298
Journal of Applied Economics
Survey (LFS). They find a decline in the average public sector premium in the post-
devolution period but also show that the pay gap varies along the earnings
distribution. The authors argue that bringing public sector pay in line with private
sector wages may therefore increase efficiency but may also increase earnings
inequality.
In this paper, new data from the BHPS is used to analyse wage differentials in
the public and private sector in Scotland for the post-devolution period. In contrast
to other UK-wide and regional studies the paper accounts explicitly for possible
sample selection bias from both sector choice and labour participation and explores
its impact on the pay gap.1
The paper is structured as follows. The next section will describe the data set
and provide some descriptive statistics. Section III outlines the econometric
framework and discusses identification issues. Finally, Section IV reports the results
on wage equations and decomposition and the final Section discusses policy
implications.
II. Data
The data is drawn from the BHPS. Since its introduction in 1991, over 5,000
households made up of roughly 10,000 individuals have been interviewed annually.
While it has always been a nationally representative sample, only since 1999
extension samples for Scotland and Wales have been launched, aiming to increase
the relatively small sample size - approximately 500 households in each country - to
1,500 households. The main objective has been to enable independent analysis of
the two countries on a representative level.
The sample of employees contains only individuals who are, at the date of the
interview, full-time employees, aged 16 to 64 (16 to 59 for women), not self-employed
and not working in either agriculture, non-profit organisations or for the armed
forces. Since the econometric framework requires identification variables which
are only available for some waves, the paper makes use of data from wave 10 in
2000.
The resulting overall cross-section sample consists of 1,054 males and 1,230
females of which 61% of males and 42% of females are participating in the labour
1 Even though Bender (2003) controls for sample selection from the sector choice, the
selection term is not separately stated in the decomposition and it is also unclear whether that
component is part of the explained or unexplained part which might have a significant impact
on the results (Gyourko and Tracy 1988).