Thursday, January 19, 2012

Modeling Certain Mayhem

In America's deep south, a region not so far away, hides a new foe threatening otherwise intelligent people's ability to decide. The Louisiana Insurance Commissioner, Jim Donelone, has rung the alarm putting homeowners on alert of  "The looming threat of the new cat model, RMS 11".  This is the newest addition in the catastrophe model rogue gallery challenging the gallant efforts of state insurance regulating offices.  The kryptonite in their coding is the incredible capacity to produce scientifically supported uncertainty thereby weakening the ability to control rates by politically hopeful insurance commissioners everywhere.   A past episode between dueling regulating powers and risk predicting machinery demonstrated the societal cost inflicted by these dastardly foes creating uncertainty whenever plugged into a wall.  In 2006, RMS rolled out an arbitrary change to their trusty hurricane catastrophe model in RiskLink 6.0, costing Florida homeowners $82 billion.  Stay tuned to state regulating offices for the latest updates on the battle between man and machine...

In the mean time, let's take a closer look at these new trade secret rascals (Submitted during 2010/2011 to meet 2009 standards).  

New models vary in characterizing the hurricane risk.  

Vertical axis: Probability of landfall; Horizontal axis: Saffir- Simpson Category 1-5
From this view, RMS seems not so different from the other models when considering landfall probabilities for the entire state of Florida.  While model estimates within each category "look" somewhat similar, because they are dealing in percentages they are in fact, substantially different.  

For example, consider a Cat 5 storm, the most damaging.  AIR and EQECAT guess 1% and RMS, ARA, and the state's own Public Model (also known as FIU), guess 2%.  While the difference is only 1% it is also the difference between experiencing a Category 5 landfall in the state of Florida once in 100 years or once in 50 years.  A similar situation exists for a Category 4 storm.  ARA estimates an annual probability of 4% or once in 25 years and RMS and EQECAT estimate 8% or once in about 13 years.  All of these estimates are grounded in "sound science."    

What sort of punch is packed into these probabilities?

There are several time frames by which to measure a probable maximum loss (PML) or how much do you probably stand to lose.  A common time frame is 100 years.  Below is a table showing each new model's estimated 100 year PML. 

According to the suite of models, the 100PML is anywhere between $18 billion and $146 billion.  Now, let's say you are charged with choosing one model on which to base rates (such a scenario is highly unlikely), which one would you choose and how would you make that decision? *Remember, each one is as good as the next.

The decision depends on something other than "facts," it depends on who you are, what you fear, and your goals.  Do you fear insolvency?  Stock holder dissatisfaction?  The cost of living month to month?  Losing an election?  Are you seeking to establish a new national natural disaster insurance regime?  Expand affordable housing mortgage lending?  Enliven fears of climate change? 

Together, these models create a great deal of uncertainty about the risk being insured against.  In the world of insurance, uncertainty about the risk is risk in and of itself.  If uncertainty increases, then the cost will too and vice versa.  So, a reasonable question to ask would be, "Has the modeled risk changed?"

What follows is a comparison of the most recent model reports, submitted over 2010/2011 to meet 2009 standards, and model reports meeting 2008 standards submitted during 2009.  All dollar estimates in reports meeting 2008 Standards have been adjusted for inflation to 2010$. 

When considering landfall probabilities, ARA and RMS have increased the estimate of a Cat 5 landfall by 100% in their most recent submissions.  There are other changes evident here too.     

What do such changes mean for estimated 100 year MPL.  Below is a graph showing estimated loss  changes from 2008 Standards to 2009 Standards as a percentage of 2008's estimate.  We see that overall, changes in estimated loss is fairly small EXCEPT for the Public Model.  They have increased their estimated losses by nearly 40%.  

(Whoa!  Look out Citizen's policyholders, your rates are directly tied to this model.)   

Look at the table below.  Recent estimated losses are fairly consistent with 2008 standards estimates which guessed a range of $20- $140 billion.   

However, if we consider changes in overall model uncertainty from 2008 standards to 2009 standards we see a notable increase of 34% in uncertainty born predominantly by the state's public model.

% change of uncertainty interval from 2008 Standards to 2009 Standards
RMS: 16.23%
AIR: -4.29%
ARA: 2.51%
EQECAT: -13.38%
Public Model: 33.07%

Presented here is only half of the uncertainty machine team as  reinsurers have a whole other set of models up their sleeves.   

If new models guess the same losses but increase uncertainty about those losses is this progress?  I don't think so.  For those familiar with a hurricane forecast, imagine that the skinny black line remained on a a constant path but the white "cone of uncertainty" around it got bigger and bigger.  

(Louisiana does not have its own state funded model.)

No comments:

Post a Comment