Operationalizing Sen’s Capability Approach Using Bayesian Item Response Theory

Abstract Amartya Sen’s capability approach (CA) has achieved wide theoretical recognition but remains difficult to operationalize due to the abstract nature of capabilities as opportunity sets rather than achieved outcomes. This paper proposes a novel measurement framework that applies Bayesian item response theory (IRT) to directly estimate latent capabilities from survey items designed to capture individuals’ perceived opportunities. We develop a multidimensional graded response model that accounts for interdependencies across capability dimensions and items, and allows for a probabilistic representation of latent capability structures. In contrast to conventional applications that proxy capability through achieved functionings, our approach aligns more closely with Sen’s original emphasis on freedom and choice. Using cross-sectional survey data from four Southeast European countries, we validate the model through posterior predictive checks and demonstrate its ability to recover meaningful latent capability structures. The results reveal substantial cross-country and age-related differences. For example, while Kosovo ranks highest in aggregate capability, it performs comparatively weakly in environmental capability. Conversely, Moldova shows the opposite pattern, with a low capability level but superior performance in the social dimension. These findings highlight that aggregate indicators may obscure critical variation across dimensions. The paper demonstrates that Bayesian IRT provides a rigorous and transparent measurement framework for operationalizing capability, and offers a promising pathway for more nuanced welfare assessment across populations.