Changes in voting behavior pose new challenges, theoretically, but also methodologically. As a consequence of “dealignment”, that is the decline of long-lasting, stable predispositions, short-term factors and momentary information become – by implication – more important. In order to cope with these changes theoretically, electoral research has undergone cognitive and affective turns, regularly employing theories and concepts from political communication and political psychology. In the wake of that, we have seen methodological imports and innovations. Election studies now regularly include dynamic elements, such as rolling cross-section and short-term panel designs. Moreover, laboratory studies (like information boards or real-time-response measurements) are used to study information processing and its consequences, often accompanied by experimental designs. Both trends have increased our potential to analyze and understand voting decisions. But shortcomings remain: Neither one of those approaches creates opportunities to study voting decisions and their determinants while they are taken. Large-scale surveys, even if run in a dynamic way, usually end at least one or two days before Election Day. Post-Election surveys start with a delay, too. Laboratory studies, on the other hand, can only simulate campaigns and voting decisions, potentially lacking mundane realism and externa validity. But in times of increasing vote switching and late deciding, it seems necessary to get as close as possible to real-world voting decisions while they are taken. Stated differently: Are exit poll – conducted on Election Day – the last exit for electoral research?
Exit polls create opportunities to test theories of voting behavior using measurements taken only minutes after voting decisions are taken. Still, they are only used for commercial, media-driven election coverage, hardly for academic electoral research. To assess the power of exit polls for electoral research, we have conducted a pilot study on September 24th, 2017, the day of the most recent German Bundestag election. We have invited voters to take part in a survey when they left their polling stations, ultimately yielding about 2000 completed interviews. Contrary to media-driven exit polls, which include only a very limited number of variables, our questionnaire included a wide range of variables related to potential short-term factors, e.g. pertaining to mood, emotions, and top-of-the-head issues. To further investigate the usefulness of exit polling for academic electoral research, we furthermore collected email addresses from our participants on Election Day. That way, we were able to re-interview a considerable share of our respondents with varying time gaps between election day and follow-up, enabling us to test how stable attitudes, emotions, moods and their explanatory power for voting decisions turned out to be.
In doing so, we aim to will provide a rigorous test of the potential power of exit polls for electoral research.