1. See The Fact Book: 1994 Property/Casualty Insurance Facts 5 (New York: Insurance Information Institute, 1994).Google Scholar

2. See Franklin Delano Roosevelt, Communication from the President to Congress, 78 Cong. Rec. S.10770, 78 Cong. Rec. H.10851 (daily ed. 8 June 1934); William H. Simon, “Rights and Redistribution in the Welfare System,” 38 Stan. L. Rev. 1431 (1986).Google Scholar

3. “Clinton Answers Questions about Health Care Reform,” Washington Post, 12 April 1994, Health, at 28.Google Scholar

4. See, e.g., William Julius Wilson, The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy (Chicago: University of Chicago Press, 1987).Google Scholar

5. See, e.g., the essays collected in Georges Dionne & Scott E. Harrington, eds., Foundations of Insurance Economics (Boston: Kluwer Academic Publishers, 1992). See also Kenneth Abraham, Distributing Risk: Insurance, Legal Theory and Public Policy (New Haven, Conn.: Yale University Press, 1986) (“Abraham, Distributing Risk“).Google Scholar

6. See, e.g., Ian Ayres, “Fair Driving: Gender and Race Discrimination in Retail Car Negotiations,” 104 Harv. L. Rev. 817 (1991) (discussing the implications for neoclassical economics of research finding gender and racial bias in the new car market).CrossRefGoogle Scholar

7. See Pierre Bourdieu & Loic J. D. Wacquant, An Invitation to Reflexive Sociology 94-100 (Chicago: University of Chicago Press, 1992) (on the concept of a field).Google Scholar

8. See, e.g., Richard Abel, “A Socialist Approach to Risk,” 41 Md. L. Rev. 695, 697–98 (1982); Richard Epstein, “Products Liability as an Insurance Market,” 14 J. Legal Stud. 645 (1985).Google Scholar

9. H. Laurence Ross, Settled Out of Court: The Social Process of Insurance Claims Adjustment (Chicago: Aldine Publishing Co., 1970) (“Ross, Settled Out of Court”).Google Scholar

10. William C. Whitford, “Strict Products Liability and the Automobile Industry: Much Ado about Nothing,” 1968 Wis. L. Rev. 1983.Google Scholar

11. See Kenneth S. Abraham & Lance Liebman, “Private Insurance, Social Insurance and Tort Reform: Toward a New Vision of Compensation for Illness and Injury,” 93 Colum. L. Rev. 75 (1993). It should be clear that we are using the word “private” in the colloquial sense indicating that the insurance companies, in contrast to, say, the Social Security Administration, are not state owned or operated.CrossRefGoogle Scholar

12. Daniel Defert, “‘Popular Life’ and Insurance Technology,” in Graham Burchell, Colin Gordon, & Peter Miller, eds., The Foucalt Effect: Studies in Governmentality 211 (Chicago: University of Chicago Press, 1991) (“Burchell et al., Foucault Effect”); Pat O’Malley, “The Prudential Man Cometh” (presented to 1995 Law & Society Association Annual Meeting, Toronto).Google Scholar

13. Pat O’Malley, “Legal Networks and Domestic Security,” 11 Stud. L. & Pol. Soc. 171 (1991).Google Scholar

14. Jonathan Simon, “The Ideological Effects of Actuarial Practices,” 22 Law & Soc'y Rev. 771 (1988); id., “In the Place of the Parent: Risk Management and the Government of Campus Life,” 3 Soc. & Legal Stud. 15 (1994).CrossRefGoogle Scholar

15. Jonathan Simon, “The Emergence of a Risk Society: Law, Insurance, and the State,” 89 Socialist Rev. 61 (1987); Francois Ewald, “Insurance and Risk,” in Burchell et al., Foucault Effect 197.Google Scholar

16. Michael Lipsky, Street Level Bureaucracy: Dilemmas of the Individual in Public Services (New York: Russell Sage Foundation, 1980).CrossRefGoogle Scholar

17. Joel Handler, The Conditions of Discretion (New York: Russell Sage Foundation, 1986).Google Scholar

18. See, e.g., Roy Sainsbury, “Administrative Justice: Discretion and Procedure in Social Security Decision-making,” in K. Hawkins, ed., The Use of Legal Discretion 295 (New York: Oxford University Press, 1992) (“Hawkins, Discretion“).Google Scholar

19. Keith Hawkins, Environment and Enforcement: Regulation and the Social Definition of Pollution (Oxford: Clarendon Press, 1984).Google Scholar

20. H. Laurence Ross & John M. Thomas, “Housing Code Enforcement: A Study of Law in Action” (1995) (MS. on file with the first author) (“Ross & Thomas, ‘Housing Code Enforcement’ ”).Google Scholar

21. See, e.g., Richard Lempert, “Discretion in a Behavioral Perspective: The Case of a Public Housing Eviction Board,” in Hawkins, Discretion 231.Google Scholar

22. Carol A. Heimer, “The Racial and Organizational Origins of Insurance Redlining,” 10 J. Intergroup Rel. 42 (1982).Google Scholar

23. See, e.g., Gregory D. Squires & William Velex, “Insurance Redlining and the Process of Discrimination,” Rev. Black Pol. Econ., Winter 1988, at 63; id., “Insurance Redlining, Agency Location, and the Process of Urban Disinvestment,” 26 Urb. Aff. Q. 567 (1991).CrossRefGoogle Scholar

24. Anthony D. Taibi, “Banking, Finance, and Community Economic Empowerment: Structural Economic Theory, Procedural Civil Rights, and Substantive Racial Justice,” 107 Harv. L. Rev. 1465 (1994); Ruthanne DeWolfe, Gregory Squires, & Alan DeWolfe, “Civil Rights Aspects of Insurance Redlining,” 29 DePaul L. Rev. 315 (1980); David I. Badain, “Insurance Redlining and the Future of the Urban Core,” 16 Colum. J.L. & Soc. Prob. 1 (1980).Google Scholar

25. Joel Handler, “Power, Quiescence and Trust,” in Hawkins, Discretion 333 (“Handler, ‘Power’ ”) (criticizing the discretion literature for insufficient attention to distributional issues).Google Scholar

26. In the disaster literature, Bolin and Bolton noted race/ethnicity differences in insurance coverage after natural disasters, but they did not explore whether these differences resulted from purchasing patterns or disparate treatment in the claims process. See Robert Bolin & Patricia Bolton, Race, Religion and Ethnicity in Disaster Recovery 16, 169 (Boulder: University of Colorado Institute of Behavioral Science, 1986).Google Scholar

27. The first author has compared the homeowners insurance policies issued by the leading homeowners insurance carriers in the Florida market (Allstate and State Farm), as well as the standard Insurance Services Office, Inc., policies issued by most of the rest of the carriers. These are the same insurance policies sold in most of the United States. While there were minor differences in the homeowners policies sold by various insurers, those differences are not significant for present purposes. Similarly, while the standard Florida policies differ slightly from those sold in other states, those differences are also not significant.Google Scholar

28. This practice is authorized by state law. See, e.g., Fla. Admin. Code § 3D-160.025 (12). It is also mandated by federal regulation. See, e.g., 12 C.F.R. § 571.4 (1994) (requiring each savings association “to include in its loan contracts provisions which require the maintenance of such hazard insurance as will protect the savings association from loss in the event of damage to or destruction of the real estate securing the savings association loan”).Google Scholar

29. For example, the race/ethnicity composition of our sample was 65% white non-Hispanic, 8% black non-Hispanic, and 27% Hispanic, compared with the Dade County population of 32% white non-Hispanic, 19% black non-Hispanic, and 49% Hispanic. Dade County figures are from U.S. Bureau of the Census, Census of Population and Housing, 1990, Summary Tape File 3, Florida.Google Scholar

30. The qualitative research was conducted by and under the direction of the first author.Google Scholar

31. This finding is consistent with survey results using a random sample of South Florida residents contacted by telephone four months after the hurricane. See Walter Gillis Peacock, Chris Girard, & Hugh Gladwin, “A Summary of Findings from the FIU Hurricane Andrew Survey” (1994) (unpub. report on file with the authors).Google Scholar

32. Logistic regression differs from the more commonly used linear regression in that logistic regression can be used to analyze associations with a dependent variable that is yes/no in character, such as having insurance or not. The beta weights and odds ratios reported in our tables express the strength of the association between a particular independent variable and the dependent variable under analysis. Here, the dependent variable is having insurance and the independent variables are race/ethnicity, income, etc.Google Scholar

A logistic regression equation predicts the log of the odds of an observation being in one category of the dependent variable versus the other. This is the beta weight (b) reported in the tables. An odds ratio is the antilog of the beta weight (e ^Beta), and it always has a positive sign (i.e., it is greater than zero). An odds ratio represents the amount by which a unit change in an independent variable multiplies the odds of an observation being in one category versus the other, holding the remaining independent variables constant. See David W. Hosmer, Jr., & Stanley Lemeshow, Applied Logistic Regression 41 (New York: Wiley, 1989). This means that the farther an odds ratio for a given independent variable is from 1.00, the more strongly associated that variable is with the dependent variable, either in a positive or negative direction. In the first equation, the highest odds ratio is the 12.44 reported for homeownership, which tells us that owning a home is very strongly associated with having insurance. The lowest odds ratio is the .17 reported for living in a trailer, which tells us that respondents who lived in trailers were quite unlikely to have insurance. We calculated the odds ratios only for the significant variables.

In comparing odds ratios among independent variables, it is important to keep in mind the difference between dummy variables (i.e., yes/no variables, such as homeownership in this study), interval variables (i.e., variables with categories composed of ranges, such as income in this study) and continuous variables (such as age in this study). Because the odds ratios predict the effect of increasing or decreasing the independent variable by one unit, regardless of the kind of variable under analysis, an important dummy variable can be expected to have an odds ratio further from 1.00 than an important interval variable, just as an important interval variable can be expected to have an odds ratio further from 1.00 than an important continuous variable. See id. at 56. Among our variables, age is continuous; income, education, and damage are interval, and the remainder are dummy variables. A further description of the variables appears in the technical appendix.

33. Among renters, the percentages having insurance by home type were 25.9% for apartments, 29.5% for townhouses, and 43.8% for houses; among owners, the percentages were 69.4% for apartments, 88.7% for townhouses, and 97.0% for houses.Google Scholar

34. See Vince E. Showers & Joyce A. Shotick, “The Effects of Household Characteristics on Demand for Insurance: A Tobit Analysis,” 61 J. Risk & Ins. 492 (1994). The study did not consider education or the other independent variables used in our equations, so no comparison of those measures can be made.Google Scholar

35. Respondents who did not report their ethnic identity also differed significantly from those who reported this item. This finding suggests that missing race/ethnicity data might have biased the results for this equation had we not accounted for the omission.Google Scholar

36. A comparison of renters and owners by race/ethnicity revealed that among white non-Hispanics, 61% of renters, 16% of condominium owners, and 2% of house owners were not insured; among black non-Hispanics, 80% of renters and 9% of house owners were not insured (there were too few condominium owners for comparison); among Hispanics, 73% of renters, 33% of condominium owners, and 3% of house owners were not insured.Google Scholar

37. While there is no difference in the incidence of home insurance among white non-Hispanic and Hispanic owners of houses, there is a substantial difference in the incidence of home insurance among owners of apartments, as the text in note 36 shows. This result is consistent with the mortgage “compulsion” because in the case of an apartment-type condominium the important coverage, from the lender's perspective, is the insurance that the condominium association maintains on the building structure, not the insurance the individual unit owners maintain on their personal property.Google Scholar

38. See, e.g., Squires & Velex, 26 Urb. Aff. Q. 567 (cited in note 23).Google Scholar

39. Kimball has described “fair discrimination” as follows: “to measure as accurately as is practicable the burden shifted to the insurance fund by the policy holder and to charge exactly for it, no more and no less.” Spencer L. Kimball, “Reverse Sex Discrimination: Manhart,“ 1979 A.B.F. Res. J. 83, 105 (1979). For a perceptive review of the literature and a discussion of the principle of actuarially fair discrimination in the context of gender-based life insurance and annuity premiums, see Simon, 22 Law & Soc'y Rev. 771 (cited in note 14).Google Scholar

40. See, e.g., R. W. Swegle (Vice-President of Safeco), Letter to the Editor, Nat'l Underwriter, Prop. & Casualty Ed., 21 March 1994, pp. 23-24 (responding to “redlining” allegations by stating that “there are loss patterns which correlate with geography, construction, replacement cost and age of construction”).Google Scholar

41. Cf. Abraham, Distributing Risk 94 (cited in note 5) (“[w]hether the use of territorial variables is considered an alternative to discriminatory classification or a mere subterfuge depends on whether the purpose of the substitution is symbolic or practical”).Google Scholar

42. There are allegations that blacks have less access to the conventional mortgage market than non-Hispanic whites. See, e.g., Roberta Achtenburg (Assistant Secretary, United States Department of Housing and Urban Development Office of Fair Housing and Equal Opportunity), Prepared Statement before the House Committee on the Judiciary, Subcommittee on Civil and Constitutional Rights Oversight Hearings on Fair Housing, 28 Sept. 1994; Peter P. Swire, “The Persistent Problem of Lending Discrimination: A Law and Economics Analysis,” 73 Tex. L. Rev. 787 (1995). At least some alternative financing arrangements, such as loans from family members or religious organizations, lack the institutional mechanism for requiring insurance exercised by banks and mortgage companies. As a result, reduced access to the conventional mortgage market can be expected to result in a lower incidence of homeowners’ insurance.Google Scholar

43. Such research would have to take into account the fact that the label “Hispanic,” like all such labels, hides a great deal of cultural diversity. See Virginia Dominguez, “Samenesses” (1994) (MS. on file with the author).Google Scholar

44. Showers & Shotick, 61 J. Risk & Ins. 492 (cited in note 34).CrossRefGoogle Scholar

45. As Kent Syverud's recent, provocative article suggests, all individuals without significant assets to protect might well be better off without such insurance, or at least without large amounts of such insurance, because having insurance makes one a more likely target of litigation. Kent Syverud, “On the Demand for Liability Insurance,” 72 Tex. L. Rev. 1629 (1994).Google Scholar

46. To screen out meetings at an insurance field office, which would not give the adjuster an opportunity to assess the damage to the home, the survey specified a meeting “at the home.”Google Scholar

47. See Barry Schwartz, Queuing and Waiting: Studies in the Social Organization of Access and Delay (Chicago: University of Chicago Press, 1975).Google Scholar

48. The odds ratio of 1.29 for the damage variable in table 4 means that one step up the seven-point damage scale increased the likelihood of being paid within the month by nearly 30%. For a description of the damage scale, see the technical appendix. For an explanation of logistic regression and odds ratios, see note 32.Google Scholar

49. The percentages of claimants in each race/ethnicity group who received a payment within 1 month after the hurricane were 42% (white non-Hispanic); 39% (black non-Hispanic); and 25% (Hispanic). For an explanation of why the timing measures are dichotomized, see the technical appendix.Google Scholar

50. There was no significant interaction effect between damage and ethnicity.Google Scholar

51. Our geographic variable was defined according to preliminary estimates of damage and could not be adapted to distinguish the relative prestige of the address.Google Scholar

52. We are indebted to Terence Halliday for the observation that, by virtue of their relationship to the university and those within it, our survey respondents were comparatively advantaged with regard to social capital.Google Scholar

53. These factors would not have as significant an effect on the timing of the first insurance payment of the first check, because a claimant could go to a walk-in claims center to pick up a check.Google Scholar

54. Florida Department of Insurance Hurricane Andrew Fact Sheet (16 Dec. 1994) (on file with the first author).Google Scholar

55. Florida Department of Insurance Hurricane Andrew Mediation Reports (27 Dec. 1994) (on file with the first author).Google Scholar

56. The first author reviewed the court's list and, together with a research assistant, examined the court files of all the insurance cases. While there is some reason to believe that the court's list of cases is incomplete, the order of magnitude is still informative. The trend line is important because it provides some confidence that a substantial number of the claims that will be litigated have already been filed, notwithstanding the applicable five-year statute of limitations. In February 1993, 21 complaints were filed; in March 1993, 20 complaints; in April 1993, 8 complaints; after that time, the trend was flat or down (data on file with the first author).Google Scholar

57. See, e.g., Handler, “Power” (cited in note 25).Google Scholar

58. Examples of such policy requirements include the requirement that the insured “mitigate” damages, that the damage not be the result of a failure to maintain the property, that the cost of rebuilding not include costs of improving the building to meet current building code requirements, that damaged items be replaced within 180 days of the loss, and that the home not be underinsured. There was widespread agreement that insurance adjusters often did not pursue such “technicalities.” As one adjuster put it, “If I and the other adjusters kept to the letter of the policy, the number of complaints would have been at least three times as much.”Google Scholar

59. See Tom Baker, “Constructing the Insurance Relationship: Sales Stories, Claims Stories, and Insurance Contract Damages,” 72 Tex. L. Rev. 1395, 1411 (1994). This story serves much the same function as similar moral evaluations made by social workers and governmental enforcement officers. See Keith Hawkins, “The Use of Legal Discretion: Perspectives from Law and Social Science,” in Hawkins, Discretion (cited in note 18); Ross & Thomas, “Housing Code Enforcement” (cited in note 20) (describing how housing code inspectors distinguish between “bad apples” and “good apples”).Google Scholar

60. Erving Goffman, “On Face Work,” in Erving Goffman, Interaction Ritual (Garden City, N.Y.: Anchor Books, 1967).Google Scholar

61. See Handler, “Power” (cited in note 25).Google Scholar

62. We do not endorse these views; we simply report them. Like xenophobia in the political arena, concern about the fraudulent tendencies of recent immigrants has a long history in the insurance field. See Tom Baker, “On the Genealogy of Moral Hazard,” 75 Tex. L. Rev. (forthcoming 1996).Google Scholar

63. See Susan T. Fiske & Shelley E. Taylor, Social Cognition 119–24 & 145 (2d ed. Reading, Mass.: Addison-Wesley Pub. Co., 1991). We are grateful to Karyl Kinsey for suggesting the contact salience hypothesis that led us to test the interaction between ethnicity and adjuster visits.Google Scholar

64. Ross, Settled Out of Court (cited in note 9).Google Scholar

65. For additional anecdotal reports of racial, ethnic, and gender bias in the insurance claims process, see Michelle Saadi, Claim It Yourself: The Accident Victim's Guide to Personal Injury Claims 17–18, 50–60 (New York: Pharos Books, 1987).Google Scholar

66. See supra note 28 and accompanying text.Google Scholar

67. See Louis Kaplow, “Incentives and Government Relief for Risk,” 4 J. Risk & Uncertainty 167 (1991).CrossRefGoogle Scholar

68. A more politically palatable, but slightly more complicated, approach might be to mandate that landlords and condominium associations purchase insurance of a given sort protecting their tenants’ or members’ property and to provide that the sanction for the failure to do so would be liability for the destruction of that property.Google Scholar

69. The institutional pressure to insure could be increased if the liability were made a lien on the property. This would induce lenders to monitor landlords and condominium associations. That lenders are capable of such monitoring is apparent in the homeowner context.Google Scholar

70. Because of the problem of insolvent landlords and condominium associations, however, even this solution would not eliminate the need for disaster grants and loans. The lien suggestion in note 69 would reduce this solvency concern.Google Scholar

71. For a theoretical analysis of why mandatory insurance is superior to government relief, see Kaplow, 4 J. Risk & Uncertainty 167. It should be noted that Kaplow addresses a situation in which the benefits provided by compulsory insurance are the same as the benefits provided by relief. Because home insurance provides more extensive benefits in a much wider range of circumstances than does federal disaster relief, mandating home insurance would, therefore, increase the total social resources devoted to protection against property loss. It would do so, however, on behalf of those who now need that protection most.Google Scholar

72. For a recent effort to address the related problem of discrimination in the lending market, see Peter P. Swire, “The Persistent Problem of Lending Discrimination,” 73 Tex. L. Rev. 787 (1995).Google Scholar

73. See Spencer L. Kimball, “The Role of the Court in the Development of Insurance Law,” 1957 Wis. L. Rev. 520.Google Scholar

74. For an early description of this aspect of insurance department activity, as well as a prediction that it would increase, see Edwin W. Patterson, The Insurance Commissioner in the United States 283–307 (Cambridge: Harvard University Press, 1927). For a more recent (and pessimistic) study, see William Whitford & Spencer Kimball, “Why Process Consumer Complaints? A Case Study of the Office of the Commissioner in Wisconsin,” 1974 Wis. L. Rev. 639.Google Scholar

75. At least some insurance companies maintain detailed demographic information about their policyholders. For companies that do not, the insurance department could select a sample of claimants and conduct a telephone survey to obtain the necessary information. Alternatively, the insured's name can be used for distinguishing gender and ethnic identity, and policy limits can be used as a proxy for income. Together with census information, address can also be used to identify demographic characteristics. While such indicators would be far from perfect, they could be used for the large-scale mapping and screening purposes we envision.Google Scholar

76. In addition to educating regulators about systemic effects of insurance adjuster discretion, claims performance measures could serve a more general purpose. Most personal-lines insurance advertising today consists of idealized “sales stories” that provide consumers with little basis for distinguishing among insurers. See Baker, 72 Tex. L. Rev. 1395, 1403-6 (cited in note 59). Claims performance measures would permit consumers to go beyond both the sales stories and the diffuse reputational information now available from intermediaries such as insurance agents, realtors and mortgage brokers. While not a panacea, the measures would give consumers an objective basis to use, in combination with price, in deciding where to invest their premium dollars.Google Scholar

77. The survey asked the following race/ethnicity question: “Which best describes you? (MARK ONE ANSWER ONLY).” The options were (1) White, non-Hispanic; (2) Black, non-Hispanic; (3) Asian or Pacific Islander; (4) Other (please specify); (5) Hispanic, regardless of race.Google Scholar