Stability and change in the Freedom House political rights and civil liberties measures

Abstract

Scholars have long attempted to make distinctions among states based on regime type. Two of the most commonly used measures by both scholars and policymakers are the Freedom House Freedom in the World political rights and civil liberties measures. Despite their popularity, little is known about the measurement properties of these variables. In 2006, Freedom House began releasing the subcategory indicators used to generate the seven-point political rights and civil liberties scales as well as the freedom status indicator. I investigate the measurement properties of these scales using Bayesian measurement models (factor analysis and latent class analysis) that explicitly incorporate the goals of the Freedom House organization (which I argue are rigor, appropriate precision, and temporal stability). I find that (a) there is considerable variation hiding in the seven-point political rights and civil liberties classifications (e.g. many countries that are coded one by Freedom House are interestingly different from each other), (b) some countries coded in different categories by Freedom House are not interestingly different from each other (e.g. some twos are not different from some threes), and (c) the subindicators are all reliable indicators of political rights and civil liberties. I show that these differences can be substantively and statistically consequential when using the Freedom House data if the researcher’s goals include comparison across space, comparison over time or predictive statistical modeling.