The shifting rhetoric of insurance denial

I. Introduction

As recent court cases and academic studies have revealed, discrimination against certain groups by the insurance industry still remains an unfortunate practice (United States v. American Family Mutual Insurance Company [1995]; "Insurer To Revise Its Urban Business," New York Times, 1 Feb. 1997; Treaster 1998). By definition, discrimination occurs when two otherwise identical individuals are treated differently by virtue of a particular characteristic. Paired testing using black and white applicants has revealed that illegal discrimination (i.e., using distinctions based on criteria banned by law) can take many forms, from agents refusing to return customer phone calls, to offering higher prices and weaker policies, to denying outright coverage for members of unwanted groups (Lynch 1997; Smith & Cloud 1997; but see Wissoker, Zimmermann & Galster 1998). Discrimination can also take the more subtle form of the insurance company being slower to handle the claims of ethnic minority claimants than of whites (Baker & McElrath 1997; Chan 1999). Insurance companies desire certain groups more than others as customers, and as a result, those who do not fit the underwriter's vision of the ideal member of society, such as certain racial and ethnic minorities, the poor, gays and lesbians, and people with alternative lifestyles, can have a difficult time obtaining desired coverage.

The biggest concern about discriminatory practices for those trying to eradicate them is how difficult they are to detect. Depending on the state, insurance commissions can be grossly under-funded, and in any event they tend to focus almost exclusively on the financial aspects of the industry, such as whether companies have sufficient loss reserves and whether their rates are adequate (Abraham 1995: 99). For most discriminatory practices to be exposed then, the affected group must either register a complaint with the state's insurance commission or file a suit in court. These individuals must somehow first become aware that they are being discriminated against, and this can be difficult to do if they do not understand the underwriting process or if they are unaware that the insurance company categorizes them in a certain manner (Austin 1983; Simon 1988). In fact, they may not even realize that they have been discriminated against at all, since their agents will most likely manage to cover up the effects by placing them in other, less advantageous, programs.

Underwriting is the process of determining which applications the company should accept, and for those who are accepted, in which program they belong. The process of risk selection (as underwriting is also known) has two faces, one that is presented to regulators and applicants, and a second that is used by underwriters. I argue that the Janus-like aspect of underwriting is what has allowed a great deal of discrimination to continue for as long as it has. The outward face is one of numbers, statistics, and objectivity. The inward face is that of narratives, character, and subjective judgement. The rhetoric of insurance exclusion-numbers, objectivity, and statistics-form what I call "the myth of the actuary," a powerful rhetorical situation in which decisions appear to be based on objectively determined criteria when they are also largely based on subjective ones.

The remainder of this section explains how insurance companies make use of narratives about applicants in their underwriting decisions and the disparate impact this can have on members of different groups. Section II covers the decades surrounding the turn of the twentieth century, when insurance companies formed guidelines that mirrored social cleavages of the era. Rather than accept or decline applicants solely on the basis of factors such as health and age or the fire risk of their homes, insurers made distinctions based on the applicant's character and morality. They constructed narratives to explain how certain applicants lacked the necessary character to be insured properly, and hence needed to be excluded from the risk pool. When state legislatures eventually began requiring insurers to validate their criteria with statistics, the narratives about character and morality were submerged in the language of economics and statistics, where they remained influential but unseen.

Section III examines contemporary underwriting guidelines. What we see is that insurance underwriters still attempt to understand where the applicant fits into society, only now the criteria they use are not explicitly told to the public.

The final section reveals the rhetorical process by which the subjective aspects of discrimination are buried under a veil of economic objectivity, which I call "the myth of the actuary." The myth of the actuary is predicated on the economics literature that states that insurers should be expected to provide coverage for any group that they can write profitably. Thus, the current approach to regulation focuses mostly on the actuarial data provided by the companies, ensuring that rates are justified and that marketing practices are fair. But fair marketing and justified rates are irrelevant if applicants are not accepted for coverage due to criteria that are never tested statistically, nor approved by the polity. The failure of regulation to eradicate the present methods of illegal or socially unacceptable forms of discrimination calls for a new approach to regulation, which I call "the narrative approach." The narrative approach to insurance regulation focuses on the subjective side of risk selection, forcing insurance companies to reveal the stories they tell about applicants during the underwriting process. Once these stories are brought to light, they can be challenged and tested, allowing market forces to function properly, while also facilitating a discussion of what underwriting criteria are socially acceptable.

The Importance of Insurance to Society

Insurance has been helping individuals, families, and commercial ventures protect their assets from accidental loss literally for thousands of years. Burial insurance traces its roots back to at least as far as the Roman Empire, being provided through quasi-- religious burial societies known as collegia. Marine insurance is equally as venerable, having provided protection to sea captains plying the Mediterranean waters in search of trade with ancient city-states. Although life insurance had a difficult time finding acceptance in America until the 19th century, fire insurance was being provided commercially as early as the 1790s. The reason that insurance has been around for so long is obvious: owning property, engaging in trade, and pursuing the good life are simply much easier to do when one has the comfort of knowing that his or her burdens will be shared by others in the unfortunate event of a loss. Insurance gives a peace of mind that few other financial instruments can confer. Yet it does more than just facilitate the good life, insurance also defines the boundaries of it. Without insurance one cannot get a mortgage, legally drive a car, or share the burdens of health care costs with others. In contemporary America, being a member of a community of risk is no longer merely an opportunity, it is a necessity.

Given the importance of insurance to individuals and their families, the question of who is allowed to share their risk and who is excluded becomes drastically important for a polity. The issue is simple when coverage is universal, since no one is excluded. The situation becomes more complicated once we drop below that level, since one must then create distinctions that group some together and exclude others. The method of determining which factors to use and which to ignore is rarely obvious and is highly political (Calabresi 1970; Douglas & Wildavsky 1982), thus statistical methods alone are of little use in determining which criteria to use. In an authoritative actuarial text, Walters (1981) states that the key task for insurance companies is to find variables that group similar risks in such a manner that there are statistically significant loss differences between the rating classes. But from a social science perspective, this method in itself is clearly insufficient, since categorizing applicants along certain criteria may be unacceptable to the polity. For example, Americans have decided to prohibit the use of distinctions based on race, regardless of whether they are significant. Distinctions based on gender elicit occasional challenges (cf. Geduldig v. Aiello [1974]; General Electric Co. v. Gilbert [1976]; City of Los Angeles Depa-ament of Water and Power v. Manhart [1978]; Hartford v. Commonwealth Insurance Commissioner [1984]), and the use of residential rating categories (in which rates are determined by the location of the home) also periodically draws criticism for being illegally discriminatory in its effects (City of Los Angeles v. Farmers Insurance Exchange [1982]).

If the criteria companies use to sort and select applicants into their various programs are explicitly stated and statistically supported, then any discrimination that does exist is at least justifiable from an economic standpoint. It is then up to the polity to decide whether the categories are acceptable. This article focuses on a key issue that prevents market forces from functioning properly, namely that the criteria used by many insurance companies to determine who is allowed to share their risks with others is neither explicitly stated nor statistically tested, and therefore will not be corrected by market forces. The contemporary focus on numbers masks the subjective nature of underwriting-the very place that holds the greatest potential for illegal discrimination.

The Subjective Nature of a See%iy Objective Process

The issue of whether or not an applicant is acceptable for coverage depends on the methods the insurance company uses to categorize and label that individual. To categorize individuals in one manner and not another is often anything but an objective process. For example, in the 1960s life insurance companies were well-aware that smoking had deleterious effects on health and longevity but intentionally chose not to include it as a rating factor (Huebner & Black 1969: 477), while today whether the applicant smokes is a highly significant factor. Currently, some home insurers rate on whether the house has a wood stove or not, while other insurers ignore this question completely. Although the question of whether or not the applicant is a smoker or uses a wood stove to heat his or her house are themselves objective, the decision to base coverage on those questions instead of focusing on other potentially significant factors is not, and this decision becomes significant when the criteria have a disparate impact on certain groups. Depending on the type of insurance for which the applicant is applying, health, life, and property-- casualty underwriters can have the leeway to decline coverage on such factors as whether the applicant is a smoker, lives with an unmarried domestic partner, has seasonal employment, resides next to a vacant building, or owns a pit bull terrier. On one level, each of these factors might contribute to a loss, and hence could contribute to the actuarial soundness of an insurance program that used them. On another level, each of these factors can be used to define the applicant's position in society. They draw boundaries between those who are the "Accepted" members of society and those who form the "Other." The guidelines, which at first glance appear to be objective, may actually be informed by highly subjective notions that are used to exclude the Other from sharing their risk with the Accepted.

The Making of an Insurance Underwriter

The categories that insurance companies utilize and the manner in which the underwriters implement them is often largely based on narratives describing the characteristics of various groups. When I was going through underwriter training in the mid-1990s for a preferred risk property casualty company (one that insures the homes and autos of highly desirable "low risk" individuals), I was taught the process of "character underwriting." Rather than simply applying a set of objective characteristics (such as age, years driving, and claims history) to a formula, character underwriting is an attempt at a holistic examination of the applicant's job, possessions, and lifestyle in order to assess whether he or she fits a vague set of standards known as underwriting criteria. For my company, the underwriting criteria was whether the applicant was "mature, stable, and responsible." Mature, stable, and responsible individuals were those who comfortably fit the stereotype of the middle-class American: acquisitive, hard-working, respectable, moderate, and mainstream. (For a more precise list of criteria, see Section III.) Those who did not meet the "mature, stable, responsible" test were denied coverage.

Those who fit the underwriters' conception of mainstream middle class will have a much easier time finding coverage from a preferred-risk company than those who do not.2 For example, how might a preferred-risk company handle new drivers? The "experience story" assumes that new drivers are more likely to get into accidents than are experienced drivers; hence, underwriters should not grant coverage to anyone who has not had at least two years of driving experience. After going through two full seasonal cycles of wet leaves, snow, rain, and glare from the sun, drivers eventually learn how to drive under a wide variety of conditions. Underlying this guideline is a view of causation that says inexperience leads new drivers to mishandle situations that more experienced drivers would know how handle properly. But not all preferred-risk companies (including the one for which I used to work) underwrite solely in this manner, because there was another story, the "responsibility story" that also needed to be considered. The responsibility story allows for the following exception, "no new drivers except those who are coming out of professional or graduate school, such as doctors, lawyers, and engineers." The reason: these individuals had to exhibit a significant amount of maturity and responsibility to get through graduate school, and now they have something to lose should they drive recklessly. The perspective has changed to one stating that losses occur not only because the driver lacks experience but also because he or she is seen as irresponsible.

Although the experience story affects all new drivers equally, the responsibility story mostly excludes those at the lower end of the socioeconomic ladder, and this exclusion is not by accident. Members of the insurance industry are instructed by textbooks to sort applicants according to characteristics that demarcate the mainstream lifestyle from the alternative or less financially successful ones: "Insurance presupposes a yearning for achievement, a drive of acquisitiveness, a desire to preserve the valuable, a bit of apprehension about the future, a readiness to obey the law, a sense of honesty, a fondness for tradition, a willingness to accept personal responsibility and accountability, and a measure of charity towards others in society" (Long 1971: 43).

Those at the lower end of the wage scale are less likely to meet these standards than those at the top, even though, as far as actual loss experience goes, there might not actually be any difference between them. Thus the issue of whether members of a particular group are able to find coverage will be influenced by the narratives told about them. Companies using the responsibility story will exclude a far different group of applicants than the companies using the experience story, and actuarially derived numbers will not be the deciding factor.

Methodology

In addition to published underwriting guidelines, I draw extensively from insurance underwriting textbooks. My assumption is that these manuals reflected the practices of the industry at the time of their publication and that they were intended to teach the rising generation of insurance underwriters to make distinctions among individuals in a particular manner along particular lines. The extent to which trainees actually proceeded to underwrite in the manner prescribed is an issue that I do not address empirically. It is certainly a fact of history, however, that in the past many insurers overtly and explicitly discriminated against many of the ethnic and minority groups mentioned later. The contemporary underwriting guidelines listed in Section III provide strong evidence that these practices continue to be used in a widespread manner to the current day.

II. Keeping the Immoral at Bay

Character underwriting has existed since the earliest days of life insurance in America, which coincided with the beginning of the Industrial Revolution. In a time of tremendous social upheaval, insurers struggled to create underwriting guidelines that would help sort out the good risks from the bad. Assessing the general health of an individual was a challenge, and initially, even basic mortality tables were unreliable. Insurers had to invent criteria for judging good risks from bad ones, and rather than simply relying on factors such as age, weight, and general health of the applicant, for example, life insurance companies also developed guidelines that coincided with the social norms of the era. Underwriting standards of the 19th and early 20th centuries were informed as much by definitions of membership in society as they were by ones relating to the actual health of the applicant.

Attempts to provide commercial life insurance began in the 1790s, but met with little success because of concerns that insuring one's life ran counter to the teachings of Christianity. It took the creation of programs covering the lives of Presbyterian and Episcopalian ministers finally to break the ice, and by the middle of the 1800s the life insurance industry was growing at a stunning pace. In the two decades between 1840 and 1860, insurance grew from just fifteen companies writing under $5 million in coverage to 43 companies writing almost $205 million worth of insurance (Zelizer 1979: 6). The phenomenal growth of insurance continued straight through the Civil War, with the amount of insurance purchased tripling between 1860 and 1865 alone (p. 6). With the onset of the Industrial Revolution and the concomitant movement off the farms and into the cities, life insurance became almost a necessity since families of workers could no longer rely on farm income when the main breadwinner passed away unexpectedly.

From its introduction, insurance was heavily purchased by the poor in order to provide for a proper burial. One contemporary author noted, "Only those who are familiar with the life and the labor of the industrial masses can fairly grasp the deeper meaning of the abhorrence of a pauper burial of a member of the family in the potter's field" (Hoffman 1900: 4). It was not only the urban industrial poor who demanded insurance, but the rural agrarian poor as well. As Weare (1973: 183) notes, sociologist Hylan Lewis (1955: 45) found that, in South Carolina, purchasing insurance was as common a cultural practice as was "church going, cotton, whisky, burying, hunting and fishing." Hortense Powdermaker noted that the insurance envelope hanging on the wall was an omnipresent feature of cabins in the Mississippi delta (1939: 63).

The latter half of the 19th century saw unprecedented immigration and population growth in the cities. Between 1860 and 1900, the urban population grew from roughly six million to 30 million residents, with most of the growth coming from rural migration (Census 1949a). However, even the vast movement of farm workers into the cities could not satisfy the need for labor, and factories began pulling in massive numbers of immigrants from abroad. For the same 40-year period, more than six million foreign-born individuals migrated to America (Census 1949b), with the result that by 1900 roughly 40% of the population of America's 12 largest cities had been born outside the country (Hays 1957: 95, cited in Skocpol 1995: 103). The vast influx of workers and families from other cultures, complete with their own religions, languages, and lifestyles quickly led to a concern by those already established about what it meant to be an American. The wealthy funded settlement houses in immigrant neighborhoods not only to lift the immigrants out of poverty but also to "Americanize" them, creating individuals who would act, speak, think, and work just like the northern European immigrants who had come before them (McClymer 1980). Needless to say, the immigrants proved to be remarkably resilient to change (see Scott 1998), and continued to look, act, worship, and speak in a manner that threatened to alter the definition of who was part of society. The newly-freed slaves also moved off the farms and into the cities at a very high rate. Since the inception of slavery, one excuse for its existence was that freedmen were an innocent people in need of protection by the "benevolent" white slaveholder, who would watch over their souls. After Emancipation, the narrative surrounding blacks rapidly changed from one of blacks being innocent children-of-God to one in which they were rapacious predators, threatening the white race through both the avoidance of work and the rape of white women (cf. Dixon 1902, 1905). Many urban whites simply were not prepared for the unexpected level of exposure and interaction with blacks that industrialization and urbanization brought about.

Thus the urban centers that held the headquarters of the major life insurance companies experienced tumultuous social change in the decades leading up to the 20th century. Massive immigration of rural farm workers, blacks, and immigrants changed the face of the cities and challenged the established definition of who was a member of society and who was not. This debate immediately found its way into the insurance community as well, and was reflected in the language of insurance exclusion of the era. Given the novelty of life insurance, companies had to invent rules that would help their agents and underwriters determine which applicants to accept and which to reject. While a physician's examination helped weed out those with preexisting conditions, companies still feared that there was another class of applicants who would threaten the risk pool because of their character. The real question was how to sort out these bad risks from the good ones.

The Threat of the Moral Hazard

A central concern for the members of any insurance company is determining which applicants they can accept and which ones they cannot. Those who can be covered are labeled as "insurable" and those who are not are "uninsurable." As mentioned previously, the process of sorting applicants into the two categories is known as underwriting and is done at the time of application, and also when policies come up for renewal. Underwriters use a set of guidelines produced by the company, and the standards may differ from firm to firm, allowing the "insurability" of a given individual to shift depending on the company with which the application is placed.

If an applicant is deemed insurable, he or she is then placed in a class with others who have similar characteristics, allowing for the "law of large of numbers" to take effect. The law of large numbers works something like this: Given a group of individuals who are similar in the characteristics that cause losses (such as age, weight, dietary habits, smoking, dangerousness of job, etc.), we may not be able to say that any one particular individual is going to suffer a loss, but we can be confident that a certain percentage of the individuals in the class will, allowing for the determination of accurate premiums. As long as individuals in the group are statistically similar and are equally risk-averse allowing for an accurate premium to be determined, the rating class will be considered "actuarially sound," and the insurance company should be able to insure them profitably. The law of large numbers fails when applicants who are not similar to the others are allowed into the class, or when there are individuals who are not as risk-averse as the others. The latter group are known as "moral hazards."3 Thus insurance companies place great importance on proper underwriting to avoid providing coverage for moral hazards.4

Why "moral" hazards? The idea was that there were certain types of individuals whose characteristics or immorality led them to be less cautious, or worse, to cause losses in search of payments (cf. Smith, Trieschmann & Wiening 1987: 208; Vaughan 1992: 629). The immoral homeowner might try to burn her home down for the insurance money, or the dishonest ship captain might dump his cargo, knowing that it was covered. The Aetna Guide to Fire Insurance for the Representatives of the Aetna Insurance Co. (Aetna 1867: 21, also cited in Baker 1996: 250) wanted its agents to be the first line of defense against allowing such individuals into the insurance pool, instructing them to ask

What is the general character borne by the applicant? Are his habits good? Is he an old resident, or a stranger and an itinerant? Is he effecting insurance hastily, or for the first time? Have threats been uttered against him? Is he peaceable or quarrelsome-popular or disliked? Is his business profitable or otherwise? Has he been trying to sell out? Is he pecuniarily embarrassed? (emphasis in original)

The moral individual was a responsible one who kept his finances in order. Thus, moral hazards were also depicted as individuals whose irresponsibility led to financial problems, and who saw an insurance payment as a way out (Huebner, Black & Cline 1976: 636; Heimer 1985: 35). Richard M. Bissell (1904: 155) explained that "[f]inancial embarrassment and the pressing necessity for ready cash often create the most serious kind of moral hazard. A merchant with notes overdue or who sees failure ahead, or a farmer who cannot pay interest on his mortgage, is often in a position where the ready money obtainable from his insurance policies, even if not equal to the value of his property, would nevertheless help him tide over a pressing emergency."

The difficulty for the insurance underwriter was determining exactly who the immoral were, since in most cases financial problems or immorality might be well-concealed. Bissell (1904) warned that moral hazards "are indefinite, incapable of analysis, separation, or estimation, yet they are of the greatest importance." (p. 154). One trick for the underwriter was to take a shortcut, eliminating entire groups that were considered more likely to harbor moral hazards than other groups were. One insurance textbook noted that the underwriter's job was made easier by the risk report that contained information "as to [the applicant's] racial descent . . . and it must be specifically stated whether he is Anglo-Saxon, Greek, Hebrew, Italian, Negro, or of other racial or national origin" (Magee 1951: 77). Knowing the applicant's nationality, one can only suppose, provided the underwriter with useful information about whether or not the applicant was a good risk.5 The highly respected Insurance Monitor noted that Jews needed to be underwritten with caution because of fire hazards: "There are honorable Jews as there are honorable Gentiles, but that the evil disposed among the race to gravitate to incendiarism is a notorious fact, and the underwriters who close their eyes to moral hazard wrong the companies and wrong their communities" (Insurance Monitor 1907, also cited in Baker 1996: 250).

Good Christian clergymen were a different story. Stone (1993: 297) quotes the following passage from the Transactions of the Society of American Actuaries: "Take, for example, a clergyman, an occupation which is conducive to longevity, whose build is most favorable, whose family is very long lived and whose habits are first class... Undoubtedly the stock from which such a risk springs has expressed its moral and its physical energy in the occupation and the temperate life of this individual" (see also Rogers & Hunter 1919: 71-72).

It is clear from many of these passages that some groups were classified as good or bad risks based on little more than prejudice. From the insurance companies' perspective, the difficulty with this approach was that different underwriters might not universally hold the same stereotypes. Arguments based on hearsay are only effective when they are widely subscribed. In 1896, a work came out that took a different approach, attempting to use statistics to prove immorality. The result of this new method was a publication of the American Economics Association entitled "Race Traits and Tendencies of the American Negro," by Prudential statistician (and future president of the American Statistics Association) Frederick L. Hoffman. In the book-length article, Hoffman (1896: V) attempted to prove "scientifically" that the immoral traits and tendencies of the black population were going to lead to their eventual extinction:

At the commencement of my investigation, especially in regard to longevity and physiological peculiarities among the colored population, I was confronted with the absence of any extensive collection of data free from the taint of prejudice or sentimentality. Being of foreign birth, a German, I was fortunately free from a personal bias which might have made an impartial treatment of the subject difficult.

Hoffman made use of a wide array of tables, statistics, citations from "authorities," and rhetoric to convince the reader that the true cause of the increase in blacks' mortality since Emancipation was not their poverty or unhealthy living conditions, but rather their excessive immorality. Data on lynchings was used to show that immorality was on the increases and since immorality led to increase in illness, it was clear that slavery was much better for the health of the black man than freedom (Hoffman 1896: 236) .7 With blacks, attempts at education were a waste of time (p. 249), since decades of freedom had only proved that the race was incorrigible (p. 241). The following paragraph succinctly summarizes Hoffman's (1896: 95) argument, and also the tone of the entire work:

For the root of the evil lies in the fact of an immense amount of immorality, which is a race trait, and of which scrofula, syphilis, and even consumption are the inevitable consequences. So long as more than one-fourth (26.5 per cent. in 1894) of the births for the colored population of Washington are illegitimate,-a city in which we should expect to meet with the least amount of immorality and vice, in which at the same time only 2.6 per cent. of the births among the whites are illegitimate,-it is plain why we should meet with a mortality from scrofula and syphilis so largely in excess of that of the whites. And it is also plain now, that we have reached the underlying causes of the excessive mortality from consumption and the enormous waste of child life. It is not in the conditions of life, but in the race traits and tendencies that we find the causes of the excessive mortality. So long as these tendencies are persisted in, so long as immorality and vice are a habit of life of the vast majority of the colored population, the effect will be to increase the mortality by hereditary transmission of weak constitutions, and to lower still further the rate of natural increase, until the births fall below the deaths, and gradual extinction results (emphasis in original).

The goal of Hoffman's article-praised in the statistical journals (Dawson 1896) and declared by one of its critics as "the most important utterance on the subject [of the black situation] since the publication of "Uncle Tom's Cabin" (Miller 1897: 1)-was to prove that blacks were uninsurable. If they really were dying out, especially if the cause were immorality, then blacks had no business being insured in the same risk pools as white applicants. Although by 1894 many states made it illegal to charge blacks higher premiums solely based on their race, Hoffman (1900: 153) reassured his readers that "the companies cannot be compelled to solicit this class of risks, and very little of this class is now being written" (also cited in Stone 1993: 313-14).

Immorality as the Tool of Exclusion

In an era of tremendous demographic changes, the insurance industry struggled over the question of whom to include and whom to exclude from the community of risk. The way in which certain groups were eventually labeled as moral hazards mirrored their perceived place in society. Old residents, AngloSaxons, and those who were peaceable and popular and who obeyed the norms of morality also happened to be those who were considered insurable. Not only were strangers, itinerants, those who had been threatened (perhaps for challenging the status quo?) or who were "pecuniarily embarrassed," Negroes, Chinese, Japanese, and Mexicans considered threats to other policyholders, they were also the very same groups who were considered threats to society.

Underwriters often had little information about the applicant at their disposal other than his race and socioeconomic status (women rarely purchased insurance unless they were the bread-- winners). There was very little actuarial data on which to base many decisions, and what did exist often seemed to be highly unreliable. Rather than collect data on the loss patterns of different nationalities to see if there really were any differences, many companies simply excluded these individuals from coverage altogether. In time this came to cause the insurers problems. Because guidelines were based on perceptions rather than data, companies found them hard to sustain under the increasing scrutiny of state legislators, who began looking for votes from many of these excluded groups. Since they automatically denied coverage to many of these groups, the insurers never had the opportunity to develop loss data that could be used to support their arguments.

It is clear from several contemporary reviews of his article that Hoffman failed to convince everyone that blacks were too immoral to insure. Reviewing the article for Political Science Quarterly, Professor Gary N. Calkins (1886) of Columbia University accepted the mortality data as correct, but argued that Hoffman's own charts suggested that the mass migration into the unsanitary urban slums alone might account for the numbers. Kelly Miller (1897) also quite thoroughly countered Hoffman's arguments regarding the demise of the black race, at one point noting that the mortality of black Americans in New Orleans was still lower than the death rate of Germans in Munich, Kronigsberg, and Breslau (p. 13).

Although his use of statistics failed to convince many scholars, the lasting significance of Hoffman's article is that it redefined the standards of insurance rhetoric. "Race Traits and Tendencies of the American Negro" was an attempt to support normative claims with objective data. In the future, underwriting guidelines would only present the objective side to the polity, keeping the subjective component out of the public glare. While the concepts of character and morality would still inform underwriting decisions, they would from then on be veiled in the language of threats to the actuarial soundness of the rating classes. The rhetoric of insurance exclusion changed to appear more scientifically based, keeping the underlying narratives about groups buried and undisclosed.

III. Statistical Data as a Tool of Exclusion

By the latter half of the 20th century, the process of classifying risks into categories that predict statistically significant differences had developed into a true science. Actuaries now train for years and sit through a series of incredibly demanding exams in order to learn their trade. Complex rating systems can have as many as 234,360 categories in which an applicant can be placed (Austin 1983: 547), and in theory, each category must have its rate supported through statistics. The idea of judging an applicant by her or his race or standing in society appears to have been replaced by mathematically justified matrices that rate individuals according to their risks and charge them the appropriate premium. But even though the stereotypes have formally disappeared from the rating systems, they still exist and are used to exclude certain groups from coverage." Rather than replacing these stories about undesired groups, the numbers, data, and forms merely hide the fact that applicants are still judged according to their standing in society. The difference between the old era of underwriting and the current one is that the appearance of subjectivity has been hidden behind a process that appears objective, mitigating the denied applicants the opportunity to develop an awareness that they have been excluded on the basis of subjective opinions about their lifestyles.

The Application as an Objective Process

Applying for insurance in contemporary times has the air of a scientific process about it. The agent asks a series of seemingly straightforward questions of the applicants, such as the value of their home, the last time they received a speeding ticket, or if there have been any recent health problems. The answers are recorded on a standard application, and with luck, the agent will phone back shortly to congratulate the applicants that they have qualified for coverage with Company X, and that according to the rating system, they will have to pay premium Y. The unlucky applicants are informed that the company lacks a program to fit their particular characteristics and that they should look elsewhere. In either event, the emphasis is on the fit between the applicants and the insurance program, as if the criteria that went into the creation of the program somehow existed outside the reality of the individuals who are placed in it. From this perspective, the decision to accept or deny the applicants and to assign the premium they are given is nothing more than the result of the fit between the program and the applicants, based on the information provided by the application. The entire process comes across as the objective application of scientific actuarial principles, matching characteristics of the applicants to the appropriate rating categories of the insurance program.

The Inspection Report and the Moral Hazard

Often, the most important information about an applicant is found not on the application but rather on a risk report, described by one older textbook as "a secret investigation of the conduct and habits of an applicant" (Dawson 1911: 175). The risk report can take varying forms, depending on the type of insurance the applicant is seeking. Applicants for health or life insurance will have to sign a waiver to allow their physicians to submit their medical histories, for example. Applicants for homeowners insurance may have to allow for an inspection of the property. Auto insurance companies will pull applicants' driving records directly from the police or Department of Motor Vehicles. While they ostensibly are looking at other factors, these reports are also designed to give the company's underwriters an understanding of the applicant's character, allowing for a focus on the applicant's morality, just as they did in the previous underwriting era.

The risk report begins with the agent, who is frequently referred to as the "frontline underwriter." The agent is expected to develop a detailed knowledge of each applicant, ranging from their health, the condition of their car and the upkeep of their home, to their lifestyle and habits. One guide for agents explains, "In determining whether an applicant for insurance is a standard insurable risk, the underwriter must take into consideration not only physical condition but also financial worth and moral condition as reviewed by a study of habits, environment, mode of living and general reputation" (Levy 1968: 337).

The risk report defines which applicants are part of society, and which remain outside of it. "It calls attention to any bankruptcies and fire losses, and comments on the intemperate use of alcohol and other departures from normal social behavior" (McGill 1967: 402). Another textbook explains that "[i]nvestigations are concerned primarily with the moral character of the prospective insured, in recognition of the fact that moral considerations are often of more importance than physical features associated with a risk situation" (Denenberg et al. 1964: 449, emphasis added).

Even with the introduction of highly sophisticated classification systems, insurance companies to the present day still decline coverage on the basis of applicant's morality. "[M]oral hazard is a subjective characteristic of the insured (or applicant) that tends to increase the probable frequency or severity of a loss.... Some indications of a moral hazard might be a weak financial condition (indicated by a financial report) or questionable moral character (perhaps indicated by a police record)" (Smith, Trieschmann & Wiening 1987: 208).

The term moral hazard has always been a blend of moral and economic components. For an economist, a moral hazard is a condition in which a rational actor faces a set of incentives that reward him or her for acting in a less risk-averse manner once insurance coverage has been granted. This is why underwriters take such special care to ensure that insurance will never pay the policyholder more than the value of the loss. If a policyholder could make money from a loss, this would indeed be a moral hazard situation, since insurance would then actually offer an incentive to file a claim. Contracts are carefully drawn up to guarantee that the insured always has a financial incentive to act in a risk-averse manner. In contrast to the financial component, the moral side, as we have seen, is highly subjective and rarely obvious. Where one is based on rational choice thinking, the other is based on social constructions.

The rhetorical strength of the manner in which the insurance community defines moral hazards is that it conflates the moral and economic components. The economic component is based on a rational actor model in which risk can be mitigated through contract design. However, there remains to the current day a "moral" aspect to the moral hazard, and this component is largely influenced by current divisions in society. In the past era, applicants were aware that they were discriminated against because of their profession or position in society, and in theory they could take legal or political steps to end it. In the current era, discrimination based on subject depictions still exists, only now applicants are not aware of it.

Current Underwriting Standards

A section of a company's underwriting guidelines will list characteristics of the applicant or of the insured item that will disqualify the individual from coverage. The goal is to exclude risks that present higher-than-average opportunities for loss. For example, dangerous animals such as attack dogs can cause extremely large liability losses if they escape from the property and harm someone. Thus, companies have the ability to state that they will refuse to insure any homeowner who owns an attack dog, such as a rottweiler or a pit bull terrier. Companies can also disqualify homes that have nonregistered vehicles on the property, or that are adjacent to vacant buildings. It is certainly the case that these factors may significantly increase a loss. Nonregistered vehicles (i.e., inoperable ones) can be dangerous to children who play in or on them, and the same can be said about vacant homes. Yet, another story can be told about such guidelines.

Many of the seemingly objective criteria used are based on a stylized depiction of the lifestyles of group members who find themselves on the fringe or lower end of society. While children might hurt themselves while playing on a junk car in the driveway, is there any less danger from playing on a motorboat trailered in the same location? The difference between the two is that poorer families are likely to have the junk car and better-off families a boat. Even though the risk posed by the two may not be different, the former group will be denied coverage while the second group will not. We can see class distinctions in the other examples as well. Rottweilers and pit bulls are frequently associated with the type of people who drive motorcycles, have long hair, and sport tattoos. One also finds very few vacant properties in better-off neighborhoods. Thus there are two stories being told. The one about hazards is told to the outside world, while the one about class and lifestyles is not.

The Stories Guidelines Tell

The seemingly objective nature of actuarial tables veils the subjective "character" underwriting process that is used to categorize applicants. If the process were as cut-and-dry as simply matching characteristics to a matrix, there would be no need for underwriters at all. An agent could take an applicant's information, match it to a chart, and guarantee coverage on the spot without it then having to be passed on to an underwriter. There would be no need for the agent to conduct a secret risk survey with the goal of establishing the applicant's character and place in society. There would not be a need to tell stories. Yet agents and underwriters do tell stories, and the applicants are not allowed to participate in their creation.

In 1994, the Texas Office of Public Insurance Counsel (OPIC) conducted a study of the underwriting guidelines various property-casualty companies use to determine whether applicants for homeowners and automobile policies are granted coverage. Below are some of the criteria they discovered.

Location of home

Many companies disqualified homes that fell under the following criteria:

* "Dwelling located in a commercial neighborhood is not eligible."

* "Persons whose property is located in a high crime area are ineligible risks."

* "Dwellings located in deteriorating and/or high crime areas are not eligible."

* "Property in recognized or known theft areas is not eligible."

* "Avoid low valued rental dwelling, rental dwellings in deteriorating neighborhoods" (Powers 1997: 128).

According to these standards, homes in high-crime areas should be associated with larger losses than those in low-crime areas, and thus they should not be written. But what is a "high-- crime" area? Is it an area that is more likely to be the target of crime (which may actually be the wealthier neighborhoods, wherein residents can afford expensive electronics, computers, and art), or is it the neighborhoods in which criminals are believed to live? The first is a story about victims, the second is about victimizers. The victim story may exclude wealthy neighborhoods from coverage, while the victimizer story will most likely exclude poor ones. What begins as an empirical association between location and loss has the potential to evolve into a process of stereotyping residents of a neighborhood on their aptitude for being criminals. In the end, the effect is to allow the underwriters to construct a narrative about the applicants. Instead of excluding the risks most likely to be associated with losses, it becomes their job to exclude those who cause losses, even though those applicants themselves may be the ones least likely ever to file a theft claim. The use of the victimizer story by property insurers may help explain why many in minority neighborhoods have such tremendous difficulty finding coverage (for a similar argument, see Heimer 1982; Klein 1997).

Employment Stability

Personal lines insurance companies have been found to have the following criteria regarding employment:

"Personal and employment stability."

"Evidence of stability and financial stability."

"[Applicant] must be employed or retired."

"Applicant must be employed for four years."

"Favorable employment record" (Powers 1997: 131).

"Applicant must be a person of integrity and financial stability who takes pride in his property" (Squires 1997: 6).

One well-known textbook has the following to say about employment stability:

In most cases where the occupation is characterized by unstable earnings, work performed at home, irregular and seasonal work, or any connection with illegal or dubious activities, underwriters will be hesitant to approve the risk. In the event the personal history of the applicant shows evidence of fraud or bad debts, a poor risk is indicated. Such activities as extramarital entanglements, criminal activity, and poor personal habits all present danger signals (Huebner & Black 1969: 509-10).

The seasonal worker (such as carpentry laborers), the craft worker operating out of the home (such as quilt makers), and those with unstable earnings (such as migrant workers) can be denied coverage because of their employment status, regardless of their ability to maintain their homes and pay their premiums. Thus the story being told here links employment instability with personal instability and a lack of integrity. Those who do not fit the mold of the average worker are thus construed as possessing characteristics that will make them more likely to incur a loss. The wealthy actress who works occasionally will find herself able to purchase coverage for her home, while the carpenter who has the winters off may not, even though both of their homes may be equally well-maintained.

Marital Status

"Moral hazards to avoid: recent divorce or marital problems."

"Applicant must be married."

"[Do not consider applicants] in the midst of separation or divorces."

"All persons over 25 must be individual members unless husband and wife living together" (Powers 1997: 132).

On their face, guidelines requiring all residents to be married may be concerned about litigation losses. Family members may in fact be less likely to litigate against each other than those who are not related, but it is doubtful that many companies have the data to support such a claim. Firstly, if it were true, companies could simply charge a larger premium to offset the extra risk. Secondly, it is difficult to collect data on groups the company does not insure. If a company never writes policies for unmarried couples, it will not be able to compare them to married losses. The litigation story is the one told to the outside world; the one told to the underwriters is quite different, focusing once again on character. The idea that divorced or separated individuals are uninsurable lies in the idea that they are unstable, or worse, immoral: "Infidelity and other departures from the code of sex behavior are seriously regarded, none the less so because they are frequently found in combination with other types of unsocial behavior, such as overindulgence in alcoholic beverages, gambling, and the use of drugs.... The discarded mistress constitutes a menace, as do jealous competitors and gambling associates" (McGill 1967: 393).

New underwriters are presented with a depiction of the gay divorcee, overindulging in drink and "other types of unsocial behavior" that mature, stable, responsible individuals avoid. The lesson is that the habits that led to the divorce may be the same that lead to insurance losses. Since the underwriter can rarely get into the insured's mind, he or she has to rely on the applicant's habits. "A person's habits are often constructed as a guide line to his character, and loose or intemporate habits are often taken as signs of a questionable risk" (Denenberg et al. 1964: 452-53).

Gay and lesbian households fair quite poorly under these guidelines. The fact that a couple may be committed to each other for life makes little difference under hard and fast rules that declare "Applicants must be married." Homosexual couples are thus forbidden from telling how they met, how long they have been together, and the strength of their commitments to each other. The heterosexual newlywed couple applying for insurance is benighted as "stable" and granted coverage; the gay couple, regardless of their time together, remains "unstable" and hence "uninsurable."

Character of Applicants

The 1994 OPIC study found that 29% of insurers had guidelines pertaining to the lifestyle and character of the applicant (Powers 1997: 129). Among those guidelines were ones such as these:

"The key for successful underwriting of all personal lines insurance is that the property to be insured is an above-average risk and the moral character of the insured is above reproach."

"[Applicant must be a] stable member of the community."

"Risk may be unacceptable if doubt be cast on the personal integrity, stability, or habits of the owner. Preferred risks: total abstainers."

"[Avoid] persons with known bad morals."

"[Preferred risks include] applicants whose moral character is above reproach."

"[Preferred risks include] an honest, average person, with high moral standards" (Powers 1997: 130).

These guidelines paint a far different picture than the one seen by the applicants sitting in their houses and filling out a standardized form. Not only are they judged by information they themselves never knowingly provided, they will never learn the real reason for their denial should they be rejected for reasons of character. Instead, they will be informed that the company lacks a program to fit their needs, as if the denial was based on nothing more than a failure to fit into a matrix. (See Table 1.)

Is Exclusion the Only Alternative?

One insurance textbook from the 1960s boasted, "Underwriters often possess an added sense: an uncanny intuition or ability to `read between the lines' of information gathered about an insured and the risk he desires to transfer to the insuring organization" (Denenberg et al. 1964: 449). What underwriters do with this information is clear: they create narratives about different groups that often coincide with contemporary popular opinions concerning which groups are accepted members of society and which groups are excluded. On the basis of an applicant's socioeconomic status and lifestyle, the underwriter then determines if the applicant constitutes a moral hazard, and declines coverage for those who do. Unlike other applicants that pose a greaterthan-average risk of loss, the moral hazard is always uninsurable. As Magee explains: "If, as a matter of fact, the insurance rate could be made to reflect all factors of a risk, the problem of underwriting could be reduced to the happy situation of compiling statistics. Moral hazard is the incommensurable factor that cannot be incorporated in the rate; hence, successful underwriting is largely contingent upon the ability of underwriters to recognize its presence" (1951: 73).

While the middle-class family with the boat in the driveway and the swimming pool in the backyard may present increased hazards, especially to young children, they are still granted coverage, albeit perhaps with an added premium for the pool. Meanwhile, the household with the inoperable car in the driveway and the pit bull in the backyard is denied coverage altogether and is labeled uninsurable. What is it about the latter group than makes them uninsurable: is it the greater risks they pose, or is it something about their character? Depending on the way in which the tale is told, these individuals could be interpreted either as insurable (although perhaps at a higher premium) or uninsurable (because of their character).

In a survey of life insurance guidelines used at the turn of the 19th century, the actuary William T. Standen noted that applicants in certain professions were frequently banned from purchasing insurance. Barge drivers, coal miners, many types of metalsmiths and machinery operators, saloon keepers, and those who handled horses often had difficulty sharing their risks with the rest of the community. The actuary had this to say about what types of people could be insured and which could not:

I am of the opinion that if we exclude those engaged in the Liquor Business, and also such exceptionally hazardous risks as Aeronauts, Divers, Manufacturers of Explosives and a very few others, nearly all the other occupations... can be safely written upon at ordinary rates, and without doing violence to the interest of other policy-holders, provided the physical risk be perfect and the applicant be a man of good intellectual power and moral force (Standen 1893: 116, emphasis in original).

The academic writing on moral hazard has evolved over time from one in which the individual's personal qualities were the primary determination of his or her risk into one in which the rational actor follows the incentive structure created by the insurance contract (see Baker 1996). But the practitioner's perspective on moral hazard up to the current day continues to conflate the two, despite their occasional awareness that even those in dangerous or nonstandard occupations could still be individuals of intellectual power and moral force, in other words, that they could be insurable. A contemporary of Standen noted that even though it was the opinion of his company's chief medical officer that many of these individuals could be profitably written, the actuaries refused to create risk categories for them (Dawson 1906: 250). Even the motivation for profit will not induce companies to create new rating categories if they operate under the belief that the current ones are designed to restrict moral hazards. If actuaries and underwriters continue to see the difference between a boat in the driveway and a junk car as being about issues of character-rather than one of class-then the argument that both types of risk could be insured will fall on deaf ears. Unless those who have been denied coverage because of issues of character challenge the narratives being told about them and demand that insurers provide empirical evidence to support their moral hazard guidelines, there is no reason to believe that current underwriting practices will change. The difficulty for these individuals is, first, to become aware that they are being depicted as moral hazards, and, second, to believe that they can do something about it.

IV. The Myth of the Actuary

Deborah Stone has expressed the opinion that "[t]he numerical rating system, and the underwriting guidelines and rating manuals it spawned, have all the trappings of scientific objectivity-medical terminology, elaborate matrices of diseases and point values, and numbers-but they often seem to be based as much on social prejudices and stereotypes as on empirical knowledge" (1993: 296, emphasis added).

The issue for those concerned with equity in public policy is that, as far as the mass public of insurance consumers is concerned, underwriting guidelines are rarely depicted as prejudiced or based on stereotypes. The power of the current underwriting system is its ability to mask the subjective process of underwriting as being objective and scientifically derived. What on its face appears to be a process of matching the characteristics of the applicant with those in a matrix turn out to be far more complex and subjective. Before one is ever fit into the box of an insurance class, an underwriter must first tell a story about the applicant, inquiring into the individual's morality, character, and place in society. Those who align with the belief system of the underwriter will find themselves able to share their risk with the other members of the community. Those who are on the outside will often end up being branded as moral hazards without ever themselves being allowed to challenge the stories being told about them. Thus, the system of exclusion is propagated by the fact that the discourse is one-sided (see Simon 1988: 776). Even though the insurance industry is regulated, an issue of great importance for the entire polity is hidden from scrutiny through a process I call "the myth of the actuary."

The myth of the actuary is a rhetorical process that conveys the impression that whether or not an individual is deemed insurable depends on nothing more than a fit between the objective characteristics of an applicant (such as age, weight, number of speeding tickets, or whether they smoke) and the rating classes of the insurance company. Applicants are left unaware of the subjective side of insurance underwriting because of the secrecy that surrounds it. It is unlikely that they will ever find out what actually goes on, since only nine states require insurance companies to file underwriting guidelines with their Insurance Commissioner, and only seven states allow consumers access to them (Powers 1997: 121). The average consumers of insurance are most likely entirely ignorant of the manner in which their policies are underwritten, and as Murray Edelman (1972) noted, when individuals lack a substantive knowledge about a subject, they are highly amenable to symbolic language.

The myth of the actuary is bolstered by the sales stories that companies and their agents tell consumers. The sales story is that the company wants the consumers' business and will be there for them in times of need (Baker 1994). The symbolism of sales stories leads to the obvious conclusion that a company would not turn down an applicant if it could make money off that policy. If this were true, the issue of exclusion should never arise. Thus the idea that the very categories themselves may be used to exclude certain groups from coverage is hidden by a myth. Americans are so used to being sorted and categorized on a daily basis that the methods of classification often go unnoticed. Years of education, voting districts, pay grades, and tax rates constantly divide the polity, using methods so common that they end up being taken for granted. Having been categorized so many times in so many different ways, one is not led to question how insurance programs work. Given the mystique that surrounds the idea of "actuarial soundness" (whatever that may be), the consumers' ignorance does not qualify them to question how they are being sorted.

Insurance companies sort people in ways those same individuals might never do themselves. For example, few policyholders probably believe that their marital status indicates they have characteristics that influence their chance of getting into an auto accident, although insurers do. In their textbook on insurance, Glendenning and Holtrom (1977: 68) explain that "[m]arried male drivers under thirty usually pay less than their single counterparts. This recognizes the fact that, generally, married persons at these ages tend to be more mature and responsible than single persons of the same age." If consumers would be surprised at how insurance companies categorize them according to objective characteristics, such as their credit rating for homeowners insurance or marital status for automobile insurance, they would probably be outraged at how insurers rate them on character and morality. Yet the myth of the actuary renders such discoveries not only unlikely but also unbelievable. Without reading the guidelines and the textbooks that few outside the field ever read, who would believe than insurers would decline an individual because he or she lacks "moral character above reproach?"

V. Conclusion: The Narrative Approach to Regulation

I have shown that an applicant for insurance can be denied coverage because underwriters construct stories that depict the applicant as uninsurable. These stories frequently appear to be based on stereotypes and prejudice, and often coincide with contemporary popular views about which groups are part of mainstream society and which are not, especially in relation to class. Rather than being able to shape the narratives being told about them, the excluded remain the subjects of the stories, but never the authors. The result is a one-sided form of discourse that virtually guarantees that the exclusionary effects will continue into the future.

If it is true that subjective underwriting is actually a powerful source of discrimination against certain groups, then the polity should be made aware of this process and participate in the regulation of how these narratives are constructed and used. The polity, and not the insurance companies, should be the ultimate authority over which groups are insurable and which are not, especially if the decisions are being based on subjective criteria. Regulation will fail to prevent discrimination if it does not engage companies at the cause. I have suggested that the myth of the actuary diverts attention from the subjective nature of underwriting, giving the process of insurance underwriting an aura of scientific objectivity. The result is a focus on rating charts, insurance factors, and premium matrices, when the source of exclusion lies elsewhere.

Theodore Porter (1995: 11) has noted that the credibility of numbers or any set of criteria is both a social and a moral problem, and therefore, they are valid only to the extent that they are accepted by the larger community. When we apply this view to the case of insurance, the idea that character underwriting is subjective does not mean therefore that it needs to be secret, in fact I believe just the opposite: its standard should be made known to the polity. Many other forms of subjective judging benefit from an open set of guidelines that all those involved can participate in creating, and so it should be in underwriting as well. If we allow the polity to shape the definition of moral hazard, insurers would be more likely to create a set of guidelines that are more fair and productive: fair in the sense that the entire community has had the opportunity to participate and comment on its creation; productive in the sense that the process creates a public dialogue over the nature of risk and its distribution in the community. Given that guidelines created by the entire polity are likely to be far more intolerant of stereotyping than the current ones in use, we would expect more individuals currently branded as uninsurable to be able to find coverage under the new system.

The idea that moral hazard standards should be made known in advance is not new. Writing in 1951, John Magee argued:

An insurance company canceling insurance on a risk when it is believed that moral hazard is a factor would scarcely be expected to give the reason. Yet it can readily be seen that difficulty in securing a proper insurance coverage might jeopardize credit and even precipitate a financial collapse. For this reason it is essential that every insured have a comprehensive knowledge of the elements that contribute to moral hazard. He may recognize them thereby when he himself is exposed to a loss from neighboring premises where the element of moral hazard is present, and he will also be quick to recognize the reason for cancellations that may seem to be unjustly made on his own property (p. 72).

Magee went on to warn underwriters of the dangers of resorting to stereotypes, exhorting them to free themselves of such sentiments and "appraise each case on its individual merits" (p. 78). The fact that his admonitions have been largely ignored for over four decades should not deter the current generation from adopting them as a practice.

The failure of regulation to eradicate the present methods of illegal or socially unacceptable forms of discrimination calls for a new approach to regulation, which I call "the narrative approach." The narrative approach to regulation would focus regulators' attention on the subjective guidelines companies use to exclude applicants, just as much as current regulation focuses on numbers. Under this approach, underwriting guidelines would have to be explicit about what characteristics of an applicant or his or her situation render the individual uninsurable, and would have to justify them statistically, just as they currently do for rating classes. Underwriting decisions based on stereotypes or prejudice would eventually be challenged by the lack of data to support them, just as they were a century ago under the previous underwriting era, when biases and stereotypes were openly stated. Regulators would more actively monitor and alter guidelines in order to prevent companies from creating narratives that are offensive to the polity. Finally, all underwriting guidelines would have to be filed in advance with the state Insurance Commissioner, and these guidelines would be made available for public scrutiny.

To see how this might work in practice we can take the case of a long-term homosexual couple who are applying for homeowners insurance. According to the criteria for marital status previously listed in Section III, they would typically be denied coverage. Under the narrative approach to regulation, the first question an insurer would have to answer is why it makes a distinction between married and unwed couples. Referring to Table 1, some type of litigation story will probably be the reply, and thus the second question will be to ask for data comparing married couples to unmarried ones. Since unmarried couples are often denied coverage, the insurer might be at a loss to provide the necessary data to justify the distinction. This situation might not only call for a study of the loss histories of gay households, for example, but also, in a best-case scenario, lead to a public discussion as to whether the polity finds it acceptable for insurance companies to discriminate on the basis of marital status at all.

Regina Austin (1983) has criticized the insurance rating procedure as reinforcing the existing socioeconomic hierarchy. Not only do rating systems judge applicants on their position in society, but they also serve to divide one's consciousness of being a part of a group of people who share that position (Simon 1988). Since applicants are not invited to help construct their own narratives, they are often left unaware that they even have a story to tell, and hence will not take action in the political arena to define which types of stories are acceptable to the polity and which are not. The narrative approach to regulation addresses these concerns, allowing individuals to develop an awareness of their relationships to others in the community. At its best, the narrative approach may help the polity to engage themselves in the issue of what they owe each other as members of a shared community of risk. Unable to support some of the social constructions they use, members of the insurance industry might finally realize that many of the risks they have been excluding might be able to be written profitably. At the very least, the narrative approach should begin to eradicate discrimination based on stylized and prejudiced notions of character and morality.

There are clearly situations in which the threat of economic moral hazard is a real and justifiable concern, but they can be left to the insurers to handle through contract design and are already covered by current methods of regulation. The ability of a financial industry to discriminate on the basis of an applicant's character or morality is a different matter altogether. This must be left for the entire polity to decide upon, and as such should be open to public debate and scrutiny. If stories must be told, then their subjects should be able to participate in the telling.