Intergenerational Effects of Parents’ Math Anxiety on Children’s Math Achievement and Anxiety

438 children (243 girls, 195 boys), mean age 7.21 — 185 in first grade, 253 in second grade — and their primary caregivers; starting sample 868

First- and second-grade children in a large US field study, with parent-reported math anxiety and parent-reported frequency of help with math homework.

Psychological Science 26(9), 1480-1488. Not an experiment: parents' math anxiety and homework-help frequency are both measured, not assigned, so this is a longitudinal design with prior-achievement controls rather than a causal test. What it does well is control the obvious rival explanations. Children were assessed at the start and end of the year on Woodcock-Johnson Applied Problems (math) and Letter-Word Identification (reading), and the moderation model controls for grade, gender, beginning-of-year math achievement and beginning-of-year math anxiety (R-squared = .68). The pattern survived controlling for parents' highest education level (the authors' proxy for parental math KNOWLEDGE) and for teacher math anxiety, teacher math knowledge and school-level SES. And it is outcome-specific: the same interaction on end-of-year READING achievement is flatly null.

This is the archive's clearest documented case of parent-delivered instruction having a NEGATIVE association with achievement. Parents' math anxiety predicted less math learned across the school year and more child math anxiety by year's end — but only among parents who reported frequently helping with math homework. The interaction was significant, F(1, 430) = 4.59, p = .033, b = -0.16 [-0.30, -0.01]. At one SD above mean homework help the relation between parent math anxiety and child end-of-year math achievement was significantly negative (95% CI [-4.62, -1.05]); at one SD below, nothing (95% CI [-1.82, 2.77]). Parents' math knowledge did not explain it — the pattern held controlling for parental education, and the authors note early-elementary maths is easy enough that most parents are competent at it. The effect was specific to maths: on reading achievement the same interaction was F(1, 425) = 0.114, p = .736. The authors' own reading is that "even when math-anxious parents have good intentions, their homework help may backfire", and they explicitly note that a conservative interpretation is that math-anxious parents should not help with math homework at all.

HIGH, and this is the study's ceiling. Math anxiety is heritable and correlated with maths ability; the parents who are anxious about maths are disproportionately the parents who were bad at maths, and they transmit alleles as well as attitudes. Controlling for parental education is a weak proxy for parental numerical ability and no proxy at all for the shared genetic variance. Two features push back: the effect is moderated by a behaviour (homework help) rather than being a flat parent-child correlation, and it is absent on reading achievement in the same children — a pure heritability story would not obviously produce either. But no design here can separate transmitted trait from transmitted teaching, and the finding cannot support a causal claim on its own.

Whether this specific moderation has been directly replicated was not established in this pass, so `replication` records the ignorance rather than asserting a gap. The surrounding literature converges on the same sign from independent designs: Barger et al. 2019 (448 studies, 480,830 families) finds parents' homework assistance is the ONLY involvement type negatively related to achievement (r = -.15) while every other type is positive, and Fernandez-Alonso et al. 2022 finds d = 0.23 in favour of students receiving LESS family homework help across four PISA cycles. All three are observational and share the same confound.