‘Barton Keyes’ is a pseudonym (which I picked) as the insurance industry insider who penned this article needs to remain anonymous. This article, which tells of how insurance companies will raise prices because of exaggerated, unrealistic threats, deserves wide circulation.
It has become commonplace to hear dire predictions about earth’s fate if mankind does not do something to stem the increasing threat of climate change. Climate change is predicted to cause massive destruction to the natural environment, damage health, shorten human lifespans, and to be a threat to national security. Calls to stop climate change have been made by the United Nations, corporations, environmental groups, and even the Pope.
The insurance industry has not been given a pass when it comes to doing their part to combat climate change. I work for an insurance company and recently participated in a meeting where our chief pricing actuary shared that he has been asked by regulators to explain what we are doing to account for the increased risk of loss due to climate change. This immediately caught my attention and I began to do some research on the topic which I would like to share with you.
The best single source I found is the “Insurer Climate Risk Disclosure Survey Report & Scorecard: 2014 FINDINGS & RECOMMENDATIONS” (from now on I will call it “the report”) which discusses the state of U.S. insurers’ preparedness for the expected consequences of climate change. The report was released by the Ceres organization that describes itself as:
Ceres is a nonprofit organization advocating for sustainability leadership. It mobilizes a powerful network of investors, companies and public interest groups to accelerate and expand the adoption of sustainable business practices and solutions to build a healthy global economy. Ceres also directs the Investor Network on Climate Risk (INCR), a network of over 100 institutional investors with collective assets totaling $13 trillion.
The report is a cornucopia of information about not only what is happening in the U.S. insurance space but also what is happening internationally. It answered my questions and left me concerned. Along with other similar organizations, the Ceres organization is absolutely convinced that man-made climate change is occurring and recommends a series of invasive and expensive solutions.
My first question was: which regulators are pushing our chief pricing actuary to describe how we are accounting for climate change? Page 4 of the report provided me with the answer:
This report summarizes responses from insurance companies to a survey on climate change risks developed by the National Association of Insurance Commissioners (NAIC). In 2013, insurance regulators in California, Connecticut, Minnesota, New York and Washington required insurers writing in excess of $100 million in direct written premiums, and licensed to operate in any of the five states, to disclose their climate-related risks using this survey.
There were five states requiring this information last year and by now there may very well be more requiring the survey be filled out. Please notice that there are no weasel words in that paragraph, these state regulators believe that climate change is happening and want to know what we are doing about it! Their certainty is addressed on page 3 of the report:
However, much of the insurance industry is still lagging on this important issue, particularly those in the Life and Annuity and Health sectors—and they cannot afford to. If 97 percent of our actuaries concluded there was going to be a decline in public health due to a medically identified epidemic, we would expect the firms we regulate to adjust their forecasts, premiums and policies accordingly. Failing to do so would be imprudent. With climate change, 97 percent of scientists in the field agree that it is a reality and are more focused on the timing and magnitude of changes and related damage we can expect. This industry should be focused less on what is causing climate change and more on how we respond to and mitigate it.
This is the famous 97% consensus statistic about climate change. I have never been a fan of “consensus science” which is an oxymoron to me. Wasn’t the phlogiston theory of combustion the consensus at one time? It was also completely wrong. However, even those people who put some credence in scientific consensus should know that the 97% consensus statistic has been debunked (please see “Cooking stove use, housing associations, white males, and the 97%” by José L. Duarte for the details). There is no 97% consensus! [Editor: also see the original papers debunking this; here and here.]
Evidence to the contrary notwithstanding the 97% consensus concerning climate change is the justification for the remainder of the report. The following sections provide the survey questions, how those questions were analyzed, and action items for U.S. insurers and regulators. Discussion of three broad insurance segments is provided: Property & Casualty Insurance, Life & Annuity Insurance, and Health Insurance. If you are a glutton for punishment I recommend that you read the report. However, for those of you who would like to spend you free time doing something else here is the short version:
- Due to increased severe storms and sea level rise there will be massive property damage going forward. Property & Casualty insurers are doomed unless we do something!
- Due to increased drought and environmental degradation there will be a lot more sick people. The Health insurers are doomed unless we do something!
- Due to losses in the financial markets (due to climate caused destruction of property) and the increase in disease in the general population the Life & Annuity insurers are doomed unless we do something!
Those few insurers taking climate change seriously are mentioned by name and congratulated in the report.
I want to point out that it is the view of the report writers that the cost of all insurance will dramatically increase due to climate change. It would be nice if the report’s writers were advocating that all insurers increase premiums to account for this and leave it at that. That way I might get some really fat bonuses for doing my part to stem the looming climate change disaster. Alas that does not appear to be the intent. The intent is to use that extra premium money for mitigation such as working with state and local governments, and not-for-profit organizations like Ceres, to prepare for the coming disaster. Oh well, in the event of massive increases in insurer profits the U.S. Congress would probably pass a “windfall profits tax” on the insurance industry. So my dreams of purchasing a yacht have slipped by once again.
An example of the potential cost impacts of the report’s recommendations can be seen if we unpack this recommendation from page 52:
Integrate Climate Change Considerations Into Catastrophe Models
As the end-users of catastrophe model software, P&C insurers should ensure that the latest climate science and projected climate impacts are being taken into account and modeled appropriately by their vendors. Accurately communicating the risks associated with climate change through pricing and underwriting is essential, and accurate catastrophe modeling is crucial in this regard.
To someone outside of property insurance this may seem innocuous enough but it isn’t. To increase the modeled losses due to hurricanes based on “projected climate impacts” will have serious consequences. Several years ago one of the leading vendors of catastrophe modeling software, Risk Management Solutions (RMS), released an updated hurricane model that increased modeled losses. The excellent article “Catastrophe Modeling: Why All the Fuss?” summarizes the effects of these changes as:
[The model update to] RMS v11 has been considered to be the equivalent of a $25 billion to $35 billion capital event in the property market, which in turn has led to an increased cost of doing business in CAT-exposed areas as insurers need to maintain higher capital adequacy ratios as measured by the rating agencies. This has led to carriers raising more capital, raising premiums, reducing their portfolio accumulations in high CAT areas, and/or purchasing more reinsurance.
The Ceres report talks about the impact of climate change in cataclysmic terms. If their projected climate impacts are incorporated into catastrophe modeling software I am confident that the change will cause much more than a “$25 billion to $35 billion” capital event in the property insurance market.
Many of the report’s recommendations will have effects similar to the catastrophe model recommendation mentioned above: to increase cost and decrease supply of insurance products. Implementing these recommendations will affect every person and business in the United States. Unless there is true scientific evidence (the debunked 97% climate change consensus statistic does not count as scientific evidence) our society should not misallocate so many resources in the name of climate change mitigation.
Buy insurance company stock now before the run up in profits and prices, but be ready to dump when the turn shows signs of arriving. You can’t teach a fool, but there are ways to protect yourself from his foolishness.
All,
More abuse of regression: ‘Economic losses from US hurricanes consistent with an influence from climate change’ from Francisco Estrada et al., in Nature Geoscience, 19 October 2015. The abstract:
I live in CT and in August my homeowner’s insurance went up 50% – a very noticeable increase. I did some shopping around and a few carriers quoted me similar prices to the increase, and a couple quotes were at my previous premium level. Looking at Appendix A of the report, I see my previous provider – Liberty Mutual – has a “Developing” score. Allstate also has a “Developing” score and gave me a similar quote. My new provider is State Farm and they have a “Beginning” score. Appendix A is a potentially useful guide for those shopping for home insurance.
Thank you, Matt, and thank you anonymous insurance insider, for bringing this insidiousness to public attention. I’m especially grateful for the warnings, in particular these two: 1) that insurers are not permitted by the states in question to disagree about the risk. They cannot make their own assessments and ‘bet’ on the risk being unreal, or less than advertised, and thus offer a lower rate than competitors; 2) that rate hikes are diverted into ‘mitigation’, which implies more funding for the phony-baloney jobs of the very people (government agencies, and the right kind of specially-selected NGOs) who have concocted this scheme to begin with .
This is a very serious misappropriation of resources, some of which will probably be used for advertising and ‘awareness’ programs to assure us ever more strenuously that this phony problem is real, that the higher rates are necessary, and that we should all be grateful that our officials are being suitably pro-active and requiring insurance companies to charge us higher rates.
And ironically the problem is real — because insurers are forced to act as if it is real. And that at least has real-world consequences.
Ms. Paige St. John and the “Herald Tribune” of Sarasota, FL won the 2011 “Pulitzer Prize” for investigative journalism. The multi-part, two year dive into the Reinsurance Industry. [Wiki still has a live link of the series. You will get an error code. However, the links are live.] I’ve copied one them below.
http://www.heraldtribune.com/article/20101114/article/11141026
The net of the story was: the Reinsurance Industry changed its historical actuarial data analysis to a computer model for storm damage potential. No surprise, the computer models raised the prices of insurance and changed the risk. Before I go on, the Reinsurance Industry are the companies that insure insurance companies. In other words, if Liberty Mutual wants to spread a potential risk, it would put bids to the Reinsurance firms. The majority of these firms are EU and Bermuda based.
So, what happened? 18 states and the District of Columbia approved the “Hurricane Deductible” Rider to storm damage policies. [ I was told following Sandy that my policy was grand-fathered and not subject to the deductible. However, my “Traveler’s” boiler plate described the Hurricane Deductible. The risk change is: “If a storm exceeds CAT 1 speed, the insured would be liable for the first 2% to 5% of the damages in addition to the deductible agreed to at the time of purchase. So, if a family bought a $ 2,000. deductible, they would have a surprise. A $200,000.00 homeowner would be responsible for his/her $2000.00 deduction plus the added $4,000.00 to $10,000.00 for the Hurricane Deductible.
When “Sandy” came ashore, New Jersey Gov. Christy and New York Gov. Cuomo were quick to announce that the storm was 1 MPH below CAT 1. The New Jersey newspapers wrote: New Jersey home owners “dodged a bullet”.
Dr. Roger Pielke, JR. observed on his blog that, in his opinion Sandy’s speed was reduced.
The irony is: the U.S. is at an all time record for the lack of land fall CAT 3 and above storms.
There is an argument for the insurance risk. I grew up on the Jersey Shore. In the 1950s and 1960s, hurricanes were a norm. People and developers understood the risk associated with construction on the “Barrier Spits” along the Atlantic Coast. Beginning in the late 90s, we now have multi-million dollar estates on these multi-state stretches. A close in CAT3 would be ruinous.
Regards
I’m a retired reinsurance actuary. Commenter Robert Doyle points out that, “the Reinsurance Industry changed its historical actuarial data analysis to a computer model for storm damage potential.” This change happened quite a few years ago. The main reason was that the historical data analysis approach sucked. There were too few enormous losses in the data to make a credible basis for prediction. A lot of guesswork went into the probabilities for large hurricanes. From a technical POV, today’s models are much superior. (Note, there are several models used for this purpose, not just one.) The historic actuarial data base showed costs only. The new models are based on a data base of actual wind events with a function to translate each event into cost per event.
The year 2005 was disastrous. Not only Katrina, but several other very large hurricanes hit the US. It was natural to believe that climate change had dramatically shifted the hurricane risk. The various models were juiced up to assume greater frequency of hurricanes. OTOH since 2005, no Catagory 3, 4 or 5 (called “Major”) hurricane made landfall in the US. This is the longest such hiatus in Major Hurricanes.
Since I’ve been retired for some years, I don’t know whether the models have been adjusted to reflect lower risk because of the hiatus in Major Hurricanes. It would seem that they should be.
David,
Thanks for your comment re. my comments.
I would suggest that the core discussion is, as you write the comparison of observed actuarial data vs. computer models is the essence of the debate. My point is models are not at the point of assimilating all of the factors of weather events. IMHO, models are not yet capable to exceed the old actuarial process. I would invite you to research the Katrina relationship to MR-GO. This case relates the case of the Mississippi Gulf Outlet. This multi-year trial, struggled to asses MR-GO as a contributor to the destruction in New Orleans could not have been modeled. The Army Corps of Engineers completed the project in 1955. No one had a premonition MR-GO would some day create a high speed lane for a cyclone to devastate New Orleans.
My opinion is: we are not advanced enough to see models as better than the actuarial models that sucked. The computational models suck! The difference is we can understand the “suckiness” of actuarial models. Computer models require belief. Memo to self: belief may have risk.
For our citizens, let’s agree to argue for more accuracy. Our fellows can do better with more money in their pockets.
Regards,
Hi David and Robert,
This is an interesting discussion. I have worked with property cat models for 19 years now and while they do suck I think it’s fair to say they suck the least of any option. The old actuarial approach was alright but there were a lot of gaps in it. What the models attempt to do is combine the best parts of the actuarial approach with predictions of hurricane movement, wind speed, dissipation, etc. The predictive skill is still poor but it is an improvement. As George E. P. Box said “essentially, all models are wrong, but some are useful”. Without some way of predicting potential losses due to hurricanes, earthquakes, etc. these perils probably can’t be insured.
My first experience with using a catastrophe model to predict losses of a real time event was in 1999 with Hurricane Floyd. After running my company’s property book through “the model” I sent our chief actuary an estimate of a $30M loss. A couple of days later I had a chance talk to him and he said the “right” number for Floyd was $18M. How he came up with that I never found out. I believe the final figure ended up being somewhere north of $50M with the majority of those losses being due to riverine flood that no cat model can predict.
The models are already terribly political. If an insurance company wants to charge homeowners premium for wind coverage in Florida it must be based upon a model certified by the State of Florida. Do you want to bet there is a lot of pressure on the modeling vendors to keep modeled losses down in Florida?
Glenn
Here’s how the old actuarial model worked. You started with the dollar cost of prior catastrophes for quite a few years. These amounts were adjusted to current costs by a factor to reflect the fact that over time, the cost of a given same cat event would rise. This factor tried to measure the combined effect of inflation, greater amounts of property at risk, more structures in exposed areas, more insurance purchased, changes in handling of claims, etc. Then some long-tailed distribution was assumed and the these inflated values were fit to this distribution.
I think you can see how uncertain this technique was. The annual adjustment factor was somewhat of a guess. Yet, it had a large impact, because old losses were adjusted for many years.
The shape of the fitted function was merely a guess. But, it had a big effect on enormous losses. Since few enormous events occur, the choice of function was a big determinant of the supposed frequency of enormous losses. If you chose a distribution with a fatter tail, you predicted a lot more enormous events. Of course, these enormous events have very low frequency, but their high cost makes them important anyhow. The probability of enormous events plays a big role in reinsurance pricing. The reinsurance costs then get added to the primary insurance rates.
Another huge advantage of today’s models is that they’re more precise for estimating the catastrophe risk for a specific body of structures. In particular, the properties’ location and certain other characteristics can be taken into account. The old method didn’t allow that to be done.
I also came across this same information when looking up information on Ceres a couple of months ago.
Insurance companies are a potential source of funding for renewable energy projects. One large North American insurance company has already funded renewable energy projects in Canada.
If insurance rates are increased, then more funding would be available for renewable energy projects.
But the public only has a limited amount of money to purchase insurance coverage. Back to being only able to afford fire insurance coverage?
Hi David,
That sounds about right about for the old actuarial models (I am not an actuary so I don’t pretend to know the nuances). What the modeling vendors do every time there is an earthquake or hurricane is check their models against the new event to see if the model results are close to reality. They are rarely “close enough” so the vendors then proceed to back-fit their models with the new claims data. Twelve to twenty-four months later the new and improved models are released.
The issue I have with that is the models have been back-fit so many times they are probably over specified. That makes sense from a vendor standpoint because they want users to get accurate results for previous events (reinsurers and regulators still ask for modeled losses of historical events). However it is an almost a complete guarantee that the models won’t be predictive for the next cat event.
Glenn
All,
Well ain’t this some kind of coincidence:
http://www.bloomberg.com/news/articles/2015-10-27/insurers-unprepared-for-global-warming-un-climate-chief-says
At least four misleading statements in what the UN climate chief said.
1. As of today there’s been NO increased frequency and severity of major weather events (except for heat waves).
2. Future increases in major weather events are hypothetical and also wouldn’t occur for a long time.
3. It may be true that the agreement that the UN expects to seal in Paris in December won’t solve climate change because of the greenhouse gases already in the atmosphere. However, this statement isn’t valid if the catastrophic models are wrong. IMHO there are good reasons to doubt these models.
4. Omitted from the UN climate chief’s statement is the fact that the amount of CO2 reduction hoped for isn’t nearly sufficient to cause the atmospheric concentration of CO2 to stop rising rapidly.
David,
It seems from a layman’s vantage point that you underwriters should know about how to assign risk. That’s how you make a living. But I don’t think you are at all adequately informed on the basics of climate science, such as it is understood in 2015.
What is coming to the fore is that the CO2 sensitivity will turn out to be far smaller than what has been advertised heretofore by the IPCC. CO2 sensitivity is defined as the amount of temperature rise attributable to a doubling of atmospheric CO2 (from whatever causes that gas to increase). You say your risk models are laden with an accumulation of back-fitted parameters? But that is the same rubric used by climate modelers to achieve any semblance of accuracy in hind-casting past temperature history. To hind-cast a period of temperature in a time-series of interest they must insert arbitrary constants. The upshot is that the doctored models are rendered next to useless for the purpose of their future predictions. But that realization never deters them from their appointed rounds. In this political climate they get away with making outlandish claims of future warming that the insurance industry is ill-equipped to deal with. Who will still be around in the year 2100 to criticize them when they get it wrong? We may be into another ice age by then. The ice-core data from Greenland and Vostok (Antarctica) consistently show CO2 as a lagging variable. Translation: temperature change (of the oceans) controls atmospheric CO2, not vice versa. The climate change discussion suffers from an advanced case of myopia.
HL
It is no surprise at all that insurance companies would go whole-heartily into climate change alarmism, as it gives them every excuse for raising premiums, even if, in reality, it is not happening. Their free pass is that they are only being prudent and following the lead of the majority/”scientists.”