Observed real wages are not constant over the cycle, but neither do
they exhibit consistent pro- or counter-cyclical movements. This
suggests that any attempt to assign systematic real wage movements
a central role in an explanation of business cycles is doomed to
failure. (Lucas, 1977)
1 Introduction
Recent years have witnessed a marked shift in economists’ views of the behav-
ior of real wages over the business cycle. The prevailing wisdom, emanating
largely from aggregate time series investigations, is that wages are at most
weakly pro-cyclical. Abraham and Haltiwanger (1995) conclude, “ Correcting
for all of the measurement problems, estimation problems, and composition
problems does not lead to a finding of systematically procyclical or counter-
cyclical real wagesT In contrast, evidence based individual-level longitudinal
surveys (e.g. Bils, 1985; Solon et al., 1994) supports the notion that wages
are strongly pro-cyclical. This paper shows that another type of data disag-
gregation adds signihcant new insights into aggregate wage cyclicality. This
concerns observing wage behavior in the frequency domain. We are interested
in real hourly earnings in U.S. manufacturing where earnings are separated
into those dehated by consumer prices and those by producer prices. We hnd
that our frequency approach provides particularly valuable insights if we break
down earnings into constituent parts. We provide a method of earnings de-
composition into the standard hourly wage, the overtime mark-up, and the
proportion of overtime workers in the total workforce.1
To appreciate the potential value-added of employing frequency methods,
consider three cycles of relatively short, medium and long periods. Although
by no means hard and fast or exhaustive representations, these might consist
of (respectively) a wage contract cycle, a business cycle and a product cy-
cle. Each type may associate systematically with the real wage. The relative
strength and direction of the associations may differ, however. The start of a
three-year United States wage contract, for example, may coincide with wage
1To obtain this breakdown we employ unpublished data relating to hours of work provided
by the U.S. Bureau of Labour Statistics. These data are provided on an annual frequency
and are available from 1959 to 1997. Further details on data sources and methods can be
found in Appendix A.3.