This paper examines the issue of causality in cross-sectional empirical models of economic growth. Using an approach to determining causal structures based on tests for conditional independence in sets of variables, we uncover alternative causal structures that are consistent with the correlation pattern of the variables in the data. We use these methods to develop alternative causal empirical models of economic growth. One of our consistent findings is that we can rule out the possibility that equipment investment causes growth. Our search procedure leads naturally to a structural model with latent variables which we then estimate. The results of our estimation are broadly consistent with traditional models of economic growth augmented for human capital.