MARSHALL SWIFT BOECKH Advisory Board Property Characteristics Correlations

MARSHALL & SWIFT / BOECKH Advisory Board Property Characteristics Correlations Claims vs. Property Characteristics An Analysis of Claims Frequency & Severity Predictors

MARSHALL & SWIFT / BOECKH Advisory Board Objectives » » To correlate property claims to property characteristics » To attempt to “Drill down” to implications such as age and location » To determine if the correlations are reliable enough to be used in premium differentiation To measure the claims dollar implications of differing property attributes » 2

MARSHALL & SWIFT / BOECKH Advisory Board Data Set » Property records • 1. 6 Million MS/B data records • Full “RCT type” property characteristics • 376, 120 with Choice. Point claims activity over zero dollar » Choice. Point records • Peril type and claims amount • Some properties had multiple claims • Claims occurred over a 5 year period » 3

MARSHALL & SWIFT / BOECKH Advisory Board Claims Summary Peril Description Claim Count % of records Average Claim Amount Total Claim Amount % of Amount Water 107, 188 28% $3, 283 $351, 941, 443 28% Wind 72, 349 19% $2, 329 $168, 507, 751 13% All Other Physical Damage 40, 302 11% $2, 173 $87, 593, 307 7% Hail 32, 817 9% $5, 596 $183, 630, 856 14% Theft 26, 980 7% $1, 745 $47, 082, 129 4% Other 23, 725 6% $2, 155 $51, 136, 165 4% Lightning 20, 633 5% $1, 541 $31, 796, 831 3% Mysterious Disappearance 16, 304 4% $1, 740 $28, 367, 518 2% Fire 15, 514 4% $15, 994 $248, 124, 090 20% Extended Coverage Perils 6, 170 2% $2, 031 $12, 530, 391 1% Liability 6, 138 2% $6, 162 $37, 822, 892 3% Vandalism & Malicious Mischief 6, 066 2% $1, 809 $10, 973, 363 1% Watercraft 1, 120 1% $2, 566 $2, 873, 910 0% 814 0% $5, 494 $4, 472, 014 0% 100% $3, 901 $1, 266, 852, 660 Dog Bite TOTAL 376, 120 100% » 4

MARSHALL & SWIFT / BOECKH Advisory Board Characteristics Measured – Property Records Roof Type Siding Type Flooring Type Foundation Type Location Number of Stories Year Built » 5

MARSHALL & SWIFT / BOECKH Advisory Board Computation Avg. Claim $ Claim Frequency Avg. sf Loss Index Premium Index Variance Fav. /<Unfav. > Nationwide $3, 368 23. 51% 1, 825 0% 0% 0 BASEMENT $3, 283 26. 16% 1, 757 108% 96% (12) CRAWL $3, 137 23. 42% 1, 688 93% 92% (0) SLAB $3, 518 20. 66% 1, 906 92% 104% 13 » 6

MARSHALL & SWIFT / BOECKH Advisory Board Findings Property Characteristics – Nationwide foundation $3, 368 age = 100% (Average Claim) » 7

MARSHALL & SWIFT / BOECKH Advisory Board Findings Property Characteristics – Nationwide floor covering $3, 368 ext. walls roof covering = 100% (Average Claim) » 8

MARSHALL & SWIFT / BOECKH Advisory Board Macro Findings » » Two-story homes consistently incur more claims dollars than one-story homes, but premiums correlate relatively closely Age of home has a measurable claims implication and is not correlated with premium differentiation » » » Floor covering also has a distinct claims implication Slab-on-grade is distinctly better than basements Multiple characteristics cause cumulative risk implications » 9

MARSHALL & SWIFT / BOECKH Advisory Board Next Steps » » Expand this trial to a much larger dataset » Increase statistical granularity for multiple factor analysis » Annualize the loss data for inclusion in annual premium assessment » Determine the loss implications of “Underwriting Questions” » Underwriting Analytics Use the larger dataset to generate deeper analyses at regional and state levels » 10

MARSHALL & SWIFT / BOECKH Advisory Board What is Underwriting Analytics? • An analysis of a book of business with policy records • Analysis of trends within the carriers records • Comparison of a given carrier’s records to an industry reference database • Comparison of a given carrier’s records to normative sources (Census, DQ, etc. ) • A set of recommendations based on above analysis and comparisons • Recommendations by home type (property characteristics) • Recommendations by home size or age • Recommendations by geography • Recommendation by source • A wealth of data from your own portfolio of properties • Geocoding for underwriting against any criteria • Identification of areas of under/over insurance • Identification of Agent’s practices Turning Data into Knowledge » 11

MARSHALL & SWIFT / BOECKH Advisory Board ABC Insurance has nearly twice the number of Slab foundation homes compared to the MS/B TES database. If these homes are in fact homes with a full basement, this represents a significant under valuation risk Foundation Type ABC Insurance Data MS/B TES Data Basement 43, 830 66. 13% 337, 953 77. 59% Slab 16, 690 25. 18% 59, 626 13. 69% Crawl 3, 980 6. 00% 19, 087 4. 38% Other Types 1, 780 2. 69% 18, 905 4. 34% 66, 280 100. 00% 435, 571 100. 00% Totals ABC Insurance and MS/B TES data are reasonably consistent with regards to roof type. Roofing Type ABC Insurance Data MS/B TES Data Composition Shingles 62, 990 95. 04% 409, 839 94. 09% Metal/Steel/Tin/Copper 1, 320 1. 99% 4, 039 0. 93% Built-up/Tar & Gravel 710 1. 07% 9, 784 2. 25% Wood Shake or Shingle 500 0. 75% 2, 440 0. 56% 50 0. 08% 1, 250 0. 29% 710 1. 07% 8, 219 1. 89% 66, 280 100. 00% 435, 571 100. 00% Clay or Concrete Tile Other Types Totals Turning Data into Knowledge » 12

MARSHALL & SWIFT / BOECKH Advisory Board ABC Insurance has a significantly higher percentage of homes with brick veneer offset by a much lower percentage of vinyl and aluminum siding exterior walls. Exterior Wall Type Stucco ABC Insurance Data 900 Vinyl 17, 290 Brick Veneer 39, 770 Aluminum 4, 380 Other types 3, 940 Totals 66, 280 MS/B TES Data 1. 36% 26. 09% 60. 00% 6. 61% 5. 94% 100. 00% 8, 133 155, 295 200, 229 42, 721 29, 193 435, 571 1. 87% 35. 65% 45. 97% 9. 81% 6. 70% 100. 00% ABC Insurance has a slightly larger percentage of “One” and “One and a half” story homes than the TES control group Stories ABC Insurance Data MS/B TES Data One 18, 920 28. 55% 112, 372 25. 80% One and a half 15, 330 23. 13% 91, 638 21. 04% Two 28, 300 42. 70% 211, 527 48. 56% 3, 720 5. 61% 20, 009 4. 59% 10 0. 02% 25 0. 01% 66, 280 100. 00% 435, 571 100. 00% Three and above Totals Turning Data into Knowledge » 13

MARSHALL & SWIFT / BOECKH Advisory Board Summary of Outliers by Count and Replacement Cost (in dollars) Test criteria for validation Number of outliers (records violating test criteria, out of 66, 280 total records) Outliers percentage of total records (8, 670 outlier records) Replacement cost of outliers (in millions) Cost per square foot of $70. 00 or less 690 1. 04% $9. 0[1] Total living area greater than 5, 000 sq. ft. 370 0. 56% $27. 8[2] Replacement cost greater than $700, 000 450 0. 68% $37. 7[3] Erroneous data collection and inaccurate input of property characteristics (excluding foundation) 6, 690 10. 09% $147. 1[4] 470 0. 71% $12. 6[5] Excess foundation Turning Data into Knowledge » 14