Measuring the cyclicality of real wages: how important is aggregation across industries?

There is a growing consensus among economists that real wages in the postwar U.S. have been moderately to strongly procyclical, particularly in panel data on workers. From the point of view of hiring decisions of firms, however, this conclusion may be premature or even erroneous. Whether a firm's labor demand curve is stable or shifting at business cycle frequencies should be tested with a wage that is deflated by the firm's own price of output, with appropriate controls for the prices of intermediate inputs, and with respect to the cyclical state of the firm's own industry, as opposed to the state of the aggregate economy. I find that failing to control for these factors has led to a substantial procyclical bias in previous estimates of wage cyclicality. In two-digit and four-digit level (SIC) industry data on wages, with controls for changes in worker composition, I find that a substantial majority of sectors have paid real product wages that vary inversely (i.e., countercyclically) with the state of their industry.