How pathogen spillover influences host community diversity and composition is poorly understood. Spillover occurs when transmission from a reservoir host species drives infection in another host species. In cheatgrass-invaded grasslands in the western United States, a fungal seed pathogen, black fingers of death (Pyrenophora semeniperda), spills over from exotic cheatgrass (Bromus tectorum) to native perennial bunchgrasses such as squirreltail (Elymus elymoides). Previous theoretical work based on this system predicts that pathogens that spill over can favor either host coexistence, the exclusion of either host species, or priority effects, depending on species-specific transmission rates and pathogen tolerance. Here, these model predictions were tested by parameterizing a population growth model with field data from Skull Valley, Utah, USA. The model suggests that, across the observed range of demographic variation, the pathogen is most likely to provide a net benefit to squirreltail and a net cost to cheatgrass, though both effects are relatively weak. Although cheatgrass (the reservoir host) is more tolerant, squirreltail is far less susceptible to infection, and its long-lived adult stage buffers population growth against seed losses to the pathogen. This work shows that, despite pathogen spillover, the shared pathogen promotes native grass persistence by reducing exotic grass competition. Counterintuitively, the reservoir host does not necessarily benefit from the presence of the pathogen, and may even suffer greater costs than the nonreservoir host. Understanding the consequences of shared pathogens for host communities requires weighing species differences in susceptibility, transmission, and tolerance using quantitative models.