State and National Polls in Forecasting Presidential Elections Jay A. DeSart Utah Valley State Thomas M. Holbrook University of Wisconsin-Milwaukee ABSTRACT This paper is an analysis and update of the DeSart and Holbrook (2003) presidential election forecast model. It compares the performance of our original state-level forecast model (with national-level extrapolations) to an updated model that takes into account national-level forces. We also assess the performance of various models that take into account varying lengths of time in our state electoral history variable. While our original model performed remarkably well in predicting the outcome in 2004, the addition of a national-level poll variable and extending our history variable out over four elections (up from two in our original model) generally makes the model perform better across elections. Given these modifications, we update the model with the data from 2004 so an a priori forecast can be generated for the 2008 presidential election when the data become available. Paper prepared for presentation at the Annual Conference of the Western Political Science Association, San Diego, CA March 20-22, 2008.