Microeconomic Adjustment Hazards and Aggregate Dynamics

The basic premise of this paper is that understanding aggregate dynamics requires considering that agents are heterogeneous and that they do not adjust continuously to the shocks they perceive. We provide a general characterization of lumpy behavior at the microeconomic level in terms of an adjustment hazard function, that relates the probability that a unit adjusts to the deviation of its state variable from what would be its optimal level if frictions were momentarily removed. We argue that adjustment hazards that are eventually increasing with respect to the magnitude of this deviation are likely to be realistic. This allows for testable restrictions and a simple comparison with the partial adjustment model, which corresponds to the constant hazard case. We show how non-constant hazards - in particular, increasing hazards - generate non-linearities and history dependence in aggregate equations. We present an example based on U.S. Manufacturing employment and job flows, and find that increasing hazard models outperform partial adjustment models in describing aggregate employment dynamics; the improvement is most notorious during deep recessions and brisk expansions