Spatial and Temporal Aggregation in the Dynamics of Labor Demand
The paper demonstrates the general difficulty of inferring the structure of adjustment costs from aggregated, including industry data, except in the unlikely case that costs are symmetric and quadratic at the micro level. The implications of this difficulty for cross-national comparisons of adjustment costs, and for attempts to infer the structure of these costs without micro data, are examined. In the voluminous literature on dynamic labor demand studies based on annual data generally find longer lags than those that use quarterly data, which in turn produce longer lags than models estimated using monthly data. However, when a consistent set of U.S. industry time series is used, and quadratic symmetric costs are assumed, the estimated length of the lag is independent of the frequency of observation. This conclusion is clearly not general: If we assume the costs of adjusting labor demand are lumpy, inferences about their structure differ greatly depending on how often the data are observed.