The dynamic equilibrium model of species diversity predicts that ecosystem productivity interacts with disturbance to determine how many species coexist. However, a robust test of this model requires manipulations of productivity and disturbance over a sufficient timescale to allow competitive exclusion, and such long-term experimental tests of this hypothesis are rare. Here we use long-term (27 years), large-scale (8 × 50-m plots), factorial manipulations of soil resource availability and sheep grazing intensity (disturbance) in grasslands to test the dynamic equilibrium model. As predicted by the model, increased productivity not only reduced plant species richness, but also moderated the effects of grazing intensity, shifting them from negative to neutral with increasing productivity. Reductions in species richness with productivity were associated with dominance by faster growing (i.e. high specific leaf area) and taller plants. Conversely, grazing favoured shorter plants and this effect became stronger with greater productivity, consistent with the view that grazing can lead to weaker asymmetric competition for light. Our study shows that the dynamic equilibrium model can help to explain changes in plant species richness following long-term increases in soil resource availability and grazing pressure, two fundamental drivers of change in grasslands worldwide.