Different time different answer? Determinants of economic growth with error control via group knockoffs

Variable selection methods are commonly used for determining important variables in cross-country growth models. However, recent works show instability in the selected growth determinants. This paper addresses two important issues that could lead to this instability. Prior studies largely neglect the crucial fact that many growth determinants can be effectively organized into groups by growth theory. In addition, we emphasize that exact selection error control, which is largely unavailable in the literature, is essential to produce statistically and economically meaningful growth determinants. To close this gap in the literature, we adapt the knockoff filter at the group level in the cross-country growth regression models. We establish the first direct evidence of relevant theory groups in explaining cross-country growth across different growth periods from 1960 to 2010. In addition, the growing influence of religion on economic growth that is typically neglected receives significant support in our study. Contrary to past studies, our findings suggest that demography is an important economic growth determinant across different times.