This paper investigates the link between development, economic growth, and the economic losses from natural disasters in a normative analytical framework, with an illustration on hurricane flood risks in New Orleans. It concludes that, where capital accumulates through increased density of capital at risk in a given area, it is optimal for (i) the probability of disaster occurrence to decrease with income; (ii) the capital at risk – and thus the economic losses in case of disaster – to increase faster than economic growth; (iii) the average annual losses to grow faster than income at low levels of development and slower than income at high levels of development. In that case, increasing risk-taking reinforces economic growth, and improving protections transfer risks from frequent low-intensity events to rarer high-impact events. These findings are robust to a broad range of modeling choices and parameter values, and to the inclusion of risk aversion. Risk-taking is both a driver and a consequence of economic development, and should not be indiscriminately suppressed. The observation of a trend in disaster losses should not be confused with the presence of excessive risk taking. In a descriptive framework, suboptimal decision-making (the introduction of prospect theory's decision weights, biases in risk perception and myopic expectations) may amplify these trends and lead to excessive or insufficient risk taking. In all instances, the world is very likely to experience fewer but more costly disasters in the future.