The Part-time Pay Penalty:

Earnings Trajectories of British Women

Sara Connolly* and Mary Gregory+

* School of Economics, University of East Anglia, Norwich NR4 7TJ, UK

+ Department of Economics, University of Oxford, Manor Road, Oxford OX1 3UQ, UK

Email:

Abstract

Part-time work among British women is extensive, and the (raw) pay penalty large. Since part-time work features most prominently when women are in their 30s, the peak childcare years and a crucial period for career building, its impact on subsequent earnings trajectories is important from a social as well as individual perspective.

We find that part-time work experience gives a very low return in future earnings, particularly when acquired in lower-skill jobs. In addition, one-quarter of women in high-skill jobs downgrade occupationally on switching to part-time work, rising to 43% among those who also change employer. In combination these effects give an immediate earnings drop of 32%, followed by a permanently lower trajectory. It is these accompanying changes, rather than part-time status itself, which damage earnings. Return to full-time work, even with reversal of the occupational downgrading, brings only a partial recovery; without it the earnings losses continue to grow.

JEL classification: C23, C25, C33, C35, J16, J22, J62.

The Part-time Pay Penalty: Earnings Trajectories of British Women

1. Introduction

Recent decades have seen a major expansion of part-time employment among British women as increasing numbers of them seek to combine paid employment outside the home with family responsibilities. 5.7 million women are now working part-time, an increase of over one-half since the mid-1970s. The numbers in full-time work have also been increasing, but more slowly. As a result, the proportion of working women who now work part-time has risen to around 40%, one of the highest rates in the advanced economies. The increase has been particularly marked among younger women, from their mid-20s to early 40s; their rising participation in each successive cohort has now fully filled in the dip in the traditional M-shaped age-profile of women’s labour market participation. But while women in full-time work have been closing the gender pay gap through their rising educational attainment, labour market attachment and occupational diversity, women working part-time have conspicuously failed to match this progress. In 1975 average hourly earnings of women working part-time were 84.6% of the earnings of women in full-time work, a pay gap of 15%; by 1985 this had widened to 21%, in the early 1990s to 25%, and in 2001 to 29% - one of the widest gaps among the advanced economies.[1] This deterioration in their relative position has led to the designation of women working part-time as ‘the new underclass’ (Humphries and Rubery, 1995), with the gender pay gap increasingly characterised in terms of ‘the part-time pay penalty’ (Manning and Petrongolo, 2008).

Since part-time work is predominantly engaged in by women (leaving aside students) its status clearly presents a major issue in gender equality.[2] More subtly, and even more significantly, around two-thirds of women work part-time at some stage of their adult careers, many then returning to full-time work. The impact of a spell in part-time work on subsequent earnings and career trajectories is therefore extremely important. That is the focus of this paper. We build on our previous work which identified occupational downgrading as an important concomitant of the switch to part-time work. Connolly and Gregory (2008) found that one-quarter of women in Britain who switch from full- to part-time work experience occupational downgrading. This includes over 20% of professional women, half of whom move to jobs classified as low-skilled, while two-thirds of nurses who leave full-time nursing become part-time care assistants, utilising only a limited portion of their specialised skills. This gives a new, sharper perspective on the ‘crowding’ into low-level occupations traditionally viewed as a major source of the gender pay gap. The fact that two-thirds of women work part-time swamps self-selection by the less skilled or less motivated. Most tellingly, significant numbers of educated women from higher-level occupations are observed switching into low-level part-time jobs.

In the present paper we extend the analysis to earnings, examining the implications of switching to part-time work for women’s subsequent earnings trajectories, and in particular the role of occupational downgrading in this. Conventional channels can be identified. Part-time employment reduces the accumulation of work experience; this can be expected to have a permanent depressive effect on future earnings, even after return to full-time work. When the switch to a part-time job also involves a change of employer, as often occurs, seniority and firm-specific capital are lost. Most challengingly, particularly for more educated women, when a part-time job is not simply a reduced-hours version of a previous full-time job but involves occupational downgrading, the return to higher-level skills is lost and career-building likely to be damaged.

Using data from a 27-year unbalanced panel which records the earnings, hours of work and occupations of over 60,000 women, we find that the pay penalty to part-time work can be fully explained through identifiable channels. The earnings return to the human capital accumulated in part-time work is lower than for full-time work, even on a full-time equivalent basis. Further, only part-time experience in a high-skill job brings a positive earnings return (still lower than for full-time experience); part-time experience accumulated in lower-skill jobs gives positive returns only at longer durations. Similarly, the marked negative impact on earnings associated with the switch from full- to part-time work largely derives from the occupational downgrading and change of employer that accompany it. Simulations indicate that a woman in a high-skilled occupation who switches to part-time work, downgrading to a lower level occupation for five years, and then returning to full-time work in the same job, will have permanently lower earnings and earnings growth; after six years in part-time work and a further five back in full-time work her earnings will be almost 40% below their potential level had she remained in full-time work in her high-skilled job.

The paper is structured as follows. Section 2 describes our dataset. Section 3 profiles the role of part-time work in the first half of the life-cycle of the women in our selected sample. In Section 4 we present estimated human capital earnings equations focusing on the respective contributions of full- and part-time work. Section 5 introduces the switch between full- and part-time work, including occupational downgrading and job change. Section 6 concludes with a discussion of some policy implications.

2. The New Earnings Survey Panel Dataset (NESPD)

To trace individual earnings trajectories over time requires panel data. Ideally the panel would be large and long, giving complete work histories and a rich set of personal and household characteristics in parallel with labour market status. The switch from full- to part-time work is a relatively infrequent event, involving fewer than 9% of those in full-time work in any year; a large sample is therefore required to give a sufficiently sizeable number of switches. Since a spell in part-time work often lasts for several years, a significant portion of the life-cycle should be covered. Experience in both full- and part-time work must be directly observed, not imputed. Labour supply decisions, including the choice of part-time work and its duration, are influenced by educational level and household situation; data on presence of a partner, number and ages of children are also desirable. No available dataset for Britain meets these requirements in full. As the best overall match we use the New Earnings Survey Panel Dataset (NESPD).

The NESPD is the panel dataset generated from the sequential annual New Earnings Surveys (NES), an administrative enquiry which requires employers to report the earnings, hours of work, occupation and other employment details for employees identified for the sample. The sampling frame comprises 1% of all adults, based on a selected pair of terminating digits in the National Insurance number. For each year this generates a random sample of individuals of all ages, in all occupations and types and sizes of firms. Since NI numbers are retained for life and the same digits are used to identify the sample in each year, the cross-sectional sampling frame automatically generates a panel; this forms the New Earnings Survey Panel Dataset (NESPD).

The NESPD offers a number of major advantages for present purposes. The sample is very large, over 74,000 women each year. Within the sampling frame new workers are automatically added as they enter employment, and individuals who have been out of the Survey in any period, due to non-employment or any failure of sampling, are re-identified in subsequent employment. Response is compulsory, with earnings and hours of work reported from payroll records. The length of the panel, up to 27 years, gives the opportunity to trace women’s earnings and occupational trajectories for over half the employment life-cycle in the case of older cohorts and for substantial periods even for younger cohorts. Labour market experience and employer tenure, in full- and part-time work separately, are available directly from the Survey. However, several limitations should be noted. Part-time workers are under-sampled as employees falling below the income tax threshold need not be included in their employer’s tax return (the basis of sample location) and therefore may not be identified for the Survey. More significantly, spurious employment interruptions are generated. At least a month elapses between the date at which individuals are located for the sample and the Survey pay week. Those changing employer in this interval are lost to the sample, even where there is a direct job-to-job move; the previous employer is no longer relevant, while the new employer cannot be identified. As a spot survey relating to a specified week, the NES may misclassify women relative to their average employment status over the year. Finally, as an administrative dataset drawn from payroll records, the NES gives only a limited set of personal characteristics (sex, age and occupation) and no information on educational attainment and qualifications, or the presence of partner and children. With the exception of the final item, however, the limitations of the NESPD are minor relative to its strengths.

We restrict our sample to women whose full labour market history to date can be observed. This confines the selection to the birth cohorts of 1958 or later. The oldest are aged 16 (the minimum school-leaving age) in 1975, followed until age 43 in 2001; potentially the youngest are born in 1985, aged 16 in 2001. Since part-time work is used by women in two different roles - supporting employment continuity until full-time employment is resumed, or alternating with spells of unemployment or inactivity in a part-time/non-employment cycle[3] - and we wish to confine the focus to women with significant labour market attachment, we further restrict the sample to those who record at least three years in work, starting from their first year working full-time. All prior observations of part-time work are dropped, along with all those for women only ever recorded working part-time. This selection also helps to eliminate part-time work by students who are not identified in the Survey.

The NES reports contractual (basic) hours for the specified week. In line with most British and OECD (although not US) practice we define part-time work as fewer than 30 hours per week. Earnings are reported gross, with overtime earnings separately. Our measure of pay is hourly earnings, calculated as gross weekly earnings, excluding pay in respect of overtime hours, divided by basic hours, deflated to 2001 prices using the RPI.[4] Where the individual’s pay is recorded as affected by absence, including cases of zero pay (which may be periods of unpaid maternity leave) the record is retained in the sample with pay recoded as ‘missing’.

To analyse occupational downgrading Connolly and Gregory (2008) construct a 15-occupation ranking, based on the average level of educational attainment and qualifications of men and women working full-time in these occupations. Since this classification is too detailed for current purposes, we aggregate the occupations to three levels: high, intermediate and low-skilled. Low-skill occupations are those where the average age of completing education is 16, the legal minimum; medium-skill occupations are held by individuals leaving education at ages 17-18 i.e. completing high school or a comparable period of vocational training; high-skill occupations are typically held by those educated beyond age 19. Indicative occupations are:

High-skill occupations - major groups 1-3 of the Standard Occupational Classification 1990 (SOC90), and occupations 1-6 of Connolly and Gregory: corporate managers and administrators; educational, legal, medical and scientific professionals; teachers; nurses; medical-related associate professionals.

Intermediate-skill - major groups 4, 6 and 7 of SOC90, occupations 7-11 of Connolly and Gregory: buyers and sales representatives, clerical and secretarial occupations, skilled trades, nursing and educational assistants.

Low-skill - SOC90 major groups 5 and 8, Connolly and Gregory occupations 12-15: sales assistants, waiters, bar staff, shelf-fillers, drivers, domestic and office cleaners.[5]

This sample selection results in an unbalanced panel of 596,160 observations on 62,061 individuals. Individuals are in the Survey between three and 27 years, at ages 16 to 43, and for an average of 9.6 years. Loss of pay due to absence affects 14.0% of the individual-year observations.

3. The profile of full- and part-time work

The age-profile of part-time work by women, shown in Figure 1 for five-year cohorts, has two striking features. The first is the remarkable consistency in the role of part-time work once women reach their later 20s, even across cohorts separated by 20 years. While overall participation has risen, the share of part-time work at each age has remained virtually constant. Secondly, although it remains small, the proportion of young women working part-time has been increasing; one source of this is the growing interleaving of work and further education.

[Figure 1 about here]

Figure 1 Percentage of Working Women who are in Part-time Employment, by Birth Cohort

The more detailed patterns illustrated in Table 1 for the longest cohort, that of 1958-62, show the total numbers in work peaking at ages 19-22. Full-time work predominates at first, then declines steadily as the women move through their later 20s and 30s, before beginning to turn upwards again as they reach their 40s. Among working women in their late 30s fewer than half are in full-time work. Part-time work, on the other hand, rises sharply from the mid-20s, peaking in the later 30s then beginning to decline. Over the entire age-span 30-43 well over one-third of working women work only part-time. Even within a four-year window as many as one woman in seven may engage in both full- and part-time work.