A heteroscedastic generalized extreme value discrete choice model

Of the commonly used discrete choice models, the probit class allows flexib le covariance structures for disturbances but is computationally burdensome for problems with more than a few alternatives. The generalized extreme va lue (GEV) class, including the widely used legit and nested legit models, h as the advantage of computational ease but suffers in general from the rest riction of homoscedastic disturbances. This article generalizes the GEV cla ss to allow heteroscedastic disturbances across decision makers as well as across choice alternatives. The resulting models include the heteroscedasti c extreme value model as a special case, which is a generalized logit model , with heteroscedasticity across choice alternatives. Particular attention is paid to the heteroscedastic legit and nested logit models because of the ir widespread use in practice. An empirical application reanalyzing data fr om the 1980 presidential election tests the hypothesis of information-induc ed heteroscedasticity across voters and finds support for a heteroscedastic logit model that reveals stronger effects of voter information on the turn out decision than suggested by the original standard legit model in Ordesho ok and Zeng.