In this editorial the author asserts that opinion polls are little better than random guesses to predicting outcomes of presidential elections. The authors basis for this assertion is that opinion polls measure only the preferences of voters at the time of the poll and that many voters change their preferences several time before voting―some remaining undecided until the moment they cast their vote. The authors reasoning is unconvincing in two critical respects.
First of alt the predictions based on random guessing are such that the greater the number of candidates, the less likely the prediction will be correct. The reason for this is obvious: random guessing requires that no outside information be allowed to influence the guess. Predictions based on opinion polls, on the other hand, will differ considerably from those based on random guesses simply because outside information will influence the result. For example, in a four-person race, random guessing would yield the correct prediction 25 percent of the time, whereas the percentage of correct predictions based on opinion polls would be much higher. The reason for this disparity is simple. Opinion polls enable us to narrow the choices. That is, opinion polls serve to reduce the number of viable candidates in the voters mind and thereby increase the likelihood that the prediction based on them will be correct.
In addition, while it is true that many voters change their minds several times before voting, and that some remain undecided until entering the voting booth, this is not true of everyone. Moreover, people who do change their minds frequently or wait until the last moment to decide have typically narrowed their choice to a few candidates