The evolution of inequality of educational opportunity in USA: a machine-learning approach

This study contributes to the literature on inequality of educational opportunity (IEOp) by advancing both the measurement and the empirical analysis of how social background shapes educational attainment in the United States over time. The study makes several key contributions. First, whereas existing studies typically rely on a limited set of circumstance variables, this study incorporates a more comprehensive and multidimensional set of background characteristics. Second, the paper introduces a novel composite measure – the Social Advantage Factor (SAF) – which summarizes multiple dimensions of social and economic background into a single index and overcomes the existing trade-off between capturing heterogeneity and maintaining statistical precision. Third, motivated by recent developments, this study employs machine-learning techniques to reduce reliance on ad hoc model selection, flexibly capture nonlinearities and interactions among circumstance variables, address multicollinearity, and enhance the robustness of opportunity estimates. In addition, this study provides a long-term perspective on inequality of educational opportunity by examining trends since 1920 across four distinct educational stages. The findings show that, over time, inequality of educational opportunity has decreased for ‘high school’ and ‘some college attainment’ levels. Evidence of persistent inequality remains at the college graduation level, while estimates for post-college attainment should be interpreted more cautiously due to weak model performance at this educational stage.