Predicting The 2000 Election
June 5, 2000
Political scientists are already predicting the 2000 presidential election based on mathematical models of voter behavior. Economists and political scientists have been working on mathematical equations to predict elections at least the early 1970s.
Probably the best of them is Ray Fair of Yale University, who regularly posts the results of his model on his website.
- Fair's April 28 results show the Democratic candidate -- the model is indifferent as to who the candidate is -- winning with 50.8 percent of the popular vote.
- Fair quickly adds that the margin of error in his model is 2.1 percentage points. Therefore, the results essentially are meaningless.
- For an analysis based on state-by-state results, one can turn to the Dismal Scientist web site, where the same sort of formulas used on a national basis are applied to the 50 states using state data.
- As of April, this model concludes that Gore will win with at least 359 electoral votes in November, with only 270 needed to win.
There are some problems with these efforts to predict elections. First, they use a limited amount of data, mainly relying on changes in real per capita gross domestic product. This is assumed to capture economic well-being, which is thought to be the driving force in voting.
Second, the models assume only the most recent economic conditions matter. Nothing earlier than 18 months or so before the election seems to affect the outcome of elections.
Third, although models may be useful as descriptions of past voting behavior, it is not necessarily valid to use them for prediction because the economic data used either consists of forecasts that can turn out to be wrong, or actual data that may be significantly revised subsequently.
Source: Bruce Bartlett, senior fellow, National Center for Policy Analysis, June 5, 2000.
For Fair model:
For the Dismal Scientist model:
For University of Iowa political futures market:
Browse more articles on Government Issues