Empirical models of decision making in industrial organization and public choice

This thesis consists of three empirical essays. The first two essays address microeconomic questions related to the election and pharmaceutical markets in the context of a discrete choice individual behaviour framework. The third essay applies recent techniques in likelihood asymptotics to the seemingly unrelated regression problem.; The first essay formulates a model of voter party choice and provides a unique application of an estimation technique to address important economic issues related to the empirical voting literature. The individual voter's utility maximization problem is specified in a discrete choice framework using a random coefficients model; the underlying preference parameters are estimated from the aggregate party vote share function estimates. The empirical results demonstrate the importance of political campaign spending during the campaign period; spending not only redistributes voters across parties but it also impacts on the size of the abstaining group of the electorate. Additionally, candidates that held incumbency status and professional occupations were found to be preferred by voters and were thus rewarded at the polls.; The second essay provides a large sample characterization of an anti-ulcer drug market using individual level data. The essay addresses the issue of market power for producers in a price regulated oligopolistic pharmaceutical industry where the market is composed of a number of molecules where several copies of the same molecule exist, each being sold by a pioneer and its licensed competitors. Market demand functions derived from underlying discrete-choice models of consumer behavior are estimated. Two specifications are considered: a mixture of the multinomial and conditional logit as well as a two-level nested logit. The results indicate the existence of little segmentation between the pioneer and its licensed cluster with a correspondingly low level of pioneer market power. Further, the results indicate no substantial substitutability between drugs belonging to different molecules.; The third essay uses recent likelihood analysis theory to examine inference in the seemingly unrelated regression context. The likelihood analysis theory used involves implicit but appropriate conditioning and marginalization for intrinsic measures of departure. Highly accurate p-values are obtained for scalar interest parameters, particularly for the difference between two regression coefficients of primary interest. The results indicate that these methods provide substantial improvement on the first-order likelihood ratio procedures. The central coverage for confidence intervals is improved and the accuracy is retained for very small samples. A simulation and various examples are presented that document this.