This paper presents the results of a data-based comparative study of several hundred catchments across continental United States belonging to the MOPEX data set to systematically explore the connection between the flood frequency curve and mean annual water balance. Mean annual water balance is expressed in terms of two similarity measures: (i) the climatic aridity index, AI, which is a measure of the competition between energy and water availability and (ii) the base flow index, BFI, which is a measure of total runoff partitioning into surface and subsurface components at the annual time scale. The data analyses showed that the aridity index, AI, has a first-order control on the shape of the flood frequency curve (also known as the growth curve), as expressed in terms of both the mean and coefficient of variation (C_v) of the annual maximum floods, once normalized by catchment size (i.e., specific flood discharge) While the mean annual (specific) flood discharge decreases with increasing aridity, C_v increases with increasing aridity. On the other hand, the BFI was found to be a second-order control on the flood frequency curve. Higher BFI, meaning higher contributions of subsurface flow to total streamflow, leads to a decrease of the mean annual (specific) flood discharge, and vice versa. The statistically significant relationship between AI and the flood frequency curve and the consistent shift of the growth curves with AI support the use of AI as a similarity measure for regionalization of flood frequency.