’I-A .9: .3 3 ; 2.: , Iran... .. .: A.,. El .7135) pt. .5... 5... .i : .. .. is; . : tn... n... 5 . “v.3...“ .2. 3... radar. 5.1.28! 1.. 4 1 ._ p! yfiiwnmu§di 22...... . , . .‘I‘i. . n. .L “I l... a .0” {3325553.}! Eliza: I... 3K3 il‘ 3 y 7 IV? '3" cl .9 t. I? {.336 xiv. fink.» .55. any. v3.0.4.3»); :IISSIP II: 7- | ..- y ms NWIi‘IIfilfiTIWIIflIflIT 2 3 1293 01388 3 This is to certify that the dissertation entitled DIMENSIONS OF RACE BIAS IN THE JUVENILE COURT: THE INFLUENCE OF LEGAL, DEMOGRAPHIC., SOCIAL, AND COMMUNITY CHARACTERISTICS ON DETENTION AND DISPOSITIONbDECISIONS presented Madeline Hordes has been accepted towards fulfillment of the requirements for Ph.D. degree in PsychoTOQy / 4/ "”‘ Zx/Z/yy/ Major professor Wiiliam S. Davidson MS U i: an Affirmative Action/Equal Opportunity Institution 0-12771 LIBRARY Michigan State University PLACE II RETURN BOXto Marathi. checkout from your record. TO AVOID FINES return on or More data duo. DATErfiDUE DATE DUE DATE DUE DIMENSIONS OF RACE BIAS IN THE JUVENILE COURT: THE INFLUENCE OF LEGAL, DEMOGRAPHIC, SOCIAL, AND COMMUNITY CHARACTERISTICS ON DETENTION AND DISPOSITION DECISIONS By Madeline Wordes A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1995 ABSTRACT DIMENSIONS OF RACE BIAS IN THE JUVENILE COURT: THE INFLUENCE OF LEGAL, DEMOGRAPHIC, SOCIAL, AND COMMUNITY CHARACTERISTICS ON DETENTION AND DISPOSITION DECISIONS By Madeline Wordes One possible reason for the disproportionate representation of Black and Latino youths in secure confinement is biased decision-making in the juvenile court. These decisions are based on multiple components including legal issues, demographic factors, social and family characteristics, and community composition. This study examined racial disparities in detention and disposition decisions in the juvenile court. The sample consisted of 455 youths living in unique census tracts who were charged with felonies. Data were garnered from case files in five courts within one state. Race bias was viewed as a multi-dimensional phenomena that can be seen directly, as well as through other individual, family, and community characteristics. A series of hierarchical logistic and ordinary least squares linear regression analyses were conducted with relevant legal, demographic, social, and community variables as predictor variables and detention and severity of disposition as outcome variables. Also, the relationships among these variables were assessed separately for Black and White youths. Study findings indicated that Black and Latino youths were more likely to be detained after controlling for prior legal status, offense behavior, social problems, and community variables. Further, while White youths were more likely to receive more severe dispositions after controlling for legal variables, when social variables were introduced to the model, this finding was not supported. Thus, a younger age and having more emotional problems were predictors of a more severe disposition. Separate analyses indicated that models for White youths were better predictors of detention and disposition than models for Black youths. Two interaction effects were found between race and other predictor variables. The interactions of race with percent Black in the census tract and household provider added to the evidence that detention decisions were partially based on race or racial stereotypes. A conflict perspective was used to interpret the findings. Future directions for research and action include exploring court differences and using a bureaucratic perspective to create interventions aimed at reducing the operationalization of conflict theory in the juvenile justice system. ACKNOWLEDGEMENTS There are several people I would like to thank for making this work possible. First, the staff at the juvenile courts who allowed us to spend many hours digging through their files and their brains. Their openness to work on the issue of minority overrepresentation gives us hope that the system can and will change. I would also like to acknowledge the work of the consultation group composed of juvenile justice professionals throughout Michigan who guided this study and contributed to its design and interpretation of the findings. Also, the many data collectors and the data manager, Beverly Dickinson, were invaluable. There are several fellow students and friends who helped me tremendously throughout this process. First, I'd like to acknowledge my parents and my family for their unending supply of love and support. Thank you to Susie Putman for becoming my Michigan family, Steve Cox, Kurt Ribisl, and Barb Schillo for continuous support, and Maureen Rumptz for continual faith and just about everything else. Finally, I want to thank Carolina Espinosa for giving her love, putting up with me, and helping me to finish this chapter in my life. lmportantly, I'd like to thank and underscore my appreciation for my mentors, Bill Davidson and Tim Bynum. Without Tim, I would not have been able to participate in the juvenile justice research that became my career and without Bill, I might never have finished. I want to express my deep appreciation to them for their support through the years and through this process. iv TABLE OF CONTENTS List of Tables .............................................. vii List of Figures .............................................. x Introduction ............................................... 1 Overview ............................................ 1 Bias Amplification ................................. 3 Functions of the Juvenile Court ....................... 4 Theoretical Perspectives ................................. 7 Consensus Perspective ............................. 7 Conflict Perspective ............................... 1O Applicability of Theory ............................. 14 Review of Empirical Literature ............................ 15 General Issues .................................. 17 Individual Level Studies ............................ 20 Mixed Level Studies .............................. 26 Structural Level Studies ............................ 30 Summary of Findings on Race Effects on Detention and Disposition Decisions .............................. 33 The Present Study .................................... 33 Theoretical Rationale .............................. 35 Methodological Rationale ........................... 36 Research Question 1: Was Race a Factor in Juvenile Court Decisions for Youths Who were Similarly Charged ......... 41 Research Question 2: What Influence Did Social Characteristics Have on Court Decisions and How was the Independent Effect of Race Impacted by Social Factors ..................... 43 Research Question 3: What were the Influences of Community Characteristics on Court Decisions .................... 44 Research Question 4: Were Factors Which Explained Court Decisions Differentially Weighted for Black Youths and White Youths ........................................ 47 Method .................................................. 50 Design and Data Sources ............................... 50 Setting ............................................. 50 Sampling ........................................... 51 Defining the Population ............................ 51 Sampling Individuals from the Population ............... 53 Sampling Individuals within Unique Census Tracts ......... 55 Power Analysis ....................................... 59 Procedures .......................................... 59 Measures ........................................... 60 Community Data ................................. 60 Case File Data Collection Forms ..................... 60 Description of Sample by Race/Ethnicity ..................... 68 Results .................................................. 74 vi Data Analysis Strategy ................................. 74 Bivariate Relationships Among Variables .................... 75 Zero-Order Correlations ............................ 75 Summary of Bivariate Relationships ................... 82 Assumptions of Normality, Linearity, and Homogeniety of Variance . 83 Findings from Logistic and Linear Regression Analyses .......... 83 Research Question 1 .............................. 85 Research Question 2 .............................. 89 Research Question 3 .............................. 93 Research Question 4 .............................. 101 Discussion ............................................... 121 The Relationship Between Race/Ethnicity and Detention ......... 121 The Relationship Between Race/Ethnicity and Disposition ........ 125 Methodological Implications .............................. 127 Theoretical Implications ................................. 128 Implications for Intervention and Policy ..................... 134 Limitations .......................................... 136 Future Research ...................................... 138 Appendix ................................................ 141 References ............................................... 143 vii LIST OF TABLES Table 1 Multivariate Studies Which Examined the Effect of Race on Detention and Disposition Decisions (1980 to 1995) .................... 34 Table 2 Summary of Samples, Variables, and Analyses in Empirical Literature (1980-1995) Examining the Effect of Race on Juvenile Court Decisions ........................................... 38 Table 3 Sample Characteristics from Original Sample and Derived Sample with Non-missing Data in Unique Census Tracts .................. 57 Table 4 Means and Standard Deviations (in parenthesis) of Offense Variables from Original Sample and Derived Samples .................. 58 Table 5 Definition of Variables and Data Sources for Variables In All Analyses 64 Table 6 Items, Item Total Correlations, and Alpha's of Scales Measuring Social Constructs .......................................... 67 Table 7 Demographic Description by Race/Ethnicity of Intake Sample ..... 68 Table 8 Description of Legal Characteristics by Race/Ethnicity of Intake Sample ............................................. 70 Table 9 Description of Community Characteristics by Race/Ethnicity of Intake Sample ............................................. 71 viii Table 10 Demographic Description by Race/Ethnicity of Social Data Sample . 71 Table 11 Description of Legal Characteristics by Race/Ethnicity of Social Data Sample ............................................. 72 Table 12 Description of Social Community Characteristics by Race/Ethnicity of Intake Sample ........................................ 73 Table 13 Zero-order Correlations of Variables from Intake Sample ......... 76 Table 14 Zero-order Correlations of Variables from Social Data Sample ..... 80 Table 15 Hierarchical Logistic Regression of Legal Variables, Demographics, and Race/Ethnicity on Detention .............................. 86 Table 16 Hierarchical Linear Regression of Legal Variables, Demographics, and Race/Ethnicity on Severity of Disposition .................... 88 Table 17 Hierarchical Logistic Regression of Legal Variables, Demographics, Social Variables, and Race/Ethnicity on Detention .............. 90 Table 18 Hierarchical Linear Regression of Legal Variables, Demographics, Social Variables, and Race/Ethnicity on Severity of Disposition ......... 92 Table 19 Hierarchical Logistic Regression of Legal, Demographic, Community, and Race/Ethnicity Variables on Detention ...................... 95 Table 20 Hierarchical Linear Regression of Legal, Demographic, Community, and Race/Ethnicity Variables on Severity of Disposition ............. 97 Table 21 Hierarchical Logistic Regression of Legal, Demographic, Social, Community, and Race/Ethnicity Variables on Detention .......... 99 Table 22 Hierarchical Linear Regression of Legal, Demographic, Social, Community, and Race/Ethnicity Variables on Severity of Disposition . 100 Table 23 Hierarchical Logistic Regression of Legal, Demographic, and Community Variables on Detention for Black Youths (n=232) ............... 103 Table 24 Hierarchical Logistic Regression of Legal, Demographic, and Community Variables on Detention for White Youths (n=204) .............. 104 Table 25 Hierarchical Linear Regression of Legal, Demographic, and Community Variables on Severity of Disposition for Black Youths (n=232) ..... 106 Table 26 Hierarchical Linear Regression of Legal, Demographic, and Community Variables on Severity of Disposition for White Youths (n=204) ..... 107 Table 27 Hierarchical Logistic Regression of Legal, Demographic, Social and Community Variables on Detention for Black Youths (n=232) ...... 110 Table 28 Hierarchical Logistic Regression of Legal, Demographic, Social and Community Variables on Detention for White Youths (n=204) ...... 111 Table 29 Hierarchical Linear Regression of Legal, Demographic, Social and Community Variables on Severity of Disposition for Black Youths (n=151) ............................................ 113 Table 30 Hierarchical Linear Regression of Legal, Demographic, Social and Community Variables on Severity of Disposition for White Youths (n=136) ............................................ 114 Appendix Sampling Frame and Sampling Weights for Each Juvenile Court . . . 141 LIST OF FIGURES Figure 1 Heuristic Model of Factors Affection Juvenile Court Decisions - Consensus Perspective .................................. 9 Figure 2 Heuristic Model of Factors Affection Juvenile Court Decisions - Conflict Perspective .......................................... 12 Figure 3 Juvenile Court Decision Points ............................ 18 Figure 4 Interaction of Race and Percent Black in Census Tract for Detention Decision Controlling for Legal, Demographic, Social, and Community Variables ........................................... 116 Figure 5 Interaction of Race and Household Provider for Detention Decision Controlling for Legal, Demographic, Social, and Community Variables ........................................... 118 xi Dimensions of Race Bias in the Juvenile Court: The Influence of Legal, Demographic, Social, and Community Characteristics on Detention and Disposition Decisions Introduction merview The fact that Black youths are disproportionately represented in juvenile justice systems within the United States is undisputed (Pope & Feyerhen'n, 1992; Bishop & Frazier, 1990; Kempf, Decker, & Bing, 1990; Krisberg, Schwartz, Fishman, Eisikovits, Guttman, & Joe, 1987). In fact, the arrest rate for Black youths in 1993 was more than twice the arrest rate for White youths (Krisberg, DeComo, Rudenstine, & Rosario, 1995). Further, Black youths were four to five times more likely to be placed in State operated secure facilities (Krisberg et al., 1995). The trend of increasing disproportionate representation has been accelerating. In a recent study, McGarreII (1993) demonstrated that there were higher rates of secure detention for Black youths than White youths across the country. Further, he noted that the rate of detention for Black youths significantly increased between 1985 and 1989 while the rate for White youths remained constant. There are three main explanations for the phenomena of disproportionate representation. First, some believe that there are differential base rates of offending. Often arrest data were used to support the notion that disproportionate representation was due to Black youths committing more crimes or more serious crimes (Hindelang, 1978). It should be noted, however, 2 that data which were officially entered into the system (Uniform Crime Reports or local arrest data) already may have been biased because of earlier discretionary police decisions. Some people, such as Wilbanks (1987), contended that there is no racial prejudice or discrimination in the justice system. He stated that some individual decision-makers are biased, but the system itself is not biased; individual decision-makers are biased in different directions and cancel each others biases out. The second main explanation for disproportionate representation was that penalties were stiffer for offenses more likely to be committed by minority youths. For instance, McGarrelI (1993) stated that the increasing propensity to detain youths charged with drug offenses disproportionately affects minority youths. The present study, however, did not focus on differential offending or discriminatory legal issues as explanations for disproportionate representation. This study was an exploration into possible systematic biases in decision- making. Specifically, how the race/ethnicity of the youth acts in concert with other salient factors to affect juvenile court processing. The third explanation is there is race bias in juvenile justice processing. There are three ways to test whether there is racial bias in the system: a) compare self-report data to official data, b) statistically hold relevant factors constant and examine the influence of race, and c) determine the importance of various factors In decision-making differentially by race. The latter two methods 3 of determining race bias were the focus of the present study and will be discussed in detail below. However, the first method will be discussed briefly now. One way to test whether disproportionate representation is due to system bias is by comparing data on self-reported delinquency to data in official records. Studies which have compared self-report delinquent behavior with official records have found little difference in self-reported behavior, yet disproportionate representation in official records (Huizinga & Elliot, 1987; Krisberg et al., 1987). Huizinga and Elliot (1987), for example, used self-report delinquency data from the National Youth Survey and official data from police records on the same youths. These authors concluded that "there are few if any substantial and consistent differences between the delinquency involvement of different racial groups" (p.215). There were no Significant differences in the self-reported delinquency for most of the serious offenses either, including felony assaults, felony thefts, and index offenses in general. In examining arrest rates, they found that the overall arrest rate for Blacks for serious offenses was about 2 or 3 times the rate for Whites. According to these authors, the negligent differences in delinquent behavior across racial categories could not account for the differences in arrest or incarceration rates. Bias Amplification Due to the many points of discretionary decision-making in the system, there are many points in which bias may enter the system. For instance, the police may differentially patrol an area and arrest more Black youths (Sampson, 4 1986). The police decision to refer the youth to court or detention may also be biased (Wordes & Bynum. 1995). The pool of youths referred to the court is then disproportionately represented by Black youths. The court process also contains several stages (which may each be biased) which finally produces a population of incarcerated youths who are mainly youths of color (Krisberg, 1993). Some have used the term “bias amplification” to indicate that as juveniles are processed further into the system, small biases at each stage can lead to large race differentials at the most severe sanctions (Bridges & Crutchfield, 1988; Farrell & Swigert, 1978). Although there may be biases at both the police and court stages of the juvenile justice system, it is important to explicate the process at the juvenile court for two main reasons. First, court processing can lead to the most severe penalties. For example, detention of a juvenile can have severe emotional and physical consequences (Sarri, 1974). The juvenile court has the power to restrict the living arrangement of a youth and can be coercive or deprive liberty (Sampson & Laub, 1993). Second, juvenile courts collect data which can provide a much richer source of social information than police data. These ”extra-legal“ factors may play a role in the decision-making process and the researcher can examine much of the same data available to the decision- maker. Eugctionfis; of the Juvenile Court Studying bias in the juvenile justice system is complicated by the fundamental functions of the system. The Progressive era reformers who 5 created the modem juvenile justice system designed it as both a means of protecting the community as well as providing treatment to youthful offenders (Platt, 1969). According to Bell and Lang (1985), "The juvenile justice system was created in order that special circumstances and problems of youth could be accommodated by a wider variety of formal and informal procedures and ‘treatment' alternatives“ (p.309). Accordingly, the reformers created a juvenile court system in the United States which now serves three main functions: a) keep social order, b) extend due process to juveniles, and 0) provide ”treatment” (Fenwick, 1982). These three fundamental functions may actually add to disproportionate representation. The first function, to keep social order, can also be interpreted as maintaining the status quo. The Progressive reformers of the early 1900's were generally people with many resources who wanted to "help" others, but not at the sacrifice of their own status in the capitalist system (Rothman, 1980). According to Rothman, the reformers did not want to change the American economic and social system because they benefitted from it. Thus, maintaining the status quo, which meant keeping the “rabble class“ in ts place was one implicit function of the juvenile court. Disproportionate representation would arise because the lower economic class contains a disproportionate number of racial minorities and young people (Quinney, 1977). The second function, to extend due prOcess, may also lead to greater disproportionate representation. Some have stated that due process is more likely to be extended to those with resources (Huizinga & Elliot, 1987). It is 6 possible that extending due process is a function of access to resources. Accordingly, youths who have access to attorneys and resources for private mental health treatment may be disproportionately White, leaving a greater percentage of minority youths to be handled by the juvenile justice system. The third function, providing treatment or being therapeutic, led to the doctrine of 'in loco parentis' which allows courts to fulfill the parental role when it deems the parents unfit or unable to do so. Research is then "complicated by the juvenile court philosophy of 'parens patriae' which, contrary to the rule of law that governs adult criminal cases, not only accepts, but justifies relatively high levels of discretion at all stages of the juvenile justice process' (Johnson & Secret, 1990). Thus, the juvenile justice system was designed to take social factors into account when making dispositional decisions. It should offer no surprises that “extra-legal” factors play a role in decision-making. According to Cohen and Kluegel (1978), “In systems such as the juvenile court, where the boundaries and limits of discretionary power regarding the application of rules are not very clear, the possibility of unequal treatment of clients increases“ (p.147). The fundamental purposes of the juvenile court as discussed above (to keep social order, extend due process to juveniles, and dispense “treatment" through maintaining discretion), may each contribute to racial disparity in the juvenile justice system. The two theoretical perspectives outlined below address both the nature of juvenile justice process and the specifics of how race may influence court decisions. 7 Theoretical Perspectives All theories regarding formal social control in the justice system begin with the premise that the main function of the system Is to preserve or create social order (Bridges & Myers, 1994). There are two major perspectives which have been derived from this main assumption: consensus and conflict. The perspectives differ in that the first purports that the social contract benefits society as a whole, whereas the second contends that the system serves the purposes of those in power. Consensus Perspective The first perspective is called consensus theory. Durkheim was one of the originators of this perspective which emphasized the issue of just punishment for norm violations. This perspective contended that the severity of the punishment was in direct proportion to the importance of the norm that was violated (Durkheim, 1964). Bridges, Cmtchfield, and Simpson (1987) interpreted the consensus perspective to mean that ”punishments are representations of broadly based norms and values, and constitute a line of defense against crime“ (p. 346). Presumably these norms are agreed upon for the benefit of the society as a whole. Most of the literature asserted that if the consensus perspective was to be supported, then differential treatment by race would not be apparent in the empirical literature. Consensus theory has been interpreted to mean that legal factors will be the sole determinant of disposition (Johnson & Secret, 1990). These legal factors included offense characteristics and prior offense history. 8 Some have claimed that Black males are more likely to violate the law and commit more serious crimes and are thus more likely than other groups to fall under the purview of the juvenile justice system (Hindelang, 1978). Thus, when examining racial disparity it is important to control for offense characteristics and prior offense history. Contrary to the assumptions of equal treatment under the law, Tittle (1994) asserted that unequal treatment within the justice system is a necessary conclusion it the logic of the consensus perspective is followed. He stated, "There will be differential sanctioning of individuals who commit similar forms of behavior, and those individuals whose life-styles, past histories, or personalities suggest greater threat to the group as a whole will be more severely sanctioned regardless of the current offense of which they are accused" (p.25). He interpreted the consensus perspective to mean that people who belong to groups or categories which seemingly threaten the collective good of society will be more likely to be the objects of control and receive more severe sanctions than those people who are non-threatening to the “collectivity”. Thus, social and family characteristics which deviate from societal norms will affect court decisions. Figure 1 is a conceptual model of the consensus perspective. It shows that legal issues will be related to court decisions. The dotted lines Indicate the ' relationships proposed by the new branch of consensus theorists (e.g., Tittle, 1994). These theorists believe social factors will affect decisions because social characteristics violate social norms and threaten the collective good. 820$ 2E3 w .98... 30568 taco 320$ Ham. ozfioamaon. macaw—Loo A 307.800 .500 o__co>:_. 9285‘ 328“. .o .822 25:31 F 939“. 1O Regardless of whether the consensus perspective assumes differential treatment due to legal and/or “extra-legal” factors, it is based on the idea that society is unified by shared goals. They believe that youths who become part of the juvenile justice system do so because they violated social norms which damages the collective good. Conflict Perspective While the consensus perspective holds a notion of a collective good, the conflict perspective views society as containing various groups vying for power. There are a variety of specific theories within the general perspective, however the overriding theme is that power is a commodity that is struggled over and that there are inherent inequalities in power among groups. Thus, groups have conflict with one another in order to promote their own interests (Chambliss & Seidman, 1982). A Marxist interpretation of conflict theory has often been used to explain bias in the justice system. This argument can be summarized as a view of criminal justice in the United States as a system designed to maintain social control for the benefit of capitalists. As Quinney (1977) stated, "In capitalistic society the healthy order is the one that benefits the capitalistic class, the class that owns and controls the productive process“ (p.3). Quinney purported that the legal system and the 'rational' administration of laws are the capitalists method of controlling who becomes part of the justice system. Sampson and Laub (1993) described the conflict perspective succinctly. They stated, "Conflict theory views society as consisting of groups with 11 conflicting and differing values, and posits that the state is organized to represent the interests of the powerful, ruling class. Criminal {law is thus viewed as an instrument to protect the interests of the powerful and the elite, with punishment based largely on extralegal variables (e.g., race, social class, etc.)" (p.288). Thus, those groups (e.g., people of color, poor people) that threaten the hegemony of those in power are subject to greater social control. In the juvenile justice context this implies more formal processing and harsher sancflons. Thus, traditional conflict theorists believe that sanctions against less powerful groups will be harsher because their powerlessness makes them more vulnerable to justice system intervention (Turk, 1969; Quinney, 1977). This can be interpreted to mean that fewer resources make a powerless group more vulnerable to formal social control. To the contrary, Hawkins (1987) argued that it is not fewer resources which precipitates more formal social control, but the threat posed by less powerful groups to the dominant group. Thus, minority groups having more power would be a precursor to harsher sanctions. A simple conceptual model of the conflict perspective is presented in Figure 2. This model shows the main difference between the consensus and conflict perspectives. The differential groups In society (defined here by race/ethnicity and community characteristics) are integral in determining court process. The figure indicates that these factors also indirectly influence legal and “extra-legal" factors that affect decision-making from the consensus perspective. 12 6253:3905? bEaEEoo 20303 :38 / 9.6333 65:00 - 29280 :30 0:55.. 958.? 808mm ho Boos. 23:32.. N 239“. 3203 SEE. a .98... 390$ .32 13 Frazier, Bishop, and Henretta (1992) have used both the traditional conflict perspective and the revision offered by Hawkins to offer specific hypotheses. They cite two main differences. First, the traditional conflict perspective would suppose that the fewer the number of Blacks in a community the less power they would have. Therefore, in a mostly White community, Blacks would be most vulnerable to formal social control. On the other hand, Hawkins supposed that the greater the proportion of Blacks in the population, the greater the threat to dominants because Blacks would have more power. A second difference between competing conflict perspectives is the issue of economic resources. The traditional theory would suppose that the lower the economic resources, the harsher the sanctions. Hawkins' revision .of this theory, however, would imply that the less economic power differential between Blacks and Whites, the greater justice system intervention (Frazier, Bishop, & Henretta, 1992). In other words, the more economic power of Blacks, the greater the need for social control to maintain White hegemony. McCarthy (1991) suggested similar distinctive hypothesis stemming from the conflict perspective. One is that the elites will respond to the threat of a large subordinate population by increasing social control efforts. The second is that subordinates are less effective at resisting social control by elites when their population is small. Consequently, the larger the population, the more power Blacks will have to resist social control efforts. Most research using the conflict perspective Is based on the former idea; the larger the subordinate population, the greater the efforts at social control. 14 Sampson and Laub (1993) combined the points of view into one general hypothesis about the relationship between race and court disposition. They stated, "Counties characterized by racial inequality and a large concentration of the 'underclass“ are more likely than other counties to be perceived as containing offensive and threatening populations and, as a result, are subject to increased social control by the juvenile justice system“ (p.293). This hypothesis combines the previously disparate conflict perspectives by adding a class dimension. Thus, power was defined by racial minority status as well as economic resources. Thus, Sampson and Laub believed that a large concentration of Blacks does not indicate that they have power unless they have economic resources. Applicabilitv of Theory The consensus and conflict perspectives provide the context for examining bias in juvenile court decisions. The consensus perspective suggests that individuals will receive sanctions based on the offense they committed or the social norms they violated. Race, then, would not be a factor in decision-making except for being related to norm violating characteristics or legal factors. The critical issue in the conflict perspective of structural inequality stems from the power relationships between groups. Basically, the perspective contends that those in power want to keep control of those who have less power and therefore, define appropriate behavior for people of low status (Quinney, 1977). Thus, people who are part of less powerful groups (racial 15 minorities) will receive harsher sanctions on the basis of their membership to the group. Unfortunately, researchers who use the conflict perspective as a basis for their studies or as an explanation for their findings do not link the theory to individual level decisions. They often cite macro-level reasons for disparate treatment, but do not connect the decision made about the individual to macro- level phenomena. The characteristics of the community in which the youth resides (representing group power or threat) may impact decisions about the individual differentially by race. Before addressing specific theoretical propositions and hypotheses, it is first necessary to review the extant empirical literature. Prior empirical research along with the theoretical background described above will provide the groundwork for examining the complex relationships among a youth's race, individual offense and social characteristics, community factors, and juvenile court decisions. Review of Empirical Liteirat_u_r_e The literature was searched using the NCJRS database, Criminal Justice Abstracts, Sociological Abstracts, and Psyc Info using the terms “race and juvenile justice“. This review focuses on the empirical literature in the past fifteen years (1980-1995). Prior to this period, most studies used bivariate analyses to examine decision-making (for review see McCarthy & Smith, 1986 or Pope & Feyerherm, 1990). Studies supporting discrimination hypotheses were criticized because they did not control for various offense factors which 16 may be related to differential decision-making (Wilbanks, 1987). Studies prior to 1980 rarely used multivariate techniques and thus few important variables ‘ (e.g., offense behavior, family functioning) could be accounted for when examining the factors affecting disposition. Some earlier studies established a relationship between race and detention (Liska & Tausig, 1979; Thomas & Sieverdes, 1975). Others, found no such relationship (Cohen & Kluegel, 1978). Seminal studies such as Thomberry (1973) found relationships between being a minority and being processed further in the system, yet others found the opposite relationship (Smith, 1980). The main criticism of these studies was the lack of statistical control because they did not account for other important factors. Other reasons for seemingly opposite findings were differences in jurisidictions, offenses sampled, control variables used, etc... In sum, in early studies there was much disagreement in the literature regarding whether any race effects were present. Since the recent move toward the convention of using multivariate statistical methods to analyze data, there has been more empirical support for the notion of differential treatment related to race of the youth. A recent summary of research suggested that “there is substantial support for the statement that there are race effects in operation within the juvenile justice system, both direct and indirect in nature“ (Pope & Feyerherm, 1990). While this general statement was supported by some of the recent literature, the nature of the relationship between a youth's race and juvenile justice processing was not clear. 17 General issues Several studies have examined the effect of race on various juvenile court decision points. The studies varied in the number of decision points examined, the method of data collection, the number of jurisdictions sampled, and the number and type of independent variables used as controls. Before reviewing the literature it is important to clarify these several issues that make understanding and examining the juvenile justice process complex. The first main issue is that the detection of bias in the juvenile court is further complicated by the several discretionary decisions made within the processing of each case. Before reviewing the literature it is necessary to explain the commonalities in decision points across courts. There are two basic decision points within each court: intake and formal hearing. Figure 3 shows the common decision points across courts. Intake includes making the initial decision of intake into the court and whether or not to file a petition. At intake the intake worker (usually a case worker or probation officer) makes the decision of whether or not to detain the youth prior to a formal hearing, whether to drop the case, divert it, or set it for a formal hearing. The one or more formal hearings are used for adjudication purposes (determination of guilt) and disposition purposes (type of sanction imposed). At a formal hearing (usually conducted by a referee or judge) the main decision choices are: to drop the case, send it to a diversion program, place it on the consent calendar (no penalty, but a formal record), make the youth a ward of the court (e.g., for 18 E m>> 95w Em>> .80.. , c8560 ..r 2380 EEO”. «ow . 520 520 M no.0 / no.0 EmEmomEA ........ . ........... ZOE—Owe 03210“. ZO.m_Ow0 w¥<._.z. $500 56600 560 2:82;. m 9:9”. 19 probation purposes), or place the youth in the custody of the State Department of Social Services. The second main issue is that many studies have not addressed the juvenile court system as a process. In early studies each decision point was viewed as separate. More recently the importance of estimating the cumulative effects of earlier decisions on later decisions has been discussed (Bridges & Crutchfield, 1988; Frazier & Cochran, 1986). For example, Frazier & Cochran (1986) studied the effect of detention on subsequent decisions. They found that "detainees are more likely to receive more harsh intake recommendations, more severe action by the State Attorney, and both more formal and more severe final dispositions of their cases“ (p. 300). Thus, controlling for other legal factors, they found that detention had an independent effect on disposition. Another issue adding to the complexity of study design is the data sets used. Often times, researchers use data which is convenient and may not include variables which have been previously shown to be important. For instance, many researchers used computerized record systems which often contain only a few variables pertaining to the case (e.g., current offense, prior record, age, gender, race). Current offense is often only coded into large offense categories such as felony or misdemeanor and specific and possibly relevant offense behavior is not taken into account. Further, depending on the decision point one is examining certain types of data are not available. For instance, the intake decision is usually made with very little formal record 20 keeping. Thus, at intake or pro-trial detention, the records do not usually contain data on family or social related issues. In the review that follows, it is necessary to keep in mind the central issues in juvenile court research discussed above: variety of the decision making points, possible amplification bias in processing, and the data set used. This review of the empirical literature is organized into three sections. First, studies conducted with individuals as the unit of analysis will be examined. Second, studies which mixed individual and group level data will be discussed. Finally, macro-level studies which used aggregate data across jurisdictions will be reviewed. Within each of these three levels of analysis, the literature is organized by complexity of design and analysis. Individual-level Studies Early studies almost exclusively used single jurisdictions and single decision points. More recently, however, studies have used single jurisdictions to study several different decision points. For instance, Fagan, Slaughter, and Hartstone (1987) used data from a B-level SMSA‘. Within that site, they examined six decision points from both police and court records: apprehension, detention, prosecutorial charging, adjudication, probation, disposition. They used separate samples from police, prosecutor, court, probation, and correction agency. Due to this sampling procedure they were unable to follow the same group of youths through the process. Using contingency tables (with three variables) they purported that racial disparity existed in almost all decisions either directly or indirectly. However, they only controlled for two or three 21 variables at a time (i.e., offense type, race, and various single individual level factors) and had very small (single digit) cell sizes for some of their analyses. McCarthy and Smith (1986) conducted a similar, but more methodologically sophisticated study. These authors also used data from a Single B-Ievel SMSA, but they followed a single group of youths through juvenile court processing. They used path analysis to examine data at three decision points: all referrals, all petitions, and all adjudications. The independent variables they used were race, sex, social class (median income of postal district), prior record, seriousness of offense, and days detained. Although this study was primarily conducted at the individual level, the authors did use an indicator of community level socio-economic status. The dependent variable, disposition, was coded on a 15 point ordinal scale of severity of sanction. They found that legal factors (offense seriousness, priors) had direct effects on all referred cases, but declined in importance when modeling the process for adjudicated youths. They interpreted this finding to mean that as the sample became more legally homogenous (more serious cases were adjudicated), extra-legal factors became more important. Thus, race and social class were found to be significantly related to disposition and have more of an effect in the legally homogenous population. In fact, being a racial minority had one of the largest direct effects of any variable on disposition when looking at petitioned and adjudicated youths. Being detained also had an independent effect on final disposition. Due to their data source, they were unable to control for specific 22 offense behavior or any social factors which may also have influenced disposition. Kempf, Decker, 8. Bing (1990) conducted a thorough study using case file data from eight jurisdictions in Missouri. They used logistic regression to examine seven dichotomous outcome variables (dismissal, informal handling, detention, petition, adjudication, disposition, and recidivism). They controlled for gender, offense type (dichotomous coding of felony, violent, and status offense variables), presence of counsel, referral source, parental willingness, household provider, and youth alcohol abuse. They conducted separate analyses for rural and urban jurisdictions. In the urban jurisdiction sample (n=1739) they found that race was a Significant factor in two of the seven analyses. Being Black was related to being detained and recidivating after controlling for other factors. Also, if the parent was unwilling to accept the youth, she or he was more likely to be detained. Out of home placement was influenced by not living with two parents. For the rural jurisdictions (n=881), race was an independent predictor of outcome in five of the seven models. Whites were more likely to be dismissed and more likely to be placed out of home. Blacks were more likely to have an informal disposition, be detained, and be adjudicated a delinquent. Detention was also influenced by a record of the youths alcohol abuse. Disposition was influenced by alcohol abuse and not living with two parents. Fenwick (1982) also found race effects in decision-making in a single court in a major urban center. He collected data on two decisions: petition and detention. The observational methodology used in this study was rare and 23 adds insight into court decision-making. Cases were randomly sampled from the population at intake (N=350), the hearings were observed, and the case files were coded. The main independent variables included sex, race, age, prior record, seriousness of offense, family disaffiliation and demeanor. Disaffiliation was measured as a composite variable consisting of: family member present at hearing, caregiver interested or disinterested in youth's welfare, behavior problems at home, and family structure. Demeanor was coded as negative, positive, and neutral. Fenwick used multiple classification analysis to Show that all the independent variables were significantly related to the petition decision. Black youths were more likely to be petitioned as were youths who were socially disaffiliated and showed a negative demeanor. However, the youth's family disaffiliation was the sole statistically significant determinant of the decision to detain. This finding is interesting because family information is not available at the intake stage in data sets that do not contain observations and most studies are unable to assess demeanor with case file or computerized data records. Kurtz, Giddings, & Sutphen (1993) conducted the only other study of juvenile court decision-making using observational data from eight counties in a single state. They examined multiple decision points by using data collection forms filled out by the different decision-makers at each stage of processing. This method allowed them to track individual youths through all stages in the process, yet their sample size was reduced considerably at each stage (n=346 at police, n=180 for court intake, n=118 adjudication, n=89 at disposition). Four 24 separate path analyses were conducted with the dependent variables at four decision points: law enforcement, intake, adjudication, and disposition. The independent variables included: seriousness of offense (Wolfgang, Figlio, Tracy, & Singer, 1985), harm to person, harm to property, prior record, race, demeanor, victim characteristics, socio-economic status, and representation by attorney. The family characteristics measured were not entered into the model because bivariate analyses showed no direct relationships with disposition. Their results suggested a complex relationship between race, social class, and disposition. Socio-economic status had strong indirect and direct effects on disposition. Youths with lower socioeconomic status received more severe dispositions. Race also showed direct and indirect effects. Blacks were more likely to have been charged with more serious offenses and to have had a more negative demeanor. Through these variables, race had an indirect effect on disposition. The zero-order correlations were not presented so it was difficult to determine the degree of multi-collinearity among the variables. It seems likely that having access to an attorney or being represented by a public defender would be highly correlated to socio-economic status, offense severity, and possibly race. Summary of Individual-level studies. Almost all of the studies prior to 1980 (for a review see McCarthy & Smith, 1986 or Pope & Feyerherm, 1990), were completed in one jurisdiction using individual level data. The findings regarding differential treatment by race were mixed primarily due to the methodological shortcomings of most studies. The more recent studies 25 described above have all used multivariate analyses at multiple decision points. Each of the studies found race effects in juvenile court decision-making, however, the nature of the relationship is still unclear. The main reasons for the lack of certainty regarding these findings are methodological issues. First, some studies did not focus on following a cohort of youths through the process (Fagan et al., 1987). Important process variables including previous decisions, such as detention, have been shown to affect subsequent dispositional decisions and are often not account for (Frazier & Cochran, 1986; McCarthy & Smith, 1986). Second, specific offense behavior was not adequately controlled for in these studies. For instance, Fagan et al., (1987) measured specific behavior, but their choice of statistical method did not allow the simultaneous control of more than a few variables. The other studies mixed various types of offenses into one statistical equation. A single measure of offense behavior (e.g., an ordinal variable of status, misdemeanor, felony) may not be adequate to measure actual harm done or seriousness of the offense. Criticisms aside, these studies pointed to several important variables for further study. The central findings were that race and social class affected decision-making at various stages and not a single stage conclusively. Thus, it is important to examine both an intake and dispositional decision for race and class bias (McCarthy & Smith, 1986). Also, family variables are important to measure because family disaffiliation was the strongest predictor of detention (Fenwick, 1982) and household provider influenced outcome in the study by Kempf, Decker, & Bing (1990). Finally, specific offense behavior is important to 26 measure. Fagan et al., (1987) found that there was a strong bivariate relationship between using a weapon and being a racial minority. Thus, without adequately controlling specific offense, individual, and family variables, the influence of a youth's race on juvenile court decisions cannot be tested. Mixed-level Studies Some studies have suggested the differences in individual-level study findings may be, in part, a function of differences in the court or community context (Hawkins & Pope, 1993; Pope & Feyerherm, 1990; McCarthy & Smith, 1986). Hence, there has been an increased emphasis upon conducting multijurisdictional studies. In fact, Frazier, Bishop and Henretta (1992) stated “researchers should not expect race bias to be either ever-present or uniform across legal jurisdictions. Instead research should focus on identifying the social contexts most likely to produce race differentials“ (p.449). Some of the studies which examined the social context of juvenile courts did so by using both individual and county level data. Frazier and Cochran (1986) used data from 32 jurisdictions throughout Florida with a computerized database comprised of State of Florida delinquency cases (N=9794). The independent variables used to predict the detention decision were age, race, gender, offense seriousness (coded misdemeanor or felony), number of offenses, priors (coded yes/no), and percent urban in county. The model indicated race was not significant in detention decision, however significant influences included being older, being female, charged with a felony, having priors, and being in a rural county. They suspected that law enforcement 27 officers in rural counties had closer access to detention facilities. They believed that detention facilities are normally located in rural counties. They could not test that hypothesis. One major problem with this study was that they did not adequately control for the offense type. Global categories such as misdemeanor and felony did not give an indication as to the nature of the offense. Another study using both individual and court level data in one state was undertaken in several jurisdictions in Nebraska to predict decisions at multiple decision points. Johnson and Secret (1990) analyzed data from the Nebraska Crime Commission Juvenile Court Statistical data sets for 1982 through 1987. They examined four stages in the court process with dichotomous dependent variables: detention, petition, adjudication, and disposition. Using logistic regression, they regressed race, age, gender, seriousness of offense ( an eight point scale), priors, decision made at prior stages (i.e. detention), and court type (coded juvenile or county court) on each of the dependent variables. They found that Blacks were approximately two times more likely to be detained than Whites. Whites, however, were more likely to be adjudicated. They explained this second finding by the notion that there might be less evidence against Blacks because of earlier police biases and judges or referees would be likely to take earlier biases into account making an adjudication decision. The detention decision would contain no such accountability. Dannefer and Schutt (1982) also examined only one county level variable (urban vs. suburban). They tested the hypothesis that “the larger the proportion 28 of minority group members in a population, the greater the likelihood of discrimination by official agencies against juvenile justice offenders who belong to minorities“ (p. 1116). To test this hypothesis they used data from two counties. They used log-linear analysis with a categorical disposition variable with values of dismissed, probation, and incarceration. The independent variables were offense type (minor, drug, property, & violent), priors (yes, no), race, family configuration (two parents, other), sex, county (urban, suburban). They found that while Black youths were treated more harshly by police, there were no significant differences in court processing. They found no differences in court processing by type of county, however, they only used two counties in their sample. These authors, like Johnson and Secret (1990), believed that the court made up for bias in the police decision because Blacks were discriminated against at the police stage and the court had to make decisions based on evidence. They concluded, "Future research in this area should concentrate on comparative studies in which social environments are sampled on direct measures of characteristics of social environments“ (p.1130). In the study by Frazier, Bishop, and Henretta (1992) better community data were used than in the previous studies reviewed. They used population, economic and crime profiles of counties (N=32), along with individual level data of Florida delinquency cases (N=48,961). They used three stages of decision- making as their dependent variables: intake recommendation, formal/informal handling, and final disposition (secure, non-secure). The Independent individual-level variables included race, age, gender, offense severity (seven 29 point scale), number of prior offenses, and prior disposition. County level variables included: percent White, percent Black living below poverty level, racial income inequality, index crime rate, and juvenile crime arrests. Using logistic regression they found that for each decision point, Blacks were more likely to receive the more severe outcomes when other variables were held statistically constant. Also, the greater percent White in the county, the more severe the disposition of Blacks compared to Whites. Summary of mixed-level studies. The studies which mixed levels of analysis offered no firm conclusions regarding the existence of racial bias against minority youths in the juvenile courts. These studies were also not able to establish any consistent relationships between disposition and county or community level characteristics. In fact, in some of the studies reviewed, Whites either received more severe dispositions or there were no race differences. As stated earlier in the summary of the individual level studies, a methodological shortcoming of each of these studies was their inability to control for specific offense behavior. Another issue was the lack of specification of community level variables. In some studies there was only one dichotomous variable measuring a court or county level influence. The final study presented in this section contained the most community level variables (Frazier, Bishop & Henretta, 1992). They found that being a racial minority was a predictor of more severe dispositions at all levels of the juvenile court process after accounting for county level characteristics. Thus, 30 specifying community characteristics or social environment may help to clarify the nature of the relationship between race and juvenile court processing. Structural-level Studies Most individual and mixed-level studies included in their discussions a need for more macro-level research. For instance, Fagan, Slaughter, and Hartstone (1987) concluded, “It is important to remember that inequality and disparity are endemic in our society, and that juvenile justice agencies are no less a part of that society than any other institution. The processes that yield economically and socially isolated minority communities also contribute to the systemic responses documented in this study“ (p.253). While these authors concluded with this statement, their data were insufficient to support this contention. There were very few studies which have examined juvenile justice decisions on a macro-level. Although much research has been conducted on the relationship between social structure and delinquency, only two recent studies have been conducted on juvenile court decision-making. The first used data from the Minnesota juvenile court system to explore the differences between urban, suburban, and rural contexts on juvenile court dispositions (Feld, 1991). Fold contended that because racial heterogeneity in a county “decreases the effectiveness of informal social controls, then urban counties may need more formal mechanisms of control" (p.160). Unfortunately, all of his analyses were bivariate. Thus, he only explored the relationships between urban/suburban/rural environments and other variables individually. 31 He concluded, however, that in urban counties which he characterized as having a more heterogeneous population, the juvenile court took more formal action, whereas in rural counties there was less formality and more leniency. Sampson & Laub (1993) conducted a more comprehensive macro-level study of the relationship between inequality, underclass community, and juvenile court dispositions. Their main question was, “How does structural context- especially racial inequality and the concentration of ‘underclass' poverty- influence formal petitionings, predisposition detention, and placement of juveniles?“ (p.287). Their stated goal for their research was to lay out the "groundwork for a better understanding of the relationship between larger societal forces of increasing poverty and inequality (Wilson, 1987,1991) and formal systems of juvenile social control“ (p.287). They used data from National Juvenile Court Data Archive and the City and County Data Book. They aggregated data from 538,000 individuals into 21 states and 322 counties. Offenses were grouped into four classes (crimes against property, crimes against persons, drug offenses, and public order offenses). Several structural variables were compiled: underclass poverty (six items), racial inequality (Black/White poverty & proportion of Blacks below poverty), county wealth, urbanism, density of youths population, region of country, county dollars spent on criminal justice system, residential mobility, and referral rate to court. They used separate logistic regression equations to predict detention for each of the general offense categories listed above. Their findings present a 32 complex picture. In sum, they found that underclass poverty predicted detention for all non-petitioned youths for all offense types. Along with underclass poverty, a higher referral rate, being from a western state, being less urban, being a wealthier county, and spending less resources on criminal justice each contributed significantly to higher detention rates for personal offenses. They also ran separate linear regression analyses for Blacks and Whites. The model used was significantly better at predicting detention for Blacks than Whites. This indicated that structural characteristics of the county were more predictive for Blacks than Whites. This type of analysis has major implications for future research. In prior research, the detection of bias was often only measured in terms of whether race had a direct or indirect effect on disposition. Examining how the variables relate to each other differently for Blacks and Whites improves our understanding of decision-making. Furthermore, examining the amount of differential influence offense, social, or community factors have for youths based on their race is another manner in which to examine bias in decision-making. Summary of structural-level studies. While these studies have contributed a great deal to push forward the study of context on juvenile justice decisions, many issues still remain unresolved. First, Feld (1991) did not look at multivariate relationships so the independent impact of an urban setting cannot be discerned. Second, the results of the Sampson and Laub (1993) study were difficult to interpret due to the separate offense categories being 33 broad. For example, personal offenses could include behavior such as throwing something at a person without hitting them, as well as, committing murder. It is thus difficult to disentangle the affect of individual level behavior data on only macro-level variables. From these sources, however, it seems county and community characteristics play a great role in determining detention and disposition decisions. These studies show the importance of including measures of social structure in examining decision-making in the juvenile court. Summary of Findings on Race Effects on Detention and Disposition Decisions There have not been many recent studies using multivariate techniques to study race effects in the juvenile court decision-making. Table 1 summarizes the most recent literature for both detention and disposition decisions. Overall, there were no clear trends defining whether race of the youths had a direct effect on decision-making at these stages. In some studies race had no effect, in others Black youths were more likely to receive more severe sanctions, and in two Studies (one examining detention and the other examining disposition), White youths were more likely to receive the more severe option. The Present Stpgy It is not difficult to document the enormity of the problem of disproportionate representation of Black youths in the juvenile justice system, especially secure confinement. It is more difficult, however, to understand the complex issues involved in exacerbating or ameliorating the problem. The present study assessed the impact of demographic, offense, social, and community characteristics on the juvenile court decisions to detain a youth and Table 1 34 Multivariate Studies Which Examined the Effect of Race on Detention a_ng Disposition Decisions (1980 to 1995) Authors Were there direct effects of youth's race? Damien Disposition Fenwick (1982) Frazier & Cochran (1986) Frazier & Bishop (1985) Johnson & Secret (1990) Kempf, Decker, & Bing (1990) Kurtz, Giddings, & Sutphen (1993) none Blacks more likely Whites more likely Blacks more likely Blacks more likely none Dannefer 8 Schutt (1982) Frazier, Bishop, & Henretta (1992) Johnson & Secret (1990) Kempf, Decker, & Bing (1990) Kurtz, Giddings, & Sutphen (1993) McCarthy & Smith (1986) none Blacks more severe none Whites more severe Blacks more severe Blacks more severe dispose of the case. This study used theory and prior research to explore this multi-dimensional issue and guide the methodological improvements. First a general theoretical and methodological rationale for the present study will be presented, followed by more specific rationale for each research question. 35 Theoretical Rationale Two theories can help guide progress into understanding these issues. Following the logic of the consensus perspective, it could be argued that Black youths violate the norms of society more often than White youths and are thus over-represented in the juvenile courts. For this theory to be substantiated, it must be shown that controlling for norm violating behavior (i.e., offense characteristics and other socially threatening individual or family characteristics), race is not a factor in juvenile court decision-making. The conflict perspective predicts the opposite of the consensus perspective. It would conclude that race and social class will impact decision- making independent of legal and social characteristics because the juvenile court is seen as a tool of those in power to continue to oppress those without power (people of color and poor persons). Not only will the individual's race affect decision-making, but the characteristics of the community environment in which the youth is embedded will also influence case outcome. Traditional conflict theorists (like Turk, or Quinney) believed that sanctions are harsher against less powerful groups because they do not have resources and are thus more vulnerable to justice system intervention. Thus, people with less economic resources are the most vulnerable to social control as are people of color when they are a small minority of the population. In 1987, Hawkins operationalized the conflict perspective differently. He stated that it is not fewer resources which precipitates formal social control, but the threat posed by the less powerful groups to the dominance of the more 36 powerful group. Some have operationalized this power in terms of population. The greater the population of Blacks, the more threat. However, population size does not necessarily imply power; there are large populations of Blacks with little economic power. Thus, some have added a class dimension to the issue of racial group power. For instance, Hawkins' revision has been operationalized by Frazier, Bishop, & Henretta (1992) to mean that the less economic power differential between Blacks and Whites, the greater the justice system intervention. Similarly, Sampson & Laub (1993) used dimensions of racial population and economic indicators to see the effect of these county level variables on formal social control. Methodolpgical Rationale Several methodological issues have hampered this area of inquiry and have left theoretical propositions inadequately tested. Table 2 lists a methodological description of the recent major studies exploring the relationship between race and juvenile court decisions. The table includes number of jurisdictions, sample composition, offense variables, social variables, community variables, and types of analyses. Each of the categories listed in Table 2 will be further explicated below. First, most research has focused on single jurisdictions. The few studies which have looked at more than one site have not explored community level influences on decision-making in a methodologically sound manner. Each study assigned court or county level variables to individuals. For example, an original 37 sample of approximately 9000 cases was actually reduced to 32 because macro-level characteristics were assigned to individual cases (e.g., Frazier & Cochran, 1986). Other studies assigned a designation of urban or suburban to each youth, which actually limited the true sample size to two (e.g., Dannefer & Schutt, 1982). Thus, a major issue is combining individual and structural level data in a methodologically appropriate manner. Second, findings may also have been influenced by the selection of a study sample. Almost all studies have included various types of offenses in their samples and in single analyses. This presents a problem because offenses which may warrant completely different sanctions were included in the same analysis. For instance, when studies include a large proportion of misdemeanor cases that face little risk of detention (e.g., shoplifting), it is not surprising to find that felony cases were more likely to be detained. Third, although many studies have posited the importance of offense seriousness, this variable may have only consisted of a dichotomous designation of felony or misdemeanor charge. There was obviously considerable variation in seriousness within large offense categories. In most studies, offense seriousness was measured by an ordinal variable reflecting more specific Offense categories (e.g., property felony). However, even within these categories there was considerable variation. Such measurement did not consider actual offense behavior. Thus, it may be important to measure the degree of harm or damage done by the offender, whether a weapon was involved, and the number of co-offenders. 38 300.5 0.55 0::0>:.q .28 255 .805 5:355 2:85 .508 .55: 5:80: E 0.20.: .828 .233 :58 t .38 as: £88: 02502 .505: ”:o 30: >55... 05: 0.000. 30:35:00 55:65.0: :0 mm Q 0250 .550...”— 59205 5:580: 5.55850 65.5.0005: €090 0 808 0.505.: 05: 055.. bemoan. 0005:5000. 55:05.0: :0 _ Ago: 0.535: 855.055 55> $55 0038 £8003 A0090 3 0005 2.3: 0:805: 3505.50 05: 25:5 5:8,. 00:50.00 30:055.: 55:55.0: :0 _ 00 5205.0 .5000 0:55 :58 ES: 3 8.8 $3: 55. mo. 55.55:» :0 50:: 0:30.50 55.. 00500.00 30:35.00 55305.0: :0 N 3:50 .0 50050 :5ng 0590 3 0.0000 380 83: 0:500. 05: 05: 00500.8 05:85.0... 55:65.0: :0 05:8 565:: .058... .0 5:50 0855 22.23:... .3 09C. 835...; 55—550 0.0355> .580 003500> 005.00 00000 .8 0.950 be 5:52 Eo5:< 05.5000 :50 0:55... co 0000 So 855 05055.53 6007000: 05.0.05 .moEQEw 5 00020.? 0:: .00_nm_l..w> .moEEmm .o ZmEEsm N 0300. 39 «58 9 80.. .500: .05 5:308 50:00.00: 800000 _.0 .5 0005000.. 0.80500 .530: 5.000: .8555 50.003 3:50 525.005 .508 090 5000 5.. 0000.05 5.00050.— .550 000.808.: 05:00 - 90.5 v. 00000 300.. 0.0030. - 005550, 30:00 0:0: 00.55000 0005:5000 55:05.00 :0 mmm 5:3 0... 5008.5 850% 052. 0: 0008 G00: 0.0.5.05 50.. .8000 .5 08005 5.008 0:0: 0.000 000005.000 55:05.0: :0 . 5.8m .0. 35003. 00.00.500.550 8.00? $20005 0880 2 :5: .0005 2 0008 .000: 5:050 0.0.5.05 500 0:0: 5805005000 5.. .2000 00055.50 55:05.00 :0 a 5 05.050 5:3. 00:50 0.: 5:0» 5.00050.— :0..80500 0:050 00:0...5 0880 00000 800.. 0.0030. .8355 5555 20:00:05 5 $5.». d:0.0.> 00555.00 :0 w 55 5 000.000 .583. 5.00050: 35:00 3500 8 00000 800.. 0.00.5. 5 0.825... 0.5 500 0:0: 0.000 000005.000 55:05.0: :0 m0 .0000m a. 5050.. 5.00052 505. 00000 $00.. .05 0.00.5. .0500 5 5...: 0:00:00 0:0: 50.0.. 5 60.8 55:05.00 :0 mm 55.000 d 00.08“. 000382 0530505.. .5 0%... 00.5555 558800 00.55:; .500m 0035:; 0000000 00000 .5 0.083 .5 038:2 055:... €35.80 N 035 40 The fourth limitation of most prior studies is the dearth of social variables. Some studies have included some measures of the social situation of the youth, but usually the scope is limited to family structure and social class (Fagan et al., 1987; McCarthy and Smith, 1986). The omission of social factors from analyses of decisions may lead to erroneous conclusions because due to the traditional philosophy of the court, social and family issues play an important role in juvenile court decision-making. The fifth major methodological limitation is that prior studies which have examined structural level influences used county level data and did not explore differences within counties. Mean scores for poverty or inequality may not be indicative of the actual situation in the part of the community where the youth lived. Further, research has looked at the structural influences of decision- making and the individual level influences of decision-making, but has not had the appropriate data sources to combine the two in a methodologically sophisticated manner. Finally, the analyses used in most mixed level studies was not appropriate to address the research questions. In all of the prior research which examined both individual and community level variables, individual variables were nested within the county variables. The sampling procedures (many individuals within a few counties) and limited analytic techniques did not allow for the independent estimation of effects at the individual and community level. One way to deal with this problem is to conduct statistical analyses which account for nested designs (i.e., hierarchical linear models). Another way is to 41 create a sampling procedure which insures that the design is not nested (i.e., sample one juvenile per county or community). With the theoretical and empirical rationale summarized above, four general research questions were developed. Within each research question, two or more hypotheses were generated. Specific theoretical and methodological support for each question is presented below. Also, improvements to prior research are listed. Research Question 1: Was race a factor in juvenile court decisions for youths who were similarly charged? Theoretical Suggort. Most adherents to the consensus perspective would contend that controlling for type of offense, race would not be a factor in decisions. The conflict perspective, would obviously support just the opposite notion. Thus, for youths who are similarly situated legally, the conflict perspective would predict that minority youths would receive more severe sanctions than White youths. Methodolgical Support. While the majority of studies have supported the notion of differential treatment by race, the studies can be criticized for not having enough methodological rigor to control adequately for offense behavior. For instance, Fagan, Slaughter & Hartstone (1987) measured specific offense behavior (degree of harm, weapon usage), but did only crosstab type analysis and could not control for various factors at once. Other studies used all types of offenses together. A mix of felonies and misdemeanors would likely produce strong effects of offense (i.e., felons will be more likely to be detained). 42 Moreover, studies rarely measured offense behavior directly. Only one of the studies reviewed, Kurtz, Giddings, & Sutphen (1993) measured offense behavior directly and used these data in multivariate analysis. Unfortunately that study had a very small sample size at disposition (89). Improvements. The present study used more detailed measures of offense behavior than in previous research. First, the analysis was confined to felonies. Second, specific behaviors were measured including: amount of victim injury, whether some kind of weapon or a gun was used, number of co- offenders, and whether or not it was a drug charge. Also, other legal factors were controlled. These include: number of priors, whether or not there was a concurrent (always less serious) offense, and legal status at time of offense. Further, detention status was controlled when predicting disposition because earlier court decisions may influence the later decisions. The specific hypotheses stemming from this rationale are listed below. 1a. Blacks and Latinos will be more likely than Whites to be detained after accounting for number of prior offenses and current offense behavior (victim injury, weapon use, number of co-offenders). 1b. Blacks and Latinos will receive more severe dispositions than Whites when accounting for detention status and number of prior offenses and current offense behavior (victim injury, weapon use, number of co-offenders). 43 Research uestion 2: What influence did social characteristics have on court decisions and how was the independent effect of rage impacted by social factors? Theoretical support. Some might argue that if race effects are found in the juvenile court, it is only because judges and referees take into account social factors that may be related to race. Due to the treatment focus of the court, some proponents of consensus theory argue that the breaking of social norms (e.g., dysfunctional families, school problems) may contribute to differential treatment (Tittle, 1994). Consensus theory would content that it is not actually racial bias that causes these effects, but differential treatment based on family or personal issues. Given this argument, it is thus important to control for social variables, some of which may be related to race. In fact, Fenwick (1982) found that family disaffiliation was the strongest predictor of detention. The disaffiliation variable was a composite of items consisting of family structure and functioning. Methodological suppprt. Previous research has either not measured family and social variables, or has not measured or analyzed them adequately. For instance, Dannefer & Schutt (1982) only measured family structure and found no effect on disposition. Fagan, Slaughter & Hartstone (1987) measured family variables, but did not conduct multivariate analysis. Improvements. This study is an addition to the literature in that it measured several social variables including: family structure, family problems, school problems, and personal problems. These variables were composites of 44 many items regarding abuse, neglect, substance abuse, school performance and attendance, etc. Table 5 in the methods section provides the items. The hypotheses tested conflict theory, which would predict that Black and Latino youths would be more likely to receive harsher sanctions than White youths after controlling for legal and social variables. 2a. Blacks and Latinos will be more likely than Whites to be detained after accounting for legal factors and social characteristics (family problems, school problems, emotional problems, family structure). 2b. Blacks and Latinos will receive more severe dispositions than Whites after accounting for legal factors and social characteristics (family problems, school problems, emotional problems, family structure). Research Question 3: What were the influences of commpnitv cmaeifijpp on court decisions: Theoretical support. The consensus perspective would predict that community characteristics would not influence decision-making. The conflict perspective, however, suggests that contextual factors influence decisions. The specific contextual factors center around power of the group. Competing hypotheses from conflict theory can be tested. Traditional conflict theory would predict that the fewer number of Blacks in the community, the more vulnerable they would be to formal social control (less power in small numbers). Traditional theory would also predict that the fewer economic resources in the community, the more severe the sanctions (due to greater vulnerability to formal 45 social control). Hawkins' revision of this theory would predict that the higher the population of Blacks in a neighborhood, the more formal social control (because large numbers are more threatening). Similarly, the revision of conflict theory would predict the more economic power of racial minorities, the more severe the sanctions (i.e., more threat more sanctions). The revisionist theory regarding economic differentials could not be directly tested in this study because census tract data is aggregated and was not categorized by race/ethnicity and income together. Methodological support. There are no studies which have looked at specific communities or census tracts. All studies which have examined contextual effects have used county level data. One reason for using county level data is it is relatively easy to collect. Community level or census tract level data are much more difficult to obtain. Another reason researchers have used county level data is that their focus was often on examining differences among courts (which are situated in individual counties). The few studies which measured contextual effects in counties did so in different ways and produced mixed findings. Frazier, Bishop, and Henretta (1992) found that the greater the percent White in the population, the more severe the dispositions of Blacks compared to Whites. This supported traditional conflict theory. Sampson and Laub (1993) found to the contrary, however, that the higher the level of underclass poverty (a six variable composite) the greater the likelihood of detention. The underclass poverty scale included a dimension of the percent Black in the population. 46 Thus, there was a mix of findings regarding the influence of contextual factors on decisions. One of the main reasons for this problem was methodological inadequacies to answer the question. First, there may be large differences within one county. Segregated neighborhoods with a high population of Whites may be situated close to areas with a large population of Blacks. Second, most of the studies looking at race disparity, assigned contextual variables to individuals. Thus, the sample size was restricted by the unit of analysis. Improvements. Two main additions to the literature on community influences on juvenile court processing were integral to this study. First, data at the level of the census tract was used account for the variability of community characteristics across county jurisdictions. Second, the problem of mixing levels of analysis was dealt with in a unique manner. Only one youth was sampled per census tract. Thus, there was no mixing of levels of analysis. The sampling procedure will be discussed in detail in the methods section. The hypotheses listed below test traditional conflict theory. However, there is little empirical support for this conceptualization. 3a. Youths from poorer communities will be more likely to be detained. 3b. Youths from communities with higher population of Whites will be more likely to be detained. 3c. Youths from poorer neighborhoods will be more likely to receive more severe dispositions. 47 3d. Youths from neighborhoods with higher population of Whites will be more likely to receive more severe dispositions. Research uestion 4: Were factors which ex lained court decisions differentially weighted for Black youths and White youths? Theoretical support. The consensus perspective would support the notion that legal and social factors are equally predictive of detention and disposition for both Blacks and Whites. However, the conflict perspective may suppose that these factors may be better predictors of court decisions for White youths than Black youths. Enforcing the “collective good“ may be predictive of how justice is meted out for White youths because the reliance is on measurable factors. Thus, conflict theory might predict that legal and social factors would be given more weight for White youths than Black youths because decisions for Black youths are based on race and not as much on other factors. Methodrological Support. Contextual variables were examined separately for Black and White youths by Sampson & Laub (1993). in this macro level study, they ran separate logistic regression equations for Black youths and White youths. They found that contextual variables were better predictors of court decisions for Blacks than for Whites. Detention of juveniles was predicted by the level of underclass poverty in the county for Black youths, but not for White youths. They also found that wealthy counties detained more Black juveniles, but county wealth was unrelated to detention for Whites. The finding that county level variables of underclass poverty and county wealth were not 48 highly correlated points to the variability within counties. Thus, many questions regarding the impact of community were left unresolved. Improvements. Sampson & Laub's (1993) study was the only one that examined how relevant variables influenced decisions differentially for Blacks and Whites separately. Their study, however, was at a macro-level, wherein the present study was at an individual level with contextual effects. Thus, this study controlled for individual level issues (offense behavior and social characteristics) while still assessing community effects on decisions. This new approach to exploring racial bias (separate multivariate equations for Blacks and Whites), along with an examination of interaction effects, could improve our current understanding of the factors important in decision-making and how factors may be differentially weighted for each racial group. These hypotheses were theoretically, not empirically derived. Prior research has not used individual level data to multivariately test racial groups separately and explore interaction effects. 4a. Legal, demographic, social, and community variables will be more predictive of detention for White youths than for Black youths. 4b. Legal, demographic, social, and community variables will be more predictive of disposition for White youths than for Black youths. 4c. Legal variables will interact with race/ethnicity to produce a greater likelihood of detention for Black youths. 4d. Social variables will interact with race/ethnicity to produce a greater likelihood of detention for Black youths. 49. 41‘. 49. 4h. 49 Community variables will interact with race/ethnicity to produce a greater likelihood of detention for Black youths. Legal variables will interact with race/ethnicity to produce a more severe disposition for Black youths. Social variables will interact with race/ethnicity to produce a more severe disposition for Black youths. Community variables will interact with race/ethnicity to produce a more severe disposition for Black youths. Method Design and Data Sources These data were collected as part of a larger study on disproportionate representation in the state of Michigan (Bynum. Wordes & Corley, 1993). The larger study was funded by the Office of Juvenile Justice and Delinquency Prevention through the Michigan Committee on Juvenile Justice. The data used for this study were collected from case files in five juvenile courts and from 1990 and 1991 census data. Letting The selection of counties began with a survey of the juvenile court in each county in the state to determine the type of records kept, the data system used, and willingness to cooperate in the study. Over the telephone, 47 counties agreed to participate, 16 declined, and the remainder were undecided. Of the 83 counties in Michigan, 19 juvenile courts used some form of the State Court Administrative Office computerized data system. A few counties used their own computer systems, and most did not have individual level data computerized. Also important in site selection was the census data on population characteristics of each county. It was necessary to obtain data from sites that had sufficient numbers of racial minorities, varied in size, and varied in geographic representation. With input from the consultation group of juvenile justice professionalsz, seven counties were selected to participate. The consultation group chose these counties as a representation of the diversity in courts and counties in the 50 51 state of Michigan. They were representative in that they differed in size of the community population, size of the caseload, community composition (including racial population breakdown and socio-economic status), geography, and in juvenile court philosophy. Five counties were chosen for inclusion in this study, due to the very small number of felony cases in the two small rural counties. §a_rr_12|i_ng Since both individual and community level variables were implicit in this research design, the sampling procedure was somewhat complicated. Wherein previous studies merely assigned county level data to individuals, this study did not mix levels of analysis in that manner. Due to the fact that it was not possible to base an original sample on census tract (because computerized addresses of the court population were mostly unavailable because addresses were kept only in written records), individual data were gathered first to determine the community in which the youths resided. Individuals were then randomly sampled within each census tract when more than one youth resided in a census tract. The detailed sampling procedure is documented below. Defining the Popplation The original study population were all youths referred to the five juvenile courts in 1990 for delinquency. Although one youth may have been involved in more than one case during the year, the unit of analysis was the individual. Thus, the most serious offense charged with in 1990 was the offense behavior specifically coded. In order to best predict the court decision, it was assumed 52 that decisions would more likely be based on the most serious charge. Thus, other offenses were coded as concurrent offenses. In order to obtain the population from which to sample several methods were used. In some courts the data were computerized and in others it was kept in written log books. In the three sites with computerized data, determining the population was relatively easy. There was one record in the database for each youth. In the two sites without computerized record keeping systems, defining the population was more difficult and the process is delineated below. In one court, the cases which were dropped or handled formally were written into a log book. This book contained a record of each youth entering the court who was not assigned to a diversion program. It contained, name, race, gender, and offense. These data were entered into a computerized database. To ensure that the diverted population was sampled, records from the agency who handled all diversion cases were used. Paper files for 1990 were coded for name, race, gender, and offense and input into the database. In the other non-computerized site, the record keeping system consisted of files organized by name. They could not access by year to determine the population in 1990. Thus, the population was determined from log books kept by the prosecutor's office and the intake coordinator at the court. Fortunately, in this county the police automatically referred cases to the prosecutor, rather than the court, so the prosecutor had a record of all cases treated formally. The log books only contained the youths names and offenses, race and gender 53 were unavailable on the population. The population of names and offenses from the prosecutor's office was input into a computerized database. In order to include the cases handled informally by the court, a second method of obtaining the population was used. Private records of the intake coordinator at the court were collected. He handled approximately 85% of the cases not referred from the prosecutor‘s office. The other 15% were handled by another intake worker who did not keep a log book of cases. The intake coordinators log book only contained dates of official contact with the court and names. Knowing the names of these cases, the paper files were pulled from the file drawers. The most serious offense for that youth in 1990 was then coded. These data were then added to the prosecutors data to represent the population (less the 15% of informal cases handled by the other intake worker). In total, there were 10,264 cases referred to the five courts in 1990. Of these cases, about one-half (5,088) were referred for felony offenses. Out of the 5,088 youths in the felony offender population, 1,690 were eventually coded. Samplinlof Individua_ls from the Pooulation Prior to describing the process of obtaining the sample of 1,690, the intention of the original sampling procedure of individual juveniles must be understood. The original sampling procedure included youths who were charged with all offense types. The goal of the original sampling procedure was to: a) accurately represent the population in order to document disproportionate representation and b) to allow for comparisons across offense categories, racial 54 categories, gender categories, and final dispostions. In order to accomplish the second goal, youths were stratified on offense, race, gender, and disposition in order to ensure adequate representation of the smaller groups. Without this process, a pure random sample might not represent small groups. For instance, a random sample might produce many White male misdemeanants (because they comprise a large portion of cases in most courts), but very few White female felons (because they are typically a small proportion of court cases) While the intent of the sampling procedure was clear, actually determining a sampling strategy was a difficult task. The common demominator among population data was that every court was able to provide a list of youths who were referred to the court in 1990 along with the specific offense for which they were charged. These specific offenses were then coded into the general categories of felony, misdemeanor, and status offenses. Only felony cases were examined in this particular study. Full stratification of the population was only possible in those sites which could provide the necessary data elements on each youth in the population. Therefore, the population at each court was stratified by race, gender, offense type, and disposition (when available). Cases were then sampled from each cell. If the population in the cell consisted of less than 25 youths, all cases were sampled. If the population was greater than 25, a random sample was taken from the cell. 55 To best describe the somewhat complicated sampling process, each site will be referred to separately. Tables in the appendix provide the sampling frame and sampling weights for each of the participating juvenile courts for the felony offense category. The sampling weights should be interpreted as the multiplicative factor for the sample in order to reach the population number. In other words, when the sampling weight was 1.0, every youth in that cell was sampled. If the sampling weight was 3.0, a random sample of one-third of the youths were sampled. The higher the sampling weight, the higher the number of youths in population within that cell. The populations in two courts were stratified by race, and gender. In two other courts more data elements were available in their computerized record system and allowed for stratification on race, gender, and type of disposition. The population cases in one court was not stratified due to the problems in their record keeping system. Sampling of Individuals Within Unigue Census Tracts There were 1690 felony cases originally coded. After removing youths who were previously State wards (because there was no dispositional decision except to remand them back to the State), omitting cases with a race designation of ”other", and removing any cases which had missing data on address, demographic, prior offense, or current offense behavior variables, the sample size was 1060. In this sample of 1060, more than one youth lived in some of the census tracts. Almost one-half (48%) of the census tracts had only one case. Forty-three percent of the census tracts had 2, 3, or 4 cases in 56 them. Only 9% of the census tracts had more than 5 cases. If a census tract had more than one case in it, one case was randomly sampled. There were 455 unique census tracts. A similar sampling procedure was used to extract the appropriate sample for examining formal hearing dispositions with social and family data. From the same original sample of 1690, 551 cases had complete demographic, prior record, offense behavior, and social variables. Out of the 551 social cases, 305 lived in unique census tracts. One case was randomly sampled from each census tract with more than one youth. The final samples used for the analyses were a complete intake data set with unique census tracts (n=455) and a complete social data set with unique census tracts (n=305). Table 3 is a descriptive representation of each of the samples. It demonstrates that there were few differences between the original sample and the derived samples except in the expected directions. These differences were tested for statistical significance. In general, the percentage of cases in the samples were of the same race or ethnic background, the same age, the same number of prior offenses on record, and having similar offense type profiles (none of these differences were statistically significant). The statistical differences in the samples were expected. For instance, more of the cases were from court 1 in the samples with individual census tract cases because that county contained a larger number of census tracts than other counties. Similarly, the data of social cases contained youths who were more likely to have a formal hearing and were more likely to be detained. This was 57 Table 3 Sam le haracteristics fr m ri inal S m Ie and Derived am I s with Non-missin ta in Unigue Census Tracts Actual Sample Used in Analyses original sample all cases with all cases with cases with (contains non- missing non- missing non-missing missing data) data‘ data from social data distinct census from distinct tracts census tracts Number of 1690 1060 455 305 cases Percent cases from each court court 1 46% 45% 62% 52% court 2 22% 19% 15% 18% court 3 10% 11% 10% 14% court 4 12% 14% 9% 11% court 5 10% 11% 6% 6% Percent of race/ethnic group White 40% 37% 45% 45% Black 55% 58% 51% 50% Latino 4% 6% 4% 6% Age (mean) 15.07 14.99 15.14 15.23 Prior offenses (mean) 1.33 1.11 .90 1.31 Type of felony property 61% 58% 56% 53% drug 7% 8% 10% 9% personal 27% 28% 28% 29% misc. 5% 7% 7% 9% Percent having formal hearing 75% 80% 84% 96% Percent in detention 33% 35% 39% 47% ‘ Contains complete data on demographics, prior offenses, offense behavior, and census tract. Social data may be incomplete. 58 due to the fact that social data were collected and stored in the files on most youths who had formal hearings and not usually on youths who did not have formal hearings. One concern with this sampling procedure was to be aware of the possible limitations in the variance in offense behavior due to the reduction in sample size. Given that felony offenses contained a large range of offense behavior (e.g., no physical harm in an assault to great bodily harm or death), it is important to not restrict variance. Table 4 contains descriptive statistics for the original data sets and the derived samples in order to demonstrate that the variance in offense behavior was not overly restricted in the samples used in the analyses. In fact, there were virtually no differences in the variance of the items comprising offense behavior among the samples. However, as expected, youths with social data (those who had a formal hearing) had higher mean scores for amount of victim injury. Table 4 Means and Standard Deviations (in parenthe§ipi of Offense Variables from Original Sample and Derived Samples original sample all cases with Actual Sample Used in Analyses cases with non- cases with non- (N=1690) non-missing missing data from missing social data (n=1060) distinct census data from tracts (n=455) distinct census tracts (n=305) Amount of .47 .49 .47 .56 victim injury (1.10) (1.11) (1.10) (1.20) Weapon used .46 .40 .37 .42 (.70) (.66) (.66) (.69) Number of co- 1.21 1.22 1.17 1.23 offenders (1 .45) (1.49) (1 .56) (1.59) 59 Power Analysis To determine if the sample size was adequate to detect effects if they were present, a power analysis was conducted. The significance criterion of .05 was chosen and the effect size was estimated at .15. Using the power analysis equations supplied by Cohen and Cohen (1983), the power for the equation with the most independent variables (16) and the smallest sample size (136) was .85. The power for all the other equations was obviously higher as the sample size was bigger and the number of variables was often reduced. Given these estimates, the sample sizes in this study should have provided the necessary power to detect effects if they were present. Procedures Three teams of case file coders coded data from the five sites. Case file coders were undergraduates or recent graduates from Michigan universities. Each team of case file coders received two days of on-site training. The training consisted of a staff member intemal to the court explaining the processing of cases and their record keeping procedures followed by detailed coding instructions from the research staff. The on-site coders were supervised on a weekly basis. To further ensure accuracy of coding, each data collection form was reviewed by two independent persons. The forms were checked for logical consistency and missing data. If it was questionable that the data might truly not be missing or the data did not follow in a logically consistent manner, the form was returned to the site to be re-coded. 60 A random sample of five percent of the cases was coded twice to assess inter-rater reliability. In each site the percent agreement was slightly different. These differences were most likely due to the record keeping practices of the specific court. Some court files were easier to interpret than others. The range of percent agreement between coders was a low of 83% to a high of 93%. Across sites, the average inter-rater agreement was 88%. Measures Communitv I_D_at.§ The community level variables were derived from the MAPINFO program containing Michigan Census tract data for 1990 and 1991. The MAPINFO computer software allowed addresses of juveniles to be geo-coded into the appropriate census tract. This software links the address to the population demographics of each census tract. Community variables are listed in Table 5 and include racial residential segregation or percent Black in the population, and meaflousehold income (both were converted to standard scores). Case File Data Colleption Forms Case files were coded on court data collection forms. The forms were developed by reviewing actual case files from several sites and noting all the information that was available. The data collection forms were designed to reflect the most comprehensive files. The court records contained a variety of data depending on the specific record keeping practices of the site and the level of formality of the case. Some of the court records contained several types of information including: 61 police report, petition, social history (e.g., family background, school reports, psychological tests), and adjudication and disposition records. Defining the decision and disposition variables was a difficult task, given the differences in terminology across courts. To circumvent terminology problems, standardized definitions for the troublesome terms were developed. A consent disposition was different across counties. The definition of “consent“ adopted for this study was: a formal adjudication with a review of the case within a short specified time period (usually 30 to 60 days) at which time, if the youth followed a set of informal rules, there was no sanction associated with the adjudication. In some courts, the consent disposition was decided at an informal hearing, however, the case was still adjudicated, thus defined as “consent“. Another term that varied across sites was ”local ward“. A local ward was defined as a youth who was adjudicated a ward of the county court. Generally, local wards were placed on probation, but in some counties there was no sanction attached. For example, a disposition of “warn and dismiss“ in one county actually fulfilled the definition of local ward. In another county "warn and dismiss“ meant the case was dropped. Thus, for the purposes of this study, the “warn and dismiss" in one county was coded “local ward“ and the same term in another county was coded "dropped”. This coding procedure allowed for comparison across counties on severity of disposition. The variables gleaned from the court data collection forms were demographic and prior record, offense characteristics, social characteristics, 62 and detention and disposition decisions. Each variable, its definition, and source is listed in Table 5. Qapjaion Variables. Two decision variables were used in these analyses corresponding to two levels of court intervention, the court intake detention and severing of disposition. These two decision variables chosen represent two important decisions (the initial detention decision at intake, and the final disposition on the case). The court irit_a_ke datention decision is usually made by an intake worker at the court who is first contacted by a police officer via telephone to lodge the youth. If accepted, the officer would transport the youth to the detention facility. The second decision variable is a compilation of both the adjudication and disposition decision and will be referred to as severity of disposition. Thus if a youth was dropped or diverted, she or he was not adjudicated a delinquent. If the youth received consent probation, regular probation, or intensive probation, the youth was adjudicated delinquent. Also, if the youth received the most restrictive disposition, commitment to the State Department of Social Services, the youth was adjudicated delinquent. The disposition decision examined in this study was made in a formal hearing by a judge or referee. The formal decision was examined because social data were not recorded at intake and thus could not be accounted for. Dempgraphic Variables. Three individual level demographic variables were used. Race/ethnicity was coded as either Black, White, or Latino. Due to the small numbers of cases belonging to other racial/ethnic categories, these groups were omitted from the analyses. Age was used as a control variable 63 because it may affect sanctioning or treatment options. Aga was coded as a continuous variable. am was also used as a control variable because it was also thought to influence juvenile court decisions (Chesney-Lind, 1973). Legal Variables. Most prior empirical research had not been able to control for specific offense behavior. In this study, only felony cases were examined to avoid mixing completely different types of offenses into one analysis. The specific offense variables included whether of not there was a weapon involved, the number of co-offenders, and an ordinal variable measuring physical harm to the victim or victim injury. The amount of property damaged or stolen was not used as it had a strong negative relationship with victim injury. The other legal variables assessed were the juvenile's prior offenses, legal status, and concurrent chjargpa. Prior offenses refers to the number of previous court referrals and was coded dichotomously into none or one or more due to a highly negatively skewed distribution of the ratio scaled item. Lgal atagg measured whether at the time of the current offense, the court had disposed of a prior charge at all, in an informal manner, or placed the youth on probation. Concurrent chapgg (or having another pending charge) was also dichotomously coded into zero or one or more for the same reasons. The concurrent charge was always less serious than the current offense coded due to the nature of the sampling strategy. Table 5 64 Definition of Variablea Used In All Analysas Variable Decision Variables Court Intake Detention Severity of Disposition Demographic Variables Age Gender Race/ Ethnicity Legal Variables Prior Offenses Legal Status Concurrent Offenses Victim Injury Weapon Involved Co-offenders Community Variables % Black in Population Mean Household Income Definition held in secure confinement at intake into court (0=no, 1 =yes) (1 = dropped, 2 = diversion, 3 = consent, 4 = local ward, 5 = state ward) (date of incident - date of birth) (1 =female, 2 =male) (1 =Latino, 2 = Black, 3 =White) number of prior offenses charged in court records (0, 1 or more) (1 =none, 2=informal, 3=formallon probation) number of charges on petition (0, or 1 or more) (1 = none, 2 = minor, 3 = medical treatment, 4 = hospitalization, 5 = fatal) I0=none, 1 =weaponlnot gun), 2 =gun) I0, 1, 2 or more) number of Blacks/total population in census tract mean household income of census tract 65 Table 5 (continued) Variable Definition Social Variables Household Provider (1 =other, 2 =mother only, 3=two parents) Family Problems mean scale score of: child physical abuse (0=no, 1 =yes) child sexual abuse (0=no, 1 =yesl child neglect (0=no, 1 =yes) domestic violence (0=no, 1 =yes) substance abuse by father, mother, or siblings (0=no, 1 =yes) criminal record of father, mother, or siblings (0=no, 1 =yes) School Problems mean 2 score of: attendance (1 =almost always attend, 2 =often, 3 =sometimes, 4 = almost never attend) academic performance I 1 =good, 2 =fair, 3=bad) behavior problems (0=no, 1 =yes) learning disabilities (0=no, 1 =yes) suspensions (0=no, 1 =yes) violence against students or teachers (0=no, 1 =yes) Personal Problems mean scale score of: youth substance abuse (0=no, 1 =yes) psychiatric diagnosis (0=no, 1 =yes) treatment for emotional problems (0=no, 1 =yes) Social Variables. Data on several social characteristics were collected from the pre-sentence investigations done by the probation officers at the court. Household provider was coded as two parents, mother only, or other. A scale assessing the amount of family problems was also compiled. The variables in the scale were all dichotomously coded (either yes or no) and included: recorded history of child physical abuse, child sexual abuse, domestic violence, child neglect, substance abuse of a family member (separate items for father, mother, and siblings), and criminal history of a family member (separate items 66 for father, mother, and siblings). The scale measuring school problems contained six items: an ordinal measure of academic performance, an ordinal scale of school attendance, the presence or absence of known Ieaming disabilities and behavior problems in school, and dichotomous measures of violence at school and whether or not the youth was suspended from school. The personal problems scale consisted of three dichotomously coded items: youth substance abuse, psychiatric treatment, and history of emotional problems. The internal consistency of the scales computed for social variables is presented in Table 6. The three item scale of firsonal problems had an alpha of .62. The item total correlation ranged from .27 to .52. The second scale, school problems, had six items. The Cronbach's alpha was .72 with corrected item-total correlations ranging from .42 to .58. The final social scale constructed, was called family problems. This scale consisted of ten items and had an alpha of .70. The item total correlations ranged from .27 to .49. Coding social data from pre-sentence investigations and social case files was a difficult task. The records were not uniform and thus there was much missing data. To be able to measure social constructs with this missing data problem, scale scores were computed if one half of the items were present in the data set. Scales scores were thus computed as mean scores on one-half or more of the items. 67 Table 6 Items Item Total Correlations, and Alpha's of Scales Measuring Social We Corrected Item-Total Scale and Item Correlation Alpha Peraonal Prfialefima .62 Psychiatric Diagnosis .50 Treatment for Emotional Problems .52 Youth Substance Abuse .27 School Problema .72 Academic Performance .47 Learning Disabilities .21 Behavior Problems .58 Violence Toward Others .52 Suspensions .54 Attendance .42 Familv Problems .70 Mother's Substance Abuse .49 Father‘s Substance Abuse .49 Sibling's Substance Abuse .26 Sibling's Criminal Record .22 Father's Criminal Record .45 Mother's Criminal Record .41 Physical Abuse .40 Sexual Abuse .27 Domestic Violence .35 Child Neglect .29 68 Description of Sample Cases by Race/Ethnic'mr Prior to presenting results, a description of the cases in the sample for the predictor variables is warranted. Table 7 presents a description of youths in the intake sample by race/ethnicity. Even in trying to oversample Latino youths, there were relatively few in the sample. Due to the fact that Latinos make up only 4% of the sample, these percentages should be interpreted with caution. Black youths comprised approximately one-half the sample. The majority of the youths were male (72%) and only one-quarter were 14 years of age or younger. The distribution of gender across racial group varied. Females were most likely to be represented in the Black racial group (31%) whereas the vast majority of Whites were male (81%). The ages of the youths across racial categories were similar and varied only slightly. Table 7 Demographic Daacnption bv Racfie/Ethyiicitv of Intake Sample (n=455) Race/Ethnicity Latino Black White Row Total 19 (4%) 232 (51%) 204 (45%) Gender Female 0% 31 % 19% 129 (28%) Male 100% 61% 81% 326 (72%) App Under 13 11% 4% 10% 32 (7%) 13 5% 8% 7% 34 (8%) 14 11% 13% 12% 57 (13%) 15 32% 23% 29% 1 18 (26%) 16 21% 30% 24% 123 (27%) 1 7 21 °/o 21 °/o 1 9% 91 (20%) 69 Table 8 shows that the majority of youths in the sample had no official legal status and no prior offenses in the court records. The legal status of the youths across racial groups did not vary and White youth were only slightly more likely to have no prior offenses on record (73% for Whites, 65% for Blacks). Similarly, Blacks and Whites varied little in the percentage of cases which had multiple or concurrent charges. However, Latino youths were more likely to have more than one charge. Blacks in the sample were more likely than other groups to be charged with a drug offense. Also, a greater proportion of Black youths had a weapon and committed the offense alone. Table 9 describes the community characteristics of youths in the intake sample. The mean household income in the census tract was approximately $12,500 less for Black youths than White youths. This table also indicates that Black youths lived in predominantly Black census tracts (80% Black) and White youths lived in census tracts with few Blacks (7% Black). Tables 10,11, and 12 present a description of the social data sample. The descriptive comparisons among racial groups in demographics, legal characteristics, and community composition are similar to the intake data description. Table 12 presents the social data from the sample. Forty-five percent of White youths lived with two parents, whereas 20% of Black youths and 22% of Latino youths lived with two parents. The mean number of family problems was lowest for Latino youths and highest for White youths. Personal problems were also highest for White youths. The mean number of school Table 8 70 Description of Legal Characteristics by Race/Ethnicity of Intake Sample (n=455) Race/Ethnicgy' Latino Black White Row Total 19 (4%) 232 (51 %) 204 (45%) Legal Status No Official 84% 84% 83% 380 (84%) Informal 5% 6% 7% 30 (7%) Probation 11% 10% 9% 45 (10%) Brigg None 63% 65% 73% 310 (68%) 1 or more 37% 35% 28% 145 (32%) Concurrents None 47% 59% 63% 274 (60%) 1 or more 53% 41% 37% 181 (40%) Drug Offense Drug Felony 5% 16% 3% 44 (10%) Victim Injugy None 79% 78% 87% 374 (82%) Minor 0% 2% 3% 11 (2%) Moderate 16% 7% 4% 28 (6%) Medical Treatmt 0% 7% 4% 26 (6%) Serious 5% 4% 1% 12 (3%) Fatal 0% 0.4% 0.4% 4 (1%) Weapon None 79% 67% 78% 330 (73%) Knife or other 11% 19% 16% 80 (18%) Gun 11% 14% 5% 45 (10%) Co-offenders None 11% 46% 35% 180 (40%) 1 53% 28% 33% 144 (32%) 2 or more 37% 25% 32% 131 (29%) 71 Table 9 Description pf Communig Characteristipa by Race/Ethnicm of lntaka Sample (n=455) Race/Ethnicig Latino Black White 1 9 (4%) 232 (51%) 204 (45%) W house income 28,247 (9,558) 24,520 (8,532) 36,917 (13,209) (mean, s.d.) % Black in pop. .11 (.22) .80 (.26) .07 (.15) (mean, s.d.) Table 10 Demographic Description by Race/Ethnicity of Spcial Data Sample (n=305) W Latino Black White Row Total 18 (6%) 151 (50%) 136 (45%) 939$. Female 0% 25% 13% 55 (18%) Male 100% 76% 87% 250 (82%) Aga Under 13 6% 1% 4% 8 (3%) 13 1 1% 9% 6% 23 (8%) 14 22% 15% 13% 44 (14%) 15 22% 30% 32% 93 (31 %) 16 28% 27% 28% 83 (27%) 17 11% 19% 17% 54 (18%) 72 Table 11 Description of Legal Characteristics by Race/Ethniciy from Social Data Sample (n=305) Race/Ethnicig Latino Black White Row Total 19 (4%) 232 (51%) 204 (45%) Legal Status No Official 83% 77% 78% 237 (78%) Informal 6% 12% 7% 29 (10%) Probation 11% 11% 15% 39 (13%) m None 56% 58% 67% 188 (62%) 1 or more 44% 42% 33% 117 (38%) Concurrents None 33% 50% 54% 155 (51%) 1 or more 67% 50% 46% 150 (49%) Drug Offense Drug Felony 0% 18% 2% 28 (9%) Victim lnjug None 61 % 73% 88% 240 (79%) Minor 0% 4% 5% 13 (4%) Moderate 17% 7% 4% 18 (6%) Medical Treatmt 6% 10% 2% 19 (6%) Serious 6% 5% 2% 11 (4%) Fatal 11% 1% 0% 4 (1%) Weapon None 50% 64% 79% 213 (70%) Knife or other 11% 24% 13% 56 (18%) Gun 39% 13% 7% 36 (12%) Co-offendera None 11% 41% 32% 108 (35%) 1 50% 31% 34% 101 (33%) 2 or more 39% 29% 34% 96 (32%) 73 Table 12 Description of Social and Commpnity Characteristics by Race/Ethnic'gy of Social Data Sample (n=305) Race/Ethnicay' ' Latino Black White Row Total 18 (6%) 151 (50%) 136 (45%) Caregiver Two Parents 22% 20% 45% 95 (31%) Mother 67% 60% 39% 156 (51%) Other 11% 20% 16% 54 (18%) Social Factors Family Problems -.13 (.36) -.03 (.43) .00 (.57) (mean, s.d.) School Problems -.06 (.55) .02 (.66) -.15 (.65) (mean, s.d.) Personal Problems -.16 (.68) -.20 (.66) .14 (.81) (mean, s.d.) Community Factors household income 25,906 (9,329) 23,818 (7,925) 36,409 (13,446) (mean, s.d.) % black in pop. .28 (.35) .79 (.27) .07 (.15) (mean, s.d.) problems were lowest for White youths and highest for Black youths. When reviewing these tables it is important to remember that these tables were intended only as descriptors of the sample and not as representations of the population in the courts. Results Data Analysis Strategy Data analysis was conducted on two samples: an intake sample and a social data sample. The intake sample consisted of 455 youths from unique census tracts. There were 305 youths in the second sample within unique census tracts who had complete social data (social data were collected on youths who had a formal hearing). Of course these samples were not independent because they represented two stages in the juvenile court process. Each hypothesis was examined using either logistic regression or ordinary least squares linear regression techniques depending on the type of dependent variable. All significance tests were at the p<.05 level. The dependent variable in the equations predicting detention was dichotomously coded (not detained, detained) and required the use of logistic regression techniques. For the equations predicting the severity of disposition, linear regression was used because the dependent variable was an ordered variable with five levels of sanctioning. Each of the first three research questions examined the notion that race had an influence on decision-making after accounting for other known influences. To test whether race had an independent effect on detention and disposition decisions, hierarchical logistic regression and hierarchical linear regression techniques were used. In the first block, legal factors were entered into the equation, the second block contained demographic variables (not race), the last block contained race/ethnicity. In the equations examining the influence 74 75 of social and community factors, these sets of variables were entered in a third and fourth block before entering race/ethnicity. The fourth research question addressed Blacks and Whites separately and was examined using separate, yet identical, equations. Subsequently, interaction effects between race/ethnicity and each predictor were tested in separate equations. If significant interaction effects were found, analysis of variance allowed for an estimation of cell means while holding other main effects constant. The cells means for the significant interaction effects were then plotted. Bivariate Relationships Among Variablea Prior to testing specific hypotheses, it was first necessary to examine the bivariate relationships among independent variables as well as the bivariate relationship of the independent variables with the two dependent variables. There are three main reasons for presenting the zero-order correlations are: a) to inspect the relationships for multicollinearity, b) to determine which predictor variables were related to race/ethnicity, and c) to understand the bivariate relationships related to the outcome. Zero-grder Correlationa Table 13 presents the bivariate correlations of the variables used in the intake sample (n=455). Out of the 91 possible unique relationships among the 14 variables, only four had a correlation coefficient over .40. As would be expected legal status at time of incident was related to the number of priors (i.e., a youths with priors was more likely to have formal legal status with the 76 00. P c0539.... . n F. 00. P c3582. 00. .3”.- 00. P 2.55 00.- . mu: .mv. 00. F oEooE E. No.- .. mm. 1.0.. .3: 00. p goo... * 00. 00. .0 P . . N F. . Na: 00. F Becca mo. . N p. 00.. m0: . 00. No. 00. P can 00.- .N P: 00. no. .0 pr .5 F. .0. 00.. «Lorenzo-co 00.- 00. . m P: 50.- . 00. 0%. N0.- 00: 00. — >53 8.83 no... .0N. .3 p .- 00.- . h ... .0. no. 00.- . Na. 00; canoes No. .3. .m..- .m..- .2. mo.- .9. mo: .3.- .«7 8.. 8.2.0 9% m0. .0 P. 00.- po. 00.- no. .N F. «0.- m0. 00. No. 00. p 825260 . K. 00. 5.- N0: 5.- 00. . P p. "0.. v0.- «0.- #0.. ho. 00. P «38m .30. .mu. .3. 00.- .2: «0. .NP. .3. 00.- 00.- 00.- mo. 50. .00. 004 20.5 don acute _c_ 6920 «38» 03.0 £23 323 oEooE x83 50:00 can .00 .83 duo? «020 .280 30¢. 22:. mp 0300. 77 court). The individual race variable was related to the community variables as expected. Being White was highly negatively correlated with the percent Black in the census tract. Also being White was related to higher median income in the census tract. Finally, the higher the percent Black in the census tract was related to lower household income. The correlations among the other variables caused no reason for concern regarding multicollinearity. There were some interesting bivariate relationships among independent variables to note, however. Demographic variables were related to several legal and community variables. The race of the juvenile was significantly related to six independent variables. Being White was associated with not having a drug charge, not using a weapon, and less victim injury. Being White was also related to being male, higher household income, and smaller proportion of Blacks in the census tract. Age was significantly related to five independent variables. As would be expected being older was related to having more priors, having formal legal status with the court, multiple charges and being charged with a drug felony. Being older was also related to a higher concentration of Black population. Gender was the final demographic variable. It showed significant relationships with six independent variables. As previously stated being male was associated with being White. Being male was also related to having more priors and more co-offenders. As might be surprising to some, being female was associated with more victim injury. A closer inspection of this relationship showed that while 30% of the girls in the sample caused some kind of injury, 78 only 13% of the boys injured their victim(s). Furthermore, over one-half of the serious injuries or fatalities in the sample were committed by girls. The two remaining relationships between gender and other variables were with the community variables. Being male was related to living in an area with higher mean income and an area with fewer Blacks in the population. The community variables were significantly related to other variables as well. Table 13 indicates that the percent Black in the population was related to several offense behavior variables. A higher proportion of Blacks in the population was related to a drug charge, having a weapon, having greater victim injury, and committing the offense alone. Lower mean income was also related to having a drug charge. The dependent variables also showed some significant bivariate relationships with the predictors. Being detained was associated with having priors, being charged with a drug offense, having a weapon, multiple charges, and committing the offense alone. Demographic and community variables were also significantly related to the detention decision. Being older, not being White, living in a census tract with a higher number of Blacks, and lower mean household income were each associated with being detained. The second dependent variable, severity of disposition, had significant bivariate relationships with only prior legal history and being detained. Thus, being given a more severe disposition was related to having more priors, being on probation and being detained. The correlations between severity of 79 disposition and all other variables (including offense variables) ranged between .02 and .09. Table 14 presents the zero order correlations of the variables in the social data sample (n=305). The relationships among variables were similar in this sample to the intake sample. Some of the variables which were significant in the larger sample, were not significant in the smaller. However, the direction of the relationships remained the same. One significant change was in the relationship between age and severity of disposition. In the intake sample, age was related to severity of disposition at a non-significant .05. However, with the social data sample, being younger was significantly related to being given a more severe disposition (-.12). It may be that social factors play a more important role for younger juveniles. Family problems was significantly related to 5 of the 15 possible variables. Family problems was significantly related to being younger, having a formal legal status, having school problems, having personal problems, and not living in a two parent family. Living in a two-parent family was also associated with formal legal status, as was not having a dmg charge. Living with two- parents was also related to having a lower percent Black in the population, higher mean household income, and being White. Having school problems was not only related to having personal problems, and family problems, but to five other independent variables as well. Having more priors, a formal legal status, being male, not being White, and living in a census tract with lower mean household income were each related to 8283:. 00. p 00. 00. F c8888.. «0.. .mm: 00. P 823 00. . 0m: .00. 00. P oEeuc. cc #0.. .00. .00.- . P0.- 004 820 .x. 8. me. P c . mo. . m c- 8.. 8:8 . N F: "0. 00. N0. N0. 50. 00. P emu 90.- p P: 00. No. . N P: . m P. 00. 00. P «80:28.3 no. 0.. .mm: .3: .09. .09.- 00. c0: 00; 23:. E83 F0. .2. .0..- .0..- .9. 00.- 00. 0P: .0P. 00.— cocoa; N0. .NN. LVN.- .0P.- .mu. 00.- .0p. .03- .0..- .03. 00; 0088 020 00.- . - p. 00.- 00.- #0.. 00.. N0. 00. N0. e0. 00.- 00. F 825080 .0N. no. No. 00. 00.- no. 00.- 00.- 00.- 00.- 00. 00. 00. P «:88 .30. .mp. .09. 2.- 00.- P0. 00. e0. 5.- 90.- 00.- N0.- N0. .50. 00.- 30.8 den acute E. e998 «:88 2.20 c380 323 2:85 820 80:00 one be .83 due; 0020 8:8 .30. Beta .000 n c. oEEem 080 iced E00 3328) 8 accuse: cecceeceN c— 2an 81 00. . «no... .23 . 00. 00. . «no... .38.» .8. . N .. oo. . 39.. 2.53 00.- 3.- .0... 00.. «823 oz: .vN. . - .. 00. 00.- 50.33:. 00. 00. 00. .0N.- cezceaeu .NN. .N..- 3. .NN. 823 .0.. .h..- 00.- .mN. eEeoE cc .0..- 00. 00. .0N: 8e... 8 . _. .u .. S.- No. 323 00.. B: . m . .. 8. on. 00.- .0.- 00.. 00. 23:33-00 .- mo.- ~o.- 8.. ca... 5.83 . 0 ..- .0.- 00.- 00.. ceeeecs N0: .0. 00. . N ..- 008.8 020 .0. 00. v0. 00.- “cotaecee .2. .2. .2. .2. «a... .82 . 0.. . 0.. no. 00.- 223 «c.2090 mEeEeca mEeBeE 3:38.. .098... £63 8:23 95. .8258. S 2%. 82 having more school problems. Having personal problems also showed significant zero order correlations with several variables. More personal problems were also associated with being White, not having a weapon, fewer Blacks in the population, and higher mean household income. Each of the dependent variables had significant zero-order correlations with at least one of the social variables. Being detained was only related to not living with two parents. Severity of disposition was related to having more school problems and more personal problems. Summapy of Bivariate Relationships The main reasons for testing examining the bivariate correlations in this study was to demonstrate that the independent variables were not highly correlated and that several predictors were related to race as well as to detention and disposition. In addressing the first issue, most of the relationships did not present a multicollinearity problem. The race variable was related, however, to the community variables and the community variables were relatively highly correlated with each other. The second issue dealt with the relationships between race/ethnicity and other predictors. The results showed that race/ethnicity had bivariate relationships with several predictors. Thus, indicating the need for multivariate analysis. As would be expected, several bivariate relationships were established between predictors and the outcome variables. A greater number of variables had significant bivariate relationships to detention than to disposition. The correlation of disposition with other 83 variables was relatively low, demonstrating that these variables do not have strong bivariate relationships to this outcome. Assumptions of Normal'gy, Linean_ty‘ and Homoganiem of Variance The data were tested to determine whether it met the assumptions necessary to run and interpret regression models. To test the linearity and homogeniety of variance assumptions residuals were plotted against predicted values. If the relationships were linear, the plots of residuals should show a relatively random dispersion. Moreover, to meet the homogeniety of variance assumption, the pattern of residuals should be equal across scores. In these samples, the tests showed no discemable patterns in the plots which indicated that the relationships were linear and the variance was relatively equal across scores. Histograms were created to test for normality in the distribution of each variable. While all variables did not follow an exact normal distribution, they mostly conformed to a kurtosis and skew of between 1 and -1. The variables which were not normally distributed were transformed by categorizing responses to better approximate a normal distribution. No log transformations were necessary. Findings from Logistic and Linear Regression Analyses Prior to examining the findings, a brief explanation of the analyses used is necessary. Logistic regression allows one to create a predictive model with a dichotomous dependent variable. It is similar to discriminant analysis, however 34 the assumptions are relaxed for the logistic regression and it is easier to interpret when comparing the results with linear regression models. Given that linear multiple regression is a fairly common analysis, it may be useful to compare the logistic model to the linear model. There are three major comparisons necessary to be able to interpret both types of equations. First, in logistic regression, a maximum likelihood method of predicting the odds of an event occurring is used. It is not based on a linear model and thus, ordinary least squares (typically used in linear regression) cannot be used because no regression line is estimated. The second comparison involves interpreting the regression coefficients. ln linear regression, the regression coefficient indicates the amount of change in the outcome for one unit change in the predictor. These coefficients when standardized within the equation are called beta weights. The logistic regression coefficient, however, is interpreted as the change in the log odds of the event occurring given a one unit change in the predictor. The log odds is merely the natural logarithm of the odds of the event occurring. The third comparison is in the determination of goodness of fit of the model. In linear regression, the R2 indicates the amount of variance accounted for by the model. A change in the R2 is tested with an F test to determine whether the change is statistically significant. Assessing the model fit in logistic regression is slightly more complicated. One method of assessing goodness of fit is to inspect the classification table. The percent correctly classified is an indication of how the predictions fit the observations. The higher the percent 85 classified, the better the predictive model fits the data. The chi-square test is another indicator of goodness of fit. The chi-square tests the null hypothesis that the coefficients for the variables in the model are zero. In hierarchical logistic regression, one can also test the significance of including additional variables through examining whether the chi-square improvement shows significant change. Research uestion 1 The first research question focused on the influence of race on court decisions for youths with similar legal status and offense behavior. To adequately test the effect of race on the detention decision, legal and offense variables were entered in the first block, demographic variables were entered in the second block, and race was entered in the third block. Hierarchical Logistic Regression of Legal Variables, Demographics, and *- Bace/Ethnicitv on Qateml. Table 15 presents the three steps in this hierarchical logistic regression model. Included in each step are three columns representing, the regression coefficients, the standard errors, and the partial correlations (R) for each independent variable. The regression coefficient is interpreted as the log of the odds of the event occurring. The probability of the detention can be calculated from these regression coefficients, given specified values of the predictor variables. The standard error column contains the standard error of the logistic regression coefficients. The R column contains the partial correlations (or independent contribution) between the independent variable and detention. 86 0N. . .. ..0- 888:8 mm. a. .3. x85 .0. N0. .0 ... 2.8-. .... goo-.0 3.. .8.»- 3388 No. 8. 8. 8. .N. . .. 8.88 8. B. 8. S. B. 2. on. u x090 8. .3... 2588 3.- 3. 8.- 8.- 2. 8.- 3.- 2. . E.- 9885.8 8. o. 8. 8. 2. v c 8. 8. o .. >5... 5.8: 2. t. .5. t. 2. .8. 2. 9. .8. 8: :88; m... 8. .8. om. 8. .3. 9... B. .3. 89.8 man m: «N. .2.- nt 5. .8. 9. &. ...... .5588 8. 8 u: 8. 9. mo. 8. 2. 2. was... .82 8. R. m... 8. mm. 8. No. mm. 8. .88 . 830 c .0.... m c .00 m c .00 m .83. 8.88.0 .... co. 8.88.0 $32. 8:888 ..0.Nnu.ceEe>eaE. .X 00.Nu.coEe>e:aE. «x .0N-00u.ceEe>eaE. «x n :08 w :05 . :05 30¢"... 00:00.00 :0 8.500803... 0cm 8.: m: 0E3 850...; .m 0.. .e :2mequ 0.8.3.. @2888... 0. 0.00... 87 Step 1 indicates that having a multiple charges, a drug charge, and using a weapon were each related to detention when controlling for other legal and offense variables. Step 2 indicates that there was no significant improvement in the model by adding the demographic variables. The results of step 3 indicate that adding race to the equation significantly improved the predictive model. In the full model, being detained was significantly related to having multiple charges, a drug charge, having a weapon, being Latino, and being Black. Hierarchical Linear Regression of Legal Variables, Demographics, and Race/Ethnic'fiy on Disposition. Using linear regression, the same hierarchial structure used to predict detention was applied to predict severity of disposition as was used to predict detention. However, detention was added as a predictor of disposition. Table 16 displays the results of this analysis. Step 1 produced an R2 of .10 and an adjusted R2 of .08 and showed that having more priors, a formal legal status, and being detained were each related severity of disposition. The model did not change when demographic variables were added. When race was added to the equation, the model changed significantly (an F change of 5.88) and the R“2 increased to 12% and an adjusted R2 of .09. The final model indicated that severity of disposition was related to having more priors, a formal legal status, being detained, and being White. Offense behavior was not significantly related to severity of disposition. 88 mm. . rm. .- 9:80:00 S. 8. .8. 8.55 m .88 mm. cum-r ucmumconu mo. 0 ... _. ... mo. 0 ... 0... 30:00 8.. 8. 8.. 8.. 8. 8.- 88 a 88m 2. .2...“ 8888 8. m e. . 8. n c m c . .m. h c m c . .m- 8.888 8.. 8. 8.. 8.. 8. 8.. 8.. 8. 8.. 2888.8 8.. 8. 8.. 8.. 8. 8.. 8.. 8. 8.. E:.:. 8.8... 8.. 5. o...- 8..V .... a... 8.. .... a... 2.... :88; 3.. 8. B.- 8.. 8. m... 8.. 8. t.- 888 8.6 8. 3. 8. 8. 3. 8. 8. .... 8. 82.588 ... 2. .8. 3. . S. .8. ... S. .8. .33. .82 8. 8. .8. t. 8. .8. ... 8. .8. no.8 . 82m 38 .8... m 88 .8 m 88 , .8 m . 00.0 a. 09.020 n. 00. .3 000020 n. . 0N0 I 000020 u. 8..-.... .8 8..... 8..-.... .8 8..-... 8ft ...8 8..... 8...”. 22222 8.1. 222...... 8.... 222:... 33w a Jul-W mmwflc cog 0.0—0 "—0 ...m>m .... 0 ccuw\momm 0:0 .0_ . . .080. m n~t~> 5.04 00 :20 0....0. ...wcz 3250.500? 0.. 0.00... 89 Summary of Findings for Research Question 1. Race had an independent affect on detention after controlling for prior history, offense behavior, and other demographics (age and gender). Hypothesis 1a was confirmed because being Black or Latino was a significant predictor of detention after controlling for legal and offense variables. Hypothesis 1b which stated that Blacks and Latinos would be more likely to receive a more severe disposition after accounting for legal and offense variables was not confirmed. In fact, being White was a significant predictor of a more severe disposition after accounting for legal and offense variables. Research Question 2 The second research question addressed the impact of race on court decisions after controlling for the effect of various social factors. Regression techniques were used in the same manner as in the equation addressing the first research question. Hierarghical Logistic Regession of Legal Variables, Demographics, Social VariablesI and Race/Ethnic'fl on Detention. Table 17 presents the results of the hierarchical logistic regression analysis. The same legal and offense variables used in prior equations, were entered in the first block, the same demographics were added in the second block, variables measuring school problems, family problems, personal problems, and family structure were entered in the third block, and race/ethnicity was entered in the fourth block. The variables which significantly contributed to explain the detention decision in step 1 were having more priors, having multiple charges, being 90 cm. —. No. v. ucsmcoo MN. mm. c be. r xuflm v v. NO. ¢ OO. r 0503 v 0.00.5 mm. P OO.. ucfimcou no. NN. a Mn. 00. m —.. N v. OQOOQ 300300 00. MN. O .... 00. mm. 50. $320000 .09.—um 00. 0m. O0. 00. ON. v0. 20.5.0000 >=Ewu OO. Ne. m ... OO. Oe. NO. 002028 .550 OO. Nm cm. OO. Om. c NO. .8505 m x000m mm. —. NOR ucR—mcou OO. Om. QO. no. mm. mm. OO. mm. OO. 3950 OO. O r. OO.. 00. O r. NO; OO. OO. O r: mum N 0.00.5 um. c O0. F. 0:38.00 OO. N r. m 9.. OO. O ... O .... OO. O ... N p... OO. O ... _. W. 300:9:900 OO. N r. OO. NO. r v. c NN. OO. _. v. ¢ QN. OO. —. v. ON. 32:. 0.503 _. v. PN. .. mm. m ... ON. c OO. O r. O ... .. MO. m ... O ... . mm. 3: coammg m ... mm. c vb. —. ON. NO. c OWN NN. Om. . med m r. em. c 0N.N OO.—~20 03.5 v v. NN. . mm. m r. ON. . OO. O _.. ON. a OO. O ... ON. .. mu. 0:9... 30:00 OO. NN. MN.. OO. rN. ON.. OO. ..N. ON.. OO. ON. m V. «33» .wmm. NO. NO. OO. 5.. Om. c Nb. N... Om. . NO. N... ON. can. 30:0 —. 0.00.5 m .m.m O m .m.m m a .m.m O a .w.m m X.» ...NN 022320 .XROKO Omzfimflo AXEOOO OmEmmmU $OOOO 00:33.0 uhflfifl $0005.: ~N mod $0002.: Ln ooé $0002.: Ed onfldm 5000?: Lu. «a a deflm. ..lauww. mom“ :1 c. 00 0. Oman ._.> _0_- uuw\00m~_ 0cm _m_00m 0.2. m.— .OEOD .m.m._ “.0 00.30.. . -m 0.02 . ... 50.50.5021 2. 2pm... 91 charged with a drug offense, and having a weapon. Adding demographic variables (age and gender) did not significantly improve the goodness of fit of the model. However, when demographics were added to the equation, the amount of victim injury became a significant predictor of detention. Similarly, in step 3 when the social characteristics were added, the model did not significantly improve. However, having a mother as the sole household provider was a significant independent predictor of detention after accounting for legal variables, offense behavior, and individual demographics. Adding race/ethnicity as the final step in the equation did significantly improve the percent of youths correctly classified. In the full model some variables were not significant that had previously been and others became significant. Having multiple charges, a drug charge, and having a weapon were the significant predictors of detention from block 1. Having personal problems was the only significant predictor from all of the social variables. In the final step, being Latino and being Black significantly predicted detention after controlling for legal, offense, demographic, and social variables. Hierarchical Linear Regression of Legal Variables, Demographics, Social Variables, and Race/Ethnicgy' on Detention. Table 18 is a representation of the hierarchical linear regression model used to predict severity of disposition. In step 1 (when legal and offense variables were entered) only legal status at time of offense was predictive of disposition. Only 7% of the variance (with an adjusted R2 of only .04) was accounted for by this equation. In step 2 demographic variables were added to the equation which significantly improved 92 mm. «3.0 0:32.00 8.. 8. 8.. 8.55 e 0.88 8. .8... 8388 8. 8. .nn. 8. 8. .8. 82¢ .8088 NO. Mr. O... OO. m... Or. 8000.00.00 .0058 8.. 8. 8.. 8.. 8. 8.. 8.28.6 2.83 8.. 8. 8.. 8.. 8. 8.. 888.“ m 0.88 8. .8... 8388 8. a. 8. 8. 8. 8. 8. 8. 8. .288 8.. 8. .8: 8.. 8. .8... 8.. 8. .8: 8.. a 0.88 s... 8. 8388 8. 8. 8. 8. 8. 8. 8. 8. 8. 8. 8. 8.888 8. 8. 8. 8. 8. 8. 8. o... 8. 8. o... 8. 888.88 8. 8. 8. 8. 8. 8. 8. B. 8. 8. 8. 8. >38. 8.85 8. 8. 8. 8. 8. 8. 8. 8. 8. 8. 8. 8. 8: :88; 8. on. 8. 8. 8. R. 8. 8. 8. 8. 8. 8. 888 on... .. 8. 8.. 8.. 8. 8.. 8.. 8. 8.. 8.. 8. 8.. 82588 8. 8. .8. 8. 8. .8. 8. 8. .8. 8. 8. .8. 8.03... .82 8. 8. 8. 8. 8. 8. 8. 8. «.... 8. 8. S. 95.8 .. 83m 38. mm. m 38 .3 m 38 .m.m m 38 mm m 3:888 . .8.nl888 . and-888 n. .«zn 888 . 8.3m .08 8.4. 8...... .8 8..... worm. .8 8..... 8...... .8 . 8&8 .8.-. 222:: 8...... 222:: 8...... 222:: 8.3. 222:... a 85 .ulmummw a mom u :0 00.00.00 m. _0_::uw\000~_ 0:0 .0000m 2: 0.0 0:000 4200.. 00 :0_mm00m0~_ 000:: .00—£00009...“ Or 0.000. 93 the model. Again, legal status was significant, but being younger was also predictive of a more severe disposition. Adding social variables also significantly improved the predictive power of the equation to an R2 of .13 and an adjusted R2 of .08. Having more personal problems was predictive of disposition beyond the effects of legal, offense, and demographic variables. Adding race to the equation did not significantly add to the variance accounted for by the model. §gmmgrv of Findings for Research Question 2. Hypothesis 2a which predicted race/ethnicity would have an independent effect on detention after controlling for social variables was supported by the logistic regression analysis. Further, offense behavior and having personal problems predicted detention. Hypothesis 2b was not supported by the linear regression analysis. Race did not have an effect on severity of disposition after accounting for social factors. However, having personal problems (emotional or substance abuse documentation) predicted a more severe disposition. Research Question 3 The third research question addressing the impact of community factors on decisions was examined in a similar manner. Logistic and linear regression techniques were used to explore the impact of demographic, offense, and community variables on detention and disposition, respectively. These analyses made use of the larger intake sample. In addition, these equations were used with the social data sample to include the effects of social variables. Social 94 variables were entered prior to community variables to determine the impact of community factors after controlling for other variables. Hierarchical Logistic Regression of Legal, Demggraghic, Commun'gy, and Race/Ethnicfiy Variables on Detention. A similar regression model was used as in the first logistic regression equation reported (for hypothesis 1a). However, community variables were entered after controlling for legal and demographic factors to determine their impact. Table 19 presents the results of that analysis. Step 1 and step 2 are identical to the first two steps in Table 16. Step 2 is enumerated in Table 19 because it serves as a starting point to determine the effect of subsequently entered variables. The addition of community variables to the equation did significantly improve the predictive model. The percent classified correctly improved from 70% to 73%. The significant legal predictors of detention included: having concurrent or multiple charges, having a drug charge, and having a weapon. The percent Black in the population also significantly improved the model. After accounting for legal status, offense behavior, gender, and age, the higher the proportion of Blacks in the census tract where the juvenile lived, the more likely they were to be detained. Mean household income was not a significant community predictor of detention. Thus, hypothesis 3a was not supported, but hypothesis 3b was supported. 95 8.. .86. .538 3. 8. 3.. :85 co. ...m. 8.. 05.3 8 0.00.8 8.. .86. 8588 8. 8.. .8. 8. am .8. 8.. c. .8... .... 8 3. 3 - 8. 3. 8.. 38:. 29.33.. a x00... ..5 .88. .588 8 8. 3.. 8. 8. 8. 8. 8. Z. 8:8 8. 8. 8. 8. 8. 8. 8. 8. 8. 8a a 508 8.. 3. 8.. 8.- 8. 8.. 8.- 8 8.- 288.38 8. 8. 8. 8. 8. z. 8. 8. 3. >58. 53> 8. t. .8. 8. 8. .3. t. 8. .8. 8: :88; 8. 8. .9; 8. 8. .8; 8. 8. .8; 8.9.0 9% 8 mm .8. 8 8. .8. 8. a. .8. 22.880 8. 8.. 8. 8. 8. 8. 8. a F. 8. 8.2m .82 8 8. 8. 8 8. 8. 8. 8. 8. 28.. . x008 c .w.m m a .md m m ma m $8.8 8.88.0 $8.8 8.88.0 .8 .8 88.820 3.8805829... ~x .mudquoEoSaE. ...x omdquEoEaE. Nx q 85 n 8.5 a ..wa mmvuc 5.8080 :o 38.. can 8.8.8.08ch 2.5.8.800 .mo_:oflooEmoJmm.nmtm> .m m. .o 5.88 mm 0.8. 0 602880.: mp 0.00... 96 To explore this issue further an exploratory analysis is also pictured in Table 19. Although race/ethnicity had a fairly high zero-order correlation with the community variables (.48 for household income and -.67 for percent Black in population), it is instructive to see if race/ethnicity showed an independent effect after controlling for community variables. This analysis is exploratory because some would argue that this relatively large zero-order correlation may pose a multicollinearity problem. Thus, the results should be interpreted with this caveat in mind. In step 4, race/ethnicity was added to the equation and did not significantly change the predictive model. Even with the addition of race/ethnicity to the model, percent Black in the population continued to impact the detention decision. Hierarchical Linear Regression of Legal, Demographic, Community. and Race/Ethnicgy' Variables on Disposition. Table 20 presents variables entered in the same order as in the previous equation, however a linear model was used to predict severity of disposition. As stated earlier, the results for this first two steps of this equation are the same as in Table 16 (hypothesis 1b). Community variables did not significantly change the amount of variance accounted for by the equation. However, in step 4 when race/ethnicity was added to the equation, it significantly added to the predictive model. In the full model, having more priors, having a formal legal status, being detained, and being White were each predictors of a more severe disposition. Adding race/ethnicity to the equation raised the R2 to .12 and the adjusted Fit2 to .09. 97 Oh. «63800 8. 8. .8. 2...; .. 50.6 8. .88 .828 8.. 8. 8.. 8.. 8. 8.. ...8 s .8... 8 8.- 8. 8.. 8. 8. 8.- 858. 8.9.3.2. a 50.6 8. .38 .8388 8. 8. S. 3. 8. 8. 8. 8. 8. .288 8.. 3. 8.. 8.- 3. 8.- 8.- 3. 8.- 8o 3 v.00...“ 8. 8. .8. 8. 8. .8. 8. 8. .8. 8828 8.. 8. 8.. 3.- 8. 8.. 8.- 8. 8.. 28883 8.- 8. 3.. 3.. 8. 8.- 3.- 8. 8.- as... 5.8... 3.. z. 8.. 3.. 8. 2.. 8.. 2. 8.- 8: :88; 8.. 3. 8.. 8.. 3. 8.. 8.- 3. 8.. 8.20 9% 8. 3. 8. 8.- 8. 8.- 8. 8. 8. .8888 2. 8. .8. S. 8. .8. Z. 8. .8. 8.3.. .82 8. 8. .8. 8. 8. .3. 8. 8. .3... as... 8 508 33 .m.w m 28 .m.m m 88 .w.m m 8.38.86 u. 8.»... ..8 8.3880 n. 8...... ..8 8389.9 a 8.8m .8 8...& 3."... 22.3.2 2...: .88. 28...: 8."... 8...... 22.3.2 8 ...um o ..wa 3 $5 . o>om co comm 8.18. £28588sz £26600 mo... 80ng .mm_na_.m> _m 3 .o 5.88 mm .88.... 80.5288 ON 220... 98 Hierarchical Logistic Regression of Legal, Demggraghic, Social, Commungy’ , and Race/Ethnic'gy Variables on Detention. To further explore this issue of the impact of community factors on detention and disposition, analyses which controlled for social factors were conducted (Table 21). For the logistic regression analysis of detention, Steps 1 through 3 in the equation were previously reported in Table 17. After controlling for legal, demographic, and social variables, the percent Black in the census tract was still a significant predictor of detention. In fact, adding community variables to the equation significantly improved the predictive power of the model. The final block of variables (race/ethnicity), in step 5, also significantly improved the predictive model. Interestingly, when race/ethnicity was added to the equation, the community factors were no longer significant and the individual race of the youths was significant. In other words, being Black and being Latino significantly impacted the detention decision after controlling for legal, demographic, social, and community factors. Hierarchical Linear Regression of Legal, Demographic, Social, Commgnitv, and Race/Ethnicity Vaiables on [Letentigy The same structural model was used in a linear regression equation predicting severity of disposition. Table 22 indicates that the addition of community factors did not significantly change the predictive model. Moreover, the addition of race/ethnicity in the last step of the model, did not add to the predictive power of the model. Thus, when legal factors, demographics, and social factors were 99 Se 8.... 8.888 8. S. .36 .8... 8. 8. .3... 853 m .88 R... 8.... 8888 8. mm. mm. P. MN. .30. don c. 530 .x. 8. 8. 8.. 8.. 8. 3.. 8.8:. 88.. c v.88 Be 8.. 8888 8. 3. .8. 8. 3. R. 8. 8. 8. 8:28... .8088. 8. 3. 3.. 8. 8. 8.. 8. 3. 8. 8.292.. 68...... 8. on. 3.. 8. on. 8.. 8. 8. 3. 8.2.8... 2.58 8. mm. 8. co. 8.. 8.. oo. ow. um. 322:8 .850 8. 8. 8. 8. an. 3. 8. 8. .3. .658: ... .60.... 8. S. E. B 8. 8. 3. 8.. R. 88.8 8. 8. 8.. 8. 8. 8.. 8. 8. 8.. m8 3 89m 8. 8. ...... 8. t. 8.. 8. 8. 8.. 285898 8. 8. 8. 8. 8. ...... B. S. .3. >53. 5.8... 8. 3. .3. 8. 8. .3. 8. on. .8. 8.. :88; 8. mm. .36 8. 8. .8... 8. E. .3." 8.8... on... 8. 8. .3. .... R. .3. 8. 8. .8. 35:38.8 8. 3. 3.. 8. 3. 3.. 8. 3. 8.. .38. .82 B. an. .«o. 8. S. .8. S. 8. .R. 96.... .. .88 a .3 m a ...... m a ....m m 88.3 3:8... 88.8 8.283 8...... 3.9.20 .38 ucEEaE. .x .882. 8888658. .x 8... ...E3. ~x m ..8... III... ..me n ..me moan: cozc8oo co $3885 comm new .>._c:EEoo ._m_oom ..o...a8doEmo ..mch .o 5.88qu o_.ml_ao._ 30.2889: 5 29a... 100 8% orfi.‘ «sump—00 8.- 8. «N- of; m 200.... 8.. .8... .588 ...- 8. a: 8. .8: ...... ... .02.. .... 8. 8. 8. .o. 8. .... 2.8:. 88.. v .60.... 8. Le... .588 8. 8. .8. 8. 8. .8. ... 8. .8. 8.28... .88.... 8. 8. ... 8. 8. 8. 8. 8. 8. 2.28... .88» ...: ... 8.- 8.- ... 8.- 8.- ... 8.- 32.2.. .23. 3.- 8. 8.- 8.. 8. 8.- 8.- . .. 8.- $28.... a 200.6 8. .N 8. 3. .u. 8. ..o. .N 8. .88.. 8.- 8. .8: u... 8. .8: 8.. 8. .8: 8.. m .60.... 8. 8. ... 8. 8. 8. 8. ... 8. 5.8.2. 8. 8. 8. 8. 8. 8. 8. 8. 8. 2.82.98 8. 8 8. 8. B. 8. 8. .... ... 81... EB... 8. 8. ... 8. 8. 8. 8. 8. 8. on. .589; 8. .... R. 8. 8. 8. 8. 8. R. 8.2.. as... 8.- 8. 8.- 8.- 8. ..N- 8.. 8. 8.. .5588 8. 8. .8. 8. 8. .8. 8. .... .8. was... .82 8. 8. cm 8. 8. 8. 8. 8. 8. 22... . v.00... 28 ma m 28 .m.m m 28 .m.m m 8.385.... n. .8»... ...... 6.882.. n. .8»... ...... ......»0820 n. .8»... .2. 8.»... ....»m 23.3.2 8.»... .8»... 283.2 8.»... .8.».. 283.2 m ..um .. ..um n ..um Gonna. 5:30qu .0 233m co 8.3%.; 88. new EcaEEoo ._m_oow 6:80 .33 .o c9383.. 82.... .32220... 8 2...: 101 taken into account, community variables and race/ethnicity did not significantly predict severity of disposition. Summary of Findings for Research Question 3. The first set of equations tested hypothesis 3a and 3b which focused on the impact of community factors on detention decisions. Hypothesis 3a was not supported in that mean household income was not a significant predictor of detention. Hypothesis 3b, however, was supported. The higher the percent Black in the census tract, the youth was more likely to be detained. When social variables and individual race/ethnicity of the youth were accounted for in the model, these findings held. Hypotheses 3c and 3d predicted that community variables would influence disposition. These hypotheses were not supported by these data; community variables did not significant impact disposition. However, the exploratory analysis without controlling for social factors showed that an individual's race (being White) was a significant predictor of a more severe disposition. When social variables were accounted for, race/ethnicity was no longer a significant predictor of disposition. Having more personal problems was related to a more severe disposition. Research uestion 4 The final research question addressed the effects of all the aforementioned variables on youths of different races in separate analyses. Thus, all analyses mentioned previously were conducted separately for Blacks and for Whites. These analyses determined the differential influences of 102 demographic, offense, social, and community variables on Blacks and Whites separately. Hierarchical Logistic Regression of Legal, Demographic, and Qommun'fly Variables on Detention for Blacks and White Separately. Tables 23 and 24 represent the logistic regression analysis predicting detention from legal variables, demographics, and community variables for youths separately by race. A significance test comparing the proportions correctly classified by the model indicated a significant difference between the models. The model predicting detention for Black youths was significantly better than the model predicting detention for White youths. The separate models are delineated below. Table 23 describes the model for Black youths (n=232). In step 1 of the model with only prior legal and offense variables included, having multiple charges, a drug charge, and using a weapon each added significant variance to the model. When demographics were introduced, the overall model did not significantly improve. However, being male was a significant independent predictor of detention. The model did not significantly improve in the third step, either. Community variables did not seem to influence detention for Black youths beyond the effects of legal and demographic variables. The overall model only correctly classified 64% of the youths. Table 24 presents the findings from a logistic regression analysis using the sample of White youths from the intake sample (n=204). In step 1, having a drug charge was the only significant predictor of detention. Adding the 103 8; 8.9 2288 00. mm. 5. .000 c. x003 8 00. mm. 3. 0500:. 0.23:0... m x0040 mm. P .00.“. .05ch 8. 8. .8. 8. 8. .8. 8:8 8. o F. 8. 8. 8. 8. 8a N 200.5 8. 8.- £288 8. 8. 8.- 8. S. 8.- 8 t. 9.. 9.88:8 8. 2. 3. 8. 2. up. 8. NF. 8. 8.5 5.83 8. mm. .8. 2. a. .8. NF 5. .8. 8: :88; 9. 3.. LN... 3. 3. .8; E. 3.. .8; 855 86 0.. 8. .8. 8. 8. .8. F F. 8. .8. .5588 8. 8. 8. 8. mm. 8. 8. 8. 2. «as... .82 8. 8. 2. 8. 8. 8. 8. 8. 8. 225 . . 0.00.5 m .m.m m m .m.m m m .m.m m 38.8 3:300 8&8 3:320 $8.8 85820 0 F.mu.coEm>an_ LA 0 mwhw RéufioEQEaE. ...x N mmkw .vm.vuu.coEo>0.0E. ~X p awhm mm 030... 104 acmamcoo 3. o... .8. .80 ... 0.8... .... 00. Na 00.. 0500... 0.020000: 0 x003 00.0 00.0- $0.800 00. cm. mm. 00. 5.. no. .00000 00. 0 F. 00. 00. N F. o F. 000 N v.00..m 0.. 00. .8; 30.88 «or mm. 00.- oo. vm. 00.- 00. «N. «0.- 200003000 00 cm. up. co. 2. mp. 00. 9. m P. b3... 8:03 8. mm mm. mo. 5. mm. so. on. um. 00.. 000003 0.. 8. .00.. 3. 8. .8." 0.. 8. .2." 0980 8.0 or. 00. .05. 00. on. 5. so. no. me. 20.50000 oo. 00. co. 8. mm. 00.. 00. «0. mo. 020.0 .000. 8. S 8. 8. m... 8. 8. 8. 8. 90.... P x003 m .m.w m m .w.m m m .w.m 0 800.3 00580.0 $00.2. 00580.0 800.2. 00580.0 .0 p. 9 7.0060355. ...x m mwhw $.u.c0E0>o.QE. «x N mmkw .—u.0pn.c0E0>0.aE_ Nx — mmkw .3 u... 05:01:55 .0. 00.25.00 :0 m0.0fl...|0> EcaEEoo 0:0 00.... 00.00800 4000.. .0 00.80.00”. 0...... .00.. 00290.0... cm 0.00... 105 demographic variables of age and gender did not improve the model. In step 3, when community variables were introduced, the model improved significantly and correctly classified 82% of the youths. In the complete model of the detention decision with intake cases, having a drug charge was predictive of outcome as well as the percent Black in the census tract. Thus, White youths who lived in areas with higher concentrations of Black population were more likely to be detained. Hierarchical Linear Regression of Legal, Demographic, and Communig Variablgsgn Disposition for Blacks and Whites Separately. Tables 25 and 26 present the linear regression models on the intake sample for Black and White youths separately. Although the multiple R in the model for Blacks was lower (.30) than in the model for Whites (.43), this difference was not statistically significant. In general, the linear regression equation predicting disposition for Black youths was not a good model (see Table 25). Using the adjusted R2, it accounted for 4% to 5% of the variance depending on the step examined. The addition of demographic and community variables as steps in the equation did not significantly add to the predictive power of the model. In the full model including legal, demographic, and community variables, not a single variable was a significant predictor of disposition. The same predictive model used on the intake sample of White youths (n=204) accounted for twice as much of the variance as the model for Black youths, although this difference was not statistically significant. As Table 26 indicates, In step 1, having more priors and being detained predicted a more 106 0... .... 80.0000 8. 00. 8. .08 ... v.8... .... 00.- m .. .0.- 0500... 0.2.0000: 0 0.00.0 vo. . .00. . E08000 0.. mm. 0.0. 0.. mm. 0.0. 30:00 00. 0.0. 00. No. 00. 00. 000 0 0.00.... .0. .20 20.88 . .. .0. 8. . .. .0. 8. 0 .. 00. 8. 8.80.8 . ..- 0.. 00.- ...- 0.. .0.- 8.- 0.. 0 ..- 208898 0o. 8. 8. 0o. 8. 8. .0.- 8. 00.. 0...... 5.8.) 8.- 0.. 0... 8.- .... 0..- 8.. 3. o... 8.. :88; .0. 00. 8. .0. 00. 8. .o. 00. 8. 8.08 02.. 8. 00. 8. 0o. 00. B. 8. 00. . .. .8888 o .. .... ..0. o .. .... v0. ... ... 00. was.» .08. 0.. v0. 8. ... 00. 8. 0.. 00. 0... 28.. . x090 0.0m .w.m m 0.00 .m.m m . 0.0m .w.m 0 0.8805 n. 8...... ..8 0... .8808 n. .00.»... ...... £00382... n. 8."... .8 8."... .8.um 0.8.22 8."... .0.»... 29.35. 8."... 8..... 28.22 0 00.0 0.0mm Wanda» mm 0.00... 107 00. . .00-0 00.0000 00.- 00. 0... .000 0. 0.00.0 .x. 00.- 0.. 00.- 2080. 29.88.. 0 0.00.8 8. .00-0 .0288 8.- 00. 0..- 8.- 00. 0... .0008 .o. 8. .o. .o. 8. .o. 80 0 0.00.0 00. .0..0 .0288 .0. 00. .3. 00. ..0. .0... 00. ..0. .00. 02.8.8 8. 0.. 8. 8. 0.. 8. 8. 0.. S. 20880.8 8.- ... 8.. 8.- ... 8.. 8.. . .. 8.- 03.0. 0.8... ...... 0.. o..- ...... 0.. ...- 8.. 0.. 0..- 8.. 088; B.- 8. 8.- 8.. ..0. 8.. 8.. 8. 0...- 0920 02.. B.- 00. .0.- 8.- 00. .0.- B.- 00. 00. 0.2.8.8 0.. 0.. .0. 0.. 0.. 00. 0.. 0.. .0. 8.0.0 .82 00. 00. .0... 00. 00. .2.- 00. 00. .0... e2... . 0.0000 200 .00 0 0.00 .00 0 200 .00 0 0.8808 n. .0..»«0 ...... 00.8808 u. .....uw. ...... 00.0882... 0 .....u... ...... 0..".0 .00."... 22...... 0.."... 0..."... 22.3.2 0.."... .00-"0 22.3.2 lull... 00.0 3'03 3040 000": 0050 00 02.000.00.000 2.0204000 000 00... 0. 0.000 00.0000> .0. 0. .0 00.000. 00 .000... 00.02000... mm 0.00... 108 severe disposition. As in other models, adding the demographic variables did not significantly improve the model. In the final step, when community variables were added, the overall predictive power of the model did not change. The only variables which continued to predict disposition were priors and detention. Hierarchical Logistic Regression of Legal, Demggraphic, Social and Commun'fiy Variables on Detention for Blacks and Whites Separately. Tables 27 and 28 present the results of a full models predicting detention for Black and White youths using the social data sample. Significant differences were found between the percent correctly classified in the models for Blacks compared to the models for Whites. The model for White youth was significantly better at predicting detention, than the same model used on the Black youths sample. Table 27 contains the results for the model for Black youths. The sample size was reduced to 151 because Black youths comprised about one- half of the social data sample. As before, prior legal status and offense behavior variables were added in step 1. Step 2 contained age and gender as demographic variables and step 3 included social variables. Finally community variables were added to the equation in step 4. In step 1, having more priors, multiple charges, and a drug charge were significant predictors of being detained. When demographics were added in block 2, the model did not significantly improve. However, being male emerged as an independent predictor of detention. Similarly, in step 3, when social variables were added, the model did not improve. Adding community variables did not significantly enhance the predictive power of the model for Black youths. 109 In sum, legal and offense variables were the best predictors of detention for Black youths although the model only classified 68% of the youths correctly. The model did not improve with additional variables added. The results presented in Table 28 show a very different process of which factors were predictive of the detention decision for White youths. Interestingly, the final model classified 81% of the White youths correctly, while only classifying 68% of Black youths correctly. For White youths, prior legal status did not predict detention. The amount of victim injury was the sole legal or offense variable predicting detention. As in most other analyses, age and gender did not enhance the prediction model. When social variables were added in step 3, the chi-square (or fit of the model) significantly improved. Whereas amount of victim injury dropped as a significant predictor of detention, using a weapon became significant, as well as having a mother as sole household provider and having personal problems. When community variables were added in the final step, the model again significantly improved. In the full model the significant predictors of being detained were: more victim injury, having a mother as sole household provider, having personal problems, and living in a census tract with a higher percentage of Blacks. Thus, a White youth who lives in a predominantly Black neighborhood was more likely to be detained after controlling for legal, demographic, and social factors. 110 MQN 00.. 8. mm. 0.4... 8. an. ear 8. cm. 3. 00. mm. NF... oo. 5. rm.. 8. on. mm. 8. ..m. 3.. no. we. as. oo. 3.. no. 8. cm. our 8. m... 8.. 3. mm. me. 5.. 00. .mn.—. 3.. cc. .3. 8. an. war no. me. an. m .m.w m co. mufi omé. 8. mm. cc. 8. mm. 9... 8. cm. mm.. 8. mm. cm. 00. cm. 9%. 8. on. em. 8. a... No. 00. cu. ..N.. 8. m... 8.. co. an. 3. N... we. .0».—. 3.. cc. .00. 5.. en. mvr no. 3. mm. a .m.m m mrd SN. 2... 3. . 90. oo. 3.. 3. 8 mu. 3... 8. 3.. ..or no on. 5. Mr. N0. .8.—. mr. mm. .30. 8. ...... me.. 00. me. . «a. a .m.m 0 mm. 8. an. co. 0... no. mu. m0. mm. Ne. mm. 8. on. me. we. .8.—. .00. mm: . no. ~— .w.m 0080000 .000 ... 0.003 8 0500.... c 0.00.5 00000000 000.0 0000.00 «0.0.00.0 .00000 8.0.00.0 2.0.0. 8200.00 .050 .0505 0 .0000 0080000 .00000 mow N 0.00.5 0080000 000000.000 >53. :50... 00: 00000.... 00.000 00.0 000000000 0330 .000. 0.2.0 .. 0.00.5 $3.8 00.0.0050 , B. 0000.0>0.0E. .0 E $3.00 not—000.0 00.0 0000.26.08. 8 $00.00 .00—2000.0 00.0 0000.05.05. .N N awhm $3.00 00.0.0050 .MQ.OP ucmrcm>00aF= ~N p awkm 000.600 000 .0800 620.00.000.00 .0 0.. .0 00.000. 0C 000. 0... 2900: um 0.50... 111 cm.» and. ucfimcou 2. 00. .00.. .000 0. .000 .... 8. 2. 8.. 00.80. a 0.0000 2.0 0...... 008008 2. 8. .00. 3. 00. .0.. 000.0 0000.00 co. 3. mar oo. 00. m.... 05030.0 .02—00 3.. av. 2... 8. me. 8.. 05030.5 2.800 co. 5. m1. 8. mm. «0.. 00200.8 .050 2. 00. .00.. E. 00. .00... .0500. 0 .00... mod 8. 3.80.000 8. .0. 8. 8. 2. 2. 8 8. .0. .00000 8. 2. 3.. 8. 2. 2.- 8. 2. 2.. 000 0 0.0000 00. .00.“. 00.0008 8. mm. 8. 8. 3. 8.. 8 0.... 8.. 8 00. 2.. 20000.38 2. 00. .0.. 8. 00. 3. 2 00. .00. 2 00. . 8. .020. 0.00... 2. 0.. 00. 2. 3. .00. 8 0... 0.. 8 00. 3. 00: 00000.0 8. 00.3 8.2 8. 2.00 00.2 8. 2.8 00.... 8. 00.8 00.0 00.000 0.00 8. 00. 00. 8. 2. 0.. B 3. E. B 3. 0.. 0000808 8. S. 2. 8. 00. 0.... 8 S. 0.. 8 00. 0.. 0380 .80. 8. 00. 8.. 8. 8. 8.. 8 .0. 00. 8 .0. 2. 0.000 2 .6000 0 0.0 0 0 .0.... 0 0 .00 0 0 0.0 0 300.8 000.0000 33.0. 000.0000 38.... 000.0000 :8... 000.0000 .095 2050.509:- mn .oufir uc0E0>o.nE_ mu me... uc0E0>o.QE_ Ln .mndr uc0E0>00nc= t4 ..de ... 000... 0 0000 . .00.... mm 0.00... 112 Hierarchical Linear Regression of Legal, Demographic, Social and Community Variables on Disposition for Blacks and Whites §egarately. Table 29 presents the results of the hierarchical linear regression equation conducted on the sample of Black youths only. These findings indicate that no legal or offense behavior, demographic, or social variables predicted disposition. The model had low R2 of .05 (adj. R2 = .00) with only legal variables in the model to .11 (adj. R2=.02) with demographic and social variables added. However, in step 4 when community variables were added to the equation, a significant change in the predictive power of the model occurred. The full model had an R2 of .17 and and adjusted R2 of .07 which indicated that having more school problems and having a lower percentage of Blacks in the population were most predictive of severity of disposition. In other words, the fewer Blacks in the population, the more severe the disposition for Black youths. The same analyses conducted with the sample of White youths showed a different predictive pattern as shown in Table 30. However, when the amount of variance accounted for was compared between the model for Blacks and the model for Whites, no significant differences were found. Although, similar to the model for Black youths in that prior legal and offense variables were not predictive of disposition, age significantly added to the predictive power of the model. The younger the youth, the more likely s/he was to be given a more , severe disposition. When social variables were added to the equation in step 3, the model was also significantly improved. Thus, youths with only one charge, who were younger, and who had more personal problems were more likely to 113 pm. w 31v 582.00 2.. 8. 5... dog 5 x8... .x 9. 2. 3. 8.8:. v 50.5 «3 .8.» £288 8. on. 8. 3. om 8. 32.. .888 2. on. .8. h F. on B. nausea 32.8 8.. 8. 8.. 8.. 8. 8.. 25.83 2.52 a... 8. 8.. 8.. 8. 3.. E28.“ a 50.5 9... .8.“ .598 9. 8. 3.. 2. 8. we. 9. mm 8. .88.. 8.. 8. 8.- 8.. 8. 8.. 8.- 8. 8.. 8. N 50.5 8. .8.« .5238 8.. 8. o..- 8.- 8. :.- 8.- 8.. 9.- 5.- 8. 8.. 8.228 8. m... 3. 3. 9. B. 8. 2. 8. 8. m... 2. 2888.8 up. 8. 9. 8. 8. z. 2. 8. z. B. 8. B. as... .58.) 3. 2. 8. «a 2. 8. 8. 3. 2 8 3. m. 8: :88; m... 8. 8. «F. 8. 8. :. 8. 8 8 8. 8 835 can 8. cm. 8.. 8.- 8. 8.. 8. 3 3. 8. a. 8. .8528 9. 2. 8. 2. cm R. 9. 8. R 2 2. 3. 3.2» .82 8. 3. S. 8. a. a. 8. ~« 8 8 a 3 22... . cage 28 .m.w a sum .m.w m 98 .m.m m «.8 ma m 8338.23 n. 8.....“ .8 8.70820 u. «out ...: {Tomcat n. 8...». ..8 3.3820 n. 8...: ..8 cities. 9.3.3.2 :.u& 8.8 28...: B....m.&.u¢ 22.52 8.u.m .Nmum 233.2 a a .35 N 85 . “.um Spa: «53> xosm 8. coEm m5 .0 33.5w :0 333:; .caEEoo new. .m_oom of 8 9.50 .m 3 .o co. m2 cm .695 302223 mm min... 114 av; .86 2.59.00 8.- 8. 3.. .8... ... 28... .... 8.- 8.. 8.. 2.8... 8 200.5 3... .36 .5228 8. .... .8. 5. 5.. .8. 82.. .882. 8.. 8. 5.. 8.- 8. 8.- 28.8... .88.. 3.. 8. 8.- 3.- mm o... 2:28... 2.52 8.- 8. 5.- 8.. ...w 5.. .5238 a 200.5 ...... .8... .5228 8.. 8. 5.. 8.- 8. o..- 5.- 8. 8.- .88.. 8.- 8. 8.. «- 8. .8.. 8.. 8. .8.- ..8 a 200.5 8. .28 .9228 5. 8. 8. 8 8. 5. 8. 8. t. 5. 8. 8. 52.528 5. 2. 8. 8. 2. 8. 8.. .... 8.. 5.. 8. 5.. 22.888 8. 5.. 2. 8 .... 5.. 8. 8. m... ... t. 3 >5... .....o... 5. 8. 8. 5. .N 8. 3.. .5 ...- 8.. 5. 8.- 8.. 882... 8. 3.. B. 8. 8. 8. 8.. 8.. 5.- 8.. 5.. 8.. 8.2... 9.... 2.. 8. .2... t.- 8. .2... 8.. 8. m...- m... ...... 8.. 2.2.858 .... 8. 8. 8. .... 8. .... .... 8. 5.. 8. 8. 8...... .82 :- 8. .... :- 8: 8. .... 8. N... ... 8. 8. 22... . 200.5 28 mm m 28 .m.w m 22. .m.w m 28 .m.m m 8.882... n. it... ...... .5882... n. ......u... ..8 82.382... .. 5..»... .8 8.88520 ... .84.. ...... Kat-8...”. 23...... 8...... 8...... 28.35. ......m.....um22..az ...»... 8..... 283.2 mama n ..um N 85 . 85 .m>o co mo_na_.a>J£c:EEoo pew .28m 0... a. 9.50 .m m... .o c .88 m .5 5.133282... 8 28. 115 get a more severe disposition. The model including legal, demographic, and social variables accounted for 14% of the variance (R2=.24) for White youths. The final step, adding community variables, did not significantly add to the predictive power of the model. Interaction Effects Between Flace and L_e_gal, Demographic, Social, and Commgnitv Predictor Variables. To test for interaction effects in the detention decision, all legal, demographic, social, and community variables were input into the models in one step in a logistic regression. After controlling for these variables, an interaction term was created between race and each predictor variable. Separate equations were run for each interaction effect, thereby assessing the specific interaction after controlling for main effects only. The same process was used to test for interaction effects in predicting disposition using a linear regression model. There were two significant interaction effects in the detention models and none in the disposition models. After controlling for all the possible main effects in previous equations, there was a significant interaction between race/ethnicity and household provider and between race/ethnicity and percent Black in the census tract. To interpret the interactions, analysis of variance was used to predict cell means while covarying out all main effects. Figure 4 is a plot showing the interaction between race/ethnicity and the percent Black in the census tract for the detention decision. The percent Black in the census tract was categorized to approximate a normal curve and for ease of interpretation. 5.85qu c. 285 Emu-.0.“. 8... 83.8... 26. 85.3 w 8.0ng 8.3.5... 116 aaaaaaaaaaaaa OOOOOOOOOOOOOOO 0 rd Nd md vd md 06 , No mm_nm_.m> 353E800 new ._m_oom diam-.6052. .89."... .8 9:58:00 9.8... =8 85.85 cozcofio ..o. .08.? 26.80 5 202m Emeon. ucm 83. .0 5:09.85 v 8:9“. 117 The figure shows that when the youth lived in a census tract that had a high concentration of Blacks, the likelihood of being detained was similar for both racial groupings. However, when the youth lived in a census tract which was predominantly White, there was a very large difference in the likelihood of being detained. White youth were two times less likely to be detained when they lived in a predominantly White census tract. Figure 5 presents the significant interaction between household provider and race. The figure shows that when the youth lived with two parents, White youths were approximately two and one-half times less likely to receive detention. The difference between youths of different race/ethnicity groups was lessened considerably when the youths did not live with two parents. Summarv of Findings from Research Question 4. In sum, the analyses conducted separately for Blacks and Whites yielded quite different results. The most striking finding from the separate analyses of Black youths and White youths was that the models predicting detention for White youths were significantly better. The models for White youths consistently fit the data better than the models for Black youths. However, while the trend pointed toward better predictive models for disposition for White youths, no statistically significant differences were found. Hypotheses 4a which stated that the predictors of detention would better fit the data for White youths than Blacks youths were confirmed. Legal variables were the only predictors of detention for Blacks, whereas detention for Whites was also predicted by a higher 118 550 5.030.“. Eocmmzo... o ...o 853.. w 9.03m I 8.2;: Nd md v.0 md 0.0 - Nd wozmtm> 3:58:80 ocm ..w_oom 6.5330960 .33.. .6. 9:65:00 c2280 cozcflmo ..o. ..mnSEn. 29.330: new comm. .0 22899:. n 9:9“. .3:an 22 mamme =w0 119 percentage of Blacks in the census tract, and more personal problems. Hypothesis 4b which predicted better prediction of disposition for White youths was not confirmed. However, the disposition decision for Blacks was significantly predicted by having more school problems and living in a predominantly White census tract. For Whites, a more severe disposition was significantly related to having a single charge, being younger, and having more personal problems. It seems that for both Blacks and Whites detention was better predicted by prior legal and offense factors, and disposition was more likely based on other factors. Out of the various equations used to detect interaction effects, only two were significant. No interaction effects were found in predicting disposition. The interactions between between race/ethnicity and percent Black in the census tract, as well as household provider in predicting detention were uncovered. The lower the percent Black in the population, the greater the difference between the likelihood of detention for Blacks and Whites. Black youths were more than two times more likely to be detained than White youths when they lived in predominantly White neighborhoods. Living with two parents also benefitted White youths more than Black youths. A White youth with two parents was much less likely to be detained than a Black youth with two parents. The difference in detention probability was less when the youths were not living with two parents. In sum, the hypotheses regarding interaction effects were generally not confirmed. There were no interaction effects in predicting disposition, thus 120 hypotheses 4f, 49, and 4h were rejected. There were also no interaction effects between race and legal variables in predicting detention, thereby rejecting hypothesis 4c. Hypotheses 4d and 4e were confirmed because interaction effects were found between race and a social variable, as well as race and a community variable in predicting detention. Discussion The primary objectives of this study were to determine whether race/ethnicity influenced juvenile court decisions and to ascertain the importance of various factors in the decision-making process which may affect youths differentially by race/ethnicity. Two court decisions were examined: detention and disposition. The results suggest that a youth's race/ethnicity was a determining influence on juvenile court decisions. However, the relationships among race/ethnicity, social factors, community characteristics, and juvenile court decisions are complex. The solutions to this problem may be as complex as the relationships. The follow sections are constructed to better understand this complex issue and bring the findings, the implications, and future directions together. The Relationsfihifitween Race/Ethnicity and Detention Prior to reviewing the findings, it is first necessary to demonstrate that the statistical models were better predictors of detention than would be predicted by knowing the base rate of detention. Knowing nothing about a youth except that she or he was in the court records, we would correctly predict that the youth was not detained 61% of the time. In other words, the base rate of detention in the intake sample was 61% non-detained. The logistic regression analysis showed that the model had more predictive power (84% correctly classified as non-detained). In the social data sample, the base rate of non-detention was 53%. The logistic regression models correctly predicted 75% of the youths non-detained. Given that the models 121 122 were better than random prediction or base rate prediction, there is much to learn from the statistical models predicting detention decisions. The results of these analyses support some previous research suggesting that being Black or Latino was significantly related to being detained (see Pope & Feyerherm, 1990 for a review). Several analyses were conducted which point to complexities in the data and in detention decision-making. For instance, the findings initially indicated that being from a family with the mother as the sole household provider was indicative of being detained. However, when race/ethnicity was put into the equation, household provider was no longer significant. This was probably due to the strong relationship between being Black and having a mother as the sole household provider. Also, when race/ethnicity was added to the social data model, having emotional problems emerged as another predictor of detention. Emotional problems may be apparent to the police officer who originally took the youth into custody and/or the intake worker. Beyond social problems, however, being Black or Latino was still predictive of detention. Interestingly, besides race/ethnicity, other variables that would likely be known at the time of detention were also statistically significant predictors of detention. They include: 1) offense behavior, 2) prior known offenses, 3) address of the youth (indicating their neighborhood), 4) household provider (guardian would be notified, and 5) any outwardly apparent emotional, psychological, or substance abuse problems. 123 The findings were also somewhat complicated when analyzing the significance of community factors in detention decisions. Without including social variables as independent predictors, the percent Black in the census tract was related to detention. This finding held even after the individuals' race/ethnicity was added to the model. When the same model was run, but included social variables as well, the findings were slightly different. After controlling for social factors, living in a community with a higher Black population, the youth was more likely to be detained. But when individual race/ethnicity was added, the community variable dropped out and the individual's race was more salient. This finding was probably due to the relatively high correlation between an individual's race/ethnicity and the racial population of the census tract. Separate analyses were conducted for Black youths and White youths to determine the effects of various factors on detention differentially by race/ethnicity. First it can be demonstrated that these models were also better than base rate predictions. The base rate for non-detention in the intake sample for White youths was 79%. The logistic regression models for White youths, however, correctly predicted non-detention for 96% of the youths. For Black youths the models were also better than base rate predictions. The base rate for Blacks youths in the intake sample for non-detention was 46%. The percent correctly classified by the model was 56%. The comparison of base rates with model predictions showed the same pattern using the social data 124 sample. The comparison of base rates to correctly classified rates, shows that the models were instructive in predicting the factors influencing detention. For Black youths (model with legal, demographic & community variables) having multiple charges, a drug charge, a weapon, and being male were significant predictors of detention. Thus, gender and offense behavior were considered in the detention decision for Black youths. However, the model only correctly classified 64% of the youths. The same model for White youths correctly classified 82% and only two variables were significant predictors: ! having a drug charge and the percent Black in the census tract. Similarly, in the separate models which included social variables, having more priors, multiple charges, and drug charges were predictors for Black youths. Still only 68% of the youths were correctly classified. For Whites, 81% were correctly classified with this model. The variables predictive of detention for White youths were: having a weapon, having a mother as sole household provider, having personal problems, and living in a census tract with a higher percentage of Blacks. Thus, it seems that factors traditionally thought to influence decisions (using the parens patriae model) were important for White youths, but not for Black youths. To further explain these differences, interaction effects were found between race/ethnicity and household provider, as well as percent Black in the population. The significant interaction between race and household provider, along with the independent main effects of each, show the importance of this social variable. Having two parents for White youths was significantly different 125 than having two parents for Black youths. With two parents, Whites were less likely to be detained, whereas for Blacks they were slightly more likely. Thus, having no father in the house is more salient in detention decisions for White youths than Black youths. The interaction between race and percent Black in the population was also interesting. Blacks were likely to be detained at the same rate regardless r of the racial population breakdown, however, Whites were much less likely to F be detained when they lived in a predominantly White census tract. “ ‘5 These findings suggest that, in part, detention for White youths depended on whether they had the characteristics usually ascribed to Black youths such as a one parent family structure and living in a predominantly Black community. White youths were more likely to be detained if there was no father in the house and they lived in predominantly Black neighborhood. The Relationship Between Race/Ethnicity and Disposition The first model tested (with only legal and demographic variables - including race/ethnicity) showed that having more priors, a formal legal status, being detained, and being White were related to a more severe disposition. Actually, adding community variables did not change the predictive model. In both models, the actual offense behavior did not significantly predict outcome. Given that the disposition decision was made with social factors as part of the decision, the equations assessing these issues were more enlightening. The actual offense charged with was still not predictive of outcome. However, prior legal status, being younger, and having more personal problems each 126 significantly predicted a more severe disposition. Unlike the model without social variables, when social factors were taken into account race/ethnicity was no longer a significant predictor of disposition. These analyses showed that the courts put more emphasis on what they had tried before with the youth, what the social situation was like, and the age of the youth than other measured factors when making dispositional decisions. Qualitative data garnered from interviews of juvenile court decision-makers support this finding. In general, they concurred that their perception of the youth's amenability to “treatment“ was especially important in disposition decisions. The analyses conducted separately for Black and White youths further support the notion that the typical factors used to predict outcome were not useful for Black youths. None of the variables typically used in other studies significantly predicted disposition for Black youths. The analyses in the present study, however, also included measures of school problems and community variables, which were both significant predictors of disposition. When social and community variables were added to the model, Black youths were more likely to receive a more severe disposition when they had school problems and were living in a predominantly White neighborhood. There were no significant differences in the predictive power of the models for Blacks compared to Whites. However, for White youths, having a single charge, being younger, and having more personal problems were predictive of disposition. Again, the actual offense behavior did not predict 127 disposition. Also, no significant interaction effects between race/ethnicity and individual predictor variables were detected for the disposition decision. Like the detention decision, it seems that judges and referees follow the traditional parens patriae model of the juvenile court for White youths rather than Black youths. A young White juvenile with personal problems was more likely to receive more ‘treatment' or the more severe disposition than other White juveniles. Methodological Implications The methodological implications of these findings are numerous. First, this study demonstrated the need to assess both detention and disposition decisions. The findings were quite different depending on the stage of processing examined. Second, detention should be accounted for when examining later decisions. Detention had a significant effect on the disposition decision. Third, data regarding the community in which a youth lived should be added to models assessing contextual effects. This study showed that community context aided in the prediction of outcome, especially detention. Fourth, not only is it important to examine racial disparity by including race as a variable, much information is gained from examining racial groups separately. Fifth, previous studies left interaction effects among race and other predictor variables unexamined. This study showed that there were significant interaction effects between race, family structure, and community composition, which predicted detention. Finally, this study showed that the general predictive power of the models of detention and disposition were low. The small effects 128 sizes found in this study were similar to previous research (Kurtz, Giddings, & Sutphen, 1993; Kempf, Decker, & Bing, 1990; McCarthy & Smith, 1986). There are obviously other variables which researchers are not measuring which have an influence on decision-making. These variables could range from a youth's physical appearance, to space availability in programs, to individual philosophical differences among court staff. Theoretical Implications In attempting to view detention and disposition separately, some may explain the finding of no direct race effects in the disposition decision as supporting the consensus perspective. On first inspection it seemed that Tittle's (1994) interpretation of the consensus perspective best fit these data on disposition. He stated that there will be differential treatment in the juvenile court based on the amount of deviation from societal norms. Thus, the current offense was not as salient as how well the youth conformed to societal expectation. Corresponding to Tittle's interpretation of consensus theory, the findings showed that the significant predictors of disposition were what the court had tried before, the age of the youth, and the amount of personal problems. It is also easy to see why some studies have found that White youths had more severe dispositions (Kempf, Decker, & Bing, 1990). This was probably due to the fact that many of the previous studies did not adequately operationalize family and social problems and thus could not determine the effect of race beyond these issues. In this study, being White did predict disposition, until social variables were controlled. The spurious relationship 129 between being White and severity of disposition was influenced by the personal problems of the youths. If juvenile justice decisions are viewed in isolation from each other, the consensus explanation of disposition decisions may seem plausible. However, youths are processed through the juvenile justice system in stages. Each stage is affected by the stage before. One typical scenario is as follows. First, a youth would be apprehended by the police. The police then make several other discretionary decisions, one of which is referring the youth to court and/or taking him or her to a detention facility. If the police officer takes the youth into custody and transports to detention, the intake worker at the court must approve. After a youth is detained, Michigan law requires that the youth must have a formal hearing within 24 hours in order to be released from detention or be held for a longer period of time. Thus, a youth who is detained is automatically processed in a formal manner. In this scenario, it is clear that early biases would then have an impact on later stages. When viewing juvenile justice as a process, the findings for both decisions support the conflict perspective. The conflict perspective framework will be used first to discuss detention decisions, followed by dispositional decisions. The findings regarding detention provided considerable support for the revision of the conflict perspective offered by Hawkins (1987). Hawkins' revision suggested that the more threatening the group, the more subject they were to social control. One way that threat can be operationalized was the higher percentage of Blacks in 130 the population. Traditional conflict theory would implicate low resources of the group as the their vulnerability to social control. These analyses did not support the idea that lower resources in the community impacted juvenile court decisions. In this study, the higher the percentage of Blacks in the census tract, the more likely it was that youths would be detained. Even for analyses run separately for White youth, the higher the proportion of Blacks in the community, the greater the risk of detention. This could be interpreted as "ecological contamination in which all persons encountered in bad neighborhoods are viewed as possessing the moral liability of the area itself" (Smith, 1980). A "bad“ neighborhood seems to be defined as one possessing a high concentration of Blacks, rather than a an area of higher poverty. However, these issues are related. The "ecological contamination“ explanation in conflict theory may be more appropriate to police decisions than court decisions as the court has less direct contact with the community. The interaction between race/ethnicity and community composition (Whites less likely to be detained when they lived in White neighborhoods, but detention for Whites is almost as likely as for Blacks in predominantly Black neighborhoods), make the explanation of police bias even more plausible. From these data, however, there is no way to definitively show whether the court followed the recommendation of the police (which would be racially differentiated) or whether the court intake workers make racially differentiated decisions. 131 Other studies have concluded that the police decision to detain a youth is racially biased (Wordes & Bynum, 1995). There are many factors which may be incorporated into the police decision. Observational studies, most notably Pillavin and Briar (1964), have found that demeanor and appearance were associated with police decisions. Further, Black youths were perceived by officers to have a more negative demeanor and be less appropriately dressed. In addition, more recent research has noted the impact of the complainant's desire for formal court involvement on the police decision to further the case. Some complainants may be more likely to request such action for offenses involving youths of color (Smith & Visher, 1981; Black & Reiss, 1972). Complex indirect relationships involving such factors as demeanor and social and family situation may then interact with race/ethnicity to exert an influence on police decisions. Another explanation may be that the police were more likely to patrol areas with a higher population of Black youths (Sampson, 1986). The finding in the present study which indicated that community composition influenced the detention decision supports the notion that the racial composition of the neighborhood one lives in plays a significant role in the decision to detain a youth. Once detained, a youth is more likely to receive a harsher disposition. Thus, the conflict perspective was supported by the direct relationship between detention and disposition. The bias amplification effect was apparent (Bridges & Crutchfield, 1988; Farrell & Swigert, 1978). Thus, even though direct effects 132 of race were not found in predicting disposition, differential treatment was evident at this stage partly due to the early detention decision. Another theoretical orientation which may shed light on the findings is called the bureaucratic perspective. This perspective may help guide future research aimed at creating interventions to reduce racial disparity. The bureaucratic perspective suggests that organizations will attempt to do what is necessary to survive and to grow. For example, Pawlak (1977) found that the more bed space available, the more youths were sent to detention. He also stated that "juveniles with similar characteristics have a higher probability of detention in courts with a detention facility than in courts without a detention home" (p.154). Possessing a detention facility is just one component of juvenile court structure and process. There is a large degree of non-uniformity in the justice process across juvenile courts (Stapleton, Aday, & Ito, 1982). In clustering various characteristics of metropolitan juvenile courts, Stapleton, Aday, & lto (1982) found two general typologies of courts: “traditional“ and “due process”. The traditional courts were more treatment focused and handled status offenders, whereas the due process courts tended to be more formal and did not usually handle status cases. Juveniles are thus handled by agencies within different contexts. The bureaucratic perspective suggests that the bureaucracy of urban courts would lead to more formalized procedures, whereas rural courts would be more likely to use more discretion (Bridges, Crutchfield, and Simpson, 1987). Certain procedures in highly bureaucratic courts may serve as 133 institutionalized discrimination against minorities (Kempf, Decker & Bing, 1990; Myers & Talarico, 1986; Miethe 8. Moore, 1986). For instance, how drug cases were handled in a particular court may add to disproportionate representation of minorities (Sampson & Laub, 1993; McGarrell, 1993). The bureaucratic perspective may thus provide a bridge between the conflict perspective at a sociological or macro-level explanation and a more concrete approach to addressing the problem. The critical issue in the conflict perspective of structural inequality stems from the power relationships among groups. Basically, the perspective contends that those in power want to keep control of those who have less power and therefore, define appropriate behavior for people of low status (Quinney, 1977). The role of the middle class is to accept these definitions and enforce them through the justice system. A mechanism for this macro-level conflict among groups has not been proposed. While conflict theorists cite macro-level reasons for disparate treatment, they do not connect the decision made about the individual to macro-level phenomena or suggest interventions to address these problems. The bureaucratic perspective may allow for that link. The structure of the court, the policies of the police departments, and the characteristics of the community in which the youth is located play a great role in decisions made by individual decision- makers. Enforcing the structure of race and class inequality in society comes under the purview of the juvenile court decision-makers who work under the bureaucratic constraints of their particular juvenile justice system. 134 Implications for Intervention and Policy From these analyses and this theoretical framework, it is clear that White youths were being judged by different standards than Black youths in making the detention decision. Both the direct effects of race/ethnicity in detention decisions and the differences in the separate predictive models for Blacks and Whites support this conclusion. While police decisions may play a large role in the detention decision, the youths were still under the jurisdiction of the courts when they were detained. it seems that White youths were being treated in the traditional "parens patriae" method of the juvenile court. The traditional decision process took into account social, family, and age factors as being salient in the decision process. An issue for White youths seemed to be whether or not they needed treatment. However, the data indicated that Black youths were being judged by different factors. The salient variables for Black youths were either legalistic or they were not measured. It was difficult to ascertain which factors were important in the detention and disposition of Black youths because the models were not strong. Thus, factors other than the ones traditionally thought of as important in juvenile court decisions were taken into account. The policy implications of these findings center on the practices of the juvenile court. These findings suggest that the court does not adequately screen police decisions which may have been racially biased. Thus, Black and Latino youths may have begun their entrance into the juvenile justice system in a racially biased manner, and no rectification of that decision was made. 135 Further, this early detention decision became amplified in the disposition decision. There are several types of intervention which may aid in lessening of disparate treatment by race/ethnicity. The intervention can be directed at individuals, courts, or society as a whole. The individual level interventions suggested from this research are mainly educational. If decision-makers are educated regarding the biases in their decisions, they may choose to alter their decision-making process. For instance, if decision-makers know that youths from predominantly Black neighborhoods are more likely to be brought to detention by the police, they may review the case more carefully when making an intake decision. Another example would be to use training to show decision- makers their own biases in making decisions. They may not be aware that basing a decision partly on whether the youth lives in a single parent family adds to minority over-representation in secure settings. Court level interventions may be aimed at changing the bureaucratic goals of the organization. If the distribution of funds for the organization were tied to reductions in minority over-representation, the courts may respond. Courts may institute a review board which would make it necessary for clear justification of each detention and disposition decision of youths from predominantly Black neighborhoods. An alternative to a review board may be hiring one person who serves as a youth advocate. Another court level intervention may be to create non-secure alternatives to detention. At this 136 point, there is no research to support whether the use of non-secure detention influences disposition in the same manner as secure detention. Finally, societal level interventions would most likely occur in the grassroots community. Conflict theory would support the idea of that if justice intervention were taken out of the hands of the people in power, more equitable treatment would occur. One suggestion is to form alternative dispute resolution forums. If a youth does not become involved in the juvenile court system, and is handled within his or her own community, disparate treatment may not be involved. Another possible intervention would be to develop community commissions who have the power to review patterns of decisions in the juvenile justice systems in their communities. According to the conflict perspective, the courts would have no reason to change discriminatory practices unless they have no choice. Limitations While these findings were suggestive of necessary reforms within the juvenile justice system, they are not conclusive. As in any study, there were limitations that must be taken into account. In general, studies using linear regression techniques were not able to account for much of the variance in the decision variables. This study attempted to account for more variables than in previous work, but still was unable to capture many of the variables which predict disposition. This may be due to both methodological and theoretical limitations. The limitations discussed below encompass data restrictions, poorly defined constructs, and theoretical under-development. 137 The first limitation was due to the data set. Although the data collected for this study were much more comprehensive than in most, it was still inherently limited to the information in the case files. Many potentially relevant factors are not routinely captured in case file records. The second limitation centered on the accuracy of the race/ethnicity variable or meaning of the construct. As in almost any study using official records, race/ethnicity was difficult to define. Individual staff members determine the race/ethnicity of the youth by some undefined or possibly varied standard. Some staff members may define youths with Puerto Rican heritage as Black, White, or Latino. It may depend on the surname, language style, or skin color, but there were no well defined categories for staff members to follow. Another limitation that may have inhibited finding significant interaction effects, was the size of the sample. This limitation arose due to the levels of analysis issue. In sampling only one youth per census tract, which was necessary to avoid mixing levels of analysis, the sample size was considerably reduced. Moreover, the sample size was considerably reduced when examining Blacks and Whites separately. Power may have been a problem because the effect sizes were small. A larger sample may have been better able to detect these small effects. Another issue was the difficulty in exploring the inter-relationships among variables of theoretical importance. For instance, it was difficult to determine the effect of race/ethnicity and community factors simultaneously because the 138 census tracts were highly segregated by race/ethnicity and income. This issue is further complicated because the juvenile was not only embedded within a community context, but also within the juvenile court context. Thus, there may be great differences among courts which influenced disposition. This study only contained five court sites, which didn't allow for comparison among courts. Exploratory analysis including site as dummy coded independent variables showed that the court in which the case was handled significantly influenced both detention and disposition decisions. Future research should examine the influence of site on court decisions by including more sites in the study and collecting data reflecting court differences in resources, philosophy, and policy. Future Bpsgarch Two types of research are necessary to help ameliorate racial disparity in juvenile court decision-making. First, further research documenting, describing, and explaining the problem is necessary. This study showed that the examination of racial groups separately is warranted due to the differential emphasis placed on various factors for White youths and Black youths. Further, due to the small effect sizes, these findings suggest that many other variables may play a role in court decisions. Several possibly relevant variables which may be important to measure in future research on racial disparity are: a) the ability of the youth's attorney (Aday, 1986), b) whether the parent/guardian was home when youth was brought to detention (from personal communications with intake workers), c) the demeanor/appearance of the youth 139 and family (Kurtz, Giddings, & Sutphen, 1993), d) the number of detention beds available (Pawlak, 1977), e) choices of programming for the youth (from personal communications with judges and referees), f) funding structure of system (Belknap, 1984). Second, the field is bereft of evaluations of interventions aimed at ameliorating the problem of disparate treatment by race/ethnicity. Several of the intervention strategies listed above could be evaluated using quasi- experimental or experimental designs. The outcomes would range from changing attitudes and knowledge of decision-makers, to actual reduction in discriminatory treatment. The complexity of the system and the myriad factors which affect decision-making make the task of reforming or redefining the juvenile court system seem daunting. While these analyses and theoretical interpretations do not point to easy solutions, the problem deserves critical attention. Allowing discriminatory treatment of minority youth in the juvenile courts is unacceptable in a society which defines justice as equal treatment under the law. The serious consequences of racial disparity in juvenile justice are not only detrimental to individual youths, but to society as a whole. Therefore, the resources of the community, academia, and policy-makers must be used to document, explain, and address this crucial issue. 140 Notes 1. A B-level Standard Metropolitan Statistical Area (SMSA) has a population between 250,000 and 1,000,000 and consists of a metropolitan area including a central city and surrounding communities. 2. The consultation group consisted of professionals from around the state including: judges, police chiefs, defense attorneys, juvenile court administrators, department of mental health officials, department of social services staff, and private non-profit treatment providers. APPENDIX 141 Appendix Sampling Frame and Sampling Weights for Each Juvenile Court COURT 1 Race/Ethnicity Black White Other Male 8.64 4.63 1.00 Female 2.32 2.12 1.67 COURT 2 Race/Ethnicity Black White Other Disposition Male informal 2.82 3.09 1.00 consent 1 .88 2.31 ----- formal 1.00 1.07 1.20 Female informal 1.00 1.00 1.00 consent 1 .22 1 .17 ----- formal 1.00 1.00 ----- COURT 3 - sampling weight for felony offenses was 1.41. 142 Appendix (continued) COURT 4 Race/Ethnicity Black White Other Disposition Male informal 1 .00 2.43 1 .00 l formal 1.64 1.15 1.50 Female informal 1 .17 1 .18 ----- formal 1 .00 mu 1 .00 COURT 5 Race/Ethnicity Black White Other Male 2.61 2.70 1.00 Female 1 .00 1 .25 ---- REFERENCES 143 References Aday, DP. (1986). Court structure, defense attorney use, and juvenile court decisions. The Sociological Quarterly, 27(1), 107-119. Belknap, J. (1984). The effect of local policy on sentencing patterns of state wards. Justice Quarterly, 1(4), 549-561. Bell, 0., Jr., & Lang, K. (1985). The intake dispositions of juvenile ‘offenders. Journal of Research in Qrime and Delinguency, 22(4), 309-328. Bishop, D.M., & Frazier, CS. (1990). A study of race juvenile justice processing in Florida. Report, University of Florida. Bishop, D.M., & Frazier, CS. (1988). The influence of race in juvenile justice processing. Journal of Researclm Crime and Delinguency, 25(3), 242-263. Black, D.A. & Reiss, A.J. (1978). Police control of juveniles. American Sociolmgzal Review, 35. 63-77. Black, T.E., & Smith, GP. (1981). A preliminary national assessment of the numbers and characteristics of juveniles processed in the juvenile justice system. Report of the National Juvenile Justice Assessment Centers (pp. 1-192). Bridges, 6.8. 8 Crutchfield, RD. (1988). Law, social standing and racial disparities in imprisonment. Social Forces, 66, 699-724. Bridges, Crutchfield, & Simpson (1987). Crime, social structure, and criminal punishment: White and nonwhite rates of imprisonment. SM! Problems 34, 345-360. 144 Bridges, 6.8. 8 Myers, MA. (1994). lnegualfiy‘ , crime, 8 social control. Boulder, CO: Westview Press, Inc. Bynum, T.S., Wordes, M. 8 Corley, C.J. (1993). Disproportionate representation in juvenile justice in Michigan: Examining the influence of race and gender. Technical Report. East Lansing, MI; Michigan State University. Chambliss, W.J. 8 Seidman, R. (1982). Law, order, and power. Reading, MA: Addison-Wesley Publishing Co. Cohen, J. 8 Cohen, P. (1983). Applied muniple regression/correlation analysis for the behyioral sciences. Hillsdale, N.J.: Lawrence Erlbaum Associates, Publishers. Cohen, L. E. 8 Kluegel, JR (1979). The detention decision: A study of the impact of social characteristics and legal factors in two metropolitan juvenile courts. Social Forces, 58, 146-161. Dannefer, D. 8 Schutt, R.K. (1982). Race and juvenile justice processing in court and police agencies. American Journal of Sociology, 87, 1113-1132. Durkheim, E. (1964). The division of labor in society. New York: Free Press. Eisenstein, J., Flemming, R.B., 8 Nardulli, PF. (1988). The contours of justice: communities and their courts. Boston, MA: Little, Brown, 8 Co. Fagan, J., Slaughter, E., 8 Hartstone, E. (1987). Blind justice? The impact of race on juvenile justice process. Crime and Delinguency, 33(2), 224-258. 145 Farrell, R. 8 Swigert, V. (1978). Prior offense record as self-fulfilling prophecy. Law and Society Review, 12, 437-453. Feld, BC. (1991). Justice by geography: Urban, suburban, and rural variations in juvenile justice administration. Journal of Qriminal Law and Qriminology, 82, 156-210. Fenwick, OR. (1982). Juvenile court intake decision making: The importance of family affiliation. Journal of Criminal Justice, 19(6), 443-453. Frazier, CE. 8 Bishop, D.M. (1985). The pretrial detention of juvenile and its impact on case dispositions. Journal of Criminal Law and Criminology, _7__§, 1132-1152. Frazier, C.E., Bishop, D.M. 8 Henretta, JC. (1992). The social context of race differentials in juvenile justice dispositions. The Sociol ical uarterl , 447-458. ' Frazier, C.E., 8 Cochran, JG. (1986). Detention of juveniles: Its effects on subsequent juvenile court processing decisions. Journal of QgiminaLLayy and Qriminology, 76(4), 1132-1152. Hawkins, D.F. (1987). Beyond anomalies: Rethinking the conflict perspective on race and criminal punishment. Social Forces, 65, 719-745. Hawkins, D.F., 8 Pope, CE. (1993). Race and criminal/juvenile justice processing: Implications of site specific research. Paper presented at the Western Society of Criminology, Monterey, CA. Hindelang, M.J. (1978). Race and involvement in common law personal crimes. American Sociological Review, $3, 993-1009. 146 Huizinga, D. 8 Elliot, 0.8. (1987). Juvenile offenders: Prevalence, offender incidence and arrest rates by race. Crime and Delinguency, 33, 206- 223. Johnson, J.B. 8 Secret, PE. (1990). Race and juvenile court decision making revisited. Criminal Justice Policy Review, 4, 159-187. Kempf, K., Decker, S., 8 Bing, R. (1990). An analysis of apparent disparities in the handling of black youth within missouri's juvenile justice systems. Technical Report. St. Louis, University of Missouri-St. Louis. Krisberg, B. 8 Austin, J. (1993). Reinventing juvenile justice. Newbury Park, CA: Sage Publications, Inc. Krisberg, B., DeComo, R., Rudenstine, S. 8 Rosario, D. (1995). Juveniles taken into custody research program. 1994 Annual Report. National Council on Crime and Delinquency. Krisberg, B., Schwartz, I., Fishman, G., Eisikovits, E., Guttman, E. 8 K. Joe (1987). The incarceration of minority youth. Crime and Delinguency, 33, 173-205. Kurtz, P.D., Giddings, MM, 8 Sutphen, R. (1993). A prospective investigation of racial disparity in the juvenile justice system. Juvenile and Family Qourt Journal, 44, 43-59. Liska, A.E. 8 Tausig. (1979). Theoretical interpretations of social class and racial differentials in legal decision-making for juveniles. The Sociological Quarterly, 20, 197-207. 147 Liska, A.E. 8 Chamlin, MB. (1984). Social structure and crime control among macrosocial units. American Journal of Sociology, 90, 383-395. Marshall, l.H. 8 Thomas, CW. (1983). Discretionary decision-making and the juvenile court. Juvenile and Family Qoug qumal, 34, 47-60. Maplnfo. (1993). Maplnfo Corporation, Troy, NY. McCarthy, BR. (1991). Social structure, crime, and social control: An examination of factors influencing rates and probabilities of arrest. Journal of Criminal Justice 19, 19-29. McCarthy, BR, 8 Smith, BL. (1986). The conceptualization of discrimination in the juvenile justice process: The impact of administrative factors and screening decisions on juvenile court dispositions. Criminology, 24(1), 41-64. McGarrell, E. (1993). Trends in racial disproportionality in juvenile court processing: 1985-1989. Crime and Delinguency, SQ, 29-48. Miethe, TD. 8 Moore, CA. (1986). Racial differences in criminal processing: The consequences of model selection on conclusions about differential treatment. Sociological Quarterly, 27, 217-237. Myers, MA. 8 Talarico, SM. (1986). Urban justice, mral justice? Urbanization and its effect on sentencing. Criminology, 24, 367-391. Pawlak, E. (1977). Differential selection of juveniles for detention. Journal of Research in Qn’me and Qelingueng, 14(2), 152-165. Pillavin, l. 8 Briar, S. (1964). Police encounters with juveniles. Amgrican Journal of Spgiology, SQ, 206-214/ 148 Platt, A. (1969). The child savers: The invention of delinguency. Chicago: Chicago University Press. Pope, CE. 8 Feyerherm, W.F. (1992). Minorities and the juvenile justice syst_erp. Report for the US. Department of Justice Office of Juvenile Justice and Delinquency Prevention. Pope, C.E., 8 Feyerherm, W.F. (1990). Minority status and juvenile processing. Qriminal Justice Abstracts, 22(2), 327-336 and 22(3), 527-542. Quinney, R. (1977). Class, state and crime: on the theory and practice of criminal justice. David McKay Company, Inc.: New York. Rothman, DJ. (1980). Conscience and convenience: the asylum and its alternatives in progressive america. Harper Collins Publishers. Sampson, R. (1986). Effects of socioeconomic context of official reaction to juvenile delinquency. American Sociological Review, 51, 876-885. Sampson, R., 8 Laub, J.H. (1993). Structural varations in juvenile court processing: Inequality, the underclass, and social control. Law and Socier Review, 27, 285-311. Sarri, R. (1974). Under lock and ke : uveniles in ' iI nd detention. National Assessment of Juvenile Corrections. Ann Arbor, MI, University of Michigan. Smith, DA. (1986). The neighborhood context of police behavior. In Reiss, A., and Tonry, M (Eds.). Qrime and justice; An annual review of research (Vol 8), Communities and crime. Chicago, IL: University of Chicago Press. 149 Smith, D.A. 8 Visher, CA. (1981). Street level justice: Situational determinants of police arrest decisions. Socia_l_Pro_blem§,_29L2_), 167-177. Stapleton, W.V., Aday, DP. 8 Ito, J.A. (1982). An empirical typology of American metropolitan juvenile courts. American Journal of Sociology, 88, 549- 564. 1 Thomas, CW. 8 Sieverdes, CM. (1975). Juvenile court intake: An analysis of discretionary decision-making. Criminology, 12, 413-432. Thomberry, T.P. (1973). Race, socioeconomic status and sentencing in the juvenile justice system. Th Journal of Criminal aw riminolo and Police Spienc_e. 64. 90-98. Tittle, CR. (1994). The theoretical basis for inequality in formal social control. in Bridges 8 Myers (Eds) Inegual'mj, crime, and social control. Boulder, CO: Westview Press, Inc. Tittle, CR. 8 Curran, D.A. (1988). Contingencies for dispositional disparities in juvenile justice. Social Forces 67, 23-58. Turk, AT (1969). Criminality and legal order. Chicago: Rand McNally. Wilbanks, W. (1987). The mflh of a racist criminal justice system. Monterey, CA: Brooks/Cole. Wolfgang, M. Figlio, R.M., Tracy, PE, 8 Singer, SJ. (1985). Ih_e national survey of crimLseveritv. Washington, DC: US. Government Printing Office. 150 Wordes, M., 8 Bynum, TS. (1995). Policing juveniles: Is their bias against youth of color? in K. Kempf, C. Pope, 8 W. Feyerherm (Eds.), Minorities in Juvenile Justice. Beverly Hills: Sage. "Illlllllllllllllllllllllls