This study tests Lewis’s two-sector growth model in Afghanistan between 1991 and 2020, a period marked by prolonged conflict, political instability, and fundamental regime change. Following the Mankiw, Romer, and Weil (MRW) augmented Solow framework, we specify a growth model that includes economic structure as the main explanatory variable, along with human capital, physical capital, and labor productivity as additional determinants of real GDP growth. Given severe data limitations and multicollinearity, ridge regression and seemingly unrelated estimation (SUEST) are employed to reduce estimation uncertainty. Official statistics are supplemented and validated using nighttime lights data. Two specifications are estimated: a fullsample ridge regression, in which all variables are positive and significant, and a comparative model using SUEST to test for structural differences between the Taliban (civil war) and Republic periods. The results indicate that the effect of economic structure on growth is not constant but regimedependent: negative during the Taliban period and positive during the Republic. Labor productivity remains the strongest and most persistent driver of growth across both periods. These findings suggest that while Lewis’s framework is broadly valid in the Afghan context, its predictions are conditional on institutional stability and governance effectiveness