Why are women more likely than men to extend paid work? The impact of work–family life history

Abstract

Extending working life beyond the state pension age is a key European Union policy. In the UK, women are more likely to extend paid work than men, indicating that factors other than the state pension age play a role in working longer. Women are less able to build pension income due to their role as carer within the family. It, therefore, follows that gender inequalities over the life course continue into older age to influence need, capacity and desire to undertake paid work after state pension age. This paper explores how work, marital and fertility history impact upon the likelihood of extending employment. It uses the British Household Panel Survey’s retrospective data from the first 14 waves to summarise work–family histories, and logistic regression to understand the impact of work and family histories on extending paid work. Findings show that, on the one hand, women are extending paid work for financial reasons to make up for ‘opportunity costs’ as a result of their caring role within the family, with short breaks due to caring, lengthy marriages, divorcing and remaining single with children all being important. Yet, there is also evidence of ‘status maintenance’ from working life, with the women most likely to extend paid work, also those with the highest work orientation, prior to state pension age. But lengthy dis-attachment (due to caring) from the labour market makes extending working life more difficult. This has implications for policy strategies to entice women into paid work to make up for low independent financial resources.

Introduction

Extending working life, and enabling people to work beyond state pension age (pension age hereafter), is a key European Union (EU) policy, being a response to population ageing and the pressure that this places upon pension systems. For policy to successfully promote extended working, it is important to understand why people work longer. In the UK, that the average age of retirement is currently above pension age for women (62.4) and beneath it for men (64.5)1 (Office for National Statistics 2011) indicates that factors other than the pension age play a role. We do know that tenure, ethnicity, caring status, health status, partner’s working status, regional unemployment levels and financial position are associated with working after pension age, regardless of gender (Smeaton and McKay 2003; Humphrey et al. 2003; Sainsbury et al. 2006; Barnes et al. 2004; Phillipson and Smith 2005). However, there has been little UK research focusing on the factors associated with working beyond pension age, and why women are more likely to extend paid work than men.

Decisions surrounding retirement are based upon current circumstances, and long-term work and family histories (O’Rand et al. 1992). Theories of ‘cumulative stratification’ propose that accumulated unequal allocation of opportunities and resources over the life course result in increasing variation in financial resources and decisions surrounding retirement (e.g. O’Rand and Henretta 1999; Raymo et al. 2010). That women are more likely to work beyond pension age may, therefore, be explained by gender differentials in work–family histories. However, women are more likely to have heterogeneous work histories, reflecting variations in marital and care histories, and thus the reasons why women extend working years are likely to vary (O’Rand 1996; O’Rand and Henretta 1999).

There is UK evidence that women’s role as carer within the male breadwinning/female carer model of the family leads to broken work histories, part-time work and low pay. This in turn leads to limited capacity to build up an independent state, private or occupational pension (Bardasi and Jenkins 2002; Department for Work and Pensions 2005; Arber and Ginn 1991, Rake 2007; Ginn 2003; Evandrou and Glaser 2003; Sefton 2011), meaning women have lower individual incomes in older age (Bardasi and Jenkins 2002). International literature has demonstrated that these ‘opportunity costs’ result in women extending paid work for economic reasons (Pienta et al. 1994). Career breaks during childbearing years and a history of part-time employment have been associated with retiring later due to lower levels of pension wealth (e.g. Yabiku 2000; Pienta 1999). Also, that German mothers retire later than their childless counterparts is likely to be due to financial reasons (Hank 2004). Therefore, that women are more likely than men to extend paid work may be explained by the gender role difference in family life, and the impact this has upon work histories and pension accumulation.

The extent that women can make up these ‘opportunity costs’ via sharing their (ex)partners pension income if their individual entitlement is low differs across marital groups, with widowers fairing better than divorcees (Ginn 2003). There is evidence that unmarried, and specifically divorced, women (but not men) remain in the workforce for longer, whilst married women retire early (e.g. McDonald 1996; Smeaton and McKay 2003). Pension sharing may be a more feasible, and desirable, option than working longer for married women, with a reduced work history. For divorcees, extending working years may be the only option to increase retirement income, but also a potentially more desirable one for social reasons (Smeaton and McKay 2003).

Another hypothesis for why women extend paid work is the ‘attachment theory’ (Pienta et al. 1994): women with strong attachment to the labour market during working life will maintain this status in later years by extending, whilst those with weaker attachment will be less likely to work longer (Pienta et al. 1994). There is evidence that women, with more stable labour force attachment, fewer disruptions due to caring (Pienta et al. 1994) and who remain in employment during childbearing years (Henretta et al. 1993; Pienta 1999; Pienta et al. 1994) are more likely to work in old age. In addition, women who start families relatively late, potentially with more established careers (Hank 2004), are more likely to work longer (Pienta et al. 1994). Thus, whilst mothers generally have a financial need to work longer, it appears that those remaining attached to the labour market during childbearing years actually do. There appears to be ‘status maintenance’ in old age (Hardy 1991), with work orientated women with established careers most likely to extend working years.

Our knowledge about the impact of work–family history on extending paid work is mainly from international, and especially North American, rather than UK, literature. This will not necessarily translate into the UK policy and labour market context (Loretto and Vickerstaff 2013). This paper, therefore, seeks to understand how marital history and work/caring history are important for working beyond pension age in the UK. Specifically, it will explore two hypotheses: (1) that women work longer to make up for ‘opportunity costs’ as a result of their caring role within the family and (2) the ‘status maintenance’ theory that women more attached to the labour market during working life continue this by extending working beyond pension age.

Method

To do this secondary longitudinal data analysis is undertaken using retrospective life history data for the first 14 waves of the British Household Panel Survey (BHPS). The data were obtained from the BHPS’s retrospective employment, marital and fertility history files. Retrospective self-reported employment status data have been collated since leaving full-time education (Halpin 1997, 2000). The family history data are contained in a separate dataset, which includes dates and current status of marriages, number of natural children and the date they were born. Work, marital and fertility histories were summarised from the two datasets in a similar manner to Sefton et al. (2008) and Sefton (2011).

To be included in the sample, individuals must have complete work and family histories between the ages of 20 and pension age (60 for women and 65 for men); have non-missing information on whether they were in paid work after pension age, be aged over pension age at some point during the panel (1991–2004) and have non-missing personal income data from at least one of the panel years. The sample was trimmed to exclude observations with very low or very high income data. As individuals are observed at multiple points in time incomes are adjusted to May 2010 prices according to the retail price index.

Individuals can be observed up to 14 times during the panel period, depending on how many years they appear in the survey. Whilst work/family life histories remain the same over the panel, other factors may change after pension age, which may impact upon a decision to work. So this information is not lost, all observations of the same individual are included in the sample. The data were weighted to allow for multiple observations of the same individual. The weight used is equal to 1/n, where n is the number of times each individual appears in the dataset to ensure that equal weight is given to each person. The total sample is of 21,773 observations of 2,680 individuals, 7,684 observations of 997 men and 14,089 observations of 1,683 women.

To examine the relationship between work–family life history and employment in late life, it is important to control for other factors that may be correlated with both. The variables controlled for were both background variables and post-pension age controls:

• Birth cohort and years since reaching pension age. These account for the different pension systems and rules, which may influence working longer.

• Socio-demographic characteristics (sex, marital status and tenure);

• Partners’ work status (if a partner is present, whether or not he/she is currently in paid work);

• Health status (limiting disability, income from disability living allowance, income from attendance allowance and health over the last year);

• Access to job opportunities (access to a car, region and educational qualifications);

• Financial resources (total personal monthly non-labour income, whether receives personal income from private pension, personal income from occupational pension, personal income from investments or savings, personal incomes from income support and future financial expectations).

Gross individual income was used for the analysis.2 There is an argument that household income is a better reflection of material living standards—married women, who can share their partner’s pension, appear better off using the household measure (Ginn et al. 2001; Ginn 2003). However, household measures assume equal distribution of resources within the household. Individual income is a direct measure of personal wealth, and so using this measure demonstrates the difference between men and women’s incomes in old age (Sefton et al. 2008).

Binary logistic regression was used to examine how work–family life history influence the likelihood of extending paid work, whilst holding other factors constant. To assess how important work and family history was in predicting the odds of extending paid work, separate regressions were run for each way of categorising work and family histories in order to explore the extent that each significantly impacted upon extending, after controlling for other factors. Separate models were fitted according to gender.

Working after pension age

Given that retirement has become a more gradual process, working longer has taken on different meanings. It can be partial, include ‘bridge jobs’ from career into retirement, re-employment after retirement, spells of retirement within a career, as well as continuing in the same job undertaken prior to pension age (Brown 1998; McDonald 1996; McDonald and Donahue 2011). Indeed, there are differential characteristics of people following different retirement patterns (McDonald 1996; McDonald and Donahue 2011). Thus, whilst work–family life history may impact upon working after pension age, whatever form it takes, different work–family histories will probably result in different paths to retirement. Due to space limitations, however, this paper does not explore the impact of work–family history upon post-SPA employment patterns. Thus, older people were defined as ‘extenders’ if they undertook paid work for at least a month3 at any point after pension age, whatever the form. ‘Non-extenders’ include those who retired at or before pension age, and undertook no paid work beyond pension age up to the most recent observed wave (prior to 2004). The data showed that women (34.1 %) were more likely than men (23.1 %) to have been ‘extenders’. Women however, on average, worked for slightly longer, with a mean of 3.11 years for men and 4.05 for women, and median of 1.33 years for men and for 3.08 years for women. But, the mode for men was 0.92 years spent in paid work, and only 0.08 years for women. So, whilst the majority worked for under a year after pension age, a few extended work for many years. Women were also more likely than men to work for a continuous full year, with 8.5 % men and 17.3 % women extending for a full year. Are these gender differentials explained by work history?

Work history

Duration of employment prior to pension age is likely to increase propensity to extend paid work—either as a result of economic incentive to make up for career interruptions or due to greater labour market attachment in working life continuing into old age. The odds of extending paid work according to duration of employment prior to pension age is shown in Table 1. In the study, men could work for a possible 45 years before pension age, whilst women could work for 40 years—reflecting the difference in pension age. For women but not men there is a statistically significant association between duration of employment and extending working years, after accounting for socio-economic factors. For women, spending more years in paid work during working life significantly increases the odds of working longer, relative to those employed for less than 25 years. It appears that women who spend many years out of the labour market have fewer opportunities to make up for it in old age, whilst those with more work experience are in a better position to do so. But if labour market attachment was the only explanatory factor, we might expect men with more work experience to extend paid work, but there was no significant association. Thus longer duration impacts differently for men and women. One explanatory factor may be gender differences in contract of employment.

Table 1.

Logistic regression of the odds beyond state pension age (any) by number of years in employment for men and women, after controls

	Men odds ratios	Women odds ratios
Number of years in employment
Less than 25 years	Ref	Ref
25–30 years	0.466	2.804***
30–35 years	0.190	2.892***
35–40 years	0.231	3.400***
40–45 years	3.952	–
Log likelihood	636.588	1,494.242
Nagelkerke R 2	0.333	0.145
N observations	6,096	10,977

* p < 0.05, ** p < 0.01, *** p < 0.001

Contract of employment

Part-time work has historically attracted lower wages and (pension) benefits. So working part time for much of ones life may influence propensity to extend paid work (Sefton et al. 2008; Sefton 2011). Table 2 shows the impact that the proportion of employed life spent in full- and part-time employment has upon the odds of extending working life. For women, working full time for all of their employed life does not significantly impact upon working longer whilst men doing the same are significantly less likely to. However, this does not account for number of years worked. So someone working 100 % of his/her working life in full-time employment may still have only worked for a few years overall. If we control for duration of employment, we can explore how important full-time employment is, regardless of the length of labour market attachment. Now women as well as men are significantly less likely to extend paid work if working full time for all of their working life. Nevertheless, for women working full time for less than 75 % of their working life is still related to increased odds of extending paid work. Thus, for women, it appears that both duration of employment and employment contract impact upon propensity to extend working life.

Table 2.

Logistic regression of the odds of working beyond state pension age (any) by type of employment for men and women, after socio-economic controls, and controlling for years in employment (in brackets)

	Men odds ratios	Women odds ratios
Proportion of working life employed full time
Never full time	Ref	Ref
Under 25 %	0.371 (0.331)	2.924*** (1.994**)
25–0.50 %	0.284* (0.231*)	2.315** (1.692*)
0.50–0.75 %	0.405 (0.317)	2.491*** (2.009**)
0.75–0.99 %	0.243** (0.243**)	1.738* (1.238)
All working life	0.154*** (0.151**)	0.650 (0.488**)
Log likelihood	706.759 (622.683)	1,492.174 (1,427.600)
Nagelkerke R 2	0.219 (0.354)	0.154 (0.208)
N observations	6,096 (6,096)	11,029 (1,0977)
Proportion of working life employed part time
Never part time	Ref	Ref
0–25 %	3.931*** (3.270***)	2.682*** (2.453***)
25–0.50 %	10.629 (8.320)	3.181*** (3.202***)
0.50–0.75 %	7.195 (12.717)	3.610*** (3.402***)
0.75–0.99 %	–	3.821*** (3.334***)
All working life	–	3.807*** (4.162***)
Log likelihood	700.930 (620.236)	1,491.421 (1,426.284)
Nagelkerke R 2	0.229 (0.358)	0.155 (0.209)
N observations	6,096 (6,096)	11,029 (10,977)

* p < 0.05, ** p < 0.01, *** p < 0.001

() Controlling for years employed

For both men and women, working part time increases the likelihood of extending paid work, holding socio-economic factors constant (Table 2). This indicates that this group attempt to make up for the reduced pay and pension benefits associated with part-time work even if their personal income is high. So, that women in our sample spent 31.36 % of their working lives in part-time work compared to only 0.07 % of men goes some way to explain why women are more likely to extend working life than men. Nevertheless, for those who do it, working part time impacts more for men than women, even after controlling for duration of employment. Men employed 25 % or less of working life in part-time employment were 3.270 times more likely than those never working part time to extend paid employment. This is compared to an odds ratio of 2.453 for women. Men working part time are, on average, paid less than women working part time (Hicks and Thomas 2009) but also, that men have been unable to claim derived benefits may mean that men without full work histories are penalized more heavily for this than are women.

Inactivity

Next, we examine the impact that periods of inactivity have upon extending working life. Table 3 shows that long durations of inactivity decrease the odds of working beyond pension age compared to experiencing inactivity of less than 6 months, for both men and women. That this holds even after income, and other factors, are controlled for, indicates that even those with very low incomes are less likely to extend paid work if they have experienced long periods of inactivity. Whilst for men short periods of inactivity are more significant in reducing the propensity to extend employment, for women very long periods of inactivity (15 years or more) are most important. This may be explained by the different types of inactivity undertaken between the genders with further analysis, not presented, showing that men in our sample were more likely to be inactive due to unemployment/sick (31.5 % compared to 15.0 % of women), and women more likely to have cared for the family—82.3 % versus 0.5 % of men (29.2 % for 20 years or more).

Table 3.

Logistic regression of working beyond state pension age by number of years inactive, after controls

	Men odds ratios	Women odds ratios
Number of years inactive
Less than 6 months	Ref	Ref
6 months to 2 years	0.208***	0.442*
2–15 years	0.053***	0.489**
15 or more years	0.119**	0.154***
Log likelihood	574.296	1,476.571
Nagelkerke R 2	0.425	0.169
N observations	6,096	11,003

* p < 0.05, ** p < 0.01, *** p < 0.001

The impact of different types of inactivity upon extending paid work is shown in Table 4. Being unemployed/sick for more than 2 years compared to never being unemployed/sick significantly reduces the chances of working after pension age for men, but not women. Given that men are more likely to experience unemployment, this partly explains why they are less likely than women to extend working life.

Table 4.

Logistic regression of working beyond state pension age by type of inactivity, after controls

	Men odds ratios	Women odds ratios
Did not retire early	Ref	Ref
Retired early <2 years	0.092***	0.069***
Retired early >2 years	0.039***	0.045***
Was not unemployed/sick	Ref	Ref
Unemployed or disabled <2 years	1.155	1.901*
2+ years	0.107***	0.506
Never other inactive	Ref	Ref
<2 years	1.555	0.708
2+ years	0.426	1.754
Log likelihood	520.484	1,246.969
Nagelkerke R 2	0.473	0.363
N observations	6,059	11,003
No family carea	–	Ref
<5 years	–	2.215**
5–10 years	–	1.863*
10–20 years	–	1.474
20+ years	–	0.291***
Log likelihood	–	1,144.865
Nagelkerke R 2	–	0.447
N observations	–	11,003

– Numbers to small for analysis

* p < 0.05, ** p < 0.01, *** p < 0.001

^a Controlling for all other types of inactivity

Very few men had undertaken family care and thus it was not possible to explore the gender differences in terms of the impact this kind of inactivity has upon extending paid work. For women, family care is important for working longer, but the relationship is not straightforward. Those with a shorter period of time undertaking family care (under 10 years) have higher odds of extending paid work compared to those never undertaking family care, with those caring for less than 5 years more than twice as likely to work longer. However, those undertaking family care for more than 20 years are significantly less likely to be working longer. Those with a longer break may find derived pension benefits a more attractive option than trying to make up many years out of the labour market.4 Whether this is a realistic option will depend upon marital history.

Marital history

For this generation, marriage brought with it a breadwinning role for men and a wifely/caring role and reduced time spent in the labour market for women: Further analysis (not presented) shows that men who marry and remain so for the duration of their working life work on average for the longest number of years in full-time employment relative to other men, but women with the same marital history work the shortest duration. Table 5 shows that marital history has a more significant impact upon working longer for women than men, but it also impacts differently between the sexes. For men, ever being married significantly decreases the chances of working after pension age compared to never being married. For women, however, ever being married increases the likelihood of working beyond pension age (by 3.456). Indeed, for women, even shorter periods of marriage (between 25 and 50 % of their working life) significantly increases the propensity to work longer (odds ratio = 4.031). The odds are also significantly increased regardless of the age women married. For women, the periods of labour market inactivity associated with marriage are thus translated into extending paid work to make up for this.5 Men’s ‘breadwinning’ role is important for lowering their propensity to work longer, although they have to be married for at least half their working life for marriage to make a significant reduction in the odds of extending paid work. Once married, men participate in paid work and additional pension schemes as part of their ‘breadwinning’ role (Price and Ginn 2003), reducing the need to extend working life.

Table 5.

Logistic regressions for the odds of working beyond state pension age by marital history for men and women (after controls)

	Men odds ratios (odds	Women odds ratios
Life marital status
Never married	Ref	Ref
Ever married	0.167*	3.456**
Never divorced	Ref	Ref
Ever divorced	1.106	1.226
Never widowed	Ref	Ref
Ever widowed	0.937	0.980
Log likelihood	698.493	1,553.600
Nagelkerke R 2	0.191	0.084
N observations	6,096	11,029
Proportion of years married
Never married	Ref	Ref
Married for under 25 %	0.129	2.643
Married 25–50 %	0.185	4.031*
50–75 %	0.159*	3.581*
75–100 %	0.153*	3.361*
Log likelihood	697.982	1,554.365
Nagelkerke R 2	0.192	0.083
N observations	6,096	11,029
Timing of marriage
Never married	Ref	Ref
Early 20 s	0.145*	3.477
Later 20 s	0.198*	3.701*
Over 30	0.180*	3.419*
Log likelihood	696.676	1,554.365
Nagelkerke R 2	0.194	0.083
N observations	6,096	11,029
Pattern of marriage
Never married	Ref	Ref
Divorced, remarried	0.154*	3.993*
Widowed early, remarried	0.140*	2.256
Divorced, stayed single	0.269	5.038**
Widowed early, stayed single	0.182	4.121**
Ever married, stayed married	0.167*	3.451*
Log likelihood	697.718	1,549.971
Nagelkerke R 2	0.193	0.088
N observations	6,096	11,029

* p < 0.05, ** p < 0.01, *** p < 0.001

For both men and women ever being divorced or widowed does not significantly impact upon working longer. What does seem to matter, however, is marital status post divorce (see Table 5). For men, remarrying post divorce or widowhood has a similar impact as being married, and never divorcing—significantly reducing the odds of extending paid work. This reinforces the finding that for men having a partner during working life is important for a decision (not) to extend paid work.

For women, the picture is more complex. Whilst we have adjusted for personal income, this does not account for the experience of losing a spousal income, and the derived pension rights associated with this. This will effect marital groups differently, which, coupled with variable work history, may impact upon propensity to work longer. Table 5 shows that women staying single after divorce are most likely to extend employment—being five times more likely than never married women. Those widowed early and staying single have lower odds of extending paid work than those divorcing and staying single, but higher than other marital groups (odds ratio = 4.121). This presumably is because, whilst, unlike divorcees, they are entitled to derived pension rights from their previous husband, they still have experienced broken work histories and no shared spousal income in old age. Divorcees who remarry and women who stay married, and never divorce, are the least likely to extend working life of all marital groups—which is likely to be because they will have shared income in old age, despite broken work history. Unlike other marital groups, women who widow early and remarry are not significantly more likely to extend paid work relative to those never married. Not only does this group have a more full work history relative to other marital groups (except never married), but loss of derived pension rights from their previous partner due to re-marrying may cancel out any gain from derived pension rights/shared income gained from their new partners. Thus, for women, as well as work history, derived benefit rights appear to influence a decision to extend employment.

Fertility history

Given the gender division of labour, fertility history may have more impact upon extending work for women than men. Analysis undertaken shows that children impact upon work history for women, but not for men: the more children women have, the fewer years employed (full time), on average. For men, number of children has no impact upon extending employment. For women, children have some impact but the impact is not as might be expected, given the impact of children on work history. The odds are increased significantly for those with one (odds ratio = 1.601), two (odds ratio = 1.885) and four children (odds ratio = 1.875), although the odds ratios for one and four children were only just significant whilst that for two children was significant at the 99 % level. It may be that mothers with two children, on the one hand, did not work enough years in (full time) employment to mean that they were financially stable at retirement but at the same time had remained attached to the labour market during childbearing years, gaining enough work experience to enable them to make up for years lost by working longer. Having three children or very large families (five or more children) did not significantly increase the odds of working longer. Generally, the picture is unclear, and it appears that number of children is not a good predictor of propensity to extend paid work.

In terms of timing of children, the age women have their first child does not impact upon the odds of extending working life. But timing of family completion (age when last child was born) does. Women with no children (odds ratio = 0.567), and those completing their family in their early 20s (odds ratio = 0.525) are significantly (at the 99 % level) less likely to extend working life than those completing in their late 30s. Despite low educational qualifications, women who have completed a family early, like those with no children, also spend the longest in full-time employment, with the associated lower odds of working longer.

Marital and fertility history

It is important to examine how marital and fertility history work in combination to impact working beyond pension age. For women, having children might exacerbate the impact that marital history has upon extending paid work. We examine the impact of various marital patterns for those with and without children.

For men without children, being married, after divorce is just significant—with lower odds of extending paid work (odds ratio = 0.054)—but having children makes no significant impact. Women with children who have either remained married (odds ratio = 3.636) or stayed single after divorce (odds ratio = 4.753) are significantly more likely to extend paid work, but not those without. This is likely to reflect the difficulties in juggling paid work and care, especially as a single parent, with mothers working longer to make up for this.

Conclusion

This paper has shown that work–family histories, and cumulative stratification throughout working life, are important in determining who works beyond pension age. In line with international evidence (e.g. Yabiku 2000; Pienta 1999), we have found that UK women extend working life to make up for ‘opportunity costs’ during working life. Women’s tendency towards part-time work, and the financial sacrifice associated with this, increases their propensity to extend paid work. But full-time work does not protect UK women, unless undertaken for all their working life. This is partly a reflection of the gender gap in earnings (DWP 2005), but also that women working full time take breaks from the labour market to care, and extend working life to make up for this. The ‘opportunity costs’ theory also seems to be the most likely explanation for why divorced mothers (but not fathers) remaining single are significantly more likely to work longer. If they were extending paid work for social reasons (Smeaton and McKay 2003), we would expect male divorcees and those female divorcees without children who remain single also to be more likely to extend working life. But this is not the case. Thus the triple disadvantage of broken work history (even post divorce), lack of derived pension benefits and no partner’s income to sustain them in old age pressures single divorced mothers to work longer.

We might expect that married mothers are more likely to make up for ‘opportunity costs’ by pension sharing rather than extending employment. Indeed, other studies have shown partners to influence retirement decisions (Loretto and Vickerstaff 2013; McDonald 1996). However, our study has shown that women remaining married with children (but not those without) are more likely to extend paid work than never married women. This suggests that married mothers’ work histories and the detrimental impact of the carer role matter in an explanation for why they extend paid work.

This paper also provides some evidence of the ‘status maintenance’ theory (Hardy 1991), with longer periods of employment in working life continuing into old age. Indeed, this study supports the international literature (e.g. Pienta et al. 1994) that lengthy labour market attachment and shorter childcare breaks increases the ability to extend employment. It is likely that these women are more work-orientated, with more rewarding careers, and thus have better negotiating power within the labour market, as well as greater desire to work longer.

But, given the measure used to identify ‘extenders’, it may be that this study has captured women returning to work after an initial early retirement. McDonald (1996, 1997) found that married women, more highly educated women, those in professional occupations and those with adequate incomes were more likely to retire early and then return to the labour market in later years.6 These characteristics are likely to be indirect measures for advantageous attachment in the labour force observed in our study. Thus married women in an advantageous position may retire early, influenced by their partners in that decision, with the benefits accumulated through marriage making earlier retirement more appealing (McDonald 1996). At the same time, their relatively advantageous position in the labour market and perhaps a strong work ethic may enable this group to return to work at a later date to make up for low personal independent income accrued due to caring responsibilities. But returning to work may also be a desirable preference for this group (McDonald 1997). More research is needed to understand the impact of work–family history on different patterns of working longer, especially for married mothers.

Of greater concern, however, is that those with lower attachment to the labour market in working life maintain this status in late life through decreased propensity to extend working life. In particular, women with long years out of the labour market to care are associated with lower odds of extending paid work. It is likely that derived benefit rights are a more attractive and a potentially financially more rewarding option for this group than trying to negotiate paid work. Nevertheless, this assumes the presence of a partner with a strong work history, and a good pension. Even fulfilling this requirement does not negate that these women will have little or no independent personal income in old age, with the associated risks. So, women potentially in most financial need to extend employment are not necessarily doing so. These groups remain disadvantaged after pension age by a continuation of social roles, and the associated inequities in accessing the labour market (Dannefer 2003).

This paper has shown that women are extending working life for financial reasons, to make up for labour market inequalities as a result of their caring role within the family. Yet, the women seemingly most likely to extend paid work are also those with the highest work orientation, with established careers. This has implications for policy strategies to entice women into paid work to make up for low independent financial resources. Those without a history of attachment either do not want to extend paid work or are not able to. This needs to be unpicked. This paper has not studied the impact of occupational history nor controlled for structural unemployment, both likely to impact upon ability to extend employment (see e.g. Hayward et al. 1994). In addition, more needs to be understood about how couple’s work histories inter-relate to impact upon decisions to work longer.

Arguably, policy changes introduced too late to impact the cohort studied in this paper will influence which women work longer among more recent and future cohorts: carers are compensated more comprehensively via home responsibilities protection allowance (see footnote 4) and a reduction in the number of years required to qualify for a full pension to 30 years has been introduced. Also, the pension age for men and women is being equalised to 66 by 2020 (HM Government 2013), which may have implications for those in more disadvantaged positions—unable to work longer and having to wait longer for any state pension to which they are entitled (McDonald 1997). In any case, the majority of women extending paid work in this study did not work much beyond the current pension age. Given the associations between work–family history and working beyond pension age, increasing the pension age is unlikely to enable those to work longer who find it difficult to do so. At the same time, regardless of the pension age, women are still more likely than men to work part time, to be paid less, and to take breaks from the labour market to care for children. Until gender inequalities in the labour market are addressed directly, some women are likely to need to work longer to make up for them. The state needs to make sure that women who need to, can.