Managing the Risks of Extreme Events in a Changing Climate
March 18, 2009
Some research suggests that climate change may increase the magnitude or frequency of extreme events, such as floods, hurricanes, and droughts. Measuring and managing this increase in risk is a major challenge to policymakers, civic planners, and private-sector strategists. In a new series of papers, RFF Fellow Carolyn Kousky and Senior Fellow Roger Cooke explore how the risks of extreme events may be changing with climate change and how our management of these risks can be improved.
In "Climate Change and Risk Management: Challenges for Insurance, Adaptation, and Loss Estimation," (RFF Discussion Paper 09-03) Kousky and Cooke examine three aspects of damage distributions: micro-correlations, fat tails, and tail dependence. First, micro-correlations are tiny, positive correlations between variables, such as insurance policies, that could be introduced from climate change. These correlations could be so small that in isolation they would be undetected, but when policies are aggregated, the correlations balloon. As an example of climate dependencies, they point to El Niño years, where "precipitation is likely to be more extreme in California, leading to mudslides and floods; nutrient-poor water is likely to cause fish catch declines in Peru; and drier conditions are more likely in Australia, increasing the chance of bushfires."
The second aspect of damage distributions they examine are "fat tails" (when the probability of a really bad outcome occurring falls less rapidly than on a normal distribution or bell curve). In this context, they are distributions where extreme events may be much more likely than we think. Damages from many natural disasters have been shown to be closely approximated by fat-tailed distributions, and Kousky and Cooke document them in flood and crop insurance data in the United States.
Finally, they also examine tail dependence, or the likelihood that bad outcomes can occur together. For instance, some insurance lines may be largely independent except when there is a catastrophe and then they are seen to be dependent. Tail dependence may also balloon with aggregation as the authors demonstrate with insurance claims data.
If risk managers are not cognizant of the presence of micro-correlations, fat tails, or tail dependence the authors warn, "Risks will be woefully underestimated." More problematically, traditional methods of diversification may not operate in situations of micro-correlations and tail dependence, creating the need for innovative risk management strategies. The authors mention the possibility of catastrophe bonds with multiple triggers for dealing with tail dependence, for example. They also discuss the possibility of de-coupling tail dependent risks through appropriate adaptation policies. In sum, "These intricate dependencies should make us cautious of simple or ad hoc consideration of climate-related risks," they write.
A Focus on Floods
In "Improving Flood Insurance and Flood Risk Management: Insights from St. Louis, Missouri," (RFF Discussion Paper 09-07) Kousky and Howard Kunreuther of the Wharton School review the history of flood risk management in St. Louis, Missouri over the past 15 years. They identify six challenges to the continued management of riparian flood risk:
1. many property owners don't buy flood insurance;
2. people underestimate flood risk;
3. we need better flood maps;
4. we have a "love affair" with levees;
5. flood risk is increasing over time; and
6. we take deep pride in rebuilding after a disaster.
The authors offer recommendations for improving our management of flood risk—particularly the National Flood Insurance Program—in light of these challenges.
Climate Change and Risk Management: Micro-correlations, Fat Tails, and Tail Dependence