Australian nursing home inspection teams are partitioned into those with an enforcement ideology supportive of reintegrative shaming and those with a more stigmatizing ideology toward noncompliers. A specific test of the shaming theory shows that when interdependency exists between the nursing home and the inspection team, reintegrative shaming has a stronger positive effect on improving compliance.

Australian nursing home inspection teams are partitioned into those with an enforcement ideology supportive of reintegrative shaming, those who believe in being tolerant and understanding when confronted with noncompliance with the law, and those with a more stigmatizing ideology toward noncompliers. Nursing homes visited by teams with a reintegrative shaming ideology display significantly improved compliance in the period following the inspection. Nursing homes visited by inspection teams with a more stigmatizing attitude toward noncompliance display an approximately equal drop in compliance. The performance of homes visited by tolerant and understanding inspection teams falls between these two extremes. A more specific test of the theory shows that when interdependency exists between the nursing home and the inspection team, reintegrative shaming has a stronger positive effect on improving compliance. In nursing homes with no link between the home and the inspection team, reintegrative shaming has no effect on compliance levels. These effects demonstrate both the importance and the limitations of reintegrative shaming as a theoretical concept worthy of further empirical investigation.

The theory of reintegrative shaming was developed during the 1980s as a response to what was seen as a theoretical malaise in criminology. Criminologists had given up on theory, or so it seemed, long before the postmodernist critique of grand narratives swept the academy. This description of the field no longer seems true, however, because the theory of reintegrative shaming (Braithwaite, 1989) was only one of a number of criminological explanations of a rather general sort that have emerged since the mid-1980s (e.g. Cohen and Machalek, 1988; Colvin and Pauly, 1983; Gottfredson and Hirschi, 1990; Wilson and Herrnstein, 1985).

The theory of reintegrative shaming was not highly original. Essentially it was a way of attempting to salvage and contextualize the explanatory power of a number of long-standing theoretical traditions, none of which enjoyed consistent empirical confirmation in the literature or widespread support from criminologists, but all of which had some claim to a degree of empirical support. These theoretical traditions were labeling theory (e.g., Becker, 1963), subcultural theory (e.g., Cohen, 1955), opportunity or strain theory (e.g., Cloward and Ohlin, 1960; Merton, 1968); control theory (Hirschi, 1969); differential association (Sutherland and Cressey, 1978), and social learning theory (e.g., Akers et al., 1979).

The key idea of reintegrative shaming is to partition shaming into two types (actually, a continuum with two poles) whereby shaming that is reintegrative forms one end of the continuum and shaming that is stigmatizing forms the other. Thus when shaming is more reintegrative, the explanatory framework of control theory works (reducing crime). When shaming is stigmatizing, however, the explanatory frameworks of labeling and subcultural theory come into play (increasing crime). Frameworks that were viewed previously as fundamentally incompatible (Hirschi, 1979) are synthesized through the simple device of the partitioning of shaming. So what is the difference between reintegrative shaming and stigmatization?

Reintegrative shaming involves the following:

Disapproval while sustaining a relationship of respect;

Ceremonies to certify deviance terminated by ceremonies to decertify deviance;

Disapproval of the evil of the deed without labeling the person as evil; and

Not allowing deviance to become a master status trait.

Stigmatization involves

Disrespectful disapproval, humiliation;

Ceremonies to certify deviance not terminated by ceremonies to decertify deviance;

Labeling the person, not only the deed, as evil; and

Allowing deviance to become a master status trait.

Shaming, according to the theory, is more likely to be reintegrative and more likely to be effective in conditions of high interdependency between the disapprover and the disapproved. A variety of individual and structural characteristics were posited as conducive to interdependency.

Thus constructed, the theory was able to account for many of the things known about the patterning of crime--why women should be a low-crime group, why the young and the unemployed should be high-crime groups, why marriage should reduce crime, why there should be more crime in large cities or in areas with high residential mobility, why a society such as the United States should have a higher crime rate than a society such as Japan, why crime should have decreased in a number of Western societies during the Victorian era but risen again toward pre-Victorian levels in the period since World War II. Although it is nice to have a theory that accounts for much of what we think we know, the most important test is the capacity to predict and account for new findings that emerge after the construction of the theory.

One opportunity to do this is the evaluation of the impact of the juvenile justice reform strategy based on "family group conferences" that has been national policy in New Zealand since 1989, and now is being implemented more widely as community accountability conferences in Australia. This involves replacing court processing of juveniles with conferences attended by the citizens who care most about the young offender (mostly interdependent family members), by the victim, and by supporters who care about the victim. These conferences are aimed at reintegrative shaming, more or less in the way commended in Crime, Shame and Reintegration (Braithwaite, 1989:173-174). The origins of the New Zealand reform are not to be found in Braithwaite's book, however, but in Maori culture, which had thought through many of the principles of reintegrative shaming centuries ago. The interesting feature of the reintegrative shaming experiments in Australia and New Zealand is that the ideas have been adapted to large, ethnically diverse urban sites such as Auckland and Sydney in a way that rejects a monocultural or neighborhood-based conception of how shaming is transacted in the metropolis. Valuable but preliminary evaluations of these programs are available (Maxwell and Morris, 1993; Moore, 1992; Morris and Maxwell, 1992; O'Connell, 1992); these do not permit systematic testing of the prediction that these programs will do less harm or more good (in the terms defined in the theory) than traditional court-based intervention.

The most important application of the theory to new data has been in interpreting the series of six domestic violence experiments funded by the National Institute of Justice. This research was big-budget science involving genuine experimental designs with random assignment of perpetrators of domestic violence to arrest, to no action, to warrant but no arrest, or to counseling by responding police. The large investment was occasioned by the extraordinary impact of the first experiment in Minneapolis (Sherman and Berk, 1984) in influencing domestic violence policies across America and around the world toward more strongly pro-arrest policies. The finding that arrest reduced subsequent domestic violence by those arrested was embraced by the women's movement and was used in campaigns that secured mandatory arrest policies in many U.S. states in the years following publication of the results.

One of the authors of the Minneapolis study, however, recently has released a book in which he concludes that the finding that arrest reduces domestic violence has not been replicated (Sherman, 1992). Indeed, Sherman concludes that if we bear in mind the methodological shortcomings of the original study and the new results, it now seems likely that mandatory arrest will do more to increase than to reduce violence. Theoretically he concludes that a simple specific deterrence model, which so many feminist enthusiasts for the Minneapolis study found so attractive an interpretation, simply does not fit the data. For one group of men, Sherman concludes that arrest reduced subsequent domestic violence; for another group, the effect of arrest was more pronounced, but in increasing domestic violence.

Sherman's interpretation is that the first group found arrest shameful. These were men with high interdependencies--married and employed. The other group, which manifested a counterdeterrent effect, was disproportionately unemployed (in four studies) and disproportionately black (in three). They were people who had lived a great deal of stigma; their reaction to further shame was rage and vindictive escalation of violence rather than remorse. In Sherman's interpretation, "Defiance is a means of avoiding shame in the face of any effort to cut one down to size, including an arrest" (1992:203).

The theory of reintegrative shaming offers a better interpretation of these initially perplexing results than do simple deterrence, rational actor, or control theory models, or any of the other standard theories in the literature that give no account of the tipping points that shift interventions from being productive to being counterproductive. Even so, the domestic violence studies were experiments not designed explicitly to test the proposition that when shaming is reintegrative, it will increase compliance with the law; when shaming is stigmatizing, it will reduce compliance. The purpose of the present study is to provide just such an opportunity to test the strategic hypothesis, albeit in a rather obscure area of law--compliance with quality of care regulations for nursing homes.

DATA AND METHOD

The data used here to test the hypothesis that reintegrative shaming will increase compliance with the law come from a study of compliance with 31 standards that regulate the quality of care provided in Australian nursing homes (Braithwaite et al., 1993). In 1987 the Australian federal government took over the major regulatory role of nursing homes throughout Australia from the individual states. At this time they also commissioned a major evaluation of this new initiative; as part of this evaluation they ensured that teams of inspectors were sent to 242 nursing homes over a 20-month period, from May 1988 to March 1990. These homes had been selected on the basis of a proportionate sample stratified by size, ownership, and residents' disability. Ninety-six percent of the selected nursing homes agreed to participate in the evaluation.(1) The sampling of the homes was restricted to regions around the four major urban areas in Australia--Sydney, Melbourne, Brisbane, and Adelaide--where two-thirds of all nursing homes in the country are located. In addition to the sample homes, all other homes that were visited by an inspection team within these geographical areas were included in the study as a supplementary sample (N = 168).

No systematic difference between the two samples was shown by comparative analyses between the random and the supplementary homes on a variety of geographical and organizational characteristics of the nursing home, on the sociodemographic characteristics and attitudes of the director of nursing, and on the nursing home's compliance ratings. For the purposes of the analyses, we combined the two groups and treated them as a single sample, which resulted in an overall sample of 410 nursing homes. Nevertheless, a control for homes that were part of the random sample, versus the supplementary sample, was included in the analyses presented later in the paper.

A second inspection of the randomly selected 242 nursing homes was undertaken 18 to 24 months after the first inspection. Eleven of these homes had closed by the time their second inspection was due. Fifteen homes were visited outside the 24-month time frame. These homes are still included in the analyses; the number of months between the first and the second inspection is controlled for in the regression models. Of the supplementary 168 homes, 110 also had received a second inspection within the time frame; three had closed. This resulted in 341 homes that had been inspected twice, 14 homes that had closed, and 55 that had not been inspected a second time. In effect an 86% response rate was achieved with the second inspection process. The compliance ratings assigned by the inspection teams at each inspection provide the measures for assessing change in compliance.

Comparisons between homes that had been inspected a second time (N = 341) and those that had not (N = 69) indicated that on seven characteristics of the director of nursing, three characteristics of the nursing home, and three characteristics of the proprietor, only two of the seven characteristics of the director of nursing revealed a significant difference at the .05 level. Because only two out of 13 significance tests showed a statistical difference between the second wave sample at Time 2 and homes not visited for a second time, we believe that the sample at Time 2 is not a biased selection of nursing homes from the original sample at Time 1.

Although this is an obscure domain, previous work on the reliability and validity of compliance ratings with the 31 Australian nursing home standards suggests profound measurement advantages for testing theories of regulatory compliance in comparison with measuring compliance with standard criminal laws, other regulatory laws, and, indeed, U.S. nursing home laws. A separate study that examined compliance ratings given by an independent inspector in 50 nursing homes at the same time as a government inspection produced test-retest reliabilities between .93 and .96, depending on the stage of the regulatory process at which they were calculated (Braithwaite et al., 1992). Quality of care standards for Australian nursing homes have a complex legal status (McDonald and Bates, 1989). They are federal standards; failure to meet them can result in the suspension or termination of federal government benefits to support residents in the nursing home. For most of the nursing homes in the country, however, some or all of the standards have been incorporated into state laws with a criminal status. These standards therefore cannot be described unqualifiedly as the predatory criminal laws, about which there is substantial community consensus, which were the explanatory target of the theory of reintegrative shaming. Empirically, however, these laws enjoy a high degree of consensus support from the industry (Braithwaite et al., 1993). Detailed work has shown that for 27 of the 31 standards, more than 96% of chief executives thought the standards were clear, 100% of chief executives thought that 18 of the 31 standards were desirable, and more than 95% thought the remaining 13 standards were desirable (Braithwaite et al., 1993, Figs. 5.1 and 5.2). Such consensus is also reflected in the extent to which chief executives agreed with inspection teams' assessments of the nursing homes; across the 31 standards, the average level of chief executives' agreement with the ratings given by the inspection team was 92%. This high degree of consensus indicates that these regulatory standards share with laws against predatory crime the theoretically central element of widespread agreement on the undesirability of lawbreaking. A subsidiary agenda of this paper is to expand the explanatory scope of a theory developed for one domain of law to a related, though different, domain.

In addition to the compliance ratings collected from the inspection process, we collected data from individual inspectors. In May 1990 we mailed questionnaires to all persons who had worked as inspectors between 1987 and 1990. Two follow-up letters were sent in July and August; telephone follow-ups were conducted in September. Two hundred and fifty-eight inspectors were contacted initially. Of these, 14 refused to participate, 32 questionnaires were returned to sender, 21 inspectors failed to return a questionnaire, and 191 returned usable schedules. Eighteen of those who returned questionnaires were managers who had no prior experience as inspectors in the monitoring program. Their inspection was a one-time training exercise; they have been excluded from the analyses that follow.

THE MEASURE OF COMPLIANCE AND MEASURING CHANGE

The 31 nursing home standards that inspection teams use to assess compliance with the law cover health care, social independence, freedom of choice, privacy and dignity enjoyed by the residents, and safety (including risks from fire, violence, infection, and the use of restraints) (Braithwaite et al., 1993). On each standard, the inspection team gives the home either "met" (1), "action required" (.5), or "urgent action required" (0). In a separate study, the standards have been shown to be reliable, valid, and comprehensive in their coverage of the medical, personal, and social needs of the nursing home residents (Braithwaite et al., 1992).

We summed these standards to form a total measure of compliance ranging from 0 (no compliance) to 31 (absolute compliance). Thus there are two measures of compliance, one for the first inspection and one for the second. Although formally the model specifies compliance for the second visit as the dependent variable, compliance for the first visit is entered as a control. In practice, then, compliance at Time 1 has been partialed out, leaving only the difference between Time 1 and Time 2 to be explained by the other independent variables in the model.

There is some debate about whether this is the best way to test a model that predicts change in a variable. Some argue that a simple change score (T2-T1) is the most appropriate dependent variable. Cohen and Cohen (1983:413), however, point out that "the study of change is not the simple, straightforward proposition it appears to be. Indeed, this is a methodological area fraught with booby traps, where intuitive 'doing what comes naturally' is almost certain to lead one astray." They argue that a fundamental problem with a simple change score is that this constructed variable is correlated with, or dependent on, the initial Time 1 score. Thus part of its variance is due to T1. Consequently, if reintegrative shaming is correlated with Time 1 compliance, it will have a stronger impact on a change score because Time 1 compliance is part of the variance that it is explaining. (This is shown to be the case in Appendix Table A2.) (Appendix Table A2 omitted) The way to avoid this problem is to use Time 2 compliance as the dependent variable and then partial out the effect of Time 1 compliance, leaving only regressed change to be predicted. This method also ensures that the relationship between the other independent variables and regressed change is net of their relationship to Time 1 compliance (Cohen and Cohen, 1983:413-422).

MEASURING REINTEGRATIVE SHAMING

Ideally, reintegrative shaming in encounters between regulatory inspectors and firms would be measured observationally. Indeed, this has been done qualitatively during observations of 132 nursing home inspections in Australia, the United States, and England. In this work, however, we used no systematic observational protocols that would enable measurement of the extent to which reintegrative shaming actually was used during regulatory encounters, so that its impact on compliance then might be estimated. On a systematic quantitative basis, however, we measured the extent to which the inspectors' attitudes were supportive of reintegrative shaming as a regulatory strategy. It was then possible to assess whether nursing homes visited by inspectors who were supportive of reintegrative shaming improved, while nursing homes visited by more stigmatizing inspectors showed worsening compliance.(2)

Reintegrative shaming has two facets--reintegration and shaming. The questionnaire included two questions that referred to the extent to which inspectors ("standards monitors" in the Australian jargon) believe in openly disapproving of poor performance, versus presenting as tolerant and demurring from overt expression of disapproval. Standards monitors were asked the extent to which they strongly agreed, agreed, neither agreed nor disagreed, disagreed, or strongly disagreed that "Standards monitoring teams should not hide their disapproval of poor practices in nursing homes" and "It is important for standards monitoring teams to appear tolerant at all times, to rate nursing homes without openly expressing disapproval when the ratings are poor." The correlation between these two items is .30. We summed individual responses to form a measure of disapproval. Because the scale had no natural metric, we rescored it so that it ran from 0 to 10.(3) The final measure had a mean of 4.48 and a standard deviation of 2.20.

Never before has anyone attempted to measure the concept of reintegration. This is no small challenge because it is a rather nuanced concept. Consequently the first attempt at operationalization is a somewhat limited measure--six items with a Cronbach's alpha of .56 (see Table 1). (Table 1 omitted) In the field of nursing home regulation in Australia, there was no point in seeking to operationalize that facet of the stigmatization-reintegration continuum which involves allowing deviance to become a master status trait. The fieldwork had shown that it is extremely rare for Australian nursing home inspectors to view nursing home managers as criminals, even when they are responsible for serious breaches of criminal laws. Indeed, the master status "director of nursing" always dominates the status "criminal" or "crook" or "shonk," even on the rare occasions when the latter labels are used. Concomitantly the facet of focusing on the evil of the person, rather than merely on the evil of the deed, is not operationalized here. This facet also is rendered problematic because in this regulatory domain, the offender in law is always found to be an organization rather than an individual.

The measure captures the notion of humiliation (disrespectful disapproval) versus avoidance of humiliation (respectful disapproval) in Item 4. Other items of this type, however, did not scale well with the six items listed. The scale is dominated by the "burying the hatchet" facet of the reintegration concept: the idea of terminating disapproval by forgiveness, of ending a disapproving encounter (or sequence of encounters) by pointing to something positive, or by offering praise for moving to fix that which was disapproved.(4) We summed the items in the same manner as for the disapproval measure and rescored them from 0 (low reintegration) to 10 (high reintegration). The overall mean (6.39) and standard deviation (1.48) indicate that individual team members are slightly more likely, on average, to agree with reintegration.

These scales are based on the responses of individual team members. The inspection process requires that a team be composed of no fewer than two members; this is the modal size of inspection teams in practice. Three-person teams are common, however; teams of four or more are rare. To determine whether individual team members' views affect change in compliance at the nursing home level, the analysis required that individual team members' responses be matched to the nursing homes they inspected. To do this, we averaged disapproval and reintegration scores across the multiperson teams and then matched them to the nursing home that the team inspected. Although the response rate for individual team members was good, missing data causes special problems with team inspections.

For 394 homes or inspections, we had information for at least one member of the team. These homes in turn can be broken down into those for which we had individual questionnaires for all team members (N = 187), those for which questionnaires were returned on 50% or more of the team members (N = 169), and those for which questionnaires were returned but by fewer than 50 percent of the team members (N = 38). Missing data obviously can become a serious source of error when data are available for only one member of a two-person team. Accordingly we undertook

series of analyses comparing those homes where complete information for the team was available with those homes where information for only some of the team members was available (see Braithwaite et al., 1993). These analyses showed that homes where complete data for the team were available were more likely to be located in Melbourne and less likely to be found in the Adelaide sample. Restricting the analysis to homes with complete information on all team members would effectively bias the sample so that it reflected the Melbourne region. Thus we decided to use all available information and to control for whether full data were available for the team. Table 2 shows that when a control is entered in the relevant regressions for having data for the full team (no missing data), it has no effect in the model. (Table 2 omitted)

CONTROLS

Theoretically, in this particular analysis we are interested not so much in predicting compliance as in predicting change in compliance after an inspection by teams with higher and lower levels of commitment to reintegrative shaming as an enforcement style. Do teams with a reintegrative shaming philosophy improve the compliance of the homes they visit, while teams with a more stigmatizing philosophy actually worsen compliance? To ensure that the model is properly specified so that we can examine this issue, a number of important controls are required.

We selected control variables on two bases: variables that had been shown in previous research to be significant factors for predicting Time 1 compliance (Braithwaite and Makkai, 1991; Makkai and Braithwaite, 1991) and variables that were thought to be of theoretical or methodological importance in affecting change in compliance. The scoring, means, and standard deviations for these variables are shown in Appendix Table A1. (Appendix Table A1 omitted) The variables are grouped into four categories: controls for the geographical location of the nursing home, controls for the various social and structural characteristics of the nursing home, controls for the inspection team and the inspection process, and two methodological controls for the nursing home sample and the inspectors' sample.

Many of the control variables in Table 2 are shown to be of little importance in explaining change in compliance. Therefore, when the model is reestimated for two subsamples within the data set (see Table 3), we exclude these variables from the estimations. (Table 3 omitted) Preliminary runs showed that their inclusion or exclusion did not affect the relationship of substantive interest, namely reintegrative shaming and change in compliance. Thus, because the sample sizes for the partitioned data sets are relatively small, only the results for the reduced models are presented in Table 3. So many of these controls fail to make any difference in the change of compliance model because their effects on compliance are captured in the control for Time 1 compliance.

RESULTS

The theory of reintegrative shaming does not predict a reintegrative main effect or a disapproval main effect on compliance. This is because it predicts negative effects of disapproval on compliance when the disapproval is stigmatizing, and positive effects of disapproval when it is reintegrative. the first column of Table 2 shows that neither of these main effects is significant. The right-hand side of Table 2 tests the theoretical prediction: that there is a significant effect On compliance of the interaction between reintegration and disapproval.(5)

The significant interaction effect in Table 2 is exactly of the type predicted by the theory of reintegrative shaming.(6) The greatest improvement in compliance scores between Time 1 and Time 2 occurs in homes that are visited by inspection teams with high average disapproval scores and high average reintegration scores (teams with a reintegrative shaming philosophy). This point can be seen most easily by graphing the effect of reintegration on compliance for different levels of disapproval. In Figure 1, teams that are average or low on expressing disapproval have intermediate effects on changes in compliance. Among these intermediate groups, the teams that perform worst are those which are tolerant and understanding (low on disapproval, high on reintegration) and those which tend toward stigmatization (average on disapproval, low on reintegration). The teams with the worst performance on changes in compliance scores are the stigmatizers who are high on disapproval but low on reintegration.(7)

Figure 2 shows that homes visited by these teams with the most stigmatizing attitudes actually display a reduction in compliance with the standards in the period following the inspection. (Figure 2 omitted) In fact, the reduction in compliance associated with an inspection by the most stigmatizing teams is as great as the improvement in compliance among homes visited by teams with the strongest belief in reintegrative disapproval. Figure 2 depicts the cases that are jointly in the top 33% of cases on disapproval and the top 33% on reintegration (reintegrative shaming), compared with those which are jointly in the top 33% on disapproval and the bottom 33% on reintegration (stigmatization) and with those in the bottom 33% on disapproval and the top 33% on reintegration (tolerance and understanding). These groups contain only 19 strong reintegrative shaming cases but 59 strong stigmatization cases. Despite the small number of strong reintegrative shaming cases, the difference is statistically significant. The comparative rarity of having the two joint conditions for reintegrative shaming shows the potential for increasing regulatory effectiveness through shifting the majority to strong reintegrative shaming, if the analysis in this paper is correct.

To put the levels of change in Figure 2 in perspective, the reintegrative shaming cases experience a 39% reduction in the level of noncompliance; the stigmatization cases show a 39% increase in noncompliance.(8) Figure 2 shows that the low disapproval/high reintegration cases ("tolerance and understanding") fall between these two extremes. The teams that are most tolerant and understanding are also associated with declines in compliance at the homes they visit.

A further elaboration of the model involves the key concept of interdependency. Braithwaite (1989) argues that reintegrative shaming will be most effective in situations in which interdependency exists between those who shame and those who are shamed. Individuals are interdependent when they participate in networks wherein they depend on each other to achieve valued ends. To some extent, interdependency characterizes all regulatory encounters--regulator and regulatee depend on each other to get their jobs done. Obviously, however, interdependency is greater when managers interact with the same inspector in repeated encounters than it is in one time encounters. Moreover, interdependency is greater when managers have dealings with inspectors in some other aspect of their lives outside the regulatory encounter. Therefore a simple measure of interdependency in this context is the reports by directors of nursing that they knew some of the inspectors before their regulatory encounter. Fifty-nine percent of directors of nursing reported that they had previously met some or all of the inspectors; 41% reported never having previously met any of the inspectors.

We divided the sample into two groups on this basis, and estimated a model that included the controls plus the reintegrative shaming interaction. The results are presented in Table 3. The effect for reintegrative shaming on regressed change in compliance is strengthened greatly in homes where the director of nursing knew one or more members of the inspection team before the inspection visit. Where the director of nursing knew none of the inspectors (low interdependency), we found no reintegrative shaming effect. This is both good and bad news for the theory. The good news is that the theoretical prediction is supported; the bad news is that the theory does not work in one-time interactions with anonymous agents of the state.

Figure 3 aptly demonstrates the effect of reintegrative shaming on compliance levels for the two groups of nursing homes. (Figure 3 omitted) Reintegrative shaming works only in situations of interdependence, as indicated by the almost parallel lines for reintegration on compliance for different levels of disapproval in homes that had no interdependency with the inspection team. Significance tests for the three simple slopes indicated that none differed significantly from 0. In strong contrast, where interdependency is greater, highly disapproving teams with high reintegration had higher predicted compliance at Time 2, controlling for performance at Time 1. As the theory predicts, the significance test for the simple slope of compliance on reintegration at a high level of disapproval (b = .96, se = .43, t-value = 2.25) differs significantly from 0. Although the slope of the line for low levels of disapproval (about which the theory makes no prediction) appears to be relatively steep, the statistical test showed that it did not differ significantly from 0 on this reduced sample (b = 1.00, se = .54, t-value = -1.85).

DISCUSSION

Qualitative fieldwork observing the training of Australian nursing home inspectors, and indeed participation in being trained as inspectors ourselves, leave no doubt that these inspectors are trained to be reintegrative rather than stigmatizing. Fieldwork observing 58 Australian nursing home inspections suggests that inspectors generally appear to be, and certainly try to be, reintegrative. This fieldwork, however, also suggests that managers are exquisitely sensitive to criticism and are capable of imputing stigmatizing intent to feedback from inspectors which is intended to be "factual" rather than to "put them down." There were inspections which both we as observers and the team as actors would have described as rather neutral, nonauthoritarian feedback on poor nursing home performance, but which moved directors of nursing to respond that they had experienced "a Spanish Inquisition" or had been confronted with inspectors who were "little Hitlers" or "a Gestapo."

The qualitative lesson may be that being "neutral" in providing negative feedback is never really possible because of human beings' propensity to read stigma into strong criticism. In this regard, being a law enforcer may be not unlike being a marker of students' essays; unless one performs active repair work to communicate respect for the evaluated actor, fair and objective feedback that is strongly negative is likely to be interpreted as humiliation, as degrading the poor performer's total identity.

Experiences of being treated as "a Gestapo" for clear communication of poor performance causes many teams to go soft on disapproval instead of going strong on reintegration. Inspectors do not have to be bribed to suffer this psychological capture: we all love to be loved; we all crave a daily working life in which we are treated as something other than pariahs. As a result, more of these Australian inspectors opt for being tolerant and understanding than for being reintegrative shamers.

Thus we find here that even in an inspectorate with a generally reintegrative ideology, even in a domain where master status traits such as "criminal" or "junkie" are rarely invoked, the most stigmatizing inspectors manage to cause significant reductions in the compliance of the nursing homes they inspect. At the same time, tolerant and understanding inspectors who believe in saying nice things at all times are almost equally ineffective because these tolerant, understanding inspectors fail to express disapproval when the standards set down in the law are not met. The effective inspectors are those who believe in strong expressions of disapproval combined with strong commitments to burying the hatchet once such robust encounters are over, to terminating disapproval with approval once things are fixed, to tempering disapproval for poor performance on one standard with approval for good performance on other standards, to avoiding humiliation by communicating disapproval of poor performance within a framework of respect for the performer. In short, effective law enforcers in this domain have high interdependency and have mastered the subtle art of reintegrative shaming.

As the theory predicts, the effectiveness of reintegrative shaming is observed most clearly in situations of interdependency. Teams espousing a reintegrative shaming philosophy are much more likely to engender a significant improvement in compliance when managers report some familiarity with the inspection teams before the inspection process. In a regulatory situation in which inspectors become known to chief executive officers over the course of time, respectful disapproval of noncompliance will be more successful in improving compliance. It follows that the common regulatory policy of rotating inspectors to reduce capture and corruption may cause reduction in compliance.

The finding that reintegrative shaming does not work with inspectors who meet managers for the first time supports Findlay's (1993) critique of the New South Wales implementation of the theory through community accountability conferences coordinated by the police. Findlay (1993:39) points out that when there is "the absence in relationships between police and the community of features essential for the success of shaming" particularly connection and respect, it is implausible that reintegrative shaming by the police could work. Yet Findlay (1993:39) goes on to discover the remedy to his own critique when he suggests that "police might more safely adopt the role as a catalyst for community organization so that dialogue can occur within a relevant community about a particular crime and its offender." This indeed is the idea of New South Wales community accountability conferences: it is not the police convening the conferences who are relied on to do the reintegrative shaming; it is the family members, friends, and football coaches, selected for attendance precisely because of their special bonds of care for the offender.

Similarly, with the nursing home regulatory process, we can ponder a parallel answer to the reviewer of this paper who wondered whether inspectors "can really have such an impact in one inspection." Indeed, this seems so implausible that it might be read as strikingly strong support for the theory that just one reintegrative shaming encounter can be associated with a notable percentage reduction in noncompliance two years later. On the basis of our fieldwork, however, we conclude that effective inspectors with a philosophy of reintegrative shaming do not rely heavily on themselves to do that shaming. Compliance discussions are important when the inspectors catalyze dialogue among highly interdependent people--proprietors and managers, managers and staff, staff and residents' committee representatives. In addition, one often hears inspectors respond to a problem by suggesting that a staff meeting be held to make sure everyone understands why what was done was wrong and to brainstorm for alternative ways of dealing with it. We hypothesize that focusing on the problem rather than on the wrongdoer, in the context of a community of care and understanding for the wrongdoer, creates the structural conditions conducive to reintegrative shaming.

Caution is warranted on two fronts, however. First, Australian nursing home regulation may be an unusual domain of law enforcement. The history of criminological theory can be read as one of scholars developing general theories that have explanatory purchase in the areas they know, but limited explanatory power beyond those areas. Thus students of gangs in the slums of cities such as Chicago developed subcultural theories that were useful there, but had little explanatory power when researchers used self-reported delinquency measures in middle-class suburbia. Conversely, scholars such as Travis Hirschi (1969), who worked with the predominantly trivial episodic delinquency revealed by self-report data, developed theories that looked useful only when viewed through the limiting prism of such data. Psychologists developed theories of psychopathology and crime geared to the problems that presented in psychologists' consulting rooms. These theories, however, did not seem very plausible to Sutherland when he assessed his white-collar criminals against them. Braithwaite (1989) developed a theory of reintegrative shaming that he concedes was shaped by what he knows best--business regulatory enforcement--even though he claims that it fits the facts known about other types of crime better than do competing theories. Skepticism is needed until supportive evidence accumulates from other domains. The experiments with reintegrative shaming of juvenile offenders in New Zealand and Australia provide outstanding opportunities for testing whether more stigmatizing community accountability conferences are counterproductive and whether more reintegrative conferences (and follow-throughs from those conferences) are productive.

Second, although this study is conducted in a domain with the virtue of an exceptionally high reliability measurement of lawbreaking, the reintegrative shaming scales have acceptable but not strong measurement properties. A great deal of work is needed to sample the facets of the reintegrative shaming domain more systematically with attitude items, to explore its dimensionality factor analytically, and to investigate its reliability. Again, this task would be simpler in the realms of juvenile delinquency and the reintegrative shaming of adult drunk drivers or wife beaters than it is through quantitative research on organizational compliance with regulatory laws.

REINTEGRATIVE SHAMING AND COMPLIANCE WITH REGULATORY STANDARDS*

* This project has enjoyed the funding support of the Australian Department of Health, Housing and Community Services, The Australian Research Council, the American Bar Foundation, and the Australian National University. The authors are indebted to the support of their colleagues on the Nursing Home Regulation in Action Project, Valerie Braithwaite, David Ermann, Diane Gibson, and Anne Jenkins. We would also like to thank Bruce Biddle for his comments on an earlier draft.

1. This response rate was achieved because of the legitimacy bestowed by government and industry associations' support for the project.

2. It follows that the limitation of this approach is that we cannot be sure whether a nursing home visited by a team with a stigmatizing ideology actually has a stigmatizing encounter with the team.

3. Before making the composite measure of disapproval, we rescored the second item. Then we divided each item by its standard deviation. This effectively ensured that each item had a variance of 1 while still retaining its mean.

4. We do not view this as a final, satisfactory measure of reintegration. A worthy research enterprise would be to factor-analytically explore a larger pool of items designed to tap the four facets of the definition of reintegration in a domain that is more contextually hospitable to all four facets. It may be that the domination of the measure by the "forgiveness ceremony" facet of the domain reflects simply the special context of Australian nursing home regulation, where criminal master statuses and evil individuals do not make much sense to inspectors. Again, it may be that reintegration is a multidimensional domain. The face validity of the items, the fact that Cronbach's alpha is a conservative measure of reliability (Carmines and Zeller, 1979:45), and that this is a first attempt at trying to develop such a measure point to the scale being acceptable.

5. The problem of multicollinearity between the components and the interaction term has led some scholars to argue that interactions cannot be adequately tested for in multiple regression (Jaccard et al., 1990). This form of multicollinearity is due to the scaling of the variables; one suggested solution to this problem is to center the components before creating the interaction term (see Aiken and West, 1991; Jaccard et al., 1990). Failure to center the data may result in increased sample variances and thus may ultimately affect the tests of significance. To center the data, we transformed the scales by subtracting the mean from each score. The correlation between disapproval and the interaction term was reduced from -.90 to -.17, and the correlation between reintegration and the interaction term was reduced from -.02 to -.00.

6. Because of a lack of invariance in the data, the traditional standardized centered coefficients from the regression model cannot be used when interactions are incorporated into the model (see Aiken and West, 1991:40-48, Jaccard et al., 1990:33-34). The correct method is to standardize all the variables before estimating the model and to enter the product of the two standardized main effects to capture the interaction term. The unstandardized coefficients from this model are the appropriate standardized coefficients. Because the cross-product term does not have a mean of 0, "the regression intercept will typically be non-zero, though in traditional standardized regression analyses this coefficient is always zero" (Aiken and West, 1991:44).

7. The figure depicts three simple regression equations for the slope of compliance on reintegration for high, average, and low values of disapproval. We adopted Cohen and Cohen's (1983) guidelines (also see Aiken and West, 1991; Jaccard et al., 1990), with low estimated at one standard deviation below the mean, average at the mean, and high at one standard deviation above the mean. Significance tests (see Aiken and West, 1991:14-22) for the simple slopes indicated that the slope at high reintegration (b = .71, se = .34, t-value = 2.12) is significantly different from 0, but the slopes for average reintegration (b = .23, se = .26, t-value = .86) and low reintegration (b = -.34, se = .98, t-value = -.90) are not.

8. These percentages are much more modest, however, if we calculate them with compliance as the base rather than noncompliance, because most nursing homes comply with most standards. Hence the percentage increase in compliance for the reintegrative shaming cases is only 6.7%. We report it as we do in the text because criminologists normally state their policy objectives in terms of percentage reductions in crime rather than percentage increases in law observance.

REFERENCES

Aiken, Leona and Stephen West 1991 Multiple Regression: Testing and Interpreting Interactions. London: Sage.

Akers, Ronald L., Marvin D. Krohn, Lonn Lanza-Kaduce, and Marcia Radosevich 1979 Social learning and deviant behavior: A specific test of a general theory. American Sociological Review 44:636-655.

Becker, Howard S. 1963 Outsiders: Studies in the Sociology of Deviance. New York: Free Press.

Braithwaite, John 1989 Crime, Shame and Reintegration. Cambridge: Cambridge University Press.

Braithwaite, John and Toni Makkai 1991 Testing an expected utility model of corporate deterrence. Law and Society Review 25:7-40.

Braithwaite, John and Toni Makkai 1993 Can resident-centred inspection of nursing homes work with very sick residents? Health Policy 24:19-33.

Braithwaite, John, Toni Makkai, Valerie Braithwaite, and Diane Gibson 1993 Raising the Standard: Resident Centred Nursing Home Regulation in Australia. Canberra: Australian Government Publishing Service.

Braithwaite, John, Valerie Braithwaite, Diane Gibson, Miriam Landau, and Toni Makkai 1992 The Reliability and Validity of Nursing Home Standards. Canberra: Australian Government Publishing Service.

Carmines, Edward and Richard Zeller 1979 Reliability and Validity Assessment. Beverly Hills: Sage.

Cloward, Richard A. and Lloyd E. Ohlin 1960 Delinquency and Opportunity: A Theory of Delinquent Gangs. Glencoe, Ill.: Free Press.

Cohen, Albert K. 1955 Delinquent Boys: The Culture of the Gang. Glencoe, Ill.: Free Press.

Cohen, Jacob and Patricia Cohen 1983 Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. 2nd ed. Hillsdale, N.J.: Erlbaum.

Cohen, Lawrence E. and Richard Machalek 1988 A general theory of expropriative crime: An evolutionary ecological approach. American Journal of Sociology 94:465-501.

Colvin, Mark and John Pauly 1983 A critique of criminology: Toward an integrated structural-Marxist theory of delinquency production. American Journal of Sociology 89:513-551.

Findlay, Mark 1993 Police authority, respect and shaming. Current Issues in Criminal Justice 5:29-41.

Gottfredson, Michael R. and Travis Hirschi 1990 A General Theory of Crime. Stanford: Stanford University Press.

Hirschi, Travis 1969 Causes of Delinquency. Berkeley: University of California Press. 1979 Separate and unequal is better. Journal of Research in Crime and Delinquency 16:34-38.

Jaccard, James, Robert Turrisi, and Choi K. Wan 1990 Interaction Effects in Multiple Regression. London: Sage.

Makkai, Toni and John Braithwaite 1991 Criminological theories and regulatory compliance. Criminology 29:191-220.

Maxwell, Gabrielle M. and Allison Morris 1993 Family, Victims and Culture: Youth Justice in New Zealand. Wellington: Social Policy Agency.

McDonald, Tracey A. and Philip W. Bates 1989 Commonwealth Nursing Outcome Standards: A Practical Guide to the Duty of Care. Canberra: Department of Community Services and Health.

Merton, Robert K. 1968 Social Theory and Social Structure. New York: Free Press.

Moore, David B. 1992 Facing the Consequences: Conferences and Juvenile Justice. National Conference on Juvenile Justice. Canberra: Australian Institute of Criminology.

Morris, Allison and Gabrielle Maxwell 1992 Juvenile Justice in New Zealand: A new paradigm. Australian and New Zealand Journal of Criminology 26:72-90.

O'Connell, Terry 1992 It May Be the Way to Go. National Conference on Juvenile Justice. Canberra: Australian Institute of Criminology.

Sherman, Lawrence W. 1992 Policing Domestic Violence. New York: Free Press.

Sherman, Lawrence W. and Richard A. Berk 1984 The specific deterrent effects of arrest for domestic assault. American Sociological Review 49:261-272.

Sutherland, Edwin H. and Donald R. Cressey 1978 Criminology. 10th ed. New York: Lippincott.

Tabachnick, Barbara and Linda Fidell 1989 Using Multivariate Statistics. 2nd ed. New York: Harper and Row.

Wilson, James Q. and Richard Herrnstein 1985 Crime and Human Nature. New York: Simon and Schuster.

Toni Makkai is a Research Fellow in the Sociology Program, Research School of Social Sciences at the Australian National University. Her research interests include regulation and compliance, drug use and public opinion, and the sociology of the professions.

John Braithwaite is a Professor in the Law Program, Research School of Social Sciences at the Australian National University. His most recent publication, co-authored with Brent Fisse, is Corporations, Crime and Accountability (Cambridge University Press).