This dissertation consists of three relatively independent chapters that study individual choice behavior from various angles. Chapter 1 is aimed at improving the existing discrete choice models. Of the commonly used models, the probit class is computationally infeasible for problems with more than a few alternatives, and the GEV class, including the widely used logit and nested logit models, suffers from the restriction of homoscedastic disturbances. We relax the homoscedasticity restriction on the GEV class to achieve both functional flexibility and computational feasibility. The heteroscedastic logit/nested logit models are of particular practical interest.
Chapter 2 studies voting behavior in mass elections using data from the 1968 and 1980 presidential elections. We discuss theoretical and methodological issues in the specification, comparative study, and empirical testing of the rational voter models, and explore the methodological treatment of voter heterogeneity. While the standard models do not predict voting turnout well, we obtain clear evidenceof strategic voting in the candidate choice decision in three candidate elections. The data suggest voter information as one source of voter heterogeneity which introduces heteroscedasticity. The heteroscedastic logit model developed in Chapter 1 is therefore applied and is shown to outperform the standard logit model and to reveal strong effects of voter information on the turnout decision.
Chapter 3 studies choice behavior in congressional career decisions. Previous research largely focuses on the binary choices of retiring vs. seeking reelection orseeking higher office vs. seeking reelection. Using data from the 80th through the 99th congresses, we rigorously explore the congressmen's choice from all availablecareer options, and discuss the effects of variables on both pairwise comparisons of the alternatives and on the unconditional probabilities of choosing the congressionalcareer options. Our findings suggest that formal positions held and previous vote margins do not figure into House members' career decisions, and being a Republican per se does not encourage progressive ambition. We also seethat a number of factors previously identified as predisposing House members to seek higher office also affect retirement decision.
