Is pay discrimination against young women a thing of the past? A tale of two cohorts.
Equal treatment of men and women in the labour market has been enshrined in British law for nearly 20 years. The Equal Pay Act came into full force at the same time as the enactment of the Sex Discrimination Act in 1975. It was amended in 1983 to allow equal pay to be claimed in comparable rather than identical jobs. By the 1990s, therefore, pay discrimination against women ought to have become a thing of the past. This article inveStigates whether this is so, taking evidence on men and women in their early 30s at two points during this period.
Differences in the rate of pay received by men and women are of clear interest from the point of view of both equity and efficiency. Horizontal equity within the labour market would require equally productive workers to receive equal pay independently of personal characteristics such as gender, age or race. Moreover, allocative efficiency requires workers at the margin to receive a wage equal to their value product so that for the labour market to operate efficiently, equally productive workers should receive equal pay. Failure to do so will, among other things, reinforce, if not influence, the time each group allocates to paid and unpaid work, to acquiring skills and accumulating human capital. Low returns on the latter for women may inhibit the development of their full potential to contribute to the economy. Obviously, neither the requirements for allocative efficiency nor the principle of horizontal equity described above, imply that women and men should receive on average equal pay, or that equal pay legislation is ineffective if a gender wage gap exists. It may just be that women and men are not homogeneous workers and bring unequal productive assets to the marketplace.
A major source of human capital differential between the two sexes is in their attachment to the labour market. To assess whether the lower pay of women is due to unequal treatment, it is therefore advisable to control for differential work experience. Such evidence is not regularly collected for nationally representative samples. This article reports one of the first attempts to bring the new evidence from the fifth sweep of National Child Development Study (NCDSS) to bear on this question. The findings from this survey of a cohort born in 1958 are compared with an earlier cohort, the MRC's National Survey of Health and Development (MRC), who were born in 1946. The latter provides evidence on pay and employment history around the beginning of 1978, by which time the Equal Pay Act seemed to have achieved most of its initial impact. NCDS reveals the crude excess of male hourly wages among 33 year-old employees born in 1958, to have fallen, at the mean to 39 per cent (from 56 per cent in the MRC cohort).
One may be tempted to interpret this decrease as further success of policies intended to reduce discrimination via the implementation and strengthening of equal pay and equal opportunity legislation. This conclusion would assume homogeneity between the young workers in the two cohorts (or that any inter-cohort change in the characteristics of the labour force had left the male/female differential in those characteristics unchanged). Workers from the two cohorts are, however, different. In standard human capital variables, the later cohort, particularly the women, are better endowed. Perhaps the closing wage gap is due less to any change in the extent of discrimination than to educational expansion, the shortening of career breaks and the increased numbers of women pursuing uninterrupted careers. The latter stems from the private "policy" of the postponement of childbearing; public policy to spread maternity leave and benefits, as in the Employment Protection Act (1976); and despite the lack of any positive Government policy to subsidize child care in the period concerned.
The purpose of this article is to establish the extent to which the inter-cohort increase in the female/male wage ratio is the result of changes in the labour market characteristics of the cohort members and to assess the magnitude of any remaining unequal treatment reflected in the unexplained residual. At this stage the emphasis is on human capital, rather than occupational characteristics. The focus is on the most recent sweep of the later study, when the members of the 1958 cohort were 33, in 1991, and the sweep of the 1946 cohort near age 32 in 1978. We use these surveys to compare the hourly pay of three groups of employees: men employed full-time, women employed full-time and women employed part-time. We find such difference in treatment between the two groups of female workers that it seems appropriate to make the threefold comparisons. The closeness of the male/female wage ratios emerging from the two cohorts with that for employees of all ages (at the median) derived from the New Earnings Survey suggests that the analyses of the cohorts probably does have some applicability to a wider age group.
Data sources, model specification and results
The cohort studies, two used here out of three national follow-up studies of a week's births, are unique to Britain. The MRC's National Survey of Health and Development follows up the legitimate singleton births in Britain during a week in March 1946 (see Wadsworth, 1991, for full details). A sub-sample of children of the urban working class were followed up, leaving a study size of 5,362 cases known at birth. In our analyses, cases have been weighted appropriately. There have been periodic follow-ups across childhood and into middle age. The data used in this study derive from a postal contact with the survey members around the end of 1977 and beginning of 1978 when they were approaching the age of 32. (For convenience we refer to this evidence as coming from 1978 at age 32.) Evidence on ability, education and employment experience came from earlier sweeps of the cohort. For more details on the data used in this article see Joshi and Newell (1989).
The other cohort study we utilized was the National Child Development Study (NCDS) of a cohort born in 1958. They are therefore exactly 12 years younger than the MRC cohort. The most recent sweep, aged 33 in 1991, provides the occasion for this study. Further details about the study can be found in Ferri (1993). As no under-sampling took place, and as immigrants were recruited in childhood, the full size of this study is greater than the MRC. Over 18,000 people have at some stage been in contact with NCDS, though the number taking part in the 1991 interview (NCDSS) was 11,407, not all of whom gave complete information. Our estimates are based on retrospective employment history information given at 3.3, educational attainments (as reported in NCDSS) and ability as measured at age 11, in an equivalent test administered to the MRC cohort.
The earnings function is estimated for all cohort members who were currently employed as employees and had valid information on wages, i.e. gross pay per hour. Its specification is based on a pure human capital model where the only determinants of differences in wages are heterogeneities in worker's specific human capital characteristics, and whether or not the job is part-time. Separate earning functions are estimated for (full-time) men and full-time and part-time women, since these two sectors have been shown to remunerate women's human capital unequally (Ermisch and Wright, 1992). The hypothesis of selectivity bias among women members is tested using a Heckman procedure extended to allow for an Ordered Probit selection into full- and part-time employment.
Our preliminary results are encouraging. The probability of women participating, and participating full-time is reduced by marriage and having children, but it increases with the age of the youngest child. Both ability and the educational dummies have significant and positive effects on participation, and so does the fact of having a husband who left school before 17 years of age. More surprising is the result that being an owner-occupier reduces the probability of participation in either of the two states for the MRC women, while it increases participation among the 1958 cohort. The significant threshold parameter, [Mu], indicates a systematic discontinuity between part-time and full-time participation. The estimates generate a separate lambda term for each sample member, reflecting her probability of being selected into part- and full-time employment respectively (Greene, 1992).
The wage equations explain the data well with [R.sup.2] ranging from 0.24 to 0.37. All the t-ratios are based on standard errors which have been corrected for heteroscedasticity. The estimation of wage equations for both part-time and full-time women leads to the unsurprising result that education and ability increase wages. More surprisingly, it also suggests that, for the earlier cohort, early work experience and job tenure do not have a significant effect on women's wages, irrespective of whether they work part-time or full-time, and that recent work experience only affects positively the wages of full-timers. However, the effect of both recent and earlier work experience and of job tenure becomes significantly positive for the NCDS women who are also significantly affected by being based in the South East. The different degree of significance of the coefficients of earlier and more recent experience for all workers in both cohorts supports the intuition behind the split of the experience variable into the two periods. For neither category and neither cohort is the lambda term significant (though verging on it for part-timers). This suggests a lack of self-selection into the labour market for women from both cohorts, a result in common with Joshi and Newell (1989). Men's wages appear to be positively affected by standard human capital variables and by being located in the South East.
Decomposition of wage gaps
The decomposition of the log wage gap breaks it down into a portion explained by differentials in attributes and a portion explained by difference in parameters. The latter expresses the unequal treatment of workers with like attributes and is conventionally used as an indicator of discrimination. If there were only one attribute and its parameter involved, the calculation would be straightforward. With multiple explanatory variables it is necessary to calculate weighted sums of attribute and parameter differences to arrive at the decomposition. It is conventional to express the experience of labour market discrimination by women by weighting parameter differences by the distribution of characteristics of the female mean, which entails a corresponding weighting of attribute differences by men's parameters. We pay most attention to this conventional form of weighting, taking full-time females as the reference point, though we also show the alternative decomposition, weighting parameter differences by attributes of men and part-time females. For the first decomposition of the gap between full-time and part-time wages, differential attribute differences are weighted by the mean values for part-time women.
On this basis, the drop in the crude log gap between men and women, of 0.11 between the two cohorts was almost entirely due to a catching up by women on human capital attributes, so that a smaller part of the difference can be attributed to different characteristics in 1991. In 1978, men's superior attributes, notably in experience and education accounted for 0.093 of their higher mean log wage. By 1991 only 0.026 was due to more favourable attributes. Unexplained parameter differences continued to be the major component, and had changed less, from 0.211 to 0.175. There seems to have been some improvement in unequal treatment, but not much.
In our analyses of the gap between pay in the full-time and part-time sectors of the women's labour market, the earnings of full-time women grew more than those of part-timers so that the overall gap grew larger. Attributes account for a bigger part of the gap at each date than for the male/female comparison, and as in that case, there is some sign of a convergence in the characteristics of part- and full-timers. By 1991 women with higher qualifications were slightly more likely to report part-time employment than in 1978 (despite the under-representation of graduates in the NCDS wage regression sample). The widening gap which is accounted for by parameter differences, 0.110 in 1978, became 0.139 in 1991. On the basis of these decompositions, female gains in the full-time labour market appear to have occurred over the 1980s, but the wages of women working part-time have moved in the opposite direction, partly offsetting these gains.
When the decomposition of the full-time male/female gap weights parameter differences by male attributes, the improvement in unequal treatment is of the same order of magnitude as for the selection-adjusted estimates quoted above. Weighted parameter differences fall from 0.252 in 1978 to 0.115 in 1991. From the perspective of men's attributes, the closing of the gender gap seems to be more a story of equalizing treatment than women gaining more human capital.
Looking at the part-time/full-time gap for women from the perspective of the attributes of part-timers, the increase in parameter differences is bigger than the difference due to attributes, whether or not selection adjustments are made. Explained components diminish over time on either treatment of selectivity and of weighting - confirming at least one of the features of the initial decomposition. Whatever the precise gains to the full-time labour market, they have been at least partly offset by deteriorating terms for part-time work.
The pattern of unequal treatment of women observed in 1978 was not uniform. On many characteristics, full-time women employees had an advantage over men, for example educational qualifications and most of the more auspicious social backgrounds. Men's wages benefited most from recent work experience, residence in the South East and from being in the reference category. The net improvement for women in 1991 came from a mixture of cross-cutting trends. The advantage of most educational qualifications and social backgrounds seems to have swung to men, but their relative advantage from recent employment experience and residence in the South East has weakened, and the advantage of being in the reference category has swung to females. This is not a straightforward case of equalizing rewards to human capital; hence the sensitivity at the discrimination estimates to which weights, are used.
The penalty of part-time work in 1978 consisted mainly of better returns to almost all characteristics in full-time than in part-time work, offset by a higher constant in part-time work. By 1991, the picture had partially reversed. Returns to individual characteristics became closer if not better in part-time work, but the constant term revealed a generalized penalty to part-time employment for those in the reference category. The changes in wage ratios for graduates is noteworthy. In 1978 graduate women would be paid, all else equal, 0.183 more on the log wage than a part-timer. By 1991 the hourly rate of graduates doing part-time work was, ceteris paribus, 0.289 higher in part-time work than full-time work. This indicates that there were a few graduate women in 1991 (2.6 per cent of the part-time sample) who managed to maintain highly paid work on a part-time basis. There were even fewer such opportunities in 1978 and even fewer graduate women working part-time (1.6 per cent).
The disadvantages of being female in the full-time British labour market seem to weigh less heavily in the 1990s than they did at the end of the 1970s. This applies at least as far as possible to people in their early 30s, though we have shown that their relative wages were fairly representative of a broader average. For women in the full-time labour market, the disadvantages have by no means been eliminated. The average full-timer could expect to be paid up to 19 per cent more if she were a man. The disadvantages of being a woman are compounded if she is constrained to take a part-time job. Although less well qualified women concentrate in the part-time sector, they are penalized relative to comparable full-time women, and this penalty has worsened over the period. While the slow progress of equal opportunity policies in closing the full-time wage gap can be applauded with at least two cheers, the deterioration of opportunities for good pay in the part-time sector continues to cause concern about fairness and the efficiency of resource allocation. To the extent that part-time participation is a response to child care and other domestic constraints, women will continue to be trapped in low wage jobs with low training content until public or private policy may ease these constraints.
Future work will explore further the effect of family responsibilities on pay in the two cohorts, and on gender differences in the job characteristics for the 1958 cohort.
This work is supported by a grant to the SSRU by the Employment Department. The data from the MRC National Survey of Health and Development are used with the permission of Professor M. Wadsworth, its Director. The authors are grateful for his co-operation and that of his staff for this access.
Ermisch, J.F. and Wright, R.E. (1992), "Differential returns to human capital in full-time and part-time employment", in Folbre, Bergmann, Agarwal and Flor (Eds), Women's Work in the World Economy, Macmillan, London, pp. 195-212.
Ferri, E. (1993), Life at 33: The Fifth Follow up of the National Child Development Study, National Children's Bureau, London.
Greene, W. (1992), Limdep Version 6.0. User Manual and Reference Guide, Econometric Software, Inc.
Joshi, H. and Newell, M.-L. (1989), "Pay differentials and parenthood: analysis of men and women born in 1946", Institute of Employment Research Report, University of Warwick, Coventry.
Wadsworth, M.E.J. (1991), The Imprint of Time: Childhood, History and Adult Life, Clarendon Press, Oxford.
Pierella Paci and Heather Joshi, City University, London, UK
Gerry Makepeace, University of Hull, Hull, UK
Peter Dolton, University of Newcastle upon Tyne, Newcastle, UK
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||EMRU Conference Papers; pay discrimination in Great Britain|
|Author:||Paci, Pierella; Joshi, Heather; Makepeace, Gerry; Dolton, Peter|
|Publication:||International Journal of Manpower|
|Date:||Feb 1, 1995|
|Previous Article:||Union membership and coverage: a study using the nested multinomial logit model.|
|Next Article:||Migration equilibria/disequilibria and the natural rate of unemployment in a regional context.|