Political Methodology Comprehensive Examination
Department of Political Science
The George Washington University
May 2005
Part I. To be completed in 5 hours (open book)
1. Use the provided Stata dataset, which contains data from the 1996 American National Election Study, to answer the following questions:
a) Fit a probit model in which the response variable is vote choice and the covariates are age, educ, and income. Interpret the results both in terms of substantive effects and significance.
b) Are the coefficients on age and educjointly statistically indistinguishable from 0? Explain how you test this.
c) Test the hypothesis that people who are both young and well educated are disproportionately likely to vote in favor of Clinton. Explain how you did this.
d) Suppose we instead want to explain the effect of age, educ and income on party identification. How do we do this? Interpret the results both in terms of substantive effects and significance.
e) What happens to the parameter estimates and standard errors if the data are heteroskedastic? How could you address problems resulting from heteroskedasticity?
2.Since the late 1990s, the Department of Energy has used polygraph machines to screen for potential spies who should not be eligible to have access to classified material. Assume that the polygraph machines have two percent false positive and false negative rates, where false positives are employees identified as spies who are not spies and false negatives are employees not identified as spies who are spies. Further assume that one out of 1000 employees actually is a spy. What is the probability that an employee identified by the polygraph as a spy actually is a spy?
3.You have a measure of ideology for members of the House of Representatives, with data for two years. Both year 1ideology and year 2 ideology have a mean of 50 and a standard deviation of 10. The correlation between the two variables is .90. If a member has an ideology score of 55 in year 1, what would be her predicted ideology for year 2?
4.You are interested in assessing the relative impact of challenger spending and incumbent spending in House elections (i.e., proportion of the vote going to the incumbent). Assume the following equation is estimated:
Explain how you could test whether a dollar spent by an incumbent is cancelled out by a dollar spent by a challenger.
What regression assumptions are likely to be violated given the nature of challenger and incumbent spending? What, if any, other assumptions are likely to be violated, given the substance of the problem?
5.You are conducting a cross-sectional country-level study of voter participation, with your dependent variable as the country level participation rate, with a theoretical range from 0 (no one votes) to 1 (everyone votes). If you simply estimated a model using OLS, what assumption necessary for Gauss-Markov to hold would you be violating? How could you transform your data to avoid violating this assumption?
6.Refer to the figure on the following page, where the payoff for player A is the first figure in each pair, and the payoff for player B is the second. Assume that the form of the game is common knowledge to both players, and that each player seeks to maximize his/her individual payoff.
(a) Find both the Nash equilibria for this game in its normal form and the subgame perfect equilibria for this game, and state them using the usual terminology. Hint: remember that a strategy is a complete plan for playing the game.
(b) If you find that there are Nash equilibria that are not subgame perfect equilibria, are these equilibria more ‘credible’ than those Nash equilibria that are also subgame perfect equilibria? Are they less credible? In either case, state why, in a one-paragraph answer.
Part II. Submit an empirical research paper along with the exam.
Good Luck!