¹ This is, of course, not true of most anthropological studies of law (Nader and Yngvesson, 1973; Koch, 1979; Gulliver, 1969). It is also not true of some who are interested in disputing as a dynamic social process (Felstiner et al., 1981).

² Recent revisionist work on mediation suggests that mediation, at least in the United States, all too often fails to achieve conciliation. As a result, revisionists believe that either it must be reformed or still other techniques tried (see Felstiner and Williams, 1980; Tomasic, 1980; Merry, 1980).

³ Disputes may be bilateral, at least in our view, wholly contained within dyadic relationships. Indeed, most of the dispute processing experiences in any society consist of disputes which have no public aspect to them unless or until some third-party processing is invoked. By accepting that disputes may exist within dyads, we depart from a tradition of anthropological research which defined disputes as having a public aspect. Gulliver (1969: 114), for example, argues that “... no dispute exists unless and until the right-claimant, or someone on his behalf, actively raises the initial disagreement from the level of dyadic argument into the public arena. . . .” In another place Gulliver writes (1979: 75) that disagreements are conflicts concerning relationships or what to do in particular matters of interest which are dealt with:

⁴ To forestall misinterpretation of the data obtained from our Household Screening Survey, it is appropriate to set forth clearly and openly what is not claimed or intended. The survey does not constitute a definitive estimate of households' incidence rates of all grievances, claims, and disputes, for at least the following reasons:

⁵ The use of a fixed list of problems to inquire about disputing (e.g., Curran, 1977) has been subject to some important theoretical and methodological criticism. Raymond Marks (1976: 195) argues, for example, Such a technique has a built-in bias that cannot easily be avoided: a problem is legal because the researcher says it is. The researcher inevitably draws heavily upon his knowledge of or sense about what people have used lawyers for in the past. This also has elements of a legal intelligence test. . . . The approach, in other words, is norm referenced for legal services users. It carries with it the possibility that those who did not take problems to a lawyer will not admit to having had the problems.

Marks' argument suggests, first, that the fixed list approach is inherently conservative in that it concentrates on problems already defined as legal and does not inquire about problems which might someday become legally cognizable. We think that this criticism is important but not dispositive. The household screening survey from which our data is taken was designed to find and identify bilateral civil legal disputes—that is, disputes for which there were available legal remedies, but which were not brought to courts or nonjudicial dispute processing institutions. This is a kind of market research, an inquiry into how much of the universe of currently defined dispute processing business is being handled by existing institutions. We recognize that the parameters of that business are quite volatile, and we do not believe that only currently recognized disputes are legitimate. Our interest, however, remains one of assessing disputing and its precursors within the frame of an interest in legal institutions.

Not only do the problems included in a fixed list represent a potentially biased subgroup of problems people face, but Marks also argues that those who have not used the legal system will under-report the incidence of problems that are on the list. Indeed, there is reason to believe that even some legal services users are reluctant to discuss their problems with a telephone interviewer. Our interest in legal institutions makes the question of bias most salient, however. We tried to avoid the bias Marks describes by careful question sequencing. All probes were reserved until the end of the entire list of problems so that respondents were not cued to our interest in dispute development and dispute processing (and to the consequent lengthening of the interview's duration) until after they had identified their grievance experiences.

Another source of bias is more difficult to address. It is perceptual in nature, arising from the factors which lead some people to label a situation as a problem while others perceive or define the identical situation as something else (Felstiner et al., 1981). We inquired about objective events when possible (“Were you involved in any auto accidents?”), but most problems have to be labeled as such by the respondents, and for some problems, such as discrimination, the labeling process may be quite subjective.

⁶ Households differ in both degree and type of exposure to risks of grievances, depending upon the amount and the kinds of interaction they have with the outside world. People who do not rent, for example, cannot have landlord-tenant problems: they are not in a relationship from which such problems could arise. The more a person drives a car, the higher the risk of an auto accident, all else being equal. We ascertained the following kinds of risks: owning real property, owning a home built within the last five years, holding a mortgage, having recent home repair work, renting a home or apartment, being divorced, and owning property jointly with someone outside the household.

⁷ Many grievances, such as those involving torts or debts, had clear monetary stakes. We asked respondents to estimate the potential or actual value of some other types. For example, those reporting consumer grievances were asked “Would it have taken $1,000 or more to resolve the problem, including other expenses it may have caused?” Appendix I notes which grievances were monetized; grievances involving less than $1,000 were screened out. Some other kinds of grievances could not easily be monetized. We judged that discrimination grievances involving employment—being denied a job or promotion, losing a job, being paid less, and so on—were very likely to entail damages of at least $1,000. Similarly, problems collecting government benefits or obtaining government services and post-divorce problems with child support or alimony were likely to involve the minimum stakes over time. Other types of discrimination (schooling or housing) or post-divorce (child custody, visitation) problems and civil rights violations were judged to be of a serious enough nature to qualify as middle-range. Similarly, property problems such as questions about zoning or boundary lines were included. Landlord grievances were limited to “serious problems.” We reviewed many cases with nonmonetary stakes, and relatively few had values below the minimum, as judged, for example, by their appropriateness for a small claims court.

⁸ This operational definition of a dispute differs from the conceptual definition offered earlier, which added that a compromise agreement indicates a partially rejected claim and thus a dispute, however brief. The result is a conservative estimate of dispute rates.

⁹ Problems involving less than $1,000 were usually screened out early in the interview (see Appendix I and note 7). There were very few which involved more than $10,000. Our probing of problems was limited to those which occurred during the past three years in order to minimize difficulties of recall, which a pretest showed begins at about that point.

¹⁰ The household was the aggrieved party in most cases for several reasons. Fully twenty-two of the thirty-three specific problems for which we probed were household grievances by their nature; eight could involve a grievance both of and against the household; and three involved grievances against the household. This apparent bias largely reflects our focus on disputes arising from members acting in a private non-business capacity. It also reflects our methodological expectation that households would under-report grievances against themselves, an expectation that seems to have been accurate. For example, 2.8 percent of the households reported some property damage or personal injury other than auto accidents 'through the fault of someone else“ which involved over $1,000. In contrast, only 0.5 percent reported that a household member had ”been accused of injuring anyone or of damaging someone else's property, either accidentally or on purpose.“

¹¹ One reason for this relatively high grievance rate may be that the survey was careful to remind respondents both of a number of potentially illegal discriminatory grounds (“ ... race, sex, age, handicaps, union membership”) and of discriminatory actions (“Have you or anyone in your household been denied a job or promotion or lost a job because of discrimination? . . . had any problems with working conditions or harassment, or being paid less? . . . had any other employment problem because of discrimination? . . . any problem with discrimination in schooling or education? ... in buying or renting housing? . . . any other problems of discrimination because of race, sex, age or anything else?”).

¹² Only grievances which were settled or abandoned at the time of the interview were pursued in later sections of the questionnaire because of the sampling goal of locating terminated disputes. Thus the proportions are probably underestimated, since making and processing a claim extends the time a grievance is active. Respondents who asked were told: “By ‘over’ we mean either an agreement has been reached ending the problem, or no one is doing anything more about it.”

¹³ As we suggested earlier, the most appropriate baseline for measuring disputes seems to us to be provided by the existence of claims. But, one might also want to examine disputing as an outcome of the experience of, or the willingness to acknowledge, having a grievance. While most grievants made a claim for redress, and most claims generated disputes, the overall disputes to grievances ratio is slightly less than 1/2 (44.9 percent). When the disputes to grievance ratio is examined for each type of problem, two types are found to have particularly low ratios: torts, 20.1 percent, reflecting the low rate of disputes relative to claims, and discrimination, 21.6 percent, reflecting the low rate of claiming.

¹⁴ We recognize, of course, that lawyers and courts do more than process such disputes; much of their activity is administrative or aimed at dispute prevention. We also recognize that the role of lawyers and courts may be very different in small or large disputes than it is in the area of middle-range disputes. Nevertheless, our data provide a first, albeit tentative and limited, overview of their role in those disputes.

¹⁵ The logit model transforms the dependent variable into the log of the odds ratio. When the dependent variable is dichotomous, a logit model is preferable to an ordinary least squares regression model for three reasons. First, the OLS model is statistically inefficient and yields biased significance tests. Second, the logit model constrains estimated probabilities to the meaningful range between 0 and 1. Finally, the effect of an independent variable varies over the probability range, being greatest over intermediate values and decreasing as the probability of the dependent variable approaches 0 or 1, thus more plausibly modeling the phenomenon of interest. See Hanushek and Jackson (1977).

¹⁶ The maximum likelihood chi-square—G²—is not an adequate overall goodness-of-fit measure in the case of maximum likelihood logit regression. DuMouchel (1976) describes a measure of the reduction in predictive error, which compares the probability of misclassification of the dependent variable without the help of the model to that probability with the help of the model. If y_i is the variable equalling one if the i^th household reports a grievance and zero otherwise, and p = P(y_i=l), then the reduction in predictive error

where

the predictive error under the assumption that all coefficients are zero, and

the predictive error under the estimated model.

¹⁷ The lack of income effects was somewhat surprising to us. It may be that this result is a function of the particular types of grievances we studied. It may be that income effects would be clearer were we to focus on the total number of grievances experienced, or were we to weight each grievance in terms of its potential impact upon a household. This would require determining what was at stake relative to the resources or assets of each household. Were we to do so we would expect that lower income would be associated with more substantial grievance impacts.

¹⁸ The judicial districts, which accounted for some variation in grievance rates, are deleted from this and subsequent models because their conceptual meaning is not clear. They did not add to the explained variation in any subsequent model.

¹⁹ As obvious as this is, or perhaps because it is so obvious, such indicators are rarely employed in studies of claiming, disputing, or litigation (see, for an exception, Rosenthal, 1974). They were omitted from this research quite simply because the question of accounting for success of claims was not, at the time the research was designed, a major concern.

For references cited in this article, see p. 883.