Thus far, election forecasting models have never been applied to Belgium. Perhaps not without reason, as forecasting Belgian elections might be a daring exploit. The complex institutional settings of this federal country are quite discouraging. Finding a well-defined and politically relevant variable is a challenge. Moreover, the limited clarity of responsibility makes it doubtful whether Belgian political parties are held accountable for the state of the economy, one of the main prerequisites for developing a classical statistical forecast model. In this paper, however, we take up the challenge and develop a statistical model to forecast Belgian election results. We start with a short assessment of whether the state of the economy is relevant for vote choices in Belgium. This is a prerequisite for developing a classical forecast model. We do so by means of an analysis of the 2009 Belgian Election Survey, in which respondents were asked not only about their vote choice but also about their perception of the state of the economy. Further, they were also asked what impact the economy has on their vote choice. The forecast model we develop is focused on the party of the Belgian prime minister, this is a theoretically and politically relevant option. With the choice for this dependent variable, because of historical evolutions, we focus solely on the Flemish party system. Since 1974, all Belgian prime ministers have been Dutch-speaking. This has the advantage that both federal and regional election results can be combined, increasing the cases included in the analyses from eight to ten. Although the sample size remains small, such a small dataset is not unusual within the practice of scientific election forecasting. We start from a classical forecast model, including both a popularity and a traditional economic variable. Further, we take into account the long-term declining trend of the vote share for traditional parties in Belgium.
