Structural Estimation and Experiments: Applications to Contracting Models

Structural estimation requires the specification of behavioral models that can be used to conduct ex ante policy evaluations and welfare analysis. Experiments generate data by exogenously varying key variables to measure outcomes under various treatment conditions. Gains from combining experiments and structural estimation can be considerable. We illustrate these gains through two recent papers (Bellemare and Shearer, 2011, 2013) in the area of contracts and compensation systems. In both papers, the combination of structural modeling and experiments is essential – experiments cannot be conducted to implement all possible treatments of interest, while structural estimation using naturally occurring (payroll) data either is infeasible or requires restrictive modeling assumptions.