I study how attitudes towards risk and risky behavior adapt over the long-run to changes in risk in individuals' environments. In the first chapter I build a dynamic model of choice in which agents' endogenous foreground risk-taking adapts to evolving beliefs about exogenous background risk. The key features of this model are that foreground and background risk are substitutes, and that the agent learns about both the mean and variance of the background risk from observing its realizations over their lifetime. My model makes sharp predictions about the dynamics of risk-taking, the three most important being that (1) risk aversion increases monotonically in variance increases relative to the prior variance, and is (2) decreasing and (3) convex in mean increases about the prior mean. In the subsequent two chapters I test the predictions of my model for two sources of background risk, using large-scale panel surveys from Indonesia and Mexico containing repeat elicited measures of risk aversion for the same subjects years apart. In chapter two, coauthored with Daniela Vidart, we link changes in measured risk aversion to state-level real GDP growth time series capturing subjects' lifetime macroeconomic experiences. In line with the model's predictions we find that measured risk aversion increases in growth volatility and decreases in mean growth. These findings are robust to controlling for changes in subjects' personal economic circumstances and experiences of violence and natural disasters. Decreases in measured risk aversion correlate with substantial increases in risky behavior in the domains of health, migration, and occupational choice. In chapter three, coauthored with Wesley Howden, we link changes in measured risk aversion to state-level temperature and precipitation time series measuring subjects' lifetime experiences of climate change. We find that measured risk aversion increases in the volatility of temperature in Indonesia and precipitation in Mexico, that it decreases in the mean of temperature in both countries and precipitation in Mexico, and that it is convex in the mean of temperature in Indonesia. These findings are robust to the inclusion of controls. Decreases in measured risk aversion correlate with large changes in migration, occupational choice, and agricultural investment behavior.