Then, you realize that NC is a swing state where college basketball is arguably the most popular sport and Obama's picks make a little more sense. I'm not alone in noticing this, but so far no one has really put this idea to a strict, quantitative test to determine whether Obama is pandering to swing voters with his NCAA bracket. So, what's a bored nerd to do, you ask?
Why, design a rinky-dink statistical analysis to determine if the relationship exists, of course! Below I'll first explain the theory and methods I used to conduct this test, then provide the results, and finish with a little discussion of what I found.
Stated explicitly, the question I hope to answer here is: "Does Obama systematically favor teams from "swing states" when filling out his NCAA bracket?" My theory is that Obama does, indeed favor teams from swing states because, by doing so, he only stands to gain. As long as his predictions are not absolutely outrageous, Obama can favor a battleground state's team over the team of a state that he either expects to win outright or expects to have no chance in in order to curry favor with that state's voters. This might seem like a silly political move at first, but if you think this wouldn't make a difference in the way people might vote, you've never met someone from rural North Carolina. On a more serious note, it makes an intuitive sort of sense that during an election year a presidential candidate should waste no opportunity to squeeze out a little more favor from voters in strategically valuable states. I think that's exactly what Obama's doing here, which leads me to my hypothesis:
Hypothesis: Obama will predict more wins for teams from swing states than the "average" individual who fills out an NCAA bracket.
The first and most obvious step in designing a quantitative test for this hypothesis is to be clear about which states I consider to be "swing states." I picked 12 states that I (and most reasonable people, I believe) consider to fall under the category. Listed by abbreviation, they are as follows: CO, FL, IA, MI, NV, NH, NM, NC, OH, PA, VA, and WI. You might be able to argue that some of these aren't swing states, and there might be a state or two that you could argue that I should add to the list. However, I seriously doubt that adjusting this list would meaningfully change the results. With this out of the way, I can explain how I operationalized the dependent and independent variables.
To test bias towards swing states I developed a very simple measure that I call the "pander score." This score assigns a value to each state for a given year that is equal to the difference in Obama's bracket and the ESPN national aggregate bracket's predictions about the total number of wins that teams from a given state will achieve throughout the tournament. By using ESPN's national bracket as a control group, this test compares Obama's predictions to the aggregated predictions of a large group, allowing me to identify in a quantifiable way to what degree Obama's predictions differ from that of the "average" individual.
The independent variable was a dummy indicating whether or not the state was one of the dozen
swing states that I identified above. A value of 0 indicates that the state was not a swing state whereas a value of 1 indicated that it was.
I then ran a simple linear regression, with my "pander score" as the dependent and the swing state dummy as the independent. A screen cap of the results are presented below:
As the results indicate, there is a positive relationship between a team's location in a swing state and the number of wins that Obama predicts that that team will achieve throughout the tournament. Furthermore, this relationship is strongly statistically significant well beyond the 95% level of confidence. Considering that the pander score measures the difference between Obama's predictions and the national average, this positive and strongly significant correlation provides strong support for the notion that Obama's bracket picks were influenced by his desire to appeal to swing state voters.
Clearly, there are some problems with this test. First, although there are 50 "observations" in the statistical analysis, I'm essentially working on an "n" of one, as I haven't taken into account Obama's brackets for any years other than 2012. I didn't consider these for two reasons. First, I couldn't find ESPN's national aggregated bracket for any year other than 2012. Second, I'm only willing to spend so much time on a dumb (albeit fun) endeavor like this one. So, if anyone wants to extend the analysis, go for it! I've provided a link to the data that I used at the end of this post, so have at it and if you do, please send the results along I'd love to see them. Furthermore, there is the possibility that swing states may tend to have a higher proportion of teams that would be considered "underrated" by someone who really knows a lot about basketball. If this is the case, then the reason for Obama's swing-state basketball bias may not be political in nature. I doubt that this is the case, but it still requires consideration.
To conclude, this hastily conducted statistical analysis suggests that Obama's NCAA predictions are indeed politically motivated. Now sit down, grab a drink and enjoy the tournament knowing that all those non-swing state teams might stand a chance after all.
And GO HEELS.
Finally, you can download the dataset I created for this post from my dropbox.