Expectation Formation and Electoral Decision-Making

Research question/goal: 

Incentives for strategic voting under proportional or mixed electoral systems are the research topic of this project. Strategic voting is defined as a conscious choice not of the most preferred party in order to attain a better final outcome by voting for a less preferred party. The incentive most investigated is the avoidance of a wasted vote for a party or candidate whose entry into parliament is not expected. This incentive plays a role especially under plurality systems in single member constituencies, but is, of course, also a possible motive in the plurality part of a mixed system (topic 1). To this research question we added as a second topic the expectation formation of voters concerning national election results and its consequences for coalition formation. A logical third topic is then coalition voting, that is choosing a party in order to advance the chances of one’s most preferred coalition. Both national surveys and experiments were used to test our hypotheses. (1a) German surveys for the 1998 and 2002 Bundestag election (Herrmann/Pappi 2008, Herrmann 2008a under review): We analyzed the complete range of potentially strategic situations voters may encounter in their local constituencies. We found that voters engage in strategic voting, given the chance to influence the outcome of the constituency race. The overall rate of strategic voting was estimated at about four percent. To explain this rather low incidence of strategic voting, we considered recent advances in strategic voting theory that demonstrate how voter uncertainty about competitors’ true support levels in the constituency undermines strategic voting, thereby leading to incomplete desertion of trailing parties. Calibrating a more flexible model of strategic voting (Myatt & Fisher 2002) to data on voting results from simple plurality elections in German constituencies, we found, for the first time, that strategic voters indeed operate under high levels of uncertainty. These results have important implications for the general belief held in the literature that, in equilibrium, strategic voting must lead to outcomes in which only two parties receive votes (Duvergerian equilibrium). (1b) First experiment (Meffert/Gschwend 2007b): It used an innovative design that embedded a laboratory experiment in two real election campaigns, allowing the manipulation of poll results and coalition signals in a realistic environment. The findings suggest that voters try to avoid wasting their votes when their preferred party is not likely to be represented in parliament. Moreover, coalition signals are found to have a surprisingly strong effect on the likelihood of casting a strategic vote in order to get a preferred coalition into government. (2) Pre-election surveys in Belgium and Austria (Huber et al. 2008, Meffert et al 2008 under review): A small set of questions on voters’ expectations about election outcomes was asked in the Belgian survey and a large optimal set was included into the Austrian survey. We identified several factors that influence expectation formation beyond simple "wishful thinking", among them political knowledge and involvement, rational considerations (expectations become the more accurate the smaller the scalometer distance between most and second most preferred party) and regional context (expectations are more accurate in regions whose election results do not deviate much from the national average). The evidence about coalition expectations suggests that, at least in the aggregate, voters have a fairly sophisticated understanding of the likelihood of various coalitions, and that they can differentiate between expectations about majorities and the likelihood of agreements between the parties. At the same time, voters clearly engage in wishful thinking, though mostly for specific coalitions, in particular if the preferred party or some truly disliked party are involved. (3a) Second experiment (Meffert/Gschwend 2008 under review): Strategic voting in multiparty systems with coalition governments is not a simple choice between first and second preference but might involve any of the parties, or even strategic abstention. Identifying the optimal vote decision quickly becomes a highly challenging task. In an economic experiment to avoid pre-existing biases toward certain parties we developed a strategic voting game model that consists of four hypothetical parties that compete for 15 voters in a two-dimensional policy space where we manipulated poll results and coalition signals over 25 elections. The results show that voters are frequently able to make strategic vote decisions, but that voters also rely on simple decision heuristics and are highly susceptible to coalition signals by parties. (3b) Austrian pre-election survey (Meffert/Gschwend 2007a, Herrmann 2008b, Pappi 2008): These survey data clearly showed that voting for coalitions is not a task of insurmountable complexity but in the realm of the possible: voters have a fairly sophisticated understanding of the likelihood of various coalitions and about one third of the respondents actually rated a coalition higher than the most preferred party. Furthermore, voters differentiate between expectations about majorities and the likelihood of agreements between the parties. Building on these findings, we stipulated a simple spatial logic for coalition voting whereby left and right voters who perceive their preferred left or right coalition as least likely to win were predicted to strategically cast their ballot for a centrist party. By contrast, those who perceive a chance for their preferred coalition to become the next government were predicted to strategically vote for the respective non-centrist party. Testing these predictions against the standard model of sincere proximity voting, analyses showed that believing one’s preferred coalition is non-viable raises the probability of voting for a centrist party while believing one’s preferred coalition to be viable lowers the probability of voting for a centrist party.

Fact sheet

2005 to 2008
Data Sources: 
Survey Data, Aggregate Data
Geographic Space: 



Behnke, Joachim, Thomas Gschwend, Delia Schindler and Kai-Uwe Schnapp (Eds.) (2006): Methoden der Politikwissenschaft. Neuere qualitative und quantitative Analyseverfahren. Baden-Baden: Nomos. more
Gschwend, Thomas (2004): Strategic Voting in Mixed-Electoral Systems. Reutlingen: SFG-Servicecenter Fachverlag. [Statistik und Wissenschaft; 2] more