Predicting Elections

Abstract

This chapter covers the major types of election forecasting, as they have been applied in the United States and other democracies. For the most part, electoral forecasts have relied on three approaches. In the order in which they have emerged, these are prediction markets, pre-election polls, and structural models. In the last few years, Google searches, Facebook, and Twitter have also been exploited by election forecasters. The handbook chapter describes the various methods and evaluates their successes and failures. Of particular interest is the widespread failure of nearly all approaches to predict the election of Donald Trump. The postmortem is still ongoing. It remains to be seen what can be done to make election forecasting more accurate and thus more trustworthy.