Climate Change and Risk Management: Challenges for Insurance, Adaptation, and Loss Estimation

Executive Summary Adapting to climate change will not only require responding to the physical effects of global warming, but will also require adapting the way we conceptualize, measure, and manage risks. Climate change is creating new risks, altering the risks we already face, and also, importantly, impacting the interdependencies between these risks. In this paper we focus on three particular phenomena of climate related risks that will require a change in our thinking about risk management: global micro-correlations, fat tails, and tail dependence. Consideration of these phenomena will be particularly important for natural disaster insurance, as they call into question traditional methods of securitization and diversification. Global micro-correlations between variables—such as investment products, insurance policies, or catastrophe bonds—obtain when all variables in a relevant class have small positive correlations that would be imperceptible if the variables were considered in isolation. When such variables are aggregated, the correlations of the aggregations will balloon. If micro-correlations are ignored when calculating reserves under a Value-at-Risk management regime, risks will be woefully underestimated, as seen in the figure to the left. A manager may think his residual risk is 0.1%, when in reality it is much higher. Moreover, risk underestimation grows with the number of products securitized. Fat tailed loss distributions resulting from uncertainty in parameters of " thin tailed " distributions have recently captured the imagination of economists, thanks to the work of Martin Weitzman. Catastrophe modeling firms study the ways in which extreme events compound damages to create fat tails and damages from many extreme events have been shown to fit fat tailed distributions. Financial risk managers identify fat tails by " mean excess " plots, which show the expected value of damage variable X above threshold x, as a function of x. Thin tailed distributions have constant mean excess, but fat tails push this excess ever upward. This increasing mean excess is seen in the figure here, ii which is based on flood damages in Saint Louis County, Missouri between 1991 and 1995. Fat tails increase the solvency risk, thus mooting the " actuarial fair price " of insurance. Tails may become fatter under aggregation, but they may not; and indeed aggregation is sometimes an effective abatement strategy. Tail dependence is distinct from simple correlation and tail fatness. It occurs when dependence concentrates in high values. Distributions with high correlation may nonetheless be relatively independent except for very high values. Tail …