Introduction
The link between previous lifecourse experiences and later-life outcomes is one of the central research questions of the lifecourse perspective (Elder and O'Rand Reference Elder, O'Rand, Cook, Fine and House1995). Unfortunately, when it comes to differences in economic wellbeing between older people, the lifecourse processes of which this so-called ‘aged heterogeneity’ (Dannefer Reference Dannefer1987) is the outcome, are still not very well understood. Most empirical research has been devoted to analysing the causes and consequences of single events and transitions, rather than to analysing long-term careers and the ways these are intertwined (e.g. Mayer Reference Mayer2009).
In this paper, the focus is on the link between previous lifecourse experiences and the incomes of older people in Belgium and the United Kingdom (UK). Prospective panel data and retrospective life-history information from the Panel Study of Belgian Households (PSBH) and the British Household Panel Survey (BHPS) are combined, allowing the construction of ‘summary’ variables pertaining to ‘life-time’ family trajectories and labour market careers of 1,083 Belgian and 1,380 British men and women, aged 60 or older. The analysis also explores whether disadvantaged positions or events over the lifecourse reinforce each other, resulting in a relatively lower income in old age. Drawing on the concept of ‘cumulative advantage and disadvantage’, two mechanisms that might contribute to income differences between older people are tested. The first mechanism reflects the idea that ‘cumulative exposure’ to adverse situations has a stronger negative effect on later outcomes compared to short-lived experiences. Secondly, the mechanism of ‘status–resource interaction’ refers to a process through which people who begin their lifecourses under more disadvantaged conditions are comparatively less able to benefit from resources or opportunities that cross their life paths. Finally, the impact of institutional arrangements is examined by taking a country-comparative perspective and by looking at the lifecourse determinants of old-age incomes for different birth cohorts.
The paper begins with a discussion on how social inequality in old age might be structured by previous lifecourse experiences, followed by a review of specific family and labour market determinants. The theory section concludes with a comparison of institutional arrangements in the UK and Belgium. The section thereafter provides information on data and measurements, followed by the empirical results. The final section presents an overview and discussion.
Conceptual building blocks: unravelling cumulative (dis)advantage
Cumulative (dis)advantage can be described as ‘a general mechanism for inequality across any temporal process in which a favourable relative position becomes a resource that produces further relative gains’ (DiPrete and Eirich Reference DiPrete and Eirich2006: 271). Applied to the development of lifecourse inequalities, this hypothesis states that, over time, sources of social, psychological, cultural and economic capital, in interaction with institutions, opportunity structures and individual life events either accumulate or become depleted (e.g. Dannefer Reference Dannefer1987; O'Rand Reference O'Rand, Binstock and George2006). As Mayer (2009: 424) points out, until recently the idea of cumulative (dis)advantage was ‘not much more than a metaphor’. It consequently has many connotations, ranging from a process where future accumulation strictly depends on current accumulation and inequality grows over time in a continuous, path-dependent way to more ‘relaxed’ conceptualisations in terms of status–resource interaction, cumulative exposure to negative situations, a higher risk of being confronted with negative life events, or the supposedly beneficial effect of being among the ‘precious and precocious’ (O'Rand Reference O'Rand1996). This process is furthermore influenced by human agency, as the accumulation of previous life experiences has an impact on the way people cope with certain situations and on labelling processes by others – leading to further differentiation. Given this complex set of mechanisms, we should not assume that lifecourse trajectories are characterised by a constantly increasing differentiation: flat or convergent trajectories are equally likely, as well as trajectories marked by more or less abrupt changes.
Although the concept of cumulative (dis)advantage is appealing as it offers a social explanation for the often-observed ‘aged heterogeneity’ in (economic) resources, it is difficult to test empirically. Often, researchers do not have suitable longitudinal data covering the prospective evolvement of lifecourse (and especially income) trajectories. Rather than trying to chart the development of income trajectories over time, in this paper the focus is on the impact of family and labour market determinants on income in old age. Two possible mechanisms arising from the cumulative (dis)advantage framework are furthermore examined. A first mechanism is easy to understand as ‘cumulative exposure’: the longer an adverse situation persists, the lower the income in old age. Recently, researchers from different disciplines have focused on the persistence of, for instance, poverty during specific life stages and on its impact in terms of subsequent socio-economic attainment and health (e.g. Wagmiller et al. Reference Wagmiller, Kuang, Aber, Lennon and Alberti2006; Willson, Shuey and Elder Reference Willson, Shuey and Elder2007).
A second mechanism relates to the idea of ‘status–resource interaction’, defined by DiPrete and Eirich (2006: 273) as ‘persisting direct and interaction effects of a status variable’, where the interaction effects imply group differences in socio-economic returns dependent on the initial position. The authors trace this mechanism of cumulative (dis)advantage back to Blau and Duncan's (Reference Blau and Duncan1967) The American Occupational Structure. The latter report how race acts as the ‘status’ variable, in the sense that highly educated blacks received less returns to their invested resources than highly educated whites. In this paper, taking educational background (both of the parents and of the respondents) as the status variables, the occurrence of status–resource interaction is evaluated by testing whether education interacts with lifecourse trajectories in such a way that ‘negative’ events or situations result in an additional income disadvantage in old age for respondents ‘starting from’ a lower initial position. Although this mechanism suggests a more or less ‘continuous’ process of cumulating (dis)advantages, a significant interaction does not contradict a more complicated process of ‘ups and downs’.
At a lower level: the lasting impact of family and labour market experiences
Apart from the theoretical literature on how lifecourse trajectories might evolve over time, there is a wealth of empirical research focusing on the short- and long-term consequences of specific life events. As the main topic of this paper concerns the impact of family and labour market experiences on incomes in old age, in this section a number of relevant mechanisms and findings are highlighted.
An important qualification with regard to existing research on economic inequality in old age is that studies are often limited to respondents with more or less stable working lives, which are usually male. Women less often figure, as their professional careers are either non-existent or difficult to chart. Female poverty in old age is hence usually related to widowhood (e.g. Hungerford Reference Hungerford2001) or to ‘atypical’ family trajectories characterised by divorce or lone parenthood. O'Rand and Landerman (Reference O'Rand and Landerman1984), however, argue that economic wellbeing in old age is for both men and women dependent on their professional and family experiences. For women, securing a decent job is hampered by the prominence of their family role. The so-called ‘child penalty’ in disposable household income following the birth of a child is strongly related to changes in mothers' labour market participation (Kalmijn Reference Kalmijn2002). Gornick, Meyers and Ross (1998) found that ‘family-friendly’ policies influence the employment decisions of mothers. From a lifecourse perspective, these decisions have an impact on the long-term employment patterns of women, with implications for their later earnings potential and economic wellbeing in later life (e.g. Sigle-Rushton and Waldfogel Reference Sigle-Rushton and Waldfogel2007).
Concerning lifecourse labour market experiences, a first straightforward hypothesis states that, controlling for occupational class, employment continuity has a positive impact on old-age income. It is expected that men and women with an unstable labour market career have a lower income in old age. Not only do people accumulate less resources when they are dependent on a replacement income or take time out, there is also a range of studies, particularly within the context of unemployment, pointing to a so-called ‘scarring effect’ (e.g. Clark, Georgellis and Sanfey Reference Clark, Georgellis and Sanfey2001; Gangl Reference Gangl2004). Proposed explanations are psychological (discouraged workers, habituation to unemployment and fading work attitudes, resulting in longer unemployment spells and leading to higher risks of subsequent unemployment), point at human and social capital loss (making it difficult to secure re-employment of a similar skill-level and wage), or look at the impact of institutions (low benefits pushing people into low-paid jobs that do not match their skills).
The consequences of early retirement for old-age income are less clear. On the one hand, every additional year of retirement results in a lower pension, because one is expected to live for more years on a pension derived from an incomplete contribution record. Also, people usually implement strategies to safeguard their income in retirement during their pre-retirement years, typically when the children have left home and/or when their mortgage has matured. The (unexpected) event of early retirement can force people to tap into their assets at a time when they should be accumulating additional resources (Pacolet and Van Steenbergen Reference Pacolet, Van Steenbergen and Cantillon2005). On the other hand, causality might be reversed, as assets such as outright home-ownership might encourage people to retire early (Doling and Horsewood Reference Doling and Horsewood2005). There might well be two ‘types’ of early retirement: the voluntary type, chosen by better-off respondents who can afford it, and the involuntary type, resulting in a lower-than-expected income, jeopardising economic wellbeing in old age.
With regard to the impact of family experiences, Wilmoth and Koso (Reference Wilmoth and Koso2002) found that marriage offers more opportunities for wealth accumulation compared to other living arrangements, including cohabitation. Next to economies of scale, marriage results in labour specialisation, a larger social network and a higher participation in ‘protective’ institutions, such as home-ownership and life insurance, that promote savings behaviour and opportunities for the accumulation of economic resources. A well-known hypothesis is the so-called ‘marriage premium’, which states that married men have better jobs, brighter prospects, higher earnings, less health problems and a higher life expectancy compared to their non-married counterparts (e.g. Kalmijn Reference Kalmijn2002). These positive effects have been ascribed to influences such as emotional support and a healthy lifestyle. Recent research, however, has shown that the education of the wife turns out to be the most important determinant of her husband's success on the labour market, as partners seem to function as each other's social capital. Turning the argument around, although many studies have shown that for men, divorce does not have major financial consequences (e.g. Andreß et al. Reference Andreß, Borgloh, Bröckel, Giesselmann and Hummelsheim2006), men do seem to suffer in other ways, through the so-called ‘divorce penalty’ (Kalmijn Reference Kalmijn2005). Divorced men run a higher risk of downward social mobility, unemployment and invalidity – all of which are likely to have an impact on later-life economic wellbeing. For women, many studies have shown that divorce results in severe financial deprivation, both in the short and long term (e.g. Jenkins Reference Jenkins, Brynin and Ermisch2009; Uunk Reference Uunk2004). The difference between men and women can be related to the gendered division of paid labour and male–female wage differentials. Also, many divorced mothers have custody of their children and are thus their main provider. With regard to the main research question, it is expected that older men and women in an intact first marriage have a higher income compared to respondents who remained single or experienced relationship disruptions such as widowhood or divorce, even when they have eventually repartnered. Furthermore, it is expected that the long-term effect of disrupting family events is stronger for women compared to men.
The effect of having children on old-age incomes is also explored. As explained above, children cost money in terms of their impact on female labour. Raising children, however, also comes with direct costs – for food, housing, health care, education and child care. In modern welfare states, these costs are partly compensated for by child allowances, tax credits and in-kind benefits. There are, however, large differences in the financial responsibility assumed by different welfare states for children of different ages (Bradshaw and Finch Reference Bradshaw and Finch2002). On the one hand, it is expected that older men and women have a lower income in old age as the number of children ever had increases. On the other hand, it is possible that the income gap between people with and without children is narrowed by mechanisms such as the above-mentioned ‘marriage premium’ or by differences in spending patterns, savings behaviour and intergenerational solidarity.
The above literature review shows that the economic wellbeing of older men and women is the result of a host of short-term and long-term influences, which can be of a very different nature, but are for both men and women related to family and labour market experiences. The following section explores how (a) labour market and family lifecourse experiences and (b) the earlier identified mechanisms of cumulative (dis)advantage – cumulative exposure and status–resource interaction – might differ between the UK and Belgium.
Different in different welfare regimes?
The Belgian welfare state is situated on the boundaries of the conservative and social-democratic regime (for a review, see Esping-Andersen Reference Esping-Andersen1990 and the ensuing research tradition). On the one hand, social and fiscal policy is based on the traditional male-breadwinner model. Historically, female labour market participation has been rather low, resulting in a high number of single-earner households. For the younger generations, this traditional family pattern is, however, limited to the low-skilled (De Lathouwer et al. Reference De Lathouwer, Cantillon, Andries and Marx1999). Social security benefits are differentiated according to professional status and linked to previous employment, with different schemes for civil servants, employees (subdivided into ‘white-collar workers’ and ‘blue-collar workers’) and the self-employed. From a comparative perspective, replacement rates for the main programmes (retirement, unemployment) are rather low. The Belgian welfare state nevertheless achieves a high level of ‘de-commodification’ or market-independence, comparable to the Scandinavian countries. Although social rights are linked to employment, this link has been loosened over time. Combined with a strong tailoring of benefits according to family type and selective measures targeted at the protection of low-income households, this uncoupling between employment record and social rights results in relatively low income poverty risks, both for dual-earner and single-earner households – be it to a lesser extent for the latter (Cantillon Reference Cantillon1999; De Lathouwer Reference De Lathouwer1996).
In the aftermath of the Second World War, British social policy was based on the same Beveridgean principles as the early Scandinavian model: universal basic benefits, a universal health-care system and a political commitment to full employment. From the 1970s onwards, the UK started to follow a more liberal course. Basic benefits were never upgraded to an adequate replacement level, pressing the middle classes to resort to private insurance. This ‘stalled development’ was reinforced by the succession of Conservative governments since 1979 (Korpi and Palme Reference Korpi and Palme2003). The electoral victory of New Labour in 1997 did not result in a major reversal of existing social policies, although various initiatives have been taken to protect the income of working families (e.g. Sutherland and Piachaud Reference Sutherland and Piachaud2001). For the respondents in this study, these changes are irrelevant, as most of them were already retired in 1997.
Although considered as a liberal welfare state, social expenditure in the UK is not much below the levels reached by ‘high-spending’ countries such as France or Belgium. Therefore, Castles and Mitchell (Reference Castles, Mitchell and Castles1993) qualified the UK as a ‘radical’ liberal welfare state, aimed at the targeted redistribution of resources in order to prevent poverty. Nevertheless, income poverty and inequality in the UK are among the highest in Europe (EU-15) (Dennis and Guio Reference Dennis and Guio2004). Furthermore, this type of welfare state produces a dual society, leaving the lower social groups to rely on less generous transfers, whereas the middle and higher classes take out private insurance. Whereas in Belgium most older people are dependent on the first-pillar public pension (and on their personal savings), in the UK about half of the income in old age is derived from occupational pensions, next to other individual arrangements (Dewilde Reference Dewilde2004). According to Walker and Hutton (Reference Walker, Hutton, Walker and Parker1988), there are two ‘nations of elderly’ in the UK: those with and without a private pension.
In the UK, more flexible employment relationships have resulted in a stagnation of real wages and growing income inequality (e.g. Esping-Andersen Reference Esping-Andersen1999). Although unemployment among the young and low-skilled is relatively low, they are often employed in lower-quality service-sector jobs generating incomes below the poverty line. Low-educated women are especially likely to hold small part-time jobs with no social rights attached and to interrupt their labour market careers following the arrival of children (Joshi and Davies Reference Joshi, Davies, Ruspini and Dale2002).
The Belgian labour market, on the other hand, is strictly regulated (Delsen Reference Delsen1997). Labour law concerning fixed-term labour contracts is fairly strict, and small part-time jobs are virtually non-existent (De Lathouwer and Marx Reference De Lathouwer and Marx2002). Long-term unemployment in Belgium is rather high. Furthermore, many older people losing their jobs never return to the labour market. For this group of discouraged workers, collectively negotiated early retirement schemes often provide a generous replacement income.
From the above it is not easy to derive specific hypotheses with regard to the impact of welfare regimes on processes of cumulative (dis)advantage over the lifecourse. On the one hand, in Britain the market is a more important determinant of the distribution of risks and resources. On the other hand, in Belgium the occupational stratification of most social security programmes implies that early lifecourse statuses might be related more strongly to later outcomes, although the Belgian welfare regime comes across as highly de-commodifying. Given the higher reliance on the male-breadwinner model, combined with an overall high level of de-commodification, it is expected that family trajectories in Belgium are more important determinants of income in old age than in the UK, whereas labour market trajectories are less important compared to the UK, where market mechanisms prevail over inequalities caused by welfare arrangements. Furthermore, it is expected that processes of cumulative (dis)advantage are more pronounced for the younger generations in the UK, who spent a larger part of their ‘active’ lifecourse under a liberalising welfare regime.
Data and measurements
Data
The analyses are based on the Panel Study of Belgian Households (PSBH 1992–2002) and the British Household Panel Survey (BHPS 1991–2005). Both are household panels, tracing a ‘first wave’ of households and individuals over time with annual interviews. In each wave, both samples are representative of the population in private households (Dewilde et al. Reference Dewilde, Bauwens, Marynissen and Lauwers2000; Taylor et al. Reference Taylor, Brice, Buck and Prentice-Lane2006).
The analysis sample consists of all respondents aged 60 or older at the time of the last panel wave for the PSBH (2002) and at the time of wave 15 (2005) for the BHPS. Rather than including interactions with gender and country, four subsamples – a male and female sample for each country – are identified. This also allows for a different operationalisation of the labour market experiences of men and women, as these are obviously ‘gendered’ for current generations of older people. While for men, the relevant focus is on labour market continuity and on the experience of early retirement, for women, the main question is whether they have ever worked and for how long. Complete information is available for 481 Belgian men and 602 Belgian women, and for 577 British men and 803 British women.
Although both panel studies were not set up as retrospective studies, it is nevertheless possible to map lifecourse family and labour market experiences. The resulting trajectories are obviously less detailed compared to studies specifically aimed at collecting retrospective life histories. Regarding family history, the available information on earlier marriages, cohabitations and fertility collected in the first wave of the PSBH and the second wave of the BHPS is used, prospectively updated for each subsequent wave depending on the registered changes in status (e.g. respondents who experienced divorce or widowhood), up to the wave on which the analysis samples are based. In wave 7 (1998), an additional Flemish sample was added to the PSBH and retrospective family histories were again collected, so that family histories could be constructed for those respondents who joined the sample between wave 1 and wave 7 (for instance, by marrying a panel member). For the BHPS, life-time family trajectories were included again in wave 8. Also, cross-checking and updating of this information was made with the British Household Panel Survey Consolidated Marital, Cohabitation and Fertility Histories, 1991–2006 (Pronzato Reference Pronzato2007). In the last wave of the Belgian panel (2002), life-time labour market careers of respondents were charted using their most important activity in each year since completing education. For the British data, similar-format information (which in this case was collected based on person-months) was derived, using the British Household Panel Survey Combined Work-life History Data, 1990–2005 (Halpin Reference Halpin2006).
Given that the available information was collected at different levels of detail for both panel studies and for the different retrospective and prospective parts of the questionnaires, it is the least detailed ‘common denominator’ that determines the content of the ‘summary’ variables describing the family and labour market experiences of the respondents. This is a drawback, as it is nearly impossible to determine the timing of important lifecourse events without losing respondents due to missing information. The independent variables are hence the result of a careful deliberation aimed at retrieving the highest amount of information for the largest number of respondents. Because of the country-comparative perspective, this also means that the less-detailed information contained in the Belgian data ‘forced’ a relative ‘under-utilisation’ of the British data.
As the longitudinal non-response weights that come with the data are by definition zero for non-panel members (i.e. those respondents who joined the panels after wave 1) and for panel members who skip an interview (which is a problem for the Belgian data), they were not used. Other longitudinal research for selective subsamples has shown that there is generally not much difference between unweighted and weighted results (e.g. Jenkins Reference Jenkins, Brynin and Ermisch2009, tracing divorced women over time). Furthermore, as the household cross-sectional weights are derived by ‘redistributing’ the longitudinal weights to all household members, the former could be considered a ‘dirty’ approximation of the latter. Although the estimates fluctuate somewhat, re-estimating the models using the household cross-sectional weights did not change the substantive conclusions (results upon request).
Measurements
The dependent variable is household income. For Belgium, the current monthly disposable household income (€) as estimated by the household reference person is used, which has a very high item-response rate (96%). In the British data set, household income variables are gross and are the sum of labour incomes, pensions, benefits, private transfers and investment income. Although net income variables are supplied (Bardasi et al. Reference Bardasi, Jenkins, Sutherland, Levy and Zantomio2008), response rates are considerably lower (around 80% compared to 96% for gross incomes). As gross and net household incomes in the UK are strongly correlated (0.75 for the subgroup of older people; 0.93 for the total sample), gross monthly household income (£) is used. Although it is known from previous research that household income estimated by the reference person tends to underestimate the total income derived from adding up all components, it is also known that this underestimation gets worse for larger households (Dewilde Reference Dewilde2004). As most respondents in the analysis sample are single or living in a couple, it is assumed that the results are not biased by using a different income concept in both countries. For older Belgian couples for instance, the average wave 1-difference between both income measures was only €12. To adjust for differences in the size and composition of households, the Modified Organisation for Economic Co-operation and Development (OECD) equivalence scale is used. This scale attributes a weight of 1 to the first adult in the household, each additional adult is given a weight of 0.5 and each child younger than 14 years of age is attributed a weight of 0.3.
Given the interest in mechanisms of cumulative (dis)advantage, wealth, as something that actually accumulates, is perhaps a more suitable dependent variable. Unfortunately, both panel studies do not contain detailed information on different types of assets. The income measures used in this research are supposed to take all income sources into account, including income from investments, but also from occupational and private pensions, which are the result of accumulation over time. Furthermore, in Belgium the public pension – which is the main source of incomes in old age – is dependent on previous earnings, and as such reflects the accumulated labour experience. From a comparative point of view, as the ‘pension-mix’ differs between countries, disposable income from all income sources is an equally relevant dependent variable compared to a measure based on wealth.
The choice to use household incomes rather than personal incomes is based on the fact that the economic wellbeing of older men and women is not only dependent on their personal experiences, but also on the household economic strategies they deployed over time. Even today, these strategies often involve a gendered division of paid labour. An extreme situation concerns the 22.7 per cent of Belgian women who never entered the labour market. Given that their husband receives a so-called ‘family pension’, their personal income often amounts to zero. For these women, the equivalised household income is obviously a better indicator.
A further complication arises as in each of the subsamples, 10–15 per cent of respondents live in a ‘complex’ household, e.g. with adult children or siblings. This is controlled for by including a variable indicating whether respondents form part of a ‘complex’ household.
OLS-regression models are estimated, with the natural log of the equivalised household income as the dependent variable. A positive effect means that a certain predictor positively influences household income.
An overview of the variables is given in Table 1. Important control variables are birth cohort, the educational level of the respondent (primary versus secondary education), and the educational level of respondent's parents (no diploma versus a primary school certificate). These dichotomies may seem an oversimplification at first sight, but one has to remember that the respondents were children during the inter-war years, while their parents were generally born before the First World War. When data are available for both parents, the value of the parent with the highest education is used. Educational levels are generally higher in Belgium compared to the UK, especially for respondents' parents.
Table 1. Description of variables
Notes: SD: standard deviation. 1. For Belgium, €; for the UK, £ (monthly amounts). Reference category for categorical predictors in parentheses. SD not shown for categorical variables.
Concerning the family history, a variable based on whether, and how many times, the respondent has lived together with a partner was constructed. The ‘number of children ever had’ was also computed. Labour market experiences are charted using the following variables: the duration of the labour market career (in categories, for men), the number of years in the labour market (continuous, for women), the number of years in unemployment or disability, the number of years of working as a blue-collar worker, the number of years of working in self-employment or as a farmer, whether the respondent retired early (for men) and whether the respondent has ever had a paid job (for women). Notable differences between both countries are that British men have longer labour market careers, but also retire early more often – private pension plans usually contain financial incentives to leave the labour market before the age of 65 (Blundell and Johnson Reference Blundell, Johnson, Gruber and Wise1999). Most British women have done some paid work over their lifecourses, whereas one-fifth of Belgian women never entered paid labour. Also, British women on average have worked more years. This difference might be partly caused by the fact that in the British data monthly job-history information was collected, so that even short labour market spells are easily recalled. In the Belgian panel, only the most important activity during each calendar year since leaving full-time education was recorded, which might mean that part-time or short-term jobs (before marriage, for instance) were not registered.
A final drawback is caused by the fact that the effects of the partner's family and labour market trajectories, if applicable, cannot be modelled. Indeed, for those respondents in the samples of analysis who are widowed or divorced, information concerning the lifecourse experiences of the partner is often missing – this is both depending on the timing of widowhood or divorce (before or during the panel period) and on the timing of the retrospective modules (before or after widowhood/divorce). Dropping these cases would have resulted in too much loss of information.
Empirical results
Lifecourse determinants of retirement incomes
Table 2 reports on the main-effects models for all sub-samples. Turning first to the control variables, with the exception of Belgian women, the older cohorts have lower incomes in old age, which of course can be linked to the progression in coverage and replacement rates over time of public and private pension arrangements (e.g. Disney and Whitehouse Reference Disney and Whitehouse2001). Originating from a family where at least one parent received some education significantly increases household income of Belgian older people, even after controlling for respondent's own education, household composition, and family and labour market experiences. In the British sample, parents' education also increases respondent's income in old age, but the effect is mediated through respondent's education (separate analyses not shown). Having a low education oneself leads to a lower income in old age. For Belgian men, the effect is no longer significant at the 0.05 level (p<0.10) when the labour market variables are added (separate analysis not shown).
Table 2. Impact of family and labour market experiences on income in old age
Notes: Ref.: reference category. Dependent variable: log(equivalised household income +1).
Significance levels: * p<0.05, ** p<0.01, *** p<0.001.
In the Belgian samples, there are no significant differences between respondents living in a ‘complex’ household versus respondents who are single or form part of a couple. In the UK, respondents in a complex household have a significantly higher income in old age. It is of course possible that the specific types of ‘complex’ households differ between countries (e.g. more older people in the UK sharing a household with adult children), but given the small size of this group of respondents, further analysis was not pursued.
The impact of marital history differs across genders and countries. For British men, no significant effects were found. Belgian single men who ever married or cohabited actually have a higher income in old age compared to men in an intact first marriage. This unexpected result could be an artefact of the way pension benefits in Belgium are calculated. The analysis examines household income, which mainly consists of a ‘family pension’ for couples where the wife never worked, or did not work enough to qualify for an individual pension. This family pension for the main breadwinner is based on 75 per cent of former wages, compared to 60 per cent for a single person. It thus comes as no surprise that single men who were ever in a partnership (most of whom are now widowed) are actually financially better-off compared to their married counterparts, who ‘share’ their slightly higher pensions with their spouses. However, indirect evidence for the so-called ‘marriage premium’ was found, as single men who ever married or cohabited have a higher income in old age compared to single men who never lived with a partner (the contrast between both categories is significant at the 0.05 level). Although it cannot be determined how this ‘marriage premium’ comes about, the fact that it persists into later life is a new finding.
For women, differences between Belgium and the UK were also found. Whereas in Belgium single women who never married or cohabited have a significantly lower income in old age compared to older women in an intact first marriage, in the UK this is the case for women who are single, but did have a partner at some time during their lives – and are now either widowed or divorced. The estimate for never-married women in both countries is controlled for their own labour market experiences (ever had a paid job and number of years in employment). Interpretation is hence not straightforward, apart from the argument that the characteristics of this particular group of women might differ between both countries. The interpretation of the country difference for women who were ever in a marriage or cohabitation (most of which are now widowed) is more straightforward. While Belgian women receive a ‘survivor's pension’ (80 per cent of the ‘family pension’) upon the death of their husband, two out of five British women lose their partners' occupational pension and hence become dependent on the (much) lower state pension, while only one in five widows has an occupational pension of her own (Disney and Whitehouse Reference Disney and Whitehouse2001). Furthermore, survivor's benefits usually amount to only 50 per cent. Widowhood is hence a well-documented correlate of low income for British older women.
The effect of the experience of divorce or widowhood as such does not seem to have persistent consequences for old-age incomes: in all sub-samples there is no significant difference between respondents in an intact first marriage compared to respondents who are currently in a couple, but have experienced relationship disruption in the past. This leads to the conclusion that income differences in old age are mainly caused by living arrangements (couples versus singles), rather than lifecourse family experiences. No further interactions with the family history variable were tested.
For Belgian respondents, each additional child significantly decreases household income in old age. In the UK, this variable has no impact, both before and after controlling for respondent's labour market experiences. From the descriptives discussed earlier, it became clear that British women had significantly more work experience than Belgian women. A tentative conclusion is hence that the ‘negative child effect’ for Belgian respondents is due to the fact that – at least for the cohorts currently examined – mothers left the labour market or never entered it. An alternative, or rather complementary, explanation would be that the direct costs of raising children were higher in Belgium. This is not unlikely, as Britain saw the development of many welfare services in kind (e.g. universal health care). However, the difference in direct costs is in all likelihood not a major explanatory factor.
Looking at the labour market variables, occupational class has a pervasive impact on incomes in old age. Each additional year in blue-collar work lowers income in old age by 0.7 per cent (eB) in Belgium and 0.5 per cent in the UK, whereas each additional year in self-employment or farming results in a proportional decrease of 0.8 per cent in Belgium and 0.6 per cent in the UK. Although these coefficients seem small, they are not: a Belgian man who has worked as a blue-collar worker for 35 years has an income in old age that is 24.5 per cent lower compared to someone who never did. Furthermore, having worked as a blue-collar worker or in self-employment or farming is associated with a ‘long’ labour market career (45 years or longer) in both countries (the latter effect is significant when the ‘duration in occupational class’ variables are removed from the model), and in Belgium blue-collar work is also associated with early retirement (separate models not reported). In the UK, an additional significant effect of labour market interruptions caused by unemployment or illness is found: each additional year in these statuses results in an income decline of 2.0 per cent. This result is in line with previous research on the scarring effects of unemployment (e.g. Gangl Reference Gangl2004), with more persistent negative effects in liberal welfare states, where the unemployed are forced into labour at the cost of lower wages, less job protection and higher risks of further unemployment. The current analysis shows that these scarring effects persist even into old age.
For British women, none of the labour market variables reaches significance. Belgian women who never had a paid job have an income in old age that is 25.7 per cent lower compared to women with labour market experience. Again, blue-collar work or self-employment/farming substantially depress old-age income.
No main effects of early retirement on old-age incomes for men in both countries are found. However, in the theoretical section it became apparent that this might be related to ‘hidden’ heterogeneity, as there possibly exist different routes into early retirement. Interactions with ‘initial social position’ and the number of years in blue-collar work were therefore tested.
With the exception of British women, the explanatory power of my models is quite high. R 2 is higher for the Belgian models than for the British models. Also, the models fit better for men than for women, which is to be expected given the fact that the incomes of older women are more dependent on their partners' lifecourse experiences. Although the results show that income in old age is dependent on family and labour market experiences for both men and women, this seems to be more so in Belgium compared to the UK. More specifically, in the UK family history only has limited explanatory power, if any, whereas for British men, more labour market variables are significant. This is in line with expectations. There is also evidence of ‘cumulative exposure’ in both countries: the more years of one's labour market career spent in a ‘lower’ occupational class, the lower household income in old age. Except for British women, these effects are quite substantial. Both the welfare state and the market are hence ‘efficient’ mechanisms translating occupational class differences into old-age income. In fact, the ‘occupational stratification’ of old-age incomes by the welfare state turns out to be a more efficient mechanism than the market. After pooling the male samples (converting £ into € for the British men), significant interactions between ‘number of years in blue-collar work’ and ‘number of years in self-employment or farming’, on the one hand, and a country-dummy, on the other hand, are found. Each additional year in these occupational statuses leads to a significantly lower old-age income in Belgium compared to the UK. Both interaction terms are significant at the 0.01-level (results available upon request).
‘Status–resource interaction’ and institutional change
Earlier, a mechanism of ‘status–resource interaction’ was put forward as referring to a process where people who begin their lifecourses under more disadvantaged conditions are comparatively less able to benefit from resources or opportunities that cross their life paths. Such a process would be consistent with a negative interaction between ‘negative’ lifecourse events or statuses and a disadvantaged ‘initial social position’, measured by the education of the respondent and of his/her parents. Interactions with parents' education are only tested for Belgium, as we saw above that in Britain, the impact of this variable is mediated by respondent's education. The focus is on the following interactions:
● Initial social position×Number of children ever had, Belgium.
● Initial social position×Number of years unemployed or ill, men in both countries.
● Initial social position×Early retirement, men in both countries.
● Number of years in blue-collar work×Early retirement, men in both countries.
● Initial social position×Ever had a paid job, Belgian women.
● Initial social position×Number of years in blue-collar work, all samples.
● Initial social position×Number of years in self-employment/farming, all samples.
Regarding the hypothesis that in Britain, the ‘liberalisation’ of the welfare regime should lead to stronger evidence of cumulative (dis)advantage for younger cohorts, the following interactions for the male sample are examined:
● Cohort×Number of years unemployed or ill.
● Cohort×Number of years in blue-collar work.
● Cohort×Number of years in self-employment/farming.
● Cohort×Early retirement.
An overview of all significant interactions is presented in Table 3. Each interaction was added separately. For British women, there were no significant interactions.
Table 3. Overview of significant interaction effects with initial social position and birth cohort
Notes: Ref.: reference category. Dependent variable: log(equivalised household income +1). All interactions added separately to the main-effects models reported in Table 2. Estimates for the variables not involved in the interactions are not reported.
Significance levels: * p<0.05, ** p<0.01, *** p<0.001.
Starting with Belgian men, a first significant interaction concerns the scarring effect caused by unemployment or illness. Whilst the main effect for this variable reported in Table 2 was not significant, the results show that, in line with a process of status–resource interaction, each additional year of unemployment or illness results in a significantly lower income in old age for lower-educated men compared to higher-educated men. Although for British men the interaction reaches statistical significance, the effect is in the opposite direction: each additional year of unemployment or illness has a significantly stronger depressing impact on old-age income for higher-educated respondents than for respondents with less education. This effect can perhaps be explained by the flat-rate nature of the benefit system in the UK. Job loss of higher-educated employees is often compensated through lump-sum pay-offs at the time of unemployment, which might not be reflected adequately in the dependent variable, the disposable household income in old age. Furthermore, higher-educated respondents have more to lose from employment interruptions compared to lower-educated respondents, as the latter group is less covered by occupational pensions. The opposite interaction term in both welfare regimes can hence be related to differences in benefit systems and in the ‘pension mix’.
A second significant interaction for Belgian men concerns the experience with blue-collar work and early retirement. Table 3 shows that with each additional year of blue-collar experience, the effect of early retirement on old-age income becomes more positive (recall that the main effect of early retirement was not significant). Removing the ‘duration of labour market career’ variable from the model hardly changes the estimates (not reported). Given the fact that early retirement in Belgium is often part of a negotiated pact between employers, unions and the government, the comparative advantage of blue-collar workers with longer job tenures, in particular of those working in large companies with a strong union representation, is visible in old-age retirement income.
Contradictory to expectations, and although the main effects (Table 2) are in the expected direction, a positive interaction between parents' education and number of years in blue-collar work is found: each additional year in blue-collar work results in a higher income in old age for men from a lower social background compared to men from a higher social background. Alternatively, we could say that each additional year in blue-collar work results in a lower old-age income for Belgian men from a higher social background (measured in terms of parents' education) compared to men from a lower social background. This could be due to selection, in the sense that men from a more privileged background who did less well on the labour market fare worse in old age compared to men from a less privileged background whose labour market experiences were more in line with their upbringing. Another tentative explanation is that a tradition of well-paid and skilled blue-collar work used to be ‘passed’ on within status and family groups, while ‘outsiders’ had to start at the bottom of the job ladder. A similar positive interaction between respondent's education and the experience with blue-collar work is found for Belgian women. Again, the main effects (Table 2) are in the expected direction, with a counter-intuitive interaction. Although the sign of the interactions between ‘initial social position’ and experience with blue-collar work is the same for both Belgian subsamples, given the sample sizes it cannot be ruled out that they are perhaps reflecting a peculiarity in the data.
Concerning the interactions with birth cohort for the British male sample, only the interaction with early retirement reaches statistical significance. The effect is as expected, as the experience of early retirement results in a comparatively lower old-age income for younger birth cohorts. This is consistent with the hypothesis that in the UK, the liberalisation of the welfare regime over time has resulted in a stronger negative effect of labour market risks on old-age income for the younger cohorts.
Conclusion and discussion
In this paper, the impact of ‘life-time’ family and labour market experiences on household incomes of older people in Belgium and the UK was analysed. To this end, panel data and life history information from the Panel Study of Belgian Households (1992–2002) and the British Household Panel Survey (1991–2005) were combined.
Drawing on the conceptual framework of cumulative (dis)advantage, two mechanisms consistent with such an underlying process, contributing to income differences in old age, were explored: ‘cumulative exposure’ and ‘status–resource interaction’. While the first refers to the cumulative impact of persistent ‘exposure’ to negative situations or life events, the second mechanism explores whether respondents who started their adult lives from a more disadvantaged position are less able to profit from opportunities that cross their life paths, and consequently have a lower income in old age. Although the focus was on disposable income in old age, rather than on accumulated wealth, empirical evidence for both mechanisms was found.
In the Belgian welfare regime, characterised by a male-breadwinner focus and the stratification of social security programmes along occupational lines, on the one hand, combined with a high level of de-commodification on the other, marital history seems to have a stronger influence on old-age income. In particular, evidence was found that the so-called ‘marriage premium’ for men persists into old age. A further difference between both countries concerns the negative impact of the number of children ever had, which is only relevant for Belgian men and women, and which persists after controlling for respondent's labour market experiences. Unfortunately, it was not possible to examine joint labour market careers of couples or labour market career patterns of women after childbirth. However, we know from historical macro-level data from the OECD that in 1984 (the earliest data point available), in Belgium only 32.0 per cent of women in the age bracket 50–54 (birth cohort 1930–34) were in paid work, compared to 66.2 per cent of British women. This suggests that in Britain, now-elderly women with children returned to the labour market in much larger numbers, which explains the non-effect of number of children ever had on economic wellbeing in old age. By 2009, the gap in terms of labour market participation rates of women aged 50–54 between both countries has more or less closed. Given the increased number of full-time working women in Belgium and the expansion of child care, one would expect that the effect of ‘children ever had’ will decrease for future retiring cohorts. Finally, if we accept the above-mentioned evidence for Belgian men as proof for the existence of a ‘marriage premium’, then this premium only compensates the cost of children to a small extent.
A final interesting result concerns the low incomes of now-single ever-partnered women in Britain, most of whom are widowed rather than divorced. Widowhood is a known correlate of female poverty in the UK, and is linked to a reduction or loss of occupational and individual pensions. Furthermore, given the transition from defined-benefit to defined-contribution occupational pension schemes and other types of personal pensions, Willetts (Reference Willetts2010) expects this gender dimension to become more pronounced. Personal pensions in particular do not require spousal protection by law. In fact, figures of men opting for an annuity that only covers themselves, leaving their widows with no survivors' benefits, amount to 80 per cent.
With the exception of British women, spending more years in blue-collar work or self-employment/farming has a pervasive negative impact on household income in old age, which is a form of ‘cumulative exposure’. An important new conclusion from this paper is that the ‘occupational stratification’ of retirement income turns out to be significantly stronger in Belgium than in Britain. In this particular instance, the welfare state is thus a more efficient stratifying institution than the market, which runs against our intuition that inequality is always higher in liberal welfare regimes. Although the Belgian welfare state reaches a fairly high level of ‘de-commodification’, this was in part achieved by protecting the incomes of the most vulnerable social groups. At the higher rungs of the social ladder, the stratifying power of existing social policies, exemplary for a conservative-corporatist welfare regime, becomes quite obvious.
When it comes to the impact of employment interruptions, we do however find a stronger ‘scarring effect’ in Britain. For Belgian men, starting out from a lower initial position results in a comparative income disadvantage for those who experienced employment interruptions. This is consistent with a welfare regime that is more aimed at income replacement than poverty prevention: higher wages result in higher unemployment benefits. The focus on poverty prevention in the UK, encouraging private provisions for the highly educated and a speedy re-employment, is in line with the finding that the latter actually suffer comparatively more from employment interruptions compared to their lower-educated counterparts. These scarring effects can be considered as evidence consistent with the hypothesised process of status–resource interaction.
Some significant interactions were in the opposite direction, pointing at a ‘lifecourse mechanism’ that warrants further research: men from a more privileged background who did less well on the labour market fare worse in old age compared to their counterparts from a less privileged background whose labour market experiences are more in line with their blue-collar upbringing. A similar interaction was found for Belgian women. Given small sample sizes, it cannot be ruled out that these effects are merely capturing random variation.
Concerning the hypothesised interactions with birth cohort, evidence is found that for the younger cohorts in the UK, who spent a larger part of their adult lifecourse under a liberalising welfare regime, labour market events (for men) have a stronger negative effect on old-age income than for the older cohorts. The only significant interaction is between birth cohort and early retirement.
Finally, note should be made of the limitations of the research. Although the data more or less map complete family and labour market experiences, they are, partly because of their panel nature, rather crude and might pick up only a small part of the ‘real’ lifecourse effects on income in old age. Also, the fact that it was not possible to include lifecourse information for respondent's partner, and the timing of important lifecourse events, is a serious limitation. A second problem concerns the dependent variable. Although, on the one hand, the income concept used should include income from assets, the focus was not on accumulated wealth across the life course. On the other hand, given the different pension mix in Belgium and the UK, disposable income might be an equally valid measure of economic wellbeing in old age.
Acknowledgements
The author thanks Dr Daniela Grunow for her input concerning some substantial interpretations, as well as the two anonymous referees for their comments and suggestions.
