Characterizing Measurement Error in Human Rights: A Bayesian Ordinal Measurement Model


Oona A. Hathaway & Daniel E. Ho


Abstract

Recent years have witnessed an explosion of interest in quantitative human rights research. Yet this research has been plagued by the inherent difficulty of measuring countries' human rights practices. Measurement error looms large. We discuss challenges of extant quantitative measures of human rights, focusing on the most widely used measure -- the Purdue Political Terror Scales (PTS), which quantify countries' political terror based on independent qualitative narrative reports by the U.S. State Department and Amnesty International. We illustrate a Bayesian measurement model that directly quantifies uncertainty about the world by combining multiple indicators of human rights. We then assess the impact of PTS measurement error on two previous studies, examining first the determinants of human rights practices and second determinants of bilateral aid allocation. The impact of measurement error can be severe. Statistical methods may thereby be crucial to assessing the sensitivity of empirical human rights scholarship to measurement error.