Exploring the private pension gender gap and occupation in later working life

Abstract

This article investigates the gender gap in private pension (PP) membership and wealth across different occupations among a cohort of employees using data from the English Longitudinal Study of Ageing. Using a Heckman selection model to correct for selection bias the results show that gender has a stronger effect than occupation on PP membership and that it is female employees’ lower rate of PP membership that has the greatest impact on their ability to accumulate PP wealth, rather than their ability to save once a member. The size of the gender gap in PP wealth is also conditioned by occupation. Analysis of the interaction of these two variables provides new insights into the heterogeneity of women’s private pension experience and the emergence of a ‘privileged pole’ among professional women.

Keywords

gender occupation private pension wealth

Introduction

Workplace pension reforms have been implemented in the UK since 2012 in response to concerns about the adequacy of future retirement incomes, particularly for employees on moderate to low earnings. These reforms are intended to increase the number of people saving for retirement in private pensions (PPs) by requiring employers to enrol employees into a workplace pension with minimum contributions from employees and the employer. According to the Department for Work and Pensions (DWP), this should go some way to tackling the gender gap in private pensions: ‘the reforms will be a genuine equality of opportunity in the access of private pension saving, enabling more women to take responsibility for their own income in retirement’ (DWP, 2011: 4).

It is timely therefore to revisit the gender gap in PPs. Compared with the extensive body of research on the gender pay gap in the UK, there is a limited literature on the gender gap in PP membership (Ginn, 2003b; Ginn and Arber, 1993, 1999), or on gender inequalities in PP income in retirement (Bardasi and Jenkins, 2010; Bozio et al., 2011; Ginn and Arber, 1991, 1996, 1999) or on PP wealth among the working population as a whole (Warren, 2003, 2006; Warren et al., 2001). Similarly it is unclear how contemporary changes in the working lives of women have affected PP membership and wealth. Increased education and greater continuity of women’s labour force participation might be expected to result in a reduction in the gender PP wealth gap (DWP, 2005). Other socioeconomic trends, however, may have sustained this gap. Women have retained the bulk of responsibility for family care and much of the growth in female labour force participation has been in part-time employment (Sefton et al., 2011). In addition, there is differentiation in the occupational pension scheme arrangements that men and women are able to access, because they tend to work in different occupations (McKay et al., 2000). Occupational gender segregation (OGS), the separation of men and women across occupations (Blackburn and Jarman, 1997), is considered a major source of inferior occupational rewards for women (Acker, 1990; Standing, 1999), including PPs (Ginn and Arber, 1993). The link between rewards and OGS may be explained by discriminatory demand-side factors or by supply-side occupational preferences associated with family roles (Stier and Yaish, 2014). Set against this, there is growing recognition that OGS may have contradictory and diverse effects in relation to gender equality (Jarman et al., 2012) and there is increased awareness of the need to explore intersections with gender, ethnicity and class (Bradley, 1999). Higher OGS may be an ‘advantage’ to women in that they are more likely to attain senior positions where male competition is weak (Jarman et al., 2012). Moreover, gender differences in the labour market experiences of some occupational groupings are diminishing, resulting in some polarization of earnings and assets among women (Warren, 2003).

It therefore remains unclear how the increasing diversity of women’s labour market experiences is affecting the gender PP wealth gap. Using data from the English Longitudinal Study of Ageing, this article explores the heterogeneity of women’s PP experience and the variation of gender disadvantage across different occupations among older workers. It utilizes a methodology that illustrates the extent to which the PP wealth gender gap is due to inequalities in pension participation and/or the capacity of participating women to accrue pension wealth.

The next section sets out the background to PPs in the UK and reviews the literature on the interplay between gender and occupation in shaping PP membership and wealth. A further section outlines the data and methodology, and then the results are presented. The article concludes with a discussion of the findings.

Private pensions in the UK and the gender gap

The UK operates a complex pensions system combining both public and private provision – for a fuller discussion, see Pensions Policy Institute (2012). PPs are voluntary non-state pensions which are intended to supplement the basic state pension. PPs are not directly funded by the state but receive state subsidies through tax relief.

PPs take two forms, either occupational schemes or personal pension schemes. Occupational schemes, including public sector pensions, are based on pension contributions made by employers and employees, categorized as defined benefit (DB) or defined contribution (DC) schemes. With DB, the level of benefit received upon retirement is defined by a formula linked to final or average salary and guaranteed by the collective pension fund. With DC, the contributions are paid into individual accounts and the level of benefit upon retirement is not determined in advance but depends on the value of the accumulated fund which has to be translated into an income through the purchase of an annuity. Personal pension schemes are also based on the purchase of an annuity but rely solely on the contributions of the scheme holder, not the employer.

Incentives for the growth of PPs have been a feature of the British welfare state since the 1940s (Foster, 2011; Ginn, 2003a). From the 1980s, PPs have been central to the strategy for reducing dependency on the state in later life (DHSS, 1985; DSS, 1998; DWP, 2006). Indeed, PPs have become an increasingly important source of retirement income for younger cohorts (Banks et al., 2005: 62; McKay et al., 2000) as evidenced by the increasing take-up of PPs and a narrowing of the gender gap in PP membership in birth cohorts aged 35–54 relative to those aged 55–64 (ONS, 2009). Elsewhere in Europe, policy is also shifting towards greater reliance on PPs (Frericks et al., 2009).

PPs remain a major source of gender inequality in retirement incomes in the UK (Banks et al., 2005). With respect to private pension wealth (PPW), women are doubly disadvantaged. They are less likely to participate in PP schemes (Ginn, 2003b; Ginn and Arber, 1999) and those females who participate accumulate less PPW than males (Warren, 2003, 2006; Warren et al., 2001). It is likely that many of the factors identified with gender inequalities in PPW, such as hours worked, will have a dual effect via women’s willingness and ability to join a PP as well as their capacity to build up pension assets once a member. However, most statistical analysis on gender and PPs is moot on this connection, the exception being Bardasi and Jenkins (2010) who reveal that the gender gap across both components rests less on women having different characteristics (education, marital status, fertility and work histories) than men but rather on these characteristics being less well rewarded.

PPs are based on labour market participation rather than citizenship, hence gender inequalities in PPW are likely to be linked to gender differences in employment histories, to hours worked and occupation and to the constraining effects of childbearing and child-rearing on women’s occupational choices and labour market activity (Ginn, 2003b; Ginn and Arber, 1993). Employees who work part-time are much less likely to be covered by occupational pensions than those working full-time (DWP, 2005; McKay et al., 2000). The numbers of years in full-time employment and years employed from age 40 to 55 have a strong impact on women’s entitlement to a PP (Ginn and Arber, 1996). Having worked full-time in their 50s also has a strong beneficial effect on women’s retirement income (Sefton et al., 2011). This effect is probably linked to the penalties incurred in final salary schemes for those who retire or reduce their hours prior to the normal retirement age and also to the increase in women’s PP membership over time (Sefton et al., 2011). Working part-time over 50, often triggered by care responsibilities, has detrimental effects on pension savings (Evandrou and Glaser, 2003).

PPs are also gendered through their link with lifetime earnings, which are reduced by discontinuous employment histories, part-time working and low pay (Ginn and Arber, 2000; O’Connell and Gash, 2003). Women are also more likely than men to be penalized as early leavers from schemes and by the costs associated with transferring pensions between organizations (Blake, 2000).

PPs also provide the highest levels of wealth for those in highly paid occupational groups (Ginn and Arber, 2000; Warren, 2003). Final salary schemes favour those whose pay rises with seniority, a more likely outcome for men in higher non-manual occupations than for women or manual workers. The PPW of women in general has been seen to be adversely affected by gendered occupations as well as gendered labour market participation (Ginn and Arber, 1993). Jobs occupied by women are only half as likely as those occupied by men to be associated with occupational pension scheme membership (McKay et al., 2000: 40), while the best occupational schemes have traditionally been concentrated in male-dominated sectors (Ginn and Arber, 1996).

However, counter to this argument, there is evidence of growing heterogeneity in women’s labour market and pension experience (Bradley, 1999; Ginn and Arber, 1999; Warren, 2003) and of contradictory and diverse outcomes associated with feminized occupations (Jarman et al., 2012). Evidence across a number of industrialized countries reveals that although the pay inequalities of feminization persist, women have more attractive occupations than men, in terms of social status (Jarman et al., 2012). In both these regards, the greater the OGS, the more the position of women is seen to improve, in part because in more feminized occupations women face less competition from men and consequently can rise to more senior positions. Indeed among full-time employees, women are now more likely than men to have an employer’s pension scheme in their workplace, partly because of the concentration of women’s employment in the public sector (DWP, 2005: 96).

Warren’s (2003) descriptive analysis suggests the emergence of a ‘privileged pole’ among female professional and managerial occupations in terms of weekly earnings and individual pension assets compared to women in other occupational groups. This was linked to greater continuity of full-time employment within these occupations, assisted by access to superior maternity leave arrangements and to higher earnings making childcare more affordable (Warren, 2003). This suggests a partial convergence in the labour market experiences of women and men across certain higher-level occupations but a growing disparity with women in ‘lower-level’ occupations that offer poorer career opportunities, wages and pension entitlements. Any study of the gender pension gap should therefore also explore the diverse experiences and actions of women themselves. Building on Warren (2003, 2006) and Bardasi and Jenkins (2010) this article explores the extent to which the gender gap in PPW is conditioned by the role of occupation by considering their impact jointly on PP membership and wealth accumulation. In so doing it aims to offer new insights into the heterogeneity of women’s PP experience and contribute to sociological debates on occupational feminization and gender inequality.

Data and methodology

The data used to analyse the impact of gender and occupation on PP membership and wealth was the English Longitudinal Study of Ageing (ELSA) Wave 1 (Marmot et al., 2013). This is a unique panel survey of individuals aged 50 years and over, looking at the physical, psychological, social and economic aspects of ageing. The first wave of interviews with over 12,000 individuals was carried out between March 2002 and March 2003. Until 2012–13 this was the only wave for which the technically complex and arduous task of calculating PPW data had been carried out (Banks et al., 2005). The analysis concentrated on employees (for whom workplace data was available) who were in the same age cohort, aged 50 to 59 (below the female state pension age of 60). This left a working sample of 2492 individuals split equally between women and men which, with the application of weights, gave a representative sample of people in this cohort. All data where there were missing values were excluded from the analysis.

Heckman selection model

Given the aforementioned discussion of wealth effects, it is evident that the modelling of PPW is problematic since the value of PPW is determined by two factors, namely, one’s ability to join a PP scheme and, once a member, the level of contributions made into the scheme. Since the decisions to join a pension and how much to contribute are related and influenced by unobserved factors, this study’s analysis needed to account for the potential selectivity bias that may arise (see Vance and Buchheim, 2005) because the dependent variable, PPW, is only observed for a restricted, non-random sample, rather than a representative sample of the population. In order to capture this feature of PPW a Heckman selection model was used. This is a maximum likelihood technique designed to eliminate or minimize sample selection bias by jointly estimating a selection and outcome.

S_{i}^{*} = z_{i} γ + u_{i}

(1)

S_{i} = 0 if S_{i}^{*} ⩽ 0 and S_{i} = 1 if S_{i}^{*} > 0

Y_{i} = x_{i} β + ε_{i} if S_{i}^{*} > 0

(2)

Where u_{i} ~ N (0, 1)

ε_{i} ~ N (0, σ)

c o r r (u_{i,} ε_{i}) = ρ

The purpose of the selection equation (1) is to understand the factors that influence the membership of PP schemes, where the dependent variable S_i is an observed dummy variable measure of pension status which is the realization of a latent continuous variable S_i which depicts the probability of having positive pension wealth, Z_i is a vector of the determinants of this status and u_i is the error term.

The outcome equation (2) models the factors that affect the accumulation of PPW where pension wealth is observed, that is where S_i =1. The dependent variable Y_i is the measure of PPW used here, x_i are the explanatory variables and ε_i is an error term. To account for the interdependence of the selection and outcome equations the distribution of the error terms (u_i, ε_i) is assumed to have a bivariate normal distribution with correlation ρ. If the pension decisions are related (ρ ≠ 0) then standard regression of the wealth equation will yield biased and inconsistent estimates. A test of whether ρ ≠ 0 therefore provides a direct test of whether sample selectivity is present or not.

A well known obstacle to implementing the Heckman model is that, for the model to be identified, the selection equation must contain at least one variable that ‘belongs’ in the selection equation but not in the pension wealth equation (Wooldridge, 2013). In order to capture the higher incidence of PPs in larger firms a measure of organizational size is included which, in the absence of a public sector marker, also acts as a proxy for more extensive provision in this sector. Also included are a number of proxy measures that reflect financial awareness (whether individuals have no financial products – current/saving accounts or investment products and a test-based score of numerical ability) alongside a measure of households’ broader financial position (housing ownership and number of vehicles owned) which might affect ability to join a PP.

Care needs to be taken in interpreting the coefficient estimates of the Heckman model, especially when an explanatory variable appears in both the selection and outcome equation (see Vance and Buchheim, 2005). In order to provide a clearer and more insightful interpretation of the results, post-estimation analysis for the two stages of the estimation was carried out. For the selection equation, the marginal effect of being in a PP scheme was calculated for each variable. For the wealth equation, the expected value of PPW, conditional on the dependent variable being observed (i.e. selected), was calculated (Neumayer, 2011). This approach recognizes that where a variable appears in both the selection and the wealth equations, the marginal effect of a variable such as gender has two components. Gender may affect wealth directly by influencing how much wealth those with a PP accumulate. It may also have an indirect effect on the value of wealth through its influence on the probability of having a PP (Greene, 2003). The predicted values of wealth therefore account for both the selection effect and the wealth effect of a particular variable.

Private pension wealth

The PPW variable captured the discounted present value of the stream of income that an individual would receive from their PP(s). It was derived from information on individuals’ current and past circumstances from the Work and Pensions module of ELSA Wave 1, along with various assumptions about past and future behaviour and economic and pensions performance (see Banks et al., 2005 for details about the assumptions made and associated sensitivity analysis).

PPW comes from four main sources: occupational pensions to which individuals are currently contributing; personal pensions to which they are currently contributing; pensions to which they are no longer contributing, but from which they are not yet receiving income; and pensions from which they are already receiving income, including former partners’ pensions received by widowed and divorced people.

These components of the PPW variable were derived in different ways, depending upon whether individuals were already in receipt of their pension income at the time of the interviews. For those who were not already in receipt of their PP(s), data on pension arrangements were used to derive PPW and different methods were used, depending on whether pension schemes were DB or DC (see Banks et al., 2005 for a detailed discussion of the methods used). The chosen measure of PPW was the wealth they would have if they continued to work and accrue pension entitlements until they reached the state pension age, which at the time of the survey was 60 for women and 65 for men.¹

Factors affecting private pension wealth

The core variables in this analysis are gender and occupation. The gender variable is coded to capture the impact of being female on PPW (1/0). The occupational variable follows the Standard Occupation Coding (ONS, 2000) which has the advantage that it classifies jobs in terms of both skill level and skill content, hence for example, distinguishing between managerial and professional occupations.

The specification also includes a range of control variables that might independently affect the value of an individual’s pension as well as influencing the size and significance of the gender and occupational effect. Modelling the gender effect in particular is complicated by the issue of omitted variable bias. For example, while variables such as the number of hours worked per week, number of children and marital status may be expected to have an independent effect on the amount of pension wealth someone can accumulate, they are also likely to affect the size of the gender effect. Women tend to work fewer hours per week and are more likely to miss out on pensionable pay because of motherhood (Ginn, 2003b). Not including these factors within the model will mean that their effect is likely to be captured by the gender variable, thus not giving a true representation of the PPW gender effect. Although omitted variable bias can never be eliminated, the richness of the ELSA data allows a wide range of variables to be controlled for, notably educational qualifications, marital status, number of hours worked per week, ethnicity (white/non-white), age and number of children.²

The core characteristics of the data are set out in Tables 1 and 2. In the sample of employees aged 50–59, there are equal numbers of working women and men. Patterns of occupational gender segregation are apparent. Women are over-represented in the administrative, personal service and sales categories and under-represented among managers, skilled trades and process workers. While women are slightly more likely than men to be in non-manual occupations (55% compared with 51%) these positions are concentrated in lower-level administration and secretariat, rather than in professional and managerial occupations.

Table 1.

Descriptive statistics – gender and occupation.

Variables	Female	Male	All	N
	%	%	%
Gender	50.52	49.48		2492
Occupation
Elementary occupations	16.75	12.11	14.45	356
Process, plant & machine operatives	3.81	13.78	8.75	205
Sales & customer service	10.39	2.50	6.48	168
Personal service	12.31	2.61	7.51	191
Skilled trades	2.18	18.18	10.10	232
Administration & secretariat	26.06	5.48	15.87	415
Associate professional & technical	9.96	9.91	9.94	251
Professional	10.98	15.38	13.15	334
Managers & senior officials	7.57	20.05	13.75	340
Positive private pension wealth
Elementary occupations	45.61	86.96	62.75
Process, plant & machine operatives	49.25	86.59	78.36
Sales & customer service	60.91	93.34	67.09
Personal service	64.75	86.34	68.47
Skilled trades	54.61	92.36	88.25
Administration & secretariat	81.50	100.00	84.66
Associate professional & technical	83.34	94.05	88.62
Professional	86.46	96.79	92.43
Managers & senior officials	80.17	96.61	92.03
All	70.10	92.9	81.38
N	945	1067	2012

Table 2.

Private pension wealth (PPW) by gender and occupation.

	Mean PPW (full sample) in £			Mean PPW (+_ve PPW sample only) in £
Gender
Female	81,094			115,682
Male	154,696			166,525
Occupation	Female	Male	All	Female	Male	All
Elementary occupations	27,333	71,403	45,603	59,928	82,107	72,670
Process, plant & machine operatives	31,704	91,962	78,689	64,378	106,210	100,419
Sales & customer service	33,052	96,438	45,134	54,264	103,319	67,273
Personal service	51,008	75,583	55,237	78,773	87,536	80,675
Skilled trades	44,152	96,516	90,810	80,841	104,500	102,905
Administration & secretariat	76,333	194,202	96,469	93,665	194,203	113,952
Associate professional & technical	123,261	194,372	158,352	147,896	206,680	178,678
Professional	227,255	234,248	231,300	262,832	242,026	250,230
Managers & senior officials	99,323	227,024	191,481	123,897	234,993	208,059
N	1344	1148	2492	945	1067	2012

Table 1 indicates that, across all occupations, women were less likely than men to have a PP and this incidence was also more variable among women. Men across all occupations reported high levels of PP membership, as did women in administrative, professional and managerial grades. Conversely only 46 per cent of women in elementary occupations had a PP; and other female-dominated occupations such as sales and personal services reported levels well below those of their male counterparts. As much as pension take-up increased up the occupational ladder, so did PPW (see Table 2). As a whole, women had less PPW than men and this effect was exacerbated by their lower take-up of such schemes. Based on the full sample, including those with no PP, the PPW gender gap was 47.58 per cent, falling to 30.53 per cent for those with a PP. This trend was also evident across occupations. The one exception was professional women who had PPW on a par with professional men across both measures of PPW indicating similar levels of pension membership and wealth accumulation.

Results

Factors affecting PP membership and wealth

The baseline model of pension wealth, in which pension wealth is only observed for those who have a PP, is reported in Table 3. The statistically significant value on ρ verifies the choice of Heckman methodology and the need to correct for sample selection bias. The model has a good fit and all categories of explanatory variables report statistically significant effects. Looking at the coefficients on each variable and their statistical significance, it is also interesting to note that they are not always consistent across both stages of the estimation, suggesting that some variables have a mixed effect in the accumulation of PPW. For ease of interpretation the results are discussed by variable across both equations rather than assessing each model in turn. In all but a few cases (where the variables are continuous), the coefficients and marginal effects are interpreted relative to the omitted group.

Table 3.

Heckman selection equation: private pension membership and wealth.

Independent variables	Positive private pension wealth (1/0)			Private pension wealth
	Coeff	t-stat	Marginal effect	Coeff	t-stat	Predicted value in £	Marginal effect in £
Gender
Female	−0.9447***	(9.63)	−0.1934***	−7824.82	(0.97)	118,829***	−26,309***
[Male]						145,138***
Occupation
Elementary occupations	−0.6576***	(4.01)	−0.1402***	−65921.77***	(4.46)	96,045***	−79,406***
Process, plant & machine operatives	0.7559***	(4.11)	−0.1662***	−69288.09***	(4.16)	90,126***	−85,326***
Sales & customer service	−0.4369**	(2.41)	−0.0863**	−55748.87***	(3.33)	111,434***	−64,018***
Personal service	−0.3756**	(2.24)	−0.0725**	−54228.20***	(3.59)	114,279***	−61,173***
Skilled trades	−0.4473**	(2.42)	−0.0887**	−73190.18***	(4.84)	93,763***	−81,689***
Admnistration & secretariat	0.0880	(0.57)	0.0140	−41375.37***	(2.80)	135,426***	−40,026***
Associate professional & technical	−0.1899	(1.07)	−0.0341	−15181.00	(0.95)	157,006***	−18,445
Professional	−0.0962	(0.52)	−0.0166	1333.18	(0.07)	175,192***	−260
[Managers & senior officials]						175,452***
Highest educational qualification
[No qualifications]						94,204***
Foreign	−0.2328*	(1.91)	−0.0549*	12589.82	(1.26)	101,452***	7248
NVQ1/CSE	0.0771	(0.44)	0.0165	1002.18	(0.07)	96,803***	2598
NVQ2/O level equivalent	0.1797*	(1.80)	0.0371*	20368.27**	(2.48)	118,161***	23,957***
NVQ3/A level equivalent	0.0583	(0.43)	0.0126	34603.51***	(3.05)	130,023***	35,819***
Higher education below degree	0.2608**	(1.98)	0.0523**	41803.74***	(4.02)	141,068***	46,864***
Degree/NVQ4/NVQ5	0.2136	(1.33)	0.0436	117604.10***	(7.22)	216,024***	121,820***
Marital status
Divorced or legally separated	−0.1193	(0.58)	−0.0231	4022.24	(0.27)	122,619***	1769
Remarried	−0.1364	(0.61)	−0.0265	12749.63	(0.82)	131,008***	10,158
Married (first marriage)	−0.1652	(0.84)	−0.0325	12372.16	(0.97)	130,050***	9200
Widowed	0.4796*	(1.69)	0.0735*	90445.21***	(3.01)	218,485***	97,635***
[Single]						120,850***
Ethnicity
Non-white	−0.2016	(0.99)	−0.0434	−1607.69	(0.05)	131,525***	−5870
[White]						125,655***
Age	−0.0247*	(1.84)	−0.0050*	1204.93	(0.87)		716
Number of Children	−0.0785***	(2.75)	−0.0159***	−2257.38	(0.91)		−3812
Hours worked per week	0.0143***	(4.29)	0.0029***	1845.84***	(5.10)		2129***
Housing
[Home owner]
Living rent free	−0.6204***	(2.74)	−0.1495**
Paying rent	−0.5114***	(4.45)	−0.1198***
Buying with a mortgage or loan	0.0263	(0.34)	0.0052
No financial products	0.4474*	(1.81)	0.0772**
No. of vehicles	0.0831*	(1.76)	0.0169*
Numeracy score	0.0379	(1.22)	0.0077
Organizational size
[2–4 employees]
5–19 employees	0.0201	(0.12)	0.0053
20–99 employees	0.1198	(0.75)	0.0306
100–499 employees	0.3102*	(1.76)	0.0756*
500–999 employees	0.3652**	(1.97)	0.0876*
1000 or more employees	0.6868***	(4.69)	0.1497***
Constant	2.2251***	(2.75)		16856.75	(0.21)
rho(ρ)	−0.4098***			F (15.56)***
Sigma	144672***
Lambda	−59292***
N	2364

***, **, *

denote statistically significant at the 1%, 5% and 10% levels.

Among the sample of employees, being female was found to be a major impediment to PP membership indicating that women were nearly 20 per cent less likely than men to be in a pension scheme. This sizeable effect persisted despite controlling for a host of other factors which are commonly perceived to lower women’s take-up of private pensions, for example having children, their marital status and working fewer hours (see conclusion for further discussion). Gender was less of an issue in the wealth equation (non-significant coefficient), indicating that the statistically significant marginal effect on PPW is largely accounted for by the low take-up of pensions by women rather than differences in their ability to accumulate pension wealth once they became a member. The wealth effect, conditional on selection, pointed to women having £26,309 less pension wealth on average than men, this figure being the difference between the predicted values of PPW for men of £145,138 against £118,829 for women – a gender gap of 18.13 per cent.

The lack of a gender effect in the wealth equation is worthy of further comment. In specifying the Heckman equation, the gender coefficient remained large and statistically significant until the ‘hours worked per week’ variable was added. Excluding the ‘hours worked’ variable gave a higher female PP membership shortfall of nearly 23 per cent and an overall PPW deficit of over £46000. This suggests that much of the expected gender PPW effect may be attributed to the lower number of hours that women worked. The inclusion of the ‘hours worked’ variable in the model therefore negates some of the gender effect.

The occupational measures revealed both selection and wealth effects culminating in some substantial and statistically significant differences across occupational values of PPW. Relative to the omitted group, managers and senior officials, a clear hierarchical split emerged between the professional/managerial occupations and all other occupational groups. Those in associate professional and professional occupations were just as likely to be members of a PP as managers and senior officials and as likely to benefit from equally favourable levels of PPW. Managers and senior officials reported the highest predicted value of PPW as £175,452 which was only marginally better (by £260) than the average professional. The associated professional and technical occupation had on average £157,006 of PPW which, although £18,445 less than managers, was not significantly different statistically.

The other occupational groupings stood in marked contrast to this. At worst, elementary occupations and process, plant and machine operatives were 14 per cent and 17 per cent less likely to have a PP. This, combined with their lesser ability to accumulate PPW, meant that, on average, they were between £80,000 and £85,000 worse off than managers and senior officials. Similar findings were revealed for the other occupations, those often held by women (sales and customer services, personal services). For administration and secretariat, which had the highest concentration of women, a pension wealth ‘deficit’ persisted. However, there was no significant selection effect, suggesting that most of the £40,000 PPW shortfall for administrators came from their inability to accumulate wealth at the same rate as those in managerial occupations.

As with occupation, categories of educational qualifications are interpreted relative to the omitted group, those with no qualifications. Generally educational qualification was not a major determinant of PP membership but had a sizeable impact on the value of PPW. The benefits of higher levels of educational attainment were particularly evident for those with degree level qualifications, this grouping being £121,820 better off, on average, than those with no qualifications. Those who had attained NVQ2 level upwards (equivalent to the school leavers’ award at age 16) were moderately better off than the reference group but those with the lowest level of qualifications were indistinguishable from those with no qualifications at all.

From the sample of individuals in paid employment the effect of marital status was remarkably homogeneous in both the selection and wealth equations. The only statistically significant difference against the reference group (single) related to those who had been widowed, their role as ‘beneficiaries’ of their partners’ pensions in all likelihood underpinning the near £100,000 additional PPW relative to their single counterparts.

Turning to the three continuous measures of age, number of children and number of hours worked per week, the number of children presented a pattern of results similar to that of gender. The number of offspring was a key influence on PP membership, each child reducing the probability of membership by 1.6 per cent; but once in a pension, the number of children did not seem to affect the accumulation of PPW. The dominance of the selection effect, however, did indicate that when both effects were considered together each child on average reduced an individual’s PPW by £3812 although this effect was marginally insignificant at conventional significance levels.

An individual’s age followed a similar trend, indicating that the oldest among the studied cohort were less likely to be members of a pension scheme (0.5% for every additional year), consistent with the increase in PP membership over time. However, this was not large enough to lead to any significant age-related wealth effect.

As has been inferred, the smaller than expected gender wealth effect was in large part down to working hours. Those working part-time or reduced hours per week were less likely to be members of a PP than those on full-time contracts and consequently would accumulate less pension wealth. Every extra hour of paid work increased the likelihood of having a pension by 0.29 per cent and increased PPW by over £2000. Among the working sample, women worked on average 13 hours less than men per week (29.8 hours as against 43 hours) which added up to a sizeable PPW deficit, even after controlling for other factors.

Other variables in the selection equation indicated the importance of organizational and economic factors. A key factor before an employee could decide whether to join a pension, or how much to contribute, was whether their employer offered an occupational pension. Organizational size was used to proxy for this effect, the evidence suggesting that PP provision is likely to be limited to the largest organizations, especially those in the public sector, that have the resources, expertise and will to operate such schemes. In the largest organizations, individuals had a 15 per cent higher probability of having a pension than their counterparts in the smallest organizations.

With regard to economic factors, those living in rented accommodation were significantly less likely to be members of a PP, by between 12 per cent and 15 per cent compared to people owning their own home. Similarly, the number of vehicles one owned was associated with a higher incidence of PP membership, each additional vehicle increasing the probability by nearly 2 per cent. However, individuals with no financial products were nearly 8 per cent more likely to be members of a PP, suggesting that for this small group PP membership is prevalent despite their lack of familiarity with other financial products.

Finally to assess people’s ability to deal with the complexity of pension issues and decisions, a proxy variable of their numerical ability was included. While this index was positively associated with PP membership the effect was not statistically significant.

Joint gender and occupation effects

The study’s analysis of the joint effect of gender and occupation on PPW involved re-estimating the aforementioned model with the inclusion of interaction terms between gender and occupation in the wealth equation.³ Rather than reporting the whole model output (the results for the control variables were nearly identical to the non-interaction model in Table 3), Table 4 reports the relevant post-estimation analysis of predicted probabilities/values and associated 95 per cent confidence intervals to give some indication of the spread of wealth outcomes within each occupation. Wald tests are also reported to discern any statistically significant difference between men’s and women’s PP membership/wealth for each occupational category. The results reveal some intriguing and unexpected findings, with several clear trends emerging.

Table 4.

Predicted private pension wealth from Heckman Model with gender/occupations interaction terms.

Independent variables	Positive private pension wealth (1/0) predicted probabilities			Private Pension Wealth Predicted values in £
	Female	Male	Adjusted Wald test	Female	Male	Adjusted Wald test
			F			F	F sign
Occupation
Elementary occupations	0.6030***	0.8698***	72.60***	110,648***	89,481***	2.73*	(0.0991)
Elementary occupations				(91,941–129,355)	(71,540–107,422)
Process, plant & machine operatives	0.5694***	0.8509***	59.95***	82,829***	104,689***	1.32	(0.2503)
Process, plant & machine operatives				(50,946–114,713)	(82,970–126,408)
Sales & customer service	0.6813***	0.9088***	58.20***	106,971***	102,000***	0.05	(0.8190)
Sales & customer service				(81,613–132,329)	(68,268–135,731)
Personal Service	0.7010***	0.9175***	70.59***	108,086***	105,832***	0.01	(0.9176)
Personal Service				(89730,126442)	(65,685–145,978)
Skilled trades	0.6715***	0.9043***	41.76***	110,504***	106,162***	0.02	(0.8791)
				(57,214–163,794)	(89,410–122,915)
Administration & secretariat	0.8312***	0.9653***	71.35***	113750***	194691***	12.21***	(0.0005)
Administration & secretariat				(102,389–125,110)	(149,078–240,304)
Associate professional & technical	0.7591***	0.9409***	47.36***	132,482***	182,977***	6.74***	(0.0096)
Associate professional & technical				(107,844–157,119)	(155,341–210,614)
professional	0.7740***	0.9464***	40.36***	193,768***	167,886***	1.59	(0.2081)
				(159,781–227,754)	(137,320–198,451)
Managers & senior officials	0.8184***	0.9613***	37.11***	116539***	203688***	15.87***	(0.0000)
Managers & senior officials				(88,551–144,526)	(170,949–236,428)

***, **, *

denote statistically significant at the 1%, 5% and 10% levels.

Clearly evident is the persistence of a gendered selection effect across all occupational groupings, although the extent of this PP membership disadvantage diminished moving up the occupational ladder. At best women in management, professional and administrative occupations reported a 15–18 per cent lower probability of having a private pension, but for large swathes of lower skilled occupations this gap increased to between 20 to 30 per cent. Even the lowest take-up of pensions among men was better than the highest take-up among women.

Looking at the wealth outcomes it is evident that male levels of PP wealth showed more conventional patterns, with a clear ‘hierarchical’ effect and a marked split occurring between the administrative and professional grades and the skilled trades, services and manual grades. In contrast PPW was clustered in a narrower range for the majority of female occupations, with the exception of the emergence of a ‘privileged pole’ among professional women.

Despite the clustering and lower incidence of PPs across all occupations there is evidence that women in many manual occupations had access to pensions that were not statistically different in value to those of men. In such occupations – process, plant and machine operative, sales and customer service, personal service and skilled trades – the average PPW of men and women, after allowing for selection effects, was concentrated in the range £105,000 to £110,000. This trend was also evident in elementary occupations, where many women and men work, except in this case it resulted in a significantly higher level of PPW for women over their male counterparts. While there was no discernible PPW gender differential for lower skilled occupations, both genders experienced low levels of PPW relative to all other occupations.

In most non-manual and higher-level occupations both selection and accumulation effects led to larger gender gaps in PPW. Despite the smaller selection disadvantage faced by such groupings the results indicate that these selection effects were further compounded by women’s inability to accumulate PPW equivalent to their male counterparts or significantly higher rates than ‘lower’-level female occupations. By far the largest gender gap in PPW was faced by female managers and senior officials, their £116,539 of PPW being dwarfed by the corresponding figure of £203,688 for men. Associate professional and technical and administrative and secretariat occupations revealed a similar, albeit smaller, gap in PPW. The anomaly in this picture is the privileged pole that is professional female employees. Despite a sizeable selection deficit, professional women accumulated the highest value of PPW across all female occupations and 15 per cent more than their male equivalents, although this difference was not significant at conventional significance levels.

Conclusion

This research confirms the salience of gender and occupation as key sources of inequality in PPW, despite controlling for a range of gendered variables, notably hours worked. It also provides new evidence on the uneven and diverse impact of occupation on the gender gap in PPW. In so doing this research contributes to sociological debates about the complex relationship between occupational gender segregation and gender inequality.

The Heckman methodology demonstrates how gender disadvantage in PPW comprises two effects, namely membership and wealth accumulation. The base model reveals no significant independent gender wealth effect, in part because the gendered nature of working hours accounts for much of the expected gender effect. By far the greatest and most uniform impact on the PPW gender gap was the lower rate of PP membership among female employees. Within occupational groupings, females may be clustered in jobs with fewer opportunities for membership, as compared with males. Thus, at a more disaggregated level, patterns of job segregation may have sustained the gender gap in PP membership. There may also be gender differences in knowledge of pensions which affect pension take-up as well as gender differences in attitudes towards PP membership, especially among married employees, given the traditional dependence of married women on their husbands’ pensions. Moreover, with lower lifetime earnings than men, women may be more likely to feel that they cannot save enough in a PP to justify membership.

Disaggregating occupation by gender highlights the heterogeneity of women’s PP experience, confirming that the feminization of occupations is a complex, uneven and contradictory process (Jarman et al., 2012). Women in manual or service occupations are doubly disadvantaged by gender and occupation, with respect to PP membership. However it is the penalties associated with occupation, rather than gender, that are significant in reducing their wealth accumulation. Occupational gender segregation may give some women access to better savings opportunities than men, possibly associated with their concentration in public sector employment. That said, because working hours are gendered, not least in lower-level occupations, women incur a further penalty in relation to PPW. Hence the conclusion that it was the socioeconomic rather than the gendered nature of occupation that had the most significant negative impact on women’s wealth accumulation in the manual and service occupations. It should be noted that men were slightly more likely than women to find themselves in manual occupations; and patterns of occupational feminization that involve a shift from manual to non-manual employment may sometimes moderate patterns of gender inequality (Jarman et al., 2012).

Women in non-manual occupations are more likely to be disadvantaged by gender than occupation in relation to both PP membership and pension savings. Thus, having a non-manual occupation had a consistently beneficial effect on PP membership and on PPW in the case of those in professional, technical and managerial occupations, though less so for the administrative grouping. However, the interaction analysis provides insights into how gender attenuates the positive effects of socioeconomic position across these occupations with the exception of the professional grouping. It is probable that, in administration and management, male advantage is associated with hierarchical patterns of occupational gender segregation, whereas within professional occupations, there is no evidence of male advantage in PPW. If anything women now appear to be advantaged in terms of the sectors and types of work in which they are concentrated.

The diverse nature of professional and managerial career trajectories (Crompton and Harris, 1999) may help explain these contrasting trends. Professional careers offer opportunities for career breaks and variation in hours worked once professional accreditation is acquired. Managerial careers offer less flexibility because expertise is evaluated by the employing organization, not by an external regulatory body. Hence women in managerial occupations are more likely than women in professional occupations to experience a negative impact on their career as a result of family care responsibilities.

A second possible explanation may stem from the different sectors in which male and female professionals are employed. Female professionals may be concentrated to a greater extent than males in public sector employment and this may have beneficial effects on their PPW, especially if they are able to access more favourable DB schemes. This public sector effect could also be relevant in explaining the unexpected gender difference found in elementary occupations where employed females’ PPW is actually higher than that for comparable males. Unfortunately data on the sector of employment are not available in the dataset but the impact of the public/private sector would be interesting to explore in future research.

Hence the conclusion that occupational feminization has contradictory effects on gender equality in PPW, in which gains derived from movement into non-manual and higher-level occupations are offset unevenly by gender differences in opportunities for wealth accumulation. Pension saving opportunities were curtailed for the majority of this cohort of women by their over-representation in manual, service and lower-level non-manual occupations, among which the largest group of women, those in the administrative occupations, experienced an additional gender penalty. For most women in higher-level occupations, those in managerial and technical occupations, the socioeconomic benefits of occupation were largely offset by gender penalties. However, a minority of women, those in professional occupations, appeared to have derived benefit from both the gendered and socioeconomic nature of their occupations, thus confirming that feminized occupations are not always characterized by inferior rewards.

This exploration of the gender gap in PPW reinforces the need for policies to address the gender gap in PP membership; and the new statutory requirement for employers to enrol employees into a workplace pension is an important step in this process. However, a narrowing of the gender gap in PPW is unlikely, without further action to address the limited opportunities for PPW accumulation in manual, service and lower non-manual occupations and to tackle the gender penalties associated with pension saving in most non-manual and managerial occupations. A policy context of public sector austerity and privatization is unlikely to be conducive to such action.

Footnotes

Acknowledgements

The data were made available through the UK Data Archive. ELSA was developed by a team of researchers based at University College London, the Institute of Fiscal Studies and the National Centre for Social Research, and funding for the ELSA team was provided by the National Institute on Aging in the U.S. and a consortium of U.K. government departments coordinated by the Office for National Statistics. The developers and funders of ELSA and the Archive do not bear any responsibility for the analyses or interpretations presented here.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Notes

Jean Gardiner is a senior lecturer in employment relations in the Centre for Employment Relations, Innovation and Change at Leeds University Business School. She has a long-established research interest in the inter-relationships between labour markets, gender and family care. Her current research is focused on the institutional and pension context for workforce participation in later life and for labour market transitions, such as redundancy and retirement.

Andrew M Robinson is Professor of Accounting and Performance at Leeds University Business School. The major focus of his research is on employee share ownership and employee-owned businesses and pension provision and pension wealth. Andrew is a co-founder of the White Rose Employee Ownership Centre.

Fathi Fakhfakh is Associate Professor of Economics at Paris II University. The major themes of his research are profit sharing, employee ownership, governance, innovation and firm performance in France and Europe.

References

Acker

(1990) Hierarchies, jobs, bodies: a theory of gendered organizations. Gender and Society 4(2): 139–58.

Banks

Emmerson

Tetlow

(2005) Estimating Pension Wealth of ELSA Respondents, Working Paper WP05/09, Institute for Fiscal Studies, London, UK.

Bardasi

Jenkins

(2010) The gender gap in private pensions. Bulletin of Economic Research 62(4): 343–63.

Blackburn

Jarman

(1997) Occupation Gender Segregation. Social Research Update, issue 16. Guildford: Department of Sociology, University of Surrey.

Blake

(2000) Does it matter what type of pension you have? Economic Journal 110(461): 46–81.

Bozio

Emmerson

Tetlow

(2011) How Much Do Lifetime Earnings Explain Retirement Resources? Working Paper WP11/02, Institute for Fiscal Studies, London, UK.

Bradley

(1999) Gender and Power in the Workplace: Analysing the Impact of Economic Change. London: Macmillan Press.

Crompton

Harris

(1999) Employment, careers and families: the significance of choice and constraint in women’s lives. In: Crompton

(ed.) Restructuring Gender Relations in Employment. Oxford: Oxford University Press, 128–49.

DWP (Department for Work and Pensions) (2005) Women and Pensions: The Evidence. Available (consulted 13 March 2015) via: http://www.dwp.gov.uk/docs/women-pensions.pdf

10.

DWP (Department for Work and Pensions) (2006) Security in Retirement: Towards a New Pensions System. Available (consulted 13 March 2015) via: http://www.dwp.gov.uk/docs/white-paper-complete.pdf

11.

DWP (Department for Work and Pensions) (2011) Workplace Pension Reforms Evaluation Strategy. Research Report 764. Available (consulted 13 March 2015) at: http://research.dwp.gov.uk/asd/asd5/rports2011-2012/rrep764.pdf

12.

DHSS (Department of Health and Social Security) (1985) Reform of Social Security. Cmnd 9517. London: HMSO.

13.

DSS (Department of Social Security) (1998) A New Contract for Welfare: Partnership in Pensions. London: The Stationery Office.

14.

Evandrou

Glaser

(2003) Combining work and family life: the pension penalty of caring. Ageing and Society 23(5): 583–601.

15.

Foster

(2011) Towards a new political economy of pensions? The implications for women. Critical Social Policy 30(1): 27–47.

16.

Frericks

Knijn

Maier

(2009) Pension reforms, working patterns and gender pension gaps in Europe. Gender, Work and Organization 16(6): 710–30.

17.

Ginn

(2003a) Gender, Pensions and the Lifecourse: How Pensions Need to Adapt to Changing Family Forms. Bristol: The Policy Press.

18.

Ginn

(2003b) Parenthood, partnership status and pensions: cohort differences among women. Sociology 37(3): 493–510.

19.

Ginn

Arber

(1991) Gender, class and income inequalities in later life. British Journal of Sociology 42(3): 369–93.

20.

Ginn

Arber

(1993) Pension penalties: the gendered division of occupational welfare. Work, Employment and Society 7(1): 47–70.

21.

Ginn

Arber

(1996) Patterns of employment, pensions and gender: the effect of work history on older women’s non-state pensions. Work, Employment and Society 10(3): 469–90.

22.

Ginn

Arber

(1999) The changing patterns of pension inequality: the shift from state to private sources. Ageing and Society 19(3): 319–42.

23.

Ginn

Arber

(2000) Personal pension take-up in the 1990s in relation to position in the labour market. Journal of Social Policy 29(2): 205–28.

24.

Greene

(2003) Econometric Analysis, 5th edition. Upper Saddle River, NJ: Prentice Hall.

25.

Jarman

Blackburn

Racko

(2012) The dimensions of occupational gender segregation in industrial countries. Sociology 46(6): 1003–19.

26.

McKay

Heaver

Walker

(2000) Building up Pension Rights. Research Report 114, Department of Social Security, London: HMSO.

27.

Marmot

Oldfield

Clemens

Blake

Phelps

Nazroo

(2013) https://dx-doi-org.web.bisu.edu.cn/10.5255/UKDA-SN-5050-6 (computer file), 20th edition. Colchester: UK Data Archive (distributor) SN: 5050. Available (consulted 13 March 2015) at: https://dx-doi-org.web.bisu.edu.cn/10.5255/UKDA-SN-5050-6

28.

Neumayer

(2011) On the detrimental impact of visa restrictions on bilateral trade and foreign direct investment. Applied Geography 31(3): 901–7.

29.

O’Connell

Gash

(2003) The effects of working time, segmentation and labour market mobility on wages and pensions in Ireland. British Journal of Industrial Relations 41(1): 71–95.

30.

ONS (Office for National Statistics) (2000) Standard Occupational Classification 2000, Volume 1: Structure and Descriptions of Unit Groups. London: The Stationery Office.

31.

ONS (Office for National Statistics) (2009) Wealth in Great Britain: Main Results from the Wealth and Assets Survey 2006/08. Available (consulted 18 March 2015) at: http://www.ons.gov.uk/ons/rel/was/wealth-in-great-britain/main-results-from-the-wealth-and-assets-survey-2006-2008/index.html

32.

Pensions Policy Institute (2012) The Pensions Primer. Available (consulted 8 March 2012) at: http://www.pensionspolicyinstitute.org.uk/default.asp?p=97

33.

Sefton

Evandrou

Falkingham

(2011) Family ties: women’s work and family histories and their association with incomes in later life in the UK. Journal of Social Policy 40(1): 41–69.

34.

Standing

(1999) Global feminization through flexible labor: a theme revisited. World Development 27(3): 583–602.

35.

Stier

Yaish

(2014) Occupation segregation and gender inequality in job quality: a multi-level approach. Work, Employment and Society 28(2): 225–46.

36.

Vance

Buchheim

(2005) On the Application of Heckman’s Sample Selection Model to Travel Survey Data: Some Practical Guidelines. Strasbourg: Association for European Transport.

37.

Warren

(2003) A privileged pole? Diversity in women’s pay, pensions and wealth in Britain. Gender, Work and Organization 10(5): 605–8.

38.

Warren

(2006) Moving beyond the gender wealth gap: on gender, class, ethnicity, and wealth inequalities in the United Kingdom. Feminist Economics 12(1/2): 195–219.

39.

Warren

Rowlingson

Whyley

(2001) Female finances: gender wage gaps, gender assets gaps. Work, Employment and Society 15(3): 465–88.

40.

Wooldridge

(2013) Introductory Econometrics: A Modern Approach, 5th edition. Mason, OH. South-Western Cengage Learning.