Urban Decline or Disinvestment

87

should be acknowledged that the only companies for which data were obtained were ones that are still writing at least some insurance in the city of Chicago: such companies are frequently criticized while those which stopped writing any insurance in the city go unscathed. Second, our analysis does not examine such subtle, but important, forms of redlining as variations in the costs, the coverage, or the terms under which insurance is available. Charging a prohibitive price, requiring full payment at the beginning of a year rather than quarterly or monthly, varying the deductibles available, and limiting the types of protection that are offered, are all examples of marketing techniques used differentially by geographic area-usually to the detriment of older urban communities. Finally, it is recognized that not every company has explicit Zip Code rating territories. The basic issue, however, is alleged unfair discrimination against specific geographic areas within major metropolitan areas in both direct and indirect insurance company practices. In order to address this issue, variations in industry marketing practices by neighborhood must be examined. The only insurance underwriting data available on a basis that even approximates neighborhoods are the Zip Code data which the Illinois Department of Insurance began collecting in 1977. At the time this analysis was conducted (summer 1978), Illinois was the only state which had enacted a disclosure law requiring companies to report their underwriting activity on any geographic basis approximating neighborhoods. Under Illinois law all insurers writing homeowners insurance must report-by Zip Code, and quarterly-the number of policies they have written, renewed, cancelled and nonrenewed. (Wisconsin, Missouri and Minnesota have since enacted similar legislation and several other states are considering such action.) With the Illinois data it is possible for the first time to begin answering the key issues that have been raised in the insurance redlining debate.

Results

Insurance underwriting activity does vary markedly among Zip Code regions within the city of Chicago. For example, the number of homeowners policies written or renewed ranged from 170 (0.6 per 100 housing units) in a predominantly black Zip Code region on the southwest side to 3,713 (10.3 per 100 housing units) in a predominantly white Zip Code region on the northwest side. The correlation matrix in Table 1 indicates that minority composition is the variable which correlates most highly with both the voluntary market activity (r = -.78, so as the minority composition of a Zip Code region increases, the voluntary insurance market activity decreases) and with the involuntary market activity (r = +.72, so as the minority composition increases, the involuntary market activity increases). Both correlations are significant at the .001 level. The other variables also correlate significantly with current marketing activity, except for theft and involuntary market activity, but none are as strongly associated with industry practices as is minority composition.

Even more revealing, however, are the relationships that exist once the relationships fire and theft rates have with regions' minority composition, age of housing, and income, have been re

6. There is still debate over the meaning of tests of statistical significance when the data base is not a selected sample from an explicitly defined universe. When the total universe is included, of course, statistical significance becomes irrelevant. The data on which this study is based falls somewhere in between a total universe and a typically small representative sample. As indicated in the text, the companies included in this analysis accounted for 70 percent of the homeowners policies written in the city of Chicago over a specified time period. An interpretation of this data base as constituting the total universe would preclude attaching any meaning to the significance levels reported or the fact that a given relationship is, or is not, statistically significant. However, an interpretation of this data base as a sample of a larger universe would view the reports of statistical significance as rough measures of the reliability of the relationships and the likelihood that they were chance occurences. To acknowledge such possible differences of interpretation, we include the results of the significance tests which were calculated, as well as the correlation coefficients.

=

moved.' The correlation between minority composition and voluntary market activity remains statistically significant (r = − .36, p < .05) as does the correlation between minority composition and involuntary market activity (r = .41, p < .05). Similarly, the relationship between age of housing and voluntary market activity is significant (r -.46 p < .05) as is the relationship between age of housing and involuntary market activity (r = .31 < .05). Income is significantly associated with voluntary market activity (r .44 p < .05) but the relationship between income and involuntary market activity (r − .18) is not statistically significant.

=

=

7. Part correlations were calculated to measure the relationship between a predictor or independent variable (e.g., minority composition) and a criterion or dependent variable (e.g., voluntary market activity) after removing that part of the predictor variable associated with another predictor variable (e.g., fire).

Both first order part correlations (which remove the effects of one predictor variable) and second order part correlations (which remove the effects of two predictor variables) were calculated.

The formula used to calculate first order part correlations between a predictor variable (e.g., minority composition) and a criterion variable (e.g., voluntary market activity) after removing that part of another predictor variable (e.g., fire) which is associated with the first predictor variable (minority composition) is:

Where variable 1 = minority composition; variable 2 = voluntary market activity; and variable 3 = fire. The formula used to calculate second order part correlations in which the effects of two predictor variables (e.g., fire and theft) are removed is:

Where variable != minority composition; variable 2 = voluntary market activity; variable 3 = fire; and variable 4 theft,

Due to the high intercorrelation among these variables, muitiple regression analysis did not answer the key question under examination, that being whether minority status, income, or age of housing, independent of

=

Urban Decline or Disinvestment

=

=

89

Given the relationship between minority composition and income (r -.70 p < .001), a question arises as to whether or not the relationship between minority composition and current marketing activity simply reflects the lower economic status of minorities rather than minority status per se. But the correlation between minority composition and voluntary market activity remains significant (r - .39 p < .05) even after the relationship between minority composition and income, is taken into account. Similarly, the relationship between income and voluntary market activity, controlling for minority composition, remains significant (r .33 p < .05); that is, both income and minority composition are related to current voluntary market activity, independent of each other. That is not the case, however, with involuntary market activity. While minority composition is significantly related to involuntary market activity (r = .34 p < .05) controlling for income, income is not significantly related to involuntary market activity (r -.22) after the relationship between income and minority composition is removed. In other words, the concentration of FAIR Plan policies in minority communities (and therefore historical voluntary marketing practices) cannot be accounted for by their association either with fire or theft rates or with the income levels of those communities.

=

In the city of Chicago, then, the variation from community to community in current insurance underwriting practices is related to the minority composition, age of housing units, and family income of residents in those (Zip Code region) communities. These relationships are statistically significant and they hold even when that portion of these variables associated with fire and theft rates is removed. While the relationship between minority composition and current voluntary market activity is accounted for in part by the lower economic status of minorities, that is not the case with current involuntary market activity. The current concentration of FAIR Plan policies in minority areas reflects historical constriction of the voluntary market in these areas. These patterns cannot be explained by their association with rates of fire, theft, or with income levels. In Chicago, the disparate impact of industry practices on older, poorer, and minority communities, exists independent of the two major causes of loss-fire and theft.

These findings confirm technically what many urban dwellers have personally experienced. Property insurance is more difficult to obtain in older urban communities, particularly those with a high concentration of minority or lower income residents: insurance redlining is an objective reality in the city of Chicago. Undoubtedly there are problems of perception and communication, as the industry generally contends. But there is also an objective problem of unfair discrimination which the industry generally does not recognize. The historical systematic, institutional denial of vital insurance services has contributed to the deterioration of many neighborhoods, and threatens to continue doing so if appropriate action is not taken.

Policy Implications

A number of steps have already been taken in attempts to alleviate insurance availability problems. Several states including Missouri, Maryland, Michigan, Minnesota and Illinois have enacted antiredlining legislation. Many others, including Washington, Indiana and Ohio are currently considering such action. Several federal agencies have launched investigations into insurance industry practices, congressional hearings have been held in these matters, and legislation has been introduced to curtail abusive practices (Hearings Before the Subcommittee on Citizens and Shareholders Rights and Remedies of the Committee of the Judiciary, 1978: U.S. Commission on Civil Rights, 1979). The industry has not stood idly by. Most public statements by industry representatives defend current policies and practices, but many voluntary efforts have been

the loss related factors, accounted for any of the variance in underwriting activity, or whether the implications of these factors could be accounted for in terms of the disparate fire and theft rates. In the regression equations, the first variable, no matter which one was selected, accounted for most of the variance, with little left to be accounted for by the remaining variables.

90

SQUIRES, DEWOLFE AND DEWOLFE

started to improve communication with consumers and to satisfy individual consumer complaints brought to their attention (Advisory Committee to the NAIC Redlining Task Force, 1978).

In most states experiencing insurance availability and redlining problems, additional legislation would be productive. To eliminate some of the more odious practices, states should adopt amendments to their unfair trade practices acts which would prohibit insurers from: using geographic area as a basis for refusing to insure or for varying the terms under which insurance is offered; using age of the property as a basis for varying the availability of insurance; asking applicants if they were ever denied coverage by another insurer or been insured through a FAIR Plan (or varying insurance availability on these grounds); and refusing to enter into a contractual relationship with an insurance agent because of the geographic location of the agent or the agent's customers. Other amendments that would also have a favorable impact would be ones requiring insurers to: provide explicit reasons for decisions to decline an application or to cancel or to not renew a policy, and information on what the applicant or insured must do to become insurable; continue to service the policies placed by an agent with a company for at least one year after it terminates any agency's contract with it; and provide all applicants and insureds with copies of any information the company has received regarding that individual (including the source of the information and the names of those with whom the information has been shared), and an opportunity for the individual to correct any inaccurate information.

One barrier to resolving insurance availability and redlining problems is the assumption that there must be a trade-off between the profit interests of the industry and the insurance needs of consumers. This is an assumption shared by industry and most critics alike. Few consumers question the right of the industry to make a reasonable profit, and most community organizations which have been fighting the insurance redlining battle explicitly acknowledge that any solution must provide for adequate industry profits (Paul and Baker, 1977: 98-99; Public Technology Inc., 1977: 4). More innovative solutions can become apparent if the issue is posed as one of the developing alternative ways for a group of people to pool their money, to spread the financial burdens of their losses and to provide more adequate protection and peace of mind, rather than as one of striking some appropriate balance between the competing profit interests of the industry and the insurance needs of communities and the society.

One possibility is the concept of a municipal insurance program developed by the Institute for Local Self Government (1977) in Berkeley. Under a municipal insurance program, fire prevention, suppression and insurance activities would all be carried out as part of one integrated system. Presently fire prevention and suppression are generally carried out by local municipalities while insurance is provided by private industry, whose major interest is not loss reduction but assuring that premium dollars exceed losses and other expenses. When all three functions are combined within one system, insurance costs can be directly reduced through fire reduction activities, thus increasing the incentive for individuals to take greater precautions. Further loss reductions would be achieved by savings in funds now going to stockholders as dividends or to agents as commissions, which could be used instead for a variety of fire reduction activities. Money that normally is accounted for as company profits would be used, for example, to purchase fire alarms for apartments, to increase inspections of homes, or for other purposes deemed appropriate by the municipality. Such actions would reduce fire losses and, eventually, fire insurance costs. Funds collected above and beyond those necessary to pay claims could be used to reduce insurance premiums or perhaps to cover all fire suppression activities, thus eliminating a major line item from the general fund.

Various kinds of public insurance programs have already been successfully operated in the United States and Europe (Institute for Local Self Government, 1977:25; Pfennigstorf, 1977: 285-290, 306; Bernard, 1976). With a grant from the Commerce Department, the Institute for Local Self Government is currently developing detailed guidelines for implementing a municipal

Urban Decline or Disinvestment

91

insurance program, with the intention of establishing a demonstration project in at least one of ten cities which have expressed an interest in the concept.

Nonprofit, community based insurance services, outside the jurisdiction of any municipality, are also being considered in some cities. Following the lead of some private corporations and government agencies which have found it more efficient to self-insure or establish their own nonprofit insurance company than to pay increasingly escalating insurance premiums (if insurance is available at all), a few community leaders are seeking alternatives to the private insurance industry for property insurance. Again, the basic difference between these proposals and the programs of a traditional insurance company is that surplus revenues would be reinvested back into the community in part for specific safety and loss reduction activities rather than paid out to sales agents and stockholders (Burack et al., 1979).

These findings demonstrate that insurance redlining is a very real problem and one with critical implications for central city communities, particularly for minority residents in those areas. The problem will not go away simply by better educating and communicating with consumers. Progressive action can and should be taken by public officials at all levels of government, by the insurers themselves, and by community organizations. There is a need for new legislation, more effective enforcement of the law, and for creative alteratives to traditional industry practices.

RESEARCH IMPLICATIONS

The insurance industry is not a passive or neutral force in society. In both its underwriting and investment capacities and activities, it exercises considerable influence on the development (or deterioration) of communities. By making essential insurance more difficult to obtain in central city neighborhoods, particularly minority neighborhoods, the industry creates barriers to equal housing and business opportunities for the nonwhite population. When a company uses its considerable financial resources to invest, for example, in a suburban shopping mall as opposed to a central city housing development, that decision certainly has considerable ramifications for the growth and decline of neighborhoods within that metropolitan area—not just influence on the profit margin of that particular company. Whether the net effect of insurance industry practices is positive or negative may be a debatable issue; but the industry does shape societal conditions, not just respond to them.

We strongly suspect that as the role of the insurance industry becomes better understood, particularly in its relationships with other institutional forces such as the lending and real estate industries, the structural/disinvestment perspective will be seen as offering a much more comprehensive and accurate way of understanding the uneven urban development than will the individualistic/natural evolutionary models. The accumulation of social science research is chipping away at the latter approach, and the foremost reason is that it is more and more evident that the uneven development of metropolitan areas is clearly a social phenomenon (Willhelm's work here, e.g., 1962, deserves much more attention than it has usually received from urban sociologists).

The idea that the quite evidently uneven development of urban areas is the result of the formation of natural communities or of inevitable market transitions denies the reality of social processes. Buildings do not naturally deteriorate simply because of age, and there is obviously no natural reason why buildings of a given age deteriorate more rapidly in one neighborhood than another. There are many well-maintained (and extremely expensive) homes in suburban Winnetka as old, or older, than many which have crumbled in Chicago's west side, yet present and future residents of Winnetka will have little difficulty obtaining mortgage or home