Who Works for Whom and the UK Gender Pay Gap
| Date | 01 March 2020 |
| Published date | 01 March 2020 |
| Author | Giovanni Razzu,Carl Singleton,Sarah Louise Jewell |
| DOI | http://doi.org/10.1111/bjir.12497 |
British Journal of Industrial Relations doi: 10.1111/bjir.12497
58:1 March 2020 0007–1080 pp. 50–81
Who Works for Whom and the UK
Gender Pay Gap
Sarah Louise Jewell, Giovanni Razzu
and Carl Singleton
Abstract
This study reports novel facts about the UK gender pay gap. We use a
representative, longitudinal and linked employer–employee dataset for 2002–
2016. Men’s average log hourly wage was 22 points higher than women’s in this
period. We find that 16 per cent of this rawpay gap is accounted for by estimated
firm-specific wage eects. This is almost three times the amount explained by
gender occupation dierences. When we decompose a pre-adjusted measure of
the pay gap,we find less than 1 percentage point or a 6 per cent share is accounted
for by the genderallocation across high- and low-wage firms. In other words, only
a small share of what is traditionally referred to as the ‘unexplained’ part of the
pay gap is explained by the dierences between men and women in whom they
work for.
1. Introduction
The gap between the average hourly pay of all male and female employees
in the United Kingdom stood at 17.1 per cent in April 2018.1This study
contributes to the literatureon the pay dierences between men and women by
assessing what rolefir m-specific wagepremiums had in the United Kingdom’s
pay gap between 2002 and 2016.
To do so, we estimate a wages model which allows for worker and firm-
specific fixed wage eects, following the contributions of Abowd et al. (1999,
2002). In our analysis sample, the observed mean log hourly pay gap (male
minus female) among employees aged 25–64 was 22.3 points.2We find that
16 per cent of this gap is accounted for by estimatedfir m-specific wage eects,
implying that men have a greater tendency than women to work for firms
which, on average, pay higher wage premiums to their employees. Although
Sarah Louise Jewell, Giovanni Razzu and Carl Singleton are at the Department of Economics,
University of Reading.
C
2019 John Wiley & Sons Ltd.
Who Works for Whom and the UK Gender Pay Gap 51
this may not appearto be a large share, to put it into perspective,the equivalent
contribution from the fact thatmen and women work in dierent occupations,
which also have dierentwage premiums, is just 6 per cent. However, the vast
majority of the pay gap is explained by the fixed characteristics of workers,
which aect their wages irrespective of what jobs they are in, such as their
education, preferences and work experience.
We also look at the importance of how workers are allocated to firms in
terms of the adjusted gender pay gap, which is typically reported alongside
the raw gap by commentators and policy makers, since it can account for
dierences in the other observable characteristics of workers or jobs relevant
for wages, such as age, tenure, occupation, industry sector and full-time status.
Wedo this by applying the Gelbach (2016) decomposition. Out of an adjusted
mean log hourly pay gap of 14.5 points, we find that an estimated share of
6 per cent is contributed by the conditional allocation of employees over
firm-specific wage premiums, with the remainder accounted for by the fixed
characteristics of workers, which are transferred across firms.
These findings contribute to a vast literature which has studied the
determinants of gender pay gaps (forreviews, see Altonji and Blank 1999; Blau
and Kahn 2017; Weichselbaumer and Winter-Ebmer 2005). Explanations
for the labour market dierences between men and women are typically
grouped into three categories: productivity, preferences and discrimination,
which are all interrelated (Altonji and Blank 1999). With the diminishing
gender gaps in education and labour market participation in the majority
of developed countries, the importance and focus on explanations from the
first category, especially human capital-based ones, has lessened. Pay gaps
nonetheless persist and are pervasive.More recent work has looked to worker
preferences and psychological attributesfor explanations, due to their impacts
on productivity, choices and beliefs (for reviews, see Azmat and Petrongolo
2014; Bertrand 2011; Croson and Gneezy 2009). The role of firms cuts across
these sets of explanations. Early work found thatUS women were more likely
to work for low-wage firms than men, and vice versa regarding high-wage
firms (e.g. Bayard et al. 2003; Blau 1977; Groshen 1991). More recent studies
have found that low-wage growth within an establishment for women plays a
bigger role in the US gender pay gap than how women are (not) sorted into
high-wage firms (Barth et al. 2017; Goldin et al. 2017).
The only studies to have previously looked at the role of where UK men
and women work, and whether this could explain part of the overall pay
gap, are by Mumford and Smith (2007, 2009) and Drolet and Mumford
(2012). These authors used cross-sectional linked employer–employee data to
disentangle the influence of observableemployer and employee characteristics.
They showed that the proportions of women relative to men in occupations
and firms of dierent types did account for part of the UK gender pay
gap. However, these studies, as well as those which have looked at other
countries using a similar method, were hampered by being unable to address
simultaneously the unobservable fixed heterogeneity over workers and firms
in the determination of wages, that is, they lacked longitudinal linked
C
2019 John Wiley& Sons Ltd.
52 British Journal of Industrial Relations
employer–employee datasets. Therefore, their results cannot be directly
compared with what we find here.
Similarly, our main findings are not directly comparable with the majority
of the recent UK-focused gender pay gap literature, which has mostly used
sources of longitudinal household survey data. This literature could not
control robustlyfor the potential influence of how male and female employees
were allocated across firm types, which could systematically dier by gender
and be correlated with the other determinants of wages. It is plausible that
omitting this factor, which generally explains a significant fraction of overall
UK wage variation, could have confounded previous results. For example,
Costa Dias et al. (2018) demonstrate the importance of accumulated years of
work experience and working hours in determining pay gaps. They find that
among UK college graduates,the majority of the gender pay gap 20 years after
the first childbirth can be explained by dierences in work experience, mostly
through accumulated working hours. But it is well known that the measured
returns to tenure and work experience are likely to be upwards biased unless
the unobserved worker–firm matchquality is controlled for (Topel 1991). Our
main findings show that the way in which employees are assigned to firms
can explain only a small fraction of the pre-adjusted UK hourly pay gap.
Therefore, we add support to some of the previous conclusions about the
determinants of UK gender pay dierences, which were based on household
survey data from a similar period, but which could not have addressed the
potential influence of how workers were matched to firms.
More direct comparisons of our main findings are possible with a few recent
studies of the gender pay gaps in other countries (e.g.Card et al. 2016; Sorkin
2017; see Section 5 for a full comparison with several relevant studies). In
particular, Cardoso et al. (2016) (henceforth CGP) looked at how much of
the Portuguese hourly wage gap could be accounted for by the allocation of
men and women over establishments and job titles. They found that these
two factors could each explain around a fifth of an adjusted measure of
the Portuguese wage gap over three decades. Our methodological approach
is close to that of CGP, although we expand on it. We use the Gelbach
decomposition to identify the role of unobservable worker and firm fixed
factors in the pay gap, after first adjusting for the influence of both time-
varying and fixed observable wage determinants, and not just the former as
in the case of CPG.
There is an active policy context related to this study. UK law recently
changed such that all British employers with at least 250 employees must
annually publish their own gender pay gaps.3Despite the potential significant
economic costs to firms of complying with this legislation, there is no robust
evidence,which is representative of the whole UK labour market,which shows
that the gender pay gap is not an issue of dierences in pay between firms
rather than within them.4It will be impossible to address this evidence gap
using the pay gap data collected from firms under the new legislation. Quite
simply, there is no robust way to address how much of the pay gaps reported
by firms, and the dierences therein betweenfir ms are explainedby workforce
C
2019 John Wiley& Sons Ltd.
To continue reading
Request your trial