- Last Updated on Wednesday, 08 August 2012 15:20
- Published on Wednesday, 08 August 2012 15:20
- Hits: 704
A Statistical Model of Election 2012
Probably one of the most dangerous things to do when it comes to forecasting the outcome of an election is to rely too heavily on statistical models. They are, after all, only representations of human behavior, and humans, when they vote and make decisions, aren’t always consistent. I know I’m not. A model may predict one outcome for an election, while the people doing the voting, may do something entirely different. It happens all the time.
However, one indicator that seems to have the most profound impact on the outcome of a national political campaign is the state of the economy. A good economy or a bad economy is often the deciding factor in who wins and who loses the race for the White House. Other issues can have a substantial impact on the result, but none seems to correlate so closely with the outcome as does the strength of the economy.
This isn’t an original idea. Ronald Reagan and Bill Clinton, the latter whose campaign pollster had a placard above his desk that said, “…it’s the economy, stupid,” understood it well. More recently, Larry Sabato, considered one of America’s most quoted political observers, included a statistical model in his newsletter that made the same argument. And it was that effort that prompted me to dust off my statistics notebook from college and see if I might try my hand at such high end political analysis. Yes, I am probably in over depth, but I was anxious to put some numbers next to my often repeated assumptions and see what happened.
What I used was multiple regression formula. Real statisticians, which I am not, use these models, carefully developed, the correlations tested several ways, to determine the relation of certain variables to a particular outcome. It’s all rather neat math, but thanks to Excel, it’s a lot easier to do than it was 35 years ago (we used punch cards in those days) and besides, my objective, and my variables were fairly simple. Just how much of a correlation was there between economics and winning the White House and could they be used to predict the outcome of election 2012?
The two variables I chose for looking at the national picture were the unemployment rate at this time of year, and the projected growth or contraction in the Gross Domestic Product (GDP) in the second quarter. I wanted to see how this data related to the margin of victory, or loss, for the person in the White House. I took the data back through several elections, Bush, Clinton, H.W. Bush, Reagan, Carter, Nixon, Johnson, and Eisenhower. All were incumbents seeking reelection. The outcome showed that the correlation, namely the relationship between the economy, as measured by the GDP and unemployment, and who won or lost the Presidency was powerful. And, taking it a step further, I ran the numbers for this election. While we didn’t have second quarter statistics, or the July unemployment data, economists had a pretty good idea of what they’d be, and the news for the President wasn’t good. Under this model, he would lose this fall’s election by 3.6%. That’s not good news for the Democrats, but it’s not a whopping margin either, the GDP growth, small as it is, moderates the outcome a little. That gives the President a little room for optimism.
Having a little more fun with this I looked at two must-win states. Ohio and Florida. In this case, the model included both the national unemployment rate (because that’s the one most people pay attention to), the state unemployment rate, and the change in national GDP. The dependent variable, like the other model, was the margin of victory for the incumbent in that particular state. In Ohio, where the unemployment rate is better than the national rate, the President was down five percent. In Florida, tracking more closely to the national numbers, it was three percent.
None of the data I used in this analysis is likely to change between now and the election. There just isn’t enough time left on the political clock for there to be any significant improvement. However, economic statistics are only an indicator. They aren’t a sentence. There are a number of mitigating factors that give the President some hope for the fall. However, Mitt Romney, no doubt with more sophisticated economic models to study, has already decided that this election is about the economy and that’s all he plans to talk about.