materials - catastrophe risk (e) subgroup...academy of actuaries (academy) on wildfire: lessons...

© 2020 National Association of Insurance Commissioners

Date: 2/18/20

2020 Spring National Meeting Phoenix, Arizona

CATASTROPHE RISK (E) SUBGROUP

Friday, March 20, 2020 10:30a.m. – 12:00 p.m.

Sheraton – Valley of the Sun AB – Level 2 ROLL CALL

Tom Botsko, Chair Ohio Robert Ridenour, Vice Chair Florida Susan Bernard California Mitchell Bronson/Eric Unger Colorado Wanchin Chou Connecticut Judy Mottar Illinois Gordon Hay Nebraska Anna Krylova New Mexico Gloria Huberman/Sak-man Luk New York Andrew Schallhorn Oklahoma Will Davis South Carolina Miriam Fisk Texas NAIC Support Staff: Eva Yeung/Jane Barr

AGENDA

1. Consider Adoption of the Joint Property and Casualty Risk-Based Capital (E) Working Group and Attachment A

Catastrophe Risk (E) Subgroup Minutes—Tom Botsko (OH)

2. Discuss the Questions Raised at the 2019 Fall National Meeting Presentation from the American Attachment B Academy of Actuaries (Academy) on Wildfire: Lessons Learned From the 2017–2018 Event —Rich Gibson (Academy)

3. Hear a Presentation from Karen Clark & Company (KCC) on its Catastrophe Model Attachment C

—Glen Daraskevich (KCC) and Joanne Yammine (KCC)

4. Discuss the Possibility of Allowing Additional Third-Party Commercial Vendors—Tom Botsko (OH)

5. Discuss Any Other Matters Brought Before the Subgroup—Tom Botsko (OH)

6. Adjournment W:\National Meetings\2020\Spring\Agenda\tf\capadequacy\pcrbc\032020 cat risk agenda.docx

1

This page intentionally left blank.

2

Attachment A Capital Adequacy (E) Task Force

3/22/20

© 2020 National Association of Insurance Commissioners 1

Draft: 2/4/20

Property and Casualty Risk-Based Capital (E) Working Group and Catastrophe Risk (E) Subgroup Conference Call February 3, 2020

The Property and Casualty Risk-Based Capital (E) Working Group of the Capital Adequacy (E) Task Force met via conference call Feb. 3, 2020, in joint session with the Catastrophe Risk (E) Subgroup of the Property and Casualty Risk-Based Capital (E) Working Group of the Capital Adequacy (E) Task Force. The following Working Group members participated: Tom Botsko, Chair, and Dale Bruggeman (OH); Richard Ford (AL); Mitchell Bronson, Rolf Kaumann and Eric Unger (CO); Wanchin Chou (CT); Robert Ridenour (FL); Judy Mottar (IL); Anna Krylova (NM); Sak-man Luk (NY); Will Davis (SC); and Randy Milquet (WI). The following Subgroup members participated: Tom Botsko, Chair, and Dale Bruggeman (OH); Robert Ridenour, Vice Chair (FL); Kim Hudson and Laura Clements (CA); Mitchell Bronson, Rolf Kaumann and Eric Unger (CO); Wanchin Chou (CT); Judy Mottar (IL); Anna Krylova (NM); Sak-man Luk (NY); Andrew Schallhorn (OK); and Will Davis (SC). Also participating were: Julie Lederer (MO); and Steve Drutz (WA). 1. Adopted the Catastrophe Risk (E) Subgroup’s 2019 Fall National Meeting Minutes Mr. Botsko said the Subgroup met Dec. 6, 2019, and took the following action: 1) adopted its Nov. 8, 2019, minutes; 2) adopted proposal 2019-14-CR (2019 U.S. and Non-U.S. Catastrophe Event Lists); 3) heard presentations from the American Academy of Actuaries (Academy) on Wildfires and the Actuaries Climate Index (ACI); 4) discussed the factor of using aggregate exceedance probability (AEP) basis vs. occurrence exceedance probability (OEP) basis; and 5) discussed modeling of projected losses. Ms. Mottar made a motion, seconded by Mr. Chou, to adopt its Dec. 6, 2019, minutes (see NAIC Proceedings – Fall 2019, Capital Adequacy (E) Task Force, Attachment Four-A). The motion passed unanimously. 2. Adopted the Property and Casualty Risk-Based Capital (E) Working Group’s 2019 Fall National Meeting Minutes Mr. Botsko said the Working Group met Dec. 8, 2019, and took the following action: 1) adopted its Nov. 8, 2019, minutes; 2) adopted the report of the Catastrophe Risk (E) Subgroup; 3) exposed proposal 2018-19-P (Vulnerable 6 or Unrated Risk Charge); 4) discussed the 2020 property/casualty (P/C) risk-based capital (RBC) working agenda; 5) discussed the possibility of using the NAIC as a centralized location for reinsurer designations; and 6) discussed the possible treatment of the R3 related to the runoff and captive companies. Mr. Milquet made a motion, seconded by Ms. Mottar, to adopt its Dec. 6, 2019, minutes (see NAIC Proceedings – Fall 2019, Capital Adequacy (E) Task Force, Attachment Four). The motion passed unanimously. 3. Adopted the Property and Casualty Risk-Based Capital (E) Working Group and Catastrophe Risk (E) Subgroup’s E-Vote

Minutes Mr. Botsko said the Working Group and the Subgroup conducted an e-vote to consider adoption of proposal 2019-14-CR (2019 U.S. and Non-U.S. Catastrophe Risk Event Lists). Mr. Chou made a motion, seconded by Ms. Mottar, to adopt their Jan. 22 minutes (Attachment SixXX). The motion passed unanimously. 4. Adopted Proposal 2018-19-P (Vulnerable 6 or Unrated Risk Charge) and Agreed to Refer the Schedule F Proposal to the

Blanks (E) Working Group Mr. Botsko said the purpose of this proposal is to modify the instructions to reflect that the factors for all uncollateralized reinsurance recoverable from unrated reinsurers be the same for authorized, unauthorized, certified and reciprocal reinsurance. W. Scott Williamson (Reinsurance Association of America—RAA) said the RAA supports the proposal and the associated Annual Statement changes. He agreed with the Working Group approach to consider moving, over time, towards a charge that is more aligned with risk-indicated factors used by the rating agencies. He also recommended that the Working Group should consider applying different charges for: 1) captives and runoff reinsurers, as they may not obtain financial strength ratings; 2)

3


3/22/20


reinsurer designation equivalent to categories 1 through 6 to reflect the most recent credit default experience and consistency with reinsurance recoverable credit risk factors in use by the rating agency capital models; and 3) lowering the cushion or margin for operational risk that is embedded in the credit risk factors. Matthew B. Vece (American Property Casualty Insurance Association—APCIA) said he is concerned that this proposal inappropriately combines two groups with inherently different risk characteristics. He recommended an alternative approach to retain the current seven categories for the RBC R3 credit risk charge, with the last two categories being: 1) vulnerable 6; and 2) unrated (whether authorized, unauthorized, certified or reciprocal). Mr. Williamson said that the unrated category includes vulnerable reinsurers in addition to solvent reinsurers. He agreed with the current proposal that eliminates the NAIC-7 designation code. If in the future, the Working Group defines a category for solvent runoff or other situations eligible for a capped factor, the NAIC-7 code could be re-activated at that time. Mr. Milquet asked if reclassifying NAIC-7 back to NAIC-6 creates more work for filing companies. Mr. Williamson replied that it is necessary to take this action to ensure that future RBC filings are not populated with “legacy” NAIC 7 codes if and when a new definition is adopted. Mr. Botsko understood the industry concerns. However, it was the Working Group’s intention to evaluate the data annually until reaching any agreed upon change to the factor and the structure. Mr. Botsko recommended that the Working Group consider: 1) adopting proposal 2018-19-P for 2020 RBC filing; 2) forwarding the blanks proposal to the Blanks (E) Working Group for consideration; and 3) documenting the industry concerns in the working agenda for future discussion. Mr. Milquet made a motion, seconded by Mr. Ridenour, to refer the Schedule F proposal to the Blanks (E) Working Group and adopt proposal 2018-19-P, subject to adoption of the Schedule F blanks proposal from the Blanks (E) Working Group. The motion passed unanimously. 5. Received Referrals from the Statutory Accounting Principles (E) Working Group Mr. Botsko said the Working Group received two referrals from the Statutory Accounting Principles (E) Working Group. The first referral is regarding Ref #2019-49: Retroactive Reinsurance Exception (Attachment XX). He said this agenda item addresses a request from the Academy Committee on Property and Liability Financial Reporting (COPLFR) Working Group to clarify both the accounting and reporting for retroactive contracts, which are accounted for prospectively. The COPLFR noted that there is diversity in the current practice due to lack of specific guidance. The clarifications requested include: 1) both the ceding entity and assuming entity, where both are members of the same group and are consolidated in the same combined annual statement; and 2) the reporting method to be used if the ceding entity and assuming entity are not in the same group. Robin Marcotte (NAIC) said the Statutory Accounting Principles (E) Working Group is currently seeking: 1) input related to the RBC impacts; and 2) volunteers to assist with developing guidance. She encouraged volunteers to contact her. Mr. Botsko said another referral is regarding Ref #2019-40: Reporting of Installment Fees and Expenses (Attachment XX). Ms. Marcotte said the purpose of this exposure is to include a minor clarification of the current installment fee guidance in Statement of Statutory Accounting Principles (SSAP) No. 53—Property Casualty Contracts–Premiums and request input from the Working Group on the questions that are included in the referral regarding if incurred installment fee expenses should be allowed to be reported in other expenses. Excluding expenses from underwriting can have an impact on underwriting ratios. Mr. Bruggeman encouraged members and interested parties to review this referral and share their thoughts during the Spring National Meeting. Mr. Botsko said the Working Group will coordinate with the Casualty Actuarial and Statistical (C) Task Force to determine how to approach these referrals and provide findings during the Spring National Meeting. 6. Discussed Other Matters Mr. Botsko said the Working Group just received another referral (Attachment XX) from the Restructuring Mechanisms (E) Subgroup on Jan. 29. He stated that the purpose of the referral is requesting that the Working Group determine if changes should be made to the P/C formula to better assess companies in runoff. He encouraged interested parties to review the referral; the Working Group will have a more in-depth discussion at the Spring National Meeting. Mr. Bruggeman said the survey included in the referral provided some example definitions of the runoff companies. He encouraged members and interested parties to review the survey and share their thoughts at the Spring National Meeting. Mr. Botsko also announced that the time of the Catastrophe Risk (E) Subgroup Spring National Meeting will change to 10:30 a.m. – 12:00 p.m. on March 20, 2020.

4


3/22/20


Having no further business, the Property and Casualty Risk-Based Capital (E) Working Group and the Catastrophe Risk (E) Subgroup adjourned. W:\National Meetings\2020\Spring\TF\CapAdequacy\PCRBC\Att01 02-03propertyrbcwg-catrisksgmin .doc

5

This page intentionally left blank.

6

1850 M Street NW Suite 300 Washington, DC 20036 Telephone 202 223 8196 Facsimile 202 872 1948 www.actuary.org

January 7, 2020

Tom Botsko, Chair Catastrophe Risk Subgroup P/C RBC Working Group National Association of Insurance Commissioners

Via email

Dear Tom:

Thank you again for inviting the American Academy of Actuaries1 to present a summary of our issue paper Wildfire: Lessons Learned from the 2017–2018 California Events at the Catastrophe Risk Subgroup’s meeting in early December. At that session there were several questions raised, which I said I would explore and get back to you and the subgroup with answers. On behalf of the Academy’s Extreme Events and Property Lines Committee, I am happy to provide the following supplemental information.

In our presentation, it was noted that between 1990 and 2010, the population in Wildland Urban Interface (WUI) zones increased 35 percent and the number of houses grew by 41 percent. You asked for a comparison with population growth in general in the United States during that same period. According to the U.S. Census, there was an overall population increase of 24 percent during those two decades. Population growth rate in the WUI was significantly greater than the national average.

In a related question, you asked about the definition of “house,” inquiring if this meant a single-family dwelling, an apartment unit, etc. In the paper and in the presentation when we referred to a “house” it would have been more accurate for us to have said “housing unit.” We were referencing U.S. Census data; its numbers include houses, condo units, apartments, single rooms occupied as separate living quarters, etc.

Regarding California’s post-wildfire regulations, there was a question about the ceasing of insurer moratoriums on writing policies in wildfire-impacted areas. In the wildfire paper, we noted that the California Department of Insurance (CDI) required the California Fair Plan to terminate the moratorium it initiated on writing new fire insurance coverage in wildfire-impacted areas (December 2017). Also, the CDI directed insurers to cease moratoriums on issuing auto

1 The American Academy of Actuaries is a 19,500-member professional association whose mission is to serve the public and the U.S. actuarial profession. For more than 50 years, the Academy has assisted public policymakers on all levels by providing leadership, objective expertise, and actuarial advice on risk and financial security issues. The Academy also sets qualification, practice, and professionalism standards for actuaries in the United States.

Attachment B

7

https://www.actuary.org/sites/default/files/2019-06/Wildfire.IssuePaper_0.pdf

1850 M Street NW Suite 300 Washington, DC 20036 Telephone 202 223 8196 Facsimile 202 872 1948 www.actuary.org

insurance in wildfire areas (December 2017). The latest news is that the CDI is banning insurers from non-renewing policyholders in wildfire disaster areas under SB 824 (December 2019). The Academy’s Extreme Events Committee looks forward to working with the Catastrophe Risk Subgroup as you continue to expand your knowledge and understanding of wildfire risk and its implications for questions of insurer solvency. If you have any additional questions, you can contact us through Marc Rosenberg, the Academy’s senior casualty policy analyst, at 202-785-7865 or [email protected]. Sincerely, Jeri Xu, MAAA, ACAS Member Extreme Events and Property Lines Committee American Academy of Actuaries

Attachment B

8

mailto:[email protected]

Karen Clark & Company Model Overview

March 20, 2020

Glen Daraskevich, Senior Vice PresidentJoanne Yammine, FCAS, Director of Actuarial Services

Attachment C

9

© 2020 Karen Clark & Company

Agenda

Introduction to Karen Clark & Company (KCC)

Innovative risk metrics for identifying threats to solvency

How KCC partners with regulators

Introduction to KCC Models KCC US Hurricane Reference Model KCC US Earthquake Reference Model Additional KCC US Reference Models

2

Attachment C

10


About Karen Clark & Company

Established in 2008 by insurance industry veterans and pioneers in catastrophe risk assessment and management Karen Clark developed the first commercial hurricane model and founded the first catastrophe modeling company, AIR Vivek Basrur architected and led the development of AIR software technology, including CLASIC/2, CATRADER, and

ISOHomeValue (now 360Value) Senior staff have extensive experience in catastrophe model development and risk management

KCC is dedicated to delivering innovative approaches and new scientific tools for addressing current challenges in estimating and managing catastrophe risk Characteristic Event methodology Physical models Live event technology

KCC supports a diverse client base with consulting services and the RiskInsight® open modeling platform Top 10 US P&C insurers Global reinsurance intermediaries Leading global reinsurers CAT Bond/ILS fund managers Regulators Academia

3

Attachment C

11


Opportunities for Improving Catastrophe Risk Management Practices

More intuitive and operational metrics for identifying threats to solvency

Improved consistency and stability in the average annual loss estimates used in ratemaking

Direct access to the data and calculations underlying catastrophe loss estimates Open platform Transparent hazard and vulnerability components Embedded visualization tools

Innovative tools for delivering actionable information to decision makers

4

Attachment C

12


Why New Risk Metrics—EP Curves and PMLs Answer Some but Not All Important Questions

5

Loss, L

Probability p(L) that losses will exceed L

Exceedance Probability (EP) Curve

.4%

1%

1-in-100 year PML

1-in-250 year PML

Attachment C

13


KCC’s Unique Characteristic Events (CEs) Address These Additional Questions

6

One Chart Summarizes What Decision-Makers Need to Know About Catastrophe LossesReturn Period Loss ($Billions)

20 50

50 120

100 150

250 205

500 245

Attachment C

14


CEs Show Large Loss Potential Can Be Very Different Between Insurers with Same PML

7

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

A351

0A3

530

A355

0A3

570

A359

0A3

610

A363

0A3

650

A367

0A3

690

A371

0A3

730

A375

0A3

770

A379

0A3

810

A383

0A3

850

A387

0A3

890

A391

0A3

930

A395

0A3

970

A399

0A4

010

A403

0A4

050

A407

0A4

090

A411

0A4

130

A415

0A4

170

A419

0A4

210

A423

0A4

250

A427

0A4

290

A431

0A4

330

A435

0A4

370

A439

0A4

410

A443

0A4

450

A447

0A4

490

A451

0A4

530

A455

0A4

570

A459

0A4

610

LOSS

($ B

illio

ns)

Landfall Location ID

Company B

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

A351

0A3

530

A355

0A3

570

A359

0A3

610

A363

0A3

650

A367

0A3

690

A371

0A3

730

A375

0A3

770

A379

0A3

810

A383

0A3

850

A387

0A3

890

A391

0A3

930

A395

0A3

970

A399

0A4

010

A403

0A4

050

A407

0A4

090

A411

0A4

130

A415

0A4

170

A419

0A4

210

A423

0A4

250

A427

0A4

290

A431

0A4

330

A435

0A4

370

A439

0A4

410

A443

0A4

450

A447

0A4

490

A451

0A4

530

A455

0A4

570

A459

0A4

610

LOSS

($ B

illio

ns)

Landfall Location ID

Company A

100 year “PML” (one percent exceedanceprobability)

Losses from 100-year hurricane greatly exceed PMLLosses from 100-year

hurricane

Attachment C

15


0.0%

1.0%

2.0%

3.0%

4.0%

5.0%

6.0%

7.0%

0

500

1,000

1,500

2,000

2,500

3,000

3,500

A351

0

A355

0

A359

0

A363

0

A367

0

A371

0

A375

0

A379

0

A383

0

A387

0

A391

0

A395

0

A399

0

A403

0

A407

0

A411

0

A415

0

A419

0

A423

0

A427

0

A431

0

A435

0

A439

0

A443

0

A447

0

A451

0

A455

0

A459

0

Mar

ket S

hare

of L

oss

Loss

(Mill

ions

)

20 Year CE Profile

How the 20 Year Event Can Cause a Loss Close to the 100 Year PML

8

100 Year PML

0.0%

1.0%

2.0%

3.0%

4.0%

5.0%

6.0%

7.0%

0

500

1,000

1,500

2,000

2,500

3,000

3,500

A351

0

A355

0

A359

0

A363

0

A367

0

A371

0

A375

0

A379

0

A383

0

A387

0

A391

0

A395

0

A399

0

A403

0

A407

0

A411

0

A415

0

A419

0

A423

0

A427

0

A431

0

A435

0

A439

0

A443

0

A447

0

A451

0

A455

0

A459

0

Mar

ket S

hare

of L

oss

Loss

(Mill

ions

)

100 Year CE Profile

100 Year PML

Attachment C

16


More Consistent and Accurate Location-level Loss Estimates is Critical for Ratemaking

9

KCC Sampling Traditional Sampling

Attachment C

17


Ratemaking – Loss Cost Comparison – Miami, FL

10

KCC Sampling Traditional Sampling

Attachment C

18


RiskInsight® is a transparent and open model to allow users to independently verify model assumptions and better understand model processes

11

View vulnerability functions and secondary uncertainty assumptions

Interrogate event parameters and underlying data

Intuitive dashboards summarize key portfolio insights

Visually verify event footprints

Attachment C

19

Introduction to KCC Models – KCC US Hurricane Reference Model

Attachment C

20


Event Catalog

What’s Different about Karen Clark & Company Models?

13

Intensity

Vulnerability

Financial

Apply policy conditions toestimate insured losses

For each event estimate intensity at each

location

Based on intensity andexposure at each location

estimate damage

Create a large sampleof hypothetical events

Where? How big?How frequent?

Based on the same science and components as traditional models.

Provides all the traditional metrics including PMLs, TVaRs, and AALs.

Advancement: Additional risk metrics for underwriting and portfolio management

Advancement: Unique sampling methodology enabling robust high resolution (location level) loss analyses

Advancement: Model components fully transparent

Advancement: Built-in tools and high resolution mapping for more efficient internal modeling process

Advancement: More accurate model loss estimates

Attachment C

21


Random Sampling Methodologies Can Result in Spatial Biases and “Blind Spots” for Threats to Solvency

14

100 Year CEs Vmax=165Randomly Generated Category 5 Events

Attachment C

22


Modeling the Frequency and Severity of US Hurricanes

15

Relative Wind Speed

High

Low

1 2 3 4 5

FLNW

1 2 3 4 5

NE

1 2 3 4 5

Gulf

1 2 3 4 5

Mid-Atl

1 2 3 4 5

SE

1 2 3 4 5

FLNE

1 2 3 4 5

FLSO

1 2 3 4 5

TX

Attachment C

23


Velocity of Maximum winds at Landfall (Vmax) Values are Assigned Based on Modeled Generalized Pareto Distributions (GPD) for Each Landfall Location

16

108

64

21

160

165

170

175

180

185

Vmax (mph)

Prob

abili

ty

GPD CDF - Vmax

Annual Frequency * Probability(Vmax range) * Catalog Length

0.012583 * 0.15558 * 50,000 = 98

0.012583 * 0.19141 * 50,000 = 120

0.012583 * 0.20572 * 50,000 = 129

0.012583 * 0.39741 * 50,000 = 250

0.012583 * 0.04988 * 50,000 = 31

26 25 25 24 23 23 22 21 21 20 20

75 77 79 81 83 85 87 89 91 93 95

Landfall Location A4250 (Miami, FL) Annual Frequency = 0.012583

Attachment C

24


Result is a Complete and Spatially Unbiased Set of Landfall Point, Vmax Pairs

17

4200

4390Model

Num

ber o

f Hur

rican

esN

umbe

r of H

urric

anes

Attachment C

25


KCC Employs a Joint Probability Method (JPM) Using a Ternary Tree Hierarchy For Generating Events

18

Model parameters are ordered in the tree according to their importance in the model output

Vmax

Track Direction

Rmax

Forward Speed

Three nodes at all levels

Symmetrical and balanced selection of values Vmax most critical value and all other variables

selected conditionally based on Vmax “Ternary” in that each parent node has three

child nodes

Attachment C

26


High-Resolution Wind Footprint Generation Methodology

19

NLCD Land Use/Land Cover (2011) used to calculate roughness lengths

Attachment C

27


CE Methodology Improves Risk Selection and Pricing with More Consistent and Accurate Location-level Loss Estimates

AALs based on uniform exposure on a 1 KM grid (open terrain)

Logical relation to risk

No spatial bias

20

Attachment C

28


KCC Multi-Peril Modeling Provides Integrated Flood, Storm Surge, and Wind Footprints To Accurately Capture the Impacts of Live Events and Future Scenarios

21

Wind Speed Footprint Inland Flood FootprintCoastal Storm Surge Footprint

Attachment C

29


KCC Has Implemented the Aerodynamic Load Resistance Component Method for Vulnerability Function Development

22

Aerodynamic load resistance-based component method (ALR) used to develop base vulnerability functions and analyze the effect of secondary building characteristics

In the ALR method, the relevant building components are first identified

The mean vulnerability function for each component is then developed using

Wind loading information from wind tunnel databases Wind resistance information from literature and expert

judgment

The component vulnerabilities are combined to produce the vulnerability of the building structure by considering

Dependency of the system of components Exposure share of each component

…

Exposure Share Dependency

Attachment C

30


Post Event Analytics Used to Inform and Validate the Vulnerability Functions

23

KCCPost Event Analytics

KCC post-eventdamage surveys

Site surveys with claims adjustors

Desktop review of individual claims

files

Portfolio claims analyses

Attachment C

31

Introduction to KCC Models – KCC US Earthquake Reference Model

Attachment C

32


Evolution of USGS Hazard Maps

25

1996 2002

2008 2014

Attachment C

33


USGS Report Evolution: Improve Best Estimation and Better Handling of Uncertainty

26

2008

2014

20021996

Attachment C

34


2014

USGS Report Evolution: Additional Faults Mapped Over Time and Enhancements in Modeling Fault-based Seismicity

27

2008

Cascading Events Grand Inversion

Hayward-North MFD

Newport-Inglewood_alt2 MFD

Attachment C

35


KCC CEs are Based on USGS MFDs and Account for Location Uncertainty

28

Select the CE Magnitude Use fault MFD (from UCERF3 in CA)

Calculate the rupture area (Wells-Coppersmith)

Calculate the rupture length Depending on depth

Assign the ruptures7

6

54

3

2

1

Attachment C

36


CEs Account for “Location Effect” by Generating Background Events on a Uniform Grid (Aligned with USGS)

29

1E-05

0.0001

0.001

0.01

0.1

1

10

5 5.5 6 6.5 7 7.5 8 8.5 9

Cum

ulat

ive

Rate

Magnitude

Fault

Background

Total

Rate

Magnitude

a

USGS/KCC Rate Distribution

Geospatially Biased Spatially Unbiased

Attachment C

37


CEs Enable High-Resolution, Location-Level Risk Metrics, and Avoid Surprise Losses

30

100 year Events:

0

100

200

300

400

500

600

700

Loss

($Bi

llion

s)

Locations

"Characteristic Event" losses from the 100 year magnitude earthquake

100 year PML

Attachment C

38


RAA 2018 Vendor Model Comparison Session: California Earthquake

31

Model Comparison: Postal code Loss Costs (GU)

RMS AIR CL

KCC IF USGS

Attachment C

39


RAA 2018 Vendor Model Comparison Session: California EQ EP Curves

32

AEP GROUND-UP LOSSES

Key Loss Metrics ($M)

NORTHERN CA

SOUTHERN CA

Attachment C

40


Multiple Sources of Information Required to Estimate Vulnerability

33

Observed Loss

Engineering KnowledgeAnalytical Approach Damage Observation

Experimental (UC Berkeley shaking table)

Experimental (Kyoto University shaking table)

Attachment C

41


Information Gained from Post-Event Damage Observation: Residential Napa Valley EQ, 2014

34

Structural Damage:Out-of-PlumsDifferential SettlementBuckled ColumnsSheared WallsCripple Wall

Non-Structural Damage:Chimneys CollapseRoom over GarageEquipment

Attachment C

42


Validation of SA (0.3 sec) Footprints

35

Northridge, 1994 Loma Prieta, 1989

KCC

USG

S

Louisa, 2011 Nisqually, 2001 Napa Valley, 2014

Attachment C

43

Introduction to KCC Models – Additional KCC US Reference Models

Attachment C

44


Major US Perils Supported

37

Winterstorm

Coastal Flood Tornado/Wind

Hail

Inland Flood

Wildfire Winterstorm

Attachment C

45


SCS Reference Model - KCC Employs a Composite Index to Capture the Physical Variables Driving Tornado and Straight-line Wind Behavior

CAPE, SHEAR, and SRH are very useful, but SCS require many ingredients

Composite Index – combine multiple parameters

EHI = 𝑓(𝐶𝐴𝑃𝐸, 𝑆𝑅𝐻)

STP = 𝑓(CAPE, SRH, SHEAR, LCL, CIN)

SCP = 𝑓(𝐶𝐴𝑃𝐸, 𝑆𝑅𝐻, 𝑆𝐻𝐸𝐴𝑅)

Enhanced Significant Tornado Parameter

38

Attachment C

46


SCS Reference Model - Model Data Sources

39

North American Regional Reanalysis (NARR) 1979 – present precipitation, wind, temp, pressure, etc. 45 vertical layers KCC has archives 20Tb of data

High Resolution Rapid Refresh Model (HRRR) 2016-present precipitation, wind, temp, pressure, etc. Hourly update cycle 10x resolution increase over NARR

NEXRAD Radars Archived since 1995 Radar Reflectivity

Multi-Radar/Multi-Sensor System (MRMS) 2016 - present multiple radars, satellites, surface observations, upper air

observations, NWP, etc. “Maximum Estimated Size of Hail” (MESH)

Attachment C

47


SCS Reference Model - Enhanced Tornado/Wind Footprint Creation: Identifying Where Storm Potential (ESTP) is Actualized

40

Attachment C

48


SCS Reference Model - Hail Detection and KCC Hail Footprint Generation

41

Reflectivity is measured in dbZ

Radars scan at different tilts (angles)

Base reflectivity is the lowest 0.5° tilt

Composite reflectivity is the maximum reflectivity at any tilt

Is combined with NWP to identify hail forming regions (MESH)

Attachment C

49


Flood Reference Model – Cellular Automata Supports High Resolution Pluvial Flood Modeling

42

Different methods have been applied to modeling pluvial flooding

1. Solving the full shallow water equations (SWEs) Generally applied in single-catchment studies Difficulty simulating flood inundation in large catchments Non-economical for large spatial scales or large event sets

2. Rapid Flood Model (RFM) Faster to run than SWEs by orders of magnitude Disregards temporal evolution of flood Requires substantial pre- and post-processing and tuning

3. Cellular Automata Accuracy is comparable to the SWE approach for pluvial flooding (e.g. Ghimire et

al., 2013; Guidolin et al., 2016; Jamali et al., 2019) Higher computational burden than the RFM but more efficient than the SWE

approach Represents the physics of water flowing over topography

Attachment C

50


Flood Reference Model - Validation of the High-Resolution Footprints using Flood Reports

43

Two vehicles drove into flooded drainage ditch

Homes in the Centre Lake

Community had at least 1 foot of water inside

Water reported in homes

Attachment C

51


RiskInsight® Directly Ingests the Latest Meteorological Information to Deliver Event Loss Estimates in Real Time

44

1 - Reads in Projected Track 2 - Calculates Intensity Footprint

3 - Estimates Damage 4 - Calculates Loss

NHC Track Data

Attachment C

52


Timely and actionable information allows decision makers to effectively deploy resources to reduce the severity and time required to settle claims

Why is Real Time Information Essential During a Live Event?

45

Losses and claims information by intensity

Losses and claims information by geography

Maps of losses and claims

Attachment C

53


CEs Can Assist Rating Agencies and Regulators in Evaluating Threats to Surplus

46

1) Gross losses by landfall point 2) Gross losses net of FHCF and private reinsurance

3) Net surplus (surplus – step 2) 4) Normalized net surplus ratio (step 3 / surplus)

FHCF + PrivateReinsurance Program

Loss

es $

(mill

ions

)

Loss

es $

(mill

ions

)

Loss

es $

(mill

ions

)

NSR

Landfall Point Landfall Point

Landfall PointLandfall Point

Attachment C

54


Summary

47

Characteristic Events are a great metric to identify threats to solvency

Improved consistency and stability

Innovative and transparent models

Live Event tracking

More accurate loss estimates

Attachment C

55


Karen Clark & Company Model Overview

48

Questions?

Attachment C

56

Research Brief

The Florida insurance market: An analysis of vulnerabilities to future hurricane losses Jack E. Nicholson, Karen Clark, and Glen Daraskevich In Journal of Insurance Regulation, 37(3)

Insurance regulators need a method to evaluate systemic risk from natural catastrophes. Current rating agency methods using traditional risk metrics account for a fraction of potential catastrophe losses and provide a very limited view of an insurer’s solvency. Furthermore, Probable Maximum Loss (PML) levels currently employed for insurance rating of individual companies do not capture the systemic risk of an entire insurance market to natural catastrophe events.

The Characteristic Event (CE) methodology provides additional risk metrics that capture the full scope of potential losses from major perils, including hurricanes, earthquakes, and severe convective storms and are better suited to evaluate the systematic risk at a state level. As a case study, CEs were used to test the Florida insurance market to determine the solvency of private and public insurers and Florida’s public risk financing entities – Citizens Property Insurance Corporation (Citizens), the Florida Hurricane Catastrophe Fund (FHCF), and the Florida Insurance Guaranty Association (FIGA).

Regulators need a consistent metric by which to evaluate the solvency of individual insurers and the insurance market

• Risk metrics need to be stable year to year to provide a robust view ofchanges in the financial strength of individual insurers.

• Insurance systems that rely on interdependent individual markets needto be stress tested collectively to ensure the solvency of direct insurersand government insuring entities alike to determine what cost would bepassed on to taxpayers in the event of a major catastrophe.

Probable Maximum Loss (PML) levels derived from catastrophe models alone are insufficient to capture the nuances of natural catastrophe impacts.

• Very little research has been conducted to test the viability of currentmethods employed to evaluate insurers.

• PMLs lack the consistency, transparency, and stability needed byregulators for a robust rating methodology.

• PMLs are not additive and cannot be used collectively to stress test aninsurance system.

• Individual catastrophe models can be updated and differ year to year,and thereby cause their PMLs to differ and thus contribute to theinability of the PML to evaluate insurer surplus over time.

Current views of risk using historical storms to stress test insurers provide a very limited perspective of potential events.

• Current stress tests conducted by regulators include requiring insurers toprovide loss estimates for specific historical storms, but this does notrepresent the full range of potential events that an insurer couldexperience as future events are unlikely to be exact replicas of historicalones.

• This method does not account for impacts on the market when multipleinsurers have their reinsurance layers triggered due to a major event andprovides no view of systemic risk.

Characteristic Events (CEs) can be applied to individual insurers and insurance markets.

• CEs are a suite of simulated events that are generated by analyzing thelikelihood of a specific event occurring at a specified time period ofinterest to insurers (e.g., the 100 or 250-year hurricane).

• They evaluate the likelihood of event impacts at every vulnerablelocation and provide complete geographic coverage of catastrophe riskpotential.

• CEs are an additive risk metric and can be used to evaluate the financialstrength of an individual insurer or an entire market.

• Using this methodology, individual insurers and entire insurance marketscan be stress tested.

Attachment C

57

The Florida Insurance Market Characteristic Event (CE) Analysis On average, Florida is hit with a hurricane every other year. Florida has been lucky for the past two decades. The 10-year period from 2006 to 2015 with no landfalling hurricanes was unprecedented in the historical record. In the 1920s, Florida experienced 10 landfalling hurricanes, including two Category 4 storms that would each result in more than $75 billion of insured losses if they occurred today.

This study illustrates a methodology that would enable Florida policymakers to more fully quantify the current vulnerabilities of the residential property insurance market in Florida in order to strengthen and enhance the resiliency of the system. Even though the natural perils and characteristics of the insurance marketplace vary from state to state, this same methodology could be applied by regulators to better quantify and manage individual insurance entities and the systematic risk to the insurance market in the state.

Study methodology The set of hurricane events used for the analysis was selected to provide meaningful comparisons between insurers. Landfall points were positioned at 10-mile increments along the entire Florida coastline. At each landfall point, the characteristics of three types of hurricanes were defined: the 20-, 50-, and 100- year hazard probability events.

Because hurricane risk changes along the Florida coast, the event characteristics must vary by landfall point in order to keep the hazard probability the same. For example,

the 100-year hurricane in Southeast Florida is a Category 5 hurricane, but in parts of Northeast and Northwest Florida, it is a Category 4 storm. Likewise, the 20- and 50-year hurricane characteristics vary by region within Florida.

As long as the events are credible from a meteorological perspective, the exact parameters selected for each storm are not critical for the analyses. What is important is that the same comprehensive set of storms is applied to each insurer. This is the only reliable way insurers can be compared with respect to hurricane vulnerability and financial solvency.

The loss estimates for this study were generated using the Karen Clark and Company (KCC) high-resolution hurricane model. In addition to the traditional EP curve metrics, the KCC model produces loss estimates for different return period events—the CEs.

Overview of individual insurer analyses and assumptions For each Florida insurer, the losses for the 333 hurricanes in the 20-, 50- and 100-year CE event sets were estimated. A fully probabilistic loss analysis was also conducted for each insurer to estimate the EP (exceedance probability) curve and the 100-year PML.

To estimate each insurer’s net loss, a number of assumptions were made regarding their reinsurance programs. It was assumed that each insurer buys risk transfer protection up to 75% of the KCC model-generated 100-year PML. The KCC hurricane model PMLs tend to be

above the midpoint range of other vendor hurricane catastrophe models. Therefore, 75% of the KCC PMLs will be close to the average PML for the five models found acceptable by the Florida Commission on Hurricane Loss Projection Methodology (FCHLPM). While this assumption will not be correct for every insurer, it should not bias the results.

Private reinsurance retentions were set at the minimum of 10% of surplus or the FHCF retention amount. Rating agency guidance indicates Florida insurers should have a retention equal to 15% of surplus or less. Companies for which their retentions were publicly available had an average retention of 10% of surplus. For several insurers, their private reinsurance programs were publicly available. The surplus figures were taken from the 2016 year-end numbers as reported in FLOIR’s 2017 annual report.

For each insurer, the following was calculated:

• Gross losses for each CE• 100-year PML• Recoveries from the FHCF• Recoveries from private risk

transfer programs• Surplus minus net losses• Normalized solvency ratio (NSR)

The NSR is the rate adjusted normalized net surplus ratio for all the 100-year events. Rather than representing an absolute value, the NSR reflects the probability of how often an insurer would become insolvent given the event set and provides a normalized metric that can be used to compare different insurers.

Attachment C

58

𝑁𝑁𝑁𝑁𝑁𝑁

= ��𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 𝑁𝑁𝑁𝑁𝑁𝑁 𝑁𝑁𝑆𝑆𝑁𝑁𝑆𝑆𝑁𝑁𝑆𝑆𝑆𝑆𝐹𝐹𝐿𝐿 100 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖

𝐸𝐸

𝑖𝑖=1

∗ �𝐸𝐸𝐸𝐸𝑁𝑁𝐸𝐸𝑁𝑁 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝐹𝐹𝐿𝐿 100 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖∑ 𝐸𝐸𝐸𝐸𝑁𝑁𝐸𝐸𝑁𝑁 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝐹𝐹𝐿𝐿 100 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖

��

Results for Insurers An insurer with an NSR > 0 has an expected positive surplus from the 100-year events. As the NSRbecomes more negative, the insurerhas a higher probability ofinsolvency from a 100-yearhurricane.

The NSR illustrates the wide disparity between Florida insurers. Twenty-six (42%) have positive NSRs and can be considered the most financially secure domestic insurers. On the other extreme, eight insurers (13%) have NSRs of -2 or less, indicating a relatively high likelihood of experiencing insolvency from a hurricane. All these insurers are rated “A” or better by Demotech. This information indicates that existing rating methodologies, which rely heavily on the PMLs, do not sufficiently differentiate insurers with respect to financial stability.

The study results also imply that the current FLOIR stress tests based on three historical hurricanes are not comprehensive enough to identify insurers that are vulnerable to hurricane losses. More comprehensive stress tests along with an improved insurer rating agency methodology would strengthen the Florida residential property insurance market.

Results for Florida’s Public Risk Financing Entities The second part of the study examined the impacts of the 20-, 50- and 100-year hurricanes on Citizens, the FHCF and FIGA.

Citizens

Notably, Citizens is financially secure due in large part to the amount of its surplus. Citizens Coastal Account has an NSR of 0.71, and Citizens Personal Lines Account (PLA)/Commercial Lines Account (CLA) has an NSR of 0.73, among the highest of all Florida insurers.

Figure 1 shows the losses from the 20-, 50- and 100-year hurricanes (vertical axis) by landfall point (horizontal axis), commonly referred to as the CE profile for Citizens PLA/CLA and Citizens Coastal Account (CA). The highest bars indicate where the insurer has exposure concentrations and is most vulnerable to hurricane landfalls. The amount of reinsurance (private plus FHCF) and surplus available is shown by the dotted line.

Figure 1 - Citizens CLA/PLA Profile

FHCF

To estimate the FHCF payout for each CE, the FHCF coverage level, coverage amount, and retention were first calculated for each participating insurer. More specifically, the coverage level and FHCF reimbursement premium reported by each company under the 2017–2018 FHCF annual

Coverage Level (%) Retention Multiple

90 5.1028

75 6.1234

45 10.2056

Each insurer’s coverage amount is calculated as 14.9294 (the payout multiple) multiplied by the insurer’s reported FHCF reimbursement premium. The FHCF recovery for each participating insurer was estimated for each event, and the cumulative FHCF payout by event was estimated as the sum of FHCF recoveries for each participating insurer. Figure 3 shows the FHCF payout by event, by landfall point.

Figure 3: FHCF CE Profile

The FHCF’s statutory maximum limit for the 2017–2018 reimbursement contract year is $17 billion. From the FHCF’s CE profile, it can be noted that a one in 500-year loss is not expected to exhaust the FHCF’s maximum limit, but would exhaust only $15.62 billion of the $17 billion limit. For the FHCF’s statutory limit to be exhausted, all participating insurers would need to exhaust their FHCF coverage limit. Additionally, this result implies that the cost of

reimbursement contract was obtained. Pursuant to the contract, each participating insurer’s retention is calculated as the FHCF reimbursement premium multiplied by the retention multiple outlined in Figure 2.

Figure 2: Coverage and Retention Multiples

Attachment C

59

risk transfer products should reflect the FHCF’s lower probabilities at the upper layers of coverage. The methodology used here illustrates an improvement over the crude methodology that has been used historically to price FHCF risk transfer coverage.

FIGA

The study results can be used to quantify the numbers of insurers likely to become insolvent, defined as an insurer having a loss exceeding the insurer’s risk transfer program, under different industry loss scenarios. Figure 4 shows the expected number of insolvencies by industry loss.

Figure 4: Companies Exceeding Risk Transfer Program

Industry Loss Range ($B)

# of Companies Exceeding their

Reinsurance Programs

<25 0

25 to 50 11

50 to 75 20

75 to 100 37

>100 48

The results indicate that at an industry loss size between $50 billion and $75 billion, 20 Florida insurers could become insolvent. This number is notable because most models agree that Hurricane Andrew would cause $50 billion to $60 billion if it occurred today. This means that more companies would become insolvent today than in 1992 from an Andrew-size loss. It is important to note that not all $60 billion events would cause 20 insolvencies. The number of insolvencies depends heavily on where the hurricane makes landfall.

FIGA is most exposed to hurricane landfalls near Miami, as noted from

Figure 5, where event losses can exceed risk transfer programs by several multiples. In extreme cases, the FIGA’s debt obligations can exceed $40 billion. However, FIGA is limited in its statutory authority to fund insolvencies. A hurricane event on the order of Hurricane Andrew could exhaust its financing capabilities (Florida Guaranty Insurance Association, 2018). Results of the analysis indicate that it does not take a one in 100-year event to stress FIGA’s capabilities to the limit. FIGA is vulnerable to the potential volatility of the financial markets following an event and by its limited assessment authority.

Figure 5: FIGA Debt Profile

This study also assessed the number of policies from the insolvent insurers that would be renewed by Citizens post-event.

At its maximum historical policyholder count, Citizens had almost 1.5 million policies in its combined PLA/CLA and CA. At the end of 2017, the combined policy count was 440,406. Figure 6 illustrates that there are a large number of one in 50-year hurricanes that could result in the repopulation of Citizens to its historical maximum policy count. Additionally, a number of the one in 50-year hurricane events could result in a surge of policies by inundating Citizens with an extra 1 million policies or more, far surpassing the historical record. For certain landfall locations, a one

Broader applications of the CE methodology The CE methodology provides a reliable risk metric with which to evaluate individual insurers and the broader insurance market. A number of natural catastrophe perils, including earthquakes and severe storms, can be evaluated in any region using a similar approach.

The results of such analyses offer insight into vulnerable insurers to determine the potential loss from an event that would be passed on to higher layers of insurance or the taxpayer. Because this methodology relies on an event set unbiased by location and offering full geographic coverage, the full range of potential events is captured by the loss metrics. CEs are the only risk metric currently in use that can be used for both rating individual insurers and evaluating the systemic risk of an insurance market.

in 100-year hurricane event could result in the Citizens policy count exceeding 4 million policyholders, which represent about two-thirds of all policyholders in the state. Citizens, FIGA and the entire Florida residential property insurance market are highly vulnerable to insurer insolvencies, which could arise from moderate to large hurricane events that are not extreme, but that could easily occur in the future.

Figure 6: Policies to Citizens from Insurer Solvencies

Attachment C

60

materials - catastrophe risk (e) subgroup...academy of actuaries (academy) on wildfire: lessons...

Documents