In their 2009 paper, Ferreira and Gyourko study whether political parties matter, in terms of policy outcomes, using a regression discontinuity design. The discontinuity comes from the fact that there can only be one elected party, the one with the majority of the votes. a. What hypothesis would one need to argue the regression discontinuity design approach is valid here? The incumbent effect Regression (2) in the paper is: I Doc+1 = 10 + D1% + MV6,411 + MV 12 + MV 13+ De MVoda + De MVA+ De, MV A6 + Veit Where Dee is whether the Democrat party won the election in yeart and MV., is the margin of vote of the Democrat party. This is the plot of this regression: 1 . .8 .6 Democratic prob. of victory, t +1 4 0.4 05 0.6 -0.6 -0.5 -0.4 -0.3 -02 -0.1 0 0.1 02 0.3 Democratic margin of victory, election FIGURE I Incumbent Effect The estimate of is 0.323, associated with a robust) standard error of 0.055. b. Interpret what this estimate means. e. Is it statistically significant? Policy outcomes Regression (1) in the paper describes their general regression: So,4 = Bo + Dc,671+ MV.Bu + MV B2 + MV3,33+ De, MV, BA + Do MV3.Bs + De MV Bo + ce Where Set is a policy outcome of interest. The following table summarizes their results for a variety of outcomes. Compare the outcomes from columns (3) and (5). What do you make of it? TABLE II OLS AND RD ESTIMATES OF THE IMPACT OF A DEMOCRATIC MAYOR (1) % diff. between Dem and Rep mayors Average OLS OLS RD RD (std) uncond. conditional cubic linear Dependent variables (2) (3) (5) Size of government Total revenues per capita ($) 1,082 0.129 0.058 -0.016 -0.014 (676) (0.029) (0.022) (0.022) (0.013) Total taxes per capita (8) 852 0.160 0.091 -0.013 0.008 (678) (0.033) (0.024) (0.021) (0.012) Total expenditures 1,067 0.131 0.060 -0.009 -0.015 per capita ($) (652) (0.029) (0.022) (0.021) (0.013) Total employment per 15.25 0.169 0.087 0.017 0.014 1,000 residents (9.52) (0.035) (0.028) (0.016) (0.011) Allocation of resources % spent on salaries and wages 0.61 0.007 0.012 0.020 0.007 (0.12) (0.006) (0.006) (0.014) (0.008) % spent on police department 0.20 -0.011 -0.003 -0.001 0.003 (0.08) (0.004) (0.004) (0.007) (0.004) % spent on fire department 0.13 -0.004 -0.001 0.006 0.006 (0.05) (0.003) (0.003) (0.005) (0.003) % spent on parka 0.19 -0.023 -0.009 0.011 0.009 and recreation (0.17) (0.009) (0.007) (0.014) (0.009) Crime indices Murders per 1,000 residents 0.08 0.019 0.008 0.005 0.011 (0.09) (0.006) (0.004) (0.007) (0.005) Robberies per 1,000 residents 2.06 0.824 0.454 0.597 0.619 (3.70) (0.200) (0.186) (0.338) (0.288) Burglaries per 1,000 residents 15.54 0.948 0.194 0.572 1.579 (12.40) (0.780) (0.732 (1.024 (0.735) Larcenics per 1,000 residents 41.49 1.923 1.389 1.798 5.424 (27.81) (1.718) (1.700 (2.489) (1.869) Covariates No Yes Yes Yes Node. Column (1) presenter and standard deviation for all independent variables, while Columna (215) report coeficients from OL and RD region of each independent variable indicated in the table an indicator variable for whether the mayor is Democrat and other controle. The RD pecification has other controle for margin of victory as described in equation (1) in the last All ofernt variables were transformed to be the wet of covariates includes city population, the type of election partnera nonpartisan, length of term status median income, percentage of white householda, texts are idade. Columns and indude a control for the respective dependent van of the year prior to the election See the lot for a more detailed planation of the fiscal and crime van while the number of crime for total employment and crime indices are 1,463 and 1,720, spect where 1 A16 is the relevant for all other varies Reported andardrer are dured by city and decade