ArticlePDF Available

The public–private sector wage differential in the UK: Evidence from longitudinal employer-employee data

Abstract and Figures

If fiscal policy exerts pressure on public services, then attention often falls on the public–private sector wage differential. Estimated with longitudinal employer–employee data for the years 2002–2016 in the United Kingdom, among men there was no significant public sector wage premium. However, women received an average 4% premium compared with working in private sector firms.
Content may be subject to copyright.
The public-private sector wage dierential in the UK:
Evidence from longitudinal employer-employee data
Carl Singleton*
School of Economics, The University of Edinburgh,
31 Buccleuch Place, Edinburgh, EH8 9JT, UK
Published at: Economics Letters. (2018; In Press). doi.org/10.1016/j.econlet.2018.11.005
Abstract
If fiscal policy exerts pressure on public services, then attention often falls
on the public-private sector wage dierential. Estimated with longitudinal
employer-employee data for the years 2002-16 in the United Kingdom, among men
there was no significant public sector wage premium. However, women received an
average 4% premium compared with working in private sector firms.
Keywords: public sector premium, firm-specific wages, gender
JEL codes: J31, J38, J45
*Permanent address: Department of Economics, University of Reading, Edith Morley, Whiteknights
Campus, Reading, RG6 6UB, UK, c.a.singleton@reading.ac.uk
1 Introduction
In 2010 the UK government imposed caps on nominal wage increases throughout the
public sector. This policy was relaxed in 2017 as real wage growth in the private sector
recovered (see Cribb (2017) for a summary). The UK is not the only country to restrain
public sector pay while attempting fiscal consolidation. Such policies are commonly
justified by the supposed existence of large public sector wage premiums which are not
allocative, due to the bargaining power of trade unions (e.g. France in 2017 and the United
States in 2019). However, pay-setting restraint could aect the ability of organisations to
hire and retain employees, having implications for services provision. Wage policies in
the public sector can also aect the level and volatility of unemployment in the wider
economy (e.g. Gomes,2015). Therefore, it is important that public sector wage premiums
are measured robustly.
The unobserved quality of workers across sectors is likely endogenous to any wage
dierential (Nickell and Quintini,2002), biasing ordinary least squares (OLS) estimates
from cross-sectional data of the public sector premium. Privatisation has been used as a
‘natural experiment’ to circumvent this (e.g. Haskel and Szymanski,1993). Estimates
were later improved using longitudinal household survey data, which could address
selection on worker unobservables, typically finding larger public sector premiums (e.g.
Disney and Gosling,1998). However, estimates from these data will also be confounded if
worker mobility between sectors is related to dierences in other wage-relevant employer
fixed characteristics. For example, Chatterji et al. (2011) used UK cross-sectional
employer-employee data to suggest that gender and sectoral wage dierences are related
to certain workplace characteristics. These included performance-related pay and
family-friendly working, which were more common in the private and public sectors,
respectively.
Using longitudinal employer-employee panel data and estimating gender-specific
Abowd et al. (1999) (henceforth AKM) wage equations, this article revisits the UK public
sector premium.
2 Method
Let there be i= 1,...,N workers, j= 1,...,J firms, t= 1,...,T years, Tiyears per worker
and N=PiTitotal observations. The AKM-type wage equations are given by:
wit =αi+φG(i)
J(it)+x0
itβ
β
βG(i)+εit ,(1)
where wit is the log real hourly wage. xit is a vector of time-varying observable worker, job
or firm characteristics. Firm fixed eects are given by φG(i)
J(it), where J(it) = jindicates the
1
firm employing worker iin year t. Similarly, G(i) = g∈ {Male, Female}indicates gender.
The remaining heterogeneity is in the residual, εit.
Equation (1) can be estimated using least squares with a strict exogeneity assumption:
Eit εit|xit , αi,φG(i)
J(it),G(i)= 0 .
Estimates of the firm fixed eects will be biased if employees switch firms according
to some component of εit , such as transitory shocks to firm-wide or worker-firm
match-specific eects. However, Jewell et al. (2018) find this exogenous mobility
assumption is not obviously rejected in these UK data.1
The worker fixed eects are transferable, aecting wages wherever and whenever
an employee works, and in whatever job. The firm fixed eects measure relative wage
premiums, which employees receive upon moving between firms. Estimates of these
eects are only comparable within mobility groups, i.e. connected sets of workers and
firms (Abowd et al.,2002). Therefore, the analysis focuses on the largest identified
mobility groups of men or women in the panel dataset.
Using estimates of Equation (1), the male or female public sector premium is defined
as:
b
δg
AKM =Eit b
φG(i)
J(it)|G(i), J(it)P ublicEit b
φG(i)
J(it)|G(i), J(it)<P ublic,
where P ubli c is the set of employers not in the private sector. This measures the expected
log hourly wage gain for an employee who switches from the private to the public sector,
keeping the characteristics in xit unchanged.
Compare Equation (1) with the reduced form version:
wit =αi+P ublicit δG(i)
W FE +x0
itβ
β
βG(i)+rit ,(2)
where P ubli cit =1J(it)P u bli c and rit is the composite error term. Least squares estimates
of the public sector premium from Equation (2), b
δg
W FE , will be unbiased compared with
b
δg
AKM only if the following orthogonality condition holds:
Eit h(rit ri)(P ubit P ubi)|G(i)i= 0 ,(3)
where bars over variables represent time averages. This will not hold when individuals
move between sectors systematically from high to low wage employers, or vice versa.
1See Card et al. (2013) and Card et al. (2018) for evidence in favour of this assumption for Germany and
Portugal, respectively.
2
3 Data
The data are from the Annual Survey of Hours and Earnings (ASHE), 2002-16. This is
drawn from administrative records each April by the Oce for National Statistics (ONS).
It is a panel of employees without attrition, where each year gives an approximate 1%
random and representative sample of UK employees. The linked employers in these data
are enterprises.2
The analysis looks at the hourly wage rate according to employer payrolls, excluding
overtime. Wages are deflated to 2002 prices using April values of the Consumer Price
Index (CPI).3Only prime-working-age employees, aged 25-64, who have non-missing
records of earnings and hours are considered. Observations with 1-100 basic paid weekly
hours are dropped, as are non-main jobs, those at a trainee or an apprentice level, and
any jobs incurring a loss of pay in the reference period for whatever reason. Unless stated
otherwise, xit contains the following job characteristics: squared and cubed employee age,
cubic polynomials for tenure and employer size (number of employees), whether a job is
part-time (thirty hours or less), and 3-digit (92 categories) occupations. xit also includes
year fixed eects. The private or public status of an employer is from administrative
records.4See Jewell et al. (2018) for further details on how this dataset can be constructed
from ASHE cross-sections.
The estimation sample contains only person-year observations which are in the largest
male or female mobility groups. This has around 80-90% of ASHE observations each
year, and is generally UK representative, though firm size is marginally higher than in the
population. Table 1gives brief sample descriptive statistics.
4 Results
Estimating Equation (1) in the years 2002-16, the public sector premiums are 0.0 and 3.8
log points for male and female hourly wages, respectively (Table 1, row 3). Estimates
accounting for only worker fixed eects are larger, being 2.9 and 4.5 log points (row
2). This suggests that, when workers move from the private to the public sector, they
tend to do so from relatively low wage employers within the former to relatively high
wage employers within the latter. Pooled OLS estimates (row 1) are consistent with
previous results from the UK, that individuals with a low permanent wage component
2An Enterprise can be defined as ... an organisational unit producing goods or services, which benefits
from a certain degree of autonomy in decision-making ... An enterprise carries out one or more activities at
one or more locations. ONS. In the public sector, schools, hospitals, government departments, the British
Broadcasting Corporation etc. are enterprises.
3Accessed from UK National Statistics, 24/4/2017.
4ASHE contains the tax authority classification: private companies, sole proprietors and partnerships
are classified as private sector. Public corporations & nationalised industries, central government, local
authorities, non-profit bodies or mutual associations are classified as public sector.
3
disproportionately work in the public sector, especially among men. Rows 4-6 of Table 1
present estimates without occupation controls in the wage equations. Comparing the
pooled OLS estimates (rows 1 & 4), the dierences in employee occupations between
the public and private sectors account for a large part of the average wage dierences
between the sectors. The estimated public sector premiums from the wage models with
worker and firm fixed eects are greater when occupation controls are left out (rows 5-6).
This suggests that part of the average wage gain experienced by workers who switch from
the private to the public sector is due to related wage gains from them also changing
occupations at the same time.
Figure 1shows kernel density estimates over employee-year observations of the
estimated firm fixed wage eects, by gender, and comparing sectors. These eects have
tighter variance in the public sector. If working in the private sector, then men and women
had similar tendencies to work in very high wage firms (>15 log points). However, there
is a stark gender dierence in so far as women were more likely to work in very low wage
firms in the private sector (<15 log points).
Public sector employers range from schools to financial services regulators. Table 2
presents estimates comparable to Table 1, row 3 for subgroups of public sector enterprises.
In local authorities, the wage premium compared with the private sector was around
2% for both men and women. In central government, men received a 3% wage penalty,
whereas women received a 5% premium. The largest premiums were for employees in
public corporations and nationalised industries (e.g. National Rail and the BBC).
Figure 2shows estimates of δg
W FE and δg
AKM from the largest mobility groups of men
and women in rolling 7-year windows, along with shares of employment in the public
sector. The wage premium for women was stable over 2002-16. For men the premium
increased by over 6 log points in this period. This was despite the public sector wage
moderation policies of government since 2010.
5 Conclusion
Previous estimates of the UK public-private sector wage dierential have not addressed
unobserved worker and unobserved firm heterogeneity. After doing so, the estimated
average public sector premium received by men between 2002 and 2016 men was zero.
The value among women was 4%.
The importance of firms in frictional wage dispersion has been increasingly studied
in recent years (e.g. Card et al.,2018). This article showcases a valuable source of
linked employer-employee panel data from the UK, which adds to sources from other
countries already extensively studied. Future research should use this to revisit what the
determinants of other UK wage dierentials are.
4
Acknowledgements
This work was based on the Annual Survey of Hours and Earnings Dataset (Crown
copyright 2017), which is funded, collected and deposited by the Oce for National
Statistics under secure access conditions with the UK Data Service (UKDS) (SN:6689). The
use of these data does not imply endorsement of the data owner or the UKDS in relation
to the interpretation or analysis of the data. Comments from colleagues at The University
of Edinburgh are gratefully acknowledged.
Funding: This work was supported by the Economic and Social Research Council (UK)
under Grant No. ES/J500136/1.
References
Abowd, J., R. H. Creecy, and F. Kramarz. 2002. “Computing Person and Firm Eects Using
Linked Longitudinal Employer-Employee Data.” Longitudinal Employer-Household Dynamics
Technical Papers, Center for Economic Studies, U.S. Census Bureau.
Abowd, J., F. Kramarz, and D. Margolis. 1999. “High Wage Workers and High Wage Firms.
Econometrica, 67(2): 251–334.
Card, D., A. R. Cardoso, J. Heining, and P. Kline. 2018. “Firms and Labor Market Inequality:
Evidence and Some Theory.” Journal of Labor Economics, 36(S1): S13 – S70.
Card, D., J. Heining, and P. Kline. 2013. “Workplace Heterogeneity and the Rise of West German
Wage Inequality.” The Quarterly Journal of Economics, 128(3): 967–1015.
Chatterji, M., K. Mumford, and P. Smith. 2011. “The public-private sector gender wage
dierential in Britain: evidence from matched employee-workplace data.Applied Economics,
43(26): 3819–3833.
Cribb, J. 2017. “Public sector pay:still time for restraint?.” IFS Briefing Note BN216, Institute for
Fiscal Studies.
Disney, R., and A. Gosling. 1998. “Does it pay to work in the public sector?.Fiscal Studies, 19(4):
347–374.
Gomes, P. 2015. “Optimal Public Sector Wages.” The Economic Journal, 125(587): 1425–1451.
Haskel, J., and S. Szymanski. 1993. “Privatization, Liberalization, Wages and Employment:
Theory and Evidence for the UK.Economica, 60(238): 161–81.
Jewell, S., G. Razzu, and C. Singleton. 2018. “Who Works for Whom and the UK Gender Pay
Gap?.” ESE Discussion Papers, Edinburgh School of Economics, University of Edinburgh.
Nickell, S., and G. Quintini. 2002. “The Consequences of The Decline in Public Sector Pay in
Britain: A Little Bit of Evidence.” The Economic Journal, 112(477): F107–F118.
Oce for National Statistics. 2017. “Annual Survey of Hours and Earnings, 1997-2016: Secure
Access. [data collection].” 9th Edition SN: 6689, UK Data Service.
5
TABLE 1: Estimates of the UK public sector log wage premium and sample descriptives,
2002-16
Male Female
With 3-digit occ. controls (preferred):
1. Pooled OLS - b
δOLS -0.020*** 0.040***
(0.003) (0.002)
2. Worker fixed eects - b
δW FE 0.029*** 0.045***
(0.004) (0.002)
3. Worker & firm fixed eects - b
δAKM 0.000 0.038***
(0.001) (0.001)
Without occ. controls:
4. Pooled OLS - b
δOLS 0.103*** 0.181***
(0.003) (0.003)
5. Worker fixed eects - b
δW FE 0.044*** 0.057***
(0.004) (0.003)
6. Worker & firm fixed eects - b
δAKM 0.013*** 0.046***
(0.001) (0.001)
Mean (st. dev.) log hourly wage:
Public 2.52 2.31
(0.48) (0.46)
Private 2.40 2.10
(0.57) (0.49)
Share in public sector 0.30 0.53
Share full-time (>30 hours):
Public 0.88 0.57
Private 0.93 0.61
Mean firm size (000s of employees):
Public 28.9 16.4
Private 14.7 25.8
N: person-year obs. 797,835 858,136
N: Persons 119,424 126,980
Public sector firms 5,607 9,690
Private sector firms 43,137 38,517
£2002. Standard errors (in parentheses) estimated robustly using person clusters. Alternative standard
errors estimated using firm-year clusters were of similar magnitude and did not aect inference;
*** statistically significant at the 1% level, two-sided tests.
6
TABLE 2: Estimates of the log wage premium relative to private sector for types of public
enterprise, 2002-16
Male Female
Central government:
Share of all UK employees 0.09 0.19
Mean log hourly wage di. 0.155 0.262
Wage Premium - b
δAKM -0.027*** 0.055***
(0.001) (0.001)
Local authority:
Share of all UK employees 0.11 0.23
Mean log hourly wage di0.127 0.174
Wage Premium - b
δAKM 0.023*** 0.022***
(0.002) (0.001)
Public corp. & nationalised:
Share of all UK employees 0.04 0.02
Mean log hourly wage di0.009 0.262
Wage Premium - b
δAKM 0.047*** 0.167***
(0.002) (0.002)
Non-profit serving hholds:
Share of all UK employees 0.06 0.10
Mean log hourly wage di. 0.119 0.179
Wage Premium - b
δAKM -0.037*** 0.019***
(0.003) (0.002)
£2002. Standard errors (in parentheses) estimated robustly using person clusters;
*** statistically significant at the 1% level, two-sided tests.
7
FIGURE 1: Distribution of estimated firm fixed wage eects, b
φj, by gender: employees in
private or public sector jobs
(a) Male
(b) Female
Notes.- kernel densities estimated with a bandwidth of one log point. Top and bottom 1% of firm eects by
gender and sector not displayed. For each gender, the eects over public and private sector firms together
have mean zero.
8
FIGURE 2: Estimates of the UK public sector wage premium, 7-year rolling sub-periods
(a) Male
(b) Female
Notes.- ‘Year’ gives the mid-point of each estimation window. ‘Raw’ gives the sample mean wage premium.
‘WFE’ gives estimates from wage models with with worker fixed eects. AKM’ refers to estimates from
wage models with worker and firm fixed eects. ‘Share Pub.’ gives the share of sample employees in the
public sector (right axis).
9
Article
Full-text available
Although wage gaps brought about by differences between sectors has been widely discussed, its specificality in the energy industry has received little attention. Due to the nature of the Chinese energy sector as a monopoly and to the influence of the Soviet Union, the wages of public sector employees in the energy industry in China are much less influenced by market mechanisms and are much higher than those in the private sector. This paper is the first article to explore the impact of the public sector on wage gaps in the energy industry using the China Urban Household Survey for 2004, 2008, and 2013. The cross-sectional regression results show a significant and continuous wage premium exists between sectors, although the size of the premium declined slightly over time. The Blinder-Oaxaca decomposition finds unexplainable part and discrimination dominates the wage gaps. The quantile regression results show that education and work experience have significant effects on low- and middle-wage groups. The heterogeneity analysis considers two sub-industries and finds that the wage gaps between the two sectors are much higher in the extractive segment of the energy industry than in the production and supply sector of the industry. The results of the study reveal that market-oriented reforms in China are not yet complete, because the obvious premium on wages in the public sector means they are not determined by market mechanisms. Finally, the paper provides policy recommendations from three perspectives: further market-oriented reform, elimination of discrimination, and education and vocational training.
Article
Full-text available
This paper argues that the study of wage gaps between public and private sector employees is sensitive to the selection of the sample. In the context of Bangladesh, Labour Force Surveys is a dominant source of employment-related data, which is disproportionately inflated with large pool of informal sector employees. Since government jobs are highly formal, the studies on wage differentials should select the groups that are as much comparable as possible on the question of formality. However, employing Oaxaca-Blinder mean decomposition method and Melly quantile counterfactual decomposition method, we find a decreasing trend in public sector wage premium as we impose more restrictions to make the sectors fitting formal. The wage differential even disappears in the entire restriction sample, and it is slightly biased towards private in the top quantile only. Therefore, we can conclude that the superiority of the public sector job does not come from wage compensation but non-monetary issues, with a strong implication for labour markets in Bangladesh.
Preprint
After the latest pay scale (National Pay Scale 2015) in Bangladesh, public sector jobs have demanded significant attention of the prospective job seekers. While literature shows that this might be attributed to the wage premium in public sector jobs, this paper argues that the result is, indeed, sensitive to sample selection. In the context of Bangladesh, Labor Force Surveys is a dominant source of employment-related data which is inflated with a disproportionately large pool of informal non-government employees. Since government jobs are highly formal in nature, the studies on wage differentials should select the groups which are as much comparable as possible on the question of formality. However, employing Oaxaca-Blinder (1973) mean decomposition method and Melly (2005) quantile counterfactual decomposition method, we find a decreasing trend in public sector wage premium as we impose more restrictions to make the sectors fitting as formal. The wage differential even disappears in the full restriction sample and is slightly biased towards private in the top quantile only. Therefore, we can conclude that the superiority of the public sector job does not come from wage compensation rather from non-monetary issues, having a strong implication for labor markets in Bangladesh.
Article
Full-text available
Abstract The aim of this research is to study the wage gap between public and private sector employees in Ecuador. The methodology used is the Oaxaca and Blinder method (decomposition method). The necessary microdata is obtained from the National Survey of Employment, Unemployment and Underemployment (ENEMDU) of the National Institute of Statistics and Censuses (INEC) for the years 2008, 2013 and 2018. The results obtained suggest the existence of a gap between public and private sectors, where public sector wages are relatively higher than those of the private sector. In addition, schooling and experience influence income levels positively, with the latter being the most determining factor. The decomposition of income gaps by sector shows that the wage gap responds to differences in human capital endowments and to discriminatory components of the labour force. The policy implications are aimed at reducing the public-private income gap. JEL code: D31, J31, J38, J45, J7 Keywords: Wage gap; Public-private income; Labour market; Public sector; Ecuador
Article
Full-text available
This study reports novel facts about the UK gender pay gap. We use a representative, longitudinal and linked employer-employee dataset for 2002-16. Men's average log hourly wage was 22 points higher than women's in this period. We find 16% of this raw pay gap is accounted for by estimated firm-specific wage effects. This is almost three times the amount explained by gender occupation differences. When we decompose a preadjusted measure of the pay gap, we find less than 1 percentage point or a 6% share is accounted for by the gender allocation 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 differences between men and women in whom they work for.
Article
Full-text available
I build a dynamic stochastic general equilibrium model with search and matching frictions in order to determine the optimal public sector wage policy. Public sector wages are crucial to achieve effcient allocation of jobs. High wages induce too many unemployed to queue for public sector jobs, in turn raising unemployment. The optimal wage depends on the frictions in the two sectors. Following technology shocks, public sector wages should be procyclical, and deviations from the optimal policy significantly increase the volatility of unemployment.This article is protected by copyright. All rights reserved.
Article
Full-text available
We study the role of establishment-specific wage premiums in generating recent increases in West German wage inequality. Models with additive fixed effects for workers and establishments are fit into four subintervals spanning the period from 1985 to 2009. We show that these models provide a good approximation to the wage structure and can explain nearly all of the dramatic rise in West German wage inequality. Our estimates suggest that the increasing dispersion of West German wages has arisen from a combination of rising heterogeneity between workers, rising dispersion in the wage premiums at different establishments, and increasing assortativeness in the assignment of workers to plants. In contrast, the idiosyncratic job-match component of wage variation is small and stable over time. Decomposing changes in mean wages between different education groups, occupations, and industries, we find that increasing plant-level heterogeneity and rising assortativeness in the assignment of workers to establishments explain a large share of the rise in inequality along all three dimensions. JEL Codes: J00, J31, J40.
Article
Full-text available
This article shows that little of the gender earnings gap in the public and private sectors in Britain are explained by differences in employee characteristics. The differences in workplace characteristics, however, make a significant, sizeable contribution. It is shown that performance related pay and company pension schemes are associated with higher relative earnings in the private sector, whilst increased provision of family-friendly employment practices in the public sector is associated with higher relative earnings for women.
Article
We synthesize two related literatures on firm-level drivers of wage inequality. Studies of rent sharing that use matched worker-firm data find elasticities of wages with respect to value added per worker in the range of 0.05–0.15. Studies of wage determination with worker and firm fixed effects typically find that firm-specific premiums explain 20% of overall wage variation. To interpret these findings, we develop a model of wage setting in which workers have idiosyncratic tastes for different workplaces. Simple versions of this model can rationalize standard fixed effects specifications and also match the typical rent-sharing elasticities in the literature.
Article
This paper uses microeconomic data from the British Household Panel and General Household Surveys to describe how the distribution of pay differs between the public and private sectors in 1983 and in the early 1990s. Separate analyses by gender and education group reveal that it is women and those with intermediate-level qualifications who do best in the public sector. The large differences between the shapes of the conditional (that is, holding age and education constant) distributions of wages in the public and private sectors are demonstrated using quantile regressions estimated separately for each education group. The paper also exploits the longitudinal structure of the data used to assess how much of these differences can be explained by the unobserved characteristics of individuals. JEL classifications: J31, J45.
Article
Following the fall in overall net public investment, the relative pay of most public sector workers in the United Kingdom declined sharply after the mid-1970s. For example, the relative pay of male teachers fell by over 10 percentage points from the late 1970s to the late 1980s. So has this generated a fall in quality? Using age 10/11 test score percentile positions as an indicator, we find that men entering non-manual public sector occupations in the early 1990s had a significantly lower test score position than those entering in the late 1970s. No such falls were exhibited by women. Copyright Royal Economic Society 2002
Article
Most agency-based privatization models are of the owner/manager relation and ignore the labor market. Yet in the 1980s there were profound changes in public sector and privatized firms to wages and employment. The authors develop a bargaining model of the manager/workforce relation that explains employment and wages when firms are privatized, government objectives become more commercial, and markets are liberalized. Using data on fourteen U.K. companies, 1972-88, that were publicly owned in 1972, the authors find: (1) employment fell following the change to more commercial objectives; and (2) wages were only slightly affected by this, but fell if the firm lost market power. Copyright 1993 by The London School of Economics and Political Science.
Article
Using a longitudinal sample of one million French workers and 500,000 employing firms, the authors decompose real total annual compensation per worker into components: observable employee characteristics, personal heterogeneity, firm heterogeneity, and residual. Unobserved personal heterogeneity is a very important source of wage variation. Unobserved firm heterogeneity, while important, is not as important as person effects. Enterprises that hire high-wage workers are more productive but not more profitable. Enterprises that pay higher wages, controlling for person effects, are more productive and more profitable. Person effects explain about 90 percent of interindustry and firm-size wage differences while firm effects explain substantially less.
Computing Person and Firm Effects Using Linked Longitudinal Employer-Employee Data
  • J Abowd
  • R H Creecy
  • F Kramarz
Abowd, J., R. H. Creecy, and F. Kramarz. 2002. "Computing Person and Firm Effects Using Linked Longitudinal Employer-Employee Data." Longitudinal Employer-Household Dynamics Technical Papers, Center for Economic Studies, U.S. Census Bureau.