Intersectional Inequality Index (Triple I)

Several studies document inequalities along race, gender or socioeconomic status that are not completely characterized by binary comparisons – when multiple social markers combine to magnify differences in outcomes, from malnutrition to educational attainment to substance abuse to victimization and incarceration. Despite progress in both language for and awareness of intersectional inequalities, to date, there are no established methods to measure them while satisfying desirable statistical properties for group comparisons. This paper introduces a new statistical method – the intersectional inequality index (Triple I) – to measure intersectional inequalities even in the absence of survey instruments specifically designed for this goal. The index captures differences in how a binary outcome is partitioned across social groups relative to fair (lottery-based) allocations. Unlike alternative inequality indicators, Triple I treats every group symmetrically regardless of group size. We discuss its axiomatic characterization, probabilistic interpretation, and statistical properties – including decompositions to study inequality sources. We illustrate its usefulness by studying intersectional inequalities in unemployment within the United States. We find that, by 2020, Midwestern States had the highest intersectional inequalities in unemployment. At the country level, such inequalities were driven by black and brown individuals above median wealth being disproportionately unemployed. As a result, the employment Equitable Access Index – which penalizes the desirable outcome by Triple I – was much more heterogeneous across States than employment rates at the time.