The question of whether financial intermediation has a first order effect on the development process has long been debated. There have also been questions about the ‘robustness’ of the empirical results that suggest financial development does indeed have a first order effect. This paper addresses the second issue within this debate by assessing the robustness of the link between financial development and economic growth to variations in the sample (countries included in the data set). Specifically, the procedure identifies the relative influence of ordered subsets in the data to determine whether or not these financial variables change sign or lose statistical significance after the removal a specific subset. The results of this exercise suggest that financial intermediation appears to have a first order effect, that is most of the variables are robust to variations in the sample, using a traditional cross-country growth regression framework and an instrumental variables framework (GMM).