The exploitation ecosystems hypothesis (EEH) makes predictions about trophic interactions along gradients of primary productivity. The EEH has been shown to apply to a wide range of terrestrial environments but its applicability to arid environments has received little attention. One reason for this is that arid environments may not satisfy the assumptions of the EEH because dearth of water may limit biological activity in both temporal and spatial contexts. The EEH predicts that herbivore biomass should increase linearly with primary productivity in the absence of predators; but when predators are present herbivore biomass will remain relatively constant due to top down regulation. We tested this prediction in an arid environment using rainfall as a proxy of primary productivity and an index of the abundance of the dominant herbivores (kangaroos Macropus spp.). We compared an index of kangaroo abundance at 18 areas situated along a gradient of mean annual rainfall in areas where a top predator (the dingo Canis lupus dingo) was rare and common. We also explored the relationship between the density of artificial water points (AWPs) and kangaroo abundance to investigate if the resource subsidy provided by AWPs allows kangaroos to persist in high numbers. Consistent with the EEH, kangaroo abundance showed a weak relationship with mean annual rainfall in the presence of dingoes but increased with increasing annual rainfall in the absence of dingoes. The density of AWPs was a poor predictor of kangaroo abundance. Our analysis of macro-ecological patterns suggests that kangaroo populations are primarily top down regulated in the presence of dingoes, but are bottom up regulated in the absence of dingoes. Our findings provide evidence that top down regulation can prevail over bottom up regulation of herbivore populations in arid ecosystems and highlights the usefulness of the EEH as a predictor of macro-ecological patterns of species abundance.