Everybody Loves a Winner: On the Mutual Causality of Presidential Approval and Success in Congress
2013; Taylor & Francis; Volume: 40; Issue: 3 Linguagem: Inglês
10.1080/07343469.2013.829891
ISSN1944-1053
Autores Tópico(s)Electoral Systems and Political Participation
ResumoAbstract A large literature argues that approval affects presidential legislative success, but Washington observers often believe that legislative success leads to higher presidential approval ratings, that is, success and approval may be endogenous. This article tests for the endogeneity of approval and success. After building a theory that links success to higher approval, annual aggregate data from 1953–2011 are used to test for the endogeneity between approval and success. All statistical tests indicate that approval and success affect each other. This article concludes by putting the findings into perspective and suggesting new research directions. Acknowledgments The author thanks seminar participants at Fordham University and at Yale University, Brian Newman of Pepperdine University, and the anonymous reviewers at this journal for their comments on earlier versions of this article. Notes 1. The first of the treaties, the Permanent Neutrality Treaty, was ratified on March 16. It declared the Canal neutral and open to vessels of all nations. The second, the Panama Canal Treaty, was ratified on April 18. It provided for joint Panama-U. S. control of the Canal through December 31, 1999; thereafter control would shift to Panama. 2. The votes were held on March 16 and April 18, with 52 Democrats and 16 Republicans voting in favor, and 10 Democrats and 22 Republicans voting against. 3. See Edwards (Citation2009) for extensive review. Recent studies include Cohen (Citation2011) and Dwyer and Treul (Citation2012). 4. Brace and Hinckley (Citation1992) have three endogenous variables, annual approval, success, and the number of presidential positions on roll calls. But their analysis of the success model, presented in Appendix B, does not indicate a unique exogenous variable which is necessary for estimation of the two-stage least squares (2SLS) technique. It is possible to perform 2SLS in a two-step fashion, first estimating the instrumented endogenous variables and then estimating the endogenous effects of the instrumented variables in separate equations. This technique may affect the standard errors in the second step, unless corrected. Brace and Hinckley do not provide enough detail to determine if they used corrected standard errors. 5. Another possibility is that vetoes indicate to the public that the president cannot get Congress to accept his policies, that he is a weak leader, which will also harm his public reputation. 6. Kriner and Schlicker also argue that there may be an endogeneity between congressional investigations and approval, as Congress may be less willing to investigate the president and executive branch when the president is popular, and more willing to do so when he is unpopular. 7. Modern president possess unilateral policy-making tools, like executive orders, but legislating is the preferred and most important policy-making method for presidents. First, legislation is harder to overturn than an executive order. Second, the authority for executive orders must come from either legislation or the Constitution, which limits the president's ability to use executive orders to realize all his policy aims. For instance, Obama's health care reform could not have been implemented in its entirety through executive orders. 8. For example, the passage of legislation during FDR's first hundred days has become a standard against which all subsequent presidents are rated, but there is still controversy over the effects of FDR's New Deal legislation, whether it prolonged the depression or not. 9. I calculate only 10 non-policy stories annually on the legislative presidency. 10. The Times Index changed its font size, requiring weighting to insure comparability. 11. See Edwards (Citation2009, 339–41) for an extensive review. 12. Studies finding an approval-support association include Barrett and Eshbaugh-Soha (Citation2007); Brace and Hinckley (Citation1992); Canes-Wrone and Demarchi (2002); Edwards (Citation1976, Citation1980, Citation1989, Citation1997); Lebo and O'Green (Citation2011); Ostrom and Simon (Citation1985); and Rivers and Rose (Citation1985). However, Bond and Fleisher (Citation1980, Citation1984, Citation1990); Bond, Fleisher, and Northrup (1988); Borrelli and Simmons (Citation1993); Cohen et al. (2000); and Collier and Sullivan (Citation1995) fail to detect such an association. 13. Legislators will not be completely responsive to constituent preferences when their and the constituents preferences diverge. Then legislators confront an optimization problem, maximizing both their personal and their constituents' policy preferences. 14. The literature on presidential approval is too large to review here fully. For reviews see Gronke and Newman (2003, 2009). For a recent study see Newman and Forcehimes (Citation2010). 15. The Dickey-Fuller statistic is –4.53, compared to a test statistic of –3.57 at the 1% level. 16. An ARIMA analysis finds the first order autocorrelation is significant with z value of 5.28, p <.001. The Durbin's h statistic at one lag is 2.63 (p = .11) and the Breusch-Godfrey is 2.64 (p = .11), indicating no disturbance in the residuals once the lagged approval term in the equation. I use the lag of approval, rather than other techniques that might deal with serial correlation, such as the Prais-Winston transformation, because the two-stage least squares estimation (2SLS), which is necessary to test the hypothesis that approval and success affect each other, employs OLS. 17. I also experimented with an event variable, the annual number of positive (+1) and negative (–1) events per year, which proved to be statistically insignificant. Events tend to have very short term effects on approval, which may not show up when using annual data. 18. House and Senate success are correlated at .56, p = <.001. 19. In the success equation, I have not included the interaction between approval and party unity voting. Not only was it statistically insignificant in OLS estimations for success, but it did not improve the fit in the 2SLS, it rendered the approval and party unity variables insignificant, and it complicates the instrumentation of approval. 20. Statistically, endogeneity may not pose a problem, as the Durbin and Wu-Hausman tests do not indicate its presence. For the success equation the Durbin Chi-square and the Wu-Hauman F tests for endogeneity are .80 (p = .37) and .68 (p = .41), respectively. For the approval equation the test statistics are 1.14 (p = .29) and .97 (p = .33). The null hypothesis for these tests is that the estimations are exogenous. These insignificant results suggest we can not reject the null hypothesis of exogeneity, but these are also unreliable tests (see Dhrymes, Citation2003). Another complication is that in the 3SLS estimates, presidential positions now becomes statistically significant predictor of success (p = .05), which violates the exclusion restriction for using the Misery Index as an instrument and the partial r of the Misery Index with presidential positions, net of approval, is .28 (p = .03). 21. The F value is 5.18, which is significant at the .05 level.
Referência(s)