Employment Histories and Pension Incomes in Europe: A multilevel analysis of the role of institutional factors

In research on pensions and retirement income, it has been frequently reasoned that the economic situation in later life is determined by an interplay of individual and institutional factors. However, previous studies in this field either focus only on individual determinants or on macro-level outcomes using aggregated data. We apply a multilevel approach to examine the impact of institutional factors on the link of individual pension income and previous employment history. The underlying research question is of how national pension systems shape this relationship; whether flexible careers and atypical employment are compensated for or, on the contrary, ‘penalised’ with a low pension income. We combine the life-history data of individuals in 13 European countries from the Survey of Health, Ageing and Retirement in Europe (SHARELIFE) with macro-data on national pension systems. While we find little cross-national variation for men, for women the strength of the relationship of employment history and pension income differs between countries and is significantly moderated by factors related to the pension system.

The income situation of the elderly population in modern welfare states is the result of an interplay of individual developments over the life course and the institutions of modern welfare states (Crystal and Shea 1990; Dewilde 2012; Hinrichs and Jessoula 2012; Sefton et al. 2011). Currently, pension systems as well as individuals' working lives undergo substantial changes which influence the economic well-being of older individuals. First, many European pension systems are in a reform process. This implies retrenchment of public benefits as well as an increasing importance of private provision (Fernández 2012; Frericks et al. 2008; Myles and Pierson 2001). Second, the ongoing destandardisation of employment careers implies a shift away from the standard of continuous full-time employment towards flexible and discontinuous working histories and atypical employment. Especially for women and for lower educated men, interrupted working biographies and atypical employment have become more prevalent over the last decades (Buchholz et al. 2009; Simonson et al. 2011; Widmer and Ritschard 2009). These developments create concerns about the future economic situation of retirees in Europe. Pension systems play a key role in preventing that working-life inequalities persist in old age because they moderate the relationship of individuals' working life and their status in retirement.

A large body of research explores either the individual determinants of the income position in later life (Bardasi et al. 2002; Dewilde 2012; Fasang et al. 2013; Sefton et al. 2011; Zaidi et al. 2005) or the economic well-being of the elderly population in international comparison using aggregate data (Disney and Withehouse 2002; Ebbinghaus and Neugschwender 2011; Gornick et al. 2009). We combine both approaches with a multilevel analysis of the effect of pension system characteristics on the relationship of individual work history and retirement. Two questions are addressed: (1) How are the current income positions of retirees in Europe related to their previous employment biographies? and (2) How do national pension systems shape the relation of employment biography and income in later life: are flexible non-standard careers compensated for or, on the contrary, ‘penalised’ with a low pension income?

The study is based on a sample of 13 European countries which are included in the third wave of the Survey of Health, Ageing and Retirement in Europe (SHARELIFE).¹ This data contains unique retrospective information on the life histories of older individuals in Europe. The impact of pension systems is evaluated by means of specific institutional characteristics, as the degree of redistribution and the public–private mix of retirement income. Furthermore, gender differences in the effect of these institutional characteristics are examined.

The article proceeds as follows: Section 2 describes the individual and institutional impact factors of the income situation in later life and derives hypotheses. Method, data and operationalization are presented in Section 3. Section 4 includes, first, a descriptive overview of the individual employment histories in country comparison. Second, the results of the multilevel regression analysis are discussed. In Section 5, we review the results and conclude.

In contemporary welfare states, most state interventions aim towards social security rather than towards social equality. Thus, the core of governmental redistribution is not the reallocation of resources among individuals but the temporal shifting of resources between different life stages (Whiteford 2008). Pension systems are a prime example of this kind of redistribution as pension benefits are mostly to some degree dependent on an individual's previous employment history. However, the strength of this link varies between countries and ranges from a reproduction of labour market inequalities in social insurance systems to a cushioning of employment biography outcomes in basic pension systems (Leitner 2001). Consequently, pension systems moderate the relation of individuals' working career and their well-being in later life on the basis of normative assumptions about working lives and gender roles (Leisering 2003; Leitner 2001). Previous labour market or gender-related inequalities may be either reinforced or compensated through regulations inherent in the national pension systems. Thus, the individual income position in old age results from the interplay of the country-specific institutional framework with the individual employment patterns over the life course (Arza 2008).

Achievements on the labour market during mid-life are the basis for the economic status in retirement. Pension entitlements in statutory pension systems typically refer to the individual employment history. Also the access to occupational pension plans is dependent on the employment situation, as the job position or the type of employer. Furthermore, the possibilities for private pension saving are determined by the level of income which for the majority of individuals stems from labour market earnings. Continuous labour market participation is beneficial in terms of pension provision, whereas inconsistent employment careers or atypical employment have adverse effects for the retirement income. On this background, two dimensions of employment histories are relevant for the accumulation of pension rights and financial resources for retirement: (1) the form of employment, i.e. regular or atypical employment, and (2) the volatility of the employment history, i.e. the frequency of employment interruptions or job changes.

Pension systems are typically oriented towards ‘the norm of continuous full-time employment’ (Leitner 2001: 103), and thereby discriminate against family care and non-employment. Some national pension systems provide pension rights for times of inactivity, especially for childcare. However, these benefits are mostly not equal to those related to labour market earnings (Leitner 2001; Lewis 1992):²

Atypical forms of employment which deviate from the norm of regular full-time work provide fewer possibilities for the accumulation of pension rights. Part-time employment, which is often performed by women to combine care work and employment, is associated with lower earnings, long-term negative wage effects, and consequently, lower pension entitlements (Fouarge and Muffels 2009). The same applies to the heterogeneous group of self-employed and farmers. Their income position is on average lower and their poverty risk is higher than of dependent employees. Furthermore, many national pension systems do not regard self-employment; if they do, membership is mainly voluntary and associated with higher costs or the possibility to opt for lower protection (European Commission 2010):Career volatility, i.e. the number of changes between different forms of labour market participation and inactivity throughout the working life, also affects the possibilities to accumulate pension rights and savings. On the one hand, the more fragmented into different episodes of labour market participation and inactivity the employment history is, the lower are the possibilities for continuous retirement saving. However, volatility also represents repeated labour market participation, for example, the re-entry into employment after a period of childcare. Thus, concerning the gender differences in career patterns, higher volatility in female careers is a sign of a higher labour market attachment. On the contrary, a higher volatility in men's careers rather indicates a marginalised labour market position. Therefore, career volatility is assumed to have a negative effect on the pension income of men but a positive effect on the pension income of women.

In previous research on pension systems and the economic situation of older people, some specific features of national pension systems are pointed out as central measures for cross-national comparisons: (1) the degree of redistribution, (2) the public–private mix of pension systems; and (3) the relative level of pension benefits compared to labour market earnings. Furthermore, gender differences regarding the retirement income are stressed (Behrendt 2000; Leitner 2001; OECD 2009; Queisser et al. 2007). In the following, we describe the possible impact of these institutional characteristics for persons with ‘non-standard’ employment histories, i.e. with a low level of labour market attachment throughout their working life.

The redistributive character of pension systems is deduced from the relevancy and the design of redistributive elements in the pension system. Therefore, we distinguish whether redistributive schemes within the first pension tier are designed as basic, targeted or minimum pensions (Queisser et al. 2007: 544). The highest relevancy of redistribution is achieved if the pension system provides basic benefits which are flat rate and paid unconditional of the previous earnings or contribution history. Usually, only a minimum number of years of citizenship is necessary to be eligible for basic pension benefits. Targeted benefits additionally involve an income test and are either paid only to pensioners whose income is below a certain threshold or the benefit level decreases with increasing incomes from other sources. Whereas basic and targeted pension schemes are clearly redistributive, minimum pension schemes are a hybrid between redistributive and earnings-related pension benefits because they require a minimum number of contribution years. Finally, in some countries, the pension system does not include a stand-alone redistributive pension tier.

Basic and targeted pension schemes are expected to mitigate the negative effects of non-standard careers, whereas earnings-related or actuarial systems with only marginal or even no redistributive elements will reward continuous regular employment while being poorly equipped to compensate for non-standard employment histories.

National pension systems differ in the relative relevance of the pillars of pension provision, which refer to the state (first pillar), the employer (second pillar) or the financial market (third pillar) as an institutionalised provider of benefits, and therefore, to the extent to which private and occupational pensions are regular sources of pension income (Ebbinghaus 2011). Whereas occupational pension plans are allocated by employers and often result from collective agreements, private pension plans typically rather follow a market logic.³ In the countries of the SHARELIFE sample, occupational pension systems play a larger role than purely individualised private pensions. High shares of private/occupational pension income can be found in The Netherlands, Denmark and Switzerland (see Table A2). In all three countries, occupational pension schemes have a high coverage. The Danish-funded occupational pension scheme is based on an agreement between the social partners and has a nearly universal coverage among employees, and thus, can be described as quasi-mandatory. The same applies to the Dutch occupational pension scheme which is also covering more than 90% of employees, although employers are not obliged to provide a pension plan in their company. In Switzerland, a mandatory occupational pension scheme was introduced in 1985 (OECD 2009).

Despite organisational differences between occupational and private pension schemes, individuals with unstable working biographies or in non-regular employment face higher barriers to participate in both. Usually, the provision of occupational pension plans is dependent on the employer's willingness or the existence of collective agreements. Employees with permanent contracts in large companies, thus, are more likely to be covered than those in precarious work. Furthermore, unlike in public pension schemes, periods of unemployment and childcare are typically not covered in occupational pension plans. Private annuity or personal pension plans stemming from investments at the financial market – although they are not bound to a certain employer – need to be steadily filled with deposit payments and therefore are dependent on the existence of a regular and sufficient income. Private personal pensions as well as occupational pensions ‘are very much linked to employment status and earnings during working life’ (Ebbinghaus and Neugschwender 2011: 395–396). Consequently, in countries, where private or occupational pension incomes play an important role, labour market inequalities are expected to be intensified in later life.

A key target of contemporary pension systems is the maintenance of a living standard during retirement which is similar to working-life. Accordingly, pension benefits should be an adequate replacement of previous earnings. The indicator usually utilised in this context is the replacement rate which relates the level of pension benefits to average earnings (OECD 2009). However, the replacement rate is an ambiguous measure. On the one hand, the higher the replacement rate of pension benefits, the more the state plays an active role in the pension provision and individuals do not need to provide privately to achieve a sufficient income in old age. On the other hand, the higher the replacement rate, the stronger the pension system reproduces the previous status during working life in old age. Considering this ambiguity, we assume that the average replacement rate will have no significant effect on the relationship of working history and pension income.

As a result of the gendered division of market and care work, women generally face a higher probability of having interrupted and/or non-standard employment biographies, and thus, have a lower capacity for an autonomous accumulation of resources during mid-life. Previous studies point to existing gender differences in the income positions in later life (Dewilde 2012; Sefton et al. 2011). Therefore, the institutional compensation of inconsistent biographies generally has a higher importance for women than for men. Women are more likely to be negatively affected by a lack of institutions which provide a sufficient basic income in old age as well as by a lack of redistributive elements in pension systems:

The Survey of Health, Ageing and Retirement in Europe (SHARE) is used as a data source for this study (Börsch-Supan et al. 2011; Schröder 2011). The third-wave SHARELIFE conducted in 2008/2009 includes unique retrospective biographical information of older individuals in the 13 European countries Austria, Belgium, Czech Republic, Denmark, France, Germany, Greece, Italy, The Netherlands, Poland, Spain, Sweden and Switzerland. As Eastern German retirees have spent the majority of their working life in the socialist welfare state of the former German Democratic Republic (GDR), we separate Eastern and Western Germany in our analysis. The analysis sample contains all individuals aged 65 years and older who are not in active employment. Furthermore, respondents with incomplete biographical data are excluded.⁴ The resulting analysis group consists of 9438 observations (4525 women and 4913 men) nested in 14 macro contexts. To account for gender differences, all estimations are calculated separately for men and women.

The dependent variable of the analysis is the annual individual income which is the sum of all public and market income sources, including regular income from private pension plans and insurances. Since the income level in this analysis is regarded as the outcome of individual achievements on the labour market, widow pensions, which solely pertain to the partner's working history, are excluded. As the analysis group contains only retirees, the income mainly consists of public and private pension benefits, so it is denoted as pension income. All income values are adjusted with the purchasing power parity (PPP) and logarithmized.⁵ To account for missing values, the multiple imputed income values provided by the SHARE group are used and analysed with the mi-package of STATA 12.0 (Christelis 2011; StataCorp 2011).

To describe the individual employment histories, the information on the respondents’ biographies is summarised into four indicators. (1) The total number of years in employment, (2) thereof the share of part-time and temporary employment, (3) the share of self-employment and (4) an indicator on career volatility which is operationalized as the number of changes between the different employment statuses regular full-time employment, part-time/temporary employment, self-employment and inactivity, which also comprises family care. The following individual characteristics are included as control variables: birth cohort in groups, educational level according to the International Standard Classification of Education (ISCED: level 0–2 indicates low, level 3–4 medium and level 5–6 a high educational level), family status, the number of children and a dichotomous variable indicating whether someone has a foreign citizenship (Table A1 includes descriptive summary statistics).

The following macro-level indicators are included in the analysis: the design of the first pension tier, operationalized according to Queisser et al. (2007), is used as an indicator of the redistributive character of the national pension system. The indicator is comprised of four categories indicating a descending relevance of the first tier: basic, targeted, minimum schemes and systems without a stand-alone redistributive pension tier. The latter group is used as reference category in the regression models.

The public–private mix of pensions is operationalized with the country mean share of private/occupational pension income on the basis of the SHARE data.⁶ Higher values of this indicator represent a higher reliance on private sources of pension provision in a country. The replacement rate of pension benefits as provided by the OECD (2009) is a percentage value that shows the generosity of the public pension benefits by comparing them to the income of an artificial ‘median earner’ with average earnings over a full, uninterrupted working career (Table A2 includes the values of the macro indicators).

Figure 1 gives a descriptive overview of men's and women's employment histories in the included countries. Whereas the employment histories of men are quite homogenous, the cross-national variation of women's biographies is larger. Regular employment is dominating the employment histories of men in all countries, apart from Greece, Italy, Spain and Poland where self-employment also plays an important role. The volatility of men's employment histories is generally lower than of women's, indicating that men have more continuous employment biographies, whereas women change more often between employment and inactivity or between full-time and part-time employment. However, women do not generally have a low labour market attachment. Only in the Southern European countries inactivity is predominant in women's careers. On the contrary, a large share of Eastern German and Czech women is continuously full-time employed. Among women in the Scandinavian countries, The Netherlands, Western Germany and Switzerland we find high shares of part-time employment.

The results of the regression models in Table 1 describe the individual-level effects on the pension income of women and men. Model 1a contains only the socio-demographic control variables, Model 1b additionally the employment history indicators. The cross-level interaction effects are included in Models 2–5 (Table 2).⁷ Country-level heterogeneity is controlled for by means of country-fixed effects in all models.

The adjusted R² values indicate the share of explained variance in the different regression models, and thus, allow assessing the relative importance of the included variables. While for women the addition of employment history characteristics increases the R² by 46% from 0.26 in Model 1a to 0.38 in Model 1b, for men this increase amounts only to 9% (0.09–0.10). Thus, differences in the individual employment histories play a larger role for the pension income position of women, whereas these differences and their effects are far less pronounced for men.

The employment history indicators show the expected effects for women (see Table 1). Each additional year in employment increases the pension income by 6.7%. On the contrary, the share of part-time employment has a negative effect on the individual pension income. Each percentage point part-time employment reduces the later income by 0.3%. The same applies to self-employment which leads to an income reduction of 0.5%. These results support Hypotheses 1 and 2 regarding the generally positive impact of labour market participation, which is, however, diminished by periods of atypical employment. Volatility in women's employment histories has a positive impact on the individual income in old age: every change between employment statuses is related to a 10.4% increase in the pension income. This result supports Hypothesis 3: for women, volatility is a sign of a high labour market attachment, indicating, for example, resumed labour market activity after a period of childcare. Women who re-enter employment have more possibilities to generate an independent provision for old age compared to women continuously in family care and dependent on their husbands’ incomes. We also find highly significant positive effects of educational level on women's pension income. Especially, highly educated women archive a high income level in retirement.

Differences related to familial factors play an important role for the individual income position of older women. Single, divorced and never-married women achieve significantly higher individual pension incomes than women who are married or widowed. Having children has a significant negative effect on the retirement income of women (Model 1a). However, this factor becomes insignificant after controlling for characteristics of the employment history in Model 1b. Thus, the lower pension income of mothers results from their reduced labour market participation.

The larger the share of years in the labour force, the better the income position of men in later life. For each percentage point increase in the share of years employed, the logarithmized income rises by 0.1% (Model 1b). However, this effect is not significant. The same applies to the negative effect of career volatility. On the contrary, periods of part-time and temporary work as well of as self-employment have a significant negative effect on the pension income of men. The negative effect of part-time/temporary work is even more pronounced than for women: every percentage point reduces the pension income by 1.3%. These results support Hypothesis 2 and show that any deviation from the career pattern of continuous full-time employment has a substantive negative effect on men's income position in later life. However, the positive effect of years in employment and the negative effect of volatility are insignificant. This might, first, stem from the high homogeneity of men's employment biographies, and thus, does not generally confute Hypotheses 1 and 3. Second, the results show that continuous full-time employment is the predominant norm among men in Europe. Hence, the accumulation of years in employment alone is not a decisive advantage for the later income position.

Compared to the effects of the educational level, the employment history of men has a rather low impact on their pension income. Those with a medium educational level have a 27.7 percentage points higher logarithmized pension income compared to men with a low educational level. On the contrary, ten additional years in the labour force lead to an average increase in the logarithmized income of only one percentage point. Consequently, the income position in old age is driven by the strong link of educational level and occupational status, which finally also determines the retirement income.

The inclusion of cross-level interaction effects leads to an increase of the share of explained variance in the models for women, while for men, factors related to national pension systems do not contribute significantly to the explanation of country differences in old age income. The cross-level interaction effects are all on a very low level and not significant. Thus, for men, no clear evidence can be found to support the general assumption that the pension system moderates the relationship of individual employment history and income in later life (results not depicted). The low relevance of the moderator function of pension systems might again result from the generally low cross-national and inter-individual variation in male careers.

On the contrary, women with their generally more volatile careers and higher shares of non-regular employment are significantly affected by the design of the national pension system (see Table 2). Only the replacement rate of the public pension benefits is of minor importance for the individual income position of older women (results not depicted). This result reflects the fact that the artificial working career of the ‘median earner’, which underlies the calculation of the replacement rate, is, as assumed, not representative for the employment histories of women.

Generally, pension systems that incorporate basic pensions significantly reduce the impact of the individual employment history on the later pension income of women (Model 2). The same applies to targeted pension schemes. On the contrary, particularly in minimum pension schemes, the link of the employment biography and pension income is strengthened, whereas this applies only to a lower degree to pension systems without a redistributive pension tier. Here, general social security benefits seem to fill the gap. Figure 2 shows the relationship of the years in employment and the estimated retirement income of women for the different types of redistributive pension schemes in the first tier. Hypothesis 4 regarding the compensating effect of basic pensions is supported. However, systems without a redistributive tier perform better than conditional minimum pension schemes.

The share of private/occupational pension income in a country has a significant negative, hence mitigating, effect on the relationship of individual employment history and later life income (see Model 3). However, this on the first glance counterintuitive result has to be viewed on the background of the design of the whole pension system. Therefore, the interaction of the public-private mix and the design of the first pension tier were tested using three-way interactions (Models 4 and 5). The results show a clear pattern: in countries with high shares of private/occupational pension income and well-developed redistributive elements (basic/targeted pension schemes), the relationship of working career and pension income is indeed mitigated (Model 4). However, private pensions have an adverse effect if they are not accompanied by adequate redistributive pension schemes (Model 5). Figure 3 shows these contrary effects for different levels of the private pension income share in a country (values of the quartiles in the country sample: 1.5%, 3.4% and 8.9%). Consequently, Hypothesis 5 is only partly supported as the effect of private pensions on the relationship of employment history and pension income is ambiguous and dependent on the existence of redistributive pension elements.

To sum up, continuous employment is, as expected, positively related to the pension income of men and women. Forms of employment which deviate from regular full-time employment, as part-time, temporary work and self-employment, have a significant negative effect on the later pension income. For women, volatility is positively related to retirement income, which indicates that it is a sign of a higher labour market attachment over the life course. Institutional factors and their interactions with characteristics of the individual employment history are more important for women. Basic pension schemes compensate for non-standard employment histories, while especially in conditional minimum pension systems, previous labour market related inequalities are reinforced. The role of private/occupational pensions depends on the interplay with other parts of the pension system. Non-public pension provision intensifies working life inequalities unless it is accompanied by sufficient basic pension schemes.

On the background of the ongoing public pension retrenchment and the destandardisation of employment histories, the results of this study provide important insights. This applies especially to the results for female retirees as they generally have more destandardised biographies than their male counterparts. Therefore, we can draw conclusions on the future development of the income situation of retirees considering the increasing prevalence of fragmented careers and atypical employment. It has become clear that non-standard employment histories do not necessarily go along with low pension income or old age poverty because the design of the pension system plays a key role in compensating for inequalities of the working life. Especially, redistributive elements in the public pension system are crucial. Basic and targeted pension schemes, which unconditionally guarantee a sufficient pension income, indeed mitigate the possible adverse effects of destandardised work histories on the later life income. On the contrary, minimum income schemes, which impose additional conditions as a minimum number of contribution years, rather intensify inequalities of the working life.

The design of redistributive pension elements is also decisive for the effect of private/occupational pensions. The latter do not generally intensify working life inequalities in later life, however, only if they are supplemented by a universal public pension scheme and have a high coverage. Multi-pillar pension systems that combine basic pensions with easily accessible occupational pensions seem to be best equipped to encounter the negative consequences of the de-standardisation of employment histories. Two conditions are related to this. First, the level of the basic pension needs to be adequate to guarantee a sufficient income. Second, the private or occupational pension schemes also need to be highly prevalent among the whole working population, for example, also among part-time working women. This applies, for example, to The Netherlands where occupational pensions are well integrated into the pension system and their coverage is nearly universal (OECD 2009).

The results of this analysis illustrate that the effect of policies and institutions only unfolds in combination with the prevalent patterns of individual behaviour. This is particularly evident in the observed gender differences. While for men the moderating function of the pension systems is rather irrelevant, for women with their more diverse and flexible employment careers the same institutional characteristics significantly moderate the relationship of employment history and pension income. However, while women generally profit from restarted labour market activity after a period of childcare, the negative effect of part-time employment on the later pension income persists despite cross-national differences in the institutional compensation of care-related part-time employment.

Furthermore, the results show that specifically designed national pension systems compensate for unstable or non-standard biographies of women. Although the general inequalities between men and women on the labour market are not completely balanced by welfare state institutions, the design of national pension systems can, and does in some countries, make a significant contribution to compensating these inequalities with regard to retirement income. However, inequalities between men's and women's retirement income remain, which are rather related to the gendered division of market and care work then to pension systems. For example, the negative effect of raising children and the related employment patterns of mothers on the pension income remains even after controlling for country differences. On the background of the also negative impact of part-time employment on pension income, this result gives reason for concern. Since part-time jobs are chosen by women to combine work and care responsibilities, working mothers face a ‘pension penalty’ irrespective of the specific national institutional context. This result fits to findings on the gender bias of pension systems and the general incapacity of the state in sufficiently supporting care responsibilities and modern family life (Daly 2005; Frericks et al. 2008). Future research, therefore, should shed more light on the interrelation of employment and family history and its relevance for later life outcomes on the background of country-specific gender differences in employment careers (Hank and Korbmacher 2012; Fasang et al. 2013).

A limitation of the analysis is the rather small number of countries included in the SHARELIFE data. For future research, an expansion of the retrospective survey of life histories to other countries with divergent pension systems would be advantageous.

I am grateful to Wim van Oorschot, Manfred te Grotenhuis and two anonymous reviewers for helpful comments on earlier versions of this paper. All remaining errors are mine.

Katja Möhring is Research Fellow at the Centre for Social Policy Research at the University of Bremen. She has worked at the Department of Social Policy at the University of Cologne and the German Institute for Economic Research (DIW) in Berlin. Her research interests are welfare state and social policy analyses, life course sociology, statistical methods and social inequality.