Senescence evolves because natural selection is less sensitive to late life than to early life. Hamilton formalized this as the sensitivities of fitness to small additive changes to age-specific mortality or fecundity; his framework has since been extended to alternative ecological and genetic assumptions. However, such forces of selection only explicitly model evolution in the short term; in the long term, as life histories evolve, their forces of selection evolve too. This thesis investigates long-term evolution of senescence by deriving conditions a population must satisfy in order to be in evolutionary equilibrium. It considers two models: a mutation-selection balance model; and an optimality model with a same-age log mortality cost of fecundity. The derived conditions are discussed, and heuristically compared to two species: Soay sheep, which senesce, and desert tortoises, which don't. Other, intermediate theoretical results are also given, including the force of selection on proportional hazards in stationary populations.