In: Ar- beit 5: 201-222. Esping-Andersen, Gøsta 1990: The Three Worlds of Welfare Capitalism. Princeton, NJ: Princeton University Press. Gatter, Jutta; Hartmann, Brigitte 1995: Betriebliche Verrentungspraktiken zwischen ar- beitsmarkt- und rentenpolitischen Interessen. In: Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 28: 412-424. Gruber, Jonathan; Wise, David 1999: Social Security and Retirement around the World. Chicago: University Press. Gruber, Jonathan; Wise, David 2004: Social Security Programs and Retirement around the World. Chicago: University Press. Gruber, Jonthan; Wise, David 2005: Social Security Programs and Retirement Around the World: Fiscal Implications. NBER Working Paper, 11290, Cambridge: NBER. Guillemard, Anne-Marie 1991: Die Destandardisierung des Lebenslaufs in den europäis- chen Wohlfahrtsstaaten. In: Zeitschrift für Sozialreform 37,2: 620-639. Hofäcker, Dirk; Pollnerová, Stepanka 2006: Late Careers and Career Exits: An Inter- national Comparison of Trends and Institutional Background Patterns. In: Blossfeld, Hans-Peter; Buchholz, Sandra; Hofäcker, Dirk (Eds.): Globalization, Uncertainty and Late Careers in Society. London/New York: Routledge: 25-53. Jabsen, Annika; Buchholz, Sandra 2009: Increasing uncertainty in old age in Germany? The development of social inequality in later life since the mid-1980s. fl exCAREER Working Paper. Bamberg University. Kaplan, Edward L.; Meier, Paul 1958: Nonparametric estimation from incomplete obser- vations. In: Journal of the American Statistical Association 53: 457-481. Kohli, Martin; Rein, Martin; Guillemard, Anne-Marie; van Gunsteren, Herman 1991: Time for Retirement: Comparative Studies of Early Exit from the Labor Force. Cam- bridge/New York: Cambridge University Press. Mayer, Karl Ulrich; Huinink, Johannes 1990: Alters-, Perioden- und Kohorteneffekte in der Analyse von Lebensverläufen oder: Lexis ade? In: Mayer, Karl Ulrich (Eds.): Le- bensverläufe und sozialer Wandel. Wiesbaden: Westdeutscher Verlag: 442-459. • Sandra Buchholz, Annika Rinklake, Hans-Peter Blossfeld906 OECD 2001: Ageing and Income: Financial Resources and Retirement in 9 OECD Coun- tries. Paris: Organization for Economic Co-operation and Development (OECD). Rinklake, Annika; Buchholz, Sandra 2011: Increasing inequalities in Germany: Older people’s employment lives and income conditions since the mid-1980s. In: Blossfeld, Hans-Peter; Buchholz, Sandra; Kurz, Karin (Eds.): Aging Populations, Globalization and the Labor Market. Comparing Late Working Life and Retirement in Modern Socie- ties. Cheltenham, UK/Northampton, MA, USA: Edward Elgar: 35-64. Riphahn, Regina; Schmidt, Peter 1997: Determinanten des Rentenzugangs: Eine Ana- lyse altersspezifi scher Verrentungsraten. In: Jahrbuch für Wirtschaftswissenschaften 48. 113-147. Schmid, Josef 2002: Wohlfahrtsstaaten im internationalen Vergleich: Soziale Sicherung in Europa: Organisation, Finanzierung, Leistungen und Probleme. Opladen: Leske + Budrich. Siddiqui, Sikandar 1997: Early Retirement in West Germany: A Sequential Model of Dis- crete Choice. In: Zeitschrift für Wirtschafts- und Sozialwissenschaften 117: 391-415. Wübbeke, Christina 1999: Der Übergang von sozialversicherungspfl ichtiger Beschäf- tigung in den Rentenbezug zwischen sozialpolitischer Steuerung und betrieblichen Interessen. In: Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 32: 103-117. Yamaguchi, Kazuo 1991: Event History Analysis, Newbury Park/London/New Delhi: Sage. A German translation of this reviewed and authors’ authorised original article by the Feder- al Institute for Population Research is available under the title “Umkehr von Frühverrentung in Deutschland – Eine Längsschnittanalyse der Auswirkungen der jüngsten Rentenreformen auf den Zeitpunkt des Erwerbsausstiegs und die Rentenhöhe”, DOI 10.12765/CPoS-2013-23de or URN urn:nbn:de:bib-cpos-2013-23de5, at http://www.comparativepopulationstudies.de. Date of submission: 13.03.2012 Date of Acceptance: 19.11.2012 Prof. Dr. Sandra Buchholz (), Prof. Dr. Hans-Peter Blossfeld. University of Bamberg, Chair of Sociology 1. Bamberg, Germany. E-Mail: sandra.buchholz@uni-bamberg.de, hans-peter.blossfeld@uni-bamberg.de URL: http://www.uni-bamberg.de/en/soz1/professor-dr-sandra-buchholz/ URL: http://www.uni-bamberg.de/en/bamberg-graduate-schools/english/bagss/people/ faculty/hans-peter-blossfeld/ Annika Rinklake. Staatsinstitut für Familienforschung an der Universität Bamberg (ifb). Bamberg, Germany. E-Mail: annika.rinklake@ifb.uni-bamberg.de URL: http://www.ifb.bayern.de/mitarbeiter/rinklake.html © Federal Institute for Population Research 2013 – All rights reserved Published by / Herausgegeben von Prof. Dr. Norbert F. Schneider Federal Institute for Population Research D-65180 Wiesbaden / Germany Managing Editor / Verantwortlicher Redakteur Frank Swiaczny Assistant Managing Editor / Stellvertretende Redakteurin Katrin Schiefer Language & Copy Editor (English) / Lektorat & Übersetzungen (englisch) Amelie Franke Copy Editor (German) / Lektorat (deutsch) Dr. Evelyn Grünheid Layout / Satz Beatriz Feiler-Fuchs E-mail: cpos@bib.bund.de Scientifi c Advisory Board / Wissenschaftlicher Beirat Jürgen Dorbritz (Wiesbaden) Paul Gans (Mannheim) Johannes Huinink (Bremen) Marc Luy (Wien) Clara H. Mulder (Groningen) Notburga Ott (Bochum) Peter Preisendörfer (Mainz) Board of Reviewers / Gutachterbeirat Martin Abraham (Erlangen) Laura Bernardi (Lausanne) Hansjörg Bucher (Bonn) Claudia Diehl (Konstanz) Andreas Diekmann (Zürich) Gabriele Doblhammer-Reiter (Rostock) Henriette Engelhardt-Wölfl er (Bamberg) E.-Jürgen Flöthmann (Bielefeld) Alexia Fürnkranz-Prskawetz (Wien) Beat Fux (Zürich) Joshua Goldstein (Rostock) Karsten Hank (Köln) Sonja Haug (Regensburg) Franz-Josef Kemper (Berlin) † Michaela Kreyenfeld (Rostock) Aart C. Liefbroer (Den Haag) Kurt Lüscher (Konstanz) Dimiter Philipov (Wien) Tomáš Sobotka (Wien) Heike Trappe (Rostock) Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet)