Individual observations of risky behaviors present a paradox: individuals who take the most risks in terms of hazards (smoking, speeding, risky sexual behaviors) are also less likely to take risks when it comes to innovation, financial risks or entrepreneurship. Existing theories of risk-preferences do not explain these patterns. From a simple model, we argue that many decisions involving risk have a temporal dimension, and that this dimension is often the main determinant of individual choices. In many real life instances, risk taking amounts to damaging the individual's capital (whether embodied capital, financial capital, social reputation, etc.), which would affect her over a long period of time after the risky decision. In evolutionary terms, the marginal cost of this type of risky behavior depends on the relative importance of the future in the individual's fitness (e.g. her time horizon). Because the cost of a degradation of their capital will be paid for an effective shorter period of time, individuals with short time horizons will give less importance to this degradation. This approach explains patterns of behaviors observed across socio-economic groups and puts forward new approaches to prevent hazardous behaviors such as smoking.