In this paper, I argue that there is no general empirical support for the neo-classical growth model. In particular, I show that Mankiw, Romer and Weil's (1992) iniluential evidence does not survive simple tests for sample pooling and that, except for the OECD countries, world wide per-capita income growth rates are significantly diverging over time. My results also iIIustrate the proposition that large groupings of disparate countries cannot meaningfully be placed in a single regression equation. This finding casts considerable doubt on the statistical validity of the findings contained in scores of papers using cross-sectional growth regressions. Finally I argue that the club convergence hypothesis is not obviously supported by the data. The OECD countries converge, but were not very likely to have done so from an ex-ante perspective. The Latin American Countries diverge, but seem far more homogeneous a group than the OECD.