COMMENTS
| DOI | http://doi.org/10.1111/j.1468-0084.1989.mp51002002.x |
| Published date | 01 May 1989 |
| Author | Andrew J. Oswald,David G. Blanchflower |
| Date | 01 May 1989 |
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 51,1(1989)
0305-9049 $3.00
COMMENTS
David G. BlanchflowerandAndrewJ. Oswald
I. INTRODUCTION
The innovative paper by Boyer et al. (1989) argues that house prices play an
important role in the wage determination process in Great Britain.1 The
authors also suggest that it is the change in unemployment, rather than the
level of unemployment, which has the dominant impact upon pay. If correct,
these findings have important microeconomic and macroeconomic
implications.
The authors' results emerge from a time-series analysis of data from the
late 1950's to the mid 1980's. The large group of independent variables
means that the number of degrees of freedom is unusually small, and this
makes it difficult to draw reliable inferences.
Our purpose in this paper is to use cross-section methods to investigate the
ideas explored by Boyer et al.2 We use the new British Social Attitude Survey
series of 198 3-86, which provide information, for a recent and interesting
period, on a sample of almost four thousand British workers. The Surveys
record the earnings and the personal and workplace characteristics of the
individuals. We graft on to this data set external information on regional
unemployment and house prices.
The Appendix describes the data. A more detailed discussion, and further
analysis, is available in Blanchflower (1988). That paper documents the
effects on wages of many variables which, because they are of marginal
relevance to the case proposed by Boyer et al. (1989), are included in later
regressions but not reported or discussed here.
II. EMPIRICAL ANALYSIS
Boyer et aL explore a wide range of issues and we do not attempt to comment
upon them all. Their conclusions about the influence of trade union power,
for example, are not particularly controversial (see the microeconomic
evidence in Stewart (1983), Blanchflower (1984, 1986) and Blanchftower,
'Carruth and Oswald (1989), following Boyer et aL, also find a house price variable to be
significant - with an elasticity of approximately 0.2 - in a time-series British real wage
equation. Their specification includes a variable for profit per employee.
2analysis assumes that regions have their own distinct labour markets. Boyer et al. make
a similar assumption.
137
138 BULLETIN
Oswald and Garrett (1988), and the large US literature surveyed in Lewis
(1985)). We also here leave to one side the interesting question of how
measures of regional dispersion - of house prices or excess demand - affect
the equilibrium wage. The exact role of labour mobility in the pay determina-
tion process is also ignored.
This still leaves many questions unanswered, and four are of particular
interest. First, do wages respond to the lagged level of house prices?3 Second,
is it primarily rises in the unemployment rate, rather than high levels of
unemployment, which dampen wage pressure? Third, how large an impact
does the level of unemployment have upon pay? Fourth, is it the case, as
Nickell (1987) and Layard and Nickell (1988) have argued, that the existence
of a high proportion of long-term unemployed increases pressure on real
wages? To examine these questions we estimate wage equations on individual
data and use regional variables to capture the effects of the price of housing
and of unemployment.
Table 1 gives the results. The dependent variable is annual earnings (a
switch to wages per hour makes no substantial difference). The unemploy-
ment and house price variables are each defined for 11 regions in each of
four years.
Equation (1) includes a set of standard human capital and other personal
and workplace control variables, plus the natural logarithm of the unemploy-
ment rate in the sampled individual's geographical region. The equation also
includes a dummy variable for London. The unemployment rate enters with
an elasticity of approximately - 0.1, which is close to that estimated both in
many recent studies (see the summary in Oswald (1986) and the new results
in Nickell and Wadhwani (1987) and Blanchflower, Oswald and Garrett
(1988)) and in some of Boyer et al.'s results. However, the addition of a full set
of regional dummies, as in equation (7), eliminates this finding. It seems that
unemployment works successfully in equation (1) because the unexplained
inter-regional pay structure is negatively correlated with the inter-regional
unemployment structure. Regional fixed effects apparently dominate.4
Equations (5) and (8) report the same exercise for the price of housing.
The (lagged) house price is highly significant in equation (5) and has an
elasticity close to the figure - after appropriate adjustment - estimated by
Boyer et al. This provides some corroborative evidence for the authors'
central argument. Nevertheless, the result in equation (5) is not robust to the
insertion of regional dummy variables (see equation (8)). As with the
unemployment variable, all that can definitely be concluded is that the
geographical structure of wages is well proxied by the geographical structure
of house prices. Once again, regional fixed effects are important.
follow Boyer et al. we use a two year lag on house prices. The results were similar when
current house prices were used.
4i (4) in Table 1 includes the regional dummy variables without any unemployment
or house price variables.
To continue reading
Request your trial