ACTUARIAL ISSUES

MAAC Fall Meeting Thursday, September 12, 2013 Stuart Mathewson, FCAS, MAAA, CPCU Sr. Property Actuary, Swiss Re Co-Chair, AAA Extreme Events Committee

     Hurricane Modeling Hurricane Insurance/Reinsurance Flood Insurance - NFIP Hurricane Sandy Questions and Answers 2

    What?

Computer simulation of catastrophic events Run against a portfolio of property risks Gives an estimate of how much loss will be incurred in the event Somewhat “black box”

   Why?

To estimate potential loss to a company To determine pricing of cat portion of account Much better than previous approaches ◦ Low frequency/High severity events ◦ Actual data is nearly useless ◦ Industry-wide use promotes market stability

  How Four modules ◦ Science 1 ◦  Simulates a cat event (e.g., hurricane) Science 2 ◦ ◦   Calculates the loss drivers (e.g., wind speed) at the various location affected Engineering  Estimates damage to structures from the event Financial Calculates insured loss from the damage Library of possible events, with probabilities

 Brief explanation of hurricane modeling

Key Parameters Forward Velocity (V F ) x Landfall Location (LF) Angle of Landfall (a) Eye Rmax Central Pressure Hurricane Hugo

Cat 5 Cat 4 Cat 3 Cat 2 Cat 1

Cat 5 Cat 4

Recorded Landfalls (1899-1996)

34 63 35 3 13

Northeast Mid-Atlantic

34

Southeast Texas Gulf Florida

Science Stochastic Event Module Science Hazard Module Engineering Vulnerability Module Financial Analysis Module

Define Hurricanes Assess Wind Field Calculate Damage Quantify Risk

 Determination of key hurricane variables ◦ Central pressure ◦ Radius of maximum winds ◦ Forward speed ◦ Landfall point and direction  Probabilities of each storm

Maximum windspeed experienced from Hurricane Hugo 1989 Peak Gusts in mph 150 + 150 - 140 140 - 130 130 - 120 120 - 110 110 - 100 100 - 90 90 - 80 80 - 70 70 - 60 60 - 50

Surface roughness and fetch affect wind speed estimation. Rougher terrain causes larger frictional effects on the windspeed .

Vulnerability Curves

Hurricane vulnerability curves represent the relationship between wind speed experienced and damage caused to the buildings 70 60 50 40 30 20 10 0 Wind speed

 Insurance or reinsurance losses can be calculated from different financial perspectives including: Ground-up, gross, net after reinsurance, catastrophe reinsurance cover  Attachment points for can be specified at location, policy, account or portfolio level  Model takes into account all affected properties in a given scenario

Average Homeowners’ Rates 1.6

1.4

1.2

1 0.8

0.6

0.4

0.2

0 TX LA MS AL FL GA SC NC VA WV MD DE PA NJ NY CT RI MA NH VT ME

      Model Input Location of building Construction Number of stories Year built Occupancy Secondary characteristics

   Model Output Annual Average Loss (AAL) Standard deviation of AAL (SD) PML levels ◦ E.g., 100-yr ◦ That is, how much loss do we expect with a probability of 1%?

Example of PML Curve Critical Prob.

0.01% 0.02% 0.10% 0.20% 0.40% 0.50% 1.00% 2.00% 4.00% 10.00% 20.00% 50.00% Pure Premium Standard Deviation Return Period 10,000 5,000 1,000 500

250

200

100

50 25 10 5 2 Loss 30,785,973 27,309,293 19,097,033 15,501,277 12,145,385 11,112,432 7,736,255 4,323,315 1,407,892 18,433 3 0 344,442 1,649,120

 

Pricing needs to include provision for the cost of potential loss Plus a load to pay for the risk

◦ For the capacity we have to allocate for the potential loss ◦ ◦ For the riskiness in having accumulations in hazardous areas For the volatility of frequency of loss

    Price based on ◦ Location ◦  Susceptibility of the area  Distance to the coast Construction Terms ◦ Higher deductibles – often percentage of value Market ◦ Near coast, often many companies exclude wind loss ◦ Wind pools are prevalent Flood (including storm surge) is almost always excluded ◦ Leads to issues of which peril caused the loss 22

◦ ◦ ◦ ◦     Property insurers determined long ago that flood was not insurable In 1968, Congress passed an act to establish NFIP By 1973, most communities were up and running Some key points Flood insurance available only in communities that established mandated controls Buildings built before establishment of flood maps were charged subsidized rates Program needs to be re-authorized periodically Later laws mandated insurance for properties with federally insured mortgages 23

◦          Some key points (cont.) Intent is that program be self sufficient In years that losses and expenses exceed premiums, NFIP can borrow from Treasury Until 2004, NFIP was able to pay back Treasury in non-catastrophe years Until 2005, borrowing limit was $1.5b

After Katrina, limit was raised to $21b After Sandy, limit was raised to $30b Current debt is $24b Current annual premiums about $3.5b

Full-risk ("actuarial") rates vs Subsidized rates 24

    Previous Flood authorization expired in 2008 AAA Extreme Events Committee started work on paper to address flood issue Congress took 4 years – and numerous short-term authorizations to pass new law

The National Flood Insurance Program: Past, Present...and Future? – 2011

◦ Purpose was to educate on the Flood program to aid in the public discourse ◦ Primary audiences   Actuaries Decision makers 25

   Capitol Hill briefing on monograph (July, 2011) Comment letters and written testimony to Congress when deliberations were ongoing (e.g., 6/28/12) Presentations to NCOIL and NAIC (2011,2012) 26

      Phases out subsidies for second homes, businesses and severe repetitive loss properties (25%/year) Any policy for property not currently covered must pay "actuarial" rates Raises cap on annual increases from 10 to 20% Sets up reserve fund Requires NFIP to set up a schedule to repay debt Requires several studies by GAO, Treasury and others 27

  Conversation with GAO on privatization   One of the studies mandated by Biggert-Waters Discussed general actuarial views on privatization Comment letter on HR1035  Community-based Flood Insurance 28

 Why was flood deemed uninsurable?

   Only those who would often get flooded were interested in buying it Premium for those properties alone would be prohibitive Small premium base wouldn't support catastrophic potential 29

 Do past issues still exist?

 Current situation      No one stepping in to profit from this niche Some companies offer excess coverage on high valued properties  Large commercial properties usually covered in all-risk policies Few buy NFIP insurance unless forced  And, enforcement not consistent when mandated Can new technologies help?

Wharton/CoreLogic Study  A Methodological Approach for Pricing Flood Insurance & Evaluating Loss Reduction Measures: Application to Texas 30

  Can it be properly rated? And would those rates be sustainable in the market?

   Expected Losses can be estimated by models Much more granular rating than NFIP NFIP has no capital requirements – therefore, no capital cost load – private companies would need a significant load  An estimate has been made that rates would have to be roughly doubled if written privately Can it develop a broad base?

 Without a mandate, it's hard to foresee increase in take-up rates 31

 Can these mechanisms be used to support the NFIP?

   At current rate levels, there is little premium to cover the cost of private reinsurers Could Federal government act as reinsurer instead of simply a lender?

One idea – Federal government pays for private market reinsurance    Would provide a stable expense for the government Private/Public partnership in vogue now for some Politically viable? 32

 Can these mechanisms be used to support a private market?

 This could be a key piece of privatization – but requires enough premium to pay for the reinsurance 33

 HR1035 – mandating a study on community-based flood insurance  What does this envision?

    Community is the insured Risk assessment and pricing for community as a whole  Covers all properties in the community Pros as per sponsors   Streamlined underwriting Increased participation  Incentive for community-based mitigation Issues   How do you decide how the taxpayers pay?

How do you get individual communities interested?

34

  At current rate levels, it would take decades to repay debt – even without further major occurrences  After Katrina, about $20B – paid down to $17B+ by 2011  After Sandy, up to $24B Biggert-Waters requires FEMA to create a repayment schedule  Must submit to Congress a report on options to eliminate debt in 10 years 35

 Reserve  Required by Biggert-Waters    1% of "total loss potential" in force Fund at 7.5% of reserve ratio until capitalized  If NFIP unable to make the minimum contribution, it must report this to Congress How is this to be paid for?

36

  Hydrological model  Basis for NFIP rates Probabilistic models     Even more difficult to model than earthquake Needs data at a very small granular detail for a large area of the country Changes to river basins often change significantly Not currently as useful for US insurers as for Europe 37

    Weird storm Hurricane or not?

Flood/Wind/Earth Movement Issues    Legal Legislative/Political NFIP 38

Cat 5 Cat 4 Cat 3 Cat 2 Cat 1

40

41

42

43

44

45

46

47

48

49

  As we have seen in the maps, it is very rare to have a storm curve into land in New Jersey NASA study    Typically, the tropical storms and hurricanes that strike the Northeast are pushed in a northeasterly direction by the prevailing upper level winds Most storms hit the coast at a grazing angle, and only areas of land that stick out into the ocean, such as Cape Hatteras, N.C., and Cape Cod, Mass., have a long history of experiencing storms that hit at a perilous perpendicular angle.

But with Sandy, the combination of a storm diving southeast from Canada, into the Midwest, and a large high pressure area in northeastern Canada and southern Greenland pinned the storm and forced it westward into the East coast 50

 NASA study    The study found that Sandy’s track stands alone in the historical record dating back to 1851, and that modeling simulations showed such a track is an event that would occur about once every 714 years "... Either Sandy was an exceedingly rare storm, or our assumption of long-term average climate conditions is erroneous, and Sandy’s track was made more likely by climate change in a way that is yet to be fully determined" The impact angle of Hurricane Sandy was its most unusual feature, ensuring the storm surge would case maximum damage, Hall said. The storm's left-hand turn put the most dangerous right-front quadrant on top of New Jersey and southeastern New York, pummeling these areas with an historic storm surge and record high waves. That, combined with astronomical high tides, led to record storm tide levels.

51

52

  Just before landfall, NOAA reclassified this as a post-tropical storm.

This meant that policies with hurricane or "named storm" deductibles (generally percent deductibles) had to revert to the base deductible in the policy 53

   One normal issue in a hurricane is the determination how much loss is wind and how much is water (storm surge or flood) In Sandy's case, much of the damage was cause by water since the wind was not quite hurricane strength and the storm surge was catastrophic Normal situations for losses on the coast   If there is flood insurance on the property, the wind carrier and the NFIP have to agree who pays what If not, the wind carrier has to determine how much of the loss is wind, while the insured is looking to maximize his payout 54

 One additional issue in Sandy was the assertion by insurers (following NFIP guidelines) that some of the damage not caused by water was caused by earth movement, which is typically excluded 55

    Legal Legislative/Political FEMA NFIP 56

  Non-insurance questions – for example   Tenants flooded out of their apartments wanted to know whether they had to continue paying rent. Homeowners needed to figure out who was responsible for removing toppled trees.  Workers at shuttered businesses weren’t sure if they were still entitled to paychecks.

Insurance Questions    Uncounted numbers of residents of New York, New Jersey and Connecticut are challenging flood insurance settlement offers that they believe didn’t properly cover damage.

Some homeowners hired their own consultants to get insurance company adjusters to revise damage estimates. If the insurer refuses to budge, claims can be appealed to the Federal Emergency Management Agency, which oversees the National Flood Insurance Program.

57

  Early on, governors in NJ, CT and NY proclaimed that this was not a hurricane, so hurricane deductibles would not apply.

Most legislative actions dealt with relaxing regulations about assessments, reserve funds, bonds, etc.

58

  Small grants to get people back in homes Any other aid is in form of low interest loans 59

   Claims Issues Congress raised borrowing limit to 30b, assuming a payout of about $13b Actual payout about $9b 60

61