Criteria to Evaluate Voting Indicators and a Recipe for a New One
This paper is made up of two interrelated parts. Part I contains a critical evaluation of the so called StemWijzer 2002, an internet application that helped and advised voters which political party to choose at the Dutch, German and Suisse parliamentary elections. Applying some validity tests and checking the discriminatory power, it is argued that the StemWijzer 2002 is seriously flawed. Part II is devoted to designing an interactive voting indicator, where the propositions to be evaluated by the voter depends on his responses to other propositions at an earlier stage. The viability of such an indicator hinges on the possibility to determine the partisanship leaning (leftist or rightist) of voters by means of a limited number of questions. For this purpose a rather minimalist voting indicator (the VoteFork) was developed. The empirical results show that it is indeed possible to ascertain for the majority of voters whether they belong to left-wing or right-wing parties. In this part I also present a simple model which can be used as a benchmark to assess the quality of voting indicators, where quality is simply defined as the fraction of correct voting advises. This paper is pioneering in the sense that yet no serious attention is paid to voting indicators in general, to the criteria they should ideally meet and to the possibility to make them interactive. Comments are welcome.