Presidential Popularity and Presidential Vote

The president's popularity rating is highly predictive of his vote share in a reelection bid, especially when popularity is assessed in June of that year. This June popularity-vote model predicts about as well as the Gallup final preelection poll, and a 50 percent approval rating will ensure reelection. Michael S. Lewis-Beck is a Professor in the Department of Political Science, University of Iowa. Tom W. Rice is a graduate student in the Department of Political Science, University of Iowa. Public Opinion Quarterly Vol. 46:534-537 ? 1982 by the Trustees of Columbia University Published by Elsevier Science Publishing Co., Inc. 0033-362X/82/0046-534/$2.50 This content downloaded from 207.46.13.48 on Wed, 12 Oct 2016 05:48:46 UTC All use subject to http://about.jstor.org/terms RESIDENTIAL POPULARITY AND PRESIDENTIAL VOTE 535 Vt= 30.53 + .41 Pt (1) (3.72) R2=.70 n=8 D-W=2.18 where Vt = president's percentage of the popular vote, Pt = percentage approving of the president's handling of his job in the last Gallup popularity poll before the election, the value in parentheses = the t-ratio, R2 = the coefficient of determination, n = the number of observations (Gallup began asking the popularity item in 1938, and incumbents sought reelection in 1940, 1944, 1948, 1956, 1964, 1972, 1976, and 1980; these data are taken from The Gallup Poll Index, Reports No. 182 and 183, October-November 1980 and December 1980, respectively), and D-W = the Durbin-Watson statistic. Surprisingly, inclusion of the unusual Carter data actually enhances the model. The goodness of fit of the regression is improved and the t-ratio is greater. However, the vote percentages for Nixon and Carter are rather poorly predicted, with errors of 7.01 percentage points and + 4.7 percentage points, respectively. Perhaps the precision of the model could be improved by slight revision, which we now propose. Sigelman measures popularity according to the last such poll before the election. This strategy has intuitive appeal. The difficulty is that these final preelection popularity polls are not always held at the same time, and thus, responses may be influenced by differing short-term forces. We are especially concerned about the impact of the primaries and the conventions, which may cause a momentary, rather large, shift in the president's job rating. To overcome such distortion, it would seem preferable to employ a survey that was close to the election, yet held in a period of relative political calm, after the primaries and before the conventions. The June poll best meets these criteria. Hence, we sought a June measure of presidential popularity, which was possible for all years except 1940 and 1944. (The exclusion of those years may be preferable, since they were war years in which an incumbent was reelected to unprecedented third and fourth terms). Here are the OLS results for this "June model": Vt = 30.80 + .42 Pt (2) (4.80) R2=.85 n=6 D-W= 2.36 where the definitions are the same as with Equation 1, except Pt = the percentage approving of the president in the June Gallup poll prior to the election. Equation 2, with its June indicator of popularity, improves upon Equation 1, with its last preelection poll indicator of popularity, in This content downloaded from 207.46.13.48 on Wed, 12 Oct 2016 05:48:46 UTC All use subject to http://about.jstor.org/terms 536 LEWIS-BECK AND RICE several measurable ways. First, the larger t-ratio supports a higher level of statistical significance for the slope estimate. Further, the explanatory power of the model is clearly heightened, as the increase in R2 from .70 to .85 shows. By observing popularity shifts six months in advance of the election, we can account for 85 percent of the variation in the incumbent president's vote share. The strong linearity of the relationship can be seen in Figure 1, where the regression line is fitted to the scatter of points.