*, ** and *** indicate, respectively, a significance level at 10, 5 and 1%. Robust two-step standard errors are reported in parentheses. All models include a full set of control variables. The Arellano–Bond test for first-order autocorrelation [AR(1)] is significant in both specifications, as expected. The second-order autocorrelation test [AR(2)] is not significant, suggesting no residual autocorrelation in the differenced errors. The Sargan test fails to reject the null hypothesis of instrument validity, supporting the overall specification of the model