Measuring error for adjacent policy position estimates: Dealing with uncertainty using CMP data

Careful users of CMP party position data should take the uncertainty of position estimates into account. We compare and evaluate two current approaches that provide error estimates for party positions. Researchers of the CMP group identify measurement error in quantitative content analysis as the cause of uncertainty about position estimates, whereas a second approach by Benoit et al. (2009) attributes the uncertainty of position estimates to a stochastic generation of election programs. We illustrate the commonalities and differences of these approaches and provide two empirical applications, the identification of the left–right order of parties and of policy shifts by parties, using CMP data for 25 countries. Despite conceptual differences, results in these applications are surprisingly similar.

► We evaluate the rival BLM and CMP approaches that provide error estimates for manifesto-based party position data.
► We provide two empirical applications, the left-right rankings of political parties and party policy shifts.
► Despite huge conceptual differences, the empirical results are surprisingly similar.
►  We show that the BLM approach has unintended, counterintuitive consequences that have not been published before.
► The CMP approach forces users to accept an illogical ‘vicious circle’ when deriving uncertainty estimates.