The reaction of the body to prolonged stress has many harmful effects. Classical theory assumes that stress responses have evolved due to their short-term selective advantages ('flight or fight'), and despite their adverse long-term effects. In contrast, we demonstrate that the adverse effects of stress responses may have a selective advantage. Using an analytical model we show that a gene that causes the early death of a relatively unfit individual can increase in frequency in a structured population even if it has no positive effect on that individual. This result offers a new perspective on the relations between stress factors, stress responses and stress-related diseases.