A model's ability to explain procyclical movements in real wages has become an important benchmark by which macroeconomists judge business cycle theories. Because Keynesian models with sticky nominal wages predict countercyclical real wages, they have been criticized and dismissed in favor of Real Business Cycle models or New Keynesian models based on price stickiness or countercyclical markups. The bulk of the evidence for procyclical real wages, however, comes from studies using panel data that estimate the unconditional, contemporaneous correlation between real wages and the unemployment rate. These studies constrain real wage cyclicality to be the same irrespective of the source of the business cycle fluctuations. This paper relaxes this constraint and estimates a structural VAR identified by long-run restrictions on the responses of hours and output to labor supply, technology, oil price, and aggregate demand shocks. It finds that real wages are procyclical in response to technology shocks and oil price shocks, but are countercyclical in response to labor supply shocks and aggregate demand shocks. The procyclicality of real wages during the periods covered by the panel data sets may be explained by the importance of the productivity slowdown and the 1970s oil price shocks. The results highlight the limitations of using the unconditional, contemporaneous correlation between real wages and business cycle indicators to sort out competing theories of the business cycle, and cast strong doubt on the appropriateness of the rejection of sticky wage models.

Empirically, real wages exhibit relatively little cyclical variation and a weak cyclical pattern. Early real business cycle (RBC) models predict, to the contrary, large, procyclical real wage movements. Incorporating efficiency wages into a RBC environment would seem promising since one prediction from the efficiency wage literature is real wage rigidity. This paper evaluates a common microfoundation for efficiency wages, the shirking model, with respect to its predictions for real wages within a RBC‐style model. Simulations of the model reveal that it can generate dampened but still strongly procyclical real wage behavior.