Do Advance Letters Improve Preelection Forecast Accuracy?

The survey methodology literature has debated whether advance letters to potential survey respondents will reduce nonresponse bias and thereby improve the accuracy of preelection forecasts. This research note analyzes the results of experiments conducted in Maryland, New York, and Pennsylvania in which advance letters were sent to a random sample of potential survey respondents to 2002 preelection surveys. We find a significant increase in the overall response rate, although notably less than in past studies. However, the advance letters did not improve the representativeness of survey respondents or the accuracy of the election forecasts.