Premium
Voting Advice Applications and the Estimation of Party Positions – A Reliable Tool?
Author(s) -
König Pascal D.,
Jäckle Sebastian
Publication year - 2018
Publication title -
swiss political science review
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.632
H-Index - 30
eISSN - 1662-6370
pISSN - 1424-7755
DOI - 10.1111/spsr.12301
Subject(s) - voting , a priori and a posteriori , advice (programming) , computer science , multidimensional scaling , ideology , reliability (semiconductor) , estimation , space (punctuation) , interpretation (philosophy) , econometrics , statistics , data mining , political science , mathematics , machine learning , law , economics , politics , power (physics) , epistemology , philosophy , physics , management , quantum mechanics , programming language , operating system
Data contained in Voting Advice Applications ( VAA s) is not only a prerequisite for the vote recommendations they provide but can also be used for estimating party positions in low‐dimensional spaces. Given that VAA s can be designed differently in terms of their number of items and their measurement level, how much can one trust the party positions obtained from this source? We tackle this question by exploiting relevant variation in a real‐world setting: three VAA s offered at the 2017 Lower Saxony election. Despite substantial design differences, the policy spaces extracted through an inductive scaling approach are highly convergent. Simulated random item removal from the pooled dataset of all three VAA s furthermore suggests that about 40 items yield satisfactory reliability of the party positions. Finally, we find that a priori assigning VAA ‐items to ideological dimensions is potentially problematic as the interpretation of resulting party spaces may differ from the ones derived inductively.