I’m against progonistication. Our job is to decide who should win the election and what he or she should do in office, not who is most likely to win. Nevertheless, near the beginning of a long and very close race, who can resist trying to predict the outcome?
Many political scientists think that an election involving an incumbent president is a referendum on current economic performance. Everything else–debates and tv spots, scandals and endorsements–is mere atmospherics. But what is the relevant way to measure “economic performance”? Bob Dole claimed yesterday that President Bush has the advantage because “the misery index?the combination of unemployment and inflation?is actually lower now than it was at this point in 1996. And less misery is a good thing for incumbents.” But Professor Larry Bartels of Princeton conducted a sophisticated meta-analysis of 48 election models and found that while no single variable can predict an election, the single best predictor is the change in disposable per capita personal income (dpi) over the twelve months prior to the election. The more buying power people have (after taxes), the more they like the powers that be. Figure 1 in this article by Bartels makes the point pretty clearly. It suggests that incumbents need to preside over at least 2% annual growth in real per capita dpi to get more than 50% of the vote. Of course, that 2% figure has a large margin of error and is nothing more than a rule-of-thumb. Moreover, in a year when most states are considered safe for one party or the other, I’m guessing that the only thing that really matters is the change in dpi in “swing” states such as Ohio. If I could find current dpi/capita statistics for Ohio, I’d risk predicting the election outcome. I cannot find those statistics, but I do see that the growth in the nation’s real annual dpi over the six months ending this February was just $163 per capita, or 0.6% percent. That is dangerously low for the incumbent.