| This work presents four independent essays applying statistical methods of causal inference to examine puzzles of interest in law and political science.; Part One, consisting of the first two essays, examines the robustness of causal claims in observational studies. The first essay reconsiders a claim in positive political economy on the causal effect of electoral systems on price levels. It illustrates that traditional parametric methods may identify causal effects solely on functional form, and shows how matching methods may provide more robust estimates.; The second essay assesses the impact of war on U.S. Supreme Court decision-making. It finds that war is associated with a substantial decrease in the probability of a liberal decision in civil rights and liberal cases. Contrary to received wisdom, the war affects cases not directly related to the war, but little evidence suggests that war impacts war-related cases.; Part Two, consisting of the last two essays, takes an experimental approach, capitalizing on randomization to draw credible causal inferences for a natural experiment and a survey with a randomized instrument. The third essay examines the natural experiment of the California alphabet lottery, which has randomized the ballot order of candidates for statewide offices since 1975. It finds that ballot order substantially affects all candidates in primary elections, and that ballot order may have changed the winner in over ten percent of all primary races from 1978 to 2002.; The final essay illustrates a Bayesian generalization of instrumental variable estimation to assess the effect of racial perceptions on political knowledge surveys. It finds that contrary to previous analyses, the race of an interviewer in a telephone survey affects political knowledge answers, irrespective of the stated perception of race, and that this effect is more substantial for white respondents than for black respondents. |
