Ballot Design Influence on Voting Behaviour: Testing the Effects of Ballot Design on Results of Multi-Option Referendums

There is a growing societal interest in the use of multi-option referendums as a more consensual alternative to binary referendums. Multi-option referendums can be designed in several ways. In real-world voting referendum ballots are universal, hindering an exploration of the effects of different design characteristics on voting behaviour. Survey experiments can fill this gap by directly comparing voting behaviour for different ballot designs to distill the effects of design decisions on referendum outcomes. This paper tests to which extent the design characteristics of multi-option referendum ballots influence voter behaviour. To this end, the paper compares two main ballot designs observed in multi-option referendums around the world: single question designs and multiple binary questions combined into a single outcome. The analysis relies on a survey experiment conducted in the Netherlands, in which respondents voted on the same policy options – two different proposals and one status quo option – in one of three designs: (a) a single question in which respondents ranked the options, (b) a single question in which respondents approved of options and (c) multiple binary questions in which respondents indicated their approval for each of the two policy proposals separately and then indicated their relative preference between the two in a deciding question. The approval voting question and the multiple binary question design are directly comparable with respect to the expressed support percentages for each of the two different policy proposals against the status quo option. Significant differences between the two point towards design effects on voter behaviour. Similarly, the relative ranking of the two proposals in the ranking question is comparable to the deciding question of the multiple binary format. The paper further takes into account respondents’ background characteristics to test whether ballot design variations manifest themselves in different ways between different groups of voters. The voting data for all three multi-option designs are compared to control groups that are presented with binary choice sets. This paper contributes to our understanding of the extent to which design influences voting results in multi-option balloting.