This study introduces and evaluates a bootstrap-enhanced Feasible Generalized Least Squares (FGLS) estimator designed to improve econometric inference under conditions of heteroskedasticity and autocorrelation. Using both simulated and secondary datasets, the performance of the bootstrap-enhanced FGLS is compared with the traditional FGLS method across varying sample sizes. Results indicate that the bootstrap approach substantially reduces bias and root mean square error (RMSE), particularly in small to moderate samples. Additionally, standard errors of coefficient estimates are more stable under the bootstrap approach, especially in the presence of complex error structures such as multicollinearity and spatial correlation. The study also validates the method’s applicability across diverse empirical domains, including macroeconomic indicators, demographic data, and spatial datasets. Findings reinforce the diagnostic power and efficiency of bootstrap resampling in improving estimator precision, making it a robust alternative to classical methods in econometric modelling. Policy recommendations emphasize the need for resampling- based strategies in economic planning and forecasting when data irregularities challenge traditional assumptions.