Beyond Mean Score Differences: Home Language and the Interpretation of Cognitive Test Performance in China

Abstract

Large-scale surveys often use cognitive test total scores to describe group differences, but the construct meaning of such scores depends on task content, administration, language experience, and opportunities to learn. The China Family Panel Studies (CFPS) word test is an oral Mandarin character reading task with education-dependent entry and stopping rules. We analyzed 16,689 adults followed from 2010 to 2018 using the full survey strata, primary sampling units, and longitudinal weights, and contrasted baseline-adjusted, gain score, geographic overlap, propensity overlap, floor exclusion, nonlinear, and measurement error sensitivity results. The fully adjusted 2018 contrast relative to Mandarin-speaking homes was −0.023 SD (95% CI [−0.077, 0.030]) for Sinitic-dialect homes and −0.292 SD (95% CI [−0.430, −0.154]) for the pooled minority-language response category. The latter estimate was reduced to −0.183 SD after exact ethnicity code fixed effects and to −0.081 SD (95% CI [−0.291, 0.129]) under overlap weighting; only 49 respondents fell in the 1st–99th percentile propensity score intersection. A covariate-adjusted gain score model yielded a smaller minority-language contrast (−0.180 SD, 95% CI [−0.356, −0.005]). Correcting the baseline under plausible reliability values of 0.85–0.95 produced point estimates from −0.226 to −0.277 SD. Floor exclusions reduced but did not eliminate the negative contrast. These results show that the pooled minority-language coefficient is strongly estimand- and support-dependent. They support a constrained interpretation of the observed CFPS score rather than a direct inference about broad crystallized intelligence or a uniform effect of minority-language use.