Essays in electoral behavior and Bayesian data analysis

摘要: The resurgence of Bayesian statistics in political research and, particular, the rising popularity Markov chain Monte Carlo (MCMC) methods, has unlocked estimation problems long thought to be considered impossible or intractable. Besides opening new terrain methodologists, these developments have allowed scholars explore revisit longstanding puzzles. This dissertation takes advantage generality and power techniques comprising MCMC methods address novel substantive methodological questions about abstention, voter choice turnout misreporting, areas where controversies remain despite rich story academic studies on electoral behavior considerable attention that been paid them. second chapter develops a statistical model jointly analyze invalid voting absenteeism, two important sources abstention compulsory systems had so far not simultaneously examined. I illustrate application using data from Brazilian legislative elections between 1945 2006, underscoring relevant differences determinants both forms non-voting. third presents study Chile’s 2005 presidential elections, examining substitution patterns voters’ preferences over competing candidates highlighting influence candidates' entry exit strategies election results, an aspect received virtually no previous analyses Chilean politics. Finally, fourth correct for misclassified binary responses information auxiliary sources, applies it analysis U.S. While main contribution is methodological, empirical clear implications researchers interested race America.