Post-retirement work

In recent years, the employment rates of people of pension age have increased considerably in many Western countries. The growing body of literature dealing with this ‘bridge employment’ finds a huge variety of reasons for and forms of post-retirement work. The majority of those working post-pension age in Germany and the United Kingdom (UK) do so part-time and receive a pension (Scherger et al. Reference Scherger, Hagemann, Hokema and Lux2012). Some work in their former job and/or for their former employer, some in a different job, which occasionally seems to involve downward mobility (Lain Reference Lain2012). More specifically and regarding the group we study in this paper, few male Britons return to work after having completely retired (Kanabar Reference Kanabar2012).Footnote ² Although no systematic information about this is available for Germany, the corresponding percentages are probably also low. For both countries, it is also well known that older workers are much less likely to be newly recruited than to be retained (Brussig Reference Brussig2011; Daniel and Heywood Reference Daniel and Heywood2007). On average, German and British post-retirement workers tend to be better educated and healthier than non-workers of their age, and men and the self-employed are clearly over-represented (Brenke Reference Brenke2013; Lain Reference Lain2011, Reference Lain2012; Scherger Reference Scherger2013). The motives for post-retirement work are as varied as its forms and include enjoyment and fulfilment through work, the social contacts and social recognition it brings, the wish to stay fit and, of course, financial motives (Hokema and Scherger Reference Hokema and Scherger2015; Scherger et al. Reference Scherger, Hagemann, Hokema and Lux2012). The latter not only cover the financial necessity of working, but, probably more importantly, also the wish to earn extra in order to maintain a certain lifestyle. This range of motives is confirmed by qualitative research which also underlines the importance of work identities and the work role, at least for a considerable sub-group of working retirees (Barnes, Parry and Taylor Reference Barnes, Parry and Taylor2004). Thus, many of the motives imply the desire not to stop working from one day to another, to keep a sense of continuity and to retire gradually. Due to a lack of sufficient data, there is no representative evidence (for the UK and Germany) on the exact pathways into post-retirement work and on the numbers of people who stop working and then take up a job again. In our empirical analyses, we focus on the group who take up a job again after a time of not working for both theoretical and technical reasons; by focusing on this group we are able to limit the complexity of working post-retirement and to make more specific and targeted assumptions.

Evidence on life satisfaction, retirement and work

Research on life satisfactionFootnote ³ and other dimensions of wellbeing has been proliferating in the last decade, increasingly using large data sets and comparing countries. A number of unequally distributed characteristics and resources are known to affect individual wellbeing. Good health, having a family and other social relationships, favourable financial and material circumstances (at least up to a certain level), and activity are among the most important factors which are likely to increase individual life satisfaction in most societies (Brockmann and Delhey Reference Brockmann and Delhey2010; Margolis and Myrskylä Reference Margolis and Myrskylä2013). Correspondingly, the life domain approach has shown that overall life satisfaction is closely related to satisfaction in different domains of life (Diener, Oishi and Lucas Reference Diener, Oishi, Lucas, Snyder and Lopez2009: 189; see also Easterlin and Sawangfa Reference Easterlin and Sawangfa2007; Margolis and Myrskylä Reference Margolis and Myrskylä2013), in particular one's economic situation, health, family relationships and work (Margolis and Myrskylä Reference Margolis and Myrskylä2013: 116; van Praag, Frijters and Ferrer-i-Carbonell Reference van Praag, Frijters and Ferrer-i-Carbonell2002).

In this paper, we focus on the effect of paid work on life satisfaction at a particular stage of life: the time after pension age. Although activity generally tends to have positive effects on wellbeing, it would be over-generalising to assume that every activity under whatever circumstances and disregarding the individual life course stage enhances individual life satisfaction. Whereas unemployment (i.e. not having a paid job involuntarily) clearly has negative effects on life satisfaction during normal working age (see e.g. Ahn, García and Jimeno Reference Ahn, García and Jimeno2004; Lucas et al. Reference Lucas, Clark, Georgellis and Diener2004), the evidence is much less clear once people have reached pension age. In all Western countries, retirement is an institutionalised transition based on which people aged roughly between 58 and 67 (depending on national regulations) withdraw from the labour market. Pension systems, whose design varies considerably, provide the material basis for this withdrawal which structures and defines later life (see e.g. Kohli Reference Kohli1987; Kohli and Arza Reference Kohli, Arza, Binstock and George2011). Under this institutional arrangement, not working ‘because of’ old age is not at all comparable to the experience of being unemployed. This is also because the ‘moral economy’ of retirement (Kohli Reference Kohli1987; Scherger and Hagemann Reference Scherger and Hagemann2014) provides valid legitimisations for not working in old age: retirement can be seen as a deserved reward for a long working life backed by contributions to pension schemes and it ensures a regular and continuous generational succession in companies and elsewhere.

Unsurprisingly, retirement is often depicted as a time of leisure to be looked forward to, especially nowadays when many retirees can expect many more healthy years after they retire. Correspondingly, research on the actual experience of retirement paints a predominantly positive picture, although the results also point to the heterogeneity of this experience (Pinquart and Schindler Reference Pinquart and Schindler2007; van Solinge Reference van Solinge and Wang2013: 315–6; Wang Reference Wang2007). Continuity or (often only temporary) increases in life satisfaction seem to be the most common consequences of the transition to retirement (e.g. Emmerson and Muriel Reference Emmerson, Muriel, Banks, Breeze, Lessof and Nazroo2008; Pinquart and Schindler Reference Pinquart and Schindler2007). Specific unfavourable conditions of retirement go together with significant decreases in wellbeing – this applies to poor health and early retirement (Börsch-Supan and Jürges Reference Börsch-Supan and Jürges2006), involuntary retirement and lack of control over the decision (Dingemans and Henkens Reference Dingemans and Henkens2014; Shultz, Morton and Weckerle Reference Shultz, Morton and Weckerle1998; van Solinge and Henkens Reference van Solinge and Henkens2008), poverty (Matthews and Nazroo 2015) and a strong work identity (for an overview of these intervening conditions, see also van Solinge Reference van Solinge and Wang2013; Wang, Henkens and van Solinge Reference Wang, Henkens and van Solinge2011). However, many of these negative effects are short-lived and it is sometimes hard to disentangle what is the actual cause of a decrease in wellbeing, such as in the case of health shocks and early retirement. By contrast, if the former job was experienced as stressful and job satisfaction was low, or someone was unemployed before retirement, retiring is associated with gains in wellbeing (Wang, Henkens and van Solinge Reference Wang, Henkens and van Solinge2011).

Just because retirement is, in the long run, a neutral or even positive experience for the majority of retirees, this does not imply that re-entering (or continuing) work is necessarily connected to decreasing life satisfaction. The opposite tends to be true in the case of unpaid, voluntary activities: people volunteering after pension age are more likely to report a higher quality of life and life satisfaction than those who do not (Matthews and Nazroo Reference Matthews, Nazroo and Scherger2015; McMunn et al. Reference McMunn, Nazroo, Wahrendorf, Breeze and Zaninotto2009; Wahrendorf and Siegrist Reference Wahrendorf and Siegrist2010). Again, there is no positive effect under certain unfavourable conditions, e.g. if the efforts of volunteering or providing (unpaid) care for someone are not rewarded adequately on a subjective level (McMunn et al. Reference McMunn, Nazroo, Wahrendorf, Breeze and Zaninotto2009; see also Wahrendorf, von dem Knesebeck and Siegrist Reference Wahrendorf, von dem Knesebeck and Siegrist2006).

This differentiation also applies to paid work in retirement. According to McMunn et al. (Reference McMunn, Nazroo, Wahrendorf, Breeze and Zaninotto2009), it is connected to higher life satisfaction as long as the relationship between effort and reward (of whatever kind) is balanced. A similar beneficial effect on wellbeing is reported by Nikolova and Graham (Reference Nikolova and Graham2014) for part-time and full-time employment of retirees in Europe and the United States of America (USA), as long as the employment is voluntary. Dingemans and Henkens (Reference Dingemans and Henkens2014) find that the unfulfilled wish for bridge employment has negative effects on life satisfaction, while realised bridge employment tends to mitigate the negative satisfaction effects of involuntary retirement; however, they also report negative effects of paid work for financial reasons, compared to intrinsic motives. In a meta analysis, Matthews (Reference Matthews2014) shows that evidence for beneficial effects of working in terms of lower indices of depression is not strong, especially in the USA compared to Europe. Matthews and Nazroo (Reference Matthews, Nazroo and Scherger2015) find negative effects of working on the likelihood of depression and positive effects for self-rated health, provided again that there is a subjectively positive balance between effort and reward. Yet they also show that a good deal of the positive relationship between wellbeing and working is due to selection effects which most other studies do not account for adequately.

Theoretical links between wellbeing and work after pension age

To argue their case, the above-mentioned (psychological and sociological) studies on the effects of retirement or working in retirement employ a broad range of theories and approaches (for an overview, see Wang, Henkens and van Solinge Reference Wang, Henkens and van Solinge2011). Of particular interest for us are role theory, the notion of perceived effort and reward, as well as the psychological concept of individual control. As an additional theoretical framework the life course perspective is useful. Altogether, these approaches can be connected to an inequality-related perspective on paid work past pension age which (for example) differentiates by class positions.

If retirement is experienced as the loss of a role which is important to one's identity, there is a higher probability of an at least temporary decrease in wellbeing – which is more often the case for men and jobs with a high status. As supported by studies on motives for post-retirement work, work identities and social recognition are important drivers of post-retirement work. This work can potentially compensate or prevent the role loss, and therefore have neutral or positive effects on life satisfaction – as found by Dingemans and Henkens (Reference Dingemans and Henkens2014) for the (mitigated) negative effects of involuntary retirement. Furthermore, a productive activity – whether paid or not – is ‘a form of inter-personal exchange founded on the notion of reciprocity’ (Siegrist, von dem Knesebeck and Pollack Reference Siegrist, von dem Knesebeck and Pollack2004), more exactly an exchange of productive activities for related rewards. Some of these rewards, such as money or social recognition, are also important motives for post-retirement work, as seen above. If the efforts that are put into an activity are not subjectively matched by adequate rewards, i.e. if the activity is not valued socially or monetarily, then being active can harm wellbeing and health. Finally, the very general psychological concept of control offers a further possible link between activities and wellbeing (see e.g. de Quadros-Wander, McGillivray and Broadbent Reference de Quadros-Wander, McGillivray and Broadbent2014), which also ties in well with sociological thinking about agency in the life course: just as involuntary retirement is connected to a decrease in wellbeing, activities such as paid work might not alter or even decrease life satisfaction when they are pursued in situations with limited or no scope of action. This is probably often the case for paid work and providing care (for supportive evidence, see Dingemans and Henkens Reference Dingemans and Henkens2014; McMunn et al. Reference McMunn, Nazroo, Wahrendorf, Breeze and Zaninotto2009). At the same time, working after pension age can also be the consequence of a high degree of control over one's work life, when it is, for example, the realisation of the desire to retire gradually from a work position with which a person has strongly identified. Seen from the perspective of sociological life course approaches, the institution of retirement changes and in most cases increases the scope of action with regard to paid work in retirement. The pension system defines retirement institutionally and provides a financial basis (for most individual actors, and to differing degrees in different countries) which at least generally and in comparison to the main working age decreases the financial ‘need’ to work. This might be related to a positive effect of work on life satisfaction in this life phase.

Based on these approaches, it is likely that the experience of re-entering the labour market and its effects on wellbeing will be heterogeneous. The latter will depend on the conditions and circumstances of the decision and of the employment that is newly started, and on the experience of retirement itself as a comparative frame of reference. In particular, the status of the job, effort and rewards related to the job, and control over the decision to work are dimensions which moderate the relationship between paid work after pension age and its effects on life satisfaction. However, none of these are directly or sufficiently measured in the data available here. Therefore, we use social class of the job done in retirement as a proxy for these characteristics of the job and the decision preceding employment. We follow Erikson's and Goldthorpe's class scheme (1993: 29–47; Goldthorpe Reference Goldthorpe and Goldthorpe2007). In doing so, we assume that the retirement job and the job before retirement are closely related, which is particularly plausible for high-class jobs because jobs in these classes are mostly only open to actors with high educational qualifications. Working in a low-class job after retirement may in some cases be the result of downward mobility because these jobs are open to actors with low education and to actors with middle or high education.Footnote ⁴

Furthermore, we assume that jobs in ‘higher’ classes, which are regulated by a service relationship, provide higher individual status (Chan and Goldthorpe Reference Chan and Goldthorpe2004). They are only available to those who had similar jobs before and thus can avoid status losses by working again. Additionally, financial rewards are higher in these jobs, and the same tends to apply to non-financial rewards, such as social recognition or enjoyment. Finally, jobs in these classes also entail higher degrees of individual control regarding the job tasks and the organisation of the work (Evans and Mills Reference Evans and Mills1998). In the age group we examine here, they also provide higher degrees of control with regard to the decision to work because people in ‘higher’ classes usually receive higher pensions which makes them less reliant on extra income after pension age, while they have better and more opportunities to work if they want to. Most of these positive circumstances also apply to the majority of the self-employed. Jobs in the classes at the ‘low’ end of the class scheme (i.e. the unskilled and skilled manual classes and the low routine non-manual class) are regulated by a labour contract and provide considerably fewer or none of the advantages just described. Additionally, working in low-class jobs might be the result of downward mobility, and hence involve potential losses in status, control and income. The advantages of high-class jobs also apply at least in part to jobs in the ‘middle’-class range, whose contracts combine elements of the service relationship and a labour contract.

Based on this, our central assumption (Hypothesis 1) is the following: taking up a new job after pension age is associated with increases in life satisfaction if the new job is not a low-class job; there is no effect or even a negative effect if the new job is in one of the ‘lower’ classes.

Additionally, including the level of satisfaction with health and household income, two of the most important specific domains of satisfaction, helps us to unpack what exactly it is about working that increases or decreases life satisfaction. Remaining effects might be an indication of the role played by non-material gratifications of work such as social recognition, social contacts, etc.

Country comparison

We examine the effects hypothesised in a comparative perspective, contrasting Germany and the UK, because they differ considerably in the institutional arrangements surrounding work and retirement. First, for the period we study below (1996–2008/2011), state pension ages and effective retirement ages have tended to be lower in Germany, with the exception of the state pension age for women in the last decade and including the exceptions and possibilities of early retirement which have been mostly discontinued or are about to cease. This is relevant insofar as we study the same age groups in both countries which nonetheless differ as to the average time lapse since their transition to retirement. Second, and most importantly, the two pension systems and their interplay with the employment system differ. Germany has a stronger first pillar of public pensions than the UK, and combined with the UK's more flexible and deregulated employment system (Organisation for Economic Co-operation and Development 2013; Schulze and Jochem Reference Schulze, Jochem, Immergut, Anderson and Schulze2007; Schulze and Moran Reference Schulze, Moran, Immergut, Anderson and Schulze2007), this arrangement leads to a higher degree of inequality of old-age incomes in the UK. Although these differences are no longer so clear-cut, old-age poverty rates have differed for most of our observation period, with higher poverty rates in the UK (Zaidi Reference Zaidi2010). Furthermore, and connected to these institutional differences, retirement cultures also differ between Germany and the UK, that is, the norms and expectations surrounding retirement and the relationship of work and old age (Flynn Reference Flynn2010; de Vroom Reference de Vroom, Maltby, Mirabile, Øverbye and de Vroom2004). Whereas retirement tends to be more positively connoted as a well-deserved reward for a working life in Germany, the idea of retirement as forced exclusion from the labour market is (somewhat) more strongly rooted in the liberal tradition of the UK where old-age discrimination issues have been debated for longer than in Germany (Scherger and Hagemann Reference Scherger and Hagemann2014). In Germany, by contrast, strong norms of a fixed age for labour market exit still apply that are institutionalised in collective agreements, for example.

It is challenging to assess how these differing institutions affect the impact of working on life satisfaction. First, assumptions can be concluded from differences in the composition of those who retire and those who pursue work post retirement. As public pensions are less generous and (unequally distributed) private pensions are more important in the UK, the inequality of old-age incomes is higher in the UK than in Germany, at least in the period observed here. It is thus plausible to assume that more people re-start working for financial reasons in the UK and that they do so more often in the low-wage service sector (for evidence pointing in this direction, see Hokema and Lux Reference Hokema, Lux and Scherger2015). This should result more frequently in unchanged or even lower life satisfaction caused by paid work beyond pension age. Second, the difference in the retirement culture, which tends to frame work after pension age in a more negative way in Germany, would indicate a more negative effect of re-starting work in Germany, and the opposite or no effect in the UK.

In brief (see Table 1), taking into account the composition of working pensioners and the retirement culture, opposing assumptions are plausible regarding the direction of country-specific effects of post-retirement work on life satisfaction in the UK and in Germany. However, the differences in the retirement cultures are nuanced and it is unclear whether the collective debates and views impact on individual life satisfaction to an extent that is measurable by quantitative data. Therefore, we assume that the ‘hard’ influence of the assumed composition of old-age workers has the more sweeping effect (including class dynamics – see below), or is more likely to have an effect at all. Based on these systematic speculations, we can formulate an exploratory hypothesis (Hypothesis 2) regarding the country differences: taking up a new job after pension age is associated with increases in life satisfaction in Germany, whereas in the UK decreases in life satisfaction are more likely to result.

Table 1. Assumed effects on life satisfaction of working after pension age

If the difference in the composition of post-retirement work is captured well by a broad (and dichotomous) differentiation of the class of the job, potential country differences should disappear after including this class differentiation. If they do not disappear, this speaks for working conditions within classes that differ between countries or for national differences in the circumstances that lead to the recruitment to the jobs of specific classes. If, for example, more people in Germany than in the UK take up low-class jobs in retirement without their finances being the main driver, this would result in a less negative (or more positive) effect of low-class jobs in Germany. A further and not unsimilar assumption is that class differences are stronger in the more deregulated labour market of the UK with its more polarised skill structure, at least at the time of our observation. In brief, we can formulate a third cautious assumption (Hypothesis 3) regarding country differences in the class effects: class differences in the effects of taking up work again are more likely to appear in the UK than in Germany.

Estimation strategy

For the analyses, we use country-specific fixed-effects regression models (FE models) in order to control for all time-constant individual characteristics. This is the method of choice as unobserved time-constant individual characteristics which simultaneously affect the life satisfaction of an individual and their propensity to work after pension age (such as optimism) are very likely to bias the effects of work beyond pension age on life satisfaction. With FE models, we avoid this endogeneity problem as we focus on the within-variation of individuals, that is, on how an individual's life satisfaction changes over time and under the influence of the time-varying independent variables. In this way, we control for all time-constant individual characteristics as well as for the included time-varying aspects. As within-variation can only be analysed for respondents who exhibit some variation in the central independent variable, we focus – for the estimation of the effect of work on life satisfaction – on respondents who start working again after having reached pension age. Other work-related patterns of change (such as stopping work) are excluded as they cannot easily be related to the effects of late employment; the same applies to respondents who continuously work and who, from a statistical point of view, have no work-related within-variation which we could analyse.

The FE models are formalised in the following way:

$${y_{it\,\,}} - {\bar y_i}\, = \,\beta ({x_{it}} - {\bar x_i}) + ({u_{it}} - {\bar u_i}).$$

In these models, the person-specific mean of the dependent ( ${\bar y_i}$ ) and the independent variables (vector ${\bar x_i}$ ) over the observed ages are subtracted from the observed values at specific ages (y _it and vector x _it ), so that only within-person variation remains to be analysed (Allison Reference Allison2009; Brüderl and Ludwig Reference Brüderl, Ludwig, Wolf and Best2015). The coefficient for the independent variable (working status) thus shows the change of the dependent variable (life satisfaction) if a respondent has started working again after age 65. Due to this demeaning, the person-specific part of the error term ( ${\bar u_i}$ ) is also cancelled out and only time-varying errors (idiosyncratic errors) are left. To get unbiased coefficients in FE models, the independent variables must not be correlated with the time-specific error term (Wooldridge Reference Wooldridge2010: 281–315). Or to put it another way: there should be no omitted individual changes which affect the life satisfaction of an individual and at the same time their propensity to work after pension age. Therefore, we control for age and periodFootnote ⁵ in the FE models, as well as for change of marital status, change of satisfaction with health and change of satisfaction with household income. The variable of crucial importance, however, is the change of the working status which relates only to starting a job here. We apply panel-robust standard errors to allow for autocorrelation and heteroscedasticity (Brüderl and Ludwig Reference Brüderl, Ludwig, Wolf and Best2015).

Data

For the analysis, we use the GSOEP and the BHPS (for a description of the surveys, see Haisken-DeNew and Frick Reference Haisken-DeNew and Frick2005; Taylor et al. Reference Taylor, Brice, Buck and Prentice-Lane2010). Both surveys provide annual information on living conditions, employment status, attitudes and many other individual characteristics of respondents, who are aged 16 or older. For the UK, we use the waves 1996–2000 and the waves 2002–2008 because no questions related to life satisfaction were asked in 2001 and before 1996.Footnote ⁶ For Germany, we use data from the waves 1996–2011.

In the FE models, we only include respondents who have at least two respondent years (that is, interviews they participated in) and started a job between age 65 and 75.Footnote ⁷ For these respondents, we only look at the jobless years before the uptake of the job and the years in the job. We exclude the respondent years after the end of the first ‘new’ job as we do not want to include the within-effect of ending a (post-retirement) job in our estimations. Furthermore, respondents without any paid job between ages 65 and 75 are also included in the models in order to improve the estimation of the effects of our further independent variables in the FE models, especially that of age (Brüderl and Ludwig Reference Brüderl, Ludwig, Wolf and Best2015). In sum, our sample consists of respondents who start to work again after age 65 and respondents without any work between ages 65 and 75.

Only with these specifications do we approximate the estimation of the causal effect of working, independently of other factors correlated to this change, and thus (at least partly) avoid the problem of endogeneity.Footnote ⁸ This set-up of our fixed-effects analyses also implies that we study a specific and selective group of working pensioners, that is, those who go back to work after a time of not working (see the section ‘Post-retirement Work’ above; see also Kanabar Reference Kanabar2012). On the one hand, we thereby exclude people who simply continue working and defer pension receipt and who are not working pensioners in a strict sense. On the other hand, we also exclude pensioner jobs which directly follow after the end of a respondent's main career, or side jobs which are continued beyond pension age (for the implications of this approach, see the last section).

After this specification and the deletion of person years with missing valuesFootnote ⁹ in one of the variables mentioned below, we are left with 6,042 respondents (33,927 respondent years) in Germany and 2,941 respondents (14,032 respondent years) in the UK. Among these respondents are 390 Germans (1,492 respondent years) and 129 Britons (500 respondent years) who started a job between ages 65 and 75. With regard to the class position, we find 227 older German respondents who start to work in low classes and 163 who start to work in other classes. In the UK, the corresponding numbers are 41 (low classes) and 88 (other classes).

Variables

The measurement of life satisfaction, the dependent variable of our study, differs slightly between the two countries. In the UK, it is measured on a seven-point scale based on the question ‘How dissatisfied or satisfied are you with your life overall?’, with answer categories ranging from 1 (not satisfied at all) to 7 (completely satisfied). In Germany, life satisfaction is measured on an 11-point scale based on the question ‘How satisfied are you with your life, all things considered?’, with categories ranging from 0 (completely dissatisfied) to 10 (completely satisfied). As their scales are different, we transform the German variable to correspond to the British one, using a transformationFootnote ¹⁰ proposed by Proto and Rustichini (Reference Proto and Rustichini2012).Footnote ¹¹

Our main independent variable is working status, which included every kind of paid employment. Respondents are defined as working if they state (a) that their current working status is full-time employed, part-time employed, self-employed or marginally/irregularly employed; or (b) that they currently have some kind of second job; or (c) if they have been engaged in paid work in the last seven days. Furthermore, we categorise the current job according to the class scheme of Erikson and Goldthorpe (Reference Erikson and Goldthorpe1993; Goldthorpe Reference Goldthorpe and Goldthorpe2007). For this purpose, we use the class variables provided by the GSOEP and BHPS.Footnote ¹² The classes are then dichotomised into low classes (semi- and unskilled manual workers, low routine non-manual and skilled manual workers) and other classes (all other classes).Footnote ¹³

We include several further variables: change of marital status (married versus not married/widowed/divorced), satisfaction with health and satisfaction with household income (both seven-point scales in the UK and transformed eleven-point scales in Germany, with positive values indicating higher satisfactionFootnote ¹⁴ ), age and period (year of observation).