Economic models to predict presidential election results

Journal Record, The (Oklahoma City), Jul 7, 2000 by Hal R. Varian N.Y. Times News Service

Economics, the most imperialistic of the social sciences, is always testing the borders of related fields. Several years ago some economists invaded the territory of political science and applied their expertise to election forecasting.

Economics now offers two distinct approaches to predicting election outcomes. One has its roots in macroeconomics, examining how national economic performance affects voting. The other is grounded in microeconomics, using a real financial market in which participants can place their bets on the election results.

Which approach is best? You can decide by using the Web sites identified below.

The first approach is most closely identified with the work of Professor Ray Fair, a macroeconometrician at Yale who has examined all two-party presidential elections since 1916.

His most recent forecasting model, which involves variables like the growth of economic output, the inflation rate and whether a candidate is an incumbent, has generated extremely accurate predictions: during the period 1916-1992, his equation had only two elections wrong.

Right now, Fair's model forecasts a narrow victory for Vice President Al Gore, with 50.8 percent of the popular vote. But he considers the election a tossup, and the prospects could change, depending on what happens to the economy in the next few months. Readers who want to see the implications of various economic scenarios can visit the Web site at http://fairmodel.econ.yale.edu and click on "Predict the 2000 presidential elections" to make their own forecasts.

A quite different approach to election forecasting has been taken by some researchers at the Tippie College of Business at the University of Iowa. On their Web site at http://www.biz.uiowa.edu/ iem/index.html, you can buy and sell political candidates B not with campaign contributions, but by participating in an online futures market in presidential elections.

Here is how it works. The Iowa Electronic Market has issued securities for the Democratic, Reform and Republican parties that are tied to the popular vote. If the Republican candidate, for example, receives 52 percent of the popular vote, participants will get a payoff of 52 cents for each Republican share they own on election day.

A security that will be worth 52 cents a little more than four months from now should be worth slightly less than 52 cents today because of the time value of money, so the current price at which people are willing to buy and sell a Republican share should give a pretty good forecast of the popular vote in November.

As this column was being written, the forecasts were 48 percent for the Democrats, 3.5 percent for the Reform Party and 48 percent for the Republican Party -- a dead heat. If this stalemate continues, it suggests that Reform Party voters and Green Party voters may have a much bigger say in the November election than most experts now forecast.

The Iowa Electronic Markets involve real money, although the maximum opening balance for a trading account is limited to $500. And, yes, it is legal: the Commodity Futures Trading Commission has ruled that as long as the IEM conforms to certain guidelines, the commission will not intervene.

The electronic market is operated by the researchers at the University of Iowa as an economic experiment, with the goal of better understanding speculative markets. As a side benefit, it has ended up being a remarkably accurate predictor of presidential elections, forecasting the correct outcome of every election since 1988. It even worked for the 1992 three-party race, which caused virtually every other forecasting model to fail.

As the election heats up, it will be fun to watch the "market response" to various political events: convention speeches, candidate missteps, and vice presidential choices. If you think you can beat the market forecasts, you are welcome to try: the odds are a lot better than at Las Vegas, and the share prices are a lot less volatile than on Nasdaq.

The most interesting thing about the Iowa market is what it teaches us about how speculative markets work. Election forecasters know that voters exhibit various sorts of biases in their responses to polls.

For example, Republicans and Democrats each tend to overestimate the fraction of the popular vote that their party's candidate will receive.

The Iowa market forecasts do not exhibit these biases, and the question is: Why not?

Most of the participants in the Iowa market are college students and faculty. Perhaps these elite and highly educated subjects are less susceptible to emotional attachments?

Alas, it is not so. When asked the same questions, the participants in the Iowa markets tend to exhibit exactly the same survey biases as the general population.

Then why is it that the market does better than the polls -- which currently show Gov. George W. Bush well ahead of Gore -- in forecasting the election outcome?

The Iowa researchers offer a convincing explanation of this seeming paradox using the notion of "marginal traders." A marginal trader, in their context, is an especially sophisticated participant in the market, as measured by the kind of trading strategies he or she uses.