Citizen Forecasting: Can Voters See into the Future?

Abstract

Political science, unlike economics, does not have a long tradition of forecasting models. However, this is changing. Currently, there is considerable interest in election forecasting. The basis for the interest is a flurry of related publications on House, Senate and presidential elections. A common goal for these studies is the development of a model, inevitably based on aggregate time-series data, which predicts election returns. The resulting models, some of which are quite accurate, can differ a good deal in specification and estimation. Also, they vary in complexity, making them more or less accessible to the engaged voter.