Identification and Estimation of Demand Models with Endogenous Product Entry and Exit

Abstract

Firms introduce products in markets where they anticipate stronger demand, using information unobserved by researchers. This creates endogenous selection in demand estimation. In differentiated-product oligopolies, multidimensional demand unobservables and strategic entry can render the ordinary propensity score insufficient for selection correction. Existing approaches either restrict firms' information at entry or jointly estimate demand, pricing, and entry under strong supply-side assumptions. We derive a new mixture representation of the selection-bias function using latent propensity scores: entry probabilities conditional on observables and a latent market state generating dependence across entry decisions. This representation yields a convenient two-step semiparametric estimator that corrects for selection and price endogeneity while accommodating richer information at entry. The approach makes weaker supply-side assumptions and is simpler to implement because it avoids repeatedly solving the full model. Applied to the US airline industry, the method yields more elastic demand and less market power than estimates ignoring endogenous entry.