From opinions to positions: Estimating potential legislators’ ideal points using data from the Comparative Candidates Surveys

The task of estimating the policy positions of political actors’ is at the heart of many theories of comparative politics, in particular those based upon the spatial approach to party competition. Several techniques have been developed for estimating the location of parties and other relevant actors in a policy space, using data derived from expert surveys, hand-coding of party manifestos, automated content analysis of political documents, mass surveys and elite surveys. Data derived from elite studies however have been of limited use as a tool for estimating policy positions as many surveys are restricted to single country analyses, a condition that does not guarantee cross-country comparability. The Comparative Candidates Survey (CCS) – a collaborative international project which collected data about candidates running for national parliamentary elections in more than 20 countries – overcomes this limitation, offering a unique chance to estimate and compare party policy positions using information about party members. By applying Poole’s Blackbox scaling procedure (1998) on a set of Likert-type questions included in the CCS in which candidates were asked to express their opinion about a set of issues such as the state intervention in the economy, abortion, same sex marriages, immigrants, environment etc., this paper aims at (a) estimating potential legislators’ ideal points, and (b) locating them in a common multi-dimensional policy space. First, we will estimate candidates’ position in a common latent space. Then their ideal points will be aggregated in order to compare party positions across the countries included in the CCS project. Finally, we will inspect the distribution of candidates’ ideal points to assess degrees of party cohesion in a comparative setting. In so doing, we will provide more accurate estimates of candidates’ positions in comparison to those obtained by simply comparing candidates’ left-right self-placements. As respondents can interpret the meaning and the content of the left-right dimension differently, our analysis enables us to address a problem known as “differential item functioning”. Moreover, our research offers the opportunity to validate party policy scores obtained by analyzing candidates’ opinions on issue scales against estimates based on traditional sources such as electoral manifestos and expert surveys data.