How Insurance Companies model the financial risks of wind damage

General Methodology

Accuracy predicting cost & frequency

THE INSURANCE INDUSTRY'S INCREDIBLE DISAPPEARING WEATHER CATASTROPHE RISK

How insurers have shifted risk and costs associated with weather catastrophes to consumers and taxpayers.

2/17/2012 Consumer Federation of America article by J. Robert Hunter
Read here

Posted 10/13/2015

THE RMS U.S. HURRICANE MODEL VERSION 11

Flip through this RMS powerpoint presentation to learn how they evaluate risk.

Posted 7/29/2015

PRICING TORNADOES: USING CAT MODELS FOR GRANULAR RISK UNDERWRITING

Kevin Van Leer, Product Manager – Americas Climate National Tornado Summit – Oklahoma City, OK February 24, 2015

Power Point presentation on how RMS proposes evaluating tornado risk addressing the question

* WHAT IS THE PROBABILITY OF A TORNADO CAUSING 30% DAMAGE TO A 2-STORY WOOD FRAME SINGLE FAMILY DWELLING IN OKLAHOMA CITY, OK?

According to AlDOI's Charles Angel, this RMS model said the 2011 tornado outbreak in central Alabama was a 1-10,000 year event.

New info aded 8/15/2015

Experts warn against complacency as U.S. endures hurricane drought

Read May 6, 2025, article by Artemis

Posted 6/8/2015

ARE THE INSURANCE COMPANY'S MODELS SCIENCE?

The models are not "Science." That suggests far too much authority. They are Hypotheses. The most precise definition of Science, in the way they are using the term, is the Scientific Method. What is Science? The Scientific Method.

There are seven-to-thirteen steps in the Scientific Method (depending on which book you read). One of the steps is "Form a Hypothesis." The step after that is "Craft and Conduct an Experiment" to test the hypothesis.

When setting up an experiment to test a hypothesis, the scientist uses the data of her hypothesis to predict what the experiment will produce, what its outcome will be. If the predicted outcome occurs, then the hypothesis is considered experimentally valid at least in one case. If the predicted outcome doesn't happen, then the hypothesis is considered false. Scientists in the latter case will go back to the drawing board and modify their hypothesis or throw it out.

If their experiment was testing a potential new drug, and it failed to produce the predicted results, the scientist wouldn't even think about putting it on the market. The FDA and whole world would laugh out loud at the thought. Some scientist and drug companies might bull ahead and insinuate the hypothetical but experimentally-failed drug on the public anyway, claiming, after all, that they've done "Science." If they went ahead, using the social mechanisms and legal loopholes available to them, they'd be thrown in jail and the public would be allowed to sue the pants off of them.

The crafting of an experiment to test the insurance companies' hypothetical hurricane models is relatively simple, conceptually. See if the models' predicted outcomes manifest during an actual hurricane and/or see if the costs predicted by the models accrue over passage of actual years. Do the models accurately predict a the cost of a forthcoming hurricane (not a past one)? Do they accurately predict the annualized cost of hurricanes in an area over a period of time?

To date, they've always been wrong in both. They've been wrong in their predictions of loss due to a forthcoming hurricane. And, as Karen Clark's study showed, they have also been very wrong in their predictions of costs over time. (Both the short and long-term models have been very wrong, according to her).

So, the models are not only not Science, they are just one step out of the many that are part of the scientific method: just hypotheses.

Hypotheses that have so far failed in experimental tests, to boot.

Revised 8/29/2012

DO ALABAMA & MISSISSIPPI HAVE A
UNIQUE WIND AND HAIL PROFILE?

Both states receive hurricanes from the Gulf of Mexico, and both have their Tornado Alleys.  In order for a hurricane to hit them, it is most-often on a track that runs on up through the rest of the state, so hurricanes also spread damage everywhere in the state.

The uniqueness is that Alabama and Mississippi wind-and-hail damages seems fairly even all over the state. That is, there is less difference in the amount of damage done by wind and hail over a period of time when comparing the coast with inland Alabama.

This seems different from Louisiana, Texas and Florida. These three have their hurricane susceptibility, but a less active Tornado Alley.  Average tracks of hurricanes tend to take the hurricanes on inland journeys to places other than the whole state.  So, there is a greater difference in damage due to wind and hail found on their coastlines, than found inland in their states.

Is this a valid conclusion?  Leave your opinion on the HHII Forum.

Posted 10/18/2011

Hurricane Forecast Computer Models

Click the above link for lots of information about hurricane risk modeling

Posted 10/5/2011

Presentations & Audio from Public Hearing Regarding Catastrophe Modeling and the Use of Short-Term vs. Longer-Term Horizons to NAIC Catastrophe Insurance (C) Working Group

Posted 9/7/2011

Model indicates coastal insurance rates should be lower,
but will they drop?

Read complete 9/4/2011 article by Jeff Amy, Press Register

Are most residents of the Alabama and Mississippi coasts due big price cuts for homeowners insurance — maybe as much as 34 percent in Baldwin County, for example?

That’s one of the findings suggested in an analysis presented by modeling firm Risk Management Solutions to Alabama Insurance Commissioner Jim Ridling and Mississippi Insurance Commissioner Mike Chaney.

California-based RMS sent employees to Montgomery in April to present projections for the states to the two commissioners.

The firm had previously declined to give specific numbers to the Press-Register. The Homeowners Hurricane Insurance Initiative, an Alabama citizens’ group lobbying for lower rates, obtained a copy of the presentation and provided it to the newspaper.

The presentation shows that when the new RMS model is compared to the calculations used since 2003, the anticipated average annual loss declines steeply from Gulfport eastward into the Florida Panhandle. In both Baldwin and Mississippi’s Jackson counties, expected hurricane losses are projected to be more than 40 percent lower under RMS’ new model.

Alabama’s annual losses fall by 34 percent under the new model, and Charles Angell, the staff actuary for the Alabama Department of Insurance, said he believed that "there’s a real good chance" rates could go down in coastal areas because of the model change.

Already, the Alabama Insurance Underwriting Association, which sells policies to people in Mobile and Baldwin counties who have trouble getting coverage from traditional insurers, lowered average rates by 6.2 percent this summer.

Members of the Homeowners Hurricane Insurance Initiative said that the RMS analysis supports their demands for lower coastal premiums. "It tends to validate the belief that the risk and costs of repair to Alabama’s coastal counties has been overstated," said Dan Hanson, one of HHII’s leaders. "Premiums ought to come down some." Beyond that, Hanson and his group insist that the insurance industry as a whole is too reliant on computer modeling.

HHII says the swing in modeled losses is an argument in favor of its demand that insurers make public years’ worth of premium and loss data by ZIP code.  A proposed law, which the group calls the clarity bill, failed in the Legislature last year, but is likely to be considered again by Gov. Robert Bentley’s Affordable Homeowners Insurance Commission. "The need for the historical data that the clarity bill requires is adamantly demanded as a way to check the results produced by these models," Hanson said.

Though HHII would like to force insurers to use only historical data to calculate rates, a return to that older method appears unlikely. Advocates for modeling say far too few hurricanes have hit any particular location in recorded history to produce reliable loss projections.

Posted 9/4/2011

RMS Presentation to AL DOI

This is the same risk modeling program rigorously evaluated by Florida Catastrophic Storm Risk Management Center and used by the State of Florida.  RMS preliminary modeling indicates wind damage risk in Baldwin County is overestimated by as much as 50%.

Posted 6/15/2011

Insurers Cautious About Adopting New RMS Hurricane Model

Property/casualty insurers appear to be in no rush to adopt a controversial revised catastrophe model that dramatically raises certain estimates of potential hurricane losses in the country.

California-based modeling firm RMS released its revised hurricane model, version 11, in February. Some in the industry have predicted that it will lead reinsurers and insurers to revise rates and underwriting. It has been predicted that the new model, along with recent catastrophes around the globe, could trigger an end to the prolonged soft market.

Others have suggested that the RMS release has shaken insurer confidence in the use of catastrophe models.

But any real effects from RMS 11 could be some time in coming. According to RMS, carriers and brokers, the adoption of RMS 11 is going slowly. Also, just because the model suggests higher rates does not mean the market will follow, especially given plentiful capacity.

Read complete 5/10/2011 Insurance Journal Report
Read more about risk modeling

Posted 5/13/2011

New Hurricane Model is Broader

A hurricane-modeling company that helps insurers predict the cost of megastorms will launch a new , more sophisticated model on Monday that shows some homeowners living hundreds of miles from the neares ocean are at greater risk than previously thought.

Read Erik Holm's complete article in 2/28/2011 WSJ
(WSJ digital subscribers can click here to read and comment)

Posted 3/1/2011

Short-term hurricane modeling driving up coastal insurance costs, say critics

Big hurricanes hit the United States mainland relatively infrequently. But when they do, every policyholder in a given area might suffer damage on one day. That’s different from fires, thefts, or car wrecks, where insurers see a steady stream of relatively predictable, smaller losses.

Modeling has helped insurers define the financial risks of hurricane damage. But some in the industry say that the relatively new practice of short-term modeling leads to excessive damage predictions and is likely driving up coastal insurance costs.

Read complete 2/6/2011 article by PR reporter Jeff Amy
See related post below

Posted 2/6/2011

Coastal homeowners may have less hurricane risk under new computer model

A new computer model for insurers predicts that inland homeowners face higher risks than previously believed, while coastal residents could actually face lower risks. The model by Risk Management Solutions will launch at month’s end amid continuing debate about the effect that modeling has on policy costs.

Many factors are considered when a company quotes a premium on a homeowners policy. But along the Gulf Coast, models showing large potential hurricane losses have a big influence on insurance costs, and even whether carriers are willing to write wind coverage.

Read complete 2/6/2011 article by MPR reporter Jeff Amy
See related post below

Posted 2/6/2011

Inland Homeowners in Path of Higher
Hurricane-Insurance Costs

By Wall Street Journal's Erik Holm

A disaster-modeling company that helps insurers predict the cost of storms and earthquakes is rolling out a new hurricane model that will likely boost the price of insurance for some homeowners living a hundred miles, or even 200 miles, from the nearest ocean.

Property coverage is typically more expensive for the homes along the East Coast and the Gulf of Mexico that have a front-row seat when a hurricane blows ashore. However, the adjustments being made by the modeling company, Risk Management Solutions Inc., could actually give some of those coastal homeowners a slight break, while increasing the estimates for how much damage a hurricane can do farther inland.

Read complete article

Posted 1/14/2011

Climate Change and Hurricane Loss

The 2015 hurricane season forecasts are for below average activity, but it only takes one significant storm to impact a highly populated area to create large insured losses. This succinct paper by Karen Clark and John Lummis discusses how investors can think about hurricane losses with respect to short term variations in hurricane activity and long term trends driven by climate change.

Posted 8/2/2015

The Acts of God Algorithm

Why the insurance industry is headed for the perfect storm
By Art Jahnke

Read this fascinating article about Karen Clark and how she came to be regarded as the hurricane risk model guru.

Posted 6/10/2015

WHEN IS A HURRICANE RISK MODEL ACCURATE?

This video was taken at one of the AHIC educational presentations last year.  In it you will see Karen Clark, the inventor of cat models, going eye to eye with Lori Medders, a representative of the Florida hurricane loss projection methodology commission.

Basically, Medders is saying that the models have a type of accuracy that can be described as good process, but between 2006 and 2010 the models predicted $65 Billion would be need to protect coastal America and actually only $15 Billion was needed (Karen Clark). Hear that big sucking noise??? We can beat this! The crack is already forming!

Read more about risk models here

Posted 10/23/2012

HISTORICAL DATA ON HURRICANE STRIKES

Click on these links to see actual hurricane strike data, as compared to hypothetical model preditions

U.S. Mainland Hurricane Strikes by State, 1900-1996

Annual Severe Weather Report Summary (National) - 2011

Annual Severe Weather Report Summary - 2011 (AL)

Annual Severe Weather Report Summary - 2011 (LA)

Annual Severe Weather Report Summary - 2011 (MS)

Posted 10/9/2012

HURRICANE COMPUTER MODEL ACCURACY

The Hurricane "Catastrophe Models" used to jack Alabama coastal insurance premiums sky high are inaccurate. This 7-minute video was taken at Alabama Governor Bentley's Affordable Homeowners Insurance Commission (AHIC) meeting in Montgomery December 12, 2011.

Watch video

Posted 1/30/2012

KETK interviews Dr. Bill Gray about hurricanes and global warming

Read KETK's Casey Claiborne's 12/13/2011 article & watch video of interview

Colorado State University's hurricane prediction team made up of Dr. William Gray and Phil Klotzbach recently announced they won't be predicting the number of hurricanes we'll get as early as December -- like they have for many years. They say it's just too hard to predict something like that so early.

KETK's chief meteorologist Scott Chesner agrees. "Predicting the weather in the long range is an impossible feat the farther out in time you go, its just another reason why especially in terms of trying to predict man's influence on the climate -- totally preposterous!" Chesner said.

Dr. Gray himself has always been very vocal against the notion of man-made global warming. But he says that's not exactly what their change in forecast style is about. "It's just, we haven't shown in real practice, skill with it. So we said 'gee if we've tried it 20 years without skill, we'd be lying to the public to put out specific numbers.'"

Dr. Gray tells KETK, contrary to reports circulated by a Canadian newspaper, they will begin making qualitative predictions in December and will still do their other 3 forecasts later in the year that predict the number of storms. "They misinterpreted it, they were kind of said we're giving up the...throwing in the towel, that isn't right," Gray said.

On a side note, Dr. Gray says that no data points to global warming having any effect on hurricane activity. The weather expert says he doesn't believe global warming is really coming -- and in the last decade or so he says data shows the earth has actually cooled. "I'm making a prediction that 100 years from now when the history books are written, Al Gore's gonna wind up halfway between Aaron Burr and Benedict Arnold," Gray said.

Dr. Gray and Dr. Klotzbach's new December report that looks at how severe the season will be is available for viewing.

Click here to visit their website

Posted 12/15/2011

Can Science Halt Hurricanes?

By David Biello
Tropical cyclones are nature's most powerful storms. Can they be stopped? 
© 2011 Scientific American, a division of Nature America, Inc.
All Rights Reserved. Reproduction in whole or in part without permission is prohibited
.


Scientific American carries a fascinating and important article that raises some very interesting points, especially on the fallibilities of hurricane modeling:

The article starts with declaring 2011 an “active” hurricane year, a term often used by the insurance industry. Yet, as usual, it was a yawner for us on the Gulf Coast, perhaps more exciting for mid-Atlantic sailors. Why so?

See what scientists believe happened to defang Hurricane Irene.  

Get some insight into why winters and spring floods have been so cruel to the Yankee states over the past few years (isn’t that where most insurance companies have their headquarters?). 

Learn about scientific prognostications on the declining power of hurricanes. The answer may come as a surprise. 

Give some thought to the hubris being expressed by some who wish to use artificial means to tone down hurricanes—possibly not such a good idea, given the beneficial effects that hurricanes have on weather overall.

It’s hard to read an article like this from such an impartial and respected source (having no concern about insurance) without seeing a dark cloud of doubt over the validity and utility of long-range forecasting models, or even short-range, for that matter.

Posted 11/22/2011 (Hat tip to Stan Virden)

A.M. Best Group Vice President Cautions
Against Overreliance on Models

Catastrophe-modeling software should be used as a starting point, not the final arbiter, of assessing catastrophic risk, and prior loss history, experience, and the unique characteristics of a company's book of business are equally important in gauging a company's risk exposure, said Anthony Diodato, group vice president of property/casualty ratings at A.M. Best.

"If companies are dependent or over-dependent on a model, and when there are changes to the version of the model from year to year, you're trying to manage your exposures based on these changes that may not be reflective of your actual exposures," Diodato said.

Earlier this year, Risk Management Solutions Inc. unveiled a new version of its hurricane model, known as version 11, which has caused a stir in the insurance, reinsurance and brokerage industries. The update incorporates data from a three-year research project from the University of Miami as well as analysis of tens of thousands of wind-speed observations. Generally speaking with RMS version 11, wind risk increases in inland areas, as the revision considers that winds will dissipate more slowly after a hurricane makes landfall. For instance, the updated model lowers the view of risk of some coastal areas of Florida and increases the view of risk to central Florida. There is more vulnerability in Texas, especially for commercial structures. As a result, the model changes increase the probable maximum loss for some carriers.

To see the entire interview with Diodato, visit

http://www.ambest.com/video.BN-NJ-10-14-2011

Posted 10/18/2011

Karen Clark's Presentation on Model Limitations, March 28, 2011 at National Association of Insurance Commissioners
Public Hearing on Catastrophe Models

Listen to podcast of Presentation

Updated 10/302011

Insurance Companies Depend Too Much on Catastrophe Models

Catastrophe models are a great risk management tool for property/casualty insurers but even the person who created the first one is worried that they are being given more credit and influence than they deserve.

Karen Clark is an expert in the field of catastrophe risk assessment. She developed the first hurricane catastrophe model in 1983 and in 1987 Clark founded the first catastrophe modeling company, known today as AIR Worldwide Corp., which she sold to Insurance Services Office (ISO) in 2002.

The need for insurers to understand catastrophe losses cannot be overestimated. Clark’s own research indicates that nearly 30 percent of every homeowner’s insurance premium dollar is going to fund catastrophes of all types.

“[T]he catastrophe losses don’t show any sign of slowing down or lessening in any way in the near future,” says Clark, who today heads her own consulting firm, Karen Clark & Co., in Boston.

While catastrophe losses themselves continue to grow, the catastrophe models have essentially stopped growing. While some of today’s modelers claim they have new scientific knowledge, Clark says that in many cases the changes are actually due to “scientific unknowledge“— which she defines as “the things that scientists don’t know.” “

Companies should not be lulled into a false sense of security by all the scientific jargon which sounds so impressive because in reality… the science underlying the models is highly uncertain and it consists of a lot of research and theories, but very few facts,” says Clark.

Read complete Insurance Journal interview report

Posted 4/16/2011

RMS CALLS CRITICISM “A FUNDAMENTAL MISUNDERSTANDING” OF CAT MODELS

Risk Management Solutions (RMS) says that a recent report (see post below) saying that short term hurricane models overestimated losses by billions as misplaced.
 
In an emailed statement, RMS argues that a recent report by Karen Clark & Co. mistakenly characterizes catastrophe models as offering “deterministic” rather than “probabilistic” results, and demonstrates a fundamental misunderstanding of the purpose of Cat models.

Karen Clark & Co. issued its second annual study of near term hurricane models where it argued that AIR Worldwide, EQECAT and RMS projected near term insured loss levels at least 35 percent above the long-term average of $40 billion for the period 2006 through 2010. Actual losses for the period turned out to be $13.3 billion, according to Clark.

Given the technical arguments made by RMS, the full text of the statement is below:

“RMS welcomes review and debate of the timeframe over which catastrophe models should be used to characterize hurricane risk. However, this report demonstrates a fundamental misunderstanding of the purpose of Cat models by asking whether they can be used to predict actual catastrophe experience in a particular one, two, or five-year period. It’s important to understand that catastrophe models deliver probabilistic forecasts not deterministic predictions. A probabilistic activity forecast means that, on average, over many five-year periods, this is the number of hurricanes to be expected. The actual number experienced in a particular five-year period will be just one sample from a broad distribution of possible outcomes.

Hurricane activity responds to warm sea surface temperatures (SSTs), which have been above the long-term historical average since the early 1990s. While the debate about the cause of increased SSTs continues, there is scientific consensus that the proportion of intense hurricanes (Category 3-5) has increased and that overall hurricane frequency has been significantly higher than the long-term historical average since 1995. The annual average of category 1-5 hurricanes making first landfalls along the U.S. coastline from 1970 to 1994 is close to 1.3, while since 1995 it is around 2.0 per year. Therefore, most hurricane experts believe that simply using the long-term historical average number of hurricanes is not the best way to estimate future activity.”

Posted 1/24/2011

Near-Term Models Misjudge Hurricane Losses By Billions?

Catastrophe modelers, through their near-term models, have overshot actual insured losses for 2006 through 2010 by as much as $53 billion.

According to the report issued by Karen Clark & Company, modelers AIR Worldwide (AIR), EQECAT and Risk Management Solutions (RMS) each dramatically overestimated insured losses above the long-term average, even with some minor revisions to their models.  Had they been correct, insured losses for the period 2006 through 2010 would have been between $60.4 billion (predicted by AIR) and $68.2 billion (EQECAT’s estimate). RMS’s projection for the period was $67.2 billion.  Actual insured losses for 2006 through 2010 was $15.2 billion.

Read complete article by Chad Hemenway, PropertyCasualty360.com

Posted 1/20/2011

Page last revised 10/13/2015