THE VITAL SOONER SCHOONER: HOW REPUBLICAN PRESIDENTS ARE ELECTED OR "IT DON'T MATTER WHO'S IN AUSTIN, BOB STOOPS IS STILL THE KING"
Ronald Keith Gaddie The University of Oklahoma version 2.0, October 2004 In 2000, pundits and prognosticators struggled to explain the inability of Al Gore to put away George Bush in the presidential election. Entering that election year, Gore was the beneficiary of a powerful economy, massive political experience, proven fund-raising ability, and incumbency as vice-president. Throughout the campaign, however, Gore trailed Governor George Bush. The predictions of political scientists and econometricians was that a sizable Democratic victory was forthcoming. These were not being borne out in an exceptionally close and controversial vote, leaving the crystal ball prognosticators shaking their heads (see, for example, the forecasting panel at the 2000 American Political Science Association Meeting in Washington DC). There was one model that called it close -- the Sooner Schooner model, which used the history of Oklahoma Sooners football to "predict" presidential election results. So, with apologies to Earl and Merle Black, Jim Campbell, and Waylon Jennings, I offered in October 2000 an alternative explanation for the pulse of presidential elections, and predicted from that model a narrow Bush victory for president and in the electoral college. What appears below is an update of this model, which includes data on the 2000 election and football season and which again forecasts a Bush victory, albeit by a more comfortable margin than in 2000. The Model The dependent variable is the net GOP victory margin in the November election (GOP% - Democratic %). There are four independent variables. The first is the tenure of the head football coach at OU, measured in years. The second predictor variable is whether the OU Sooners won their regular season conference title in the election year (a dummy variable coded 1 if the Sooners win the conference, a feat they accomplished in seven of the last seventeen presidential election years). The third and fourth variables are rivalry measures, whether the Sooners won, lost, or tied their games with Texas and Nebraska respectively (this variable is coded 1 for a Sooner win in the respective game, -1 for a loss, and 0 for a tie). All data are gathered for the years 1932-2000. There is a logic to the selection of these variables. The structure of American presidential politics indicates a long-lasting expansion of the GOP vote starting in the 1940s and continuing until the beginning of
the 1960s; another such expansion began in the 1970s and ran until the end of the 1980s. These two eras also coincide with the tenure of the two most successful coaches in the history of the University of Oklahoma, Bud Wilkinson and Barry Switzer. There are two parallels between these coaches and GOP success that merit a more general examination of OU football as a predictor of GOP presidential success. Wilkinson was a member of the GOP, having run for the US Senate from Oklahoma as a Republican. And, of course, the collapse of Switzer's coaching career and his replacement was roughly coincident to the onset of the continuing recession that undid the presidency of the first George Bush. After the early 1990s the Sooners went through a series of short-term coaches, and GOP presidential hopes likewise declined. The longer an OU head coach is at OU, the better Republicans do (the GOP won the last three presidential elections ['80, '84, '88] of Barry Switzer's tenure after losing the first [1976]). Republicans won seven of eight presidential elections when the Sooners won their conference, but just one of ten when the Sooners did not (Chi-square p < .005). The average difference between the GOP and Democratic vote from president is +5.94 to the Republican's favor when the Sooners win the conference, and -10.322 to the Republican disadvantage when the Sooners lose the conference. Clearly GOP fortunes are tied to the ability of the Oklahoma gridiron squad to vanquish the Big 6/7/8/XII/Whatever. More specifically, in Oklahoma football lore there are two games that loom large in the psyche of the Sooner football fan: the Texas game and the Nebraska game. The Texas game, it is conceded, is the more important and more intense, played at a neutral site and, until recently, only for bragging rights and without conference implications. The Nebraska game, in contrast, was often played for the conference title, a trip to the Orange Bowl, and as part of the continued good-guy/bad-guy rivalry between Bootlegger Barry Switzer and Saint Tom of Osborne. Since 1932, when Texas beats or ties OU in a presidential election year, Republicans are 3-8 (27.3%) for president; when OU beats Texas, Republicans are 5-2 (71.4%). When OU loses to Nebraska, Republicans go 1-7 (12.5%); when OU beats Nebraska, Republicans are 7-3 (70.0%).[Note: we tried to incude Oklahoma State in the model and found that the Cowboys just were not significant].
Results These data were subjected to an ordinary least squares (OLS) regression estimate. The results appear in Table 1. The model explains over 90% of the variation in the Republican margin over the Democrat. Three predictors -- coach's tenure, beating Nebraska, and beating Texas -- are significant at at least the .10 level and are also positively associated with the success of the Republican presidential candidate. Winning an OU conference championship is a powerful negative predictor, mitigating against the power wins and shifting the GOP vote margin by -15.2%. As OU Coaches stick around, an era of sustained GOP success ensues, with the presidential vote going up over six points from one election to the next as OU keeps its coach.
TABLE 1: SOONERSCHOONER MODEL FOR PRESIDENT, 1932-2000 A. Popular Vote Model Constant COACHTENURE BEATTEXAS BEATNEBRASKA WINCONFERENCE R-Square = .898 N = 18 B. Electoral College Vote Model B Constant -25.465 COACHTENURE 7.955 BEATTEXAS 14.305 BEATNEBRASKA 56.929 WINCONFERENCE -50.345 R-Square = .829 N = 18 + p< .10 *p< .05
B -3.772 1.637 2.996 14.053 -15.242
S.E. 2.974 .258 1.388 2.759 5.296
T 6.349* 2.158* 5.093* -2.878*
S.E. 18.771 1.627 8.761 17.414 33.422
T 4.889* 1.633+ 3.269* -1.506+
So what happens when we check the predictive validity of the model? As can be seen in Figure 1, the observed and expected vote track closely together. The correlation between the expected and observed GOP margin is strong and the model usually correctly predicted a GOP win or loss, and never made a vote forecast outside the margin of error, even in 2000. In 2000, most forecast models called for a 10 to 18 point Gore victory. The SoonerSchooner model called a more modest 1.7 point Bush margin, and captured the actual -0.7% Bush "margin" within its margin of predictive error (95% confidence interval). The electoral college version of the model called 308 Bush EC votes (271 actual). These forecasts were closer than any theoretic model. So what does the SoonerSchooner model say for 2004? In sum, a more confortable Bush victory. The estimates for the popular vote say that Bush will win by a margin of about 7.4% (53.7% to 46.3%) and take 384 electoral college votes.
If Bush makes it back, he'll want to thank Bob Stoops. It is therefore imperative to Republicans that OU retain Bob Stoops, and equally imperative to Democrats that Texas fire Mack Brown and hire Stoops, regardless of cost, in order to break his tenure at OU, beat OU, and deny the Sooners a conference championship. Otherwise, a long period of Republican rule will ensue.
FIGURE 1: PROJECTED AND ACTUAL PRESIDENTIAL VOTE FROM THE SOONER SCHOONER MODEL
Caveat: The above is for entertainment purposesonly. While the above correlations and relationships are in evidence, they have no theoretic or causal value. In other words, yes, it is a spurious order relationship, I know that, and that is the whole purpose of the exercise. Sometimes we need to laugh!