LISPOP associate issues methodological comment on CBC’s Vote Compass. Click here for access to paper.
Abstract: This methodological note advances two arguments. First, random error in the measure of the two underlying dimensions in the Vote Compass is converted through the process of averaging into systematic error in the measure of a user’s position relative to the political parties on the two-dimensional plane. This argument can be demonstrated via simulation. The second argument is that the two-dimensional Vote Compass graphic is likely to “misclassify” many users in terms of their actual proximity to political parties on the questions that the Vote Compass asks about. By “misclassify” I mean that the two-dimensional graphic generated by the Vote Compass indicates to some users that they are “closest” to one political party when in fact there was some other political party with which these users agreed more often and more closely on the issues that the Vote Compass asked about. The possibility persists whether the users answer the questions randomly, ideologically, or some combination of both. My substantive conclusion is that the Vote Compass’ two-dimensional graphic is misleading for many users.