Informal Social Control of Crime-Modification of Labeling(II)

全文

(1)~'. INFORMAL SOCIAL CONTROL OF CRIME MODIFICATION OF LABELING THEORY (II) Mei-chun Liu. .. 1tl. ~. $* ~. * )( l:i:a:~ ~ '&;jiilIH~ IE:rtf±it~ iIilltEm¥ ~iIilJ l:1Wit!ltt • i!:if!. , IJ.(~JliH~iE:rtttit~lIIDlWm¥~iIilJ~fm •. Jjtl.J!I!IitUIW&~JjiH~IE. tI.J!I!IitU£*l:~~fi~~~IW.~ttittl~IT~,it~~~=~. m¥ • *)(~ljtlllfi£Jlit)tl~Wltftttittl.:t!U~~ , lJ.(1JiEtI~J!I!IitU , ~. .:E%tf±it• • ITjIj/F&:::,~~ltmmIlW:lju.1it~ Jf: , m!lH.:a: (1) "'.. 1W*$t~f£:W ; (2)1Jk"'.~1W 1"J:a:£lMJ,C'itifl!i •. J(*4*$t:fIH':frJ}tl;&t1974.&1975~~~lii!lIml1nffl~;f3.riiWm,lU~. Wm,IYf~~ilJIW"j;'~ , Jjfl£1MIYf~IW)!!ifH':EiUj;a.1f ( riiWm,,ff*~. ~~M*~~,WWm,,ff*~=~M*~~)·~»~:frm,~~=,ff. *(Att;J:l;2ooA ) • • ,ttit• •. IWmr=~r,,~*liITPanel analysis· ~H IT~~*e$IE:rt • • ~(~m~· ~f'Ffflit~*eiE:rt~.~ ( ~UW~.&ttI.~) *1W:k 0 ~;&tf±ittl. • ITjIj~~1:.IE lii1it:a lii~t\tl: ' J{IJmtl~IWf£:W -I!;&tm.mib~ J{U~. 1:.iElii.WIW~fmf£:k,£M.~~Q.1:.:aliilW • • •. l:i£~H!li!OI:!:~MT~~IE:rtf±it~iIilJ:Wl~:a:m!!l!li1il\llJl:lWm!lf£ 0. J&!*~.iIilJtrIJ.(IE:rt1!fllt~m>t·IW~~ ( ~Il!ltiWJJ ' WifIi'&$SLt±it. •• ),. f£~.) ,~i!H1Jl:j=lIYl.~jc.,{l~mf1i,\!I,~. 0. l. RESEARCH DESIGN AND DATA. ./. . 1. The Building Block of Current Research: Research by Klein et ai. The current study is based upon previous research by a group headed by Malcolm Klein at the University of Southern California. This was a continuation of their original study. However, this research is different in that a di(ferent paradigm is pursued and a new conceptualization is presented. The data set was collected more than a decade ago. Their ~rch was designed. I.. *"11::!- A *~*.:Lliff ~ FIf.llJ~:!t -287. .'. . ....

(2) The Journal of National Chengchi University, Vol. 69, 1994. .... to clarify the key /JiSsumptions of labeling theory and to correct methodological problems from P~fiOUS research. For instance, previous studies are better test of immediate impact Ithan tests of impact over time. Most labeling theorists would argue that it takes time for the progression from label application to label acceptance and eventually to subsequent behavior. Methodologically, past studies have suffered from problems related to poor sampling, unknown populations, and non-replicability, as indicated by Klein et a1. (1977). Also, they have measured different official actions to the neglect of in9ividuals' subjective perceptions. More importantly, previous research was not e,Kperimental in design; inadequate comparison groups rendered any result inclusivt. The study by Klein et a1. was experimental, with randomization of subjects to several types of official contact representing four different degrees of labeling. The main objective was to determine whether a higher recidivism rate resulted in the group that had a greater exposure to the negative societal labeling. And they did find a higher official recidivism rate in this high exposure group. But self-reports failed to show significant differences between groups. The discrepancy between official records and self reports led to their suspicion that a higher official recidivism rate in one group of juveniles might be a consequence of the fact that "the labelers are somehow responding to their own prior decision." (Klein et aI., 1977:30). The original paradigm began with initial societal reaction (i.e. arrest), followed by three kinds of dispositions: inserting further into the juvenile justice system; referring to the social service system; and releasing without further action. This distinction of various dispositions is of central importance. Ageton and Elliott pointed out that police contact is quite a crude category so that it needs further differentiation in terms of the extent of the official contact (1974:98). Different levels of label encapsulation Ii and subsequent label acceptance/rejection were then determined for each disposition. Label encapSUlation addresses the external process of societal labeling while the label acceptt.nce/rejection indicates the internal process of labeling. by !I. According to the labeling paradigm described in the study by Klein et aI., label encapsulation is mainly composed of three components. The first component is the number of contacts, namely the number of label-relevant contacts (e.g. two arresting officers, three juvenile officers, people at the referral agencies, etc.). The second component is label spread which is the number of persons informed of .the arrest (e.g. parents, people at school, etc.). The last one is label application which refers to the number of label applications. Essentially, label encapsulation distinguishes between those with more and less exposure to labeling. -288. r. 1.

(3) Informal Social Control of Crime -. Modification of Labeling Theory. specifying both internal and external processes in the labeling process, the relationship between each process with the behavior outcome could be examined. Additionally, the distinction between the two processes permitted a test as to which could better predict subsequent behavior. 12 Another objective was to test whether the invocation of identity change as an intervening process is a necessary theoretical refinement. Finally, the behavior outcomes were investigated. This paradigm clearly spelled out the labeling process: predicting behavior outcome not only from the initial official response (i.e. dispostiions), but also from the two intermediate stages, i.e. label encapSUlation and label acceptance/rejection. For a detailed representation of this paradigm, please see Appendix A. A detailed description of the original research design and the instrument employed were documented in Klein's final report (Klein et al. 1977). Only important background information is summarized here. It was an experimental study with randomization of four dispositions conducted in one cohort of respondents. The other cohort, as a natural control, had no randomization of different dispositions. The experimental cohort (N = 306) were drawn from arrest logs in nine stations (out of 18) of the Los Angeles Sheriffs Department. Members of the experimental cohort were considered referable by the policy personnel; they were not serious or habitual offenders. I3 The control cohort (N =412) 12 . IJ. Lemert posited greater importance of label acceptance and rejection than that of label. encapsulation. The research by Klein et aI., was designed to test this assumption.. By referable, we mean those who fall under the shaded area in Figure 7, bellow. Figure 7: The Experimental Design l Normal Police Project Dispositiolls. Randomized . / Dispositiolls. I. Detained Petition. 2: 'N()Jl·dctiline retiliOIl@. 3. Referred~" 4. Counseled and. ------..,)010. Released /.. 5. Not serious enough to be altcmlcd. ~. I. NOI1-dctuincd Petition. 2. Referred 3. Counseled and Rclc(lScd. Ii ~. ~r. ,. It is obvious that those who are considered serious enough to receive a detained petition and those who are not serious enough to require attention do not enter this project. -289. .. .... ..

(4) The JOI.:mal of National Chengchi University, Vol. 69, 1994. .... were drawn from,JGrest files of 33 out of 35 municipal police departments from the same metrq'Cfitan area. They were also referable but less delinquent than the experimental conort. 14 The control cohort were all referrals. Juveniles in the experimental cohort were arrested in August, September, and October of 1974. Youths in the control cohort were arrested in January, February, and March of 1975. Referrals in the experimental cohort were closely followed up, yielding a higher proportion of agency contacts for agency referred experimentals than for referred conJrols. Similarly, there was a far higher proportion of juveniles who were successn/Iy interviewed in the experimental cohort than in the control cohort. Separate analysis, therefore, was performed on each cohort in the study by Klein et aI. Three waves of data gathering were conducted for the experimental cohort, spanning 27 months, while only two waves of interviews were gathered for the control cohort, since data gathering on this cohort started at the same time when the Wave 2 interview was conducted on the experimental cohort. The time period between instant arrest and the Wave 1 interview was six months. Nine months later, the Wave 2 interview was conducted. The last interview was done a year after the Wave 2 interview. Only 185 (60% of the original pool) respondents in the experimental cohort and 129 (31 % of the original pool) in the control cohort were successfully contacted and interviewed in Wave 1. Fortunately, the loss of cases from interview refusals and unlocatable subjects does not appear to be biased. Comparison of descriptors in the experimental cohort with the interviewed subjects indicates that gender and mean age are the same while the proportion of minority subjects is 9% higher than that in the original pool. The control cohort showed some minor attrition effects; males are 6% more predominant, mean age is one year higher, and minority proportion is 4% higher among interviewed subjects. There seems to be no systematic bias related to response rate. 14 . As noted in the study by Klein et aI., neither experimental group nor control group could be described as seriously delinquent, their difference is striking. The control cohort is much like "diversion" groups, only mildly delinquent and unlikely to yield a high recidvism rate. The experimental cohort, though not composed of serious or habitual offenders, is nevertheless deliberately drawn from a more serious offender population, many of whose members might have eventually been petitioned to juvenile court. 290.

(5) Informal Social Control of Crime. Modification of Labeling Theory. In Wave 2, only 115 cases in the experimental cohort and 85 cases in the control cohort were interviewed, and Wave 3 has 80 cases interviewed in experimental cohort. The independent cohort has 85 cases interviewed in Wave 2. Table 2 shows the attrition over time.. Table 2: Attrition Over Time Cohorts. Wave 1 (6 months). Wave 2 (9 months). Experimental. 185. 115. Wave 3 (12 months) ----. -------. Independent. 80. -------. 129. 85. 2. Some Modifications in the Current Research This research advances current understanding of the labeling process in a number of important ways.. .. .". .. 1) Examining the Process Over Time. First, we follow up respondents over time, hoping to capture the mechanism leading to the behavior outcome. This approach overcomes limitations 6f crosssectional studies which fall short of capturing the dynamic process of label~. Since shaming or labeling involves change over time, the time dimension is an essential element of the whole process. Another merit of this follow-up design is to establish the causal order among variables. Attention is called nevertheless to the very limited number of follow-ups. Ideally, we hope that we can trace down any change in attitude and its associated behavior change over time so that a causal relationship could be established. Tests such as this demand observations at many time pgln~s. With two or three time points of observation, any conclusion about cau~lity should be . /' , conservative. -...,-I. 291.

(6) The Journal of National .Chengchi University, Vol. 69, 1994. 2) Distinction bet>1{een Predatory and Non-predatory Crimes , -,' . Second, we;1ifferentiate between two types of subsequent delinquency, predatory and non-predatory crimes. The reason for this distinction is that the shaming construct was originally designed by Braithwaite to apply to predatory crime. His assumption is that shaming will be more effective in societies where there is moral consensus about crime, and it will be less effective in areas where moral consensus is hard to reach. Moral consensus is greater for predatory crimes specifically. The author will argue that it it'worth testing both,. since societal consensus is also relatively clear in areas as.T fo what is "proper conduct" expected of juveniles. 3) Distinction between Formal and Informal Sources of Shaming. '" It is fair to say that the paradigm proposed by Klein et aI., underscored a process model - labeling-perception-behavior outcome, a typical focus of the labeling model. Their model made the distinction between two kinds of labeling: labeling from police and the justice system on the one hand, and labeling from social service agencies on the other. This research goes one step further by identifying a third kind of labeling which originates from interpersonal networks such as the family, relatives and significant others with whom individuals interact almost on a daily basis. The inclusion of this third kind has empirical support, which argues for a greater influence of significant others. Instead of separating the justice and social service types of labeling in Klein et al.. this research will combine them together for the. following two reasons. First, people from both judicial and social service agencies carry with them negative social stereotypes toward their clients (i.e. "bad kids" from justice system and "sick kids" from social service agencies). Second, the study by Klein et al., in their analysis of the Wave 1 interviews, did not show a statistically significant relationship between the two types of labeling and arrestees' subsequent behavior outcome (1977:46). As a result, this research combines the two as one source of shaming. the other source of shaming is individuals' interpersonal network. As stated in the previous section, informal networks (i.e. significant others) are postulated by both symbolic interactionism and reference group theory to have a greater influence upon individuals' behavior than a formal one. It merits testing out this proposition.. -292. .:J~I. ~ .".;'~ ~~': ~ :,0·..

(7) ~i,~. Informal Social Control of Crime. Modification of Labeling Theory. 4) Emphilsis on Perception and Interpretation of Societal Labeling Rather Than Types of Official Labeling An important manipulation in the study by Klein, et aI., is the different official dispositions to which arrestees are randomly assigned. The purpose of distinguishing the four dispositions is to grade the different levels of seriousness of labeling associated with each disposition. The theoretical rationale underlying such a design is that the greater the labeling, be it either system penetration or the magnitude of the label spread or the content of the label, the worse its consequence on subsequent behavior. However, this distinction becomes less relevant in our shaming model in that shaming is defined to be associated more with individuals' definition and interpretation of the societal response (Le. the four official dispositions in this case), and less with the societal response itself. In other words, only when individuals interpret an event as significant will that event become significant in its consequence. In this sense, a more serious social labling (Le. deeper penetration into the justice system) is not necessaily related to a greater shaming effect because of the individual variation in perception of societal reaction. Furthermore, research suggests that the attitude of the individual toward labelers may be more important than the dispositions in determining the degree of susceptibility to the labeling process. Hirschi has conducted a study examining studies by others on the various treatment programs. IS He found that various treatment programs failed to show differential labeling effects (1975). Even the study by Klein et al. found that the four dispositions made no significant difference in arrestees' l self-report subsequent delinquency (1977:43-44). .. Therefore, this research does not distinguish the testing out their respective behavior consequences. Instead, what matters to this study is juveniles' perception of others' response rather than the response itself.. diSPosition~4n. 5) Self-report Delinquency Self-report information instead of the official record will be ~n this research. IS . Three studies cited by Hirschi as serious blows to the labelingAJetspective are the Cambridge-Somerville Youth Study as reanalyzed by McCord and McCord (1959), the Silverlake Experiment by Empey and Lubeck (l971), and the Provo Experiment by Empey and Erickson (1972). . 293. ,,;.. .". ..

(8) The Journal of National Chengchi University, Vol. 69, 1994. ,. the rise of self-reIl9:;-measures is a response to questioning the adequacy of official statistics. For ex~#iple, Schur (1973: 156) pointed out that traditional "correlates of crime" are not correlates of criminal behavior. It is generally recognized that official record is more representative of those with more serious delinquency, and underrpresentative of a wide range of minor offenses. Therefore, self-report may serve better to discover "minor offenses" that police are less likely to respond to with an arrest. The other reas9,Ji for choosing self-report is its consistency with the data in the emergence of ttl labeling perspective. Labeling theory gained support from self report research indicating that "the behaviors we now call delinquency are extremely common throughout the entire society" (Schur, 1973:82). Although not devoid of flaws,16 the relative advantages of using self-report records have been discussed extensively in the literature (Hinde lang et. al., 1981). For instance. self-reports allow for the study of relevant behaviors prior to officiill actions (Tolan and Lorion, 1988:548). Self-report can generally serve as an index of delinquency involvement. Further. it may provide adequate variation in "delinquency" within populations that would be uniformly "non-delinquent" if official meansures alone were the criterion (Hirschi. Hindelang, and Weis, 1981:474). In addition, it avoids the problem that the police respond to their own former reactions. Lincoln et aI., (1977) who used the same data set also noted this problem: different official dispositions yielded sta~istically significant differences in official recidivism rate, a relationship which self-report fails to support. An initial examination of the self-report delinquency as reported in Table 3 reveals a striking discrepancy between self-report delinquency and self-report arrest, let alone the official arrestY. 16 . 17 . Criticism against the use of self-reports are, to name a few, its overemphasis on the trivial offenses, the underreporting of the seriousness of offenses, and difficulty in establishing reliability and validity (Nettler, 1974; Reiss, 1975; West, 1973). Self-report delinquency showed that in Wave 1 only four out of 314 juveniles reported no commission of anyone of the 18 offenses listed in the questionnaire, while the self report arrest showed that 129 juveniles reported no subsequent arrests. In other words, most juveniles committed offenses and yet were free of official consequences. This suggests that arrest records are a poor indicator of juvenile delinquency. 294.

(9) Informal Social Control of Crime -. Modification of Labeling Theory. Table 3: Simple Receidivism Over the three Waves. ----. Wave 1. Wave 2. Wave 3. 314. 200. 80. Self-report Arrestees. 183. 73. 25. Self-report Recidivists. 310. 177. 80. Subjects Total Ns. ----. Perhaps, the most important for the popularity of self-report is that it provides a more valid measure of delinquent behavior than do official records.In this research, it is particularly problematic to draw a valid causal relationship between official . reaction and subsequent behavior outcomes as officially recorded. It was suggested by Klein et aI., that "the labelers are somehow responding to their own prior decision" (1977:30) such that official records show a significant difference of official reactions in predicting officially recorded subsequent bheaviors while self-reports fail to do so. In the current research, self-reports are particularly pertinent because they shift the focus from legal-judicial reactions to the involvement in delinquent activities. As mentioned before, official records are records of legal-judicial reactions to offenses known to the police. They inevitably underestimate the extent and magnitude of crime involvement. In comparison, self-reports are better, if not perfectly accurate, indicators of delinquency involvement.. -. 3. Data and Analytical Strategy. /. 1) Data. .'. Effective Sample Size Since the distinction of different dispositions is not relevant to the shaming model, both cohorts will be analyzed jointly. Therefore, the analysis start out with 718 arrestees (combining 306 from the experimental cohort and 412 the control cohort). However, when the 718 juveniles were arrested, no inter:f1.ew was done so that a measure of their shaming state is missing. Therefore, >-vje 'are forced to choose the Wave 1 interview as the starting point of the study since only from. fim. -295. ~. .". ....

(10) The Journal of National Chengchi University, Vol. 69, 1994. then can w~ "ttitce the change of their shaming state and its effect upon their subsequent. ,~haviors. The sample size, therefore, was reduced from 718 to 314. Over half of the original sample attrited in subsequent waves. Some could not be located, while others refused to be in the study. With more attrition occurring at the second wave and third wave, only 200 cases were interviewed in Wave 2, and out of the 200, 80 cases completed the third interview. The reason for having such a few cases in Wave 3 is, asi~ from the attrition problem occurring in the experimental group, no Wave 3 in{~rview was ever performed in the control cohort. Since the sample size shrinked drastically at wave 3, it is decided to drop wave 3. Therefore, the following analysis is performed on the 199 cases (one case deleted due to missing value on shaming variable) .. .... ... 2) Strategies of Analysis Analysis of Change Specifically, the first set of models estimated is for those who participated in the first two interviews. The analysis on the first interval provides us with insight as to how and to what degree the predictors measured in Wave I relate to the subsequent delinquency as reported in Wave 2. What can not be learned in this analysis, however, is the degree of change occurring over a longer time period. For example, shaming is assumed to have an impact upon behavior outcome. But it is not clear when the effect of shaming occurs and whether that effect will hold constant over time or will fade away eventually. We can not conclude that shaming has no significant effect on behavior outcome if an effect is absent in the first time interval. It is possible to observe a shaming effect later on. However, the data can not support test of a longer time due to a small sample of remaining cases. Therefore, any interpretation based on the analysis of the first time interval should be careful. In essence, the approach to analyze this type of data is panel analysis. A typical equation is:. Y2=a+J3I YI +132 X1.+e2. [1]. where the subscript indicates the time order. -296. -.

(11) Informal Social Control of Crime -. Modification of Labeling Theory. This equation shows quite clearly the contributions of stability and change in the influence of Y! on Y2. If the distribution of Y stays quite stable over time (i.e. little change), then little is left for X to account for and bl approaches unity. the distribution of Y changes over time, we see a smaller value of bl, indicating a departure from static equilibrium. With the inclusion of the lagged values of the dependent variable on the right hand side of the regression equation, OLS (ordinary least squares) no longer yields unbiased estimates of the regression coefficients regardless of the sample size. This is because the basic OLS assumptins about the error term do not hold in the case of panel data due to the serially correlated error terms. Figure 6 is a diagrammatic representation of the serial correlation of the error terms. Figure 6: Serial Correlation of the Error Terms. m. Ct . j. 1 Yt . )0. Y,. .. . .... Y! is correlated with. e2 through the autocorrelation of the disturbances. The result is that the OLS estimate of the effect of Y! on Y2 wiiI include the spurious covariance of the V's due to the disturbance autocorrelatiol. Consider the following measurement model describing the relationship between 'observed values, Y, true values, y, and measurement errors, .€ Y=y+€ . . [2]. Since y measured at the two time points is an imperfec,t indicator of Y, the substitution of Equation 2 to 1 generates:. f' .. 'r. (Y2-€2)=a+{3t(Y! -€!)+{32X2+d •. Y2=a+ {31 YI +{32X2 + [(e2 +€2) - {3!€i) -297. [3].

(12) 1 The Journal of National Chengchi University, Vol. 69, 1994. lIThe OLS estima!tir assumes that the regressor is independent of the disturbance. But equation 3 'loes not satisfY this assumption, since the compound disturbance term consists of the measurement error contained in YI ({31tl). Ways to deal with the serial correlation are extensively discussed (see Turma and HaMan, 1984; Markus, 1979; Kessler and Greenberg, 1981; Sayrs, 1989). GLS (generalized least squares) or WLS (weighted least squares) are two of the approaches frequently mentioned that can produce the best linear unbiased estimates of structural parameters. 18. I". Modeling Strategy. , First, juveniles with high shamability and low shamability are separated. Analysis are performed on each set of case. The following analysis explicates various sources of influence upon subsequent behaviors by adopting a hierarchical modeling strategy. Thus, models are specified by introducing first the most exogenous variables _ social and demographic factors. Prior delinquency level and duration since the latest offense are added to see to what extent the outcome measure is modified by these variables. Last, external shaming, both informal and formal are investigated after controlling for the previously entered variables. SO 1 . social and demographic factors inter-wave period. +. social and demographic. SD3 = prior delinquency level + inter-wave period factors + external shaming * SO, subsequent delinquency.. +. social and demographic. S02 = prior delinquency level factors. +. With such an analysis, we can examine the variance in recidivism accounted for by each set of variables separately as well as the total explanatory power. The dependent variable is self-report delinquency. Since virtually everyone was 18 . The SAS regular regression procedure does not support GLS or WLS (weighted least squares). But one procedure called GLM does offer this option. So, the reported analyses are based on the estimation of procedure GLM. In fact, if no serious autocorrelation is present, OLS could still gererate fairly efficient estimates. To know whether any serious autocorrelation exists, comparing the results from OLS and GLS will give the answer. If results from both are essentially the same, then OLS is a good approach too. In fact, results from both estimation techniques are very close. 298 . "'~~. :.I.

(13) Informal Social Control of Crime -. Modification of Labeling Theory. a recidivist, we are not interested in distinguishing those who recidivated from those . who did not. Instead, we are interested in examining how shaming impacts upon the level and type of subsequent delinquency. In addition, we break down the dependent variable into predatory offenses and non-preatory offenses. Each model will then be estimated for both types of dependent variables.. 4. Hypotheses According to the model discussed and the analytic strategy described above, hypotheses are generated to test the shaming model. The following hypotheses posit the specific relationship between self-report delinquency as a dependent variable and its postulated determinants as independent variables, such as shaming, and social and demographic variables. Each hypothesis contains predictions for subsequent total offenses, subsequent predatory offenses, and subsequent non-predatory offenses.. Hypothesis on Juveniles with High Shamability HI. Given the same higher shamability, those with a perception of greater reintegrating shaming from informal sources will commit fewer offenses (including predatory and non-predatory) than those with a perception of greater stigmatizing shaming. H2. Given the same higher shamability, those with a perception of greater reintegrating shaming from informal sources will commilt fewer offenses (including predatory and non-predatory) than those with a ~ception of greater stigmatizing shaming. " ·H3. Given the same higher shamability, but different sources of shaming, informal reintegrating shaming has a greater positive effect than formal reintegrating shaming upon subsequent offenses (including predatory and non-predatory).. Hypotheses on Juveniles with ww shamability . ( T'. Shami::Jt~nd. H4. Given the same lower shamability, informal source of formal source of shaming alike, either reintegrating or stigmatizing, have no effect upon subsequent delinquency.. -299. " .,. ....

(14) The Journal of National Chengchi University, Vol. 69, 1994. 5. operationalizatio~. i',r:. 1) Independent Variables. .... Prior History refers to prior delinquency level. The self-reported delinquency becomes available in the follow-up interviews. It is measured as the number of times the listed eighteen offenses have been committed since last interview. This prior history variable does notlrefer to juveniles' delinquent history before their instant arrest. The first wave mbsure covered the period following the arrest that produced project involvement. However, it is causally sound to use this measure gathered in Wave 1 as the prior history to predict subequent delinquency in Wave 2. Therefore, prior delinquency level here refers to the self-reported delinquency as reported in the previous interview. Duration refers to the time interval between two interviews. The length of time juveniles received before the next interviews varies. It is measured by using week as the unit.. Shaming is of two types. Internal shaming is defined as how one cares about others' opmlons. It is an indicator of shamability. Shame felt by individuals upon committing the acts that led to the arrest, or upon the discovery of crimes by others, can be activated by imagined or anticipated opinions of others. Since the invocation of internal shaming is determined by whether one cares about others' opinions, internal shaming does not always occur. That is, some are more prone to shaming (Le. shamable) than others, depending upon the extent to which they care about others' opinions. Whether individuals are shamable or not is inferred from the answer to the following question: "Does it matter to you what - thought of you or not?" (where the blank refers to police or social service personnel in the data set). The answer to this question reflects whether respondents care about others' opinions. Those who answered "Yes" are defined as high shamable, and were assigned a value of 3, and those answering "No" as low shamable with a value -300.

(15) Informal Social Control of Crime. Modification of Labeling Theory. of l. Those whose answers are "not sure" are in the middle with a value of 2. Since the same question was asked about all police and social service workers who had contact with juveniles since their last arrest, more than one perceived opinion from justice system and social service agencies were reported by the juveniles. Therefore, scores were summed and divided by the number of times answers were given. Thus, the scores range from 1 (do not care) to 3 (do care). Unfortunately, questions asked of the juveniles' perception of their significant others (i.e. parents, friends, teachers, and so on) were absent. This indicator, therefore, is weak in determining the degree of shamability. For instance, juveniles may not care about police or counselors' opinions of them but they may care about how they were viewed in the eyes of other significant people. Therefore, if this weak variable shows any effect, we may infer a greater effect of the generic construct. Consequently, any conclusion related to this effect should be cautious. External Shaming refers to the opinions of others, both positive (reintegrating) and negative (stigmatizing), including family members, friends, priests, teachers, police, and social agency personnel. We do not have a direct measure of the others' opinions in our instrument. Instead, the juveniles were asked what they thought others felt about them. This direct measure is not important in this study since what influences the course of action more is one's interpretation of the reality (the subjective perception of others' opinion) rather than the reality itself (others' opinion).19 Social problems, as noted by Cohen, consist not only of a fixed and given condition, but the perception and definition of people to whom this condition poses a threat (1971: 14). In other words, it is our perception of others' opinion }hat counts most. To symbolic interactionists, the gesture of reacceptance vS'1ejection as perceived makes a big difference in an individual's life: a perceived refoction gesture .can drive one farther away from the conventional world to the situation where one finds no any other life alternatives. kaplan, et. al. also noted the significance of 19 . Different types of people have different interpretations of reality. For example, the fact that "others still think they will do O.K. in life" will be interpreted differently. To the unshamable juveniles, their interpretation of others' O.K. gesture may meatthat being arrested is no big deal since others still think they are O.K. As a result, theY;n"lay continue their deviant life. In comparison, those who are shamable will be more li~ly to resume a normal life after being ashamed by the forgiveness and tolerance of ot1lers. especially the significant others. Therefore, it is the subjective interpretation of others' opinions rather than others' opinions that have more impact upon people's behaviors. -301. .. ~. '".

(16) The Journal of National Cheng chi University, Vol. 69, 1994. subjective recogniti0h-of the source of self-rejecting or self-accepting feelings as motivating devian!.fehavior (1988). The perceived opinions were inferred from the following two questions: Does -. (significant other) think you're someone who will do O.K. in. life in things like school, jobs, having a family and so on, or not?" Does (signifiqmt other) think you're a person who will do something that will get y~ into trouble, or not?" 1. .... "Yes" answers to the first question and "No" answers to the second question. Both infer a reintegrating shaming since it is a reacceptance gesture from the community. In contrast, answers "No" to the first question and "Yes" to the second question indicate a stigmatize shaming, a rejection gesture. Answers of "Yes-Yes" or "No-No" to both questions connote something in between reintegrating and stigmatizig shaming. For example, there is the possibility that juveniles perceived that their significant others still believe they do O.K. in life, and yet think they will get themselves into trouble somehow. The coding for the external shaming is from 3 (reintegrating shaming) to2 (having something of both types of shaming) to 1 (stigmatizing shaming). Again, since juveniles were asked for their perception of various others' opinions, we have to sum the scores and divide by the number of answers given. Therefore, the range is from within 1 (they will get into trouble, stigmatizing) to 3 (they will still be ok, reintegrating).20 External formal shaming refers to perceived shaming from an official or impersonal settings. Police, court, and social service agencies are the major sources of formal shaming where individuals find little personal relevance and emotional 20 . By simply summing up scores, we assume an equal weight, which may not be correct. The final decision to give equal weights to each opinion is both a theoretical and empirical matter. Theory suggests more weight should be given to opinions from those to whom we are more attached. Therefore, theoretically, there should be a differential importance in terms of the opinions given by parents, teachers, peers, and other acquaintances. Previous literature does not examine the effect of the whole range of significant others' opinion. Among the few that examine significant others' importance in orienting our behavior, none except the peers has a direct influence on subsequent delinquency (Elliott, et. aI., 1985). This finding was supported by what was found in this data set. We regress juveniles' behavior outcome on the opinions of each. None of the opinions shows significance in influencing one's behavior outcome. 302.

(17) Informal Social Control of Crime. Modification of Labeling Theory. attachment. Our instrument includes arresting officer, juvenile officer, and counselor whose opinions, as perceived by juveniles, are considered to be official shaming. A higher score on this variable suggests greater reintegrating shaming. External Informal Shaming refers to perceived shaming from personally valued significant others. Family members, associates, peers, and immediate community members are those with whom one has interpersonal interaction, usually on a regular or even daily basis. In our instrument, the informal network includes arrestee's mother (or stepmother), father (or stepfather), guardian, close friends, other teenagers at school, neighbors, teachers and other school officials, and minister or priest or rabbi. A higher score on this variable indicates greater reintegrating shaming. Social and demographic descriptors that are included in this research are gender, ethnicity, age, and involvement in the conventional world such as school, church, employment, family, and adult-sponsored activity. Age refers to the chronological age, as reported by the respondents. Age is time-dependent so that we have age as reported in Wave 1 and in Wave 2 and in Wave 3. Gender is defined as male or female. Ethnicity is defined as the ethnic group with which respondents indentify themselves. There are five categories, namely, Anglo, Black, Mexican or Spanish American, Asian, and others. We combined Mexican and Spanish American together as one group, and Asians and others together as another due to the few in each. Involvement in conventional actlvItIes is measured as the combined extent of importance juveniles felt about finishing high school, going to chufch, finding a job, establishing a family, ~nd participating in adult-sponsored acti~y. 21 The extent of importance in each indicator is measured from 4 (very impori~nt) to 1 (not at all important).. 21 . Here, we use attitudinal measurement instead of behavioral ones as an indicator of juveniles' of conventional involvement. Two reasons for choosing, I~he attitudinal measure are: one, the response scale to questions asked of their behatfors is quite a subjective judgement. Whether juveniles think they do a lot, or some, ~~. a few, or not at all means something different to each juvenile. My "a lot" may·..,ije 'your "some", or even "a few"; two, attitudinal measures, though not measuring real behavior, reflect strength of basic orientations. When something me~s important to one, it will have a greater impact on one's behavior orientation. 303. .. .". ...

(18) The Journal of National Chengchi University, Vol. 69, 1994. Each indicator-is inferred from the following questions: "How i~ant is it to you that you .finish high school?" "How impOrtant is it to you that you spend time in activities related to your church or temple?" ,~ . "How important is it to you that you will be able to get a regular job in the future?" "How important is it to you that you will get married and have a family .,' someday?" "How importifnt is it to you that you spend time in school-sponsored organizationJ or clubs?" "How important is it to you that you spend time in other adult-sponsored organizations or clubs?". ..... Therefore, the importance of conventional activities has values ranging from 24 (the most attached) to 6 (the least attached). Since not everyone answer all six questions, individual scores are detennined by summing over individual answer items and dividing by the number of items answered. 22 The effective score range is from 1 (least involvement) to 4 (most involvement). 2) Dependent Variables. Subsequent Delinquency, is defined as delinquent acts committed since the last interview. In order to test the shaming effect on predatory vs. non-predatory offenses, further distinction of subsequent delinquency into these two types of offenses was performed. Self-report delinquency includes the following 18 offense' items: 1. Running away from home. 2. Taking things worth more than $50.. 22 . Again, we encounter a situation similar to that of determining whether different weights should be given to some variables that are assumed to have a greater influence on behavior outcome. In the absence of prior literature suggesting which is more important, finishing high school, getting a job, establishing a family, going to church, etc., in influencing juveniles' behavior, we run a regression by regressing behavior outcome on the five types of activities. None shows a significantly greater relationship than the others in influencing juveniles' behavior. Therefore, we assigned an equal weight to each item. -304.

(19) '". Informal Social Control of Crime. Modification of Labeling Theory. 3. Smoking marijuana. 4. Driving a car when drunk. 5. Taking something from a person by force. 6. Being out after lO at night without your parents' permIssIOn. 7. Beating up on somebody or fighting people physically. 8. Cutting classes, or staying away from school without permission 9. Sniffing glue or cocaine or taking pills and so on. 10. Taking a car without the owner's permission. 11. Buying or getting something that was stolen by someone else. 12. Going onto school grounds when you shouldn't have been there.. 13. 14. 15. 16. 17. 18.. Breaking into a place and stealing something. Disobeying your parents or guardian about important things. Taking things worth less than $50. Carrying a gun, a knife, or other kind of weapon. Drinking any liquor or wine. Setting fire to buildings, trash, or other things.. Total Offenses The operational measure of total offenses is simply the cumulative number of subsequent offenses self-reported per item, summed over all 18 items over the period since the last interview. The response category for each item has a value from o (not at all) to 4 (those who committed offenses more than 3 times) with the last category truncated. This truncation problem is recognized but was foUnd to be minor by Klein et aL (1977:42). Thus, the maximum total is 72, ind the minimum ,,' is O. Predatory vs. Non-predatory Crime / Predatory crime could be defined either as crimes against persons and property, or as crimes in which the intent is the concern (Glaser, 1978:6-8). In brief. predatory crimes are defined as crimes against persons both willful and negligent. and against property with intent. Willful predations against person include simple assault, murder, non-negligent manslau~t~r, aggravated assault, rape, and statutory rape. Criminal negligence predations a;ainst persons are of life such as crimes either in which there are unintended injuries or even auto accident, or in which victims are only potential, such as speeding and reckless driving. Crimes against property with intent (i.e: property crimes) include all willful. lois'. .-305 . ~.~ I '".

(20) ,... ~. The Journal of National Chengchi University, Vol. 69, 1994. ,r.. .... predations for rrfoney or other goods, such as theft (or larceny), burglary, fraud (I.e. forgery,' -{mbezzlement. and confidence game), robbery, criminal pollution, criminal invasion of privacy, criminal electioneering, and cruelty to wildlife. Non-predatory crimes could only be classified as such when they have a complaining audience or because of what is sold, purchased, used, or possessed (Glaser, 1978:9-11). Glaser further distinguished nonpredatory crimes into five types. Other than disloyalty, a crime not relevant to the juvenile case, the other four types are themselves qUF diverse and not easy to define. The first type is called iIlegal performance of~nses, offenses which victimize some spectators or listeners, such as public drunkenness, indecent exposure, vagrancy, and disorderly conduct. The second type is illegal-selling offenses, such as sale of illegal drugs, as well as of gambling and prostitution services. Sales of stolen goods and of government services (e.g. bribes) are also included. The third type is illegal consumption, including the purchase or use of illegal goods or services. Usually, these crimes are by consent of all participants. The last type is illegal-status offenses, which means those not chargeable to adults, e.g. runaway or habitual truancy. It should be noted that whether an act is classified as nonpredatory crime or not varies with cultures. Following the descriptions above, this research divides the 18 delinquent acts into either predatory or nonpredatory crime as follows: Predatory crimes are (1) taking things worth more than $50; (2) taking something from a person by force; (3) beating up somebody or fighting people physically; (4) breaking into a place and stealing something; (5) taking things worth less than $50; (6) carring a gun, knife, or other kind of weapon; (7) setting fire to buildings, trash, or other things; (8) taking a car without the owner's permission. Nonpredatory crimes are (1) running away from home; (2) smoking marijuana; (3) being out after 10 at night without your parents' permission; (4) cutting classes, or staying away from school without permission; (4) sniffing glue or cocaine or taking pills and so on; (6) disobeying your parents or guardian about important things; (7) drinking any liquor or wine; (8) buying or getting something that was stolen by someone else; (9) going onto school grounds when you shouldn't have been there; (10) driving a car when drunk. The measure of predatory offenses is the cumulative number of subsequent predatory offenses self-reported per item, summed over the eight items over the period since the last interviews. The score ranges from 0 to 32. The non-predatory measure is obtained in this same manner, thus yielding a maximum total of 40 and 306 .

(21) If. Informal Social Control of Crime -. Modification of Labeling Theory. a minimum of O. DATA ANALYSIS AND RESULT Table 4 presents the mean, standard deviation, and statistical significance of the differences, while Table 5 presents the correlation matrix for all cases in the first time interval. This information provides us with an initial understanding of the relationship and distribution of variables of interest. An examination of the correlation table gives an initial picture of the strength and direction of the relationship between each pair of variables. Offenses reported in the prior interview are strongly related to offenses reported in the next interview (.597). As to the degree of involvement in conventional activities, the higher involvement, the fewer offenses (- .256) and the more perceived reintegrating shaming (.322 for informal as well as .233 for formal). The perceived reintegrating shaming in negatively associated with subsequent delinquent level (- .451 for informal as well as - .2656 for formal). Age is not associated with either prior delinquent level or subsequent delinquency. Nor is it related to any types of shaming. With regard to ethnicity, being black is negatively related to the number of prior ( .221), and subsequent offenses committed (- .266). Being black is positively related to perception of greater informal reintegrating shaming (.274). In contrast, being white is negatively related to a perceived greater informal stigmatizing shaming (- .144). Being white is also positively associated with more subsequent offenses (.189) As to gender, males and females are almost at the same levels of delinquency,' commitment in coventional activities, and shaming (almost every correlation is lon-significant and .. the strength of association is nearly zero).. ,. (' ,. 307 . -. .4(. ....

(22) The Journal of National Chengchi University, Vol. 69, 1994. Variable. '". I• •. it~;. AGE PRIOR! PRIOR2 COMMIT INTSHAM EXIFSHM EXFMSHM WHITES OTHERS HISPANIC BLACKS MALE PREDI PRED2 NPRED! NPRED2 PERIOD. , .r . " . lable 4: Mean and Standard Deviation " N Mean Std Dev A B B A B A. 199 197 199 199 198 198, 198 199 199 199 199 199 197 199 197 199 199. 80 79 80 80 79 79 79 80 80 80 80 80 79 80 79 80 80. 15.982 27.228 24.789 2.861 1.751 2.468 2.192 0.412 0.045 0.352 0.191 0.764 8.30 6.68 18.93 18.10 38.36. 16.294 27.177 22.675 2.878 1.783 2.463 2.176 0.287 0.062 0.400 0.250 0.775 8.53 6.04 18.64 16.64 31.74. 1.798 16.277 15.035 0.511 0.663 0.432 0.661 0.493 0.208 0.479 0.394 0.426 8.44 7.58 9.42 9.02 8.84. 1.636 17.147 13.999 0.550 0.690 0.441 0.671 0.455 0.243 0.493 0.436 0.420 9.10 6.89 9.30 8.45 3.35. P ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns ns .001. Subscript: 1 - Refers to prior delinquency; 2 - Refers to subsequent delinquency A - Statistics for the Total Sample; B Statistics for the Sub-sample P significance test. 308 .

(23) Informal Social Control of Crime -. Modification of Labeling Theory. Table 5: Pearson Correlations for Total Sample AGE PRIOR'. PRIOR' COMMIT INTSHAM EXIFSHM. EXFMSHM. WHITES OTHERS HISPAN BLACKS MALE. PERIOD. PRED'. NPRED'. AGE PRIOR' I.lXXl a 199b 0.112 197 -0.01l 199 -0.115 199. PRIOR'. COMMIT INTSHAM EXIFSHM EXFMSHM WHITE. 0.597" 197. MALE. PERIOD. LIXXl. 199. 0184 198. -0.164' 197. -0.107 198. 0.244" 198. 1.000. 0.019. -0.496" -0.451" 197 198. 0.122" 198. 0.181' 198. I.IXXl. -0,344" -0.165" 198 197. 0.233" 198. 0.271" 198. 0.405" 198. 0.013 199. 0.085 197. -0,034 199. 0.045 197. -0.001 199. 0.074 197. 0.006 199. BLACKS. 197. I.IXXl. 0.002 199. HISPAN. I.IXXl. -0.263" -0.2)6" 197 199. 0.039 198. OTHERS. 199 198 198 1.000. 198. -0.141' 199. -0,078 198. -0.144' 198. -0001 198. 1.000. -0.009 199. 0.107 199. -0.008 198. -0.037 198. -0.027 198. -0.182' 199. I.IXXl. 0.028 199. -0.080 199. 0.049 198. -0.059 198. -0.056 198. -0.617" 199. -0.160* 199. 0.041 199. 0.214" 198. 0084 198. 0.189" 199. -0.111" -0.266" 197 199. 0.226" 199. 199 199. . .... 1.000. 199. -0.407** -0.106 199 198. -0.358" 199. 1.000. -0.050 199. 0013 199. -0001 199. ,.. 199. 0.072 197. 0.144' 199. -0.018 199. 0.025 198. -0.015 198. -0.181' 198. 0.009 199. -0.1 W" cO.0I2 199 197. 0.180' 199. 0.003 199. -0.048 198. -0.051 198. 0.003 198. 0.138" 199. -0.030 -0.094 199 199. -0.16{-0016 199 199. 1.000. -0.012 199. 0.901" 197. 0538*' -0.155' 199 199. -0.1l2 198. -0.431" -OJIO" 198 198. 0.047 199. 0.019 199. 0.070 199. -0.133 199. 0.001 199. 0.921" 197. 0.550" 197. -0.165' 197. -0.471" 197. 0.104 197. 0.102 197. 0.015 197. -0.263** -0.051 197 197. -0.429" -0:271" 198 198. 0.071 199. 0.098 199. -0.041 199. -0.186"1- 1'11 199 . 199 >, ,. -0.391" -0.114". 0.256** -0035 199 199. 0.019 199. 0.204" 197. -0.315" 197. -0.316" 197. I.IXXl. 199. I. 0.196' 199. 199. -0.021 197. i. PRED'. -0.067 199. 0.545" 197. 0.887" -0.232" -0.045 199 199 198. NPERD'. 0.035. 0537". 0.912" -0.132". -0.141'. ,-. <0.05; .. <0.01: Supeftriplnumiler i:1dicales lime point a - Correlation coefficient b - Toral Ii. 309. -0.28~ 199. 0.074 199. 0.125 199 0.195" 199.

(24) The Journal of National Chengchi University, Vol. 69, 1994. Table 6: Unstanaardized Regression Coefficients for Subsequent Delinquency (N = 107) . ,} High Shamable Juveniles TOIal Offenses ml INTERCEPT AGE MALE. .... 1_.. COMMIT ETHNICITYI White Hspnc Other. m2. 57.48**" 19A8 (l7.64)b (16.85) -.89 .54 5 (.73) ) ~02 2.68 I (3.53) (3.01) -6A7* -4.88 (3.09) (2.67). to. 1.14 (4.61) 3.66 (4.57) 5.16 (8.90). DURATION PRIOR Predatory. -1.93 (4.02) 1.72 (3.92) -.34. Predatory Offenses m3. ml. m2. 41.33* (17.31) -.57 (.70) 1.25 (2,91) -.25 (2.91). 17.73 (12.25) .07 (.65) 2.51 (2.42) -4.01* (1.84). 7.25 (12.89) -.04. -1.85 (3.04) 2.92 (2.82) -2.28 (4.87). -2.44 -3.44 (2.99) (2.83) .62 2.76 (2.75) (2.58) -4.00 -1028* (4.n) (4.71) .12 .12 (.09) (.08). .34*. -4.02 (3.88) -.75 (3.80) -4.23 (7.53) .30*. (.13). (.13). A6** (.09). .32** (.09). (7.73). Nonpredatory SHAMING Exifshm Esfmshm R2 P. .08 .33. .34 .001. -.70* (2.95) -3.16 (2.05) A3 .001. .17 .053. (.65). 2A2 (2.55) -3-88* (1.79). m3. -310. ml. m2. 26.06* 35.76** 8A7 (12.88) (10.32) (10.17) -.23 -.58 .34 (.60) (.48) (.42) -A8 .93 1.97 (2A5) (2.03) (1.88) .10 -2.57 -lAO (1.99) (1.88) (\.65). 0.26 (.\5) 0.12 (.19). 0.29* (.14) -.11 (.18). .28 .01. -4.93* (1.96) -3.44* (1.45) .42 .001. in parenthesis is standard error. .05); ** « .01); P (significance level for the overall model).. I (blacks are the reference category).. Exifshm (external informal shaming); Exfmshm (external formal shaming).. *«. Non-Predatory. 1.52 (2.72) -JI (2.73) 6.39 (5.76). .05 .53. m3 17.95 (11.11) - .32 (AI) 1.56 (1. 90) .33 (1.80). -1.28 (2A2) -1.85 (2.37) 4.61 (5.05) .23** (.08). -2.06 (2.40) -2.61 (2.36) 3.97 (5.07) .21* (.08). 0.05 (.12) 0.51** (.15). .005 (.12) AI* (.16). .33 .001. - 3.87* (1.87) -.52 (1.29) .37 .001.

(25) Informal Social Control of Crime -. Modification of Labeling Theory. Table 7: Unstandardized Regression Coefficients for Self-report Delinquency (N =92) Low shamable Juveniles Total Offenses ml INTERCEPT AGE MALE COMMIT ETHNICITYI White Hospnc Other. m2. Predatory Offenses m3. 26.98 a (l7.66)h -.01. 30.27 11.30 (18.55) (17.45) -.38 -.62 (.85) (.85) 9.82** 10.27** 13.60** (3.34) (3.35) (3.57) -4.17 -6.15* -4.5S (2.59) (2.77) (2.53) IS.05** (4.40) 6.67 (4.57) 4.47 (6.S4). DURATION PRIOR Predatory . 11.87** (4.04) 6.43 (4.12) 2.87 (S.92) .32* (.16) .39**. 10.02* (4.08) S.53 (4.08) .50 (5.94) .30* (.IS) .33**. ml. m2. 15.52 (10.84) .11 (.56) 8.82** (3.11 ) -4.4S* (1.92). !l.40 (11.55) -.53 (.60) 6.32* (3.12) -4.4S* (1.96). 22.43 (11.90) -.05 (.62) 5.48 (3.11 ) -3.34 (1.97). 4.46 (2.79) 2.81 (2.98) -.14 (4.05). 3.02 (2.66) 2.85 (2.82) -1.43 (3.85) .17. 2.23 (2.76) 2.67 (2.80) 1.74 (4.09) .16. (.ll). (.10). 0.28 (.15) (.IS). 0.32* (.14) -.04 (.16). .30 .001. -6.85* (2.59) -.49 (1.75) .37 .001. .10. Nonpredatory SHAMING Exifshm Exfmshm R2 P. .26 .001. .41 .001. -1O.lS* (3.92) 1.90 (2.35) .45 .001. Non-Predatory. .18 .016. m3. ml. m2. 11.62 1.16 (9.54) (9.55) .25 -.Il (.49) (.51) 4.66** 3.94* (1.74) (1.81) -2.33 -1.23 (1.42) (1.4» 11.29** (2.32) 5.38* (2.39) 6.05 (3.48). 9.48** (2.20) 4.8S* (2.2S) 5.53 (3.24) .20* (.08) 0.07 (.12) 0.28* (.13). .. 8.98 (10.44). -.06 (.53) 4.25* (1.87) 1.41 (1.41) 8.71** (2.23). 4.46. (2.26). 4.53. (3.26). .19* (.08). .06 (.12). .25. (.14). -3.99 (2.18) 1.31 (1.27). ,I. .42 .001. .30 .001. * « .05); ** « .01); P (significance level for the overall model).. I (blacks are the reference category). Exifshm (external informal shaming); Exfmshm (external formal shaming).. m3. .44. .001. ,. t. ('. .. ....,-I. ". ,. Due to the nature of the dependent variable and the goal of the study, regression analysis is performed to test the hypotheses: Unstandardized beta coefficients are -311. - ... .....

(26) The Journal of National Chengchi University, Vol. 69, 1994. ,r.. reported in Table f)' and Table 7. 23 Those with 'lot and hIgh shamability alike, external mformal shamIng has a significant influence upon juveniles either shamable or not shamable. The only exception is: to low shamable juveniles, external informal shaming has no effect upon the commission of non-predatory offenses, which is consisten with one of the hypothesis. However, external informal shaming has neverthelss an impact upon low shamabe juveniles in their commission of predatory crimes, which is not expected. In cases of high sha~le juveniles, reintegrating shaming can help reduce subsequent delinqency level, both predatory and non-predatory. Informal reintegrating shaming can reduce subsequ~nt overall and predatory delinquency level even in cases of low shamable juveniles . In contrast, external formal shaming has no effect upon juveniles' subsequent offenses except in high shamable juveniles cases where external formal shaming still can exert an effect on subsequent predatory offenses. In other words, high shamable juveniles are also responsive to external formal shaming. The more reintegrating the shaming is, the fewer the predatory offenses committed by high shamable juveniles. As to the relative importance of informal and formal shaming, the result supports our hypothesis, suggesting that informal shaming is more imrx>rtant than formal shaming. The magnitude of the regression coefficients of both variables shows that informal shaming has a greater impact upon predatory offenses than formal shaming. To our expectation, how significant others think of us, negatively or positively, leaves a bearing on our behavior. Also, as expected, external formal shaming has no influence upon those with low shamability. Given the variable shamability is derived from whether juveniles care about those insignificant other's opinion toward themselves. It is quite reasonable to see that external formal shaming can hardly influence low shamable juveniles, who, though not caring about insignificant others' opinion, are nevertheless sensitive to significant others opinions. The fact that external informal shaming has a significant effect upon low shamable juveniles seems to support the suspicion just mentioned above. As to those socio-demographic variables, readers can refer to Table 7 to see their independent effects. As mentioned before, the inclusion of those controls is !. ..... 23 . The reason to report unstandardized coefficients is that unstandardized regression techniques are not sensitive to changes in variances across populations, while correlational methods are. Thus, the former are generally preferred in panel analysis (Blalock, 1967).. 312.

(27) Informal Social Control of Crime. Modification of Labeling Theory. to determine to what extent and how the shaming variables affect juveniles' behaviors by controlling for those socio-demographic variables.. DISCUSSION AND CONCLUSION In this section, we will discuss which hypotheses are supported and which are rejected by the data. Later, we will reexamine labeling theory in view of the shaming model. We then will discuss the weakness in this data set and problems that hamper studies like the current one. Last, a brief summary is presented to conclude the current study.. 1. Hypothesis Review The first two hypothesis postulated that people with higher shamability will if shamed by either significant or insignificant others exhibit less deviance. These were supported by the data, especially in cases of predatory offenses. In other words, juveniles who care about how others think of them do reduce their delinquency level. Whether this effect will remain a longer time is another point in question that merits further study to determine exactly how shamability functions over time. To be noted, even formal shaming can exert an impact upon the commission of predatory crimes. Simply put, reintegrating shaming helps reduce predatory offenses, while stigmatizing push toward further delinquency. It is obvious that shaming from different sources, namely informal and formal sources, leads to different behavior outcomes depending upon the internal state of the juven~s (shamable or not) and the nature of external shaming (reintegrating or stigm£tizing). The study seems to suggest that a perception of greater reintegrating shaming leads to less deviance, especially serious offenses. This is supported in this study. We did find that the perception of reintegrating shaming from ones' significant others would lead to less deviance, both predatory and non-predatory. This conflicts with Braithwaite's shaming theory which argues for a shaming effect on pr~9atory crimes only. An important policy concern arises from the recognition o.!jthe significance of informal shaming, namely, what determines a perception of positive vs. negative shaming. Prior delinquent history? The experience .after the first official intervention? Or the cumulative effects of a growing record of public intervention? Answers need. f '. -313. .i.. ... ..

(28) The Journal of National Cliengchi University, Vol. 69, 1994. ... a detailed examimu£rr of the potential determinants of informal shaming, which is the subject for fu.re research. ,./ The third hypothesis holds that informal shaming can exert a greater effect upon behavior outcomes than formal shaming. To high shamable juveniles, informal shaming holds a greater influence upon subsequent delinquency. This finding lends quite a support to symbolic interactionism, suggesting the importance of significant other in an individual's lives. The fourth hypo~esis suggests no effect of either formal and informal shaming on low shamable jt.¥"~niles. The data basically support this position. Compared to high shamable juveniles, not only formal but also informal shaming is relatively helpless with the prevention of subequent offenses in low shamable juveniles. The only exception is informal shaming can have an effect upon low shamable juveniles, especially in their commission of predatory offenses. The reason that low shamable juveniles are still responsive to external informal shaming lies perhaps in the inadequacy of its empirical derivative. It is likely that low shamabe juveniles are relatively immune to external shaming from either informal or formal sources if the empirical indicator of shamability is adequate. Additionally, non-predatory offenses are generally not what collective sentiments coverge so that external shaming, especially shaming from formal sources usually fails.. 2. Review of Labeling Theory The shaming model present in this study refined some key constructs of labelling theory. We proposed a distinction between societal reactions into two types: reactions from the police and social service agencies (formal source) on the one hand, and reactions from one's significant others (informal source), on the other. Further, we proposed that the response varies even within each type of reaction. We introduced reintegrating shaming and stigmatizing shaming to characterize the nature of societal reactions. In addition, we introduced the concept of shamability, a very important theoretical linkage that might explain why people exposed to the same societal shaming behave differently. Overall, the shaming model places more emphasis on the perception of societal reaction than the societal reaction itself, since we think people's perception varies even though social facts may stay the same. It is the perception that is a better determinant of behavior. One major assertion of labeling theory is that societal labeling as a result of official contact is a cause of secondary deviance. The more the contact, the more -314.

(29) Informal Social Control of Crime -. Modification of Labeling Theory. serious the secondary deviance. The data in this study did not entirely refute nor support this proposition. The effect of societal labeling will take place under some conditions (Le. to those with a higher sense of shamability), and the effect is not always negative (if the labelling is reintegrating). The data show a significant effect of shaming on juveniles who are shamable. This finding helps fill a gap in labeling theory by pointing out shamability as the mechanism connecting labeling to subsequent behaviors. Further, external shaming does have an interaction effect with shamability as time progresses. Higher shamability and greater reintegrating shaming decrease delinquency over time, a finding consistent to Braithwaite's shaming model. Additionally, labeling theory does not distinguish different sources of labeling, implying labeling from both informal and formal sources have no different quality. These assumptions are not supported by the data. We find a different quality between significant others' reaction and official personnel's reaction. How juveniles perceive the reaction from their significant others is a significant determinant of their later behavior, both predatory and non-predatory. The more reintegrating shaming they perceive from thier significant others, the less deviant they will be. In comparison, the impact of formal shaming is quite limited, while that of informal shaming is much more influential and global.. " .... '. 3. Insufficiency in the Data Set for Testing Shaming Hypotheses ..... The original data set was not designed to test the shaming model. Therefore, we encountered some problems in testing our hypotheses. First, since }he perceptual process leading up to the eventual behavior manifestation involves ~rriod of time, it is methodologically ideal to conduct this line of research und~r a longitudinal framework. The data did include more than one interview to make it more longitudinal. However, given the time span between interviews, whether this time interval is good enough to trace the change is still an empirical question. Considering juveniles' delinquent pattern, the time interval between two interviews may be too long to have an accurate memory of how many times certain offen~es had been committed. This is especially so for non-predatory crimes. It is i~al to take a behavior log whenever offenses occur so that we can accurately trac~ th~ development of delinquency pattern. ...,l Also absent from the data set is the social-economic status juveniles' parents have. This variable is often postulated in criminological theory as one of important 315.

(30) '111""""~. The Journal of National 'Chengchi University, VoL 69, 1994. '". factors that speci,f{Juve~iles' life circu~stances (Cohen, 1955; G~ve. 1975; .Hewitt, 1970; Merton,. ,iJ68; Mlller, 1958; Relss, 1976). Even though hterature usmg self reports suggests' that delinquency is not just a lower-class phenonemon (Cohen and Short, 1971; Haney and Gold, 1973; Reid, 1976; Tittle and Villemez, 1978; Voss, 1966), it would have been useful to test it in the shaming model. We do assume that there is a status-specific shaming effect based on the reasoning that the consequence of shaming puts those in higher social standing in a much more disadvantageous po~ition. Lower class youth, in comparison, may have internalized conceptions of linied social-economic opportunities which further negative societal reaction can deGrease no less (Foster et at., 1972:209). Another weakness is that the perceived opinions of significant others is not measured. We only have this from the police and social service agencies. It is possible that juveniles do not care about these official personnel's opinions but do care about those of significant others. If this is true, then the data at hand provide insufficient information to let us determine whether the youth is really shamable or not. In other words, the measure of shamability used in this study reflects a partial degree of shamability. The last weakness is the absence of the length of delinquent history (i.e. years of delinquent life). A veteran delinquent should respond to shaming differently from a relatively new delinquent. Prior history that is controlled in the model refers only to offenses committed just prior to the current interview since the last interview. This is more a reflection of the current delinquent level than that of prior delinquent history. Research shows that age at first arrest or age at the onset of the antisocial behavior is an efficient predictor of later delinquency (Glueck, and Glueck, 1960; Mandelzys, 1979; Shannon, 1978; Wolfgang, et. aI., 1972). Juveniles with a long delinquent history are very likely to perceive shaming more negatively, or even be resistant to shaming. In other words, shaming will have a differential effect on juveniles with various lengths of delinquent life. From a policy point of view, strategies in dealing with juveniles with different deviant histories should be different.. 4. Conclusion The current study supports some aspects of Braithwaite's shaming hypothesis, namely the significance of informal shaming; What he predicted is that shaming is more deterring when administered by those who are significant to us. However, 316.

(31) ~. Informal Social Control of Crime -. Modification of Labeling Theory. it is a twcredged sword. A perception of shaming that is reintegrating can restore the once-wayward back to society. However, once stigmatizing shaming aoo rejection is perceived, shaming may push one further toward more delinquency. What is suggested here is that our behavior depends a lot upon our social images in the eyes of our signifi~ant others. Social approval from them is constantly needed for positive reinforcement of pro-social behaviors. This finding helps fIll a gap in labeling theory, namely, when labeling leads to more deviance vs. less deviance. One important implication arising from this study is that when people care about their significant others' opinion, that care is the prerequisite to their "reform". This study, however, does not fully support Braithwaite's shaming hypotheses about predatory crimes. To high shamable juveniles, informal shaming works not only for predatory offenses, but also for non-predatory offenses. But to low shamable juveniles, this study seems to suggest that internal shaming state, along with other social forces, is a greater control mechanism. Societal reaction from more agency personnel is not found to be an important variable in juveniles' behavior. Labeling theory attributes the development of secondary deviance to societal labeling. This study modifies this proposition by suggesting that development of secondary deviance will take place only when juveniles are shamable and perceive greater reintegrating shaming. Nevertheless, this effect is limited to total and predatory offenses, but not non-predatory offenses. One important conclusion is that labeling theory only tells half of the reality. Labeling will not necessarily incur negative consequences. It depends upon the nature of labeling itself and the situation in which labeling is administered. Further, the shaming model helps modify the labeling concept by specifically sP<711ing out when and how one kind of consequences vs. others will take place. 9verall, informal reintegrating shaming generates positive behavioral consequen~ while informal stigmatizing shaming generates negative consequences, given that juveniles are shamable. Formal shaming by itself has no effect, but when joined with shamability, it can influence certain behavior. A second conclusion is that, as an important theoretical addition to the labeling theory, shamability was found to be a significant mediator linking e~ternal societal shaming to specific behavior outcomes. However, its effect eme~es only when internal shame state of individuals exists. A third conclusion is that external informal shaming works 'for 'both predatory , and non-predatory offenses, contrary to one of Braithwaite's shaming propositions, But consistent to Braithwaite's prediction, external 'informal shaming is more important 'f. -317. ~. ..,. ....

(32) The Journal of National Chengchi University, Vol. 69, 1994. than formal shamin¥' with regards to their effect upon subsequent offenses. Last but nOLXeast, informal social control has an influence upon juveniles' behavior, as also predicted by Braithwaite. More reintegrating shaming from significant others helps reduce delinquency. This shaming effect works for both predatory and non-predatory offenses. In addition, the shaming effect is not a constant I. ... phenomenon. It has a duration. This study is not able to determine when external shaming effect is at its height and when it will decline. The external shaming effect by itself was found tr,take place in the early stage of delinquecy, but its continued effect lies in the level of internal shaming. Future research needs to be done to I explore its exact pattern for greater policy utility. In brief, the current study points to the importance of informal social mechanisms in crime control. The key to effective crime control lies not only with judicial and social service organizations, but more importantly, with the informal social organizations, such as family,. school, neighborhood, church, etc. This finding, on the one hand, supports previous cross-cultural studies, such as Adler's, noting a greater effectivenes of crime control by informal social organizations. On the other, it explains why the past focus on restructuring judicial or formal social organizations to control crime failed the task.. REFERENCES Ageton, Suzanne, and Elliott, Delbert S. 1974. "The Effects of Legal Processing on Self Concept." Social Problems. 22(1): 87-100. Ageton, Suzanne and Elliott, Delbert S. 1978. The Incidence of Delinquent Behavior in a National Probability Sample of Adolescents. Boulder, Colorado: Behavioral Research Institute. Adler, Freda. 1983. Nations Not Obsessed With Crime. Littleton, Colorado: F.B. Rothman. Aultman, Madeline G. 1979. "Delinquency Causation: A Typological Comparison of Path Models." Journal of Criminal Law and Criminology. 20(2): 152-163. Ausubel, David P. 1955. "Relationships Between Shame and Guilt in the Socializing Process." Psychological Review. 62:378-390. Bayley, David H. 1976. Forces of Order: Police Behavior in Japan and the United States. Berkeley: University of California Press. Bayley, David H. 1985. Social Control and Political Change. Research Monograph. No.49. Woodrow Wilson School of Public and International Affairs, Princeton University. Becker, Howard S. 1963. Outsiders: Studies in the Sociology of Deviance, NY: Free Press. Bernstein, Ilene N., Kelly, William and Doyle, Patricia A. 1977. "Societal Reaction to Deviants: The Case of Criminal Defendants." American Sociological Review. 42(5):743-755. 318.