Does Participation in Music and Performing Arts Influence Child Development?

Reported as 2,907 children with data at all time points - this figure comes from a secondary summary (ArtsEdSearch) and is NOT verified against the article

US children in the Panel Study of Income Dynamics, using the Child Development Supplement and Transition Into Adulthood components; outcomes measured in adolescence and early adulthood

Propensity-score WEIGHTING (not matching) on child- and family-level covariates, applied to a nationally representative panel with unusually rich background data. Included because its headline sentence is the cleanest statement in the literature of what selection does to arts effects. Hard-paywalled with no open-access copy located, so everything here is from the registered abstract plus a secondary summary; the exact estimator, covariate list, balance diagnostics and effect sizes are unknown. Two corrections to widely circulated descriptions: the dataset is PSID, not ECLS-K, and the method is weighting, not matching.

From the abstract, verbatim: "Our results show that selection into arts education is at least as strong as any direct effect on outcomes, providing no support for the causal associations between arts participation and cognitive outcomes. We do find that arts education increases arts engagement during young adulthood." Outcomes examined include GPA, self-esteem, college attendance, reading and mathematics achievement, short-term memory, behaviour problems and high-school completion. The one durable positive is arts participation in young adulthood - i.e. arts education produces adults who do art, which is the honest transfer claim and rarely the advertised one.

Addressed only through observed covariates, which propensity weighting can balance but cannot make exogenous. The finding that selection is "at least as strong as any direct effect" is a statement about the size of the confound rather than a removal of it.

Consistent with Elpus (2013), which reaches the same conclusion on a different national dataset with a different method. In tension with Guhn et al. (2020), whose adjusted associations remain positive. No direct replication of this analysis exists.