Disaster loss trends and the burden of proof

2 June 1, 2018 at 4:29 pm by

When illustrating the general trend in disaster losses just a handful of graphs tend to get used the most.

When showing the international trend, the most frequently used exhibits are those published by Swiss Re and Munich Re.

For the Canadian trend, the most frequently used graph is one published by Insurance Bureau of Canada (IBC).

The two most commonly used Swiss Re and Munich Re graphs look at both the number of natural disasters recorded per annum and the insured costs of those same disasters.

Swiss Re data begins at 1970 and Munich Re data begins at 1980.

Each company uses specific criteria with particular thresholds to determine what to include in their respective analysis.

For Swiss Re, the company uses insured loss figures (for ‘Maritime disasters’, ‘Aviation’ and ‘Other’) or Total Economic Losses or Casualties (‘Dead or missing’, ‘Injured, and ‘Homeless’). Munich Re uses similar metrics.

Many often ask why it is necessary to include dead, missing, injured and homeless in the data. Why not just use insured losses?

Casualty data is used primarily because insurance penetration is low in developing countries, so using insured loss data would only provide a good indication of loss experience in the industrialized countries. In order to get a full and accurate international picture of events and losses it is necessary to include a measure other than insured losses.

Unlike Swiss Re and Munich Re, IBC does not publish a graph showing number of Canadian loss events, only insured losses.

Bureau data begins at 1983. From that year to 2007, IBC uses data it collected itself through various company surveys conducted immediately after significant natural disaster events. It also uses various data from Property Claim Services (PCS), Swiss Re, Munich Re and Deloitte. After 2007, the Bureau only uses data from Catastrophe Indices and Quantification Inc. (CatIQ).

In all cases, the Swiss Re, Munich Re (both international) and IBC (Canada) data show a trend that is unmistakably upward: Over the periods in question, there have been both more natural disasters and higher associated costs.

But there are those out there (including climate change deniers) who maintain that the graphs are misleading, because while the data is normalized for inflation, no attempt has been made to take growth in population, economic activity, insurance premiums and development into consideration. When you factor these in, they argue, natural disaster severity may not be going up at all and, therefore, the charts may be giving a false impression.

Now, this is where the discussion can get messy to be sure.

Normalizing disaster loss data to include such factors as growth in population, economic activity and building stock is not a simple undertaking. Further, there are many problems with using simple measures like GDP or insurance premium growth as a normalizer.

For these and other reasons, I don’t want to go ‘there’ at this point, as such a discussion could not possibly be handled properly in a space like this. Further, it is something of a ‘Holy Grail‘ issue, as no one has adequately ‘cracked the code’ to be able to say with high statistical certainty that disaster losses are really going up or down when all factors are taken into consideration. It is what some might call a ‘wicked problem’.

But here is my take on why it is wrong to simply dismiss the Swiss Re, Munich Re and IBC graphs merely because they haven’t taken the above factors into consideration.

These graphs do not attempt to explain why disaster losses are going up, and no commentary is made on the apparent trend. They are simply plotting the number of events recorded and their associated costs. That is, the producers are not making any particular claims about the data, they are simply presenting the record (just as some plot trends in steel industry production, number of vehicles licenced or U.S. visitors to Canada). The disaster loss graphs are showing what insurers and reinsurers have paid out for natural disasters over a given time frame. That’s it.

Now, if the respective firms were using the graphs to make a particular argument, that would be a different story. They would then be expected to apply rigour to show a true underlying tendency and prove a cause.

However, producers of the graphs do not use them to attempt to explain the tendency or give a particular reason why the trend is going up. And when they do mention such things as climate change, they make no claim that that is the singular reason why disaster losses are rising (indeed, it is generally accepted by these companies and others that losses are going up primarily due to increases in insurable values, often in at-risk areas).

Now, if a third party claims that the graphs prove that the upward trend in catastrophe losses is solely or primarily due to climate change, increases in insurable equity or the price of beans, that’s their issue (and it is then incumbent on them to provide rigour-based evidence to back up their claim). But this has nothing to do with the original intent of the producer of the exhibit.

So how Swiss Re’s, Munich Re’s or IBC’s graphs are used (or misused) is neither the fault nor the concern of these entities, just as it is not the fault of a sporting goods manufacturer if one of their baseball bats is used to club someone over the head.

So, simply put, those who push to dismiss the graphs as inaccurate and misleading are barking up the wrong tree.

Note: By submitting your comments you acknowledge that insBlogs has the right to reproduce, broadcast and publicize those comments or any part thereof in any manner whatsoever. Please note that due to the volume of e-mails we receive, not all comments will be published and those that are published will not be edited. However, all will be carefully read, considered and appreciated.