19 August 2010

Reinsurance Innumeracy

I was intrigued to see the following in a Swiss Re press release a reader sent to me this morning (thanks FN):
Climate change could significantly increase the risk of hurricanes and storms in the Caribbean and threaten future development in the region, concludes a new study released by the Caribbean Catastrophe Risk Insurance Facility (CCRIF). Damage from wind, storm surge and inland flooding already amounts to 6% of GDP per year in some countries, according to the study’s preliminary results. Under a high climate change scenario, annual expected losses could rise by another 1 to 3% of GDP by 2030.
The statement is interesting because of the very large increase in projected hurricane losses to 2030, even under the most extreme scenario. So I took a look at the underlying report, which was produced by the intergovernmental CCRIF supported by reinsurance companies including Swiss Re. And like a lot that you find in the grey literature related to climate change, it does not hold up so well.  Here are two reasons why, and they have nothing to do with the report's cherry picking of extreme scenarios or even the validity of those scenarios.  My critique below takes the climate part of the methodology as given (I didn't even look at the climate part of the methodology, which has its own obvious problems).

First, the report (here in PDF) combines projected future damage resulting from GDP growth and projected climate change and calls the total "climate change."  Not good.  You can see this in the figure below from the report. I have placed a red circle around the report's breakdown of the sources of future increases in damage, and you can see that a significant part is due solely to an "increase due to asset growth."

Then in the left panel you can see the total increase reported (in the green oval that I added), with a 50% increase in losses as a proportion of GDP (from 6% to 9%, and the difference of 3% is 50% of 6%).  By the time this makes its way to the press release it is characterized as
Findings from the study indicate that annual expected losses from wind, storm surge and inland flooding already amount to up to 6% of GDP in some countries and that, in a worst case scenario, climate change has the potential to increase these expected losses by 1 to 3 percentage points of GDP by 2030.
That is just wrong and misleading. Not good.

The second problem with the report is that while it takes an extreme scenario for climate change, it takes a single apparently conservative scenario for GDP growth.  For instance, the report assumes a 1.2% per year GDP growth for Jamaica, as compared to a 1.8%per year increase in damages due solely to climate change.  If the report were to instead assume a 2% per year annual growth rate (as the Jamaican government does for 2001/12) then hurricane damage would decrease as a proportion of GDP by 2030, because economic growth would outstrip the independent effects of climate change.

A better conclusion from this report would be that climate change -- even under the most extreme scenarios -- might increase or decrease future Caribbean hurricane damage relative to GDP, or even have no discernible effect, but the policy options that make sense in this region are insensitive to these uncertainties.

I see that some in the media have already uncritically repeated the misleading conclusions from the report.  Let's see if anyone else does.

[UPDATE 8/20: Climate Progress fell for it.]

14 comments:

  1. This post has been corrected to properly characterize CCRIF as an intergovernmental organization.

    ReplyDelete
  2. Hi Roger,

    Yes, this is hilarious. The $844 million for in damages in 2030 is compared to the *2009* GDP.

    For example, using only numbers from the figure:

    1) In 2009, storm damage was $479 million, which is stated to be 6% of GDP. That would make *2009* GDP equal to $7.98 billion. (Note that this appears to be much lower than the actual GDP of Jamaica in 2009, which wonderful Wikipedia estimates at $23.25 billion.)

    http://en.wikipedia.org/wiki/Economy_of_Jamaica

    2) But let's stay with the $7.98 billion in 2009. Then grow it by 1.2% per year to 2030. We get $7.98 x 1.012^21 = $7.98 x 1.28 = $10.2 billion.

    3) So the damages in 2030 should be compared to the GDP in 2030...not the GDP in 2010. In fact, if they had gone far enough into the future, they would have gotten damages more than 100% of GDP (*2009* GDP, of course)! :-)

    ReplyDelete
  3. Hi Roger,

    Scratch that last comment. I don't know what's going on. Maybe they just rounded $844 million divided by $10.26 billion...or 8.22 percent... up to 9 percent.

    What a mess!

    Mark

    ReplyDelete
  4. Hi Roger,

    Did you see the headline on your link to Bloomberg?

    "Caribbean Storm Damage Costs May Rise 505 With Global Warming"

    6! 42! 35! Hike!

    ReplyDelete
  5. If Swiss Re, or any of the European Re's, really believe this, all they have to do is impose restrictive sublimits in their treaties. Primaries would not be able to write but a fraction of the risk. The outcry would be enormous forcing something to happen. In fact the Re's are derelict towards their owners if they do not act, IF they really believe this nonsense.

    ReplyDelete
  6. I dunno about innumeracy.

    I did see in the report that despite the vast majority of the damage in Jamaica coming from wind they concluded that adopting and enforcing 'wind' building codes wasn't worth doing.

    In my over-simplistic mind if someone lives in Hurricane Alley and concludes that a wind building code isn't worth the cost then I'm not gonna cry when bad old nature blows their houses down.

    ReplyDelete
  7. Well since the Insurance Companies can see so well so far into the future, they better raise their premiums they charge right now.

    Wouldn't want the insurance companies to get into financial difficulties just because they failed to raise rates soon enough.

    ReplyDelete
  8. Keep in mind to follow the money trail. Swiss Re has bought into the Chicago Climate Exchange and is a member of The Global Roundtable on Climate Change.
    Swiss Re Global Markets Limited is involved in the carbon credits trading business. Pay no attention to that man behind the curtain.

    ReplyDelete
  9. While "hurricanes and storms in the Caribbean" are features of weather, the probability of those events is a characteristic of climate. So an increase in that probability is climate change. So climate change causes changes in climate. As you Americans say, "Who knew?"

    When they substituted "climate change" for "global warming" - which is at least meaningful whether it right or wrong - they removed the sense. And when people don't care about the sense I know they intend to use the emotional impact of trigger words. And that makes it propaganda.

    Fortunately, I know what to do with propaganda.

    ReplyDelete
  10. I noticed the reinsurance industry move to use faux stats and hysterical misrepresentations of science irt climate concerns awhile ago. My bet is it is a combination of cynical exploitation and naive faith in the apocalyptic salesman of AGW.
    The larger question for me is this:
    The reinsurance companies are the bulwark of insurance reserves worldwide. If the reinsurance companies are as vulnerable to irresponsible policy making in other areas of risk management, how strong is this bulwark?
    In fact if the bulwark is weak, and a severe weather event takes place that finds reinsurers unable to perform their obligations, will they simply point to these faux statistics and use those as a rationalization for other financial failures?

    ReplyDelete
  11. Hi Roger,

    Although I agree that the Swiss Re report does not compare the right number, I do not understand how climate change might decrease Caribbean hurricane losses, as you state. As you have shown in a paper on future hurricane losses, risk is a product of GDP (proxy for exposure) and hazard probability.

    Now assuming that climate change can either have no effect, or leads to an increase in hazard probability, losses as % of GDP would either stay the same or increase. Or would you argue that in (another) extreme scenario, hurricane probabilities might decrease?

    In my view, only vulnerability reduction / adaptation can lead to decreases.

    ReplyDelete
  12. -11-Laurens

    There are two answers ...

    1. The post refers to a change in losses as a fraction of GDP, not total losses. Total losses can increase but diminish as a fraction of GDP, if GDP grows at a faster rate.

    2. The Bender et al. 2010 modeling study of the effects of GHGs on hurricanes did indeed include several model results in which human-caused climate change led to a decrease in storm activity, and they project (logically) a decrease in losses.

    ReplyDelete
  13. I agree, but

    1) the CCRIF study equates loss potential growth with GDP growth. So according to this logic the share of losses can never decline, if hazard probability remains the same.

    2) Bender et al. suggests declines, but based only on two GCMs, of which only one shows overall declines and the other still shows substantial loss increases from cat 4 and 5 hurricanes. The overall ensemble shows increases.

    Now the CCRIF study probably does not include possible declines in hurricane losses, as the ECA study for Atlantic hurricanes only mentions 3-5% wind speed increases (see link below, at page 107). Of course, providing results for different or high and low scenarios (as most impact studies do) would be better.

    http://www.mckinsey.com/App_Media/Images/Page_Images/Offices/SocialSector/PDF/ECA_Shaping_Climate%20Resilent_Development.pdf

    ReplyDelete
  14. -13-Laurens

    In the Jamaica case calculate that the CCRIF study has the GDP growth rate at 1.2% per year and the independent effect of climate change as 1.8%. Under this scenario losses will increase a a proportion of GDP. However, what if GDP increases by 2.0% per year? In this case the size of the economy will grow faster than the independent effects of climate change.

    And you are right, the CCRIF does not include a range of modeling studies. That point is simply to show that all possible climate change futures are not necessarily loss enhancing.

    ReplyDelete