INTER-INDUSTRY WAGE DIFFERENTIALS: Empirical Evidence


The theory of effort can explain a large number of empirical facts including wage differences across industries, productivity differences across countries, and the limited capital flow from high wage to low wage regions. The purpose of this paper is to breathe more life into this theory by showing that the US labor market does seem to have a wage-effort offer curve. Two data sets are employed to this purpose, industry level data from the Bartelsman-Becker-Gray Manufacturing Productivity Data (hereafter NBER)5 and data at the worker level from the March Current Population Surveys (hereafter CPS).

We will assume initially that all workers are identically productive and that there is a single wage-effort curve with higher wages compensating for higher effort levels. This wage-effort offer curve is indexed by the capital intensity of the sector, with the high-effort, high-wage contracts occurring in the capital-intensive sectors. Workers may have the same tastes and therefore be indifferent among the wage-effort contracts that are formed, or workers may have different attitudes towards effort with the industrious (or materialistic) choosing high effort-high wage contracts and with the slothful (humanistic) workers choosing low effort-low wage jobs. Later we will allow ability differences proxied by education levels.

Measurement of intensity of effort is the biggest problem we face. Effort is the product of unobservable “intensity” times observable hours. Fortunately, since effort is our dependent variable, measurement errors cause noise but not bias. Although we suspect that hours and intensity of work are positively correlated, it is enough that they are not so negatively correlated as to destroy any positive association between effort and hours.

Capital sharing from shift or temp work might cause us serious difficulties, but it doesn’t. If the same capital К is used by two different workers over the course of a day, then our measure of the capital intensity of the job correctly is equal to Kj2 , since 2 is the number of employees. Likewise, if one worker uses the equipment К for half the year, and another uses it for the other half, then the annual rental cost is also proportional to К/2 , which we appropriately measure because the number of employees is doubled. Thus both shift-work and temp-work are compatible with our theory and with our empirical work. What we don’t allow for are set-up times associated with the handing of the capital from one worker to the next. Nor do we allow for unused capacity that is not “charged” against labor. But we think both of these are relative minor concerns.