Property Insurance: Data Needed to Examine Availability, Affordability, and Accessibility Issues (Letter Report, 02/09/94, GAO/RCED-94-39). In the wake of the 1992 Los Angeles riots, concerns have been raised about the availability and affordability of insurance needed for rebuilding. This report examines several issues involving property insurance in urban areas. This report discusses (1) the types of data that are now collected for determining the availability, affordability, and accessibility of property insurance for homeowners and small businesses in urban neighborhoods; (2) the types of data that would be needed to assess these issues if available data are inadequate; and (3) options that are available for collecting these data for homeowners insurance. --------------------------- Indexing Terms ----------------------------- REPORTNUM: RCED-94-39 TITLE: Property Insurance: Data Needed to Examine Availability, Affordability, and Accessibility Issues DATE: 02/09/94 SUBJECT: Insurance companies Data collection operations Comparative analysis Information analysis operations Demographic data Property damages Statistical data Insurance regulation Homeowners insurance IDENTIFIER: Los Angeles (CA) Illinois Minnesota Missouri Wisconsin St. Louis (MO) Kansas City (MO) Chicago (IL) ************************************************************************** * This file contains an ASCII representation of the text of a GAO * * report. Delineations within the text indicating chapter titles, * * headings, and bullets are preserved. Major divisions and subdivisions * * of the text, such as Chapters, Sections, and Appendixes, are * * identified by double and single lines. The numbers on the right end * * of these lines indicate the position of each of the subsections in the * * document outline. These numbers do NOT correspond with the page * * numbers of the printed product. * * * * No attempt has been made to display graphic images, although figure * * captions are reproduced. Tables are included, but may not resemble * * those in the printed version. * * * * A printed copy of this report may be obtained from the GAO Document * * Distribution Facility by calling (202) 512-6000, by faxing your * * request to (301) 258-4066, or by writing to P.O. Box 6015, * * Gaithersburg, MD 20884-6015. We are unable to accept electronic orders * * for printed documents at this time. * ************************************************************************** Cover ================================================================ COVER Report to the Chairman, Committee on Banking, Housing, and Urban Affairs, U.S. Senate February 1994 PROPERTY INSURANCE - DATA NEEDED TO EXAMINE AVAILABILITY, AFFORDABILITY, AND ACCESSIBILITY ISSUES GAO/RCED-94-39 Property Insurance Abbreviations =============================================================== ABBREV ACORN - Association of Community Organizations for Reform Now GAO - General Accounting Office NAIC - National Association of Insurance Commissioners Letter =============================================================== LETTER B-255047 February 9, 1994 The Honorable Donald W. Riegle, Jr. Chairman, Committee on Banking, Housing, and Urban Affairs United States Senate Dear Mr. Chairman: Twenty-five years ago, the President's National Advisory Panel on Insurance in Riot-Affected Areas found ". . . a serious lack of property insurance in the core areas of our nation's cities." The Panel's report\1 outlined widespread practices by insurance companies of drawing a red line around parts of a city that were considered undesirable to insure (redlining). Because insurance, a basic necessity for a property owner, was unavailable in certain areas, the report concluded that "Communities without insurance are communities without hope." Following the report, actions were taken to provide federal backing for insurance to protect against loss due to riots. Eligible states were those that began programs to provide insurance to individuals and businesses that could not otherwise obtain it. The issue reemerged in the late 1970s but received relatively little attention or study over the subsequent decade until news reports following the April 1992 Los Angeles riots revealed potential problems with the availability and affordability of insurance. In the wake of the riots and the resulting concerns about insurance, you requested that we study several issues regarding property insurance in urban areas. Specifically, we agreed to identify (1) the types of data that are currently collected for determining whether property insurance for homeowners and small businesses is available, affordable, and accessible in urban neighborhoods; (2) the types of data that would be needed to assess these issues if available data are not adequate; and (3) options that are available for collecting these data for homeowners insurance. To address our objectives, we examined relevant literature; met with industry trade associations, the insurance commissioners' association, consumer groups, and statistical agents who collect insurance data; and monitored congressional hearings held on the topic. We found little definitive literature or consistency among the affected groups about what data would be needed to determine whether property insurance is available, affordable, and accessible in urban neighborhoods or about how to collect such data. For this reason, we developed our own analyses and conclusions with respect to the data needed to assess these issues and the options available for collecting the data. We then discussed our results with these groups and other experts to assess their validity and reasonableness. -------------------- \1 Meeting the Insurance Crisis of Our Cities: A Report by the President's National Advisory Panel on Insurance in Riot-Affected Areas, Federal Emergency Management Agency, reprint of 1968 version (Washington, D.C.: June 1980). RESULTS IN BRIEF ------------------------------------------------------------ Letter :1 Substantial amounts of homeowners property insurance data are collected from insurance companies by statistical agents to assist states in their regulation of insurance rates. However, most of the data are not useful for determining whether homeowners insurance is available and affordable in urban neighborhoods because the data are aggregated at a geographic level that is too large. In 1994 and 1995, statistical agents will begin collecting homeowners insurance data by the homeowners' ZIP code and could make data in this format available to regulators if requested. No data are currently collected by the statistical agents to analyze homeowners insurance accessibility issues. In addition, no data are generally available from statistical agents for analyzing the availability, affordability, and accessibility of property insurance for small businesses. Without adequate data for analyses, those concerned about urban homeowners and small business owners have had to rely primarily on anecdotal evidence and industry-sponsored surveys to evaluate whether affordability, availability, and accessibility problems exist. We believe that the type of company-reported data that are needed to analyze the availability, affordability, and accessibility of homeowners insurance in urban neighborhoods depends on the issue to be addressed. For example, to examine availability, information on the number of properties insured, by company and type of policy, would be needed. To review affordability-related issues, premium and coverage amounts as well as loss data would also be needed. To determine how accessible insurance is--assuming it is available and affordable--data on marketing activities and agents' locations would be needed. In general, these data would need to be collected at geographic levels, such as at the ZIP-code or census-tract level, that are small enough to be useful for availability, affordability, or accessibility analyses. They could then be analyzed in conjunction with demographic data from the Census Bureau to provide an indication of where problems might exist. In contrast, for commercial insurance, several format and definitional questions would need to be answered before any agreement could be reached on data needs because of the (1) lack of homogeneity among businesses, (2) multiple locations of some businesses, and (3) difficulty in defining "small business." For homeowners insurance, we identified three options for collecting data to analyze availability, affordability, and/or accessibility. The options are to require insurance companies to report (1) existing data at a ZIP-code level, (2) existing data at a census tract level, or (3) existing data plus accessibility-related data that are not currently collected. With any of these options, the data would be more useful for identifying problems if the data could be easily obtained by the public, as well as regulators. However, we believe that a number of additional factors, such as the ease of implementation and volume of data, should be considered before implementing any of these options. BACKGROUND ------------------------------------------------------------ Letter :2 State governments have the primary responsibility for regulating the insurance industry. Since 1945, the legislative basis for this has been the McCarran-Ferguson Act, in which the Congress declared that continued state regulation of insurance is in the public interest.\2 States are responsible for regulating rates, monitoring the availability of insurance, and assessing insurance firms' solvency. Nevertheless, the federal government retains oversight responsibility for the regulation of insurance. Insurance statistical data\3 are used as part of the rate approval process by states. These data are typically reported by insurance companies by rating territory to "statistical agents" who aggregate the data--combining data on similar risks from many insurance companies--on behalf of the states to create a statistically reliable set of data. However, a few large insurance companies compile and report their data to the states independently. Rating territories vary in size but can be as large as the District of Columbia or the city of Chicago. Insurance is a contractual agreement through which an individual transfers the risk of a financial loss associated with an uncertain future event to a company that specializes in assuming such risks. Insurance companies differentiate between risks in order to provide coverage to individuals at a price commensurate with the expected losses. Similar risks that are located within the same defined geographic territory, or zone, are grouped together into "risk classes," or categories, and purchasers in these classes are charged like rates. Determining what these rates should be is referred to as ratemaking. As a result of this process, a wood-frame house, for example, would be charged a higher rate for fire protection than a brick house in the same area. In those instances when a company believes that the risk of loss is unacceptably high given the rate that can be charged, it declines to offer coverage. The decision-making as to which risks to accept and decline is referred to as underwriting. The competitive nature of the insurance industry gives companies the economic incentive to (1) seek out risk to assume wherever they believe doing so is profitable, (2) estimate losses for each risk class as accurately as they are able, and (3) price policies in relation to the expected losses. Individual insurance consumers, on the other hand, make decisions on the basis of the rates and coverage offered by insurance companies, the degree of risk they are willing to bear, and their ability to pay for the coverage. Although, for many consumers, insurance is no more than a second thought in the purchase of a house or operation of a store, they nevertheless view insurance as important in mitigating the risks they face. A concern of community groups and urban residents is that ratemaking and underwriting are sometimes "unfair." Many urban residents, according to the groups, have difficulty in obtaining quality insurance at an affordable price. The industry, in contrast, cites evidence that homeowners insurance is available to all but at premiums that the industry believes reflect underlying risks. These conflicting viewpoints were discussed during congressional hearings in 1993.\4 Some consumer advocates and researchers believe that insurance companies are redlining. The District of Columbia and 26 states have specific laws against redlining. In addition, most state insurance laws require that rates not be excessive, inadequate, or unfairly discriminatory. Unfair trade practices acts also exist in most states to prohibit unfair discrimination in insurance rules and underwriting. Although redlining is now most commonly defined as bias because of the geographic location of the risk, there is a racial component to the controversy that stems from the belief that minorities are more likely to be affected by redlining because they are more likely to live in a redlined area or to be the target of redlining. Redlining is a term that has come to refer to a variety of related insurance issues. Three of these are the availability, affordability, and accessibility of urban property insurance. Availability becomes an issue when insurance companies refuse to write coverage or certain types of coverage for particular parts of an urban area, leaving consumers with fewer options. In markets where there is limited availability, more policies are sold under state-mandated insurance programs\5 or by companies unlicensed in the state.\6 Affordability, a major concern for insurance consumers, refers to whether the consumer can afford to purchase the insurance. A related issue is whether the premium being charged is appropriate. Some consumer groups contend that residents of low-income minority areas pay disproportionately high premiums compared with the risk. Accessibility refers to how easily urban residents can gain access to insurance coverage--assuming that insurance is available and affordable. When insurance companies do not advertise in a particular area and agents that have contracts with major insurance companies are not located in an area, consumers may be unaware of how to obtain insurance or face greater obstacles in trying to purchase it. -------------------- \2 Pub. L. No. 15, 79th Cong., 1st Sess., 59 Stat. 33, Mar. 9, 1945, 15 U.S.C. 1011-1015. \3 Data collected by states are divided into two categories--financial and statistical. Financial data pertain to the financial health of companies, while statistical data relate to the companies' practices. The statistical data contain the type of information that could be used in looking at availability, affordability, and accessibility issues. \4 The Subcommittee on Commerce, Consumer Protection, and Competitiveness, House Committee on Energy and Commerce, held hearings on March 3 and April 26, 1993. The Subcommittee on Consumer Credit and Insurance, House Committee on Banking, Finance and Urban Affairs, held hearings on February 24 and April 1, 1993. \5 These state-mandated insurance programs are usually referred to as fair-access-to-insurance-requirements plans or as involuntary or residual market plans. The plans provide essential property insurance to individuals and businesses that are unable to obtain coverage through conventional sources. \6 These companies may be licensed in a different state and typically offer specialized coverage. They generally are not subject to rate regulation. MOST CURRENTLY AVAILABLE DATA ARE NOT USEFUL FOR DETERMINING IF AVAILABILITY, AFFORDABILITY, AND ACCESSIBILITY PROBLEMS EXIST ------------------------------------------------------------ Letter :3 Although substantial amounts of insurance data are collected by statistical agents, most of these data, in their current form, are not useful for determining whether homeowners and small business property insurance is available, affordable, and accessible in urban neighborhoods. If homeowners property insurance data were collected on a smaller geographic level, such as at the ZIP-code level, some of them could be used to look at the availability and affordability of insurance. Among other things, data on the number of properties insured, the type of coverage, and the dollar amount of the insurance premiums are collected for homeowners insurance and would be useful for studies of availability and affordability. However, no accessibility-related information--such as the location of company agents and the number of solicitations by mail or telephone--is collected by the statistical agents. Within the next few years, data collected by the statistical agents for homeowners coverage are likely to be of greater use for assessing availability and affordability issues because some statistical agents are requiring that the data be reported on a ZIP-code level. One of the largest statistical agents is requiring, effective January 1994, that member companies report homeowners insurance data by ZIP code, while another will begin in 1995. Together, these statistical agents represent over 70 percent of the homeowners insurance market. In addition, another insurance company that represents about 20 percent of the homeowners insurance market and reports independently already has some data by ZIP code. Data specific to small businesses are not available from statistical agents because insurance companies do not use a standard definition of a small business. Instead, data on all types and sizes of businesses are included in commercial property insurance data. According to the two major statistical agents, commercial insurance data will continue to be collected for rating territories. The representatives and officials from the National Association of Insurance Commissioners (NAIC)\7 said that, in addition to the lack of homogeneity among businesses, another complicating factor in collecting data at the ZIP-code level is the likelihood of multiple company locations in different ZIP codes being covered under a single policy, thereby making it difficult to study availability by ZIP code. In addition to data collected by the statistical agents, industry representatives and state insurance officials identified four states that have a regular reporting requirement for homeowners insurance data on a ZIP-code basis. The states--Illinois, Minnesota, Missouri, and Wisconsin--collect data on the number of properties insured and/or policies by ZIP code. In some cases, the data are reported only for selected ZIP codes or only for selected insurance companies on the basis of their volume of business. Illinois and Minnesota collect additional data by ZIP code on the number of renewals, cancellations, nonrenewals, or applications denied. Illinois and Missouri routinely collect data by ZIP code, such as the amount of the insurance premium, that could be used to examine the affordability issues for homeowners insurance. Other states and NAIC have made periodic special requests for data to examine the availability or affordability of insurance. In addition, some states keep data on insurance agents' locations. In August 1993, NAIC requested ZIP-code level data on availability and affordability issues for several types of insurance, including homeowners, in 45 cities in 23 states. Aside from these data, anecdotal evidence and industry-sponsored surveys are what those concerned about urban consumers have had to rely primarily on to evaluate whether affordability, availability, and accessibility problems exist. The extent to which the homeowners property insurance data collected are used by states to examine availability and affordability varies also. The Missouri insurance department has used its data to compare insurance companies' market shares in low-income St. Louis and Kansas City ZIP codes with their market shares statewide. Missouri has used other data to compare the cost per thousand dollars of coverage between low-income black and low-income white neighborhoods. At the time of our review, the Illinois insurance department was studying availability and affordability in the Chicago area. Yet insurance commission officials in Minnesota said that they had done virtually no analysis of the data because of a lack of funds and staff. In each of these four states, the data are also available to consumer groups and the public. The Association of Community Organizations for Reform Now (ACORN) used data from these states in its study on the availability of homeowners coverage in four cities.\8 ACORN combined the insurance data with demographic information from the Census Bureau to compare levels of coverage in neighborhoods with different racial profiles and income levels. -------------------- \7 NAIC is an association of insurance regulators from the 50 states, the District of Columbia, Guam, American Samoa, Puerto Rico, and the Virgin Islands that promotes uniformity in state supervision of insurance matters and recommends legislation in the various state legislatures. \8 A Policy of Discrimination: Homeowners Insurance Redlining in 14 Cities, ACORN (Feb. 4, 1993). DIFFERENT DATA ELEMENTS ARE NEEDED FOR DIFFERENT ISSUES ------------------------------------------------------------ Letter :4 Examining the availability, affordability, and accessibility of homeowners insurance requires different types of data from insurance companies. We believe that deciding upon the objective is the first task necessary to determine the data needed from insurance companies. These data could be analyzed in conjunction with demographic data from other sources to more fully answer availability, affordability, and accessibility questions. In general, data collected at a smaller geographic level will be more useful because smaller units tend to be more homogeneous. The data can serve as a first step toward identifying potential problems but will not provide enough information to determine whether individual companies are unfairly discriminating. Analyzing the availability of homeowners insurance requires having data from insurance companies on the number of properties insured by each company and the locations of the properties. Also needed are data to identify the types of coverage provided. With these data, geographic areas can be examined over time to identify where policies are sold. Insurer-provided data could be analyzed in conjunction with census demographic data to compare policies sold in an area with the number of homes. The insurer-provided data could also be examined on a company-by-company basis to determine which companies are selling fewer policies in a designated area. (See app. I for more details on the types of data needed. App. II discusses examples of how these data and others might be used in practice as well as the limitations in using the data.) In addition, a review of insurance companies' underwriting guidelines could help to explain why companies may not write policies in certain areas. For example, a company with restrictions on the age of property would be less likely to provide insurance in urban neighborhoods where the majority of the buildings are old. To assess whether there appears to be a sound basis for these guidelines, regulators and consumer analysts could review them in conjunction with data on companies' losses and census data. Still, other factors that cannot be discerned through a review of statistical data and underwriting guidelines--for instance, a marketing strategy--may explain the location of a company's policies. To begin to address the issue of affordability of homeowners insurance, including the fairness of rates, requires additional data on the amount of premiums charged and the amount of insurance coverage. These data could be used to determine whether insurance is more costly in urban neighborhoods. However, insurance could be more costly because of higher risks. To assess whether the higher rates have an actuarial basis, at a minimum one would also need industrywide data on losses and risk classifications. Even then, some distinctions among risks, such as those reflecting the condition of the property, may not be captured by the classification system and, thus, would be difficult to assess. Analyzing the accessibility of homeowners insurance or evaluating whether racial discrimination could be occurring in certain areas requires different data. To examine accessibility requires analyzing data on the number of agents by location--for those companies who use sales agents--and the level of solicitation by mail or telephone-- for companies that sell directly to the public. While comparisons of insurer-provided data with demographic data from the Census Bureau would allow for some analysis of the relationship between insurance problems and race, to more directly assess whether racial discrimination could be happening would require data on the race of policyholders and rejected applicants. The greater complexities of commercial insurance make meaningful analysis of availability, affordability, and accessibility through the disclosure of data more difficult to do for a variety of reasons. First, since the type of business frequently affects coverage and price and many types of businesses exist, the data would not make for comparisons as meaningful as homeowners data would. Second, a definition of small business would need to be established to provide a basis for collection. And third, a method would have to be determined for treating businesses with multiple locations. Given the difficulties, some researchers told us that other methods of analysis, such as detailed surveys of small business owners in some selected urban neighborhoods, could be done in lieu of systematic data collection. PROS AND CONS EXIST IN COLLECTING DATA AT SMALLER GEOGRAPHIC LEVELS ---------------------------------------------------------- Letter :4.1 In general, data collected at a smaller geographic level (i.e., ZIP-code level or census-tract-level data\9 ) will be more useful because smaller units tend to be more homogeneous. Aggregating data by ZIP code rather than by the much larger rating territories increases the potential usefulness of data elements. In addition, more data by ZIP code on homeowners insurance will be available beginning in 1994. However, in some cases, ZIP codes may be too large and heterogeneous to identify problems in some individual communities or to highlight demographic similarities among people with insurance problems. Aggregation by census tract as opposed to ZIP code would ensure more accurate matching to census demographic data to determine whether common characteristics exist among underserved communities. For example, even though data on the race of policyholders are not currently collected, a comparison of the census demographic data with insurer-provided data on availability, affordability, and accessibility could provide some insight into whether racial inequities appear to exist. However, industry officials said that aggregating data by census tract, as opposed to ZIP code, would be considerably more difficult because they would have to (1) purchase software to determine a census tract designation for each policy and (2) make major revisions to their computer systems to capture and store the data. At the time of our review, no reliable estimates were available on the cost of reporting data by census tract. Aggregating data on losses by census tract for an individual company would result in too little data to be statistically meaningful, according to statistical agents. They said that the volume of company loss data needed to produce statistically valid information on the appropriateness of the rates charged would generally require aggregating many companies' data for an area significantly larger than either a census tract or a ZIP code. Even though extensive aggregation is needed, building with smaller, more homogeneous geographic units--such as census tracts--has the potential of producing more homogeneous areas for analysis, and thereby more distinct comparisons among various locations. -------------------- \9 A census tract is a small, relatively permanent division of a metropolitan statistical area or selected nonmetropolitan county designed to be relatively homogeneous with respect to population characteristics, economic status, and living conditions. PUBLIC DISCLOSURE OF DATA HAS BENEFITS AND LIMITATIONS ---------------------------------------------------------- Letter :4.2 Public disclosure of insurance data has benefits. For example, community groups suggested that they can use the data to signal where unfair discrimination may exist. Using data as a "flag," they can persuade regulators to investigate companies' practices more extensively. In addition, community organizations emphasized that public disclosure of insurance companies' practices helps the industry to police itself, the benefit of which exceeds that gained merely from regulatory review. To get the full benefit of public disclosure, community groups and researchers stressed that they must be able to obtain any data disclosed in an easily accessible and usable format. While public disclosure has benefits, certain types of disclosure also have limitations. Disclosure could infringe upon a company's right to have trade secrets. For example, industry officials indicate that underwriting criteria are a means by which insurance companies compete with one another. Although some companies may have an idea of what a competitor's guidelines are, they said that they do not know for certain. Individual insurance companies' loss experiences are also considered proprietary. Industry representatives believe that reporting loss data by company could give competitors an indication of an insurer's underwriting criteria. The collection and public disclosure of other types of data, specifically the identification of the policyholder by race, may be problematic as well. Agents, insurance companies, and regulators alike have expressed concern over introducing questions about race into an application process that has until now frequently enabled applicants to be racially anonymous. Some believe that the solicitation of information on race puts both agents and applicants in an uncomfortable position and may even encourage a perception of bias. Others point out practical problems with collecting the data, such as whether applicants would provide the information and what to do when joint applicants are of different races. OPTIONS FOR THE DISCLOSURE OF HOMEOWNERS INSURANCE DATA ------------------------------------------------------------ Letter :5 For homeowners insurance, we believe that alternatives for collecting data to analyze availability, affordability, and accessibility can be grouped under three options. With any of these options, the data would be most easily reported by insurance companies when done prospectively as opposed to retrospectively, and would be more useful if they were easily obtainable by the public. Any of these options could be required by the states--who have the primary responsibility for regulating the insurance industry--or by the federal government in its oversight capacity. One option would be to require all insurance companies to report statistical data by ZIP code, as will be done in 1994 and 1995 by those insurance companies reporting to major statistical agents. A second option would be to require all insurance companies to report statistical data but to specify that they be reported by census tract. A third option would be to require all insurance companies to report, at the ZIP-code level, existing data elements plus those not currently collected by statistical agents, such as agents' locations and the race of applicants, in order to more fully address all of the availability, affordability, and accessibility issues.\10 Of these three options, the first is the easiest and least costly to implement, since it essentially relies on data that are already available or soon to be available for companies representing about 90 percent of the homeowners insurance market. These data include key elements needed to review the issues of availability and affordability. The impact on most of the insurance market would be minimal because most insurance companies are already reporting or preparing to report by ZIP code within the next 2 years. Although we believe that availability and affordability problems could be obscured in some locations where ZIP codes are very heterogeneous, for many other locations, this level of detail may be sufficient for analyses. Furthermore, when potential problems are found, regulators could request census tract data for those ZIP codes. Drawbacks to this approach are that (1) community groups would be dependent upon regulators to collect necessary detailed data within specific ZIP codes, (2) it would be difficult to know whether and to what extent problems are being masked by insufficient detail in the data, and (3) new data needed to assess issues of accessibility and racial discrimination would be unavailable. The second option provides a level of detail that we believe, in many cases, will be the most useful for analysis, especially for doing comparisons across income levels and racial groups. This approach would, however, require insurance companies to make substantial modifications to their computer systems to collect, aggregate, and store data by census tract. The volume of data to manipulate and store would expand significantly under this approach even if the requirement for disclosure were only for currently collected data, since census tracts are more numerous than ZIP code areas. Consequently, we believe that some restrictions might be warranted to limit the volume of data and to ensure their manageability. The restrictions could include limits on the number of geographic areas covered, the frequency of reporting by insurance companies, and/or the number of data elements. Under the third approach, new data items to be collected at a ZIP code level would be used to examine questions of accessibility and/or racial discrimination more directly. However, the new reporting requirements would, in some cases, include data items that insurance companies do not currently collect. Furthermore, the possible collection of data such as racial identity and the disclosure of other data such as insurance companies' underwriting guidelines raise concerns about individuals' and companies' privacy, respectively. In deciding whether to collect data and, if so, how much, several additional factors need to be weighed, such as (1) how federal mandates would affect state responsibilities and roles; (2) what organization would collect and disseminate the data; (3) what the expected cost would be to both insurance companies and to the organization designated to accumulate and disseminate the data; (4) how the data would be made available to the public; (5) what the expected benefit and costs of the data collection would be; and (6) whether methods other than data collection, such as providing incentives to insurance companies who serve urban neighborhoods, could produce change more quickly if problems are believed to exist. -------------------- \10 A fourth option, contained in the original versions of two bills introduced in the 103rd Congress (H.R. 1188 and H.R. 1257) and regarded by many in the industry as infeasible, advocates the extensive collection of new and existing data elements, including census tract designations, by individual policy. CONCLUSIONS ------------------------------------------------------------ Letter :6 Although 25 years have passed since the nation's first presidential report on insurance in riot-affected areas, questions have arisen again about the availability, affordability, and accessibility of property insurance in urban neighborhoods. These questions have been primarily the responsibility of the state governments to address; however, the federal government does retain oversight responsibility in the regulation of insurance. Currently available data are insufficient to determine the extent of current problems. Without adequate data for analyses, those concerned about urban consumers have had to rely primarily on anecdotal evidence and industry-sponsored surveys to evaluate whether affordability, availability, and accessibility problems exist. Data that are collected for homeowners insurance will be more useful in examining availability and affordability once the data are collected on a ZIP-code level (beginning in 1994) and analyzed in conjunction with Census Bureau data. However, data on accessibility are not collected. Reducing the size of the reporting unit to census tracts would, in most cases, increase the value of the data by enabling more homogeneous units to be analyzed. Any consideration of such reporting would need to be weighed carefully against the additional burden it places on companies to comply. In contrast to having some information on homeowners insurance, no data are available to analyze availability, affordability, and accessibility issues concerning property insurance for small businesses. Furthermore, several format and definitional questions would have to be resolved before meaningful data could be collected for small business insurance. However, other methods of analysis, including surveys of small business owners in some selected urban neighborhoods, could provide useful information about the extent of problems with the availability and affordability of insurance in those areas. Whatever insurance data are reported in the future by insurance companies will be more useful in assessing availability, affordability, and accessibility issues if the public can easily obtain them. The data that can be readily reported by companies will be useful for examining availability and affordability issues, but will not be sufficient to determine conclusively whether unfair discrimination exists or why. However, such data would provide a marked improvement over the knowledge available today and could serve to point regulators more effectively in directions for further probing. AGENCY COMMENTS ------------------------------------------------------------ Letter :7 We discussed our findings and conclusions with senior officials from (1) NAIC; (2) insurance trade associations and statistical agents--the American Insurance Association, the National Association of Mutual Insurance Companies, the Insurance Services Office, and the National Association of Independent Insurers; and (3) consumer groups--the National Insurance Consumer Organization, ACORN, and the Consumer's Union. NAIC, the trade associations, and the consumer groups said that the report was well-balanced and generally accurate and that the data elements we developed were reasonable. However, there was some disagreement concerning the relative costs and benefits of reporting by census tract. Industry officials believe that reporting data by census tract would be more costly and less beneficial than do consumer groups. We have addressed the viewpoints of the various groups in the report. These groups and the statistical agents also provided technical corrections, which we have incorporated into this report. As requested, we did not obtain written comments on a draft of this report. SCOPE AND METHODOLOGY ------------------------------------------------------------ Letter :8 To determine what types of homeowners and business property insurance data are currently available and will be available in the future to examine affordability, availability, and accessibility issues, we reviewed NAIC's model plan for the collection of data and documents summarizing the data collected by the two largest statistical agents. We also examined relevant literature and monitored congressional hearings held on the topic. Through our discussions with NAIC, the statistical agents, and industry and consumer group representatives, we identified four states that routinely require homeowners insurance data at the ZIP-code level. We also obtained information from these states regarding their data requirements. To determine what types of data would be needed to address the issues of availability, affordability, and accessibility for homeowners and small businesses and how the data could be reported, we reviewed relevant insurance literature, proposed bills, and state requirements. We also interviewed officials representing industry trade associations, consumer groups, statistical agents who collect insurance data, research organizations, and NAIC. We obtained documentation regarding their positions, if any, on proposed data disclosure legislation. However, we found little definitive literature or consistency among the affected groups about what data would be needed to examine availability, affordability, and accessibility issues or how they could be reported. For this reason, we developed our own analyses and conclusions with respect to the needs for additional data to assess these issues and options for collecting these data. We then discussed our results with these groups and other experts to assess their validity and reasonableness. We conducted our review between April and September 1993 in accordance with generally accepted government auditing standards. ---------------------------------------------------------- Letter :8.1 As arranged with your office, unless you publicly announce its contents earlier, we plan no further distribution of this report until 30 days after the date of this letter. At that time, we will send copies of the report to NAIC and other interested parties. Copies will be made available to others on request. Please contact me at (202) 512-5167 if you or your staff have questions. Major contributors to this report are listed in appendix III. Sincerely yours, Judy A. England-Joseph Director, Housing and Community Development Issues SUMMARY OF DATA ELEMENTS THAT COULD BE USED TO ADDRESS HOMEOWNERS INSURANCE AVAILABILITY, AFFORDABILITY, AND ACCESSIBILITY ISSUES =========================================================== Appendix I Data element Availability Affordability Accessibility -------------------------------- -------------- -------------- -------------- Generally reported by insurance companies to statistical agents\a Type of coverage: (policy form, X X residual, or voluntary market)\b Company name X X X Number of exposures\c X X Premium amount (in dollars) X Amount of insurance coverage X Number of claims\d X Dollar amount of losses\d X Risk classification X Not reported to statistical agents, but could be reported by insurance companies Underwriting guidelines X Number of cancellations, X nonrenewals, and written declinations made by the insurance company Level of direct mail or X telephone solicitation Number of agents X Generally not collected by insurance companies Number of declinations made over X the telephone by the insurance company -------------------------------------------------------------------------------- \a These data are currently aggregated at a rating-territory level by the statistical agents. However, the two major statistical agents plan to begin aggregating the data by ZIP code within the next 2 years. \b "Policy form" refers to the type of homeowners policy. For example, an H0-3 policy covers the home's contents and the structure for multiple perils. "Residual, or voluntary market" refers to whether the policy was sold through state-mandated insurance programs or through conventional sources. \c "Number of exposures" means the number of policies over a set amount of time, such as years. For example, a policy for one house for 1 year would represent one exposure. \d Loss experience data would need to be aggregated across companies to generate a statistically credible data set. POSSIBLE USES OF DATA FROM INSURANCE COMPANIES ========================================================== Appendix II Many types of analysis could be done to examine homeowners insurance availability, affordability, and accessibility trends using data from insurance companies in conjuction with census demographic information. Below are a few examples of how data might be used in examining each of the issues, as well as some of the limitations of such analysis. AVAILABILITY One type of analysis would be to compare the total number of properties that all companies report insuring in the designated area with the total number of homes in the area as recorded in census data. If, for example, census data showed 1,000 homes in an area, while companies reported insuring 900 homes in the area, then one could estimate that about 90 percent of the homes in that area are insured. As part of this analysis, one would also want to compare the proportion of properties insured by the type of coverage. The policy types--building-only versus building and contents coverage, replacement value versus market value loss protection, and state-mandated plan versus voluntary market--might vary depending on the location. Comparing data from insurance companies with census data is not without shortcomings. First, the analysis is limited by the accuracy of the census data to reflect the true number of homes. The margin of error will also increase as the census data ages and where data are aggregated by ZIP codes rather than census tracts. Second, some insurer databases may not capture the true location of the insured property if it is being rented. And third, the analysis does not allow one to conclude why houses might be uninsured or underinsured and whether insurance is necessarily unavailable or simply not in demand by homeowners in that area. An alternative analysis of availability would be to compare an insurance company's share of the market in the designated urban area with its share statewide. In this method of analysis, one is comparing an insurance company with itself to see whether the company writes proportionately more or less of its policies in the designated urban area. These comparisons could also be made over time to assess whether insurance companies are increasing or decreasing their shares in certain areas. This second form of analysis has limitations as well. As with the first method, the market share analysis also fails to explain differences. There may be legitimate reasons why an insurance company has proportionately lower market share in urban areas than in the rest of the state. Second, since statewide market share is not generally available by type of homeowners policy, it would be impossible to detect differences in the types of policies an insurance company writes across the state. And third, even if an individual company is not writing in a designated area, this need not imply that there is an availability problem, especially if other insurance companies are writing there. AFFORDABILITY To compare trends in the cost of insurance between different geographic areas, one could calculate the average premium paid per similar property and policy coverage in the designated areas. It is important to distinguish between different types of policies as much as possible, since differences in coverage have a bearing on cost. Likewise, the dollar amount of insurance would affect the premium--more expensive homes are more costly to insure but would normally have a lower cost per thousand dollars of home value. The affordability data could also be analyzed in conjunction with census demographic information to determine whether insurance tends to cost more in certain areas. To assess whether premiums charged appear to have an actuarial basis, one would want to compare premiums with actual loss experience across different risk classifications and geographic areas. In this way, one can see, for example, whether a difference in rates charged to similar risks reflects a difference in loss experience in different geographic areas. Yet multiyear data may be needed to generate a credible database, and some differences in rates may not be clearly explained through this analysis. Individual insurance companies may differ in the premiums they charge because they base their ratemaking analysis on different data. The largest insurance companies in an area often define rating territories differently, and they may consider additional risk factors that others do not. In addition, some insurance companies base their rates on their own loss experience, while others use an aggregated database that reflects the loss experience of many insurers. As a result of variations such as these, what looks like an adequate premium for one insurance company may not to another. ACCESSIBILITY For insurance companies that use sales agents, one could analyze the geographic distribution of agents to gauge how accessible agents are to consumers. For insurance companies that market their products directly to consumers, one could analyze the geographic distribution of their telephone and mail solicitations. Coupled with census demographic information, these data could also be used to evaluate whether agent locations and company marketing campaigns tend to concentrate on certain types of neighborhoods or income groups. The analysis would not, however, explain why agents locate where they do or why insurance companies concentrate their marketing efforts in certain areas. Neither could one conclude, using accessibility data alone, that agents are not selling insurance in areas beyond their office location. It is possible that an insurance company without agents located in a specific area could still have the largest market share there. MAJOR CONTRIBUTORS TO THIS REPORT ========================================================= Appendix III RESOURCES, COMMUNITY, AND ECONOMIC DEVELOPMENT DIVISION, WASHINGTON, D.C. Jacquelyn L. Williams-Bridgers, Associate Director Marnie Shaul, Assistant Director Susan H. Beekman, Evaluator-in-Charge C. Bernard Myers, Staff Evaluator Sandra A. Potter, Intern