Industry wage differentials:
how many, big and significant
are they?
Kevin T. Reilly and Luisa Zanchi
Economics Division, Leeds University Business School,
University of Leeds, Leeds, UK
Keywords Pay differentials, Industry competitiveness, Labour market
Abstract In this paper, three implementation and interpretation issues are examined associated
with Krueger and Summers’ method for calculating inter-industry wage differentials. The
literature tends to report a less than complete set of industry wage differentials, use the wrong
standard errors, and misinterpret the meaning of the industry wage differentials. The solution to
the first two issues follows from making explicit the restriction that the employment-weighted
average of all industry wage effects is zero, the same restriction that Krueger and Summers are
implicitly imposing on industry wage effects. All industries have thus a wage effect relative to an
average worker net of any industry effect and correct standard errors are available via the Delta
method. Finally, a method is proposed for analysing inter-industry wage differentials as actual
differences between the wage levels expressed in percentage points and not as log points, which is
the current misleading standard. The procedure calculates actual average percentage wage
differences by industry and avoids the distortion in differences across industries that log point
comparisons engender. An application is provided, using the United States Outgoing Rotation
Files of the Current Population Survey for 1989 and 1996, and so updates the work by Krueger
and Summers.
Introduction
Since Krueger and Summers’ (1988) work on industry wage differentials, there
has been a proliferation of studies for specific countries and international
comparisons that examine the existence and size of these differentials[1]. In this
literature, the industry wage effect is measured as the difference between the
least squares coefficient and the employment-weighted average of all estimated
coefficients of the industry dummy variables in the context of a log-linear wage
equation specification[2]. This measure has the double advantage of being
more easily interpretable from an economic point of view (Suits, 1984) and
independent of the arbitrarily chosen omitted industry dummy. This paper is
concerned with three issues, which will improve the implementation and
interpretation of this important innovation in the industry wage differentials
literature.
The Emerald Research Register for this journal is available at The current issue and full text archive of this journal is available at
http://www .emeraldinsight .com/researchregister http:// www.emeraldinsigh t.com/0143-772 0.htm
The authors thank the participants at seminars held at the University of Manchester and
Universita
`
di Padova and at the conference “Econometrics of Wages” organised by the Applied
Econometrics Association, Universite
´
Libre de Bruxelles, 28-29 May 2002, for comments
received. All errors remaining are the authors’ responsibility.
How significant
are wage
differentials?
367
International Journal of Manpower
Vol. 24 No. 4, 2003
pp. 367-398
q MCB UP Limited
0143-7720
DOI 10.1108/01437720310485906