Averageism vs Marginalism in Research

Social sciences can describe phenomena two ways: by averages (what a typical member or aggregate looks like) or by margins (what one additional unit changes). The article argues modern empirical practice—and machine learning—tilts researchers toward estimating credible causal or predictive averages without checking whether those estimates map to the marginal quantities that older theory prized. — If researchers stop asking whether their estimates capture the theoretically relevant marginal effects, policy decisions may be driven by well‑identified correlations or predictions that don't have the causal meaning policymakers assume.