THE EVOLUTION OF CATASTROPHE MODELS AND

Transcription

1 SERA Charleston, SC THE EVOLUTION OF CATASTROPHE MODELS AND THE REGULATORY IMPLICATIONS OFTHEIR USE April 2015 Richard Piazza Chief Actuary Louisiana Department of Insurance 1

2 Warning to Regulators! Don t believe everything you just heard! 2

3 How Were Catastrophic Events Incorporated Into Rates? A Brief Review Before Hurricane Andrew 1992 Industry observed many years of low to moderate catastrophic loss Ratemaking looked at all catastrophic events together 20+ years of historical data seemed predictable Judgmental tempering of low frequency events 6% statewide profit provision was sufficient 3

4 Then came Hurricane Andrew 4

5 How Are Catastrophic Events Incorporated Into Rates? A Brief Review After Hurricane Andrew Catastrophe vs. non-catastrophe ratemaking emerged Hurricane models introduced in ratemaking Promises made Promises broken Other models introduced in ratemaking Earthquake Terrorism Severe convective storm Flood Generalized Linear Models Credit scoring 5

6 What Makes a Good Filing? Statutory Compliance Regulatory Compliance Good Documentation Actuarial Support 6

7 Statutory Compliance Rates shall not be: Inadequate Unfairly discriminatory Excessive 7

8 Hurricane Models in Ratemaking State-of-the-Art Presumed to be fully credible Uses in ratemaking Overall rate level Territory relativities Deductible relativities Wind mitigation credits Reinsurance recoveries 8

9 Regulatory Compliance There are unique filing and support requirements in many states LA Bulletin (Catastrophe Model Interrogatories) FL Commission on Hurricane Loss Projection Methodology Know state filing procedures Provide all state-specific transmittals 9

10 LA Bulletin Catastrophe Model Interrogatories Part A submitted by insurer Certification Insurer information Insurer s view of model settings and input Printed output Part B submitted by model company Certification Model evaluation Model validation Model sensitivity 10

11 Model Documentation in a Filing Model/technical overview of model How did the model change? Description of model inputs Discuss underlying data sampling/partitioning Explain how the model handles incomplete or missing information Description of model output Explain how model output is used in the company s rating plan Describe how the model results were validated Provide statistical significance measures 11

12 Some Regulatory Concerns Metrics not available (unlike GLM) Input data validation/scrubbing Near-term versus long-term model view Model stability over time EQECAT 2008 LA revision (+175% AAL) RMS 2011 LA revision (+45% AAL) Model stability below statewide level, e.g., zip code 12

13 Some Regulatory Concerns Storm surge e.g., greater than estimated hurricane wind loss Side effects of model estimates and changes e.g., deductible or mitigation credits Allocation of statewide catastrophe expected loss to rating territory or by peril Undocumented adjustments to model input/output, e.g., trend, replacement cost or LAE 13

14 AAL per $1,000 amount of insurance year Changes to Modeled AAL Modeled AAL Source: Louisiana Homeowner Filings (per amount of insurance year) % -7% % % +45% EQECAT WORLDCAT RMS RISKLINK AIR CLASIC/ % /1/2008 1/1/2009 1/1/2010 1/1/2011 1/1/2012 1/1/

15 Hurricane Risk Transfer Cost (RTC) RTC is a Louisiana metric for assessing statewide reasonability of hurricane insurer s safety cushion RTC ratio: excess profit PLUS risk load PLUS net cost of reinsurance all relative to hurricane AAL excess profit is profit required over 6% of premium Observed RTC across scores of company filings over many years falls into a range of 0.5 to 1.5 Geographic and line of business diversification should be considered in interpreting the RTC 15

16 Territory Ratemaking Cat potential varies significantly within state Territory a priori non-catastrophe expected loss ratio (used as compliment to credibility) Do not use: Statewide non-catastrophe loss ratio Do use: Starting point of 100% MINUS Statewide expense ratio MINUS Territory-specific profit provision MINUS Territory-specific reinsurance provision MINUS Territory-specific risk provision MINUS Territory-specific modeled cat provision 16

17 Territory Credibility Complements and RTC (values relative to premium) (A) (B) (C) (D) (E) (F) (G) (H) (I) Territory Premium Weight Hurricane AAL Net Cost of Reinsurance Base Profit Risk Load Expense Ratio Expected Non-CAT Loss Ratio North Louisiana 33% 1% 0% 7% 0% 20% 72% 1.00 Central Louisiana 33% 15% 8% 7% 4% 20% 46% 0.87 South Louisiana 34% 28% 15% 7% 12% 20% 18% 1.00 Total State 100% 15% 8% 7% 5% 20% 45% 0.93 RTC Where: (H) = 100% - (C) - (D) - (E) - (F) - (G) RTC = ((E)- 6%) + (F) + (D) (C) 17

18 Actuarial Standards of Practice ASOP 12: Risk Classification ASOP 23: Data Quality ASOP 25: Credibility Procedures ASOP 29: Expense Provisions in P&C Ratemaking ASOP 38: Using Models Outside the Actuary s Area of Expertise ASOP 39: Treatment of Catastrophe Losses in P&C Insurance Ratemaking ASOP 41 Actuarial Communications 18

19 ASOP 38 Using Models Outside the Actuary s Area of Expertise Determine appropriate reliance on experts Have a basic understanding of the model (user input, model output) Evaluate whether the model is appropriate for the intended application Determine that appropriate validation has occurred (sensitivity testing Determine the appropriate use of the model 19

20 Regulatory Challenges Models based on many complex assumptions Regulatory resources Regulatory expertise Company knowledge/expertise Model transparency whether open or not Ask the right questions Know what can be adjusted in model or by company Track inputs and outputs Understand model sensitivities 20

21 21