Forecasting elections in Europe: Synthetic models

Abstract

Scientific work on national election forecasting has become most developed for the United States case, where three dominant approaches can be identified: Structuralists, Aggregators, and Synthesizers. For European cases, election forecasting models remain almost exclusively Structuralist. Here we join together structural modeling and aggregate polling results, to form a hybrid, which we label a Synthetic Model. This model contains a political economy core, to which poll numbers are added (to tap omitted variables). We apply this model to a sample of three Western European countries: Germany, Ireland, and the United Kingdom. This combinatory strategy appears to offer clear forecasting gains, in terms of lead and accuracy.