Industrial Reorganization: Learning About Patient Substitution Patterns from Natural Experiments
Federal Trade Commission Working Paper No. 329
82 Pages Posted: 26 May 2016
Date Written: May 25, 2016
Despite their widespread usage, little is known about the predictive accuracy of different discrete choice demand models. To evaluate their performance, we use a series of natural disasters that unexpectedly removed hospitals from consumers' choice sets. We compare the model predictions of post-disaster behavior to the benchmark of actual post-disaster consumer behavior. Across our different settings, we find that models that allow for flexible interactions between patient characteristics and unobserved hospital quality perform the best and that it is important to use different classes of models. Further, the use of less accurate models could lead to more lax merger enforcement.
Keywords: hospitals, natural experiment, patient choice, forecasting, antitrust
JEL Classification: C18, I11, L1, L41
Suggested Citation: Suggested Citation