As someone who has spent some time looking at changes in the incidence of hurricanes around the planet, I have been asked to provide a response to Roger Pielke Jr.’s article “Disasters Cost More Than Ever — But Not Because of Climate Change,” published at FiveThirtyEight earlier this month.
Let me begin by saying that I am sympathetic to Pielke’s emphasis on the role of changing demographics in increasing damages from natural disasters. This is a serious problem that could be addressed by wiser policies. For example, in the United States, policies regulating insurance and providing federal flood insurance and disaster relief have the effect of subsidizing risk-taking, and the recent repeal of large sections of the 2012 Biggert-Waters Federal Flood Insurance Reform Act shows just how difficult it is to reform these risk-inducing policies.
Having said that, I’m not comfortable with Pielke’s assertion that climate change has played no role in the observed increase in damages from natural hazards; I don’t see how the data he cites support such a confident assertion. To begin with, it’s not necessarily appropriate to normalize damages by gross domestic product (GDP) if the intent is to detect an underlying climate trend. GDP increase does not translate in any obvious way to damage increase; in fact, wealthier countries can better afford to build stronger structures and to protect assets (for example, build seawalls and pass and enforce building regulations).1 A grass hut will be completely destroyed by a hurricane, but a modern steel office building will only be partially damaged; damage does not scale linearly with the value of the asset.
More seriously, a casual inspection of both graphs (normalized and non-normalized damage over time) presented by Pielke leads me to question the statistical significance of either. This is hardly surprising, since 23 years is not a very long time to detect trends in natural hazard damages, whether such trends are caused by demographics or by climate change. A 2012 study2 by London School of Economics researchers Fabian Barthel and Eric Neumayer looked at damage trends normalized by GDP, a measure they used because others are not universally available. For Germany and the United States, with 29 and 36 years of data, respectively, they detected “statistically significant upward trends in normalized insured losses from all non-geophysical disasters as well as from certain specific disaster types,” but for the globe as a whole, with 19 years of data available, they could find no significant trends.
Since the U.S. alone accounted for roughly half the insured losses over this period, the significance of the longer U.S. record and lack thereof in the shorter global record suggests that 20 years may be too short to detect significant trends. The increasing normalized trends in the U.S. were evident in convective storms, winter storms, flooding events and high temperature-related losses, and were almost statistically significant for hurricanes at the conventional 95 percent confidence level.3 In view of data like this, it’s very hard to accept Pielke’s confident assertion that “[n]o matter what President Obama and British Prime Minister David Cameron say, recent costly disasters are not part of a trend driven by climate change.”
There is an even more significant problem with Pielke’s analysis. In a nutshell, he addresses trend detection when what we need is event risk assessment. The two would be equivalent if the actuarial data was the only data available pertaining to event risk. But that is far from the case; we often have much more information about risk.
Let me illustrate this with a simple example. Suppose observations showed conclusively that the bear population in a particular forest had recently doubled. What would we think of someone who, knowing this, would nevertheless take no extra precautions in walking in the woods unless and until he saw a significant upward trend in the rate at which his neighbors were being mauled by bears?
The point here is that the number of bears in the woods is presumably much greater than the incidence of their contact with humans, so the overall bear statistics should be much more robust than any mauling statistics. The actuarial information here is the rate of mauling, while the doubling of the bear population represents a priori information. Were it possible to buy insurance against mauling, no reasonable firm supplying such insurance would ignore a doubling of the bear population, lack of any significant mauling trend notwithstanding. And even our friendly sylvan pedestrian, sticking to mauling statistics, would never wait for 95 percent confidence before adjusting his bear risk assessment. Being conservative in signal detection (insisting on high confidence that the null hypothesis is void) is the opposite of being conservative in risk assessment.
When it comes to certain types of natural hazards, there are more bears in the woods. For example, there is a clear upward trend in overall North Atlantic hurricane activity by virtually all metrics, over the past 30 years or so, though the cause of this is still uncertain. But given that only about a third of Atlantic hurricanes strike the U.S.; hurricanes do damage during a very small fraction of their typical lifetimes; and only intense hurricanes (a small fraction of the total) do significant damage, the amount of hurricane data pertinent to U.S. damage is a tiny fraction of the entire database of North Atlantic hurricanes. Thus it is hardly surprising that the upward trend in U.S. hurricane damage is of only marginal statistical significance, and Pielke’s own analysis shows that it takes several decades for such a trend to emerge.
This does not mean that there is no underlying change in the risk, and the priors we have in this case point to a significant increase in such risk. One would be foolish to make plans that have to deal with U.S. hurricane risk without accounting for the evidence that the underlying risk is increasing, whether or not actuarial trends have yet emerged at the 95 percent confidence level.
This is particularly so when one accounts for another form of prior information: theory and models. While some disagreement remains about projections of the weakest storms, which seldom do much damage, both theory and models are now in good agreement that the frequency of high category hurricanes should increase, as should hurricane rainfall and the flooding it produces.
Looking ahead, I collaborated with Yale economist Robert Mendelsohn and his colleagues in estimating global hurricane damage changes through the year 2100, based on hurricanes “downscaled” from four climate models. We estimate that global hurricane damage will about double owing to demographic trends, and double again because of climate change. These projections are not inconsistent with what we’ve been seeing in hurricane data and in economic damage from hurricanes. Besides this study, there are robust theory and modeling results that show increased risk of hydrological extremes (floods and droughts) and heat-related problems.
Some of these predicted trends are beginning to emerge in actuarial data. Governments, markets and ordinary people are beginning to account for the increased risk. Those who wait for actuarial trends to emerge at the 95 percent confidence level before acting do so at their peril.