Thursday, January 5, 2012

Model Me This...


The NFIP had to borrow from the US Treasury to cover losses during several flood events of 2005 and one Hurricane Ike in 2008.  This is not uncommon in the US, where supposedly, nearly everything disaster related comes from the Treasury.  In a recent report, the research team at the Wharton Center for Risk Management and Decision Processes argued that the reason for this is because the NFIP does not use probabilistic catastrophe models to establish premiums  but instead charges premiums that reflect the average annual year loss.

So the report investigates what would the premiums cost be if the method of calculating premiums were catastrophically modeled in two Texas counties.  Travis County is subject to riverine flood, Galveston County is subject to storm surge. 

They compare unloaded costs and show that "the current unloaded NFIP premiums are 'too high' in some areas and 'too low' in others relative to the probabilistic flood model results."  That is to say, they argue for a change in charges amongst risk groups, but averaging together risk groups in either county shows little difference from NFIP premiums.  For instance, in Travis County the study average is $2.42/ $1000 of property compared to the NFIP average of $2.6.  In Galveston County the study average is $4.7/$1000 of property compared to the NFIP average of $3.8.       

With this in mind it seems unlikely that a change of pricing technique would result in less chance of NFIP deficit considering roughly the same amount of money is collected annually.  What the model seems to most clearly do is change who pays what- which is, in effect, an ethical decision that should be well grounded in actual human political debate.      
The red circle and line is from the report. In reference to the idea that the NFIP is "too high" in some instances. 

I wonder, however, where the $1 increase comes from for Galveston County.  I am inclined to guess that it comes from one or both of two sources.  First, the modeled risk uses predictions of storm surge which are notoriously uncertain.  If, then, using the storm surge predictions increases the uncertainty of losses over that in the historical record, then it would necessitate increases in rates.  The other (obvious to me) possible source of the increase in rate is incorporating climate change assumptions into the model.  It is an interesting choice by the authors, considering the controversy surrounding the issue. The authors, themselves, admit that disaster loss increases are attributed to societal changes.

No comments:

Post a Comment