Citation
Female offenders in state prisons

Material Information

Title:
Female offenders in state prisons the analysis of racial effects and strains during the socialization process
Creator:
Daugherty, Stacey M
Publication Date:
Language:
English
Physical Description:
82 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Female offenders -- United States ( lcsh )
Prisons -- States -- United States ( lcsh )
Racism -- United States ( lcsh )
Female offenders ( fast )
Prisons -- U.S. states ( fast )
Racism ( fast )
United States ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 76-82).
Statement of Responsibility:
by Stacey M. Daugherty.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
55471792 ( OCLC )
ocm55471792
Classification:
LD1190.L66 2003m D38 ( lcc )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

FEMALE OFFENDERS IN STATE PRISONS: THE ANALYSIS OF RACIAL EFFECTS AND STRAINS DURING TIIE SOCIALIZATION PROCESS. by Stacey M. Daugherty B.A., University of Colorado at Denver, 2002 A thesis submitted to the University of Colorado at Denver partial fulfillment of the requirements for the degree of Master of Arts Sociology 2003

PAGE 2

This thesis for the Master of Sociology degree by Stacey Daugherty been approved by Date

PAGE 3

Daugherty, Stacey (M.A., Sociology) Females Offenders in the State Prisons: The Analysis of Racial Effects and Strains During the Socialization Process. Thesis directed by Professor Yili Xu ABSTRACT The purpose of this study is to determine the racial effects and that occur during the process of socialization of females in state prisons. This study was guided by the theoretical understanding of socialization, feminist criminology, Sutherlands' differential association, and Agnew's general strain theory. In this study, I utilized secondary data from the Survey of Inmates in State and Federal Correctional Facilities, from the year 1997. For analysis, structural equation modeling was utilized for the whole model (N = 1038), and two additional sub-models for whites (N = 467) and non-whites (N 571) were created to determine the racial differences. The results revealed that racial differences exist among female state offenders in the areas of abuse victimization, substance use, frequency of arrest, and the initial age of arrest. The most compelling fmdings revealed that peer deviant behavior had a significant effect in all categories; initial age of arrest, frequency of arrest, substance abuse, and abuse victimization for non-white respondents. However, for the white sample peer deviant behavior did not have significant effect in all the categories mentioned above. Furthermore, childhood environment had a significant effect on initial age of arrest, but did not for the non-white sample. This abstract accurately represents the content of the candidate's thesis. I recommend its publication. Signed Yili Xu

PAGE 4

DEDICATION I would like to dedicate my thesis to my wonderful fiance, Brent Donaldson for his continuous love and support. Thank you for your patience and understanding throughout my academic career. I would also like to thank my parents for their support and love throughout my life.

PAGE 5

ACKNOWLEDGEMENT I want to give like to give special thanks to my advisor Yili Xu, for his endless support, guidance, and patience throughout my academic career. Thanks to him I have been able to learn a vast amount of knowledge and he has taught me that it is through hard work and determination in life that one is able to eventually reap the rewards. I would also like to acknowledge Candan Duran-Aydintug for academically challenging me, if it wasn't for her, I would have never gone on to get my masters in Sociology. Thank you for all of support and constantly believing in me. Additionally, I would like to thank Jerry Williams, for being such a wonderful listener and for his endless support. These three instructors have truly enhanced my experience and learning process at the University of Colorado at Denver, and for that I will be forever grateful.

PAGE 6

Figures Tables CHAPTER CONTENTS 1. IN1'R.ODUCTION .................................................................................. 1 2. LITERA TORE REVIEW ....................................................................... 6 Influences of Socialization ....................................... ..................... 7 Influence of SES, Education, and Income ....................................... 8 Influence of Parental Roles .............................................................. 9 Influence of Peers .......................................................................... 11 Impact of Substance Abuse and Abuse Victimization ................... 12 Racial Effects .......................... ................................... .................. 16 3. TIffiORETICAL BACKGROUND ....................................... .............. 20 History ............................................................................................ 20 Feminist Theory ............................................................................. 23 Differential Association ................................................................. 25 General Strain Theory .................................................................... 26 Hypotheses ........................ ............................................................ 30 vi

PAGE 7

4. DATA AND ME11IODS ..................................................................... 32 Data ............................................................................................ 32 Statistical Procedures ..................................................................... 34 Causal Modeling ................................................................ 34 Direct, Indirect, and Total Effects ..................................... .35 Latent and Manifest Variables ........................................... 35 Fit Indexes .......................................................................... 36 Model Specification ....................................................................... 3 8 Measurement Model .......................................................... 38 Structural Model ................................................................ 42 Testing Strategy ................................................................. 45 5. RESULTS AND DISCUSSION .......................................................... .47 Whole Model ............................................................................... 47 Abuse Victimization .......................................................... 50 Substance Abuse ................................................................ 51 Frequency of Arrest ........................................................... 53 Age of First Arrest ............................................................. 54 Sub-Models .................................................................................. 55 White Model ...................................................................... 56 Abuse Victimization ................................................... 58 Substance Abuse ......................................................... 58 vii

PAGE 8

Frequency of Arrest .................................................... 59 Age of First Arrest ...................................................... 59 Non-White Model .............................................................. 60 Abuse Victimization ................................................... 62 Substance Abuse ......................................................... 62 Frequency of Arrest .................................................... 63 Age of First Arrest ...................................................... 63 Comparison of Sub-sample Models ............................................ 63 Significant Findings and Contributions ....................................... 69 Limitations and Future Research ................................................. 71 REFERENCES ............................................................................................ 76 viii

PAGE 9

FIGURES Figure 1. The Effects of Strains and Exposure to the Negative Stimuli during Socialization Process of Female Offenders State Prisons ..................... 44

PAGE 10

Table TABLES 1. Estimated Direct, Indirect, and Total Effects of Strains and Negative Stimuli of Female Offenders for the Whole Sample .......................... .49 2. Estimated Standardized Direct, Indirect, and Total Effects of Strains of Female Offenders for the White Sample ......................................... .57 3. Estimated Standardized Direct, Indirect, and Total Effects of Strains and Negative Stimuli of Female Offenders for the Non-White Sample ............................................................................... 61 4. Comparison of White and Non-White Models ..................................... 64 x

PAGE 11

INTRODUCTION Females are becoming increasingly involved in criminal activities, although the majority of crimes in the United are committed by males. The Bureau of Justice Statistics reported that 1.8 million females were arrested in 1986; the number climbed up to 2.3 million in 1991; and in 1998, the number jumped to 3.2 million (Greenfield Snell, 1999). From 1991 to 1998, .there has been a 38% increase in the number of females arrested. addition, more than 980,000 women were under correctional supervision in 1998 (Greeield Snell, 1999). The trend of females' increasing engagement in criminal activities presents a serious challenge not only to the criminal justice. system and law enforcement agencies, but also to social scientists, criminologists and law scholars, who are constantly pondering the and patterns underlying this phenomenon. The fundamental q1l:estions most people would ask regarding the recent increases of female criminality are: What are the major causes and conditions that underlie female criminality? When studying female offenders, the most critical problem researchers are facing the inadequate amount of research relative to female offenders. As a result, too often, conclusions that have been drawn about male offenders are used to

PAGE 12

explain and predict female offenses. The goals of this study are to bridge the gap between the studies of male and female offenders, to pinpoint the underlying factors that contribute to the increasing occurrence of female criminality, and to understand how these factors affect the age and recidivism an individual enters the criminal justice system. Strategy wise, the current study intends to examine female offenders from a sociological point of view and integrates several major theories and concepts that are pertinent to female offenders. In order to better comprehend this phenomenon,one must fully understand the process of socialization and how it pertains to females in the United States. One component of socialization is gender roles. Current literature asserts that in addition to biological differences, socially constructed differences exist between males and females, and "women's roles are learned and socially determined" (Klein, 1984, 3). It is widely recognized in the sociology field that gender is socially constructed (Henderson, 1994, Kimmel, 2000, Scott, 1986). 2000Anderson summarized: Gender is a social, not biological concept, referring to the entire array of social patterns that we associate with women and men society. Being "female" and "male" are biological facts; being a woman and a man is a social and cultural process-one that is constructed through the whole array of social, political, economic, and cultural experiences in a given society (p. 20). From this, one can conclude that gender differences between males and females are characteristics constructed by society. Through on-going socialization, society's gender role ideology has been embedded in-social institutions, such as the family, 2

PAGE 13

religion, and education, and internalized by individuals. Throughout the process of socialization, gender differences in criminality have been shaped. The worship of masculinity in our culture and the connection between masculinity and crime exemplify accurately how traditional gender role socialization could negatively affect individuals' behavior. A full comprehension of femininity and crime requires a scrutiny of the relationship between masculinity and crime. Michael Kimmel, a leading researcher the area of sex and gender, suggested that in order for boys to evade the stigma of being labeled a "sissy," boys are socialized to be tough and assert their masculinity by means of violence and dominance, (Kimmel, 2000). Our society promotes and enforces the concept of hegemonic masculinity (Flavin, 2001, Kimmel, 2000). Hegemonic masculinity denotes that males are socialized to be strong, dominant, aggressive, and emotionless, so that they can live up to society's traditional standard of being a "real man." On the other hand, the "ideal woman" that society promotes should be tender, submissive, subordinate, and emotional. This traditional way of gender role -molding may have unintentionally contributed to the historical discrepancy between the numbers of male and female offenders. our postmodem society, traditional ideologies are gradually losing their dominance. Individuaiism and gender equality have been major political and ideological themes, and significant changes have occurred to social institutions. As a result, society's standards for women have shifted. As women become more and more independent and equal to men in all social dimensions, it should be no surprise 3

PAGE 14

to see the diminishing gap in the behavioral patterns between male, and female, and consequently in the numbers of men and women behind the bars. This could be one of the reasons underlying the trend of female criminality in the United States. Given the changed social environment, women and men still face different social conditions and experience different sets of strains and stimuli. Females may well be exposed to different sets of social, political, and economical strains and stimuli and respond to them differently compared to their male counterparts. And these different sets of strains, negative stimuli, and response patterns may provide important explanations for the increase of female criminality. By utilizing the framework of socialization as a key explanatory factor, I hope to better understand the major strains that female offenders have been exposed to during their socialization process .. The current study pays particular attention to the impact of race on female criminality. The reason for that is (1) it's long been observed that the numbers of minority inmates are disproportionately high in almost all federal or state prisons. A direct interpretation of the statistics is that being a minority is more likely to be incarcerated. (2) One study in particular shows that race is a significant predictor of criminal arrest (Walker, Spohn, DeLone, 1999). Therefore, another analytical task of this study is to test the effect of race on criminal arrest in the population of female offenders to determine if there are any differential patterns of causality for white and nonwhite female offenders. This test becomes particularly important, since the significance of race founded by previous studies

PAGE 15

were mostly based on male prison populations. It is important to verify if the findings of these studies can also hold for the population of female offenders. The aim of this study is to gain a more systematic comprehension of the key agents in the socialization process, the important strains and negative stimuli, and determine the racial effects involved in female criminality. 5

PAGE 16

CHAPTER 2 LITERATURE REVIEW Research on female offenders is a relatively new area of interest, since female criminality was not a primary focus of criminologists in the past. As a result, studies specifically devoted to female offenders are few and sporadic, compared to the large volume of research directed to male criminality. This presents a challenge to the current study to search for relevant literature and to make a connection to past research. There is not a vast amount of empirical studies that specifically test specifically the negative strains in conjunction with socialization factors, to address the issue of female offenders. However, this is also an opportunity for the present research to dig into the past, and find the potential links between previous knowledge and behavioral patterns of current female offenders. essence, I will attempt to bridge the gap between the past research and current reality. This study tackles important social, behavioral, and demographic factors that could explain the formation and development of female deviance. This review will focus on the following the influences of socialization, socioeconomic background, education, parents, peers, and victimization as they pertain to deviant behavior. The research studies reviewed that fall into each of these dimensions contain samples that focused on male and female respondents or specifically on female respondents. 6

PAGE 17

Influences of Socialization There is some controversy over whether crime is caused by biological or sociological factors. Barkan addressed three main problems with biological explanations for crime. First, he focused on how the-diversity of crime is too vast to be explained by biological factors. Can biological explanations then explain violent and non-violent crimes? The second argument that Barkan made is that there are methodological problems with the studies on biological explanations of crime. Often the sample size is very small too generalize any findings. Third,he mentioned the problem of "group rate differences." He indicates that if biology is the cause of crime, how can one explain that the United States has a much higher crime rate comparison to other countries (Barkan, 1997, p. 137). One would expect crime rates to be held constant in all countries if biology was the key to explaining the causes of crime. For the reasons mentioned above, I take the stance that crime is caused by sociological factors, and that female (and male) offenders were not biologically preordained to display deviant behavior. Rather, deviant behavior is learned through socialization processes. The specific types of deviance they acquired, the frequency and severity of the delinquency they display, and situations which they act a deviant or unlawful manner all depend on the environment in which the females were socialized and socialization agents with which they interact. Individuals from the primary socializing agents; family members, peers, and school mates are especially crucial in tenns of what kinds of influence, normative expectations, and behavioral models

PAGE 18

individuals are exposed to. These agents influence how strongly individuals can resist abnonnal desires and temptations, and what forms of behavior they can eventually develop. According to James Wilson (1983) a steadily growing body of evidence suggests that the family affects criminality and that its effect, at least for serious offenders, is lasting. Influence of SES, Education and Income The 1960's Presidential Commission on Law Enforcement and the Administration of Justice pointed out that the root causes of crime are distorted, racist, and are economic injustices. The American Correctional Association conducted a national survey of state prisons in 1990 and concluded that approximately 60% of the female inmates were on welfare prior to their arrest (Miller, 1998). Previous research has found that young adults who are habitual substance users are more likely to be constantly changing jobs, or unemployed (Hartnagel, T.F., 1992, as cited in Baron, 1999). Many female offenders, who are charged with drug offences and minor property offences, have a low socioeconomic status and are pushed into selling drugs and/or stealing to make a living. has been assumed that poverty is a general factor involved in criminal behavior (Wright, Caspi, Moffitt, Miech, & Silva, 1999). Additional research suggest that poor children are more likely to represent the majority of those that commit an offence at young age (Duncan, Brooks-Gunn, & Klebanov, 1994). The previous studies support that crime and deviant behavior are more 8

PAGE 19

likely to occur among people with low socioeconomic backgrounds, poor education, high unemployment rates, and low income. However, other researchers have fOlmd that poverty is not a significant predictor for individual's negative outcomes (Duncan & Brooks-Gunn, 1997, Mayer & Jencks, 1989). Influence of Parental Roles Parents are one of the primary, most influential socializing agents in a child's life. Previous research indicates that children with deviant parents are more likely to become involved in delinquent acts (Barnes, Farrell-& Cairns, 1986). The Bureau of Justice Statistics reported that approximately 47% of the female inmates reported having an immediate family member who had spent time in prison (Snell, 1991). This statistic warrants one to conclude that parental deviant influence may be a key indicator when examining juvenile delinquency. In addition, has been supported that parental deviant behavior, such as substance abuse disorder (PSUD), increases male and female adolescents' risk of becoming involved in substance abuse and associating with peers that abuse substances (Hoffman & Su, 1998). Previous studies also suggest-that parental deviant behavior such as drugs or alcohol abuse has a significant effect on their children's delinquency (Farrington, 1990, as cited in Wallace, Goddard, Goff, & Melancon, 2000). These results show that deviant parental behavior is an important factor to consider when examining offenders. 9

PAGE 20

There have been conflicting reports on whether children raised in single-headed households are more likely to display deviant behavior in comparison to children raised by both parents. Some studies have found that an individual who was raised in a single-headed household may have an increased chance of becoming more prone to deviant behavior (Barnes & Farrell, 1992, Barnes, Farrell & BaneIjee, 1994; Thomas, Farrell, Barnes, 1996). However, additional studies have found that children raised in single-headed households are no more likely than individuals that are raised with both. parents to exhibit deviant behavior (Turner, Irwin, & Millstone, 1991, Watts & Watts, 1991). Additional studies have uncovered that children raised single headed households are more likely to experience negfltive outcomes in life, such as low academic achievement (Barnes, Farrell & Barnes, 1996, Dornbusch et al., Garasky, 1995, Newcomer Udry, 1987). Furthermore, research has shown that adolescent's who live with both biological parents are less likely to participate in acts of deviance (Thomas, Farrell, Barnes, 1996). Often the consequence of living in a single-headed household is a low socioeconomic status, especially, if it is a female headed household, since females on average earn lower salaries than males (O'Hare, Pollard, Mann & Kent, 1991). Reports indicate that two-thirds of female-headed households are poor; black female-headed households have a median income of $12, 000 in comparisons to white female -headed households who average $19,000 annual income (O'Hare, Pollard, Mann & Kent, 1991). Another study concluded that single headed homes often contribute to lower socioeconomic status, which may be

PAGE 21

indirectly related to the delinquency (Li Wojtkiewicz, 1992). Furthennore, the differences in the outcome of children raised in single-headed households experiences are often caused by a change in family structure, resulting from racial and economic inequalities (Dornbush, et aI., 1985). Additionally, parents' socioeconomic status plays a significant role in the process of child socialization. For example, the quality of neighborhood which a family resides is usually based on the parents' financial situation. Moreover, neighborhood stratification is mostly detennined by residents' personal financial worth. Studies have shown that the area of residence influences who the child's peers will be (Colvin Pauly, 1983). And since a child's peers have a major impact on his or her behavior, the location ofa family'S residence is a strong factor in shaping children's lives. Influence of Peers Peers have been found to have a compelling influence on an individual's actions and behaviors (Barnes & Farrell, 1992). As mentioned previously, parents' social status influences the type of peers that a child interacts with on daily basis. Prior research indicates that the type of peer influence an individual is exposed to may detennine whether that individual displays deviant behaviors throughout his/her life (Akers, 1985, Flom, 2001; Urberg, 1992; Warr Stafford, 1991). This implies that an individual who is associated with deviant peers is more likely to exhibit similar

PAGE 22

behaviors. contrast, if an adolescent is exposed to peers with positive influences, the person is more likely to act in a similar manner and abide by society's nonns and laws. Supplementary studies have concluded that peer influence has a strong correlation to individual alcohol use (Ary, Tildesley, Hops, Andrew, 1993; Urberg, Degirmencioglu, Pilgrim, 1997). Based on this research, one would expect to fmd that individuals, who associate with peers that consume alcohol regularly, would also partake in this behavior. addition, literature suggests that for both male and female adolescents, the influence of peer drug abuse is strong predictor of individual drug use (Akers, Krohn, Lanza-Kaduce, Radosevich, 1979, Aseltine 1995, Hoffman & Su, 1998). The previous research mentioned gives compelling support that peers have profound impact on whether adolescents become involved deviant behavior. Impact of Substance Abuse and Abuse Victimization A person's substance abuse could be an indicator of the behavioral patterns of their family members and/or her peers, and their willingness and ability to conform to the commonly accepted social nonns and to exercise self-control. is also an important predictor of the likelihood of that person to violate other social rules and to break laws. Research shows that substance abuse and crime are intricately linked (Zawitz, 1992). The 1999 Bureau of Justice Statistics reported that female offenders accounted for only 14% of all violent crimes committed from 1993 through 1997 (Greenfield 12

PAGE 23

Snell, 1999). Therefore, males are committing the majority of violent crimes, so one can conclude that overall, males and females are committing different types of crimes. It was reported that the majority of female prisoners have been convicted of drug offences, accounting for more than half of the prison grOWth (Snell, 1991). In state prisons 34% of the female offenders were charged with drug offences, therefore the majority females are committing non-violent, drug-related crimes (Greenfield Snell, 1999). One must also take abuse victimization and the cycle of violence into consideration when examining strains and conditions that female offenders have undergone. An area that is becoming increasingly studied is the impact of sexual or physical abuse female offenders have experienced on their criminal behavior. Victimization represents an important source of negative stimuli that may have occurred during socialization processes offemale"offenders. has been estimated that 34% of all girls will experience some form of abuse before adulthood (Benson, 1990). The Bureau of Justice Statistics reported that approximately 6 in 10 women in state prisons have been physically or sexually abused in their lives (Greenfield Snell, 1999). Research suggests that female inmates had experienced more physical, sexual, and emotional abuse as a child, in comparison to males (McClellan, Farabee, Crouch, 1997). Furthermore, research has shown that females subjected to sexual and/or physical abuse may increase the probability that those individuals will commit crimes throughout the course of their lives (Daly, K. 1992; Gaarder, E., Belknap, J.

PAGE 24

2001, Gilfus, 1992, Widom, 1989, Widom, 1996). These studies support why over half of the females incarcerated have been abused, and therefore, abuse victimization is a key indicator when examining the causes of female offenders. Furthermore, research supports that females who have experienced abuse are more likely to self medicate with drugs and alcohol (Beckman, 1984, Toray, Coughlin, Yamaguchi Kandel, 1991). Another study specifically supported that an individual that was sexually abused as a child is more likely to self-medicate with substances (Miller, Downs, Gondoli, Keil, 1987). Moreover, research shows that childhood victimization is correlated to their use of drugs and alcohol later life (Blane, Miller, Leonard, 1988, Miller, Downs, 1993, Rose, Peabody, Stratigeas, 1991). ResearchersToray and Coughlin (1991) found that childhood victimization is directly related to the increase in individuals' illicit drug use. Additionally, a longitudinal study concluded that childhood victimization increases the risks of an individual being arrested as a juvenile and later in life as an adult (Widom, 2000). Cathy Spatz Widom found that children who experienced abuse were involved in criminal behavior one year earlier than respondents who were not abused (Widom, 2000). Widom also concluded that children who experience abuse are often more likely to be frequently arrested (Widom, 2000). Widom examines the cycle of violence, which suggests an individual has experienced violence or abuse as a child, then they are more likely to display deviant behaviors 14

PAGE 25

later in their life. Abuse may lead females into a life of substance use, an earlier entrance into the criminal justice system, and/or into a cycle of recidivism. Over half of all women in state prisons have experienced negative strains, such as physical and, or sexual abuse in their lives (Greenfield Snell, 1999). Whether these strains were related to income, education, childhood environment, peer or parental behaviors, or a combination of them, it is likely that these factors laid the foundation for their criminal activities. In addition, some of these strains may lead females into a life of deviance at an earlier age or may increase the frequency of arrest. Although most of the socialization factors that shape a female's criminal activities are experienced earlier in life, certain factors such as substance abuse can trigger criminal behavior later life. Previous research indicates that women who become involved in crime later in life, is often the result of a substance abuse habit that started earlier in life (Pettiway, 1997, as cited in Katz, 2000). 15

PAGE 26

Racial Effects Minority inmates are disproportionally high in almost all state and federal prisons. The 1997 Survey of Inmates in state and federal correctional facilities conducted by the United States Department of Justice shows that the percentage of minorities incarcerated is about 45%, while in the general population minorities the United States constitutes less than 30%. This implies that race is probably the most important explanatory factor with regard to criminal behavior. Historically, minorities, African Americans particular, have faced great oppression in the United States. By examining the past, one can see that blacks have been oppressed from the time of slavery, throughout the Jim Crow era, and even in present day society. African Americans have been historically perceived at the bottom of the social hierarchy. Due to the historical oppression African Americans have experienced, black children may be socialized differently comparison to white children. ResearchersThomton, Chatters, Taylor, and Allen (1990) pointed out that: Parental socialization values generally reflect and complement those of the wider community, and, ideally, other major socialization agents reinforce the patterns found within families. Black parents, however, encounter unique societal proscriptions that create a dilemma for inculcating a positive group identity in their children (p. 401). Researchers Hill and Sprague (1999) pointed out that many black women admit they do not fall into the "white middle class box." They asserted that many black women, due to racial oppression and discrimination, have never had the luxury of 16

PAGE 27

solely staying home and raising their families because they have had to work to help support their families. Therefore, the Afrocentrism view is that not all African Americans have adopted the dominant-"'white middle-class values," and as a result they may not socialize their children according to these ''white middle-class" norms. An example of middle-class norms would be socializingboys to the person primarily responsible for providing their family's financial needs. contrast, females are conditioned to get a good education, but they are often expected to only be the primary housekeeper. However, researchers such as Collins suggest that black females are socialized to be strong, independent, and resourceful, and generally have more gender neutrality (Collins, 1990, as cited in Hill & Sprague, 1999). comparison to whites, black families are less likely to have a traditional family consisting of two married parents, and black children are more likely than whites to be raised in single-headed households (Taylor, Chatters, Tucker, & Lewis, 1990). In 1990 it was reported that 51 % of African American children lived in a female-headed household, in comparison to 16% of Caucasian children who lived in female-headed household (United States Bureau of the Census, 1990). Based on these statistics and the prior research, it becomes apparent that African Americans are more likely to come from single-headed households, in comparison to Caucasians. When examining female-headed households one can see clearly that racial differences exist. Black female single-headed households are among the most economically underprivileged the United States (Taylor, Chatters, Tucker, Lewis, 1990). For instance, it was

PAGE 28

reported that black female-headed households mean income in 1990 was less than $12,000, in comparison to white-female headed households who on average made $19,000 (O'Hare, Pollard, Mann & Kent, 1991). Black female-headed households face multiple challenges, but two important economic strains stand out. First a two parent family often has more potential to economically provide for their children, since there is the possibility of having two incomes instead of one. Second, black female-headed households on average make substantiallyless money than whites. Research has shown that racial differences also exist substance abuse, which as discussed above, is a significant factor for many females to break the law. Peers and substances may influence white and black adolescents in different ways. Some researchers have found that white adolescents were more likely blacks to report consuming vast amounts of alcohol and drugs (Barnes Farrell, 1992, Brannock, Schandler, Oncley, 1990, Thomas, Farrell, & Barnes, 1996). However, other research has shown that there is no significant difference between.. white and blacks consumption of alcohol (Brooks, Whiteman, & Gordon, 1983, Coombs, Fawzy, Gerber, 1986). However, this research was done in the 1980's and as a result more research may be necessary to determine whether there is a significant difference between white and black adolescents' alcohol Finally, race is a factor that may have an effect how early and often a female offender is arrested. For instance, previous research found that 8.1 % of black females had been arrested more than five times, and that white females accounted for about 18

PAGE 29

4.9% of those who were arrested more than five times (Facella, 1983, as cited Chesney-Lind Sheldon, 1998). this study I utilized relevant research that has been previously mentioned. The majority of the research mentioned above focused mainly on male and female samples, and very few specifically focused only on female samples. There is, general, a lack of research on female offenders. is not the aim of this study to compare male and female offenders; but rather to gain: a greater understanding of female offenders.

PAGE 30

CHAPTER 3 THEORETICAL BACKGROUND History In order to comprehend the current research and theoretical evolvement of female criminality, one must understand how it historically been characterized. Originally, female and male differences were explained solely on the basis of biology. Until recently, the sociological understanding of females and crime has not been a major area of concern. Sociological studies regarding females and deviance only dates back to the early 1920's. Cesare Lombroso, W.I. Thomas, and Otto Pollak were a few of the pioneering theorists from the early twentieth century that researched female deviance. Cesare Lombroso wrote (1920), and it was here that he suggested female criminal behavior was based on biological factors. Lombroso's theoretical works have been correlated to the theory of evolution. For instance, he discussed the concept of atavism, in which he suggested that human beings are situated in a hierarchal manner .. White males are at the top of this hierarchy and less evolved non-white women are situated at the bottom. Lombroso believed that deviant individuals were less evolved than law-abiding citizens, therefore deviant females were inevitable in society because they were less evolved than males (Lombroso, 20

PAGE 31

1920, as cited in Price Sokoloff, 1995). He described criminal and deviant acts purely based on sexual orientation and evolution. order to detennine female deviance he utilized a number of physical tnuts as indicators. For instance, he created a criminal profile for females based on their brain size, race, height, and any abnormality to justify female deviance (Lombroso, 1920, as cited in Belknap, 2001). Lombroso's speculations were based on how underprivileged, minority women were more likely to commit acts of deviance. He applied these speculations and generalized his findings to represent female criminality as a whole. His solution to eliminate female offenders was to prohibit deviant women from reproducing. Lombroso's work is widely identified in the criminology field, yet it is heavily discounted. Another theorist who made contributions that are currently inconsequential was W.I. Thomas. Thomas wrote (1923) in which he focused on female's sexuality and their ability to effortlessly manipulate men. Thomas's intent was to illustrate that female deviance was not only a product of biology, but was also caused by inadequate and inferior socialization. He suggested that poverty-stricken women were more likely to be deviant because they were not socialized to value purity and obedience. Additionally, he felt underprivileged females yearned for the sexual excitement that could only be obtained by committing an act of deviance. regards to female deviance Thomas (1923) wrote:

PAGE 32

The beginning of delinquency girls is usually an impulse to get amusement, adventure, pretty clothes, favorable notice, distinction, freedom the larger world ... The girls have usually become "wild" before the development of sexual desire, and their casual sex relations do not usually awaken sex feelings. Their sex is used as condition of the realization of other wishes. It is their capital (p. 109). Thomas implied that certain females are inclined to be sexually manipulative, and committed crimes for sexual and psychological purposes. Twenty-seven years later Otto Pollack published (1950). Pollack also utilized biological and physiological frameworks to comprehend female criminality. However, he expressed that female deviance is the outcome of sexual roles played by females. Two of the factors that Pollack used to base this assumption on were females' ability to conceal their menstruation cycle, as well as their sexual orgasms. He believed that females were deceitful and had the ability to effortlessly hide their crimes. Pollack (1950) stated Thus, for biological as well as for cultural reasons, woman seems to possess greater powers of concealment than does man. And for this group population which is therefore better equipped to achieve the supreme goal of most criminals, namely, to remain undetected, our culture furnishes almost ideal conditions for the perpetuation of crime (p. 151). Additionally, Pollack held the view that female crimes were under reported due to chivalry, and this was the reason for such discrepancy between male and female crime rates. Pollack (1950) stated: 22

PAGE 33

Men hate to accuse women and thus indirectly to send them to their punishment, police officers dislike arrest them, district attorneys to prosecute them, judges and juries to find them guilty, and so on (p. 151). Finally, he suggested that females often coerced males into committing crimes on their behalf. is important to gain insight about how female criminality was defined and related to sex, class, and race during the early twentieth century. By taking into account how females and <;ieviance were characterized a biological and sexual manner, we are able to step back and view the progress that has been made the comprehension of female criminality. Although great strides have been made to better understand the criminal motives of females, more is necessary in order to thoroughly comprehend the matter. Feminist Theory Feminist theory has provided a solid framework that contributes to the conception of both criminology and socialization. One of the aims of feminist theorists is to elucidate how the social construction of gender roles, such as the aforementioned, is constantly reproduced and to illustrate that one main influential agents of gender socialization are parents (Weitzman, 1979). Hoffman-Bustamante (1973) suggested that, for girls Social control in the form of informal sanctions applied by primary and secondary groups is imposed more consistently 23

PAGE 34

and for more minor deviation from accepted standards. This result a situation where females have been taught to confonn to more rigid standards and rewarded for such behavior, whereas males are told to conform, yet rewarded for flaunting many conventional standards. (p. 120) Additionally, it is necessary to understand whether both Caucasian and African American parents socialize their children accordance with the dominant society. Researchers, such as Hill and Sprague (1999), have utilized a multicultural feminist perspective to understand how race, class, and gender interact. They suggested that race and class playa vital role on gendering, however the effects vary based on class and racial structure (Hill & Sprague, 1999). Due to differences in socialization conditions, race becomes a factor determining which kinds of strains are placed on females. Finally, feminist criminologists are currently researching female criminality by taking on a "pathways to crime" approach (Gaarder Belknap, 2002, p. 38). This qualitative approach takes into account an individual life course and the negative stimuli that one has undergone, which could possibly lead an individual down a path of deviance (Gaarder Belknap, 2002). 24

PAGE 35

Differential Association Sutherland's theory of differential association (1934, 1947) is based on the framework of socialization; which addressed how to conceptualize causal factors of crime. He believed that criminal acts are behaviors that are socially learned through the process of interaction. According to this theory, for example, adolescents are more likely to participate substance use if their parents or peers use drugs or alcohol. addition, adolescents.who have family members or peers involved criminal acts such as substance abuse are more likely to learn this pattern of behavior by mimicking their parents or peers deviant behaviors. This type of environment lacks parental moral guidance, and the child's inhibi!or to deviance and crime is low or nonexistent. the book, the (1934), Sutherland wrote: First, any person can be trained to adopt and follow any pattern of behavior which he is able to execute. Second, failure to follow a prescribed pattern of behavier is due to the inconsistencies and lack of harmony in the influences which direct the individual. Third, the conflict of cultures is therefore the fundamental principle in the explanation of crime (p. 5152). Sutherland was suggesting that criminal behavior, like any other type of behavior is learned. Additionally, he believed that when an individual is associated with others that do not adopt the dominant social nonns and laws, then that individual is more likely, to embrace the illegitimate nonns of their social surroundings. However, he maintained a stance, which infers that not all individuals who are subjected to 25

PAGE 36

negative influences are affected identically. Furthermore, he discussed that cultural was inevitable once certain individuals or groups acclimate to deviant lifestyles. Sutherland proposed his fmal theorization on differential association in (1947), in which he ameliorated the theory. He suggested that two components are involved in the processes of learning deviant behaviors. First, one must learn the skills that are essential to commit crimes, and second one must learn the anti-social norms that support and reaffirm deviance (Sutherland, 1947). This theorization reinforces the process of socialization, and reiterated that through the process of socialization individuals learn behaviors acceptable in society. General Strain Theory Although the majority of the prominent criminology theories were developed solely to understand and explain male criminality, it is necessary to incorporate the theoretical conceptualization of socialization, feminist, differential association in order to better understand female criminality. Robert Agnew's (1992) general strain theory (GST) provides a critical assessment of how this can effectively be accomplished. In comparison to Merton's (1938) strain theory, Agnew's (1992) general strain theory offers a broader context by suggesting that many strains may be involved when an individual commits a crime. Whereas Merton's (1938) strain theory merely focused on the strains of making one unable to achieve ones 26

PAGE 37

economically driven goals, Agnew's general strain theory surpassed Merton's by enabling scholars to conceptualize that individuals experience a variety of strains throughout their lives. Agnew's general strain theory has been applied to many studies pertaining to male offenders. GST can also be applied to female offenders, even though women are facing different set of strains compared to their male cOlll1terpart. First, because of differential gender socialization, males and females are treated differently and are experiencing different strains and emotions in their lives (Broidy Agnew, 1997). In addition, not only do males and females undergo different strains, they also may react differently to those strains. The different reactions to the negative stimuli can be attributed to the differential gender socialization in society. During the course of their childhood males are socialized in ways that encourage them to react to certain situations and strains in a specific manner; a manner that often differs from the. acceptable behaviors of females. For example, males are taught not to be "sissies" and to fight back when they are confronted. In contrast, females are socialized to believe that similar behavior is unladylike (Kimmel, 2000). If one separates the strains that female offenders undergo, then Agnew's GST becomes of great importance and relevance to understand female criminal behavior. The aim is not to explain why males are inherently more involved crime, but to understand what gender specific strains are more likely to lead to female offense. 27

PAGE 38

Agnew defined the tenn strain as, "relationships in which others are not treating the individual as he or she would like to be treated" (Agnew, 1992, p. 28). Agnew differentiated two categories of stains. The first is an objective strain, which refers to a consensus of shared understandings of what constitutes a strain in a society. For example, if one were abused at some point in time, then society as a whole would recognize this as a negative strain. The second category is the subjective strain, which corresponds to undesirable strain in which an individual undergoes, but there is not a shared consensus that the experience is a strain (Agnew, 2001) Agnew defined three types of strain; failure to achieve positively valued goals, loss of positively valued stimuli, and the presentation of negatively valued stimuli (Broidy Agnew, 1997). The first type of strain explained by Agnew dealt with individuals failing to achieve their expected goals. Many females are worried about financial security and the potential obstacles that could prohibit them from achieving that goal. For instance, increased divorce rates and low wages for women, in comparison to men, are considered to be strains. Lower wages for women, in comparison to males, becomes a strain when one cannot achieve culturally valued goals, such as economic stability. However, Agnew recanted his stance in his revision of GST, and stated that the inability to achieve positive valued goals, such as economic success and education, are not significant when examining deviant behavior (Agnew, Additionally, he suggests that if an individual strives to 28

PAGE 39

become educationally and economically successful than that person has already adapted somewhat to the customary nonns of society, and will forego acts of deviance (Agnew, 2002). The second type of strain discussed by Agnew deals with the loss of a positive environment such as freedom, job, family, parent, or spouse (Broidy& Agnew, 1997). The third and most important and relevant strain is the endurance of a negative environment (goal blockage) (Broidy & Agnew, 1997). For example, one may experience abuse, peer deviant behavior, parental deviant behavior, disadvantaged childhood environment, substance use, or sexual/physical victimization. Broidy and Agnew theorize that when individuals have multiple types of strains present in their lives, they are more susceptible to be involved in criminal activity (Broidy& Agnew, 1997). Agnew's perspective of negative stimuli is an important concept adopted in this study, in addition to foundational works of socialization, feminist criminology, and Sutherland's differential association. Many criminology theories were developed with male criminality as the primary focus. Socialization theory has not fully penetrated the field of criminology and criminal justice, or been accepted as an influential paradigm. Sutherland's differential association does not assume any differences exist between males and females in their association patterns. It is not very easy to derive testable hypotheses from feminist criminology theories that deal with female criminality. However, by combining these theories; socialization, feminist thought, differential association, and 29

PAGE 40

general strains, we are able to grasp the key elements that propel females toward crime. Specifically, deviant peer and parental association, exposure to negative stimuli during socialization, and poor socioeconomic status could all contribute to the reality of why some females are forced to travel doWn the path of deviance. The aim of this study is to focus on the objective and to measure the effects of these strains on female offenders within the framework of socialization. Hypotheses Based on the theoretical frameworks discussed above and the review of relevant literature, the following hypotheses have been formulated. The idea is to pinpoint the key factors that explain and predict female criminality in a framework that integrates the concepts of socialization, differential association, general strain, and feminist criminology HI: Personal substance abuse will result in a younger initial age and a higher frequency of arrest. H2: Abuse victimization will lead to a younger initial age and a higher frequency of arrest. H3: Parental and peer deviant behavior will lead to a younger initial age and a higher frequency of arrest. H4: Parental and peer deviant behavior will lead to a higher level of abuse victimization and a younger initial age of substance abuse. Adverse family structure will result in a younger initial age and a higher frequency of arrest. 30

PAGE 41

H6: Adverse family structure will lead to a higher level of abuse victimization and a younger initial age of substance abuse. H7: A higher level of abuse victimization will result in a younger initial age of substance abuse. 31

PAGE 42

CHAPTER 4 DATA AND METHODS Data this study I utilized secondary data from the Survey of Inmates in State and Federal Correctional Facilities, from the year 1997. This survey was conducted by the United States Department of Justice Bureau of Justice Statistics and United States Department of Justice Federal Bureau of Prisons. The unit of analysis was the individual inmate who was incarcerated during the time of the survey, and the study applied a multi-stage cluster, stratified random, face-to-face survey design. It United States Department of Justice Bureau of Justice Statistics and United States Department of Justice Federal Bureau of Prisons, (1997) reported that the first stage of the sample design involved the selection of prisons and facilities that would be included in the study. The state prisons population included 1131 male prisons, 131 female prisons, and 147 prisons coed prisons. From this population, 13 ofthe largest male prisons and 17 of the largest female prisons were selected for further analysis. The researchers detennined that the 30 largest prisons were representative. The researchers then stratified the remaining 1265 male prisons and 261 female 32

PAGE 43

prisons into 7 strata, based on census regions. It was reported that within each stratum each prison was put in to order by facility type, size, and security level. From the state sample frame the researchers selected 223 male prisons and 47 female prisons based on the probability proportional to size. The total sample size for state prisons was 300. The state facilities selected participants from a sample frame of inmates that used abed at one of the facilities the prior night. The researchers then employed a systematic sample in which they randomly selected participants from each stratum by using and random start and a sampling interval. The state facilities reported that 12,269 males and 3,116 females were sampled. A computer-assisted personal interviewing (CAPI) was used during each interview. Each interview lasted approximately one hour. All of the respondents were informed both verbally and in writing that this was a voluntary procedure, and that they could decide not to participate in the study at anytime. In addition, the respondents were informed that all the information collected would be kept confidential (United States Department of Justice Bureau of Justice Statistics and United States Department of Justice Federal Bureau of Prisons, 1997). The researchers used a questionnaire as the measuring instrument. Upon studying it carefully I found that the questionnaire had high face validity. The researchers employed a survey design using a face-to-face administrative questionnaire, which was conducted using CAP!. By using CAPI the interviewers were able to ask each 33

PAGE 44

question in a universal fashion. addition, the CAPI allows the interviewer to give exactly the same defmition of each concept to each participant who did not comprehend. The researchers employed a stepwise sampling strategy that include the following steps; cluster, stratified random, and systematic sampling. The researchers conducted a very solid methodological sampling, and I found it to be very representative of individuals United States state prisons. However, it must be noted that the reliability coefficient of the questionnaire was not reported the data set. order to investigate my research questions and to test my hypotheses specifically, a sub-sample of approximately 1038 respondents was created, which includes only female inmates that were incarcerated state prisons. Statistical Procedures Causal Modeling I chose to utilize the technique of structural equation modeling (SEM) to analyze the data. By using this technique, I was able to model the complexity and dynamics of the causal structure with reference to female criminality and all the contributing factors, as this multivariate technique allows for the simultaneous estimation of almost any possible causal specification. The SEM procedure is based on a 34

PAGE 45

minimization function with which the discrepancy between the observed covariance matrix (from the raw data) and the model implied covariance matrix is minimized. Direct, Indirect, and Total Effects The SEM allows researcher to estimate not only direct but also indirect and total effects of exogenous and endogenous variables. The multiple effect capability provides researcher greater means examine the subject matters and to test hypotheses. So far, very few studies in criminology have investigated causal relationships regarding female criminality beyond direct effects. Latent and Manifest Variables The SEM differentiates latent and manifest variables, and allows researchers to measure abstract latent concepts by employing multiple manifest indicators. Latent variables refer to general, abstract concepts that cannot be observed directly and measured perfectly. The function of latent variables is to allow researchers to use of multiple (manifest) indicators to measure an abstract concept more accurately and precisely. For an abstract concept, multiple indicator measurement is preferable to a single indicator operation in terms of the content validity. This study will take the advantage of the multiple indicator measurement wherever the data permits. 35

PAGE 46

Fit Indexes A group of fit indices for SEM model will be used to evaluate model fit. The first index is the chi-square statistic test. It is commonly used indicator for gauging the overall model fit. A significant chi-square index indicates that there is a substantial discrepancy between the observed and the model implied covariance matrixes, and that the mode1.does not fit well. While, chi-square provides information about the fit of a model, it is very sensitive to sample size. For that reason, Joreskog and Sorbom (1979) advised that one should use the ratio of chi-square to degrees of freedom (CMINIDF) to compensate for the dependence on sample size. the CMINIDF is less than 2, the model can be considered to be an adequate fit (Byrne, 1989). Another fiUndex that been utilized is the goodness of fit index (OFI), suggested byJoreskogand Sorbom (1979) as an appropriate index to determine the overall fit of the model. It was reported that GFI is parallel to R2 in multiple regression analysis (Tanaka and Huba, 1989, as cited in Tabachnick Fidell, 1996). In essence, GFI allows a scale to be created by taking the sum of the squared .. variances of the estimated parameter population divided by the sum of the squared variances from the sample covariance matrix (Tabachnick Fidell, 1996). The general rule is that if a GIF value is greater than or equal to .90, it indicates an overall good model fit (Long, 1983; Loehlin, 1987). In addition, I used the adjusted fit index (AGFI), which is derived from OFI. AGFI allows for the adjustment of OFI by the degrees of freedom, i.e., AGFI takes 36

PAGE 47

model complexity into account. In essence, AGFI adjusts the GFI by incorporating the information on how many parameters are estimated. The general rule for the AGFI value states that.if it is greater than or equal to .80, a close model fit is achieved (Long, 1983; Loehlin, 1987). The next fit index is the root mean square residual (RMR). represents the error terms between the observed and the implied covariance matrices. A low RMR means the differences between the observed and model implied covariance matrixes are small and therefore suggests a good fit model. When a RMR value is less than or equal to .05, the model has an acceptable fit (Long,. 1983; Loehlin, 1987). The last fit index is the root mean square of error of approximation (RMSEA). lbisis a better choice when a model has large degrees of freedom. It gives a more favorable fit to a model with large numbers of variables and parameter estimates. When the value of RMSEA is less than or equal to .05, an adequate error approximation is achieved, and the model has a reasonable fit (Hu Bentler, 1999). Any fit index discussed above has its strengths and weaknesses. When evaluating model fit, it is highly recommended by Joreskog and Sorbom to use multiple indices rather than focus on any particular one. I took in all available information when evaluating the fit of all my models. 37

PAGE 48

Model Specification Structural equation model contains two main components, the measurement model and structural model. Measurement Model The measurement model establishes relationships between the observed (manifest) variables and the unobserved (latent) factors. Based upon the previous empirical research and the current theoretical framework, I used a total of 9 latent variables, which are measured, by 17 manifest variables in the general model. The first group includes factors representing the concepts of socialization, differential association, sources of negative stimuli during socialization process, and basic demographic variables. Specifically, it comprises of five exogenous latent variables): childhood environment, parental deviant behavior, peer deviant behavior, socioeconomic status, and education. The first three latent variables are the main focus of this investigation, which is guided by a socialization framework. The other two demographic factors are also important from a socialization point of view, and therefore have been included as control variables. The five exogenous factors are by no means exhaustive. However, given the theoretical objectives and availability of 1 Exogenous variables have been included in the model to assist in explaining other variables, and only have predicted effects on the other variables the model. 38

PAGE 49

variables the data set, they were sufficient to serve my purpose. All of the five factors, except for socioeconomic status and education were measured with multiple indicators. Socioeconomic status. (SES) was measured with the single indicator,.income. The respondents were asked which category best described their personal income during the month before their arrest? The response categories are: 0 no income, -1 $1-199,2=$200399 ... and all the way up to 16=$7500 or more. Education (Educ) was also measured with a single variable. The respondents were asked before their arrest what was the highest grade that they had attended. There were 8 response categories from 1 = first grade in elementary school to 8 = two or more years in graduate school. Family structure was measured with three indicators. The first measures who the respondent lived with while growing up (BotbPar): The response categories are: both parents (including one step parent), mother, father, grandparents, other relatives (including step-relation), friends, foster homes, agency or institutions (including religious institution). Since this is a nominal variable, a dichotomous variable was created, with both parents = 0, and all other categories = 1. The second indicator selected for childhood environment is whether the respondent's parents received welfare. The respondents were asked, while growing up if their parents or guardians ever receive any welfare or public assistance 39

PAGE 50

(Pwelfare), for example AFDC, food stamps, Medicaid or This is a variable with dichotomous answers: 1 =yes and O=no. The third indicator for family structure is whether the respondent lived in public housing (pubHous). The respondents were asked while they were growing up whether they ever lived in publicly assisted housing, for example, Section 8 housing. The answer was yes (l) or no (0). order to test the appropriateness of the indicators for the latent variable (family structure), a factor analysis was conducted. The results showed that the factor loadings for the three indicators were 0.63 for both parents, 0.77 for public housing, and 0.82 for parents welfare status. Apparently, all three indicators load very well on the latent variable, family structure. addition, the reliability test for the three indicators as a scale also showed an acceptable result: alpha=0.59. The second latent variable, peer deviant behavior, was measured with four indicators. The flrst one is whether the respondent's peers used drugs (Drugs). The respondents were asked, while growing up, whether their friends they hung aroWld with engaged in activities like using drugs. There are two response categories: yes=l and no=O. The second indicator (ShopLift) is whether the respondent's peers participated in shoplifting (yes = 1 and no =0). The third indicator (SSP) measures whether the respondent's peers sold.stolen properties (yes = 1 and no =0). The fourth and flnal manifest variable (SIMD) asking whether the respondent's peers sold, imported or manufactured drugs (yes = 1 and no =0). The factor loadings for the four 40

PAGE 51

indicators are 0.75 for drugs, 0.76 for shoplifting, 0.77 for selling, importing, or manufacturing drugs, and 0.80 for friends selling stolen property. These demonstrate a well-measured construct with adequate indicators, as all the factor loadings greater than 0.70. The reliability analysis reported an alpha of 0.77, another indicator of reliable measurement of the concept. Parental deviant behavior was measured with two indicators. The first one (SubAbuse) is the respondent's parent(s) abused substances. The respondent was asked, while growing up, whether any of their parents or guardians abused alcohol or drugs. There were two response categories: yes 1 and no = O. The second indicator (J ailTime) is whether the respondent's parent( s) spent time in prison or jail.(yes = 1 and no = 0). The results of factor analysis showed high factor loadings (0.82) for both indicators, indicating a well-measured latent construct. In addition, the reliability analysis reported an alpha score of 0.52, which is also acceptable. The second group of latent variables includes two variables with behavioral connotation: substances abuse and abuse victimization. Substance abuse was measured with two continuous manifest indicators. The first one is respondent's initial age of alcohol use (Ralchohol), and the second one measures respondent's initial age of drug use (RDrugs). The questions asked how old the respondents were when they first started using alcohol/drugs. The factor loadings were 0.88 for both indicators suggested that the two variables are adequate measures of the latent factor

PAGE 52

substance abuse. addition, the reliability analysis returned an alpha index of 0.69, which is very acceptable. The other. latent variable in the third group is abuse victimization. was measured with two manifest variables. The first variable (physical) asks whether the respondent was ever physically abused (yes=l and no=O). The second indictor (Sexual) records whether the respondent was ever sexually abused (yes=l and no=O). The respondents were asked, before they were admitted to prison, whether they had ever been sexually/physically abused. The results of factor analysis revealed a factor loading of 0.82 for both indicators, which are quite high. Furthermore, the reliability analysis alpha for abuse victimization was reported to be 0.51, which is acceptable. The third group of structural factors contains two most central (dependent) variables. They are age of first arrest (Agearest) and frequency of arrest (Freq 1). Since these variables are not abstract, each was measured with a single-continuous indicator. The age of first arrest asked how old the respondent was the first time she was arrested for a crime. And the question about the frequency of arrest asked the respondent how many times she had ever been arrested, as an adult or juvenile, before her latest arrest. Structural Model Structural model specifies the relationship between latent or structural variables. The aim of this analysis is to test the effects of strains presented in one's family structure, 42

PAGE 53

exposure to negative stimuli including parental and peer deviant behavior, and sexual and physical abuse suffered by the respondent, and respondent's behavior of substance abuse on the earliest age, and frequency of criminal arrest of female offenders. Additional steps were taken to detennine what differences and similarities existed between Caucasian and minority inmates. The model specification for the structural model is shown Figure 1. 43

PAGE 54

t o

PAGE 55

First, all exogenous factors, family structure, parental deviant behavior, peer deviant behavior, education, and SES have a direct link to each one of the 4 endogenous factors, abuse victimization, substance abuse, age of first arrest, and frequency of arrest. The purpose of this is to test whether structural 'factors, socialization, deviant association, negative stimUli and demographics contribute significantly to female offenders' abuse victimization, substance abuse, initialage of arrest, and frequency of arrest. Second, abuse victimization is specified to have a direct link to substance abuse, age of first arrest and frequency of arrest. As discussed previously, women that have experienced abuse, tend to self medicate by taking drugs and/or alcohol as a way to alleviate pain. However, as shown repeatedly previous studies, substance abuse can only push women down: the path of deviance further and faster. lbird, substance abuse is specified as connecting directly to age of first arrest and frequency of arrest. Testing Strategy First, I estimated a general model based on the whole sample. Its purpose is to get baseline estiinates for the female population in the state prisons, and to test the hypotheses that address general issues. Then, two additional sub-models were tested in order to detect and compare racial differences in the relational patterns described the general model. It is common many studies to use race as a control variable. However, by testing two separate racial sub models (white and non-white), I could 45

PAGE 56

see if there is a differential causal pattern determined by race factor for the relational structure that was found in the general model. Therefore, using two sub-sample models enabled me to better test racial effects. Three models were eventually tested for the entire, white, and non-white samples respectively. The whole sample included 1038 respondents and represented the entire population of female offenders in state prisons. The white sub-sample contained 467 female inmates and the non-white sub-sample was comprised of 571 female offenders. A frequency analysis of the original race variable revealed that whites accounted for 45%, blacks 49.6%, Asian or Pacific Islander 1.5%, Aboriginal North Americans 3.4% and "other" was equal to 0.5% of the sample. However, a closer look at the numbers reveals that the entire sample is mainly comprised of whites (45%) and blacks (49.6%). Therefore, the white-non-white dichotomy is very close to a white-black dichotomy. 46

PAGE 57

CHAPTER 5 RESULTS AND DISCUSSION The preliminary descriptive analyses revealed some general characteristics of the sample. For example, the average monthly income for the entire sample was $650, the average age was 35, and the average highest grade that the respondents attended was 11 th grade. addition, 60% of the respondents in this sample did not live with both parents while growing up, 24% of the females in this sample had a parent that served time ajail or prison, and 45.6% of the female inmates reported that their parentCs) abused drugs and/or alcohol. Additionally, 53.8% of the female inmates in the sample had experienced physical abuse, and 43% had experienced sexual abuse. The average initial age of arrest was 21.6 and the average of frequency of arrest for female inmates was 6. 7 times. Whole Model Table 1 (see below) shows the results for the general model based on the whole sample. Five model fit indices have been included to evaluate how well the model fits the data. These indices are 1.691, GFI 0.984, AFGI = 0.971, RMR 0.118, and RMSEA 0.026. All of them, except for the RMR, suggest that the model fits the data very well. Even though RMR was slightly higher than the average 47

PAGE 58

acceptable value, it does not dismiss the fact that the model is an adequate fit, as the majority of the fit indexes demonstrate a good fit. 48

PAGE 59

Table I. Estimated Direct. Indirect. and Total Effects of Strains and Negative Stimuli of Female Offenders for the Whole Sample (N=I038). Dependent Independent Standardized Structural Coefficient Variable Variable Direct Effect Indirect Effect Total Effect Family -0.134" 0.017 -0.117 Structure Peer Deviant -0.101 -0.168 -0.269" Behavior Parental Deviant 0.021 .-0.089 -0.069 Behavior Abuse 0.059 -0.135 -0.076 Victimization Substance 0.460 0.460 Abuse Family -0.005 -0.009 -0.014 = 0.061 Structure Peer Deviant 0.166" 0.024 0.190 Behavior Parental Deviant 0.028 0.025 0.052 Behavior Abuse 0.020 0.014 0.034 Victimization Substance -0.048 -0.048 Abuse = 0.355 Family -0.012 0.086 0.074 Structure Peer Deviant -0.325 u -0.070 -0.';95 u Behavior Parental Deviant -0.096 -0.175 -0.270 Behavior Abuse -0.294 -0.294u Victimization Family -0.292 -0.292 0.393 Structure Peer Deviant 0.238 u 0.238" Behavior Parental Deviant 0.594 Behavior Tests of Model Fit DF (P-value) = 142184 (.0003) CMINIDF = 1.691 RMR= 0.118 GFI =0.984 AGFI = 0.971 RMSEA 0.026 Note: significant at level. and significant at level. 49

PAGE 60

Abuse Victimization The results show that the R2 for the latent variable, abuse victimization, is 0.39. This indicates that 39% ofthe variation in abuse victimization can be explained by the 5 exogenous variables. Among them, parental deviant behavior (0.594**) and peer deviant behavior (0.238**) have significant direct effects on abuse victimization. These findings support part of hypothesis H4, which states that parental and peer deviant behavior will lead to a higher level of abuse victimization and a younger initial age of substance use. The implication is that parental and/or peer deviant behaviors will lead to a higher level of abuse victimization, probably by exposing individuals to more risky/criminal environment. Being exposed to a criminal environment may make these individuals more vulnerable to violence, abuse, or exploitation, thus making them more susceptible to victimization. Since the structural coefficients used here are standardized, we can compare the relative importance of the factors. The results show that that the effect of parental deviant behavior on abuse victimization is more than double the effect of peer deviant behavior This suggests that the closer the source of negative stimuli is to the child, the more susceptible the child is to various abuses. Negative family structure, theoretically, should dampen all economic resources and protection available to a child. a result, this increases the child's chance of being physically and/or sexually victimized. However, the statistical results show that family structure does not have an expected effect on abuse 50

PAGE 61

victimization, as stated in A possible interpretation of this is that the respondents gave socially desirable answers. A Speannan's correlation analysis was conducted, and it shows a lack of even a simple bivariate correlation between the indicators involved (some of these non-significant correlations are negative). Substance Abuse The results show that the R2 for substance abuse is 0.36, which means that 36% of the variation of the endogenous/latent factor, substance abuse, can be explained by the structural model, i.e., by all 5 exogenous variables plus an endogenous one, abuse victimization. Among them, peer deviant behavior (0.325**) and abuse victimization (-0.294**) both have significant direct effects on the initial age of respondent's alcohoVdrug use. These findings suggest that a respondent who were associated with deviant peers, or were physically or sexually victimized, are more likely to start using substances at an earlier age. These results support the hypothesis H7, which states that a higher level of abuse victimization will result in a younger initial age of substance abuse. addition, part of hypothesis which states that peer deviant behavior will lead to a younger initial age of substance abuse, was supported. Furthermore, the findings of the significant effects are also consistent with previous research, which found that sexual victimization had a significant direct effect on the respondent's substance abuse (Dembo, Dertke, La Voie, Borders, 51

PAGE 62

Washburn Schmeidler, 1997). Additional research found that an individual that had experienced childhood abuse is more likely to develop a habit of substance abuse in order to self medicate (Beckman, 1984, Toray Coughlin, Yamaguchi Kandel, 1991). Moreover, Miller, Downs, Gondoli, and Keil (1987) research found that females that specifically experience sexual abuse are more likely to self medicate with substances. addition, previous research suggests that females are more likely to engage in self-medication after experiencing a negative traumatic event (Arnold, 1990). According to table 1, parental deviant behavior does not have a significant direct effect on substance abuse, as is stated in the f4. Nevertheless, it both significant indirect (-0.175**) and total (-0.27**) effects on substance abuse. These indicate that the effect of parental deviant behavior on substance abuse is mainly mediated by abuse victimization. An interpretation of this could be that negative parental influence, or simply negligent parents, are more likely to cause abuse victimization ofa female respondent. This may cause the female to be more likely, according to Miller, Downs, Gondoli, Keil (1987), to choose to deal with the physical and/or emotional trauma through substance abuse. What I want to emphasize here is that the significant overall effect of parental deviant behavior on a child's substance abuse would not be discovered without exploring indirect and total effects. 52

PAGE 63

Family structure does not have a significant effect on respondent's substance abuse, according to Table 1. Therefore, the portion that pertains to how adverse family structure will lead to younger initial age of substance abuse was not supported. This may be due to the same reason as previously discussed: Flawed data or poor measurement for the latent variable, family structure. Frequency of Arrest The R2 for frequency of arrest was equal to 0.06, meaning that only 6% of the variation of this factor can be explained by 7 up-stream latent variables. The smaller R2 may partially be caused by the lack of variance in the latent factor itself The descriptive statistics show that the variance for frequency of arrest is only half the size of the variance of latent factor, age of first arrest. This is intuitively understandable, since many offenders may have only committed a crime once in their lives. However, that crime may carry a long prison term, such as murder. This eliminates their chances to cumulate higher frequency of arrest. On the other hand, people may commit crime at almost any age. Table I shows, among those predictors, only peer deviant behavior has a significant direct effect (0.166**) on the respondent's frequency of arrest. This suggests that those who have more delinquent peer association are more likely to arrested more frequently. The part ofR3 that stated that parental and peer deviant behavior will lead to a younger initial age of arrest and higher frequency of arrest, 53

PAGE 64

was supported. Nonetheless, it must be noted that I failed to find support for other hypotheses pertaining to frequency of arrest. One reason for the insignificance of the other factors may be attributed to the small variance of the variable, frequency of arrest. When there is not much variance to explain, it is harder to show the significant effect of a variable among many included and controlled for in one single equation. Age of First Arrest The R2 reported for age of first arrest is 0.32, implying that 32% of the variation in age of first arrest can be explained by 7 latent variables in the structural model. The significant up-stream variables include substance abuse (0.46**), family structure (-0.134**), and peer deviant behavior (.;0.101 *:t<). suggests that respondents who started to use substances at an earlier age, who experienced a negative family structure, and who were associated with delinquent peers tend to have trouble with laws and be arrested at a younger age. The significant effect of family structure echoes the findings of previous studies. Previous studies have examined how the impact of living with a single parent or guardian increases the chances of that individual becoming more prone to deviant behavior (Barnes & Farrell, 1992; Barnes, Farrell & Banerjee, 1994; Thomas, Farrell, Barnes, 1996). Also, living in a state of poverty was likely to increase the respondents' chances of becoming involved in criminal activities (Wright, Entner, Caspi, Moffitt, Miech, & Silva, 1999). Those who are socialized a poor 54

PAGE 65

environment are more likely to rebel against social norms and laws. Prior research has also reported that peer deviant behavior is one of the leading predictors of female criminality when other variables, such as parents and school environment have been controlled for (Aseltine, 1995). Table I also reveals that negative peer influence seems to be very robust at all levels of significant effects: direct (-.101 **), indirect (-.168), and total (-.269**). The implication of this finding is that it will be very difficult for a person to avoid negative behavioral consequences if she has socialized or interacted with deviant or delinquent peers on a daily basis. Based on these results, the following hypotheses that relate to the initial age of arrest are supported: HI. which states that personal substance abuse will result in a younger age and higher frequency of arrest; Hs, states that adverse family structure will result in a younger initial age and a high frequency of arrest; The portion ofH3 that states, parental and peer deviant behavior will lead to a younger initial age and a higher frequency of arrest, was confirmed. However, since parental deviant behavior and abuse victimization did not have significant effects on the initial age of arrest, the hypothesis R2 and the other half ofR3 fell short of empirical support in this data set. Sub-Models Two sub-models, one for white and the other for non-whites, were estimated to detect any possible racial impact on the relationships among socialization conditions and 55

PAGE 66

agents, economic and environmental strains, negative life experience and trauma, and age and frequency of incarceration for female offenders. To implement the sub model design, I split the whole sample into white and non-white sub-samples and ran the general structural equation model on each of the sub-samples. What I was for through the model comparison is to see if there are any differences in structural coefficients between the two sub-models. White Model Table 2 shows (see below) the complete SEM estimates for the White model. The model fit indexes indicate that it is a well fit model 1.475, GFI 0.971, AFGI = 0.946, RMR = 0.155, and RMSEA = 0.032). Again, all of these indices, expect for R.MR, for the model fit meet the criteria for these indices. fact, if we take into consideration the fact that the sample size of this model is less than a half of that of the whole model, the fit is better than expected. Furthermore, when the sample size becomes smaller one has less statistical power, which decreases the amount significant coefficients. 56

PAGE 67

Table 2. Estimated Standardized Direct, Indirect, and Total Effects of Strains and of Female Offenders for the White Sample (N=467) Dependent Independent Standardized Path Coefficient Variable Variable Direct Effect Indirect Effect Total Effect Family -0.199* 0.101 -0.098 Structure = 0.398 Peer Deviant -0.029 -0.304--0.333** Behavior Parental Deviant 0.099 -0.133 -0.033 Behavior Abuse -0.005 -0.165 -0.170 Victimization Substance 0.569-0.569** Abuse Family 0.163 -0.057 0.106 Structure Peer Deviant 0.114 0.075 0.189** Behavior Parental Deviant 0.084 -0.072 Behavior Abuse 0.098 0.029 -0.128 Victimization Substance -0.101 -0.101 Abuse Family 0.058 0.116 0.174 Structure Peer Deviant -0.470** -0.062 -0.533** Behavior Parental DeViant -0.047 -0:180 -0.277** Behavior Abuse -0.290** .-0.290** Victimization Family -0.398** -0.398** Structure Peer Deviant .0.215* Behavior Parental Deviant 0.620--0.620** Behavior Tests of Model Fit DF (p-value) = CMINIDF = GFI=0.971 AGFI = 0.946 RMR = 0.155 RMSEA = 0.032 Note: '*' significant at a=.05 level. and '**' significant at a=.01 level. 57

PAGE 68

Abuse Victimization. The R2 for abuse victimization is 0.33. TItis indicates that 33% of the variation in abuse victimization can be explained by the 5 exogenous variables. Among them, parental deviant behavior (0.62**) and peer deviant behavior (0.22**) have a significant direct effect on abuse victimization. The standardized structural coefficients have a similar effect order as the whole model, i.e., parental deviant behavior is much more conducive than peer deviant behavior to abuse victimization of female respondents. Adverse family structure again returned a wrong significances direction, this is the problem that occurred the whole model. Substance Abuse. The result shows that the R2 for substance abuse is 0.432, which means that more than 43% of the variations in substance abuse can be explained by the structural model, i.e., by all 5 exogenous" variables and an endogenous one, abuse victimization. Among them,peer deviant behavior (-0.47**) and abuse victimization (-0.29**) both have a significant direct effect on the initial age of respondent's alcohol/drug use. These fmdings suggest that a white female who was associated deviant peers, or was physically or sexually victimized, is more likely to start using substances at a younger age. Same as in the whole model, even parental deviant "behavior does not have a significant direct effect, it comparable indirect (-0.180) and total (-0.28**) effects on substance abuse. The absolute value of the indirect effect, -.180, though not significant, is greater than its counterpart in the whole model. It became insignificant mainly because of the reduced sample size. 58

PAGE 69

Frequency of Arrest. The R2 for frequency of arrest was still small, 0.07, meaning that only 7% of the variation of this factor can be explained by 7 prior latent variables. Continuing the pattern of lacking significant direct effects, the results do not show any significant direct structural coefficients with regard to frequency of arrest. However, the pattern of the effects of peer deviant behavior appears to be similar to the results of the whole model. That's, it has a significant total effect (0.19* *), and the absolute value of its direct effect (0.11) is comparable to its counterpart in the whole model, although not significant. The explanation for this again lies in the lowered statistical power by reduced sample size. Age of First Arrest. The R2 for age of first arrest is 0.398, implying that about 40% of the variation in age of first arrest can be explained by 7 latent variables in the structural model. These significant predictors include substance abuse (0.57**) and family structure (-0.20**). suggests that white females who started to use substances at an earlier age and who had a negative family structure tend to have trouble abiding by laws and often arrested at a younger age. Peer deviant behavior does not have a significant direct effect (-0.03) on the age of arrest. While it continues to show strong indirect (-0.30**) and total (0.33**) effects, it is inconsistent with the whole model. The attenuation of the indirect effect may also be related to the smaller sample size and lower statistical power of the model. 59

PAGE 70

Non-White Model Table 3 shows (see below) the complete SEM estimates for the nonwhite model. The model fit indexes indicate that it is a well fit model (CMINIDR=1.152, GFI=O.981, AFGI=O.965, RMR=O.104, and RMSEA=O.016). Again, all these indices, expect for RMR, meet the general criteria for these specific indices. 60

PAGE 71

Table 3. Estimated Standardized Direct, Indirect, and Total Effects of Strains and Negative Stimuli of Female Offenders for tbe Non-White Sample (N=S71) Dependent Independent Standardized Structural Coefficient Variable Variable Direct Effect Indirect Effect Total Effect Family -0.073 -0.037 -0.110 Structure = 0.278 Peer Deviant -0.122" -0.099 -0.222"" Behavior Parental Deviant -0.076 -0.043 -0.120 Behavior Abuse 0.118 -0.101 0.017 Victimization Substance 0.411 "" 0.411"" Abuse Family -0.070 0.005 -0.065 = 0.071 Structure Peer Deviant 0.188"" 0.001 0.189"" Behavior Parental Deviant 0.115 -0.009 0.106 Behavior Abuse -0.026 0.005 -0.021 Victimization Substance -0.022 -0.022 Abuse = 0.300 Family -0.084 0.038 -0.047 Structure Peer Deviant -0.252"" -0.060 -0.312"" Behavior Parental Deviant -0.128 -0.137 -0.265"" Behavior Abuse -0.246" -0.246"" Victimization Family -0.153 -0.153 Structure Peer Deviant 0.244"" 0.244"" Be!tavior Parental Deviant 0.555" 0.555"" Behavior Tests of Model Fit DF (P-value) = 96.763/84 (0.161) CMINIDF = 1.152 GFI=0.981 AGFI 0.965 RMR=Q.I04 RMSEA 0.016 Note: significant at a=.05 level, and significant at a=.OI level.

PAGE 72

Abuse Victimization. The R2 for abuse victimization is 0.386, indicating that about 39% of the variation in abuse victimization can be explained by the 5 exogenous variables. Parental deviant behavior (0.56**) and peer deviant behavior (0.24**) both have a significant direct effect on abuse victimization. It means that for non-white women, exposure to the negative socialization agents tends to lead to abuse victimization. Different from the whole and white models, the variable of family structure became insignificant in the non-white model. Substance Abuse. The result shows that the R2 for substance abuse is 0.3, meaning that 30% of the variation in substance abuse can be explained by the structural model. The significant predictors include peer deviant behavior (-0.252**) and abuse victimization (-0.246**). These results suggest that minority females who were associated with delinquent peers, or were physically or sexually victimized, are more likely to using substances at a younger age. This finding is also consistent with Arnold's 1990 study in which she interviewed a number of African American female inmates and discovered that many of them used substances to self medicate and block out their negative experiences (Amold, 1990). Same as in the white model, parental deviant behavior only has a significant total effect (-0.27") on substance abuse. I speculate that both the direct (-0.128) and indirect (-0.137) effects would become significant if the sample size were large enough. 62

PAGE 73

Frequency of Arrest. The W for frequency of arrest was 0.07, indicating that the structural model can explain 7% of the variation of this faCtor. 'Among explanatory factors, only peer deviant behavior has a significant direct effect (0.19**). It seems that minority female offenders, the presence of negative peer influence in their lives is a major factor coniributing to their younger age and high frequency of being incarcerated. Age of First Arrest. The R2 for age of first arrest is 0.29, implying that about 29% of the variation in age of first arrest can be explained by the structural model. The factors that have a significant direct effect.include substance abuse "(0.41 **) and peer deviant behavior (-0.12*). The interpretation is that minority females who started to use substances at an earlier age who with delinquent peers tend to have troubles with abiding by the law and being arrested at a younger age. Comparison of Sub-sample Models Table 4 (see below) shows a comparison6fthe white and non-white models. This is an effective way to detect any racial differences in the effects of all social and behavioral factors regarding arrest records between white and-non-white female offenders. The assumption is that if there are differential patterns for white and non white female offenders respectively, the corresponding or comparable structw"al coefficients the white and non-white SEM models should be significantly different. 63

PAGE 74

Table 4. Comparison of White and Non-White Models Dependent Independent Direct Effect Indirect Effect Total Effect Variable Variable White Non-White White Non-White White Non-White Family -0.20 -0.07 0.10 -0.04 -0.10 -0.11 Structure Peer Deviant -0.03 -0.12 -0.30 -0.10 -0.33 -0.22 Behavior Parental 0.10 -0.08 -0.13 -0.04 -0.03 -0.12 Deviant Behavior Abuse -0.01 0.12 -0.17 -0.10 -0.17 0.02 Victimizalion Substance 0.57 0.57 Abuse Family 0.16 -0.07 -0.06 0.01 0.11 -0.07 Structure Peer Deviant 0.11 0.19*" 0:08 0.00 0.19 0.19 Behavior Parental -0.16 0.12 0.08 -0.01 -0.07 0.11 Deviant Behavior Abuse 0.10 -0.03 0.03 0.01 -0.13 -0.02 Victimization Substance -0.10 -0.02 -0.10 -0.02 Abuse Family 0.06 -0.084 0.12 0.04 0.17 -0.05 Structure Peer Deviant -0.47 -0.252 -0.06 -0.06 -0.53 -0.31 Behavior Parental -0.05 -0.128 -0.18 -0.14 -0.28** -0.27 Deviant Behavior Abuse -0.30 -0.25 -0.29 -0.25 VictimizatioR Family -0.40 -0.15 -0.15 structure Peer Deviant 0.22 0.24 0.22 0.24 Behavior Parental 0.62 0.56 0.62 0.56 Deviant Behavior Note: significant at C1=.OS level, and significant at C1=.OI level.

PAGE 75

The comparison uncovered 4 significant differences between thewhite and non-white models. First, family structure has a significant direct impact on age of first arrest for the white female offenders (-0.20*), but not for the non-white female offenders (-0.07). This suggests that white female offenders are more likely to be arrested at a younger age if they did not live with both their parents while growing up, their parents were on welfare, and/or their family lived a public housing. adverse family structure seems to be more influential developing deviant behavior in white females than in minority females The second difference between white and non-white models is the effect of peer deviant behavior on age of first arrest. The results show that peer deviant behavior is a significant predictor (-0.12*) for non-white female offenders, but not for white female offenders (-0.03). The third discrepancy between the two models relates to the effect of peer deviant behavior on frequency of arrest. Table 4 shows that peer deviant behavior a significant effect on frequency of arrest for minority female offenders (-0.19**); but not for white female offenders (-0.11). When considering the last two differences together, one will find that peer deviant behavior seems to have a particularly strong and consistent impact on incarceration records of minority female offenders. However, peer deviant behavior had no effect at all on arrest records of white female offenders. This suggests that peer influence plays quite different roles among white and non-white female offenders, in terms of pushing them down the path of 65

PAGE 76

criminality. Minorities draw their behavioral influence more from their peers than from any other sources. As for why this is not the case for white female offenders, the answer may lie in the different family structures. According to National Center for Health Statistics, in 1995, about 70% of African-American children were born out of wedlock and about 90 % of them live female-headed households (National Center for Health Statistics, 1996). Consequently, African American households lack of father figures desired to help reinforce social nonns and assist with behavioral modification. As a result, the family influence, which could be the last defense to ward off negative peer influence on non-white children, was greatly weakened or non-existent. This may result in non-white female offenders seeking alternative guidance from their peers. Prior research has found, that when there is an interaction deficiency between parents and children, adolescents are more likely to seek guidance from their peers and withdraw from family communication (Barnes Farrell, 1992). As they are more identified with their peers, their behavior, according Sutherland (1934, 1947), is more similar to their peers' On the other since white female children are more likely to live in traditional two parent households comparison to black children (Thornton, Chatters, Taylor, Allen, they are more likely to be socialized in an environment where informal social control mechanisms are much stronger, and deviant peer influence is less likely to penetrate. 66

PAGE 77

The fourth difference between the white and non-white models lies in the effects of peer deviant behavior on female offenders' substance abuse. The results show that peer deviant behavior is a significant factor in both white and non-white models. However, the difference is that the effect for the white model (-0.40* *) is almost twice the size of the effect for the non-white model (-0.25**). Although, this continues the pattern of significant peer influence for non-white female offenders, it suggests that peer deviant behavior plays a particularly important role enticing white females to travel down the path of substance abuse. This is consistent with previous research that reports deviant peer influence was a stronger predictor for the delinquency of white females (Katz, 2000, Barnes & Farrell, 1992). Overall, these findings confirm the well recognized relationship between peer influence and substance abuse (Ary, Tildesley, Hops, Andrew, 1993; Urberg, Degirrnencioglu, Pilgrim, 1997). general, the major difference between the white and non-white models is the function of peer deviant behavior:. PeeJ;deviant behavior is a significant factor for every endogenous variable (Frequency of arrest, age of first arrest, substance abuse, and abuse victimization) the non-white model, while it is significant only for 2 of the 4 endogenous variables (substance abuse and abuse victimization) the white model. This implies that peer influence is a ubiquitous factor the lives of non white female offenders, and is pushing them down the path of delinquency forcefully. This result also supports previous research that suggests that peer influence an 67

PAGE 78

individual is exposed to directly effects whether that individual will display deviant behaviors throughout their life (Flom, 2001; Urberg, 1992; Warr Stafford', 1991). The remainder of the two models is basically the same. They confmned, to various degrees, socialization, differential association, and general strain theories. Females who were raised by deviant parents and socialized with delinquent peers were surrounded by negative stimuli, and this often results their traumatic experiences. The abuse victimization in may lead to substance abuse as a fonn of therapy, which is, indeed, a destructive way ofrespondirtg to the depressive personal situations. The results also show that for both sub-sample models, parental deviant behavior a significant total effect on the respondent's initial age of substance use, while the direct effect is not significant. This additional effort of estimating total effects allows us to discover the overall impact of parental deviant behavior on substance abuse, which would be missed otherwise. The inclusion of the total effects into the analytical picture certainly provides a new means gauge the real impact of parental deviant behavior on substance abuse. And the results eventually strongly supported Sutherland's differential association theory and Agnew's general strain theory. 68

PAGE 79

Significant Findings and Contributions The following section explains in detail how my research questions were answered. for the whole model and the additional two sub-models. The major findings of this study include the following. For the whole model, (1). Peer deviant behavior has a significant influence on substance abuse, abuse victimization, frequency of arrest, and age of frrst arrest (arranged in a descending order by the standardized structural coefficients) of female offenders. (2). A female offender's substance abuse and family structure are significant predictors of her age of first arrest. (3). Physical and/or sexual abuse victimization of female offenders contributes significantly to their behavior of substances abuse. (4). Parental deviant behavior subjects female offenders to victim of physical and/or sexual abuse. (5). Parental deviant behavior indirectly leads to female offenders' substance abuse. For the comparison of white and non-white models: (1). Peer deviant behavior a significant influence on substance abuse, abuse victimization, frequency of arrest, and age offrrst arrest (arranged in a descending order by the standardized structural coefficients) of non-white female offenders. And it only has a significant effect on substance abuse and abuse 69

PAGE 80

victimization of white female offenders. It seems that negative peer influence is ubiquitous for non-white female inmates, while it is not significantly related to arrest records of white female offenders. (2). Family structure is a significant predictor of age of first arrest for white female offenders, but not for non-white female offenders. (3). The effect of peer deviant behavior on substance abuse for white female offenders is much stronger" than that for non-white female offenders. This study anchors in a socialization framework that combines Sutherland's differential association, Agnew's general strain theory, and theories of feminist criminologists. It intends to explain the recent trend and behavioral patternS of female criminality in terms of socialization agents and process, family structure and socioeconomic status, environmental and economic strains, exposure to negative stimuli and behavioral models, and victimization experience and destructive behavior of female offenders. This study adopts a systematic approach that connects current behavior of a female offender to a process linking back to her childhood environment. The study also accommodates her entire social relations and interactions with the key" agents involved her socialization process. The results of this study could be unique and informative because of these theoretical and analytical considerations. The current analysis is based on a more advanced statistical technique, structural equation modeling. The strengths of the SEM lie its simultaneous estimation of complex causal models, its ability to investigate direct, indirect, and 70

PAGE 81

total effects, and its unique capability to handle latent and manifest variables. The benefit of the SEM is tremendous and this study taken a full advantage of it. The model setup, the data analysis, and the final results of this study should reflect the power of the technique. So far, only a handful studies in the field of criminology have adopted the SEM technique, and many researchers have not yet fully realized the necessity in certain situation (e.g., with abstract concepts) to use the SEM technology. The current investigation focuses particularly on the racial impact of the causal structure and the dynamics described in the general model. This is necessary in order to comprehend today's harsh reality that a disproportionally high percentage of minority females ate in our prison population. This model design is different from simply controlling for race in a regression model. provides a side-by-side clear comparison between the two populations, and it looks for the effect of race on structural relationships rather than on particular variables. Overall, this study accomplished the initial goals of determining which strains female offenders (in state prisons) were subjected to prior to their arrest. also detennined the racial differences, and how particular strains affected the respondents. Limitations and Future Research First, the theoretical framework of this study sets out to be comprehensive and systematic by combining several major theories to explain the origin, process, and 71

PAGE 82

causes of female criminality. However, the explanatory factors included in the fmal model, though among the most important, are by no means exhaustive. To some extent, it is still limited in terms of fully representing the real processes and relationships in the life course of female offenders. other words, there is still a 0 distance in truthfully modeling the reality. For example, the model failed to include any subjective measures to gauge the attitudes, beliefs, and values female offenders may have. We need these types of variables to measure the levels of moral development and normative compliance/rebellionamong female offenders. Without this dimension it is unclear under what conditions person takes drugs. For example, some people take drugs because. they are willing to, and they believe nothing is wrong with their behavior. For other-individuals taking drugs may be counter-attitudinal and they suffer serious cognitive inconsistency. Even if behavior wise, both situations involve substance abuse, they have different meanings and effects when we try to use them to predict age and/or frequency oof arrest. By utilizing secondary data, I was limited to the variables used by the original researchers and the ways in which they operationalized those variables. While I felt the study was well conducted and provided a sample that could be generalized to females state prisons, there was a lack of subjective and sociological variables (i.e. interaction and relationship variables). Additionally, when dealing with secondary data in which human subjects were involved, and especially when interviewing 'prisoners, there are validity issues. The researchers could not control one hundred percent for the respondents who may 72

PAGE 83

have given socially desirable answers, lied, or had reconstructive memory problems. In this study I took preventative measures and removed all outliers in order to effectively control for these issues. There also was a reliability issue, as the reliability coefficient was not reported for the measurement instrument. Second, as we all know crime rates tend to cluster, i.e. to concentrate in poor neighborhoods or households. For that reason, I might need to take a closer look at how different household and/or neighborhood conditions or characteristics shape individuals' deviant/criminal behavior. The fmdings of the current study may confound individuals' results with household or neighborhood effects. Given the available data set, the current model may be the best option. However, a multilevel analysis might bring out results closer to the reality. Third, measurement issues are always a key concern in any empirical study. Though I tried my best, a fewfactors the model still could have been measured better. For example, all of the indicators used for family structure are dichotomous, and the indicators for this latent concept only measured structural factors. I used all the relevant variables the data set to measure the latent concepts. Even if it this is not a serious problem, as they are all exogenous independent variables, woUld be ideal if they were measured at the interval/ratio level. Poor measurement could be the major reason for the wrong significance directions of the latent variable the model. In addition, it must be noted that most of the variables chosen from the data set represent structural factors, because the did not contain any variables on 73

PAGE 84

interaction. For instance, I specifically looked at socialization factors, yet none of my indicators related to interaction. the researcher, I am fully aware of this issue, and that the aim of this study was to look at socialization by examining structural factors and the influence of deviant peers and parents. Lastly, crimes committed by women are different from those by men, and most of the crimes women committed are non-violent crimes. It may be strategically a more effective design to use specific crime instead of crime general as dependent variables. For example, to differentiate between violent crime and property crime or to use drug related crime only. The benefit here is very clear. For instance, in the current moclel there are independent variables with reference to respondent, parent, and peer drug use. If the dependent variables were limited to drug related crime, the connection from respondent, parent, and peer drug use to the dependent variables, age and frequency of arrest during to drug offense, is very intuitive and clear. If the dependent variables include murder, for instance, the link between drug use and murder is not that certain and needs intennediate variables and extra justification. There is a general call for more studies to be conducted on female criminality, as there is an increasing amount of juvenile and adult female offenders. Future research is necessary to comprehend the significant racial differences that were found in the underlying strains that push females in to the criminal justice system. Qualitative research certainly has the potential to provide a more in-depth understanding of these racial differences. Once researchers understand these racial 74

PAGE 85

differences greater depth, then preventative programs can effectively be implemented for female adolescents. Hopefully through these programs, the strains that females experience can be identified and properly addressed. It is imperative that society understands that females experience different strains than males. Once this is accomplished, hopefully we will understand female offenders better, and have the ability to effectively deter them from committing crimes. Future research should also compare male and female based on this model. One could easily incorporate the same two sub-models by using one sample for males and another for females, in order to detennine which are for both males and females, and which strains are stronger predictors for females. This study also sets up the enabling future research to compare female offenders in state prisons to female offenders in federal prisons. This will determine which strains are significant for these two different populations Finally, by utilizing structural equation model this study found significant racial etIects. I believe that this is a major contribution to the field of criminology and sociology, since not many researchers have examined these differences using this statistic technique. I found that the most compelling fmdings and contributions were that negative peer influence a compelling effect for non-white female offenders in areas examined. I believe that future qualitative research focusing on these findings; will lead to significant gains and a more comprehensive understanding of this phenomenon. 75

PAGE 86

REFERENCES Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Agnew, R. (2001). Building on the foundation of general strain theory: Specifying the types of strain most likely to lead to crime and delinquency. 38, 319-361. Agnew, R. (2002). Experienced vicarious, and anticipated strain: An exploratory study on physical victimization and delinquency. 603-632. Akers, R.L. (1985). Belmont, CA: Wadsworth. Akers, R. L., Krohn, M. D., Lanza-Kaduce, L., Radosevich, M. (1979). Socialleaming and deviant behavior: A specific test of general theory. 44, 636-655. Anderson, M. L. (2000). (4th ed.). Boston: Allyn and Bacon. Arnold, R. (1990). Women of color: Processes of victimization and criminalization of black women. 17, 153-166. Ary, D.V., Tildesley, E., Hops, H., & Andrews, J. (1993). The influence of parent, siblings, and peer modeling and attitudes on adolescent use of alcohol. 28,853-880. Aseltine, R. H. (1995). A reconsideration of parental and peer influences on adolescent deviance. 36, 103-121. Barkan, S. E. (1997). Upper Saddle River, NJ: Prentice Hall. 76

PAGE 87

Barnes, G.M., Farrell, M.P., Cairns, A (1986). Parental socialization factors and adolescent drinking behaviors. 48,27-36. Barnes, G.M. Farrell, M.P. (1992). Parental support and control as predictors of adolescent drinking, delinquency, and related problem behavior. 54, 763-776. Barnes, G.M., Farrell, M.P., BaneIjee, S. (1994). Family influences on alcohol abuse and other problem behaviors among black and white adolescents in a general population sample. 4, 183-201. Baron, S. (1999). Street youth and substance use. & 31,3-26. Beckman, L. J. (1984). Treatment needs of women alcoholics. 1, 101-114. Belknap, J. (2001). Canada: Wadsworth. Benson, P. L. (1990). The Minneapolis: Search Institute. Blane, H. T, Miller, B. A, Leonard, (1988). Intra-and intergenerational as aspects of serious domestic violence and alcohol and drugs. Washington, D.C.: National Institute of Justice. Brannock, J. C., Schandler, S. L., & Oncley, P. (1990). Cross-cultural and cognitive factors examined groups of adolescent drinkers. Broidy, L., & Agnew, (1997). Gender and crime: A general strain theory perspective. 34, 275-302. Brooks, A F., Whiteman, M., Gordon, A S. (1983). Stages of use in adolescence: Personality, peer and family correlates. 269-277. Byrne, M. W. (1989). New York: Springer-Verlag. 77

PAGE 88

Cannines, E. G., Zeller, R. A. (1979): Beverly Hills: Sage Publications. Chesneyl-Lind, M., Shelden, R. G. (1998). Belmont, CA:Wadsworth Publishing. Colvin, M., & Pauly, J. (1983). A critique of criminology: Toward an integrated structural-Marxist theory of delinquency prevention. 89, 513-545. Coombs, R., Fawzy, F. I., Gerber, B. E. (1986). Patterns of cigarette, alcohol, and other drug use among children and adolescents: A longitudinal study .. 21, Daly, K. (1992). Women's pathways to felony court: Feminist theories of lawbreaking and problems of representation. 2, 11-52. Dembo, R., Dertke, M., La L., Borders, S., Washburn; M., & Schmeidler, J. (1987). Physical abuse, sexual victimization and illicit drug use: A structural analysis among high risk adolescents. Dornbusch, S. Carl smith, J. Bushwell, S. J., Ritter, P. L., Leidennan. R., Rastorf, A. R., Gross, T. (1985). Single parents, extended households, and the control of adolescents. 56, 326-341. Duncan, G. J., & Brooks-Gunn, J. (1997). New York: Russell Sage. Duncan, G. J., Brooks-Gunn, J., & Klebanov, P. A. (1994). Economic deprivation and early childhood development. 65,296-318. Federal Bureau of Investigation. (1992). Washington, D.C. U.S.: Department of Justice Flavin, J. (2001). Feminism for the mainstream criminologist. 29, 271-285. Flom, P. L. (2001). Recalled adolescent peer norms towards drug use young adulthood a low income, minority urban area. 31, 425-437. 78

PAGE 89

Gaarder, E., & Belknap, J. (2002). Tenuous borders: Girls transferred to adult court. GaraSky, S. (1995). The effects offamily structure on educational attainment: Do the effects vary by the age of the child? 54,89-111. Gilfus, M. E. (1992). From victims to survivors to offenders: Women's routes of entry and immersion into street crime. 4, 63. 90. Greenfield, L. A, Snell, T., L. (1999). Bureau of Justice Statistics Special Report. Henderson, K A (1994). Perspectives on analyzing gender, women, and leisure. 26, 119-137. Hill, S. A., & Sprague,J. (1999). Parenting black and white families: The interaction of gender race and class. 13, 480-502. Hoffman-Bustamante, D. (1973). The nature of female criminality. 8, 117-137. Hoffman, P., Susan S. (1998). Parental substance use disorder, mediating variables and adolescent drug use: non-recursive mode1. 93, 1351-1364. Hu, L., Bentler, P. M. (1999). Cutoff criteria for fit indexes in stucture analysis: Conventional Criteria Versus. 6, 1-55. Joreskog, G., Sorbom, D. (1993). 8 Chicago: Scientific Software International. Katz, R. S. (2000). Explaining girl's and women's crime and desistance in the context of their victimization experiences: developmental test of revised strain theoJY and life course perspective. 6,633-661. 79

PAGE 90

Kimmel, M. S. (2000). New York: Oxford University Press. Klein, (1984). Cambridge, MA: Harvard University Press. J.H., & Wojtkiewicz, R.A. (1992). A new look at the effects offamily structure on status attainment. 73,581-595. Loeber, R, Dishion, T. (1982). Early predictors of male delinquency: A review. 94, 68-99. Loehlin, J. C. (1987). New Jersey: Lawrence Erlbaum Associates. Long, J. S. (1983). Beverly Hills: Sage Publishers. Mayer,S. E., Jencks, C. (1989). Growing up poor neighborhoods: How much does it matter. 243, 1441-1445. McClellan, D. S.,Farabee, D., & Crouch,B. (1997). Early drug use, and criminality: A .comparison of male and female prisoners. 24, 455-489. R. (1938). Social structure and anomie. 3, 672-682. Miller, A. (1993). The impact of family violence on the use of alcohol by women: Research indicates the women with alcohol problems have experienced high rates of violence during their childhoods and as adults. 17, 137-153. Miller, M., Downs, W. R., Gondoli, D, M, & Keil, A. (1987). The role of childhood sexual abuse the development of alcoholism women. 2, 157-172. Miller, S. (1998). Thousand Oaks, C.A.: Sage Publications. National Center for Health Statistics. (1996). [on-line], Available: http://www.cdc.gov/nchs/. 80

PAGE 91

Newcomer, S., Udry, J. (1987). Parental marital status effects on adolescent sexual behavior. 49,235-240. O'Hare, W. P., Pollard, M, T. L., Kent, M. (1991). African-americans the 1990's. 46, 1-7. President's Commission on Law Enforcement and Administration of Justice. Washington, D.C.: U.S. Government Printing Office. Price, B., Sokoloff, N. (1995). New York: McGraw-Hill, Inc. Rose, S. M, Peabody, C. G., Stratigeas, B. (1991). Undetected abuse among intensive case management clients. 499-502. Scott, J. W. (1986). Gender: a useful category for historical analysis. 91, 1053-1075. Snell, T. (1991). Bureau of Justice Statistics Special Report. Sutherland, E. (1934). (2nd ed.) Philadelphia: Lippincott. Sutherland, E. (1947). (4ed). Philadelphia: Lippincott. Tabachnick, B. G., Fidell. L. S. (1996). (3rd Ed). California State University, Northridge: Harper Collins College Publishers. Taylor, J., Chatters, M., Tucker, M. B., & Lewis, E. (1990). Developments research on black families: A decade of review. 52, 993-1014. Thomas, G., Farrell, M.P., Barnes, G. M. (1996). The effects of single mother families and nonresident fathers on delinquency and substance abuse in black and white adolescents. 58, 884-894. 81

PAGE 92

Thomas, W.I. (1923). Boston: Little, Brown. Thornton, M. C., Chatters, M., Taylor, R. J., Allen, W. R. (1990). Sociodemograpbic and environmental correlates of racial socialization by black parents. 61,401-409. Toray, T., C. (1991). Gender differences associated adolescent substance abuse: Comparisons and implications for treatment. 43, 338-349. Turner, R. A., C. E., Millstone, S. G. (1991). Family structure, family processes, and experimenting substances, during adolescence. 1, 93-106. United States Department of Justice Bureau of Justice Statistics and United States Department of Justice Federal Bureau of Prisons. (1997). United States Bureau of the Census. (1990). 18 (Current Population Reports, Series 20, No. 447).Wasbington, DC: U.S. Government Printing Office. Urberg, A. (1992). Locus of peer influence: Social crowds and best friends. 21, 439-450. Urberg, A., Degirmencioglu, S. M., Pilgrim, C. (1997). Close friends and group influence of adolescent cigarette smoking and alcohol use. 33, 834-844. Wallace, Goddard R., Goff, B. G., Melancon, M. V. (2000). Profiles of delinquency: A comparison of delinquent behavioral groups. 15, 19-32. Walker, S., Spohn, C., DeLone, M. (1999). Belmont, CA: Wadsworth. Warr, M., & Stafford, M. (1991). The influence of delinquent peers: What they think or what they do? 29, 851-866. Watts, D. S., Watts, K. M (1991). Impact of female-headed parental families on academic achievement. 17, 97-114. 82