Under-Identification in Evaluative Social Data

This working paper develops a structure-first methodological framework for analysing gender differences in happiness, satisfaction, and related evaluative outcomes. It argues that a substantial part of the existing literature is not best understood as descriptively weak, but as under-identified at the mechanism level. The central claim is that evaluative outcomes are often analysed without jointly identifying expectation, domain salience, and structural load. As a result, observed empirical patterns—such as gender differences in subjective well-being—may remain compatible with multiple competing explanations, including adversarial, load-based, mismatch-based, and institutionally mediated interpretations. The paper introduces a formal under-identification proposition and proposes a Weighted Gap Index (WGI) as a benchmark for salience-weighted expectation–outcome mismatch across domains. It further develops an adversarial re-reading protocol designed to separate observation, measurement, mechanism, and interpretation in secondary empirical analysis. Using a structured re-reading of influential studies across happiness research, domestic asymmetry, cognitive labour, crisis-period burden, policy design, and bargaining models, the paper demonstrates how strong descriptive findings may coexist with unresolved mechanism-level ambiguity. The contribution is methodological rather than doctrinal. The paper does not reject adversarial or inequality-based interpretations, but argues that such interpretations are often not uniquely identified by the variables actually measured. It proposes a stricter inferential order in evaluative social research: descriptive closure should precede explanatory closure, and explanatory closure should precede normative judgement. This paper is part of the Structural Viability and Dyadic Systems research programme and is released as a version-controlled working paper.