Homeowners Insurance Coverages

Transcription

1 Homeowners Insurance Coverages An Actuarial Study of the Frequency and Cost of Claims for the State of Michigan by EPIC Consulting, LLC Principal Authors: Michael J. Miller, FCAS, MAAA Klayton N. Southwood, FCAS, MAAA Peer Reviewer: Richard A. Smith, FCAS, MAAA June 2004

2 Foreword EPIC Consulting, LLC was retained to study the frequency and cost of claims data in Michigan for the homeowners insurance coverages. The primary purpose of the study was to determine the degree of variation in losses by geographical area and, to the extent possible, determine the factors which were driving the losses. The study and this report were sponsored by the Insurance Institute of Michigan and the Michigan Insurance Coalition. EPIC had the sole responsibility and the independence to prepare this report and to conduct the study in the way it considered to be actuarially sound. The opinions and conclusions expressed in this report are those of the individuals on EPIC s research team. i

3 About EPIC Consulting, LLC EPIC is a privately-held Illinois limited liability corporation. EPIC provides actuarial services to insurers, insurance regulators, and self-insured business groups. Many of EPIC s clients have been served continuously by its senior consultants for nearly twenty years. The authors of this report are principals of EPIC, Fellows of the Casualty Actuarial Society and Members of the American Academy of Actuaries. Each has been actively involved for their entire career in ratemaking for personal insurance coverages. The authors are available to answer questions about this report by calling (715) , or (309) , or by contact through the EPIC website at ii

4 Reliances and Limitations The authors have relied on the accuracy of the data provided by the ten participating insurers. We did compare the aggregate loss ratios to those published by the National Association of Insurance Commissioners (NAIC) and found consistency in those loss ratios. This comparison increased our confidence in the accuracy of the aggregate databases used for this study. To the extent there are material, undetected errors in the database our conclusions could be significantly impacted. iii

5 Table of Contents Foreword...i About EPIC Consulting, LLC...ii Reliances and Limitations... iii Executive Summary...1 What the Study Is and Is Not...1 Conclusions of the Study...1 Description of the Study and Database...3 Definition of Important Terms and Concepts...4 Homeowners Insurance...4 Perils/Cause of Loss...4 Claim Frequency...5 Average Cost Per Claim...5 Pure Premium...5 Premiums...6 Claim Losses...6 Loss Ratio...6 Expense Ratio...7 Profit...7 Insurance Rates...8 Analysis and Findings...9 Industrywide Data...9 Study Database Statewide...16 Observations Residential Coverage...18 Observations Tenants Coverage...22 Observations Condominiums...26 Observations Mobile Homeowners...29 Geographical Loss Variation...31 Catastrophe Losses...32 Frequency, Severity and Pure Premium Bands...33 Population Density Bands...40 By-Peril Territories...44 Degree of Competition...45 Conclusions...46 TOC

6 Executive Summary What the Study Is and Is Not This is not a study of the adequacy or excessiveness of homeowners insurance rates. The reasonableness of rates is an insurer-by-insurer determination, because each insurer has a unique group of insureds with unique expected losses; has unique methods of operation with unique operating expenses; and has a unique cost of capital and a unique investment portfolio which translate into a unique profit margin provision in its rates. The reasonableness of any rate schedule can only be judged on the losses and expenses which were expected at the time the rates were implemented. The reasonableness of rates cannot be judged by simply looking back at actual results for two or three years. This is a study of the homeowners claim losses that actually occurred in 2000, 2001 and The goal of the study was to explain, to the extent possible, what the losses were, how often the claims occurred, where they occurred, and what perils gave rise to the losses. Conclusions of the Study The nature of homeowners insurance losses is significantly different for each of the four forms of homeowners coverage analyzed in this study (i.e., residential owners, tenants, condominium owners and mobile homeowners). The six perils, or causes of loss, included in the study have significantly different impact on the losses for each of the four coverage forms. The claim losses vary significantly by geographical areas within Michigan and these geographical variations are different for each of the four coverage forms, and for each of the six perils. The variation in losses by coverage form, by peril, and by geographical area means the homeowners losses and loss ratios will likely differ from one insurer to the next. Each insurer has a unique group of policyholders with respect to the portion of the four coverage forms which the insurer provides. Since losses vary geographically and by coverage form, an insurer s total homeowners losses and loss ratio will be uniquely affected by the location of its unique group of policyholders. 1

7 The areas within the state with the highest overall frequency of claims tend to have the highest frequency of claims for every peril of loss. The areas within the state with the highest average cost of claims tend to have the highest costs for every peril of loss. Michigan s homeowners loss ratios reached a high of 109.1% in 2001, ending a sustained increase since The primary driver of the increasing loss ratios in Michigan has been a steady, persistent increase in the average cost of claims. Homeowners premiums have been exceeded by the sum of the incurred losses and expenses in Michigan (i.e., there has been an underwriting loss) each year from 1994 through We found a significant difference in the rates being charged by each of the insurers in the study and we found this rate divergence to exist throughout all areas within Michigan. We interpreted the spread of rates, and the differences in the way the insurers assess risk, to be an indicator of a competitive homeowners insurance market in Michigan. 2

8 Description of the Study and Database Data for the study were provided by the following insurers: Allstate Insurance Company AMCO Insurance Company Auto Club Insurance Association Auto-Owners Insurance Company Citizens Insurance Company of America Farm Bureau Insurance Company of Michigan Home-Owners Insurance Company Ohio Farmers Insurance Company (Westfield Group) Secura Insurance Company State Farm Fire and Casualty Insurance Company The participating insurers write over 60% of Michigan s total homeowners insurance market, and as such, were able to provide a reliable and credible database for this analysis. Each participating insurer provided data for all its Michigan policies in effect during calendar years 2000, 2001 and The losses included in the data call were for all claims that occurred in 2000, 2001, or All losses were valued as of March 31, Data elements provided were written and earned premiums, earned house years, claim counts, and incurred claim amounts valued as of March 31, The data were summarized by zip code, amount of insurance, age of home and size of deductible. Losses were identified by cause of loss (i.e., fire/lightning, wind/hail, water damage/freezing, theft, vandalism/malicious mischief, and liability/medical). The three-year database included a total count of records equivalent to 6.3 million house years (i.e., one home insured for twelve months). 3

9 Definition of Important Terms and Concepts Homeowners Insurance Homeowners insurance, first introduced in the 1950 s, represents a combination of several insurance coverages. Coverage for physical damage to the dwelling, to contents, to outbuildings, for theft, and for the personal liability of the insureds is conveniently packaged into a single insurance policy. Special forms of the homeowners package policy are designed specifically for residential owners, for renters/tenants, for condominium owners, and for mobile/manufactured homeowners. In this report, the term homeowners insurance is used as a general term encompassing all forms of the coverage: residential owners, tenants, condominium owners and mobile/manufactured homeowners. In most instances, the analyses of the loss data are presented separately for each of the four forms of the coverage. Not included in this study are loss data generated from insurance policies commonly referred to as dwelling/fire policies. The dwelling/fire coverage is a forerunner to the homeowners package. While dwelling/fire policies can be constructed so as to be similar to the homeowners package of coverages, dwelling/fire policies often provide coverage more restrictive than the homeowners package. Dwelling/fire policies are not generally considered to be a form of homeowners coverage and, accordingly, are not included in this study. Perils/Cause of Loss The homeowners package of coverages provides insurance for losses arising from fire, lightning, windstorm, hail, water damage, freezing, theft, vandalism, medical, liability and other perils. Each peril is unique in its probability of loss occurrence and the likely cost of a claim when a loss does occur. Because the characteristics of the losses are different for each peril, a full understanding of the nature of homeowners insurance losses can be best achieved if the losses are studied separately by peril. This study provides separate analyses of losses for six groupings of perils: fire/lightning, wind/hail, water/ freezing, theft, vandalism/malicious mischief/all other, and liability/medical. 4

10 Claim Frequency Claim frequency is the ratio of the number of insurance claims to an exposure base. In this study the selected exposure base is one home insured for one year (i.e., earned house year). For example, a claim frequency of.150 means there were 150 claims for every 1000 homes insured. A claim frequency of.150 can also be interpreted as a 15% chance, or likelihood, that a particular insured will incur a claim. Average Cost Per Claim The average cost of a claim is calculated as the total dollars of incurred claim losses divided by the number of claims. This value is also commonly referred to as claim severity. Pure Premium The pure premium is the average cost of claims per insured home. It is calculated as the total dollars of incurred claim losses divided by the number of insured homes. As shown in the following algebraic formula, the pure premium is the product of the claim frequency times the average cost per claim. Let: N C D C/N D/C D/N = number of insured homes = number of claims = dollars of claim losses = claim frequency = average cost per claim (i.e., claim severity) = pure premium. Then: (C/N) x (D/C) = D/N = Pure Premium. Since the pure premium is a combination of the probability of a claim occurring (i.e., claim frequency) and the average cost of a claim once it occurs (i.e., claim severity), it is considered as the best measure of risk for an individual insured, or for a group of insureds. An insured with an expected pure premium of $450 would be considered a higher risk than an insured with an expected pure premium of $300. 5

11 Premiums Premiums are recorded as written premiums at the beginning of the policy term. The written premium is earned pro rata during the policy period. For example, an insurance premium of $500 for an annual policy written on December 31, 2003 will be recorded as $500 of written premium for year By June 30, 2004, the insurer will have earned one-half (i.e., $250) of the premium and will have recorded $250 of earned premium during the first six months of Claim Losses Claim losses include actual payments made to the claimants, plus amounts held in reserve for future payments on claims that have already occurred. Most homeowners claims are quickly paid and settled. But some claims, especially the liability claims, may not be completely paid for five years or more. The loss reserve amounts are estimates of future payments. Loss reserves change as more information becomes known and as partial payments on the claims are made. There are a variety of accounting protocols for recording claim losses. Losses can be accounted for according to the year in which the policy is written. These losses are referred to as policy-year losses. An example would be a policy written in December 2003, with a claim occurring in January In this case the loss would be considered a policy-year 2003 loss because that was the year in which the policy was written. Losses can be accounted for according to the year in which the claim occurred. These losses are referred to as accident-year losses. An example would be a policy written in December 2003 with a claim occurring in January In this case the loss would be considered an accident-year 2004 loss, even though the policy was written in Another common accounting protocol for recording claim losses is referred to as calendar-year losses. In this case any claim payments are recorded to the year in which the payments are made and any changes in the reserve for future payments are recorded to the years in which the reserve changes occur. Loss Ratio Claim losses are often expressed as a ratio to premiums. The total dollars of claim losses divided by the total dollars of premiums is a loss ratio. For example, a loss ratio of.70 (i.e., 70%) means that 70% of the premium dollars were needed to fund the payment of claim losses. 6

12 For most rate analyses, the accepted loss ratio to analyze is an incurred loss (usually on an accidentyear basis) divided by earned premiums. A ratio of incurred losses to written premiums is hardly ever used because such a ratio will reflect a mismatch of the claim losses to the premiums used to fund the losses. An example may help clarify the mismatch problem. All policies written in 2003 will have the written premium recorded to 2003, but some claims on those policies may not be incurred until All policies written in 2004 will have the written premium recorded to 2004, but some claims on those policies may not be incurred until If one were to relate accident-year 2004 incurred losses to 2004 written premium, there would be two sources of mismatch. There would be accident-year 2004 losses in the numerator that properly tie to 2003 written premium, not to the 2004 written premiums in the denominator. There would also be 2004 written premiums in the denominator that tie to accidentyear 2005 losses, not to the 2004 losses in the numerator. Expense Ratio Expenses are often expressed as a ratio to premiums. Claim settlement expenses (e.g., legal or claim investigation expenses) related to premiums is commonly referred to as a loss adjustment expense ratio. Operational/underwriting expenses related to premiums are commonly referred to as an underwriting expense ratio. Profit There are three types of profit ratios commonly used in insurance rate analyses. Each profit measure relates to a specific portion of an insurer s overall return. The underwriting profit represents that portion of the premiums which remain after funding the claim losses and expenses. For example, if claim losses in a particular year constituted 70% of premium and all expenses totaled 25% of premium, the insurer would have experienced a 5% underwriting profit. If claim losses constituted 80% of premium and all expenses totaled 25% of premium, the insurer would have experienced a 5% underwriting loss. In addition to the underwriting profit/loss, an insurer also earns investment income while the premiums are being held to pay future claims and expenses. The investment income from premiums, 7

13 plus the underwriting profit/loss, is commonly referred to as the operating profit. The operating profit represents the return generated by the insurance operations. Insurers are required to commit sufficient capital to guarantee the availability of funds to pay claims. These capital funds also generate investment income. The investment income from the capital funds, plus the insurance operating profit, represents the total return to the insurer. The total return must be sufficient to compensate the insurer for placing its capital at risk in the insurance operation. Insurance Rates Insurance rates are established so as to be sufficient to fund the claim losses and the expenses expected to be incurred during the time the insurance policy is in force. Expected future claim losses and expenses cannot be determined by merely looking at what actually happened in the past. For example, the fact that a house has not burned in the past does not mean the chances of a fire in the future are zero. Or, just because there were several years of relatively mild weather does not reduce the likelihood of future claims due to severe weather. Determining these future probabilities is an actuarial process. The future likelihood of losses, as to both the expected frequency and the expected severity of claims, is the only basis for determining the reasonableness of insurance rates and the only basis for determining if rates meet the Michigan rate standards of neither excessive, inadequate, nor unfairly discriminatory. In addition to the losses and expenses, insurance rates provide for a reasonable margin for underwriting profit. The underwriting profit provision is determined so that the insurer s total expected return is consistent with the degree of risk to which the insurer s capital is exposed. Reasonable profit provisions, expense provisions and provisions for expected claim losses vary significantly among insurers, thereby making insurance ratemaking an insurer-by-insurer calculation. 8

14 Analysis and Findings Industrywide Data The National Association of Insurance Commissioners (NAIC) annually publishes loss, expense, and profit data for each state and for each line of insurance. The NAIC data for Michigan are derived from all insurers doing business in Michigan. The NAIC loss ratios are calendar-year incurred losses divided by earned premiums. The premium and loss data are combined for all forms of homeowners coverage. Exhibit I shows that Michigan s homeowners loss ratio reached a high of 109.1% in 2001, ending a sustained increase following the relatively low loss ratio of 58.4% in Homeowners Michigan Loss Ratio Exhibit I 115.0% 110.0% 105.0% 100.0% 95.0% 90.0% Loss Ratio 85.0% 80.0% 75.0% 70.0% 65.0% 60.0% 55.0% 50.0% HO MI Loss Ratio Calendar Year 9 HO MI Loss Ratio Avg

15 Exhibit II shows the NAIC countrywide homeowners loss ratios. The countrywide loss ratios do not show the same dramatic increase in loss ratios during the period 1993 through 2001, as was experienced in Michigan. Homeowners Countrywide Loss Ratio Exhibt II 130.0% 120.0% 110.0% 100.0% Loss Ratio 90.0% 80.0% 70.0% 60.0% 50.0% HO CW Loss Ratio Calendar Year HO CW Loss Ratio Avg Homeowners losses can fluctuate widely from year-to-year because of the occurrence of natural catastrophes, such as wind losses or losses arising from freezing. In years when there are relatively minor catastrophe losses, the homeowners loss ratio may be unusually low. In years with serious catastrophe losses, the loss ratio will spike to unusually high levels. The sharp spike in the 1992 countrywide loss ratio (see Exhibit II) is an example of the effect of catastrophe wind losses. When Florida, with the Hurricane Andrew losses, is removed from the 1992 data, the countrywide loss ratio drops to 75.1%. 10

16 While we did identify some catastrophe losses in the Michigan database underlying this study, the existence of catastrophe losses does not explain the sustained increase in Michigan s loss ratio from 1993 through Exhibits III (a), (b), and (c) show the pure premiums, the average claim severity, and the claim frequencies for the residential form of homeowners coverage in Michigan. These data are generated through the Fast-Track Reporting System and represent data from a significant portion of the homeowners insurance business written in Michigan. The loss data for the tenants and the condominium owner forms of the homeowners coverage are also published through the Fast-Track system. Those graphs are not shown in this report merely because we desired to reduce the number of charts and graphs as much as possible. The loss data for the tenants and condominium owners coverages do exhibit patterns similar to the patterns shown in Exhibit III for the residential coverage. Those patterns show long-term decreases in claim frequencies and long-term increases in claim severities. Exhibit III(a) shows that the Michigan pure premiums (i.e., average loss per insured) have been generally increasing over the long-term. There are short periods of decreases in the losses, but the general trend has been upward. An increasing trend in losses will eventually translate into an increase in homeowners insurance rates. The amount of the rate increase will vary from insurer to insurer, depending on each insurer s expected loss and expense experience. Exhibit III(b) shows that there has been a long-term upward trend in Michigan in the average cost of claims. Exhibit III(c) shows a long-term decrease in the frequency of homeowners claims in Michigan. Overall, homeowners losses have been increasing because the decline in the frequency of claims has not been sufficient to offset the increase in the average cost of claims. The increase in the average cost of claims has been the primary reason that homeowners losses have been increasing in Michigan. 11

17 Residential Owners Pure Premium Exhibit III(a) $ $ Pure Premium $ $ $ $ $ Period (Year Ending Quarter) Source: ISO/NAII Fast Track Data Michigan Countrywide Exhibit III(b) Residential Owners Average Claim Severity $6,000 $5,000 Average Claim Severity $4,000 $3,000 $2,000 $1,000 $ Period (Year Ending Quarter) Source: ISO/NAII Fast Track Data Michigan Countrywide 12

18 Residential Owners Claim Frequency per 100 Exposures Exhibit III(c) Paid Claim Frequency Period (Year Ending Quarter) Source: ISO/NAII Fast Track Data Michigan Countrywide Exhibit IV presents the underwriting profit ratios published by the NAIC. As previously described, the underwriting profit is equal to earned premiums minus incurred losses and minus incurred expenses. In 1993, the Michigan homeowners earned premium exceeded the total of incurred losses and expenses by 4.3% (i.e., underwriting profit ratio of 4.3%). In each year since 1993, the Michigan homeowners earned premiums have been insufficient to fund the incurred losses and expenses. Underwriting profit/loss ratios have ranged from -4.1% to -50.9% during the period 1994 through

19 Homeowners Michigan Underwriting Profit Exhibit IV 10.0% 5.0% 0.0% -5.0% -10.0% Underwriting Profit -15.0% -20.0% -25.0% -30.0% -35.0% -40.0% -45.0% -50.0% -55.0% Calendar Year HO MI Underwriting Profit HO MI Underwriting Profit Avg Exhibit V presents the countrywide homeowners underwriting profit ratios published by the NAIC. While the countrywide data show only one year of underwriting profit since 1993 (3.4% in 1997), the average countrywide underwriting loss has not been as unfavorable as was experienced in Michigan. During the period 1980 through 2002, Michigan has experienced an average underwriting loss ratio of -14.0%, while the countrywide average underwriting loss ratio was -10.0% for the same time period. 14

20 Homeowners Countrywide Underwriting Profit Exhibit V Underwriting Profit 10.0% 5.0% 0.0% -5.0% -10.0% -15.0% -20.0% -25.0% -30.0% -35.0% -40.0% -45.0% -50.0% -55.0% -60.0% -65.0% -70.0% Calendar Year HO CW Underwriting Profit HO CW Underwriting Profit Avg We acknowledge that underwriting profit ratios represent an incomplete picture of an insurer s returns. Obviously missing in the underwriting profit calculation is the income from investments. According to the NAIC, total returns (i.e., underwriting profit plus investment income on policyholder supplied funds and capital/surplus) in Michigan have been negative in every year during the period 1996 through Even in 2002 the total homeowners return in Michigan was only 3.3%. Such a return is less than an insurer could have realized had it invested its capital in some business other than homeowners insurance. The countrywide total returns for homeowners were not favorable compared to alternate investment opportunities, but at least the countrywide total returns were positive in all but two years during the period 1996 through Michigan s homeowners insurance has underperformed the countrywide average results in every year, except one, since

21 Study Database Statewide Tables 1 through 5 summarize the premium and loss data in the study s database for each form of the homeowners coverage, and for each year. Table 1 Residential Earned Incurred Average House Earned Incurred Incurred Loss Claim Claim Pure Year Years Premiums Losses Claims Ratio Frequency Severity Premium ,704,717 $744,919,588 $806,237, , % $3,314 $ ,788, ,910, ,361, , % , ,835, ,988, ,889, , % , Total 5,329,107 $2,523,818,587 $2,201,487, , % $3,358 $413 Table 2 Tenants Earned Incurred Average House Earned Incurred Incurred Loss Claim Claim Pure Year Years Premiums Losses Claims Ratio Frequency Severity Premium ,868 $30,491,935 $16,412,930 7, % $2,081 $ ,395 32,148,717 19,379,496 7, % , ,454 33,512,230 16,121,521 6, % , Total 574,717 $96,152,882 $51,913,947 21, % $2,375 $90 Table 3 Condominium Owners Earned Incurred Average House Earned Incurred Incurred Loss Claim Claim Pure Year Years Premiums Losses Claims Ratio Frequency Severity Premium ,954 $17,687,643 $13,835,149 7, % $1,938 $ ,152 19,898,281 14,352,905 6, % , ,440 23,038,335 11,707,723 5, % , Total 304,546 $60,624,259 $39,895,777 18, % $2,106 $131 16

22 Table 4 Mobile Homeowners Earned Incurred Average House Earned Incurred Incurred Loss Claim Claim Pure Year Years Premiums Losses Claims Ratio Frequency Severity Premium ,650 $18,931,302 $14,126,437 6, % $2,268 $ ,508 20,005,574 16,861,255 7, % , ,016 22,166,018 12,736,529 5, % , Total 178,174 $61,102,894 $43,724,221 18, % $2,320 $245 Table 5 All Forms of Homeowners Earned Incurred Average House Earned Incurred Incurred Loss Claim Claim Pure Year Years Premiums Losses Claims Ratio Frequency Severity Premium ,049,189 $812,030,468 $850,611, , % $3,216 $ ,144, ,962, ,955, , % , ,192,723 1,033,705, ,454, , % , Total 6,386,544 $2,741,698,621 $2,337,021, , % $3,267 $366 The total loss ratios in the study database are compared in Table 6 to the homeowners loss ratios published by the NAIC. Table 6 Comparison of Loss Ratios Study s Database Year NAIC Actual Adjusted % 104.8% 105.8% % 91.6% 93.5% % 64.4% 74.0% Average 87.5% 85.2% 89.8% 17

23 The accident-year incurred losses in the study s database were valued as of March 31, Not all the claims in the database have been completely settled and almost certainly there will be adjustments in the reserved amounts for those claims, thereby causing adjustments to the accident-year loss ratios shown above for the study s database. How those loss reserves and loss ratios change will vary significantly from insurer to insurer. Since the study s database did not provide sufficient data for a reliable estimate of future loss reserve changes, we relied on publicly-available loss reserve development data to develop the Adjusted loss ratios shown above. These estimates are approximations of the loss ratios we expect when all the claims for accident-years are completely settled. Because much of this study deals with a comparison of losses (e.g., one peril to another, or one geographic area to another), we concluded there was no need to adjust all the losses to their estimated, ultimate levels. We have provided adjusted loss ratios here only because we are making comparisons to the calendar-year loss ratios published by the NAIC. There is a great deal of similarity in the 2002 loss ratios and also in the average loss ratios for the three-year period. This similarity is to be expected because the study s database consists of a representative sample of the entire industry which makes up the NAIC database. At first glance the loss ratios for 2000 and 2001 appear to be inconsistent between the two databases. However, the NAIC incurred losses are recorded on a calendar-year basis and the incurred losses in the study database are recorded to the year in which the loss occurred (i.e., accident-year). When the difference in timing in accounting for the losses is considered, we find the loss data in the study s database to be consistent with the industrywide loss data published by the NAIC. We are confident that the sample of loss data represented by the study s database is representative of the entire industry for the years Observations Residential Coverage The total homeowners coverage loss ratios (see Table 5) for years 2000 and 2001 were especially unfavorable due primarily to the unfavorable results for the residential form of the homeowners coverage (see Table 1). The unfavorable loss ratios for the residential coverages appear to be due primarily to unusually high claim frequencies in 2000 and

24 As shown in the following Exhibits VI and VII, the water/freezing claims and the fire claims constitute the highest proportion of claims. The proportion of claims occurring due to water/freezing was unusually high in 2000, as were fire claims. The combination of water/freezing and fire claims contributed to the unfavorable 108.2% loss ratio for the residential coverage in The high loss ratio in 2001 appears to be primarily due to a relatively high proportion of water/freezing claims. 19

25 0.050 Michigan Residential Owner Policies Frequency by Cause of Loss Exhibit VII Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Claim Frequency Year Exhibit VIII presents the average claim severities for each peril, for the residential form of coverage. There was a sharp increase in the average cost of fire claims during the period 2001 and However, as shown on Exhibit VII, the increase in the average cost of fire claims in 2001 and 2002 was mitigated by a decrease in the frequency of fire claims. 20

26 7,000 Michigan Residential Owner Policies Severity by Cause of Loss Exhibit VIII 6,000 Severity 5,000 4,000 3,000 Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments 2,000 1, Year Exhibits IX and X summarize the pure premiums by peril for the residential coverage. The greatest proportion of residential homeowners losses in Michigan have been due to water/freezing and fire claims, and those losses are primarily a function of the frequency of these types of claims. 21

27 Michigan Residential Owner Policies Pure Premium by Cause of Loss Exhibit X Pure Premium Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Year Observations Tenants Coverage Exhibits XI and XII present the proportion of claims that are generated by the various perils for the tenants form of the homeowners coverage. Theft claims and vandalism claims are the most frequent types of claims for the tenants coverage. We found no remarkable changes in the claim frequencies by peril during the period

28 0.018 Michigan Tenant Policies Frequency by Cause of Loss Exhibit XII Claim Frequency Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Year As was true for the residential coverage, we found a significant increase in the average cost of fire claims for the tenants coverage in 2001 and

29 7,000 Michigan Tenant Policies Severity by Cause of Loss Exhibit XIII 6,000 Severity 5,000 4,000 3,000 Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments 2,000 1, Year Exhibits XIV and XV summarize the pure premiums by peril for the tenants coverage. We find a much higher proportion of losses for tenants arising from vandalism and theft than is true for the residential form of the homeowners coverage and these losses are primarily a function of the frequency of these types of claims. 24

30 30.00 Michigan Tenant Policies Pure Premium by Cause of Loss Exhibit XV Pure Premium Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Year 25

31 Observations Condominiums Exhibits XVI and XVII show that water/freezing constitute a significant proportion of the total claims for condominium owners. There was a decrease in the frequency of water/freezing claims in 2001 and 2002 which contributed to the improvement in the loss ratios for 2001 and Michigan Condominium Owner Policies Frequency by Cause of Loss Exhibit XVII Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Claim Frequency Year 26

32 Exhibit XVIII shows an increase in the cost of fire claims in 2001 and 2002, which is very similar to the increases observed for fire claims for all other forms of the homeowners coverage. 5,000 Michigan Condominium Owner Policies Severity by Cause of Loss Exhibit XVIII 4,500 Severity 4,000 3,500 3,000 2,500 2,000 Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments 1,500 1, Year Exhibits XIX and XX summarize the pure premiums by peril for the condominium owners. Water/ freezing losses represent 49.3% of the total losses for the condominium owner coverage. In comparison, water/freezing claims are 24.0% of the residential coverage losses. 27

33 90.00 Michigan Condominium Owner Policies Pure Premium by Cause of Loss Exhibit XX Pure Premium Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Year 28

34 Observations Mobile Homeowners Exhibits XXI XXV summarize the claim frequencies, claim severities and the pure premiums by cause of loss for the mobile homeowners form of the coverage. The leading cause of losses for mobilehomes is fire, followed closely by liability/medical losses. Liability/medical losses constitute 25.1% of the total losses. In comparison, liability losses are only 10.5% of the total for residential, 5.5% for tenants, and 13.2% for condominium owners Michigan Mobile Homeowner Policies Frequency by Cause of Loss Fire & Lightning Wind & Hail Water & Freezing Theft Exhibit XXII VMM & All Other Physical Damage Liability & Medical Payments Claim Frequency Year 29

35 7,000 Michigan Mobile Homeowner Policies Severity by Cause of Loss Exhibit XXIII 6,000 Severity 5,000 4,000 3,000 Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments 2,000 1, Year 30

36 Michigan Mobile Homeowner Policies Pure Premium by Cause of Loss Exhibit XXV Fire & Lightning Wind & Hail Water & Freezing Theft VMM & All Other Physical Damage Liability & Medical Payments Pure Premium Year Geographical Loss Variation Homeowners losses vary geographically. The geographical variation is different for each form of the coverage (i.e., residential, tenants, condominium owners and mobile homeowners) and it is different for each cause of loss (i.e., peril). How to best analyze this geographical variation in losses and present the results, represented a formidable challenge to this study. There exist no standard, or benchmark, territory definitions in Michigan which insurers use to assess the risk and price the homeowners insurance product. We analyzed the rating territories used by eight leading homeowners insurers (representing 75% of the total homeowners market) and found a great difference in how these insurers define their respective rating territories. This significant divergence in how rating territories are defined, and how insurers assess geographic risk, is not surprising in this highly competitive marketplace. Lacking a commonly used means of geographically subdividing Michigan s loss data, we proceeded with this analysis using five different types of geographical subdivisions of the data. 31

37 1) We ranked each zip code from high to low based on its claim frequency and aggregated the loss data into ten claim frequency bands. Frequency bands were created separately for each form of the homeowners coverage. Maps of the frequency bands are presented in Appendix A. 2) We ranked each zip code from high to low based on its claim severity and aggregated the loss data into ten claim severity bands. Severity bands were created separately for each form of the homeowners coverage. Maps of the severity bands are presented in Appendix B. 3) We ranked each zip code from high to low based on its pure premium and aggregated the loss data into ten pure premium bands. Pure premium bands were created separately for each form of the homeowners coverage. Maps of the pure premium bands are presented in Appendix C. 4) We aggregated the data by cause of loss for each form of the coverage into ten groups based on the population density of each zip code. A map of the population density bands is presented in Appendix D. 5) Utilizing the residential loss data for each peril, we grouped zip codes that were most similar in the risk of loss for each peril. The maps of the territories for each of the six perils are presented in Appendices E through J. Catastrophe Losses Homeowners insurance is susceptible to the occurrence of a cluster of weather-related claims or to single, jumbo claims arising from events such as fires, severe weather, or liability awards. These catastrophe and jumbo claims can distort the losses in the year of occurrence. Actuaries ordinarily remove these losses from the historical data and substitute long-term averages for these types of claims. The way the data were provided for this study did not permit the identification of every catastrophe or jumbo claim which occurred in the experience period. However, we were able to identify at least some catastrophe losses. Before subdividing the database geographically, we inspected the losses in every zip code. 32

38 For the residential coverages we removed a total of 5,663 claims (5,610 wind claims and 35 water/freezing claims) from the database. The dollar losses removed for the geographical analysis totaled to $182,738,549. For the tenants coverage we removed no loss counts, but we did limit the severity of some claims, thereby removing $622,651 of losses. For the condominium coverage we removed no loss counts, but we did limit the severity of some claims, thereby removing a total of $196,379. For the mobile homeowners coverage we eliminated 122 wind claims and $857,363 of losses. In our judgment, the removal of these catastrophe claims, both claim counts and dollar amounts, permitted a more valid comparison of the losses across the geographical bands without the potential for distortions in the data which can arise when a period as short as three years is being analyzed. Frequency, Severity and Pure Premium Bands The study s entire database is summarized in Exhibits XXVI through XXIX for each form of coverage on the basis of three geographical subdivisions of the data: frequency bands, severity bands and pure premium bands. Although the frequency, severity and pure premium bands were developed with data combined for all perils, we found a remarkable consistency for each peril analyzed separately. With the exception of liability/medical claims, we found the claim frequencies for each peril to be decreasing from the highest to lowest all-peril frequency bands. We found the claim severities for each peril to be generally decreasing from the highest to lowest all-peril severity bands. We found the pure premiums for each peril to be generally decreasing from the highest to lowest all-peril pure premium bands. In short, the highest frequency, severity and pure premium bands tended to be the highest for each peril. The lowest frequency, severity and pure premium bands tended to be the lowest for each peril. With respect to the residential coverages, we observe the following: a) loss ratios decrease from the highest frequency and severity band (i.e., Band 1) to the lowest band (i.e., Band 10); 33

39 b) the claim severities are relatively constant across the frequency bands, indicating a degree of independence between claim frequencies and claim severities; c) the claim frequencies are relatively constant across the severity bands, indicating a degree of independence between claim frequencies and claim severities; d) fire losses and liability losses are proportionately less in the higher frequency bands; and e) theft and vandalism losses are proportionately higher in the highest frequency band. With respect to the tenants coverage, we observe the following: a) there is a greater variation in loss ratios across the frequency and severity bands than was evident for the residential coverage; b) high frequency bands also tend to have high average claim severities, and vice versa; c) fire losses and liability losses are proportionately less in frequency Band 1; d) fire losses are proportionately higher in the severity Band 1; and e) theft losses are proportionately higher in the highest frequency bands. With respect to the condominium owners coverage, we observe the following. a) there is greater variation in loss ratios across the frequency and severity bands than is evident for the residential coverage; b) water and liability losses tend to account for proportionately more of the total losses in the highest frequency bands; and c) vandalism accounts for proportionately more of the losses in the highest severity band than is true for any other band. 34

40 With respect to the mobile homeowners form of coverage, we observe the following; a) there is a significant variation in loss ratios across the frequency and severity bands; b) the claim severities are relatively constant across the frequency bands, indicating a degree of independence between claim frequencies and claim severities; c) the claim frequencies are relatively constant across the severity bands, indicating a degree of independence between claim frequencies and claim severities; d) fires are a major cause of loss in the lowest frequency band; e) water/freezing is a major cause of loss in the highest frequency band; and f) fires are a major cause of loss in the highest severity band. 35

41 Exhibit XXVI DATA BY FREQUENCY, SEVERITY, AND PURE PREMIUM BAND Residential Owner Policies PREMIUM & CLAIM DATA BY FREQUENCY BAND Distribution Of Losses by Peril VMM & All Other Liability & Frequency Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 325,009, ,375, % , % 15.7% 22.1% 7.1% 23.5% 7.0% 2 263,225, ,978, % , % 13.5% 25.8% 4.5% 17.5% 11.1% 3 265,892, ,304, % , % 13.1% 22.8% 4.5% 17.0% 11.5% 4 254,972, ,241, % , % 9.8% 25.8% 4.3% 16.7% 11.2% 5 254,266, ,688, % , % 9.0% 25.6% 4.4% 16.7% 10.3% 6 234,791, ,391, % , % 8.8% 25.5% 5.3% 18.2% 8.6% 7 234,480, ,526, % , % 8.9% 25.8% 4.4% 17.2% 10.2% 8 231,850, ,058, % , % 8.5% 24.8% 4.4% 16.9% 10.4% 9 229,085, ,473, % , % 8.2% 23.7% 3.5% 16.3% 12.7% ,242, ,709, % , % 8.9% 18.0% 2.5% 13.8% 17.0% State 2,523,818,587 2,018,749, % , % 11.3% 24.0% 4.8% 18.1% 10.5% PREMIUM & CLAIM DATA BY SEVERITY BAND Distribution Of Losses by Peril VMM & All Other Liability & Severity Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 303,924, ,086, % , % 15.1% 21.2% 3.3% 18.5% 8.9% 2 274,776, ,328, % , % 10.6% 25.0% 3.9% 16.1% 10.0% 3 288,818, ,448, % , % 9.2% 22.5% 6.7% 22.7% 8.0% 4 264,499, ,827, % , % 10.2% 23.3% 6.2% 19.5% 10.0% 5 252,118, ,070, % , % 10.5% 24.7% 5.1% 18.5% 11.0% 6 230,486, ,025, % , % 11.6% 25.2% 4.4% 16.2% 10.8% 7 227,308, ,369, % , % 10.7% 25.9% 5.0% 15.5% 10.0% 8 230,028, ,479, % , % 10.5% 25.4% 4.9% 18.5% 12.0% 9 222,128, ,045, % , % 10.8% 25.4% 4.4% 15.9% 14.4% ,729, ,066, % , % 11.6% 25.0% 4.9% 16.9% 13.9% State 2,523,818,587 2,018,749, % , % 11.3% 24.0% 4.8% 18.1% 10.5% PREMIUM & CLAIM DATA BY PURE PREMIUM BAND Distribution Of Losses by Peril VMM & Pure All Other Liability & Premium Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 346,769, ,180, % , % 15.2% 21.7% 6.1% 23.1% 7.1% 2 284,783, ,217, % , % 11.9% 23.4% 5.1% 19.1% 8.9% 3 252,821, ,037, % , % 12.4% 23.5% 3.8% 16.8% 12.1% 4 261,882, ,642, % , % 9.5% 27.1% 4.4% 15.1% 10.9% 5 251,028, ,625, % , % 9.5% 24.7% 4.7% 17.4% 11.0% 6 231,271, ,423, % , % 9.3% 24.8% 4.9% 16.3% 11.4% 7 230,811, ,468, % , % 9.2% 24.3% 5.0% 16.1% 10.4% 8 224,571, ,468, % , % 9.1% 26.0% 4.4% 16.5% 10.3% 9 218,525, ,601, % , % 8.7% 24.6% 3.9% 17.2% 13.3% ,352, ,083, % , % 11.5% 22.9% 3.7% 15.0% 17.5% State 2,523,818,587 2,018,749, % , % 11.3% 24.0% 4.8% 18.1% 10.5% 36

42 Exhibit XXVII DATA BY FREQUENCY, SEVERITY, AND PURE PREMIUM BAND Tenant Policies PREMIUM & CLAIM DATA BY FREQUENCY BAND Distribution Of Losses by Peril VMM & All Other Liability & Frequency Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 11,983,768 15,239, % , % 1.3% 17.8% 33.2% 24.8% 2.4% 2 11,152,744 6,776, % , % 0.9% 16.9% 32.0% 29.0% 4.7% 3 9,479,223 5,788, % , % 1.4% 16.4% 22.6% 28.1% 5.4% 4 9,326,402 4,885, % , % 1.8% 20.5% 26.1% 21.0% 4.4% 5 8,991,292 3,752, % , % 2.6% 18.2% 28.5% 18.5% 8.7% 6 8,889,047 3,894, % , % 1.5% 14.1% 25.1% 22.0% 8.9% 7 8,986,329 3,504, % , % 1.1% 16.2% 25.3% 21.2% 8.3% 8 9,209,807 3,063, % , % 1.4% 15.3% 26.6% 16.9% 10.5% 9 8,849,693 2,580, % , % 1.3% 25.7% 26.6% 19.0% 7.3% 10 9,284,578 1,805, % , % 1.8% 22.2% 22.5% 24.8% 6.3% State 96,152,882 51,291, % , % 1.4% 17.8% 28.6% 23.7% 5.5% PREMIUM & CLAIM DATA BY SEVERITY BAND Distribution Of Losses VMM & All Other Liability & Severity Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 11,081,077 16,009, % , % 0.8% 17.4% 22.8% 26.2% 3.2% 2 10,928,644 8,005, % , % 1.0% 17.0% 28.4% 26.0% 5.1% 3 10,233,233 5,872, % , % 1.1% 16.0% 30.4% 22.1% 6.4% 4 9,253,080 4,618, % , % 1.1% 19.5% 26.3% 24.9% 9.0% 5 9,740,378 3,938, % , % 0.9% 20.5% 34.5% 21.0% 4.4% 6 9,243,431 4,072, % , % 1.6% 20.0% 35.1% 19.6% 6.2% 7 9,067,298 3,063, % , % 1.9% 23.5% 35.3% 18.7% 5.8% 8 8,935,545 2,504, % , % 3.2% 16.5% 31.1% 18.7% 8.1% 9 8,840,022 2,044, % , % 4.1% 12.2% 38.2% 21.2% 7.6% 10 8,830,174 1,162, % % 7.1% 13.5% 25.2% 27.6% 10.2% State 96,152,882 51,291, % , % 1.4% 17.8% 28.6% 23.7% 5.5% PREMIUM & CLAIM DATA BY PURE PREMIUM BAND Distribution Of Losses by Peril VMM & Pure All Other Liability & Premium Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 11,799,600 17,896, % , % 0.9% 18.0% 26.6% 24.3% 3.0% 2 11,011,043 7,744, % , % 1.3% 14.7% 27.5% 31.1% 4.6% 3 10,147,875 6,032, % , % 1.2% 17.0% 28.3% 23.3% 8.8% 4 9,491,413 4,762, % , % 1.6% 18.4% 30.2% 22.3% 5.9% 5 9,016,020 3,863, % , % 1.8% 17.8% 27.4% 22.2% 7.9% 6 8,294,113 2,944, % , % 1.8% 18.7% 33.6% 17.8% 7.8% 7 8,966,622 2,788, % , % 1.9% 20.0% 32.9% 13.9% 6.2% 8 8,901,306 2,327, % , % 2.0% 24.7% 30.4% 21.5% 4.9% 9 9,068,463 1,881, % , % 2.0% 18.0% 31.6% 21.9% 9.6% 10 9,456,427 1,050, % % 6.2% 16.7% 33.0% 23.0% 9.3% State 96,152,882 51,291, % , % 1.4% 17.8% 28.6% 23.7% 5.5% 37

43 Exhibit XXVIII DATA BY FREQUENCY, SEVERITY, AND PURE PREMIUM BAND Condominium Owner Policies PREMIUM & CLAIM DATA BY FREQUENCY BAND Distribution Of Losses by Peril VMM & All Other Liability & Frequency Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 7,896,605 8,080, % , % 1.4% 53.1% 5.1% 15.9% 16.2% 2 6,873,960 6,772, % , % 1.6% 54.1% 7.7% 17.1% 10.5% 3 6,820,988 4,849, % , % 3.0% 51.4% 9.2% 14.5% 11.5% 4 6,216,145 3,824, % , % 1.1% 53.0% 7.9% 14.0% 9.7% 5 5,898,614 3,410, % , % 2.2% 49.2% 9.0% 14.8% 10.9% 6 5,393,886 3,343, % , % 2.4% 47.8% 8.6% 15.7% 10.0% 7 5,548,232 2,700, % , % 1.3% 41.9% 10.4% 21.2% 10.6% 8 5,403,109 3,011, % , % 1.2% 45.1% 5.9% 16.2% 14.8% 9 5,612,926 2,415, % , % 1.7% 33.2% 5.0% 33.8% 19.8% 10 4,959,793 1,290, % , % 1.7% 39.8% 4.4% 16.1% 28.1% State 60,624,259 39,699, % , % 1.8% 49.3% 7.3% 17.1% 13.2% PREMIUM & CLAIM DATA BY SEVERITY BAND Distribution Of Losses by Peril VMM & All Other Liability & Severity Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 7,238,746 8,939, % , % 1.2% 46.6% 5.5% 22.3% 13.8% 2 6,855,323 6,851, % , % 1.2% 50.1% 7.4% 13.3% 12.8% 3 5,567,588 4,128, % , % 1.5% 53.2% 7.3% 17.9% 7.3% 4 6,525,442 4,422, % , % 1.7% 51.8% 8.0% 14.6% 12.7% 5 6,674,092 4,425, % , % 1.9% 53.1% 7.5% 15.7% 10.4% 6 5,405,827 2,590, % , % 2.9% 52.0% 9.1% 14.7% 14.3% 7 5,655,281 2,702, % , % 2.4% 44.1% 9.9% 16.0% 15.3% 8 6,400,653 2,509, % , % 2.4% 50.9% 6.3% 18.3% 14.2% 9 5,768,010 2,210, % , % 2.4% 44.0% 7.7% 17.2% 19.6% 10 4,533, , % % 4.1% 36.9% 9.4% 18.0% 23.4% State 60,624,259 39,699, % , % 1.8% 49.3% 7.3% 17.1% 13.2% PREMIUM & CLAIM DATA BY PURE PREMIUM BAND Distribution Of Losses by Peril VMM & Pure All Other Liability & Premium Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 7,751,875 10,545, % , % 1.4% 49.4% 5.2% 17.8% 13.2% 2 7,041,456 6,046, % , % 1.0% 50.7% 8.3% 20.8% 11.5% 3 6,575,685 5,090, % , % 2.7% 50.7% 8.5% 11.8% 14.1% 4 5,711,730 4,036, % , % 1.1% 51.3% 7.6% 17.7% 10.3% 5 5,324,729 3,483, % , % 1.4% 53.8% 7.3% 15.2% 7.7% 6 5,637,338 2,979, % , % 1.9% 46.6% 7.5% 16.1% 16.1% 7 5,903,130 2,634, % , % 3.3% 49.8% 10.4% 17.3% 12.0% 8 5,585,250 2,218, % , % 2.1% 44.4% 8.2% 20.8% 12.4% 9 5,946,800 1,735, % , % 2.0% 44.2% 6.8% 15.2% 23.1% 10 5,146, , % % 4.2% 33.5% 7.4% 17.0% 27.9% State 60,624,259 39,699, % , % 1.8% 49.3% 7.3% 17.1% 13.2% 38

44 Exhibit XXIX DATA BY FREQUENCY, SEVERITY, AND PURE PREMIUM BAND Mobile Homeowner Policies PREMIUM & CLAIM DATA BY FREQUENCY BAND Distribution Of Losses by Peril VMM & All Other Liability & Frequency Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 6,355,445 7,872, % , % 14.9% 23.7% 3.6% 11.5% 21.0% 2 6,268,852 5,748, % , % 10.8% 28.7% 3.2% 9.9% 23.9% 3 6,338,243 5,793, % , % 12.1% 22.1% 3.6% 11.1% 23.2% 4 5,852,596 3,912, % , % 10.0% 22.7% 3.1% 11.8% 32.1% 5 5,979,027 4,233, % , % 9.2% 22.2% 3.4% 12.5% 22.4% 6 5,834,946 3,983, % , % 6.5% 21.4% 2.9% 14.4% 27.1% 7 6,208,361 3,227, % , % 10.7% 21.8% 2.9% 16.4% 26.7% 8 6,303,792 3,260, % , % 8.2% 26.6% 2.9% 11.8% 27.5% 9 5,938,055 2,798, % , % 7.6% 16.3% 3.3% 14.5% 31.1% 10 5,998,178 2,036, % , % 7.9% 10.0% 2.7% 12.3% 23.6% State 61,077,496 42,866, % , % 10.5% 22.6% 3.3% 12.3% 25.1% PREMIUM & CLAIM DATA BY SEVERITY BAND Distribution Of Losses by Peril VMM & All Other Liability & Severity Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 6,509,704 10,711, % , % 6.3% 13.5% 2.7% 13.5% 15.7% 2 6,120,848 6,074, % , % 9.5% 21.2% 2.2% 12.5% 20.8% 3 6,033,339 4,458, % , % 14.5% 22.3% 2.2% 13.8% 25.3% 4 6,153,659 4,650, % , % 13.1% 25.3% 4.3% 11.2% 29.6% 5 6,040,925 4,070, % , % 12.0% 26.1% 2.7% 10.9% 34.2% 6 6,092,127 3,768, % , % 9.7% 31.6% 4.5% 11.1% 27.6% 7 5,712,662 2,857, % , % 11.2% 27.0% 2.8% 10.6% 34.4% 8 6,023,786 2,795, % , % 10.7% 34.7% 5.1% 10.9% 28.8% 9 6,092,474 2,370, % , % 13.3% 26.0% 5.0% 12.0% 31.4% 10 6,297,971 1,108, % % 19.9% 17.4% 5.7% 14.2% 32.0% State 61,077,496 42,866, % , % 10.5% 22.6% 3.3% 12.3% 25.1% PREMIUM & CLAIM DATA BY PURE PREMIUM BAND Distribution Of Losses by Peril VMM & Pure All Other Liability & Premium Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 6,201,414 11,329, % , % 8.8% 14.3% 2.6% 12.3% 15.5% 2 6,074,710 6,599, % , % 13.2% 22.6% 2.6% 11.2% 22.2% 3 6,151,391 5,224, % , % 10.7% 30.9% 3.3% 13.1% 27.5% 4 6,275,477 4,377, % , % 9.8% 25.5% 3.3% 12.0% 27.3% 5 5,991,739 3,859, % , % 9.7% 27.0% 3.1% 11.7% 31.0% 6 5,856,361 3,371, % , % 10.2% 23.8% 4.4% 13.6% 29.0% 7 5,793,456 2,896, % , % 9.0% 26.4% 3.8% 11.8% 34.4% 8 6,140,923 2,368, % , % 11.6% 28.3% 4.8% 12.1% 30.8% 9 6,148,716 1,927, % , % 12.8% 21.9% 3.8% 12.4% 36.8% 10 6,443, , % % 18.0% 17.8% 4.6% 14.9% 33.2% State 61,077,496 42,866, % , % 10.5% 22.6% 3.3% 12.3% 25.1% 39

45 Additional frequency, severity and pure premium data, by peril, for each of the geographical bands is provided in Appendix K. Population Density Bands The study s database is summarized in Exhibit XXX for each form of coverage on the basis of each zip code s population density. The loss ratios for the residential coverage are relatively low for the two least densely populated bands and relatively high for the most densely populated Band 10. Theft and vandalism stand out as significant causes of loss for Band 10. The loss ratios for Bands 3 9 are similar and the differences appear to be insignificant. Except for an unusually high loss ratio for the most densely populated Band 10, we find a great deal of consistency in the loss ratios across the population density bands for the tenants coverage. Fire tends to decrease in its importance as a cause of loss as population density increases, whereas the opposite is true for the theft peril. The loss ratios for the condominium owners coverage are consistently lower in the four, least densely populated bands. Theft tends to be a more significant cause of loss for the more populated Bands There is a limited volume of mobilehome data for the more populated Bands 9 and 10, making that data less reliable for analysis. We find consistently higher loss ratios in the middle population density Bands 4 8. We note the significance of liability/medical as a cause of loss for mobilehomes, accounting for 25% of the losses as compared to only 10% for the residential coverage. 40

46 Exhibit XXX Page 1 PREMIUM & CLAIM DATA BY POPULATION DENSITY BAND Residential Owner Policies Distribution Of Losses by Peril VMM & Population All Other Liability & Density Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band* Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 305,743, ,540, % , % 15.1% 14.2% 2.3% 14.3% 15.6% 2 276,007, ,485, % , % 16.1% 18.4% 2.9% 14.7% 14.3% 3 262,037, ,449, % , % 12.9% 21.0% 3.4% 16.5% 13.7% 4 239,823, ,770, % , % 12.8% 24.7% 4.4% 17.8% 12.5% 5 248,404, ,705, % , % 16.2% 25.7% 4.3% 15.8% 10.1% 6 235,965, ,163, % , % 8.8% 28.7% 4.3% 16.5% 10.2% 7 254,617, ,508, % , % 7.7% 30.6% 5.0% 17.0% 8.1% 8 221,558, ,052, % , % 7.1% 29.5% 5.7% 17.7% 8.0% 9 246,056, ,851, % , % 8.4% 28.6% 5.6% 18.0% 6.7% ,845, ,369, % , % 6.2% 20.7% 10.9% 33.7% 4.4% No Data 13,757,118 10,851, % , % 16.1% 15.6% 3.8% 9.2% 11.8% State 2,523,818,587 2,018,749, % , % 11.3% 24.0% 4.8% 18.1% 10.5% Tenant Owner Policies Distribution Of Losses by Peril VMM & Population All Other Liability & Density Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band* Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 5,587,013 2,837, % , % 5.4% 6.4% 8.8% 25.1% 10.2% 2 6,809,931 3,165, % , % 3.9% 5.8% 15.9% 24.3% 8.2% 3 8,043,199 3,487, % , % 2.7% 12.6% 18.8% 21.4% 6.5% 4 9,043,171 4,564, % , % 1.5% 11.7% 25.1% 20.3% 6.3% 5 10,681,844 4,507, % , % 1.2% 16.1% 28.0% 21.8% 8.1% 6 13,452,144 6,019, % , % 1.6% 20.0% 29.4% 20.4% 6.1% 7 13,749,699 7,191, % , % 0.7% 20.3% 29.2% 29.8% 3.5% 8 13,013,415 6,343, % , % 0.7% 19.2% 32.2% 23.0% 5.2% 9 10,267,901 5,586, % , % 0.5% 22.0% 34.8% 23.5% 6.4% 10 5,072,393 7,469, % , % 0.2% 26.2% 39.2% 24.7% 0.8% No Data 432, , % , % 3.1% 10.5% 46.7% 19.9% 6.7% State 96,152,882 51,291, % , % 1.4% 17.8% 28.6% 23.7% 5.5% *Population Density Band 1 has the least people per square mile and Band 10 has the most people per square mile. 41

47 Exhibit XXX Page 2 PREMIUM & CLAIM DATA BY POPULATION DENSITY BAND Condominium Owner Policies Distribution Of Losses by Peril VMM & Population All Other Liability & Density Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band* Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 1,344, , % , % 2.7% 35.4% 2.4% 7.8% 46.5% 2 1,783, , % , % 5.6% 31.8% 1.7% 25.6% 21.3% 3 3,027,398 1,209, % , % 2.9% 46.6% 5.6% 14.3% 21.0% 4 5,152,343 2,446, % , % 1.7% 34.2% 4.7% 25.3% 26.1% 5 8,471,151 5,588, % , % 1.6% 55.4% 7.1% 13.9% 14.4% 6 12,380,836 8,902, % , % 1.5% 46.2% 7.2% 19.3% 12.7% 7 12,806,115 10,310, % , % 1.5% 53.2% 7.7% 16.4% 12.2% 8 9,008,139 5,222, % , % 2.1% 50.6% 8.6% 13.8% 8.5% 9 5,775,559 3,937, % , % 1.2% 54.1% 9.9% 15.1% 5.5% , , % , % 2.8% 27.0% 6.0% 44.6% 2.5% No Data 227,291 90, % , % 0.9% 79.8% 5.1% 1.6% 11.1% State 60,624,259 39,699, % , % 1.8% 49.3% 7.3% 17.1% 13.2% Mobile Homeowner Policies Distribution Of Losses by Peril VMM & Population All Other Liability & Density Earned Incurred Loss Pure Fire & Wind & Water & Physical Medical Band* Premium Losses Ratio Frequency Severity Premium Lightning Hail Freezing Theft Damage Payments 1 14,837,252 8,199, % , % 11.2% 14.5% 3.0% 12.5% 28.9% 2 11,435,511 6,940, % , % 10.3% 22.1% 2.9% 13.4% 23.0% 3 9,994,721 6,690, % , % 9.1% 18.0% 3.4% 12.0% 29.7% 4 8,297,499 6,746, % , % 12.5% 25.4% 3.4% 11.8% 23.1% 5 7,638,893 6,513, % , % 8.1% 26.7% 3.6% 13.5% 24.1% 6 4,227,780 3,790, % , % 10.7% 31.1% 3.0% 10.1% 25.7% 7 2,396,527 2,099, % , % 12.3% 28.5% 2.4% 9.0% 18.3% 8 1,390,510 1,142, % , % 10.6% 27.6% 4.8% 11.9% 19.4% 9 540, , % , % 9.5% 47.1% 3.1% 9.0% 9.3% 10 23,810 5, % % 0.0% 47.7% 0.0% 0.0% 52.3% No Data 294, , % , % 24.9% 16.2% 7.0% 19.1% 20.6% State 61,077,496 42,866, % , % 10.5% 22.6% 3.3% 12.3% 25.1% *Population Density Band 1 has the least people per square mile and Band 10 has the most people per square mile. 42

48 Appendices L through O provide additional details concerning the nature of losses when analyzed by population density bands. Especially for the residential coverages, we observe a remarkable consistency of the frequency of fire claims across the ten population density bands. We also observe little variation in the average severity of the fire claims. If one assumes that insureds are closer to fire protection in the higher populated areas, this data suggests a degree of independence between fire insurance claims and the availability and sophistication of fire protection. This data may suggest more uniformity in the effectiveness of fire protection over the last several decades. The fire data for the tenants coverage is very different from the residential data. For tenants coverage (i.e., apartments), fire losses are significantly higher in the less densely populated areas. We make the following additional observations of this loss data. a) Claim frequencies are relatively constant by population density group for the fire/lightning and the wind/hail perils. b) The water/freezing, theft and vandalism perils have the highest claim frequencies in the high population density bands and the lowest claim frequencies in the low population density bands. c) Claim frequencies for liability claims tend to be lowest in the high population density bands. d) Claim severities for the residential coverages tend to be lower in the high population density bands for wind and water perils and higher for the theft, vandalism and liability claims. e) Pure premiums for fire/lightning, water/freezing, theft and vandalism are among the highest for the most densely populated bands. f) Lower valued homes tend to be more concentrated in the low population density bands and higher valued homes tend to be more concentrated in the middle range of population density bands. 43

49 By-Peril Territories The loss data for each of the six perils were separately analyzed by zip code so as to determine which areas within Michigan were most similar with respect to the risk of fire, the risk of wind losses, and etc. The data were adjusted for catastrophe losses which potentially distorted either the claim frequency or the claim severity for any year in the database. After adjustment for catastrophe losses, the by-peril pure premiums were each adjusted for credibility using a commonly accepted actuarial standard for full credibility. The fully-credible pure premiums for each peril were then grouped so that each group contained zip codes of similar risk for the peril being analyzed. For example, the zip codes were grouped so that each group contained a similar risk of a fire loss. Then another grouping of the zip codes was done so that each group contained a similar risk of a wind loss, then a grouping by theft, and etc. The homogeneity of risk within each zip code group was measured by the statistical variance of the by-peril pure premiums within each group. The goal was to minimize the variance, or increase the homogeneity, of the risk of loss within each group. The loss data indicated 17 distinct, homogeneous areas within Michigan for the risk of fire loss, 15 areas for the wind risk, 18 areas for the water/freezing risk, 15 areas for the theft risk, 20 areas for the vandalism risk, and 14 areas for the liability risk. Maps of the zip code groupings for each peril are presented in Appendices E through J. The maps are significantly different for each peril. The homeowners loss exposure varies geographically and this geographic variation is not the same for each type of loss. The risk of fire loss tends to be scattered throughout the state. Within the Detroit metropolitan area we find pockets of high fire losses that are next door to areas of moderate fire risk. The pattern of fire losses throughout Michigan does not appear consistent with the fire protection class definitions. The capabilities of fire departments are graded with fire protection generally considered to be more effective in large cities than in rural areas. The data in this study do not indicate a strong relationship between the effectiveness of fire protection and the risk of fire losses. The highest exposure to wind/hail losses tends to be concentrated in Michigan s southern tier of counties, not including Wayne County. The Detroit metropolitan area is exposed to a relatively moderate risk of wind/hail losses. 44

50 The risk of loss of water/freezing tends to increase as you move from north to south in Michigan. The highest risk of a theft loss in Michigan tends to be concentrated in a corridor between Detroit and Flint, with decreasing theft risk radiating out from that corridor. The risk of a vandalism/mischief loss tends to also be concentrated in the Detroit/Flint corridor, with an additional exposure observed in and around the Battle Creek area. The risk of a liability/medical loss tends to be concentrated in the far northern and far southern sections of Michigan. The maps show that the likelihood of a loss and the causes of the losses are not uniform across Michigan and that what is likely to give rise to a homeowners claim in one part of Michigan may not be the leading cause of a claim in another part. Degree of Competition This study was not designed to provide an exhaustive analysis of the degree of competition which exists in the Michigan homeowners insurance market. However, the data necessary to study the nature of the claim losses also provide at least a partial picture of the market s competitiveness. In the process of analyzing the geographic variation of claim losses within Michigan, we found a great divergence in the rate territory definitions being used by insurers. Rating territories are the means by which insurers measure and assess geographic risk. Divergence in the assessment of risk is a sign of a competitive market. Divergence in the assessment of risk leads to a divergence of rates available in the insurance market for insurance customers. We found a significant difference in the rates being charged among the ten insurers in the study and we found this divergence in rates to be consistent across all areas within Michigan. We calculated the average residential coverage premium for each of the ten insurers, in each of the ten population density bands. The lowest average residential premium within each band was divided by the overall average residential premium for that band to determine a ratio of the lowest premium to the average premium. We also calculated a ratio of the highest premium to the average premium for each band. These lowand high-premium ratios are presented in the following Table 7. 45

51 Table 7 - Ratio of Low and High Premiums to the Average Population Density Band Lowest Premium Highest Premium Difference We found the greatest rate deviation below the average in the highest population density Band 10, but not significantly different than the downward rate deviations in several other bands (e.g., 5, 6, 7, and 8). We found the least spread in the rates from low to high in population density Band 10, but not significantly different than the spread in rates for either Band 4 or 8. Overall, we found significantly different residential coverage rates available to insurance customers and we found this rate divergence to exist throughout all areas within Michigan. We interpret the spread of rates available in the market as one indicator of a competitive market. Conclusions The study s database includes a wealth of information concerning the nature of homeowners insurance losses in Michigan. At the risk of oversimplifying, we found that: a) the nature of homeowners insurance losses are significantly different for each of the four coverage forms; b) each of the six perils studied has a significantly different impact on the losses of each of the four coverage forms; c) the losses vary significantly by geography and those geographical variations are different for each of the four coverage forms and for each of the six perils. The variation in the losses cited above means that there will likely be substantial differences in the homeowners losses from one insurer to the next. Each insurer s losses will be dependent upon where its insureds are concentrated geographically and what portion of its homeowners coverages are residential, tenants, condominium owners, and mobile homeowners. For example, an insurer focused 46

52 on providing coverage to mobile homeowners or to tenants can be expected to have significantly different claim losses than an insurer that focuses on the residential form of the coverage. We also found a significant difference in the rates being charged by each of the insurers in the study and we found this rate divergence to exist throughout all areas within Michigan. We interpreted the spread of rates, and the differences in the way the insurers assess risk, to be an indicator of a competitive homeowners insurance market in Michigan. 47

53 APPENDICES

54