This study examines how COVID-19 affected stock market returns and whether government intervention moderated this relationship across 32 emerging and developed financial markets during the acute phase of the pandemic. Pandemic intensity is measured using daily new confirmed COVID-19 cases per million population, transformed as ln⁡(1+"cases per million" ). Government intervention is captured through a composite policy index constructed by principal component analysis using four Oxford COVID-19 Government Response Tracker indicators: the Government Response Index, Stringency Index, Economic Support Index, and Health Response Index. The empirical strategy combines dynamic fixed-effects estimation and System GMM to account for return persistence, unobserved country heterogeneity, and potential endogeneity. To complement the structural analysis, Support Vector Regression and a Single layer neural network are used to evaluate nonlinear predictive performance. The results show that COVID-19 had a statistically significant negative association with stock market returns during the first pandemic wave. Government policy responses significantly moderated this relationship. Machine learning models further indicate that nonlinear specifications improve out-of-sample prediction relative to linear benchmarks. The findings contribute to crisis-finance literature by showing that pandemic information was rapidly priced by financial markets, while government intervention operated as a conditional signal that shaped investors’ interpretation of pandemic shocks.