Homework answers / question archive / Empirical methods in Economics Exercise 3 Regression Discontinuity Design Divergence theory is a slightly more commonsense view of political actors

Empirical methods in Economics Exercise 3 Regression Discontinuity Design Divergence theory is a slightly more commonsense view of political actors

Economics

Share With

Empirical methods in Economics
Exercise 3 Regression Discontinuity Design
Divergence theory is a slightly more commonsense view of political actors. When partisan politicians cannot credibly commit to certain policies, then convergence is undermined, and the result can be full policy “divergence.” Divergence is when the winning candidate, after taking office, simply pursues her most-preferred policy. In this extreme case, voters are unable to compel candidates to reach any kind of policy compromise, and this is expressed as two opposing candidates choosing very different policies under different counterfactual victory scenarios. Cunningham simplified the model of policy convergence/divergency presented by Lee et al. (2004): two candidates, one Democrat and one Republican, have their preferred policy position. The candidate who gets the minimum needed share of votes would win the election. The voters expect the candidates to choose certain policies and decide which to vote for. The game has multiple equilibria: 1. Partial/complete convergence: Voters affect policies. 2. Complete divergence: Voters elect politicians with fixed policies who do whatever they want to do. Politicians drive policies. If convergence is true, then Republicans and Democrats who just barely won should vote almost identically, otherwise if divergence is true, they should vote differently at the margins of a close race. This “at the margins of a close race” is crucial because the idea is that it is at the margins of a close race that the distribution of voter preferences is the same. And if voter preferences are the same, but policies diverge at the cutoff, then it suggests politicians and not voters are driving policy making. To better understand the theoretical background, you are encouraged to read chapter 6.4 in Causal Inference: The Mixtape. You may also try the example of Stata code in chapter 6.4.1. (https://mixtape.scunning.com/regression-discontinuity.html#replicating-a-popular-design[1]the-close-election) Part 2 Econometric framework To investigate which of the two equilibria is at work, Lee et al. (2004) used a dataset that contains a voting score of US House representatives (politicians) from 1948 to 1990. The voting score indicates how liberal a representative is. For instance, the higher the score is, the more liberal the representative is. The score shows what policy is eventually chosen by the representatives. The score shows the policy position and is the dependent variable. The dataset also includes other important variables of the representatives that can be the independent variables: - A dummy, democrat, that indicates whether Democrat won the election or not. - The variable, demvoteshare, shows the share of votes that went to Democrat. - The variable, id, indicate the group of a representative by the state and district where s/he resided and the year. - There are other interesting variables of representatives included in the dataset. Part 3 Questions Question 1 Practice the commands in the do file RDD.do. Note that to investigate whether policy convergence or policy divergence is at work, we need to check if the policy chosen (voting score) is affected by which party wins the election. From the results of the first regression, discuss which coefficient is of most interest when we check whether policy convergence or policy divergence is at work and what conclusion we may draw from the interested coefficient.

Question 2 When you practice the commands in RDD.do, you may get the graph below. The graph tells the jump of voting score around the cutoff of 50% share of votes for Democrat. That says, the representatives get more liberal in voting when Democrat wins. Modify the codes of drawing the graph in RDD.do to draw the graph that shows the local jump by restricting the sample to the vote share of Democrat between 0.48 and 0.52. Save and report the graph you draw. Briefly discuss how to interpret the graph you draw.
Question 3 Run a local regression by re-running the second regression and restricting the sample to the vote share of Democrat between 0.48 and 0.52. Report the regression results. You may use the command esttab to report the regression results. See the example in RDD.do. Discussion how the results differ from the results of the original second regression in RDD.do.

Question 4 The representatives’ policy choice (voting score, how liberal a representative is) can be also affected by their socio-economic background. Add at least two control variables into the second regression in RDD.do. Report the regression results. You may use the command esttab to report the regression results. See the example in RDD.do. Discuss how the control variables affect voting score. You may find different variables in the dataset, for example, gender, age, education, income, military service, etc