This paper shows, within a Heckscher-Ohlin version of the two-sector neoclassical growth model, that land, besides having long-run effects, is also a main determinant of the speed of convergence toward the steady state when there are cross-sector capital share differences. This result stands in sharp contrast to the predictions of standard neoclassical growth frameworks, and calls for a reinterpretation of the conditional-convergence and the resource-curse findings. More specifically, the model predicts that the former finding requires the existence not only of diminishing returns but also of relatively small differences in capital shares across sectors. With respect to the latter finding, our results imply that it may be a consequence of purely transitional effects of natural riches on growth, and that it can not be interpreted as evidence that natural inputs necessarily harm long-run welfare. We produce empirical evidence on the relationship between land, income levels, and growth rates, and present data on cross-sector capital shares. We claim that most of that evidence is consistent with the predictions of the model.