Forecasting elections may be impossible, but no more so than resisting the temptation to try. As citizens following a campaign, we all indulge in private guesses about who is going to win electoral contests. Some go further and engage in illegal acts—unless done through the services of the Iowa Electronic Market—and bet money on election outcomes. Numerous political scientists have invited fame, but also ridicule, by designing sophisticated models to forecast elections; readers of PS are no strangers to those efforts. Students of elections can hardly escape this tempting opportunity any more than they can escape elections themselves. No phenomenon in our discipline comes in such a regular, precise, and verifiable form as an election.
As Republican hopefuls announce bids to run for president in 1996, while Bill Clinton is pondering his strategy for reelection, political scientists resume their tinkering with models to predict the outcome of the 1996 presidential election. Those forecasts, however, will not arrive until two or three months before election day, when the most up-to-date readings for the various predictors are available. Some would argue that such early forecasting is presumptuous any-how, since it ignores the whole general election campaign. Others will complain that those forecasts arrive too late to have more than curiosity value.