The exploitation ecosystems hypothesis (EEH) proposes that 1) plant biomass reflects the primary productivity of an ecosystem modified by the regulating effect of herbivory, and 2) herbivore abundance reflects the productivity of plants modified by the regulating effect of predation. Primary productivity thus determines the number of trophic levels in an ecosystem and the extent to which bottom–up and top–down regulation influence the biomass ratios of adjacent and non-adjacent trophic levels (i.e. trophic cascading). We constructed an interactive model of plant (pasture), herbivore (red kangaroo Macropus rufus) and predator (dingo Canis lupus dingo), a system in which trophic cascades have been suggested to occur, and used it to test the effects of increasing stochastic variation in primary productivity and dingo culling on predictions of the EEH. The model contained four feedback loops: the predator–herbivore and herbivore–plant feedback loops, and the predator and plant density-dependent feedback loops. The equilibrium conditions along the primary productivity gradient reproduced the three zones of trophic dynamics predicted by the EEH, plus an additional zone at productivities above which the maximum density of a predator is achieved due to social regulation: that zone is characterized by increasing herbivore density and decreasing plant biomass. Culling dingoes produced trophic cascades that were strongly attenuated at primary productivities below which the maximum density of dingoes was attained. Results were robust to uncertainty in kangaroo off-take by dingoes and to the efficacy of dingo culling, but prey switching by dingoes from red kangaroos to reptiles would weaken trophic cascades. We conclude that social regulation of carnivores has important implications for expression of the EEH and trophic cascades, and that attenuation of trophic cascades increases with increasing stochasticity in primary productivity. Our model also provides a framework for understanding the conditions in which dingo-mediated trophic cascades might be expected to occur, and generates testable predictions about the effects of higher dingo densities (e.g. by stopping culling or reintroduction to former range) on kangaroo and pasture dynamics.