The design effects of voting advice applications: Comparing methods of calculating matches

Abstract

In election times more and more voters consult voting advice applications (VAAs), which show them what party or candidate provides the best match. The potential impact of these tools on election outcomes is substantial and hence it is important to study the effects of their design. This article focuses on the method used to calculate the match between voters and parties. More specifically, we examine the use (explicit or implicit) of alternative spatial models and metrics. The analyses are based on the actual answers given by users of one of the most popular VAAs in Europe, StemWijzer in the Netherlands. The results indicate that the advice depends strongly on the spatial model adopted. A majority of the users of StemWijzer would have received another advice, if another spatial model had been used. At the aggregate level this means that how often a particular party is presented as best match depends strongly on the method used to determine the advice. These findings have important implications for the design of future VAAs.

Estimating parties’ policy positions through voting advice applications: Some methodological considerations

“…26-28). Since different scaling methods may lead to different voting recommendations (Louwerse and Rosema, 2011), it is advisable to follow some of best practices outlined in the dimensionality and scaling literature.…”

VAAs as sources of volatility and fragmentation: self-selection effects and genuine effects

“…For example, parties that oppose free childcare (as this has regressive income effects) benefit from this question being framed as "households with dual incomes should pay for child care", rather than as "childcare should be freely available to everyone regardless of income" (Kleinnijenhuis et al 2007). Thirdly, specific decision rules -alternatively labeled as decisional logics (Mendez 2016), or simply as algorithms (van der Linden and Dufresne 2016) -direct voters to specific parties (Kleinnijenhuis and Krouwel 2008;Louwerse and Rosema 2014;Mendez 2016), for example a low dimensional Euclidean decision rule as compared to a high-dimensional Manhattan city-block decision rule (van der Linden and Dufresne 2016). The consequence of these three aspects of VAA designs in short, is a certain level of arbitrariness in the advices VAAs provide to voters, which also may lead to fragmentation and volatility.…”

Electoral Dioramas: On the Problem of Representation in Voting Advice Applications

“…(Gemenis 2013;Trechsel and Mair 2011;Wagner and Ruusuvirta 2012) How can the result or voting advice be calculated in a fair and unbiased way? (Louwerse and Rosema 2014;Mendez 2012a) And what is the best way to present the results? (Louwerse and Rosema 2014;Otjes and Louwerse 2014) These kinds of epistemological and methodological questions are important and will arise for any VAA.…”

“…(Louwerse and Rosema 2014;Mendez 2012a) And what is the best way to present the results? (Louwerse and Rosema 2014;Otjes and Louwerse 2014) These kinds of epistemological and methodological questions are important and will arise for any VAA. Now, suppose for the moment that we can satisfactorily address these concerns.…”

“…The increasing prominence of VAAs in the electoral landscape has brought them under the scrutiny of political scientists. There is a burgeoning literature that focuses on empirical effects that VAAs have on the behaviour of voters, on the one hand (Gemenis and Rosema 2014;Marschall and Schmidt 2010;Walgrave et al 2008;Wall et al 2014), and on methodological concerns about their design on the other (Gemenis 2013;Louwerse and Rosema 2014;Walgrave et al 2009).…”