One of the challenges to quantifying the costs and benefits of symbiosis is that symbionts can influence different components of host fitness. To improve understanding of the ecology of inherited symbionts, we developed general theory for a perennial host-hereditary symbiont interaction, in which symbionts can have independent and potentially opposing effects on host regeneration and survival. The model showed that negative effects on one component of fitness may be outweighed by positive effects on another, leading to a net positive impact of symbiosis on population growth. Model predictions depended on the availability of suitable patches, which influenced the relative contributions of survival vs. regeneration to host fitness. We then used experimental symbiont removal to quantify effects of a hereditary, fungal endophyte on a grass host. Endophyte presence strongly reduced host survival but increased regeneration. Application of the model revealed that negative effects on plant survival were overwhelmed by beneficial effects on regeneration, resulting in stable endophyte persistence at 100% frequency, consistent with field observations. Our work demonstrates the utility of a demographic perspective for predicting the dynamics of symbioses and supports the hypothesis that symbionts function as mutualists when host and symbiont fitness are coupled through vertical transmission.