Capturing Voters’ Policy Preferences? Using VAA Techniques for Experimental Research on Program Voting Procedures

Elections have two central functions: selecting the political elites, and selecting candidates’ or parties’ policy programs (Manin/Przeworski/Stokes 1999). However, there is evidence that translation of the citizens’ preferences through elections is dysfunctional: The composition of the parliament (e.g., Tangian 2013) or the political output (e.g., Achen 1978; Barnes 1977; Blais/Bodet 2006; Miller/Stokes 1963) is not congruent with the preferences of (most) citizens. In a diversified society voting on parties’ policy bundles does not capture the preferences of large groups of people accurately (Barber 1984; Bartolini / Mair 1990) and measuring preferences using votes on fixed program packages is difficult if voters have diverse policy preferences as shown with theoretical modeling (Brams 1976; Nurmi, 1999). This paper seeks to partially address this issue by proposing an innovative voting procedure for capturing citizens’ preferences accurately, and outlining how to assess it. More precisely, describes an experiment for analyzing how accurately three different direct democratic voting procedures capture subjects’ policy preferences by utilizing methods known from VAAs. The three procedures allow citizens to vote directly on policy programs that vary according to the degree of influence on the composition of the policy programs provided (allowing voters to rank policies or select policies of different programs) and different ballot layouts. The paper describes two different experimental designs conducted in cooperation with a municipality in Germany; 1) a lab experiment with 300 subjects and 2) an online experiment with 3,000 subjects. The paper explains in detail how methods used in VAAs can be applied for experimental studies on voting procedures, particularly in regard to different matching algorithms.