Article ID: | iaor20101196 |
Volume: | 26 |
Issue: | 1 |
Start Page Number: | 11 |
End Page Number: | 18 |
Publication Date: | Jan 2010 |
Journal: | International Journal of Forecasting |
Authors: | Nadeau Richard, Lewis-Beck Michael S, Blanger ric |
Keywords: | forecasting: applications |
In the field of election forecasting, France is a lead case. Recently, however, certain modelers stumbled badly in their efforts to forecast the 2007 presidential election. The difficulty appears due partly to the single-equation format that has constrained past work, and partly to a failure to fully appreciate how key standard independent variables should be measured in the French context. As a potential remedy, we offer a multi-equation model, where the first equation gives strict emphasis to prediction, the second equation to explanation. By various statistical tests, this recursive system of equations manages the forecasting of Fifth Republic presidential elections remarkably well.