Consider two models forestimating, in advance of anelection, the shares of votes that will go to rival candidates. According to onemodel, pollsters' surveys of a randomly chosen set of registered voters before an election can be used to forecast the percentage of votes that each candidate will receive. This first model relies on the assumption that unpaid survey respondents will give truthful responses about how they will vote and that they will actually cast a ballot in the election. The other model uses prices of financial assets(legally bindingIOUs) issued by the Iowa ElectronicMarket, operated by the University ofIowa, to predict electoral outcomes. The final payments received by owners of theseassets, which can be bought or sold during the weeks and days preceding anelection, depend on the shares of votes the candidates actually end up receiving.

This second model assumes that owners of these assets wish to earn the highest possiblereturns, and it indicates that the market prices of these assets provide an indication of the percentage of votes that each candidate will actually receive on the day of the election.

A researcher can evaluate which is the better model for forecasting electoral outcomes by;

a. Comparing the predictions of the two models with actual election results.

b. Comparing the number of respondents in the opinion poll with the number of the traders in the Iowa Electronic Market.

c. Evaluating which one of the two methods is easier to understand for the public.

d. Comparing the cost of collecting information about the predictions of the two models.