Aug 21, 2006 - Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, .... 1987; Sellers and Winfree 1990), have a natural link to gang re-.
This article was downloaded by: [Arizona State University] On: 12 February 2014, At: 10:59 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Justice Quarterly Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rjqy20
Youth gangs and incarcerated delinquents: Exploring the ties between gang membership, delinquency, and social learning theory a
a
L. Thomas Winfree Jr. , G. Larry Mays & Teresa VigilBäckström
a
a
New Mexico State University Published online: 21 Aug 2006.
To cite this article: L. Thomas Winfree Jr. , G. Larry Mays & Teresa Vigil-Bäckström (1994) Youth gangs and incarcerated delinquents: Exploring the ties between gang membership, delinquency, and social learning theory, Justice Quarterly, 11:2, 229-256, DOI: 10.1080/07418829400092241 To link to this article: http://dx.doi.org/10.1080/07418829400092241
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden.
Downloaded by [Arizona State University] at 10:59 12 February 2014
Terms & Conditions of access and use can be found at http://www.tandfonline.com/ page/terms-and-conditions
Downloaded by [Arizona State University] at 10:59 12 February 2014
YOUTH GANGS AND INCARCERATED DELINQUENTS: EXPLORING THE TIES BETWEEN GANG MEMBERSHIP, DELINQUENCY, AND SOCIAL LEARNING THEORY*
L. THOMAS WINFREE JR. G. LARRY MAYS TERESA VIGIL-B2[CKSTROM New Mexico State University
This paper explores the conceptual and empirical ties between membership in youth gangs, youthful misconduct, and Akers's social learning theory. The adolescent population studied, however, is unique: it consists of adjudicated and incarcerated delinquents. The data for the study were obtained by conducting a census of all youths in the custody of the New Mexico Youth Authority in January 1991; more than 85 percent of all youths incarcerated on the administration day (258 young men and women) participated in the study. We found that gang members had acquired more pro-gang attitudes than nongang youths and were more favorably inclined toward gang activities; neither gang membership nor the gang-based social learning theory variables, however, were related uniformly to all forms of selfreported delinquency. We address the theoretical and policy implications of these findings.
Research interest in gangs and gang behavior has revived throughout the 1980s and into the early 1990s (see Curry and Spergel 1988, 1993; Goldstein 1991; Moore and Vigil 1989; Spergel 1990). Various theoretical and methodological approaches have been directed toward what is called a "growing urban phenomenon" (Fagan 1989, 1990; Spergel 1990). One theoretical perspective that has gained increasing empirical support in gang research over the past decade is Akers's variant of social learning. 1 For example, a * An earlier version of this paper was presented at the 1992 meetings of the American Society of Criminology, held in New Orleans. Send all correspondence to first author c/oDepartment of Criminal Justice, Department 3487, New Mexico State University, Las Cruces, NM 88003-001. 1 The phrase social learning has been employed by both sociologists and psychologists to describe a process whereby individuals acquire certain behavioral and attitudinal responses to their social, psychological, and physical environments. Psychologists, especially behavioralists, associate the term with researchers such as Skinner (1953) and Bandura (1977); social behavioralists are careful to distinguish JUSTICE QUARTERLY, Vok 11 No. 2, June 1994 © 1994 Academy of Criminal Justice Sciences
Downloaded by [Arizona State University] at 10:59 12 February 2014
230
YOUTH GANGS AND INCARCERATED DELINQUENTS
decade ago Morash (1983:313) hypothesized, following Akers's theory, that "peers' delinquency is itself predictive of individuals' delinquency, regardless of whether the peer group is like a gang and irrespective of the community context or youth's sex." Her research supported this hypothesis. Subsequent generations of researchers have explored the ties between delinquent peers, gangs, and gang delinquency. Their work is governed largely by a widely held belief that the spread of gangs and the increase in urban violence are closely linked social phenomena (Curry and Spergel 1993; Huff 1990; Miller 1990; Spergel 1990). Although researchers who have addressed the links between gangs and delinquency have produced somewhat ambiguous findings, they tend to support the primacy of delinquent peers for understanding both phenomena (see Curry and Spergel 1993; Thornberry et al. 1993). Gang studies in general, however, even those which reference Akers's social learning, rarely include more t h a n the single variable "gang member friends" (see Curry and Spergel 1993; Morash 1993; Thornberry et al. 1993). The present research employs a far more complete social learning theoretical framework. We examine gang involvement in a group of adjudicated delinquents to determine how far it is possible to distinguish between gang and nongang delinquent youths and to predict the level of self-reported offense-specific offending. Two research questions guide our application of social learning theory to gang behavior: 1) to what extent does social learning theory allow us to differentiate between gang and nongang members? 2) To what extent do social learning theory and gang involvem e n t allow us to predict the breadth of self-reported delinquent behavior? SOCIAL LEARNING, DELINQUENCY, AND GANGS Even atheoretical gang researchers discuss the "costs" and '%enefits" associated with gang membership, and not in a strictly econometric sense. That is, membership in a social group such as an urban street gang entails costs as well as rewards: one must go their work from that of social behaviorists in the tradition of G.H. Mead (see Baldwin 1985; Michaels and Green 1978). The applicationof behaviorist concepts to sociological issues also has a long tradition (Burgess and Bushell 1969). A m o n g criminologists, the term social learning has been attributed variously to Jeffery's (1965) differentialreinforcement theory, Burgess and Akers's (1966) differentialassociation/reinforcement theory, and Adams's (1973) differentialassociation/learning theory (also see Halbasch 1979; Jeffery 1980). In the past two decades, however, social learning theory, as applied to questions of crime and delinquency, has come increasingly to mean Akers's social learning theory (Akers 1985, 1993); we employ that denotation here.
Downloaded by [Arizona State University] at 10:59 12 February 2014
WINFREE, MAYS& VIGIL-BACKSTROM
231
through an initiation, give up nongang friends, and be intensely loyal to the "hood" or "el barrio" (see Curry and Spergel 1988; Spergel 1990; Torres 1980; Vigil 1988). These gang-based systems of reinforcements and punishments exhibit close similarities to the basic learning process. For example, instrumental conditioning is an important element in social learning theory. Behavior is "acquired or conditioned by the effects, outcomes, or consequences it has on the person's environment" (Akers 1985:43; author's emphasis). Reinforcements and punishments are the primary processes by which instrumental conditioning is achieved. Behavior is said to be reinforced when repeated episodes are met with a response by others that influences the actor to engage in the behavior again under similar circumstances; as a result, the behavior is repeated (Akers 1985:44,45). Behavior can be said to be punished when the response of others is such that the actor is discouraged from engaging in the conduct again under similar circumstances; as a result, the behavior decreases (Akers 1985:44,45). In addition, gang members provide one another with definitions of reality that are not simply shared and expected; these definitions may be mandated and may serve as a source of ingroup cohesiveness (Klein 1968; Klein and Crawford 1968; Miller 1975; Moore and Vigil 1989; Spergel 1990; but also see Campbell 1987). Akers, like Sutherland, believes that definitions are "normative meanings which are given to behavior; that is, they define an action as right or not right" (1985:49). Akers differs from Sutherland, however, in viewing these definitions as "verbal behavior (overt and subvocal), which can be reinforced" (49). Akers (p.49) also views such definitions as belonging to a class he calls discriminative stimu l i . Discriminative stimuli are themselves a source of reinforcements; they are differentially reinforcing. In the context of a youth gang, we view differential definitions as a construct which implies that an individual learns, through close, intimate interactions with others, evaluations of orientations and behavior as appropriate or inappropriate. The instant content of the learning process, then, consists of gang-related orientations and behavior. Gang membership and involvement are more likely to ensue when young people develop, through a process of reinforcements and punishments, definitions favorable rather than unfavorable to gang activity. Differential associations, perhaps the single most important explanatory variable in either Akers's or Sutherland's theories (see Akers et al. 1979; Krohn, Lanza-Kaduce, and Akers 1984; Orcutt 1987; Sellers and Winfree 1990), have a natural link to gang research. Gangs typically do not encourage members to have friends, especially close friends, outside the gang (see Horowitz 1983; Miller
Downloaded by [Arizona State University] at 10:59 12 February 2014
232
YOUTH GANGS AND INCARCERATED DELINQUENTS
1975; Spergel 1990:233-36). We believe that the ties between differential associations and both gang membership and youthful misconduct should be as strong as those observed in the case of illicit drugs, especially when the associations are described in a behaviorspecific context (see Akers et al. 1979; Orcutt 1987; Winfree and Griffiths 1983). 2 These individuals are potentially the source of both criminal and noncriminal definitions. Like both Sutherland and Akers, we believe that the greater the proportion of an individuals best friends are gang members, the greater the social forces impelling that individual to become a member of that or a similar group and to engage further in conduct endorsed by the group. Our research is guided by three related hypotheses, all of which reflect Akers's inclusion of Sutherland's differential association theory under the conceptual umbrella of social learning theory. Specifically, Akers's theory suggests the first hypothesis: The degree to which a person is exposed to gang-involved peers, pro-gang values from peers and significant other adults and reinforcers of pro-gang behavior and attitudes is related to personal pro-gang definitions. The second hypothesis is also derived from social learning theory: The presence of personal pro-gang attitudes--and, to a lesser extent, gang-related differential associations and differential reinforcers--is related to identification with youth gangs. The third and final hypothesis is also derived from Akers, but owes a debt to Morash (1983) as well. Morash observed that theories such as social learning emphasize the causal effect of peers' delinquency on individuals' delinquency and place such variables as gang membership and even gender in secondary roles. Most important is the presence of differential associations and differential definitions. Hence the third hypothesis: Differential associations and definitions predict the individual's delinquency, regardless of gang membership, age, gender, and ethnicity. METHODS
Our study represents an additional test of the empirical ties between social learning theory and both gang membership and youthful misconduct. The population under study consists of 2 It would be a misinterpretation of Sutherland, however, to assume that these ties simply reflect a restatement of the old adage '~irds of a feather flock together." Rather, as Cressey observed, "the theory of differential association is concerned with ratios of associations with patterns of behavior, no matter what the character of the person presenting them" (Sutherland and Cressey 1974:79; authors' emphasis).
Downloaded by [Arizona State University] at 10:59 12 February 2014
WINFREE, MAYS & VIGIL-BACKSTROM
233
youths who were adjudicated delinquent by the State of New Mexico and were incarcerated for their offenses. We selected this population for two reasons. First, the administrators of the correctional facilities stated that the proportion of gang members incarcerated approached nearly 50 percent. Thus we were reasonably certain that we would have variation in a critical variable of the study, namely gang membership. Second, we believe that incarcerated delinquents provide a unique population in which to test social learning theory. A cynic might argue that all such youths possess prodelinquency definitions and delinquent associations; this is precisely the point. We expect the theory to distinguish between gang and nongang youths and to predict the breadth of crime-specific offending when everyone surveyed is an adjudicated delinquent. The population under study, however, has some limitations. First, minority-group representation, although deep, is not broad. That is, the institutions house urban and rural Hispanic youths, but other minority groups, especially African-Americans and AsianAmericans, are noticeably underrepresented. Insofar as these groups exhibit unique patterns of gang behavior and criminality, the generalizability of the study is limited. Second, middle-class, suburban gang members are unlikely to appear in an institutional population (Curry and Spergel 1988; Huff 1993). Moreover, few contemporary studies of youth gangs examine these important subgroups; the present study does not fill this lacuna (see, however, Takata and Zevitz 1990).
The Research Setting and Administration In early 1991 we approached the senior administrators of the New Mexico Youth Authority with the idea of studying the extent and nature of youth gang membership in the primary juvenile institutions under their control. One of these institutions, a facility for male offenders, is in a small town in northern New Mexico; the institution has no walls or fences. Approximately 200 boys, age 12 to 19, live in nine one-room cottages. Cottage residency is determined by the youth's age and the behavioral risk he might pose. The second institution, the Youth Development and Diagnostic Center (YDDC), is located in Albuquerque, the state's largest city, and contains two elements. The first, consisting of five cottages, is the reception and diagnostic center for all male and female admissions to state custody; most of the 100 or so inmates are sentenced to state custody for 90-day diagnostic screening (see Lozano, Mays, and Winfree 1990). The second element is a single 20-inmate cottage designed as a school for female offenders. As in the boys' school, the
234
YOUTHGANGSAND INCARCERATEDDELINQUENTS
Downloaded by [Arizona State University] at 10:59 12 February 2014
youths range in age from 12 to 19. The administrators of both facilities cooperated completely with the researchers and placed no restrictions on the project other than asking to see the questionnaire in advance. Given the relatively small size of the student population at each institution, we decided to conduct a census of each facility's resident population. This effort was quite successful. At the boys' school, one member of the research team met alone with the boys during the evening hours, although a staff member escorted the researcher to and from the cottages. The boys were administered the questionnaires in groups of 20 to 25. The general purpose of the study was explained to them orally at that time ("We want to learn about youth gangs in New Mexico"), and the confidentiality of their responses was assured. All of the participants received the opportunity to withdraw without prejudice at that time. (We explained that they would simply stay in the room until all the others were finished so that no one would know who did and did not participate.) If a youth could not read the questionnaire, it was read to him and he recorded his own answer. The boys were instructed to place their completed questionnaires in a locked box at the front of the room. Five youths refused to participate; the responses of another 13 were dropped from the analysis because of missing or inconsistent answers, most of which (we suspect) were due to problems with reading English or generally low reading levels. The remaining 161 respondents represented nearly 90 percent of the institutional population in residence during the three-day administration period. Only one school day was disrupted to allow students to participate in the survey. Classes were dismissed, and the residents of each cottage returned to the common living area for one period. The administrative procedure that we used at the boys' school was followed as well as YDDC and at the girls' school. Approximately 30 youths, almost all residents of YDDC, refused to participate in the survey. We obtained completed questionnaires from 99 male and female inmates; two, however, were dropped from the analysis because of missing data. The remaining 97 questionnaires obtained from this facility represented 80 percent of the youths in residence on the administration day.
Variables Delinquency. We employed a self-report delinquency (SRD) inventory. Each respondent was asked to indicate the level of his or her involvement in 22 different activities, the specific content of which
Downloaded by [Arizona State University] at 10:59 12 February 2014
WINFREE, MAYS & VIGIL-B/kCKSTROM
235
was based on earlier research inventories (Gold 1970; Hindelang, Hirschi, and Weis 1981; Osgood et al. 1989). We encountered two problems with the SRD measures. First, the subjects, by definition, were "involved delinquents." Therefore we were not surprised by the high offending rates they reported. Another factor, however, made us question the specific rates. The SRD had a problematic anchoring point: the youths were asked to indicate the number of times they had committed each offense since leaving middle school. This could have led to inaccurate reporting. We resolved both problems with a two-step method for operationalizing the SRD measures. First we assigned a value of 1 to each offense for which the self-reported offending rate exceeded "once or twice;" rates at or below that benchmark received a value of 0. The use of this dichotomization method has inherent weaknesses, particularly because it allows us to review only the breadth, and not the depth, of the misconduct (see Hindelang et al. 1981). The anchoring-point problem, however, made this operationalization necessary. In addition, by assigning a value of 1 to self-reported involvement that exceeded the "once or twice" category, we focused on serious involvement; moreover, separate analyses performed on the less reliable offense reporting rates revealed substantively identical findings. Next we elected to examine offense-specific indices (also see Fagan 1989; Osgood et al. 1989). We also excluded relatively minor rule-violating acts, including status offenses.~ Another three offenses, all reflecting personal crimes, failed to create a reliable selfreport index; they too were excluded from the analysis. 4 We grouped the remaining 12 offenses into four composite indices: 1) theft crimes ("taken things worth less than $2," "taken things worth between $2 and $50," "taken things worth over $50"); 2) drug offenses ("bought or drank beer, wine or liquor," "used illegal drugs like marijuana or cocaine," "sold illegal drugs like marijuana or cocaine"); 3)joy-riding and property destruction crimes ("borrowed someone's car without his or her permission," "destroyed private property," "destroyed public property"); and 4) group-context personal crimes ("taken part in a fight involving more t h a n two people where only fists were used," "taken part in a fight involving more s Each student was asked how often he or she had engaged in each of the following behaviors since middle school: "driven a car without a driver's license or permit," %kipped school without a legitimate excuse," ~run away from home at least overnight," ~driven too fast or recklessly in a car or on a motorcycle," %pread an u n t r u e r u m o r about someone to get even," ~been placed on school probation or expelled from school," and %pray-painted a wall or building." 4 Once again, each respondent ,was asked how often he or she had "had a fist fight with one other person," ~ 'beat up' on kids who hadn't done anything to you," and, "hurt or inflicted pain on someone else to see him or her squirm."
236
YOUTH GANGS AND INCARCERATED DELINQUENTS
than two people where weapons other than fists were used," "shot at someone because you were told to by someone else"). We summed the scores for each crime index. In addition, we summed all 12 SRD responses to create an index of overall criminality. Cronbach's alpha provided a measure of scale reliability: theft crime (.79), drug-related crime (.7), property crime (.77), group-context crime (.67), and general criminality (.86). S u m m a r y statistics for each SRD offense i n d e x - - a n d the rest of the v a r i a b l e s - - a r e reported in Table 1.
Downloaded by [Arizona State University] at 10:59 12 February 2014
Table 1. Key Variables Variable
Attributes
Percentage (N) Mean (SD)
Gender
Male Female Hispanic Non-Hispanic Measured in years 1) None or nearly none 2) Less than half 3) More than half 4) All or nearly all 1) Strongly approve 2) Approve 3) Depends 4) Disapprove 5) Strongly disapprove 1) Strongly approve 2) Approve 3) Depends 4) Disapprove 5) Strongly disapprove 1) Positive reaction 2) Neutral reaction 3) Negative reaction 1) Positive reaction 2) Neutral reaction 3) Negative reaction Scale score
85.7% 14.3% 57.0% 43.0% 17,03 24.4% 15,5% 15.1% 45.0% 21,3% 16.3% 42.2% 12.0% 8.1% 8.1% 3.1% 24,0% 24.8% 39,9% 28,7% 52.3% 19.0% 5.4 48.4% 46.1% -0.65
(221) (37) (147) (101) (1.33) (63) (40) (39) (116) (55) (42) (109) (31) (21) (21) (8) (62) (64) (103) (74) (135) (49) (14) (125) (119) (1.25)
8.11
(2.49)
54.3% 45,7% 1,76 1.61 2.16 1.61 7.13
(140) (118) (1.24) (1.23) (1.05) (1.07) (3.59)
Ethnicity
Age Differential Associations: Peer Gang Members
Differential Association: Peers' Disapproval
Differential Associations: Significant Adult Disapproval
Differential Reinforcers: Peers' Reactions Differential Reinforcers: Parents' Reactions Differential Reinforcers: Net Effects Differential Definitions: Pro-Gang Attitudes Gang Membership Theft Crimes Joyriding/Destruction Crimes Drug-Related Crimes Group-Context Crimes General Criminality
Scale score Nongang member Gang member Index score Index score Index score Index score Index score
Downloaded by [Arizona State University] at 10:59 12 February 2014
WINFREE, MAYS & VIGIL-BACKSTROM
237
Gang membership. We employed a combination of the self-definitional and the criterion methods to determine gang membership (see Klein 1971; Miller 1990; Short 1989). We selected these items on the basis of an ethnographic study of youth gangs in New Mexico (see Mays, Winfree, and Jackson 1993). The instrument contained a series of questions related to respondent's involvement with gangs. Having determined that the youth not only had been a gang m e m b e r at some point b u t currently considered himself or herself a gang m e m b e r although incarcerated, we next employed three additional criteria. First, the group had to have a name and at least one of the following cultural elements: 1) an initiation rite, 2) a specific leader or leaders, or 3) gang nicknames for members. Second, the gang had to employ at least one of the following symbols: 1) "colors," 2) tattoo(s), 3) hand signs, or 4)jewelry. Finally, we asked the youths to rank the gang's five most important activities. To qualify the group as a youth gang, at least one illicit activity (sex, drugs, or alcohol use) or one illegal activity (fighting, committing crimes, or vandalism) had to appear on this list. By this restrictive process, we identified 140 gang members, or 54.3 percent of the youths surveyed. Differential associations. Following Akers's model, this construct consists of two dimensions. The first is norm qualities, or the values of one's significant others and parents as perceived by the youth. We measured parents' (and guardians') norm qualities with a single question: "What is the attitude toward gangs of most of the adults whose opinions you value or think are important?" The respondents were asked to select one of the following responses: 1) strongly approve, 2) approve, 3) dependents on the circumstances, 4) disapprove, and 5) strongly disapprove. The smaller the score reported, the greater the disapproval. The identical question was asked about "the teenagers whose opinions you value or think are important." Generally the youths surveyed reported higher levels of disapproval among their peers than among significant other adults. The second and far more common indicator of differential association is the proportion of one's peers who belong to the target population, in this instance youth gangs. Specifically, we asked: "About how m a n y of your best friends are gang members?" Thus the question included Sutherland's "intensity" element from differential association theory. We provided the following response categories: 1 none or almost none, 2) less than half, 3) more than half b u t not all, and 4) all or almost all.
238
YOUTHGANGSAND INCARCERATEDDELINQUENTS
Downloaded by [Arizona State University] at 10:59 12 February 2014
Differential reinforcements. Social reinforcers constitute a major part of Akers's variant of social learning theory. As stated earlier, according to Akers, reinforcement is a process whereby "sometimes our behavior is met by reactions from others (or has some other consequence attached to it) which influence us to do the same again under similar circumstance" (1985:44). Akers further differentiates between social and nonsocial reinforcers: the latter are rewarded at an unconditioned or psychological level, whereas the former require other human beings to aid in their interpretation as desired states or outcomes. Similarly, Akers's social learning theory emphasizes the role of punishers as extinguishers of behavior. As with reinforcers, "punishments may be brought about by either the addition or the subtraction of stimulus events" (Akers 1985:44). Akers also recognized that punishers may be social or nonsocial, or even may combine social with nonsocial differential reinforcements (Akers 1985, 1993; Akers et al. 1979). We employed two methods to measure reinforcers. In the first method, we measured probable rewarding and punishing reactions from friends and family members faced with the idea that the respondent was a gang member. We asked the following question: "What would your parents or guardians most likely do if they thought you were a member of a gang?" We included five responses, ranging from encouragement to kicking the respondent out of the house and turning him or her over to the police; we also allowed the respondents to report other possible reactions. Most of these responses were arrayed on a continuum from positive through neutral to negative. Few youths (5.4%) believed that their parents were likely to respond positively; most (48.4%) reported that their parents would be neutral; the remainder (46.1%) stated that in all likelihood, their parents would receive the information negatively. When the youths were asked about "most of their friends," the pattern was somewhat different: most of the respondents (52.3%) believed that their friends would take a neutral position, but most of the rest (28.7%) replied that friends would be supportive. Only one in five reported the possibility of negative peer responses. Second, following Akers's interest in combined social and nonsocial reinforcers, we asked each respondent the following question: "Whether or not you are a member of a gang, what good things do you think would happen to you as a member of a gang?" This statement was followed by a checklist of possible answers. In this way we allowed each respondent to define, in his or her estimation, the good things associated with gang membership. Five items were included on the list of potentially good things: 1) "feel successful" (positive social reinforcer), 2) be "cool" (positive social reinforcer),
Downloaded by [Arizona State University] at 10:59 12 February 2014
WINFREE, MAYS & VIGIL-BACKSTROM
239
3) be more like someone else (positive social reinforcer), 4) excitement (positive nonsocial reinforcer), and 5) money (positive nonsocial reinforcer). Respondents also were asked to indicate which of a list of bad things they believed could happen to them as gang members. This list included 1) "feel guilty" (negative social/nonsocial punisher), 2) "get in trouble with the police" (negative social punisher), 3) "get in trouble with my parents" (negative social punisher), 4) "get hurt" (negative nonsocial punisher), and 5) "lose friends" (negative social punisher). We then assigned a value of I to all reinforcers and a value o f 1 to all punishers, and summed across the 10 items. The summed scale was reasonably reliable (alpha=.50). A negative summed response (-1 to -5) was taken to reflect a negative net effect of the gang-specific social/nonsocial reinforcers and punishers; a positive summated score (1 to 4) indicated a positive net effect. Nearly onehalf of the sample scored 0; that is, the positives and the negatives balanced. The distribution of scores, however, also was negatively skewed (mean = -.65; sd = 1.25).
Differential definitions. The expression of pro-deviance attitudes has been used in a number of studies to measure differential definitions (see Akers et al. 1979; Sellers and Winfree 1990; Winfree, Sellers, and Clason 1993). This measure was borrowed directly from Sutherland's differential association theory; it is conceived to be a product of the process whereby the individual, through interactions with others, learns evaluations of behavior as good or bad. As Akers et al. observe, the definitions are "themselves verbal and cognitive behavior which can be directly reinforced and also act as cue (discriminative) stimuli for other behavior" (1979:638). Deviant behavior is more likely to result when individuals develop definitions that are favorable rather than unfavorable to the behavior. The pro-deviance differential definition measure used in this study was grounded in "gang experiences." The subjects were asked whether they 1) disapproved, 2) neither disapproved nor approved, or 3) approved of the following: 1) having friends in gangs, 2) being in a gang yourself, 3) taking part in illegal gang activities like fights, 4) doing whatever the gang leaders tell you to do. These items formed a reliable measure of personal gang attitudes (alpha = .83). The higher the score, the more pro-gang the attitude.
Personal-biographical characteristics.
Three personal-biographical characteristics are of theoretical and substantive interest to the current study. First, gang behavior, especially gang-related crime, is primarily but not exclusively a male phenomenon (see Bowker, Gross, and Klein 1980; Campbell 1987; Spergel 1986, 1990). We
240
YOUTHGANGSAND INCARCERATEDDELINQUENTS
acknowledge the importance of gender in this differential status and involvement by examining the responses of residents of the girls' school. For this study we coded gender as male (1) or female (0). The sample was largely male (85.7%).
Downloaded by [Arizona State University] at 10:59 12 February 2014
Second, age is significant, if only as a control variable. Older youths typically report far more extensive crime histories and a broader range of self-reported delinquencies (Spergel 1986, 1990). We treated age as an interval variable. The average age was 17 years, with a standard deviation of 1.3 years. Finally, the gang phenomenon has long been associated with minority-group membership; in the southwest, this is the HispanicAmerican culture (Spergel 1990:212; Vigil 1988). Critics of this viewpoint, however, point out that nothing endemic to the Chicano barrio supports the view that it is more gang- or violence-prone than other subcultural communities (Erlanger 1979). Vigil (1988:7), for example, maintains that although Chicano gangs are notable for their prevalence and persistence, only about 4 to 10 percent of the barrio youths--the population perhaps at the greatest r i s k m a r e associated with gangs. Most studies of youth gangs, however, focus on the minority status of the gang membership (see Curry and Spergel 1993; Fagan 1989; Fagan, Piper, and Moore 1986; Miller 1975, 1982; Spergel 1990). Given the relatively small number of non-Hispanic minority-group members in state custody, we coded ethnicity as Hispanic (1) and non-Hispanic (0); Hispanics accounted for 57 percent of the study population.
Analytic Techniques As a result of the operationalizations of the variables, we had to employ two different analytic techniques--ordinary least squares (OLS) regression and logistic regression. First, two variables--attitudes toward gangs and level of criminal involvement--are at least ordinal measures. Thus, for these variables, we could employ OLS analysis. We were interested in the extent to which the youths' progang attitudes could be understood in terms of variables derived from social learning, while controlling for ethnicity, age, and gender. The analysis of SRD added differential definitions and selfreported gang membership: we were interested in the extent to which these variables were linked empirically to the breadth of serious self-reported delinquency (i.e., self-reported rates of more than once or twice for each of the offenses in a particular offense index). Using OLS regression, we could compare the contributions of each variable in relation to the others for offense type by examining the standardized coefficient. Also--a consideration important
WINFREE, MAYS & VIGIL-B.~CKSTROM
241
Downloaded by [Arizona State University] at 10:59 12 February 2014
for the analysis of SRD--we could compare the effects of each independent variable across dependent variables by examining the unstandardized coefficients. Finally, OLS analysis provided a standardized coefficient (coefficient of determination) for each equation. This coefficient is understood as the amount of variance in the dependent variable explained by knowledge of the combined direct effects of the independent variables. Second, we were interested in whether the youths had ever been involved in a youth gang. We wish to demonstrate that we can predict group membership if we possess certain information about all members of our study. In theory, the only relevant predictor factors should be those related to social learning theory. Factors such as gender, race, and age should not play a major role in this classification process. For this purpose we employed the SAS logistic regression analysis (Harrell 1986). Logistic regression (Cox 1970) is a nonlinear regression procedure that requires fewer assumptions than linear classification methods such as classification analysis. 5 Because the logistic model makes no assumptions about the normality of the classification variables or about the quality of the dispersion matrices for the several populations, it is often preferred to discriminant analysis (Cox 1970; Press and Wilson 1978). Model parameters are interpreted similarly to partial regression parameters. 6 The classification models presented here were estimated by the method of maximum likelihood using the SAS supplementary library procedure LOGIST (Harrell 1986). We classified cases as belonging to Group 1 (gang members) whenever the inverse logit7 exceeded 0.543, the proportion of the sample who were gang members. We also analyzed the data using the linear discriminant analysis, with qualitatively similar results. Somers' d, a product of
5 The analysis proceeded with Yi as the dependent variable (group indicator) for the ith observation, and X = (Xil, Xi2, • •., X'in) as the vector of the independent variables for the ith observation. In binary logistic regression the model used is P[Yi = 1 = 1/(1 - exp (X)i'B)] where P (Yi = 1) is the probability t h a t the ith observation belongs to the group i, and B is the vector of unknown parameters t h a t may or may not include the intercept term. If P (Yi = 1) is abbreviated as Pl, straightforward manipulation shows logit [pl = log~[p~/(1-pi)] = X~B. 6 I f we assume t h a t all other variables remain constant, a change of one unit in Xik results in an increase Ofbk units in the logit for case i. The nonlinear nature of the logit transformation must remembered, however. A change of one unit from (say) 1 to 2 is not equivalent on the probability scale to a change of one unit from 20 to 21. 7 The term logit was defined originally as a particular metametric transformation in quantal bioassays: logit (n) = logc(~]l-~). The term is used in this sense throughout this paper. The logit is a statistical construct, created from the various independent variables as a m e a n s of classifying subjects on the dependent variable.
Downloaded by [Arizona State University] at 10:59 12 February 2014
242
YOUTH GANGS AND INCARCERATED DELINQUENTS
PROC logistic, is a measure of association that generalizes Goodman and Kruskal's (1979) gamma, s Finally, we inspected for collinearity for both the logistic and the ordinary least squares regression techniques. Both techniques are sensitive to collinearities among the independent variables of the models. Hosmer and Lemeshow (1989) suggest that although collinearity analysis should be helpful in identifying the dependencies among the covariates, one normally would not employ this for logistic analysis unless the estimated coefficients and the standard errors provided evidence of degradation in fit. We found no such evidence in the present analyses. In the case of OLS analyses, we also saw no evidence of collinearity in the intercorrelation matrix (see appendix) or the adjusted R 2 values.
FINDINGS
Differential Associations, Differential Reinforcers, and Gang Definitions In the present study, personal definitions were grounded in the "gang experience." That is, the pro-gang definitions scale addressed the respondent's perceptions of and feelings toward gangs, and was coded so that ascending scores indicated increasing levels of support for gangs in general and gang-related activities in particular. Table 2 contains the hierarchical model for gang definitions. The nine independent variables can explain 45 percent of the variance in gang definitions. The most important social learning variable is the proportion of one's best friends who are members of youth gangs, or peer gang members (beta = .41) Two other differential association variables--significant peers' a~d adults' levels of gang approval--also exhibit a significant tie to differential definitions: youths who perceive that their peers and adults whose opinions they value hold pro-youth gang attitudes are themselves likely to hold such attitudes (betas = .19 and .13 respectively). Two of three social differential reinforcements exhibited the anticipated ties to personal definitions. That is, pro-gang definitions were lower among youths whose parents, in the respondent's opinion, would respond negatively to the knowledge that their child was involved with a gang (beta -.12). Pro-gang definitions were higher among those youths whose parents viewed positively the net s PROC logistic reports this measure of association between the predicted logit ~XI',B) and the population indicator, as in the correlation coefficient - 1