Two University of Maryland political scientists have just released a study that seeks to pinpoint the effect Oprah’s endorsement of Barack Obama’s candidacy. From the conclusion of the full report…
The results for primary elections and caucus elections using Oprah Magazine and state fixed effects are used to generate predictions of the vote share Obama would have received, with and without the Winfrey’s endorsement. These were then used to predict county-level vote totals, and were then summed to provide an estimate of Winfrey’s total effect on Obama’s vote share, conditional on voting. In total, it is estimated that Winfrey’s endorsement was responsible for 1,015,559 votes for Obama. The 95% confidence interval for this estimated effect is 423,123 to 1,596,995.
There are two important caveats for this estimate. The first is that this is the effect conditional on voting: it does not taking into account any participation effect. The second is that this is the estimated effect for our sample of states, which does not include Texas, Michigan, North Dakota, Kansas, or Alaska. In the elections in the 45 states and the District of Columbia included in our sample, Barack Obama received 278,966 more votes than Hillary Clinton. Given that 423,123, the lower bound of the estimated impact of the endorsement is greater than this difference, the results suggest that Oprah’s endorsement was responsible for the difference in the popular vote in our sample between Barack Obama and Hillary Clinton.
They find that Oprah’s endorsement gave Barack nearly 1 million more votes than he otherwise would have received. This amount put him over the top relative to Hillary in the popular vote column…though that in and of itself doesn’t decide the nomination, the delagtes do.
Political academics typically cannot declare any findings with as much certainty as experts in the hard sciences. However that is not to say that strict standards of statistical validity don’t go into political studies. In other words, this isn’t the musings of two guys who happen to know a lot about politics. As with all political research it is grounded in empirical information, not random pontificating.