1. Weathering The Storm: Bringing Clarity To The Unknowns Of Severe Thunderstorm Modeling Prepared by Impact Forecasting March 2015

2. 1 Proprietary & Confidential Aon Benfield: The world’s leading reinsurance intermediary Experts in utilizing catastrophe models, annually analyzing 70% of the Homeowners market The worst priced 10% of the portfolio erodes the overall ROE by 500-800 bps Tools and services designed to help clients maximize the opportunities for Profitable Growth Experts in quantifying reinsurance cost drivers, with more than $9B of annual ceded catastrophe premium Experts in retained profit loads including what can practically be implemented, based upon over 100 engagements per year in catastrophe ratemaking filing support No bad risks, only bad prices

3. 2 Proprietary & Confidential Impact Forecasting: Aon Benfield’s cat model development center Risk rating information to enhance primary underwriting Breadth of model coverage Transparent ELEMENTS loss calculation platform Customizable models Catastrophe analysis and reporting Natural disaster insured losses: USD 900B Natural disaster economic losses: USD 3.4T 85+ modeling experts 100+ catastrophe models 50+ countries 12 perils 6,000+ Events in the catastrophe insights database 200+ ELEMENTS users

4. 3 Proprietary & Confidential Reality: STS cat models haven’t performed well Comparison of Severe Thunderstorm AALs to actual company experience Analysis performed by Aon Benfield, June 2014  Historically, STS models have significantly underestimated AALs and claims –This is where the high frequency/low severity events occur (hail, convective winds)  Current feedback is that the models are also significantly overestimating PMLs –Too much emphasis on large outbreaks, NWP modeling overemphasizing instability parameters

5. 4 Proprietary & Confidential Year-to-year frequency changes lead to loss variability  Expect to pay claims frequently –2 years out of 5 companies may pay 50% above cat AAL  An accurate loss pick is required for rate making –Events can be geographically broad or geographically narrow –Account for LOB variation by state  Every year is NOT created equal  Every state is NOT created equal HO Results in Thunderstorm vs. Hurricane States = LR ~1SD above the mean = LR ~3SD above the mean

6. 5 Proprietary & Confidential Tornadoes have small footprints, rapid identity changes  Comparison between a very small hurricane wind swath (2005’s Hurricane Dennis) and the largest tornado wind swath in history (2013’s El Reno, OK tornado) - 74 El Reno tornadoes could fit in Hurricane Dennis’ wind swath  Tornadoes can change identity quickly Hurricane Dennis 65+ mph wind swath: 3,100 square miles El Reno, OK tornado 65+ mph wind swath: 42 square miles

7. 6 Proprietary & Confidential Incomplete data = big hazard uncertainty  Convective wind storm reports can be reported in two ways: –Actual gust wind speed (57 mph or greater) –Damage caused by severe thunderstorm wind gust  Some regions more prone to reporting wind DAMAGE vs. actual wind speed

8. 7 Proprietary & Confidential Many issues to deal with for claims reporting What exact peril caused the damage? How is the peril coded in claims? What exact component of the structure was damaged? Was the damage caused by thunderstorm- produced winds or winds not associated with a thunderstorm? Was only the damaged component repaired/replaced, or was other damage included in the claims payout that was wasn’t documented? Are the TIVs and construction codes for the policy correct? What if only a few roof shingles were damaged, but a full roof replacement was done? How does this affect modeling? ‘Me Too’ Syndrome: one house gets repaired, rest in the area often follow… The dreaded public adjusters and fly- by-night companies…

9. 8 Proprietary & Confidential How have insurance companies been pricing STS risk? ANSWER: 10-25 years of claims history

10. 9 Proprietary & Confidential Process of using cat models in insurance companies  INPUT DATA: Gain an accurate understanding of their exposure portfolios –exposure concentrations, predominant construction types, occupancies, etc.  SCENARIO MODELING: Compare historical scenarios to actual claims using historical exposure portfolios to test the vulnerability and financial components of the models and gain an understanding of how models are handling different exposure characteristics and to gain confidence in a model’s methodologies and tendencies  STOCHASTIC MODELING: Apply the lessons learned from scenario modeling to stochastic modeling to more accurately determine AALs and PMLs  ACCURATE RISK PRICING: Use modeled results along with other insurance and exposure information to make informed decisions on premiums + + =

11. 10 Proprietary & Confidential What’s the problem?  Claims from the severe thunderstorm peril continue to increase due to increased urban sprawl, increased material costs, increased home sizes, and increased property values  Insurance companies can no longer depend on using actuarial methods or claims experience to price the peril risk  Companies don’t trust the catastrophe models since there’s no scenario models to isolate frequency and severity with, no convergence on losses between model releases, inaccuracies in AAL results, and wide variability in PML results  Insurance companies response to Aon Benfield and Impact Forecasting: HELP!!!!!!!!!

12. 11 Proprietary & Confidential Impact Forecasting’s path to our current model suite  2009 –Request: Large Midwest insurance client wanting a better understanding of AALs and to better calibrate their exposure portfolio’s loss behavior –Core Problem: Isolate model frequency/severity to focus on vulnerability –Solution: Develop 2006-2012 historical STS event sets to compare back to their daily/annual claims  2013 –Request: Another large Midwest insurance client wanting a better understanding of AALs, wanting to better calibrate their exposure portfolio’s loss behavior, to build in custom exposure characteristics and to start pricing ACV vs. RCV for roof coverage –Core Problem: Isolate model frequency/severity to focus on vulnerability, bring specific unique exposure characteristics into the model that weren’t currently included –Solution: Develop additional 2003-2005 historical event sets to compare back to their daily/annual claims, run “what-if” scenarios through historical event sets to see how underwriting rules changed losses

13. 12 Proprietary & Confidential STS RePlay model overview  Historical recast of the last 12 years of STS events (tornadoes, hail, convective winds) –Event set is updated on an annual basis  Over 80,000 historical outbreaks comprised of nearly 7.5 million individual tornado, hail, and convective wind swaths  Retains the historical characteristics of the occurrences when possible (intensity, size, directionality)  Vulnerability module for residential, commercial and industrial risks including dealer open lots

14. 13 Proprietary & Confidential Special considerations for wind and hail perils  Limited to point estimates –No end coordinate in SPC data –No data on swath shape (length, width) –SOLUTION: Utilize current in-place Impact Forecasting research on characterization of hail/wind swaths  Multiple reports could represent portions of same event swath –Overlapping swaths over-count losses –SOLUTION: Bin events by temporal and spatial proximity and treat similar reports as duplicates of the previous reports Single Hail Swath Multiple Hail Reports

15. 14 Proprietary & Confidential Special considerations for EF3+ tornadoes  Strong to violent tornadoes (EF3+) are often long-lived with complex track shape and varying intensity  SPC LSR data characterizes start and end coordinates and single intensity rating –All EF3+ events in 2003-2014 data set, but majority didn’t have detailed info  Some tornado tracks with EF3+ intensity contained detailed information –Average of 4 additional segments for each modified single tornado track –Ad hoc research for event shape and intensity (NWS-sourced official track data) –Re-parameterized events as necessary for proper representation within the event set

16. 15 Proprietary & Confidential Position uncertainty is handled directly in the event set  Tolerance in location accuracy: adds local uncertainty in LSR event placement –Street-level address not available (cross-street assignments most likely) –SOLUTION: Sample every LSR a number of times, adding position uncertainty Less position uncertainty for tornadoes (more accurate reporting) More position uncertainty for hail and wind Nearly 7.5 million events generated in ELEMENTS from nearly 300,000 LSRs

17. 16 Proprietary & Confidential STS RePlay event losses vs. PCS (RES/COM only)  2003 – 88 (428 tornado reports, 2,083 hail reports, 1,066 wind reports) –May 2 to May 11, 2003 –AL, AR, CO, GA, IL, IN, IA, KS, KY, MS, MO, NE, NC, OH, OK, SC, SD, TN  2008 – 42 (182 tornado reports, 759 hail reports, 382 wind reports) –May 22 to May 26, 2008 –CO, IA, KS, MN, NE, OK, WY  2011 – 46 (550 tornado reports, 1,088 hail reports, 1,375 wind reports) –April 22 to April 28, 2011 –AL, AR, GA, IL, KY, LA, MS, MO, OH, OK, TN, TX, VA

18. 17 Proprietary & Confidential STS Stochastic model overview  Robust probabilistic model simulating up to 35,000 years of events  Leverages historical SPC data from 1950-2012 –Hail and convective wind frequencies are weighted towards the 2003-2012 period (more accurate reporting period)  Stochastic event set contains over 1.7M outbreaks –Includes over 266 million individual severe weather occurrences –Combined outbreaks by historical date (generally 1 to 4 days – PCS event definitions)  Includes ALL events, not just events that meet a hazard or loss threshold –Important to companies that may have localized exposure concentrations  Uses the same hazard and loss methods as well as the same vulnerability functions as STS RePlay for modeling consistency

19. 18 Proprietary & Confidential STS Stochastic model development  Event frequencies are based on the time series of observed historical severe weather events recorded in the Storm Prediction Center database –Duplicate report filtering is performed to minimize over-counting of hazard and loss  To account for the under-reporting of tornado and hail events (especially low intensity events such as F0, F1, and F2 tornadoes and H0, H1 and H2 hail) in the earlier historical years, linear regressions are performed to de-trend the annual number of single tracks per outbreak and the number of outbreaks per year –This way, most of the later years’ bias will be removed while the short term trends in the original data is maintained

20. 19 Proprietary & Confidential Spatial reassignment addresses location uncertainty  Incorporates latitude & longitude coordinates from historical STS peril originations and samples the probable spatial coordinates within a given region  Causes the perturbation of stochastic samples uniformly in a radial fashion within a given distance  Allocation process incorporates equally-sized sampling regions within the radial to assure a consistently-stratified spatial set  Severe thunderstorm peril elements within the outbreak were usually sampled up to a predetermined realistic radius limit

21. 20 Proprietary & Confidential PCS AAL vs. IF STS Stochastic model loss (using IED)  PCS AAL (2003-2014), all losses adjusted to 2014 USD  IF model loss using IF’s 2014 Industry Exposure Database (IED)  PCS AAL: $10.5 billion  IF STS Stochastic model AAL: $12.1 billion  Difference (IF vs. PCS): +15.4%  NOTE: PCS events do NOT include events that fall below their $25 million catastrophe number designation threshold, but IF keeps ALL events because they can add 10-20% to overall losses!

22. 21 Proprietary & Confidential Vulnerability is fully transparent and customizable  STS vulnerability suite considers statistical variation in wind speed and hail size as well as construction, occupancy, building age, and wind zone classes  Clients can compare their own historical loss ratios directly with IF’s vulnerability curves and make adjustments easily directly within the ELEMENTS catastrophe modeling software  Additional exposure characteristics can be easily added through ELEMENTS functionality Hail Size Loss $ EF- Scale Wind Speed

23. 22 Proprietary & Confidential Summary  No matter the methodology being taken by catastrophe modeling companies, insurance companies still are having issues with modeling the severe thunderstorm peril –Hazard variabilities and issues –Vulnerability variabilities and issues –Claims resolution and exact causes  Impact Forecasting went back to the basics with this peril and developed the STS RePlay historical model that retains the majority of the actual event characteristics, bringing in higher resolution event data into the modeling space –Higher resolution event data allows for more accuracy in both the RePlay and stochastic modeling environments –The inclusion of all events, full Monte Carlo sampling, and full transparency on all model data and assumptions helps insurance companies get a handle on their own view of risk, allowing for more accurate regional or state pricing

24. 23 Proprietary & Confidential Contacts Steve Drews Director Impact Forecasting 1 (312) 381-5888 steven.drews@aonbenfield.com

25. 24 Proprietary & Confidential Disclaimer Legal Disclaimer © Aon UK Limited trading as Aon Benfield (for itself and on behalf of each subsidiary company of Aon Plc) (“Aon Benfield”) reserves all rights to the content of this report or document (“Report”). This Report is for distribution to Aon Benfield and the organisation to which it was originally delivered by Aon Benfield only (the “Recipient”). Copies may be made by that organisation for its own internal purposes but this Report may not be distributed in whole or in part to any third party without both (i) the prior written consent of Aon Benfield and (ii) the third party having first signed a “recipient of report” letter in a form acceptable to Aon Benfield. This Report is provided as a courtesy to the recipient and for general information and marketing purposes only. The Report should not be construed as giving opinions, assessment of risks or advice of any kind (including but not limited to actuarial, re/insurance, tax, regulatory or legal advice). The content of this Report is made available without warranty of any kind and without any other assurance whatsoever as to its completeness or accuracy. Aon Benfield does not accept any liability to any Recipient or third party as a result of any reliance placed by such party on this Report. Any decision to rely on the contents of this Report is entirely the responsibility of the Recipient. The Recipient acknowledges that this Report does not replace the need for the Recipient to undertake its own assessment or seek independent and/or specialist risk assessment and/or other relevant advice. The contents of this Report are based on publically available information and/or third party sources (the “Data”) in respect of which Aon Benfield has no control and such information has not been verified by Aon Benfield. This Data may have been subjected to mathematical and/or empirical analysis and modelling in producing the Report. The Recipient acknowledges that any form of mathematical and/or empirical analysis and modelling (including that used in the preparation of this Report) may produce results which differ from actual events or losses. Limitations of Catastrophe Models This report includes information that is output from catastrophe models of Impact Forecasting, LLC (IF). The information from the models is provided by Aon Benfield Services, Inc. (Aon Benfield) under the terms of its license agreements with IF. The results in this report from IF are the products of the exposures modelled, the financial assumptions made concerning deductibles and limits, and the risk models that project the pounds of damage that may be caused by defined catastrophe perils. Aon Benfield recommends that the results from these models in this report not be relied upon in isolation when making decisions that may affect the underwriting appetite, rate adequacy or solvency of the company. The IF models are based on scientific data, mathematical and empirical models, and the experience of engineering, geological and meteorological experts. Calibration of the models using actual loss experience is based on very sparse data, and material inaccuracies in these models are possible. The loss probabilities generated by the models are not predictive of future hurricanes, other windstorms, or earthquakes or other natural catastrophes, but provide estimates of the magnitude of losses that may occur in the event of such natural catastrophes. Aon Benfield makes no warranty about the accuracy of the IF models and has made no attempt to independently verify them. Aon Benfield will not be liable for any special, indirect or consequential damages, including, without limitation, losses or damages arising from or related to any use of or decisions based upon data developed using the models of IF. Additional Limitations of Impact Forecasting, LLC The results listed in this report are based on engineering / scientific analysis and data, information provided by the client, and mathematical and empirical models. The accuracy of the results depends on the uncertainty associated with each of these areas. In particular, as with any model, actual losses may differ from the results of simulations. It is only possible to provide plausible results based on complete and accurate information provided by the client and other reputable data sources. Furthermore, this information may only be used for the business application specified by Impact Forecasting, LLC and for no other purpose. It may not be used to support development of or calibration of a product or service offering that competes with Impact Forecasting, LLC. The information in this report may not be used as a part of or as a source for any insurance rate filing documentation. THIS INFORMATION IS PROVIDED “AS IS” AND IMPACT FORECASTING, LLC HAS NOT MADE AND DOES NOT MAKE ANY WARRANTY OF ANY KIND WHATSOEVER, EXPRESS OR IMPLIED, WITH RESPECT TO THIS REPORT; AND ALL WARRANTIES INCLUDING WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE HEREBY DISCLAIMED BY IMPACT FORECASTING, LLC. IMPACT FORECASTING, LLC WILL NOT BE LIABLE TO ANYONE WITH RESPECT TO ANY DAMAGES, LOSS OR CLAIM WHATSOEVER, NO MATTER HOW OCCASIONED, IN CONNECTION WITH THE PREPARATION OR USE OF THIS REPORT.