UNIONS, INCOMES POLICY AND RELATIVE WAGES IN BRITAIN*
| Author | David Metcalf |
| DOI | http://doi.org/10.1111/j.1467-8543.1977.tb00082.x |
| Published date | 01 July 1977 |
| Date | 01 July 1977 |
British Journal
of
Industrial Relations Vol.
XV
No.
2
UNIONS, INCOMES POLICY AND RELATIVE WAGES IN BRITAIN*
DAVID METCALFt
WAGE differentials excite controversy. There is disagreement on how they are
formed, whether they are fair and whether they can be modified by fiat. Further, no
estimate exists of the efficiency implications of, for example, manipulating wage
differentials via incomes policy or by legislation which encourages the closed shop
or which disbands wages councils. The polemic surrounding the structure of wages
is
not new. Hobsbawm describes how in Victorian Britain the status of the labour
aristocracy was dependent on their wage differential, how the hierarchy of pay
emerged and how ‘employers almost certainly got their skilled labour in the
nineteenth century at less than market cost’.’
Two elements of the wage structure which have received rather little attention
in Britain are discussed here. First, the impact of unions on relative wages is
analysed (Section
I).
Data on the coverage of collective agreements have recently
become available and this has led to
a
number
of
estimates of the wage premium
associated with union coverage. These estimates are, at
first
sight, surprisingly
high, and they vary by type of agreement and by sex. The evidence that unionised
workers extract a wage differential as compared with similar non-covered workers
is drawn on in Section
I1
to examine the impact of incomes policy on the structure
of
relative wages. Conclusions are presented in Section
I11
and the data sources and
definitions for the statistical analysis of Section
I1
are contained in Appendix I.
Appendix
I1
discusses some thorny but important econometric issues.
I. UNIONS
AND
RELATIVE
WAGES
The issue studied here is the impact of unionisation on the structure of relative
wages. There are other strands of the economic analysis of unions which
are
interesting, and which we sometimes mention, but they are not our main current
concern. These include the appropriate model of union behaviour, the nature
of
the collective bargaining process, the impact of unions on income inequality,
the impact of unions on resource allocation, and the effect of unions on the
rate of change of aggregate money wages. Recent American work on these topics
has been surveyed by Johnson.2 The topics are important but are not con-
sidered here because, with the exception of unions and wage inflation, British
economists have paid little attention to them. Further, this paper is con-
cerned with the wage structure itself, rather than all facets
of
the economic
analysis of unionism.
1.
Methodology
The methodology for estimating the impact of unions on relative wages and the
conceptual problems involved in such estimates are set out in Lewis3 and in more
simplified form in Pencavel and M~lvey.~ Each analysis proceeds from
a
simple
identity. The logarithm of the geometric average wage, W, may be thought
of
as
a
*
I
acknowledge, with thanks, helpful discussions and/or comments on an earlier draft from
Orley Ashenfelter, John Creedy, John Flemming, John Gennard, Peter Hart, Richard
Layard, Charles Mulvey, Stephen Nickell, John Pencavel and Mark Stewart, and research
assistance
from
John Kun.
t
Professor of Economics, University College, Cardiff.
157
158
BRITISH JOURNAL
OF
INDUSTRIAL RELATIONS
geometric weighted average of the logarithm of the geometric average union wage,
W: and the logarithm of the geometric average non-union wage, W",
so
that
W =W(l-T)+W"T
W
"+
(W" Wn)T
=Wn+DT
where T is the fraction of workers who are union members or who are covered by
collective agreements and D=ln(l+M) is the logarithm of unity plus the propor-
tionate uniodnon-union wage differential. W and T are obtained from published
data. W"is replaced by a vector of variables which determine it (e.g. skill, experi-
ence, age) and M
is
then estimated from regression analysis.
It
is
M
we are
interested in here-the proportionate uniodnon-union wage differential (or
mark-up) for otherwise homogeneous lab~ur.~
Log of earnings is normally used
as
the dependent variable because proportion-
ate differentials have more intuitive appeal and because if we wish to make
comparisons of M over time or over local labour markets or across skill groups, the
proportionate mark-up makes more sense than the absolute mark-up (in
E
or
pence) which is calculated if the dependent variable is
not
entered in logs. Further,
if the data used to estimate M is from industry average wages and if the distribu-
tion of wages within each industry is log normal then log earnings is more 'rep-
resentative' than earnings not in logs.
M
can be estimated using two types ofdata in three different combinations. First,
cross-section data on average unionisation level, average pay, average skill mix,
etc. by industry or occupation or local labour market can be used. Second, the data
can refer to individuals. Third, a mixture of individual and industry/occupation
averages can be used. In this last case the observations are individuals but because
the researcher does not know whether the individual is a union member or is union
covered
or
not, the extent of unionisation in the individual's industry or occupation
is used
to
calculate M. While these latter studies overcome the problem of the
simultaneous determination of union membership and wage levels, they suffer
from an omitted variable problem, which
is
discussed in Appendix
11.
It
is important
to
interpret
M
correctly. Where the data refer to individuals and it
is known whether or not a person is unionised/covered, the interpretation is
straightforward: the individual who is
a
union mernberlcovered receives a propor-
tionate wage differential of
M
per cent as compared with an otherwise identical
person who
is
not a union memberlcovered. Unfortunately this type
of
calculation
has not been made for Britain because of data problems. Data which indicate
whether a person is
a
union member/covered (e.g. raw
1973
New Earnings Survey
(N.E.S.)
data on coverage of collective agreements)
do
not simultaneously give
personal characteristics such
as
age, skill and experience. Data which are good on
personal characteristics (e.g. the annual General Household Survey) do not indi-
cate whether an individual is
a
union memberlcovered or not.
In the studies discussed below the unionisation variable usually refers to union
coverage by industry. This causes a problem in interpreting M which is sometimes
neglected.
M
may be interpreted in two similar ways. We may say that the average
wage in
an
industry which is
100
per cent covered
is
ceteris paribus
M per cent
greater than an industry
0
per cent covered. Alternatively,
a
worker in the 100 per
cent covered industry earns
M
per cent more,
ceterisparibus,
than a worker in the
0
per cent covered industry. The problem is that we seldom observe an industry with
no coverage or with complete collective agreement coverage. Alternatively, there-
fore, one may say that the wage
in
an industry with
90
per cent coverage
is
MI2 per
cent greater,
ceteris paribus,
than that in an industry with
40
per cent coverage.
To continue reading
Request your trial