Inference on semiparametric multinomial response models

We explore inference on regression coefficients in semiparametric multinomial response models. consider cross?sectional, and both static dynamic panel settings where we focus throughout under sufficient conditions for point identification. The approach to identification uses a matching insight all three models coupled with variation regressors: cross?section data, match across individuals while within over time. Across models, relax the Indpendence of Irrelevant Alternatives (or IIA assumption, see McFadden (1974)) allow arbitrary correlation unobservables that determine utility various alternatives. For cross?sectional model, estimation is based localized rank objective function, analogous used Abrevaya, Hausman, Khan (2010), presents generalization existing approaches. In data settings, rates convergence are shown exhibit curse dimensionality number results model generalize work Honoré Kyriazidou (2000) cover case. A simulation study establishes adequate finite sample properties our new procedures. apply estimators scanner set.