Brief No 313

Brief No: 313
November 2001
ISBN 1 84185 630 4
THE RETURNS TO EDUCATION:
EVIDENCE FROM THE LABOUR FORCE SURVEY
Ian Walker and Yu Zhu
University of Warwick
Introduction
This report is concerned with the effect of education on earnings and, in particular, with the
financial return to education. We use large survey datasets that contain information on education,
earnings and other characteristics, to estimate the relationship between (log) wages and education.
In addition to providing estimates of the average return to education we also show how this return
varies across individuals according to their observable characteristics, and how the return varies for
reasons which we cannot observe.
Key Findings





Our results confirm existing research which suggests that the average rate of return to
education in the UK is high and has shown no tendency to fall over time despite the large increase
in participation in post compulsory education. Our range of estimates suggest that the return is in
the order of 8% for men and 10% for women.
We find that, on average, the returns to a year in higher education are approximately the same as
a year of 16-18 education but that the penalty for not successfully completing the course of
study is large.
The returns to a degree seem to vary considerably by subject of study – consistently high returns
are found in Economics and Management, Health, and Law while consistently low returns are found
in Arts and Education. However, we cannot separate out the effects of degree subject from preuniversity ability differences in our data.
The returns to education seem broadly stable over time across the wage distribution with the
exception of some fall (from around 8% to 6%) for women who are at the very bottom of the wage
distribution – say those with lowest ability. Thus, there is weak evidence here that the expansion
in education participation has resulted in a reduction in returns at the very bottom of the
distribution for women.
We find large variances in returns around the average effects that we estimate. Although we find
no significant change in the part of the variance that arises for reasons which we cannot observe
(say, ability, motivation, social skills, etc.) there seems to be some increase in returns for the
types of individuals for which we estimate quite high returns and a decrease for those where we
estimate quite low returns.
Estimated Returns in the Labour Force Survey
Data
Our empirical specifications are based on the
theory of human capital, whereby individuals
make decisions to acquire human capital, such
as education, up to the point where the returns
to education are driven down to the real return
on other assets. With additional assumptions,
such a framework implies that (log) wages are
linearly related to education (or qualifications)
and a quadratic function of work experience.
Furthermore, the coefficient on education in
such a model (i.e. the effect of a year of
education on wages) can be interpreted as the
financial rate of return to education providing
the only costs of education are the opportunity
costs of the forgone earnings.
The average return and how it changes over
time
Figures 1 show how, using a simple
methodology, the estimated return to a year of
education (controlling for other observable
characteristics) has changed over time: for
women it falls marginally, while for men it is
broadly constant. The scale on the vertical axis
indicates that a return to a year of education
is about 7.5% for men and around 8.5% for
women.
seeing how wages vary with the qualifications
that individuals have achieved. Figures 2 and 3
show how, again controlling for other
observable characteristics in the data, the
returns to qualifications have changed over
time. The figures just show the main
qualifications: GCSE’s, A-levels, and a degree.
The scale indicates that GCSE’s add around
10% to wages, A-levels around a further 15%
for women and 20% for men, while a degree
adds a further 25% for women and 15% for
men. Since a degree takes 3 years and A-levels
2 this corresponds, crudely, to an annualised
return of about 8% for women and 7% for men
– close to the suggestion in Figure 1. One
notable feature of Figures 2 and 3 is that
there is no tendency for the returns to
qualifications to fall despite the large rise in
participation, both post-16 and post-18 that
has occurred over this period. The implication
is that the large increase in the supply of
skilled workers has only just kept pace with
demand.
Figure 2 - OLS Returns to Highest
Qualifications Relative to None – Women
0.6
GCSE
2+ A levels
Bachelor degree
0.5
Figure 1 - OLS Returns to Education Years –
Men and Women
0.4
0.3
0.2
0.1
0.1
0
Return
0.09
1993
0.08
0.07
0.06
Women
Men
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
0.05
Greater sophistication can be achieved by
1994 1995
1996
1997
1998 1999
2000
Figure 3 - OLS Returns to Highest
Qualifications Relative to None – Men
0.6
GCSE
2+ A levels
Bachelor degree
Figure 4 - Returns to a Year of Education –
Men by Decile
0.12
1st
7th
0.11
0.5
3rd
9th
5th
0.1
0.4
0.09
0.3
0.08
0.07
0.2
0.06
0.1
0.05
1993
0
1993
1994
1995
1996
1997
1998
1999
1994
1995
1996
1997
1998
1999
2000
2000
In the report, we compare these results with
similar findings for many countries around the
world. We find that the returns in the UK are
high relative to most other countries.
The variance in returns
We also investigate how the return varies
across individuals. Part of the variance across
individuals might arise because individuals are
different in observable ways – the return may
be higher for women than men, for nonwhites
than whites, etc. Thus we allow for the returns
to
differ
according
to
observable
characteristics. But we also investigate the
extent to which returns vary for unobserved
reasons.
Here our findings are more complicated.
Figures 4 and 5 show how the returns have
varied over time for different parts of the
wage distribution: here split into the bottom
decile (10%), 3rd decile, 5th decile, 7th decile,
and 9th decile. For men a rise in the returns to
the top decile are apparent, but no long term
change elsewhere. In contrast there was little
change for the top decile for women but a fall
at the bottom. These trends are also reflected
in the returns to a degree relative to A-levels
for both groups.
This analysis looks at how returns vary across
individuals according to their place in the wage
distribution. That place is determined by both
observable characteristics (like race, age, etc.)
and by unobservable factors like ability and
chance. Thus, a final strand to our research is
to try to isolate the observable factors from
the unobservable. We do this by estimating a
model which allows returns to differ by both
factors and we find that the contribution of
the chance/ability elements to the variation in
returns across individuals is approximately
constant. Thus, it seems that, the increase in
participation in post-compulsory education has
not implied a growing proportion of low ability
students – else we would have found an
increase in the importance of the unobservable
source of variance.
Figure 5 - Returns to a Year of Education –
Women by Decile
0.12
1st
3rd
0.11
7th
9th
5th
0.1
0.09
0.08
0.07
0.06
0.05
1993 1994 1995 1996 1997 1998 1999 2000
Conclusion
The broad conclusion from our analysis is that:
despite the expansion in higher education that
has occurred in recent years, there is no
significant trend in the average return to
education, which remains high by international
standards. Apart from women at the very
bottom of the wage distribution where returns
have fallen somewhat, there have been no
marked changes in returns across the wage
distribution.
Copies of the full report (RR313) - priced £4.95 are available by writing to DfES Publications,
PO Box 5050, Sherwood Park, Annesley, Nottingham
NG15 0DJ.
Cheques should be made payable to "DfES Priced
Publications".
Copies of this Research Brief (RB313) are available
free of charge from the above address (tel: 0845
6022260). Research Briefs and Research Reports
can also be accessed at:
http://www.dfee.gov.uk/research/
Further information about this research can be
obtained from David Thompson, Room W631, DfES,
Moorfoot, Sheffield S1 4PQ.
Email: [email protected]