We revisit a central estimate in the economics of education: the human-capital loss associated with COVID-19 school closures. Estimates of pandemic learning loss may be affected by publication bias, p-hacking, and the mechanical correlation between standardized effect sizes and their standard errors. We conduct a comprehensive multi-method assessment of bias by applying a wide range of correction techniques — including PET-PEESE, three-parameter selection models (3PSM), Robust Bayesian Meta-Analysis (RoBMA), Meta-Analysis Instrumental Variable Estimation (MAIVE), Right-Truncated Meta-Analysis (RTMA), and multi-bias sensitivity analysis. Our preferred specifications, RoBMA and MAIVE, rely on different assumptions yet converge on an effect size of approximately −0.12 SD, equivalent to a learning loss of about 30% of a school year. Although some methods reveal signs of publication bias and selective reporting, these findings do not explain away the central finding: the COVID-19 learning deficit is economically meaningful and statistically robust.

Fig: A learning deficit that survives correction for publication bias



Funnel plot of 291 effect-size estimates of COVID-19 learning loss. Horizontal axis: effect size (Cohen's d); vertical axis: estimated standard error. The dashed line marks zero (no effect); the dotted line marks the unweighted mean of the estimates (about −0.13 SD). Most estimates are negative, indicating a learning deficit. Across a battery of bias-correction methods (PET-PEESE, 3PSM, RoBMA, MAIVE, RTMA, multi-bias), the preferred RoBMA and MAIVE specifications converge on about −0.12 SD — roughly 30% of a school year — and the deficit survives corrections for publication bias and selective reporting.


Reference: Luskova Martina, Buliskeria Nino, Elminejad Ali, Havranek Tomas, Irsova Zuzana, Jurajda Stepan, Kapicka Marek (2026), “Publication Bias and P-Hacking in the Effect of COVID-19 on Learning.” Charles University, Prague. Available at meta-analysis.cz/learning.