«my» rm. ml upw’ :x..::-. n .rn‘v Me- n 393‘? m. a. .. .4" 22::— .‘..w my mg}. 4th.... .7!- u I. . - m- .... «:L m .4 M u . ‘ ~ I- ”w .. -,,,_ a. - A. .. 5.52.41... mu uuau" u.”- 4....nnh- ‘ ”0-. 3 Lu. .. . A ’g» 1-way- y... ‘r . 3'. " ’N: {I i-tszgni‘fi; s u,— «w.»- nuuwrwqo-n .o‘uud . .u , .. ,- .43, Ar 9 '3“ '3 “5:5 ' a 12-5,, .‘H 9 ‘ .2 33";ggizi 6: ‘ _ y l T: 2:32 " . g ., ‘ . ' #1521329“! in W: 3‘8“»? it , . nu n»..— «v 8 ' a ‘i‘35 ‘ . 'y- k: 1: 3 ‘lozzi, ‘ grv i I}: want )4»- 40“...» -.v “v. on vvn‘u'Q-nl .wummvw ., u -o-‘ m .. :85!" 3 ? "vtfgr'fl; "Ii-"4m - ‘ ’ .. .. —u A ”a" .. “.m- 9.. "m ,5.» .quJ. ’58“. ..... "45‘ ph'l‘h¢ a: v “‘35" 8" ,.~L— % w: a nu w .- .¢. "L4- 0' ' ”1. .. .m {a ‘35:": pass-s L1 M Illlii/giizifliig/LiifliiiWit 5440 This is to certify that the dissertation entitled PREDICTIONS FOR PARTNER ASSAILANTS AT A PROBATION DEPARTMENT presented by DAVID CANALES-PORTALATIN has been accepted towards fulfillment of the requirements for PhD Psychoiogy degree in I Major professor William S. Davidson, II Date 7/ 14/95 MS U i: an Affirmative Action/Equal Opportunity Institution 0-12771 LiBR‘ARY‘ Michigan State University PLACE IN RETURN BOX to remove this chockout from your record. TO AVOID FINES Mum on or botoro dot. duo. DATE DUE DATE DUE DATE DUE PREDICTIONS FOR PARTNER ASSAILANTS AT A PROBATION DEPARTMENT BY David Canales-Portalatin A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1996 ABSTRACT PREDICTIONS FOR PARTNER ASSAILANTS AT A PROBATION DEPARTMENT BY David Canales-Portalatin Researchers have been investigating the problem of partner assault either by focusing on police intervention or treatment programs, but these individuals have not been studied through looking at data at the court system. The current study intends to predict the sentences and A m v>—mm‘ recidivism of partner assailants based on their demographic” _. characteristics. This study identified 182 males who had assaulted female partners during the period from 1988 through 1990, from archive files at a probation department. M— in Michigan. With the aid of the Michigan State Police, these cases were followed in order to identify subsequent EE£§§§§ for assaults within 36 months. The prediction models analyzed presented weak relationships with the Hfl - .____ -w . kw- ’__, predictor variables, although these relationships improved when criminal history variables wereincluded. These findingsflcould beinterpreted in two different ways according to the prediction philosophy that the reader prefers.fibéupporters of this concept of prediction could hold that the findings in the current study support the idea that predictions could be made if better variables were USE inte pre- pos; adva J): / used. Opponents of predictions, on the other hand, could interpret the current findings as support for the idea that hp predictions cannot be achieved. Whatever philosophical ..p.——»- % position is taken, prediction should continue in order to advance knowledge in this area. Copyright by DAVID CANALES-PORTALATIN 19 9 6 To my partner, Michael C. Ennis-McMillan, for his supporting role during my doctoral studies and ACKNOWLEDGEMENTS I am extremely thankful to my partner, Michael C. Ennis McMillan, to whom I am dedicating this dissertation because he provided me with support to complete this research. He was kind enough to take time to read and make wonderful suggestions at the earlier stages of this difficult project. He patiently listened to my topic and its limitations. I also want to say thank you to Nancy McCrohan, who also provided me with the support to complete this project and who also read some of the chapters of the dissertation and provided me with valuable suggestions. I want to thank the members of my guidance committee for their desire to work with me on this controversial topic of partner assault and for providing me with suggestions: Professors Michael Lambert, Laura Smale, Cris Sullivan, and William 8. Davidson II, Chair of my committee. Thank you to Dr. Davidson, who provided me with almost weekly feedback, and to Dr. Sullivan for her editorial comments on my manuscript. I also thank the following programs for assistance during my doctoral program: Minority Competitive Doctoral Fellowship, and Martin Luther King, César Chavez, Rosa Parks vi Fella and c Fellowship. These fellowships made it possible to continue and complete my doctoral studies and research. I want to thank the probation departments, which provided me with the opportunity to look at their files. Without their agreement, this project would not have been possible. The names of these departments need to remain confidential as part of our agreement. Also, thank you to Ms. Echo Swin of the Michigan State Police for allowing me to conduct follow-up on names of assailants in the records division. I also want to thank the students who helped me to collect data in the field. Without their help the data collection would have been very difficult. Their names include: Jill Barton, Leonard Bloch, Evette Charboneau, Jennifer Juras, Jody Martek, Kelly Salada, Alissa Tonjes, Stephen Valente, and Matthew Winstanley. I am also extremely thankful to Professor Donald H. Williams, who provided me with support during this project. When others closed their doors, Dr. Williams was able to open his and provided the material and emotional help that I needed. I also want to thank my friends in the community who supported me and my desire to finish this project. These friends were willing to put up with my lack of time for them. Thank you for your patience and understanding: Eufracio Abaya, Encarnita Figueroa, Rick Greiner, David Lounsbury, and Jerod Scholten. vii of Cr me Hi inclu Willi Coto, me to model: justi: contii of en: I want to thank the faculty members in the Department of Criminal Justice at Indiana University, who also provided me with support while revising of this document. These include: Professors Harold E. Pepinsky, Coramae Mann, William Oliver, Leon Pettiway, and Alex Weiss. Finally, I want to thank my mother, Matilde Portalatin Coto, and my father, Florencio Canales Ortiz, who encouraged me to complete an advanced degree. They have been role models for me throughout my live, always striving for justice and human dignity. Much of the motivation for my continuing my formal education is because of their example of encouragement. To all of you who helped me and love me, iMUCHAS GRACIAS! viii S; CHAPTER I: INTRODUCTION 1 . . . . . . . . . . . . CHAPTER II: METHOD TABLE OF CONTENTS LI ST OF TABLES O O O O O O O O O O O O O O O O O O O Terminology . . . . . . . . . . . . . . . . . Magnitude of Partner Assault . . . Criminal Justice Responses to Partner Assault Police Performance . . . . . . . . . . . Critique of Police Performance . . . Policies of Criminal Justice Organizations Summary of Criminal Justice Responses to Partner Assault . . . . . . . . . . . Treatment Programs . . . . . . . . . . . . . . . Assailant Characteristics . . . . . . . . . Critique of Assailant Characteristics Characteristics of Repeat Partner Assailants Critique of Characteristics of Repeat Partner Assailants . . . . . . . Summary of Characteristics of Repeat Partner Assailants . . . . . . . Summary of Treatment Programs . . . . . . . Characteristics of Repeat Offenders For All Crimes . . . . . . . . . . . . . . . . . . General Prediction of Violence . . . . . . Actuarial Predictions in Criminal Justice . Summary of Characteristics of Repeat Offenders . . . . . . . . . . . Ecological Assessment of Repeat Offenders . . . Summary of Ecological Assessment of Repeat Offenders . . . . . . . . . . . . . . The Current Research . . . . . . . . . . . . . . Hypotheses . . . . . . . . . . . . . . . . Subjects . . . . Setting . . Apparatus . . . . Pilot Studies and Instrument Construction Ontogenetic Level . . . . . . . . . . . Age at Time of Initial Offense Marital Status . . . . . . . . Education . . . . . . . . . . . ix xiii ODm-hH 19 23 26 27 30 31 4O 43 44 47 50 51 58 6O 66 66 74 77 77 77 78 79 80 81 81 81 Employment Status . . . . Income . . . . . . . Defendant's Ethnicity . . Microsystem Level . . . . . . . . . Victim and Assailant Livi 5 ‘Qoooo Together . . . . . Victim’s Ethnicity . . Victim's Age . . . . . Assaultive Incident . . Involvement of Substances Exosystem Level . . . . . . . . . . Number of Prior Arrests . Type of Last Offense . . Days in Jail . . . . . . Restitution to Victim . . Money Paid to the Court . . Days of Probation Supervision . Assignment for Domestic Violence Program . . . . . . . . . Subsequent Contact with Criminal Justice System for an Assault . . . . . . Summary of Variables in the Study . Procedure . . . . . . . . . . . . . . . »v/ Power Analysis . . . . . . . . . . CHAPTER III: RESULTS . . . . . . . . . . . . Descriptive Statistics . . . . . . . . . Ontogenetic Level . . . . . . . . . Age at time of initial offense Marital Status . . . . . . . Education . . . . . . . . Employment Status . . . . Income . . . . . . . . . . Defendant's Ethnicity . . . Summary of Ontogenetic Level Microsystem Level . . . . . . . . Victim and Assailant Living Together Relationship Between Assailant and Victim . . . . . . . . . Victim's Ethnicity . . . . Victim's Age . . . . . . . . Assaultive Incident . . . . Involvement of Substances . Racial Match . . . . . . . . Summary of Microsystem Level Exosystem Level . . . . . . . . . on Number of Prior Arrests Type of Last Offense . . Days in Jail . . . . . . Restitution to Victim . Money Paid to Court . . Days of Probation Supervisin X 82 82 82 83 83 84 84 85 85 86 88 88 89 89 90 90 90 90 92 92 96 98 98 98 99 99 99 100 100 100 100 100 101 101 101 102 102 102 102 104 104 104 105 105 106 106 106 ' I - CEZrED‘ I Assignment for Domestic Violence Program . . . . Summary of Exosystem Level . Subsequent Contact with Criminal Justice System for an Assault Data Reduction . . . . . . . . Inter-Correlation . . . . . Discriminant Analysis . . . Logistic Regression Analysis Interventions . . . . . Sentenced to Jail . Mandated to Substance Abuse Treatment Mandated for Partner Assault Abatement Treatment . . . Sentenced to Probatio Summary of Interventi Court Experiences . . . . Involvement of Substances . . Ecological Model . . . . Summary of Logistic Regression CHAPTER IV: DISCUSSION . . . . . General Findings . . . . . . . Hypothesis I . . . . . . Sentenced to Jail . n . ODS Mandated to Substance Abuse Treatment Mandated for Partner Assault Treatment . . . Sentence to Probation Summary of Hypothesis Hypothesis II . . . . . . Hypothesis III . . . . . Hypothesis IV . . . . . . Implications of the Results Policy Implications . Limitations of the Study . Future Research . . . . . Conclusion . . . . . . . . APPENDIX A: APPENDIX B: APPENDIX C: APPENDIX D: APPENDIX E: SEARCH ON PROBATIONERS PRELIMINARY DATA FORM I RECORDS Abatement LAST CODING FORM PROBATION DEPARTMENT DATA 0 O O O O O O 0 SUMMARY OF VALUE OF FACTOR ANALYSIS ROTATED MATRIX . . . APPROVAL TO OBTAIN INFORMATION FROM HUMANS . . . . . . . xi 106 106 107 107 108 114 123 124 125 130 134 134 137 139 139 140 141 144 144 155 155 156 158 159 162 164 165 165 166 169 171 174 176 179 181 185 188 189 LISTOFREFERENCES..................190 xii Tabl Tabl Tabli Table Table Table Table Table Table Tible Table Table Table Table Table Table Table Table Table Table Table LIST OF TABLES 1 - Characteristics of Partner Assailants . . 2 - Weight of Assaults According to Marshall 3 - Relationship Between Ecological Model and Variables of Study . . . . . . . - Percentages of Assaults . . . . . Reliability Analysys . . . . . . - Group Means for Court Sanctions . \l 0‘ UI uh I - Wilks Lambda Statistics . . . . . 8 - Logistic Regression for Sentenced 9 to Jail - Logistic Regression for Mandated to Substance-Abuse Treatment . . . . 10 - Logistic Regression for Probation superVis ion 7 O O O O O O O O O O 11 - Findings Comparison . . . . . . xiii 48 87 94 103 110 115 118 129 133 138 149 \ CHAPTER I INTRODUCTION An increasing number of studies have examined characteristics of partner assailants for the purpose of predicting who is most likely to commit an act of intimate violence (Hamberger & Hastings, 1990; Saunders, in press; Straus, 1993; Tolman & Bennett, 1990). Most of these studies, however, have selected their samples from treatment groups (Grusznski, 1986; Grusznski & Carrillo, 1988; Hamberger & Hastings, 1990). Although some of these samples have included men with court mandates to receive treatment, a high number have included assailants who entered treatment without such mandates (Edleson & Grusznski, 1988; Edleson & Tolman, 1992; Grusznski, 1986; Grusznski & Carrillo, 1988; Hamberger & Hastings, 1988, 1989, 1990). Thus, studies about the characteristics of men who have assaulted a partner have tended to examine those who have not had experience with criminal justice interventions. Therefore prediction of partner assault utilizing data from treatment programs excludes several men who have entered the criminal justice system but have not been mandated to attend group treatment. deg for the i of 52 inter Prose 2 Information on partner assailants at probation departments is limited. In the course of the current study, it has not been possible to find quantitative or qualitative research on partner assailants who came into contact with a probation department. Researchers have assumed that all the legal consequences a partner assailant confronts have been sufficiently investigated in experiments with police interventions and in the studies of group treatment programs for assailants. However, these studies were deficient in the examination of legal consequences and in the selection of samples. They failed to study the whole legal system’s intervention with these assailants, including the prosecutor's office, the court, and the probation department. Moreover, most of the studies with group treatment samples combined the findings from assailants referred by the court and from those who were volunteers (Rosenfeld, 1992). Therefore, findings from these studies contained biased results in that they did not include all the cases of partner assault in which criminal justice organizations typically intervene on a daily basis. One purpose of this chapter is to describe the problems with generalizing about the legal consequences of partner assaults on the basis of experiments with police. It will present the possibility of utilizing probation records to overcome some of these legal limitations. This chapter will then present the goal of the current research, which is to (a) determine which abusers are identified at a probation depart differ offici exanid Hichig T involv resear Organi Chapte 0f er that d' assail. ldndat‘ organi‘ fter . 3 department, (b) study prediction of recidivism with a sample different from those in traditional studies, and (c) predict official recidivism. The current goals are based on the examination of records from a probation department in Michigan. The current chapter addresses different perspectives involved in the problem of prediction of assault. Previous researchers have not connected these perspectives. To organize these different perspectives, this introductory chapter arranges the literature like a continuum of a series of events in a partner assault. For instance, let's assume that during a partner assault police were called, the assailant was arrested and brought to the court, and mandated for treatment. This analogy resembles the organization of the literature in the introductory chapter. After the terminology section, the first section describes the number of partner assaults in society. The second section describes the criminal justice response to the issue of partner assault. The third section describes the findings at treatment programs for partner assailants. The fourth section describes general prediction studies in criminal justice. The section before last explains the ecological model that will help organizing the variables to examine in the current research. The last section of this chapter presents the justification and rationale of the current study. were. part: Sauna 1950) of in the t1 relatj conjug Partne Various 4 Terminology Historically, studies of intimate male assaults on women have used the terms "husband and wife" to describe partner abuse (Dutton, 1985, 1987; Feld & Straus, 1989; Saunders, 1988, in press; Straus, Gelles, & Steinmetz, 1980). However, the current research recognizes a variety of intimate relationships between men and women and will use the term "partner." This term will include various forms of relationship, such as current and former legal marriage, conjugal living arrangements, and cohabitation. The term partner is more accurate because it encompasses these various relationships. Additionally, Michigan’s statutes use the terms "legal marriage," "conjugal living arrangements," or "cohabitation,” in referring to victims and assailants involved in domestic assaults (Field, 1993). These same terms are used in court documents, primarily because the Michigan judicial system follows Michigan’s statutes. Thus, the term partner is broad enough to describe the arrangements defined by the Michigan statutes. The cases selected for the current research will be those of men who assaulted a current or former female partner. Consequently, the defendants or assailants will be referred to with the pronoun "he," and the victims with the pronoun "she." Since women are rarely identified as assailants, and men are the focus of the current study, these pronoun references are correct. tré vic chi of i gene dSSd 5 Additionally, the problem of partner assault is traditionally referred to as "domestic violence" or ”family violence." These terms are general and can also include child abuse, elder abuse, sibling abuse, or the combination of all of these. This document, then, will avoid such general phrases and use the more specific term "partner assault" to describe the research problem. Magnitude of Partner Assault Due to a growing awareness of the social problem of partner assault, a number of studies have investigated the incidence in this form of violence (Dutton, 1987). Harlow (1991) estimated that each year 626,000 women are victims of violence by intimates. This average was based on violence that women considered criminal and that victims related to interviewers. Bell and Chance-Hill (1991) studied women who applied for divorce. The researchers found that between 37 and 50 percent of these women had experienced some form of violence in their relationship. Bell and Chance-Hill (1991) also found that from the number of women that come to the emergency room of hospitals, 3.4% was as result of their partner's battering. Twenty—one percent of women that required emergency surgery were battered and half of the injuries were the result of abuse (Bell & Chance-Hill, 1991). Langan and Innes (1986) presented the percentage of murders in relation to the victim from the 1984 FBI Uniform Crime Report. This report shows that the highest percentage (5%) i From - partn“ of th in th lillic Violen of 54% only ti 1989 t. 6 (5%) was from husbands killing wives (Langan & Innes, 1986). From the same data, in the category of domestic violence, partner represented 40% of the incidence, and ex-partner 19% of the incidence. Sherman (1992) later stated that police in the United States confront cases of partner assault 8 million times every year. In Michigan, the Domestic Violence Prevention and Treatment Board reported an increase of 54% in the number of assaults reported to the police in only three years--that is, an increase from 19,416 cases in 1989 to 29,891 cases in 1992. Dutton (1985) and Langan and Innes (1986) have argued that an abused woman is likely to be victimized again once abuse has occurred. Furthermore, Feld and Straus (1989) found that two-thirds of couples who experienced assaults claimed to have experienced more than one assaultive incident. This research also shows that severe assaults on women reported to the police often are the culmination of a trend of minor assaults initiated by husbands earlier in the relationship (Feld & Straus, 1989). Thus a possible reason for the high incidence of partner assaults is the repetition of assault. Therefore, as indicated by this literature, the controversy over partner assault has two dimensions. One dimension is the number of incidents, and the second is the operational definition used in the research. As explained earlier, researchers report different numbers for the incidence of partner assault, ranging from 8 million to 626 thou al., may 1 Diffi pool: sourc data marril work i Earloa Violer Obtain bY the the da Conjuc aSsaul mortal the De report random State 1 1991; Straus find in 7 thousand annually (Harlow, 1991; Sherman, 1992; Straus et al., 1980; Weis, 1989). A reason for this great disparity may be the various methods utilized to conduct the research. Different researchers obtained their samples from different pools (e.g., calls to police, surveys). The data for these various findings come from different sources. For instance, Straus et a1. (1980) collected the data from 2,143 couples living together with a partner or married on a national survey. Weis (1989) used the early work of Straus (1978) with survey method of households. Harlow (1991) obtained its rates from the amount of family violence that people considered to be criminal. Harlow obtained the data from National Crime Victimization Survey by the U.S. Department of Justice. Sherman (1992) obtained the data from a survey of 57 major city police departments conducted by the Crime Control Institute in 1990. Researchers also found disparate rates of partner assault when they utilized different data sources, such as mortality data on women from the FBI Crime Report, data from the Department of Justice Crime Victim's Report, injury reports from hospitals, applications for divorce, national random-digit-dialing surveys, or reports from the Michigan State Police (Bell 8 Chance-Hill, 1991; Emery, 1989; Harlow, 1991; Michigan Domestic Violence and Treatment Board, 1993; Straus et al., 1980). Weis (1989) explained that the discrepancy in numbers and in research methods made it difficult to determine whe‘ dec1 539? just inc: a de 8 whether the rate of partner assault was increasing or decreasing. At the national level, victimization surveys suggested little variation by year (Weis, 1989). Criminal justice, health, and social service statistics, suggested an increase in partner assault, but self-report data suggested a decrease (Weis, 1989). In summary, the problem of partner assault has a high magnitude. Although the reports did not show consistency in the source of information where the rates were coming from, the rates were alarmingly high. These high numbers are consistent with the issue of considering partner assault a major social problem. Criminal justice organizations who often face the call for domestic assault could be one of the most responsible maintaining statistics of this problem. However, the criminal justice system have not worked consistently in this issue, blaming victims for the inconsistencies of the system. Also the criminal justice system have used the issue of domestic violence to punish ethnic minorities men (Sherman, 1992). Some studies have examined the effectiveness of police actions (Dunford, 1992; Pate & Hamilton, 1992; Sherman & Berk, 1984; Sherman et al., 1991; Sherman, Smith, Schmidt, & Rogan, 1992). The following section examines the criminal justice responses to the problem of partner assault. begin durin was a; examii respor litera crinin descrii researc Policie Changes is the The fir Partner Sharman they Se inciden aSsault: aSsaila] invova‘ This Ex; and bat. 5‘ ““9. TM‘ ‘ 9 Criminal Justice Responses to Partner Assault This section continues the analogy presented at the beginning of the current chapter. The analogy says that during a partner assault police were called, the assailant was arrested and brought to court, ... The current section examines the historical development of criminal justice responses to the issue of partner assault. Doing this, the literature of police performance and policy development in criminal justice organizations is examined. After the description of the police performance, the current researcher criticizes this performance. The section on policies of criminal justice organizations examines the changes over time in criminal justice organizations. W A series of experiments were conducted to examine what is the most effective manner to confront partner assailants. The first experiment involving police intervention with partner assailants was conducted in Minneapolis in 1984 by Sherman and Berk. In order to expedite the data collection, they selected the two Minneapolis precincts with the highest incidence of reports about, and arrests for, partner assaults. These precincts had a disproportionate number of assailants who were black and unemployed, and who had been involved in prior violent incidents where police intervened. This experiment included 330 cases of misdemeanor assault and battery between partners where the assailant and victim were still present. The cases were randomly assigned to three sepa: that assau optio assau nedia1 10 three different interventions: arrest, temporary separation, and mediation. Sherman and Berk (1984) reported that for men who were arrested, the rate of repeating an assault was 19 percent. For men who were assigned the option of temporary separation, the rate of repeating the assault was 28% and for those who were assigned to police mediation, the rate was 37 percent. A caution about interpreting these findings is that Sherman and Berk (1984) included only 51% of the overall sample of the couples who originally were assigned to the experimental conditions. The researchers lost 49% of the original sample at the time of follow-up. The percentage of individuals repeating their assaults after the arrest option were the lowest of the three options. However, the differences between the options were not statistically significant. Furthermore, the internal validity of this study is in doubt because police in the field assessed participation eligibility and decided which action to take in some cases violating random assignment principles (Sherman et al., 1991). After this study, other police departments began similar experiments with the purpose of testing independently the deterrent effects of arrest in other communities (Garner, Fagan, & Maxwell, 1995). These experiments, known as the Spouse Assault Replication Program (SARP), used the following criteria: cases had to be eligible for arrest for misdemeanor spouse assault, and alter after 1 replic experi proble first i experii subject the res interve detenti three f the num: °Peratic Thai-EEO, for long 11 alternative interventions were assigned by randomization after the eligibility was determined. The study by Sherman et al. (1991) was one of the replications conducted in the city of Milwaukee. In this experiment, the researchers wanted to correct for the problem of threat to internal validity which occurred in the first experiment. The police officers in this second experiment did not know the experimental conditions for the subjects they were working with until they got the code from the researchers. The conditions randomly assigned for intervention in partner assaults were long detention, short detention, and warning. Then Sherman et al. (1991) used three forms of measuring recidivism. The first measure was the number of arrest reports regarding the same or other victims. The second measure was the number of offense reports filed by the same victim only about offenses by the same assailant. The third measure was the number of hotline records generated by police and recorded by volunteers in the Sojourner Truth House. The hotline record included offender-absent and offender-present data, and the victim's report of any domestic violence by the same suspect, both before and after the incident selected for research. The time for long and short detention was not operationally defined previous to the experiment. Therefore, the authors found that the mean number of hours for long-term detainees for whom time was collected (n=91) was 4.5 hours. However, the authors estimated the detention time a time c the av receive Shermar tern de months assaulé term, a did not with uh detenti GVerage Sherua: CustOd). Pa repliCa conduct in this Or Den. 12 time as 11.1 hours for (n=404) long-term detainees. The time of detention was obtained for short-term detainees, and the average holding period was 2.8 hours. Individuals who received warnings were not arrested for any period of time. Sherman et a1. (1991) found that both long-term and short- term detention reduced the risk of future arrests during 33 months of surveillance. The prevalence of repeat partner assault was 1.7% for long-term detention, 2.2% for short- term, and 7% for the warning. Unfortunately, these results did not differentiate significantly any of the three groups with whom the police intervened. Furthermore, time of detention was not clearly documented in the experiment. The average time for short-term detention was 2.8 hours. Sherman et al. (1991) could not keep track of the time in custody for 77% of the assailants under long-term detention. Pate's and Hamilton's (1992) study was another of the replications for police intervention. These researchers conducted their study in Dade County, Florida. Assailants in this experiment were assigned two interventions: arrest or non-arrest. Florida statute requires that couples had to be currently married or previously married for police to arrest the assailant. Thus eligible cases had this requirement. Arrested individuals repeated their assaults in 9% of the cases, and those not arrested repeated their assaults in 10.6% of the cases. Pate and Hamilton (1992) did not find statistically significant differences between individuals who were arrested for partner assault and those who fro: appe of a vict 13 who were not. The researchers obtained the follow-up data from "the Domestic Violence Continuation Report form appended to all offense reports involving domestic violence of any type" (p. 694) if assailants assaulted the same victim. Pate and Hamilton (1992) found, however, that arrest for partner assault had a statistically significant deterrent effect upon employed individuals, but increased subsequent assaults among the unemployed. The strength of the relationships between these two variables was not described in the study, nor did the researchers evaluate these data using a stepwise discriminant analysis of the Wilks' lambda to determine if these two groups discriminated between the employed and the unemployed. Dunford's (1992) study was also part of the replication program for Sherman’s and Berk's (1984) work. Dunford conducted the study in the city of Omaha. He randomly assigned cases involving calls to the police for partner assault to three interventions: arrest, separation, or police mediation. The effects of this experiment were studied 6 months and then 12 months after the interventions occurred. Dunford found that during the first 6 months, arrest did not appear to be more effective than mediation or separation in decreasing additional partner assaults. Ten percent of assailants assigned to the arrest option repeated an assault. Dunford did not mention the amount of time these assailants were kept in a detention center or jail. 14 Neither did he report the percentages of assailants assigned to mediation and separation who recidivated during the period of the study. Thus, arrest appears to have had the same impact as separation or police mediation. The results of Dunford's follow-up after 12 months did not demonstrate a change in these findings. Dunford concluded that the three interventions in the experiment did not differ from each other in decreasing further partner assaults. Dunford found, however, that recidivism was higher during the first six months after police intervention than the second six months after the intervention. Dunford (1992) used two types of measures of recidivism at six and twelve months. The first measure was "new arrests and complaints for any crimes committed by perpetrators against victims as found in official police records. The second was victims reports of three forms of repeated violence" (p. 122). The three forms of violence that victims reported included: (a) "victim fear of injury,” (b) "pushing-hitting," and (c) "physical injury" (p. 122). Dunford interviewed victims during the first week after the incident in research, then six and twelve months after the initial incident. ’c e a e The overall criticism of these experiments is that researchers selected cases that were easy for police, and excluded many cases that police routinely encounter. For instance, assailants who committed severe assaults were not inclu cc: : (Dunf the v. assail with t valid; the le the ef: the e23 scene a EXperi: Vhich a SEBetifii {056cm “Refine praCtica EXCluded CaSeload Ho: assaila: were not neigthr fam'IEr dEtecteéJ iii-tame: in the S 15 included in these experiments because of the ethical constraints necessitated by the seriousness of the assault (Dunford, 1992; Sherman, 1992; Sherman & Berk, 1984). If the victims' lives were in danger, the police arrested assailants. However, the effects of the police intervention with hard cases are unknown. This affects only the external validity of the study. More importantly, the actions that the legal system takes against these types of assailants and the effects on female survivors are also unknown. Most of the experiments excluded cases where assailants had left the scene and police could not assign them to any of the experimental conditions. Police routinely handle cases in which assailants leave the scene. Furthermore, police sometimes arrest these escapees in other locations, or the prosecutor’s office sends them letters of summons. The experimental police interventions represent only one practical component of the criminal justice system. Cases excluded from the experiment sometimes become part of the caseload of other criminal justice organizations. Most experimental police interventions with partner assailants have excluded affluent districts where police were not called. Sherman (1992) explained that in affluent neighborhood where walls are thicker and the houses are farther apart, assaults against women may not be easily detected. Furthermore, the chances of police and neighbors intervening in these cases are very slight. The experiments in the studies were conducted in districts that have the greate distri minori‘ 1991; ] distric race,“ with gh (1991) " evaluatl As vict the S324 (Reiss, tudy tx Sleeq'uE Ans experibe individu conditio Fl°rida. 16 greatest number of domestic violence reports. These districts, however, had a heavy concentration of poor, minority, and "racially mixed" couples (Sherman et al., 1991; p. 826). Sherman et al. (1991) described these districts as heavily "segregated in terms of class and race," and reported that "each district included vast tracts with ghetto poverty characteristics according to Wilson (1991)" (p. 826). Sherman's and Berk's (1984) study evaluated only repetition of assaults to the same victims. As victims may change constantly, relying on reports from the same victim could constitute a limitation of the study (Reiss, 1985; Sherman et al., 1991). With this type of study both groups are underestimated in the number of subsequent assaults. Another limitation of excluding cases is that each experiment made its selection from different groups of individuals. In Dade County, only married or formerly married couples were assigned to the experimental conditions. This was necessary because of the laws in Florida. In other locations, the researchers included more cohabitating than married couples (Garner et al., 1995). This difference in examining individuals with the same procedure expands on the criticism that Garner et a1. (1995) indicated as selection of site by convenience. A study with a rigorous measure plan would have selected sites with similar state laws. al. I polic House often domes the li concez telept study volunt 17 One measure of recidivism in the study of Sherman et al. (1991) was the number of hotline records generated by police and recorded by volunteers in the Sojourner Truth House. The hotline record included offender-absent and offender-present data, and the victim's report of any domestic violence by the same suspect, both before and after the instant incident. The police provided these data concerning arrest and experimental interventions over the telephone to volunteers in the Sojourner Truth House. The study did not include an analysis of the reliability of the volunteers obtaining and entering the data. No researcher has adequately explained the reason that the percentage of men who repeat their assaults after a contact with police is so low (Rosenfeld, 1992). Rosenfeld used the findings from Spouse Assault Replication Program's (SARP) studies to argue that the number of repeat offenses decreases independently of criminal justice intervention. According to Rosenfeld (1992), the vast majority of men who assault a female partner and receive a police intervention do not repeat the assault. However, it is argued that these assailants are not identified as repeaters. In the reported studies, repeat assaults were identified in 1.7 to 37 percent of the.cases in which there was an intervention (Dunford, 1992; Pate & Hamilton, 1992; Sherman & Berk, 1984; Sherman et al., 1991). Although there are extreme differences in findings, none indicated that over 50% of the assail; experil A: extehsj that t? not prc deterre publish site re cross-s et a1. rfiplica 0f data did not in the 1 Stated . aimp: 18 assailants repeated the assault after any of the experimental police interventions. Additionally, Garner et al. (1995) conducted an extensive review of the SARP's experiments. They stated that these studies replicating the Minneapolis findings did not provide enough evidence to conclude that arrest has no deterrent effect. One of the reasons for this is that the published results were ”a series of inconsistent individual- site reports and a few incomplete and highly selective cross-site comparisons" (Garner et al., 1995, p. 8). Garner et al. (1995) also argued that these experimental replications lacked similar methodology, comparable analysis of data, and a standard report of the findings. The studies did not provide a replica of the measures and analysis used in the Minneapolis experiment. Garner et al. (1995) further stated that ”there was no a priori consensus about the most appropriate of these possible analytical comparisons" (p. 9). Additionally, Garner et al. (1995) stated that all the SARP studies failed to report the power of their statistical comparisons. Garner et al. (1995) examined the statistical power of failing to reject the null hypothesis. Analyzing the statistical power of such comparisons allow researchers to determine whether a nonsignificant statistical difference was due to the absence of an effect or due to the research design. Garner et al. (1995) found that the statistical power of these studies using official records was higher 19 than 0.80. This level of statistical power means that these studies had a low probability of type II error. Type II error is the failure to reject the null hypothesis when it is false. Thus, it was very unlikely that the studies were not significant due to lack of power. Furthermore, the experiments with police intervention did not examine legal consequences for partner assailants beyond the arrest. It could be assumed that when police arrest an assailant, this assailant will receive the traditional criminal justice remedies of incarceration, fines, or probation (Rosenfeld, 1992; Straus, 1993). However, this is not always the case; as in the experiments with police interventions, assailants could receive police advice, mediation, warnings, or be separated from partners (Dunford, 1992; Pate & Hamilton, 1992; Sherman & Berk, 1984; Sherman et al., 1991). Moreover, the court could dismiss the charges of assault against the assailants without providing a legal sanction. Thus, not all partner assailants are subject to legal consequences after committing an assault. E 1' i s E : . . J 1 !° ; . !' n The history of the response of the criminal justice system to partner assaults reflects inconsistency in policies and practices, which have dramatic changes over time and have varied according to geographical location. At one time wife beating was approved by court rulings like that of Judge Buller in England in 1783 and those in 20 Mississippi and North Carolina in the nineteenth century. Judge Buller gave men permission to beat their wives, and in Mississippi and North Carolina wife beating was a non- punishable act (Sherman, 1992). However, in the period from 1880 to 1906, the states of Maryland, Delaware and Oregon passed legislation establishing whipping posts to punish men who battered their wives. Sherman (1992) disclosed, however, that the beating posts were often used as an excuse to beat black men. New York was another state with strong policies against wife beating. In 1844 the police force of New York City treated partner beating as a social disorder. Yet Sherman (1992) revealed that in the 18808 this policy was being discouraged by the office of the district attorney. Lawyers, claiming that women often withdrew charges, declined to prosecute batterers. Moreover, the prosecutors' policy discouraged police in New York City from arresting assailants (Sherman, 1992). More recently, an international organization and a national organization passed resolutions about police response to cases of partner assault. In 1967 the International Association of Chiefs of Police declared that arrest in cases of domestic violence should be used as the last recourse. Moreover, in 1973 the American Bar Association supported the International Association, which encouraged police to serve as mediators in partner assault situations rather than arrest the assailants (Sherman, 1992). P3 He co: vi: 7",. . .‘AH RCf‘i 21 Sherman (1992) stated that as early as 1880 prosecuting attorneys refused to prosecute wife beating. This led indirectly to the failure of police to take action against these assailants. Most of the incidents of partner abuse fall into the category of simple assault, such as slapping, pushing, shoving, or throwing an object at a victim (Feld & Straus, 1989). However, Straus and Gelles (1988) stated that a third of the incidents were severe and involved punching, kicking, and attacks with objects or weapons. Nevertheless, the criminal justice system is likely to consider violence between intimates as insignificant, unlike violence between strangers (Sherman, 1992; Weis, 1989). Thus, the lack of police intervention is part of the history of this problem, as it was widely believed that nonintervention was appropriate. As a result of these policy changes, police departments and cities have faced various law suits. For example, a nationally publicized law suit was brought against the city of Torrington, Connecticut (Edleson, 1991; Sherman, 1992). This law suit was brought by a coalition of the family members of Tracey Thurman, other battered women, and victims' rights groups in Thurman_yL_gity_gf_12;zingtgn (1984). According to Sherman's (1992) description of the incident, a police officer watched Charles Thurman, who had a knife covered with blood, kick Tracey in the head. She was on the ground and wounded in the chest, neck, and throat. The incident involved four police officers who did 22 not arrest the assailant until he approached the victim to kick her again. Two similar law suits were ugggan_yL W (1982) and W York (1982). The courts in both cases ruled that the police should have arrested the potential assailant when he made a threat against the life of the victim, and that the police could have prevented the death of the victim (Sherman, 1992). Advocates for women survivors of partner assault have encouraged police to make arrests. Their argument was based on a 1976 report from the Police Foundation (cited in Sherman & Berk, 1984). This report revealed that 85% of a sample of partner homicides had been subject to a previous intervention by the police within the two years previous to the fatal assault. The same report disclosed that the number of previous police interventions for 54% of these cases was five or more. In a retrospective study, Gondolf and Fisher (1988) found that only 15% of abused women's assailants were arrested. Approximately one-third of the women interviewed indicated that the police intervention in their cases was mediation or referral. Battered women for whom the police did nothing accounted for 20% of the most severe police cases. These women suffered severe abuse, and in 53% of these cases, the assailants were likely to use a weapon and to threaten the victim. These assailants, however, were generally less violent and had a history of less crime and 23 less alcohol abuse than other offenders. These criteria were the basis for the decision that they should not be arrested. Batterers with previous arrests who were under the influence of alcohol before the assault, who used verbal abuse, and who also caused more physical harm, were subject to more action from the police. These individuals were aggressive against the police as well, and they were more likely to be arrested under any circumstances. They were also men who were less likely to benefit from arrest than were other men (Gondolf 8 Fisher, 1988). Mini, on 'u'1., st'ce ;- '015“ 9 '. ;- l:-= -t The section of criminal justice response to partner assault presented a historical development of police performance and policies in these organizations. Both revealed their controversies and inconsistencies. The literature on police arrest of partner assailants reflected that arrest does not decrease the number of repeat assaults more significantly than other police interventions (Dunford, 1992; Garner et al., 1995; Pate 8 Hamilton, 1992; Sherman 8 Berk, 1984; Sherman et al., 1991). Although it did not demonstrate significant difference, advocates of the arrest of assailants claimed that when police arrested the perpetrators, they were less likely to repeat assaults than when they were when the police only gave advice or ordered the assailants to leave the location (Garner et al., 1995; Gondolf 8 Fisher, 1988; Langan 8 Innes, 1986; Sherman 8 Berk, 1984; Sherman et al., 1991). cc 1122 vic Pie: P€::_ asset Stan: $14556 than I 5;“ 6 .5. p 24 Sherman (1992) and Sherman et al. (1991) found that in certain cities where assailants were more likely to be black, unmarried and unemployed, arrest did not appear to significantly reduced the number of assaults (Sherman et al., 1992). Sherman (1992) reported that unemployed, unmarried assailants were identified while increasing their violence after the arrest, but unemployed, married assailants were not identified as increasers of assaults. Conversely, the number of assaults among employed men was identified as increasing when they were not arrested (Pate 8 Hamilton, 1992). Sherman (1992) and Sherman et al. (1991) also found that batterers with low socioeconomic status, those who were unemployed, unmarried, and African-American, were identified continuing their assaults on their partners after an arrest. One explanation that Sherman et al. (1991) offered for this phenomenon was the effect of continuous police harassment of people of color and poor individuals. The arrest for assault was considered to be further harassment, not a standard police practice. The findings of the police arrest were criticized by Garner et al. (1995) because they were based on a small subset of data, and the data reported was less consistent than Sherman (1992), and Sherman et al., (1992) suggested. Furthermore this theoretical explanation was not based on the precise methods used in experimental studies. The explé evidi orgai with Each depa: effe: raze: inst: law 5 Gave} perfo state. itua do no There V» ~u€ l is Fa AS833 25 explanation of the findings was not supported by sufficient evidence. The section of policies in criminal justice organizations reflected that the inconsistencies in dealing with partner assailants has continued into the present. Each police department, prosecutor's office, probation department, and court has its own rules for dealing with partner assailants. Unless authorities use consistent and effective measures against partner assailants, battered women, victims' rights groups and other advocates will force institutions to take actions against the batterers through law suits. After unfortunate incidents, many states have developed policies to deal with partner battering. Gondolf and Fisher (1988) connected the police performance and policies in criminal justice systems. They stated that police want to have control of the immediate situation that they confront upon their arrival, and police do not consistently respond to partner assault as a crime. Therefore, police use personal discretion in implementing the law (Gondolf 8 Fisher, 1988). This personal discretion is part of a personal policy to address cases of partner assault. Finally, the literature on police arrests does not contain information about the most effective punishments for these assailants. Incarceration as a punishment, rather than as an initial detention, is mentioned in this literature without indicating the actual number of days that who 6 nc.. "id assa assa 26 individuals spent in jail. Characteristics of individuals who went to jail are not distinguished from those who did not. Probation is vaguely defined. Moreover, the preferred and most effective conditions of probation for partner assailants is not widely known. The characteristics of assailants who benefit most from each particular probation condition are also unknown. The criminal justice response were deficient examining other legal consequences for partner assailants arrested. Due to the lack of additional information of partner assailants in the courts, the next section skip the court intervention to focus on treatment programs. The reason to focus on treatment programs is because it is assumed that assailants are sent from courts to treatment programs. Treatment Programs This section of treatment programs continues the analogy presented at the beginning of the current chapter. The analogy says that during a partner assault police were called, ..., and assailants were mandated to treatment. This section reviews studies of treatment programs. They have been divided into two sub-components, first the general characteristics of assailants in group treatment programs in the community, and second the characteristics of assailants who repeated an assault to partners. These studies were selected because they described in numeric forms the characteristics of partners assailants. Furthermore, the w”: GU“ 27 studies that examined repetition of assault were selected for its connection with the current research. After the studies are described, the current study provides a criticism to these findings. Wigs Fitch and Papantonio (1983) reported the demographic and clinical characteristics of 188 men who had physically abused their partners. Fifty-one percent of these assailants sought treatment at their partners’ insistence, 18% volunteered, and 31% were referred by civil or criminal courts. The authors found that 59% of these men reported abusing alcohol. In addition, 18% of the assailants abused drugs to the point of impairing their daily functioning. About 22% were considered to be unemployed because they did not hold jobs for at least 20 hours per week on a regular basis (Fitch 8 Papantonio 1983). The study of Fitch and Papantonio (1983) found that a high percentage of individuals, about 59% of the sample, were addicted to alcohol. It appears that, in accepting this large group of individuals with alcohol problems, the clinicians did not follow the rules requiring rejection of the "different" client. The rules for clinicians working with partner assailants has been to do not accept individuals with alcohol and/or drug problems, and do not accept mentally ill individuals. The authors did not report the ethnicity of these assailants. Therefore it is unknown what! of d U ‘. ‘4 1‘ ‘I‘F Ao.‘ refs 2:.» “‘ 28 whether or not these findings are different for individuals of different ethnic groups. Bernard and Bernard (1984) studied a group of men who voluntarily sought to stop their abusive behavior toward their partners. The researchers obtained data from 46 men who returned to the treatment program after the orientation week. The mean age of this group was 30.8, and the mean educational level was 13.3 years of school. The mean income was $20,000 a year, and the mean number of marriages was one and a half. Bernard and Bernard (1984) did not provide information on the percentage of individuals who were referred by the court or came voluntarily to the program, nor information about other characteristics of these assailants. Edleson and Syers (1990) reported the demographics characteristics of 283 men randomly assigned to one of six possible treatment conditions. The mean age was 31.8 years, the majority of the men (73.7%) where white, 10.6% were black, 3.8% Native American, 2.7% Hispanic, and 0.4% Asian American. The mean number of education completed was 12.7 and half of the men 50.2% were employed full time. The report of unemployment was 33.5 percent. About 34.5% were married, 24.5% separated, and 20.1% single and had never been married. Ordered by the court encompassed 38.3% and 61.7% came under some other form of social pressure. Over 50% had received some chemical dependency treatment and 50.2% reported receiving mental health treatment. and j the , ‘16:: 29 Hamberger and Hastings (1990) also presented some findings regarding men in a treatment program. The researchers found that 86% of these assailants were Caucasian, 12% were black, and 2% were from other ethnic groups. About 80% were employed, while 20% were unemployed. Their average age was 31.1 years. Eighty percent of the assailants had high school degrees or more advanced education. About 70% attended treatment on a volunteer basis, while 30% were assigned from the court. According to the researchers, about 40% used alcohol. Saunders (1992) examined 182 men during their assessment for admission to a treatment program for men who had assaulted a spouse or partner and had already attended a treatment program. Saunders found that approximately 70% of these men were court-ordered to attend. Most of the others were referred by county organizations. The average age of this sample was 30.6 years (Saunders, 1992). Almost 60% had not been educated beyond high school. About 76% were white, and 18.1% were African-American. Saunders did not describe the ethnicity of the remaining six percent. It is unknown whether there were significant differences between these demographic groups. In summary, most of studies of the characteristics of partner assailants have based their research on men who participated in treatment programs. The findings of these studies reflect a variety in sample size, from 46 to 283 male assailants. Most of the studies found that assailants 'o'EI pro 58‘. 30 were about 30 years old. Unemployment was a factor in 34% of the samples. Most of them were caucasian 74 to 97 percent. Alcohol and psychoactive drug use was reported in 18% to 59% of the samples. Not all studies reported the same variables in a standard format. The next section presents a critique of the findings reported by studies on assailants. The studies of assailant characteristics in treatment programs present that most of the individuals in these settings have education, employment, and of European American descent. These characteristics were very different from the studies of police arrests. Studies of characteristics of assailants at treatment programs cannot be considered as a part of the continuum that follows the police interventions. Reports of police intervention indicated that the communities where most domestic assaults were reported were poor and racially mixed, and that identified assailants were mostly African American, and unemployed males. Most of these individuals were selected for police interventions, including arrests and alternatives to arrest (Sherman 8 Berk, 1984; Sherman, 1992). The treatment programs in the same city where the police arrests studies were conducted, however, reported that the majority of their clients were white, and with completed high school degrees (Edleson 8 Grusznski, 1988; Edleson 8 Syers, 1990, 1991). Therefore, after the police arres uh :; lives th : the ' assi It i sen .n a L's: H“. 36ft 5%. ‘- I1 l" 31 arrest and the assignment for treatment, uneducated, unemployed black males were not in treatment. Furthermore, men from other ethnic minority groups who lived in neighborhoods with low-income housing did not have the same ethnic representation in the police studies as in the treatment programs. The reasons for their not being assigned to treatment are not spelled out in the literature. It is unknown if the probation officers did not send these men to treatment, or if the treatment program rejected them. In any event, it seems that police studies were conducted in different cities than those where the treatment programs were conducted. ’3 ' s e a t e s ' t Another group of studies has examined partner assailants who completed treatment programs and repeated their assaults (DeMaris 8 Jackson, 1987; Edleson 8 Grusznski, 1988; Hamberger 8 Hastings, 1990; Harrell, 1991; Purdy 8 Nickle, 1981). This section will review these studies. These studies are reported in chronological order. Purdy and Nickle (1981) studied 170 male partner assailants who came to a treatment program during a two-year period. About 3% of these men were mandated by the courts for treatment, while 97% accessed the program on a volunteer basis. The researchers contacted the assailants and the victims separately six months after their termination of the program. Purdy and Nickle (1981) found that 75% of these men remained with the same partner. Among these couples, data they Of 35 rate :;_ Sui 'I‘as 1' prESq 32 59% of the women had not experienced physical or sexual violence. Although the researchers indicated the percentage of couples who did not experience emotional violence (14%), it was not clear whether this low level of emotional violence was within the group that did not experience any form of violence or within the group that experienced some form of physical violence. DeMaris and Jackson (1987) assessed the number of assaults after assailants had completed the program. The data on assault was obtained from 53 men, one year after they had finished a six-month treatment program in the city of Baltimore, Maryland. The overall recorded recidivism rate of the sample was 35 percent. The researchers did not find any significant differences between individuals who reported repeating their assaults and those who did not in relation to the following variables: court-mandated versus volunteer treatment, living arrangements (living with partner versus separated), current involvement in a relationship, current involvement with the same partner who was abused, drug use (whether in the past or in the present), type of prior offense (violent versus nonviolent), or alcohol problems. In a "T-test" of the mean differences in reduction of violence from before to after treatment, men who attended treatment voluntarily had a significant smaller level of violence after treatment in comparison with court- ordered men. questi counse treat: use 01 “id 4 33 DeMaris and Jackson (1987) sent in the mail a questionnaire to all men who had attended at least one counseling session and were not currently attending treatment group. "Recidivism was defined as the return to use of violence of any kind with a female partner after counseling" (DeMaris 8 Jackson, 1987, p. 460). Edleson and Grusznski (1988) reported three studies of follow-ups for male assailants who attended a treatment program. The first study by Edleson and Grusznski (1988) surveyed 63 men who responded to a follow-up interview. Of these respondents, 32 had completed the treatment program. The survey also included responses from the victims of partner assault, 27 of whom were also interviewed. The group of non-completers included 31 men, and 30 of their partners were interviewed. The summary of findings of the first study indicates that the average age of the individuals interviewed was 29.3 years. About 97% were white, and only 9% were unemployed at the time of the intake. Additionally, 76% were married, 20% were separated, and 4% were single. About 25% reported previous treatment for chemical dependency. In comparing the group who completed the program with the group who did not, Edleson and Grusznski (1988) did not find significant differences in age, race, marital status, occupation, or religion. Furthermore, they did not find differences between the groups in their history of contact with 1 33:5 i that : ICIE i 34 with the judicial system or in the number of their previous acts of violence. Edleson and Grusznski (1988) found in their first study that individuals who completed the program had significantly more education than those who did not. Additionally, the group who completed the program reported less violence than Ithe group who did not. Edleson and Grusznski (1988) interpreted the Chi Square test as indicating that completers were "more often nonviolent at follow-up when compared to non-completers" (p. 10). In their second study, Edleson and Grusznski (1988) assessed 86 male assailants who completed the treatment program in Minneapolis. The researchers interviewed 42 female partners of these men. The number of women interviewed was low because researchers could not found all victims nine months after the men had completed the program. Additionally, in this second study, Edleson and Grusznski (1988) were unable to find individuals who did not complete the treatment program to establish a comparison between the two groups. The average age of the individuals in the program was 32 years. The majority of the assailants were white (86%), only 6% were black, 2% were Native American and 1% Asian. Approximately 5% were racially mixed and were not identified in the study as belonging to any ethnic group. About 76.7% were employed full-time, and 16.3% were unemployed. About 84.3% stated that they had received prior mental health treatment and just over 34.5% reported that they ha sources service court in 35 they had received treatment for chemical dependency. The sources of referrals for these assailants were mostly social service agencies (30%), followed by partners (24.4%), and court mandates (10.7%). In this second study, Edleson and Grusznski (1988) found that both males and females reported that 24% of the assailants were nonthreatening and nonviolent nine months after completion of the treatment. About 43% were reported to have used threats, but not physical violence. The researchers combined these groups and stated that 68% of the sample were not physically violent. About 33% of the partners of these assailants reported suffering at least one more act of physical violence after the treatment. One of the main comparisons in this study was between the assaults reported by victims and the assaults reported by assailants. The comparison between the reports of assailants and survivors did not show a statistically significant difference between these two sources of information. The third study of Edleson and Grusznski (1988) examined 112 men who completed the program and compared them to 47 men who did not. The researchers used telephone interviews to contact 84 female partners of program completers and'37 partners of non-completers between 6 and 7 months after the assailants had completed treatment. Edleson and Grusznski (1988) found that the average age for these assailants was 34 years. About 88% were white, 2.5% Hispanic, 2.5% black, 2.5 Native American, and 4.1% of 6'8: “5‘ 7*.- -\ HA 1..., . ‘v- 4., “I (I) 36 mixed race. The average education level for this group was almost two years of college. Approximately 90% had received mental health counseling and 42% had received chemical dependency treatment. The majority (29%) were referred by their partners, 26% were referred by a community agency, and 7.4% were referred by the court. Edleson's and Grusznski's (1988) findings for this third study compared the men's violent events by using the victims’ reports. They found that the categorical difference between completers and non-completers was not statistically significant in regard to these reports of violence. Fifty-nine percent of the men who completed the program were nonviolent, and men non-completers were nonviolent in 52% of the cases. Therefore, the difference in recurrence of assaults on females for males who completed a program versus those who did not was not statistically significant. The recurrence rate of violence for men who completed the program was 41 percent, versus 49% for assailants who did not complete the program. Hamberger and Hastings (1990) examined 106 partner assailants one year after they all had completed a cognitive-behavioral intervention. The researchers found that about 30%‘of the assailants had repeated at least one assault on a partner. The demographic characteristics of individuals who repeated their assaults were 78.1% Caucasian, 18.8% black, and 3.1% other. Among non- repeaters, 89.1% were Caucasian, 9.5% black, and 1.4% other. ”“P R‘- .bfll ’A "PP-\‘ “ ‘.41.} ex; \‘4 1... (7 16 greatest number of domestic violence reports. These districts, however, had a heavy concentration of poor, minority, and "racially mixed" couples (Sherman et al., 1991; p. 826). Sherman et al. (1991) described these districts as heavily "segregated in terms of class and race," and reported that "each district included vast tracts with ghetto poverty characteristics according to Wilson (1991)” (p. 826). Sherman's and Berk's (1984) study evaluated only repetition of assaults to the same victims. As victims may change constantly, relying on reports from the same victim could constitute a limitation of the study (Reiss, 1985; Sherman et al., 1991). With this type of study both groups are underestimated in the number of subsequent assaults. Another limitation of excluding cases is that each experiment made its selection from different groups of individuals. In Dade County, only married or formerly married couples were assigned to the experimental conditions. This was necessary because of the laws in Florida. In other locations, the researchers included more cohabitating than married couples (Garner et al., 1995). This difference in examining individuals with the same procedure expands on the criticism that Garner et al. (1995) indicated as selection of site by convenience. A study with a rigorous measure plan would have selected sites with similar state laws. al. 0 poliCi House offen< doses1 the i1 conce: telep? tudy volun‘ 17 One measure of recidivism in the study of Sherman et al. (1991) was the number of hotline records generated by police and recorded by volunteers in the Sojourner Truth House. The hotline record included offender-absent and offender-present data, and the victim's report of any domestic violence by the same suspect, both before and after the instant incident. The police provided these data concerning arrest and experimental interventions over the telephone to volunteers in the Sojourner Truth House. The study did not include an analysis of the reliability of the volunteers obtaining and entering the data. No researcher has adequately explained the reason that the percentage of men who repeat their assaults after a contact with police is so low (Rosenfeld, 1992). Rosenfeld used the findings from Spouse Assault Replication Program's (SARP) studies to argue that the number of repeat offenses decreases independently of criminal justice intervention. According to Rosenfeld (1992), the vast majority of men who assault a female partner and receive a police intervention do not repeat the assault. However, it is argued that these assailants are not identified as repeaters. In the reported studies, repeat assaults were identified in 1.7 to 37 percent of the.cases in which there was an intervention (Dunford, 1992; Pate 8 Hamilton, 1992; Sherman 8 Berk, 1984; Sherman et al., 1991). Although there are extreme differences in findings, none indicated that over 50% of the assai] experi extens that t not p: deter: 18 assailants repeated the assault after any of the experimental police interventions. Additionally, Garner et al. (1995) conducted an extensive review of the SARP’s experiments. They stated that these studies replicating the Minneapolis findings did not provide enough evidence to conclude that arrest has no deterrent effect. One of the reasons for this is that the published results were "a series of inconsistent individual- site reports and a few incomplete and highly selective cross-site comparisons" (Garner et al., 1995, p. 8). Garner et al. (1995) also argued that these experimental replications lacked similar methodology, comparable analysis of data, and a standard report of the findings. The studies did not provide a replica of the measures and analysis used in the Minneapolis experiment. Garner et al. (1995) further stated that ”there was no a priori consensus about the most appropriate of these possible analytical comparisons" (p. 9). Additionally, Garner et al. (1995) stated that all the SARP studies failed to report the power of their statistical comparisons. Garner et al. (1995) examined the statistical power of failing to reject the null hypothesis. Analyzing the statistical power of such comparisons allow researchers to determine whether a nonsignificant statistical difference was due to the absence of an effect or due to the research design. Garner et al. (1995) found that the statistical power of these studies using official records was higher than I studi 19 than 0.80. This level of statistical power means that these studies had a low probability of type II error. Type II error is the failure to reject the null hypothesis when it is false. Thus, it was very unlikely that the studies were not significant due to lack of power. Furthermore, the experiments with police intervention did not examine legal consequences for partner assailants beyond the arrest. It could be assumed that when police arrest an assailant, this assailant will receive the traditional criminal justice remedies of incarceration, fines, or probation (Rosenfeld, 1992; Straus, 1993). However, this is not always the case; as in the experiments with police interventions, assailants could receive police advice, mediation, warnings, or be separated from partners (Dunford, 1992; Pate 8 Hamilton, 1992; Sherman 8 Berk, 1984; Sherman et al., 1991). Moreover, the court could dismiss the charges of assault against the assailants without providing a legal sanction. Thus, not all partner assailants are subject to legal consequences after committing an assault. 'c ' t' ' t The history of the response of the criminal justice system to partner assaults reflects inconsistency in policies and practices, which have dramatic changes over time and have varied according to geographical location. At one time wife beating was approved by court rulings like that of Judge Buller in England in 1783 and those in passed who be boweve to bea Pclici 593' Yo Yi 20 Mississippi and North Carolina in the nineteenth century. Judge Buller gave men permission to beat their wives, and in Mississippi and North Carolina wife beating was a non- punishable act (Sherman, 1992). However, in the period from 1880 to 1906, the states of Maryland, Delaware and Oregon passed legislation establishing whipping posts to punish men who battered their wives. Sherman (1992) disclosed, however, that the beating posts were often used as an excuse to beat black men. New York was another state with strong policies against wife beating. In 1844 the police force of New York City treated partner beating as a social disorder. Yet Sherman (1992) revealed that in the 18808 this policy was being discouraged by the office of the district attorney. Lawyers, claiming that women often withdrew charges, declined to prosecute batterers. Moreover, the prosecutors' policy discouraged police in New York City from arresting assailants (Sherman, 1992). More recently, an international organization and a national organization passed resolutions about police response to cases of partner assault. In 1967 the International Association of Chiefs of Police declared that arrest in cases of domestic violence should be used as the last recourse. Moreover, in 1973 the American Bar Association supported the International Association, which encouraged police to serve as mediators in partner assault situations rather than arrest the assailants (Sherman, 1992). (fl attorr indire these fall 1 pushir Straus that a puh:.i Nevert consid Violen we- . ands, 21 Sherman (1992) stated that as early as 1880 prosecuting attorneys refused to prosecute wife beating. This led indirectly to the failure of police to take action against these assailants. Most of the incidents of partner abuse fall into the category of simple assault, such as slapping, pushing, shoving, or throwing an object at a victim (Feld 8 Straus, 1989). However, Straus and Gelles (1988) stated that a third of the incidents were severe and involved punching, kicking, and attacks with objects or weapons. Nevertheless, the criminal justice system is likely to consider violence between intimates as insignificant, unlike violence between strangers (Sherman, 1992; Weis, 1989). Thus, the lack of police intervention is part of the history of this problem, as it was widely believed that nonintervention was appropriate. As a result of these policy changes, police departments and cities have faced various law suits. For example, a nationally publicized law suit was brought against the city of Torrington, Connecticut (Edleson, 1991; Sherman, 1992). This law suit was brought by a coalition of the family members of Tracey Thurman, other battered women, and victims’ rights groups in Thuzman_y,_gity_g;_lggzingtgn (1984). According to Sherman’s (1992) description of the incident, a police officer watched Charles Thurman, who had a knife covered with blood, kick Tracey in the head. She was on the ground and wounded in the chest, neck, and throat. The incident involved four police officers who did 22 not arrest the assailant until he approached the victim to kick her again. Two similar law suits were ugzgan_y, Distrist_of_golumbia (1982) and Bensigners_xl_£itx_2f_8e1 XQIK (1982). The courts in both cases ruled that the police should have arrested the potential assailant when he made a threat against the life of the victim, and that the police could have prevented the death of the victim (Sherman, 1992). Advocates for women survivors of partner assault have encouraged police to make arrests. Their argument was based on a 1976 report from the Police Foundation (cited in Sherman 8 Berk, 1984). This report revealed that 85% of a sample of partner homicides had been subject to a previous intervention by the police within the two years previous to the fatal assault. The same report disclosed that the number of previous police interventions for 54% of these cases was five or more. In a retrospective study, Gondolf and Fisher (1988) found that only 15% of abused women’s assailants were arrested. Approximately one-third of the women interviewed indicated that the police intervention in their cases was mediation or referral. Battered women for whom the police did nothing accounted for 20% of the most severe police cases. These women suffered severe abuse, and in 53% of these cases, the assailants were likely to use a weapon and to threaten the victim. These assailants, however, were generally less violent and had a history of less crime and less a were ti arresti the in: abuse, to more aggres: likely also in were 0‘ 23 less alcohol abuse than other offenders. These criteria were the basis for the decision that they should not be arrested. Batterers with previous arrests who were under the influence of alcohol before the assault, who used verbal abuse, and who also caused more physical harm, were subject to more action from the police. These individuals were aggressive against the police as well, and they were more likely to be arrested under any circumstances. They were also men who were less likely to benefit from arrest than were other men (Gondolf 8 Fisher, 1988). a s 'ce 5 t er an The section of criminal justice response to partner assault presented a historical development of police performance and policies in these organizations. Both revealed their controversies and inconsistencies. The literature on police arrest of partner assailants reflected that arrest does not decrease the number of repeat assaults more significantly than other police interventions (Dunford, 1992; Garner et al., 1995; Pate 8 Hamilton, 1992; Sherman 8 Berk, 1984; Sherman et al., 1991). Although it did not demonstrate significant difference, advocates of the arrest of assailants claimed that when police arrested the perpetrators, they were less likely to repeat assaults than when they were when the police only gave advice or ordered the assailants to leave the location (Garner et al., 1995; Gondolf 8 Fisher, 1988; Langan 8 Innes, 1986; Sherman 8 Berk, 1984; Sherman et al., 1991). assail. Conver; identij Ealilti 24 Sherman (1992) and Sherman et al. (1991) found that in certain cities where assailants were more likely to be black, unmarried and unemployed, arrest did not appear to significantly reduced the number of assaults (Sherman et al., 1992). Sherman (1992) reported that unemployed, unmarried assailants were identified while increasing their violence after the arrest, but unemployed, married assailants were not identified as increasers of assaults. Conversely, the number of assaults among employed men was identified as increasing when they were not arrested (Pate 8 Hamilton, 1992). Sherman (1992) and Sherman et al. (1991) also found that batterers with low socioeconomic status, those who were unemployed, unmarried, and African-American, were identified continuing their assaults on their partners after an arrest. One explanation that Sherman et al. (1991) offered for this phenomenon was the effect of continuous police harassment of people of color and poor individuals. The arrest for assault was considered to be further harassment, not a standard police practice. The findings of the police arrest were criticized by Garner et al. (1995) because they were based on a small subset of data, and the data reported was less consistent than Sherman (1992), and Sherman et al., (1992) suggested. Furthermore this theoretical explanation was not based on the precise methods used in experimental studies. The explan eviden T crgani with p. Each pi depart] ' 'tnei effect: women, institi lav S‘;j de'telo; Gc 25 explanation of the findings was not supported by sufficient evidence. The section of policies in criminal justice organizations reflected that the inconsistencies in dealing with partner assailants has continued into the present. Each police department, prosecutor’s office, probation department, and court has its own rules for dealing with partner assailants. Unless authorities use consistent and effective measures against partner assailants, battered women, victims’ rights groups and other advocates will force institutions to take actions against the batterers through law suits. After unfortunate incidents, many states have developed policies to deal with partner battering. Gondolf and Fisher (1988) connected the police performance and policies in criminal justice systems. They stated that police want to have control of the immediate situation that they confront upon their arrival, and police do not consistently respond to partner assault as a crime. Therefore, police use personal discretion in implementing the law (Gondolf 8 Fisher, 1988). This personal discretion is part of a personal policy to address cases of partner assault. Finally, the literature on police arrests does not contain information about the most effective punishments for these assailants. Incarceration as a punishment, rather than as an initial detention, is mentioned in this literature without indicating the actual number of days that assai inter focus assai 26 individuals spent in jail. Characteristics of individuals who went to jail are not distinguished from those who did not. Probation is vaguely defined. Moreover, the preferred and most effective conditions of probation for partner assailants is not widely known. The characteristics of assailants who benefit most from each particular probation condition are also unknown. The criminal justice response were deficient examining other legal consequences for partner assailants arrested. Due to the lack of additional information of partner assailants in the courts, the next section skip the court intervention to focus on treatment programs. The reason to ‘focus on treatment programs is because it is assumed that assailants are sent from courts to treatment programs. Treatment Programs This section of treatment programs continues the analogy presented at the beginning of the current chapter. The analogy says that during a partner assault police were called, ..., and assailants were mandated to treatment. This section reviews studies of treatment programs. They have been divided into two sub-components, first the general characteristics of assailants in group treatment programs in the community, and second the characteristics of assailants who repeated an assault to partners. These studies were selected because they described in numeric forms the characteristics of partners assailants. Furthermore, the for it studie and cl abused assail 15% vo courts abasin 27 studies that examined repetition of assault were selected for its connection with the current research. After the studies are described, the current study provides a criticism to these findings. WW Fitch and Papantonio (1983) reported the demographic and clinical characteristics of 188 men who had physically abused their partners. Fifty-one percent of these assailants sought treatment at their partners’ insistence, 18% volunteered, and 31% were referred by civil or criminal courts. The authors found that 59% of these men reported abusing alcohol. In addition, 18% of the assailants abused drugs to the point of impairing their daily functioning. About 22% were considered to be unemployed because they did not hold jobs for at least 20 hours per week on a regular basis (Fitch 8 Papantonio 1983). The study of Fitch and Papantonio (1983) found that a high percentage of individuals, about 59% of the sample, were addicted to alcohol. It appears that, in accepting this large group of individuals with alcohol problems, the clinicians did not follow the rules requiring rejection of the "different" client. The rules for clinicians working with partner assailants has been to do not accept individuals with alcohol and/or drug problems, and do not accept mentally ill individuals. The authors did not report the ethnicity of these assailants. Therefore it is unknown rhethe: volunt. their 3 who re reek. educat was $2 and a ' infer: refer: nor in assail E charac Pissi‘: the ma black, Afier‘ 81c 28 whether or not these findings are different for individuals of different ethnic groups. Bernard and Bernard (1984) studied a group of men who voluntarily sought to stop their abusive behavior toward their partners. The researchers obtained data from 46 men who returned to the treatment program after the orientation week. The mean age of this group was 30.8, and the mean educational level was 13.3 years of school. The mean income was $20,000 a year, and the mean number of marriages was one and a half. Bernard and Bernard (1984) did not provide information on the percentage of individuals who were referred by the court or came voluntarily to the program, nor information about other characteristics of these assailants. Edleson and Syers (1990) reported the demographics characteristics of 283 men randomly assigned to one of six possible treatment conditions. The mean age was 31.8 years, the majority of the men (73.7%) where white, 10.6% were black, 3.8% Native American, 2.7% Hispanic, and 0.4% Asian American. The mean number of education completed was 12.7 and half of the men 50.2% were employed full time. The report of unemployment was 33.5 percent. About 34.5% were married, 24.5% separated, and 20.1% single and had never been married. Ordered by the court encompassed 38.3% and 61.7% came under some other form of social pressure. Over 50% had received some chemical dependency treatment and 50.2% reported receiving mental health treatment. Their . assail educat basis, the re this 5 m be and 18 the 9t -+ . I, e .ekne 29 Hamberger and Hastings (1990) also presented some findings regarding men in a treatment program. The researchers found that 86% of these assailants were Caucasian, 12% were black, and 2% were from other ethnic groups. About 80% were employed, while 20% were unemployed. Their average age was 31.1 years. Eighty percent of the assailants had high school degrees or more advanced education. About 70% attended treatment on a volunteer basis, while 30% were assigned from the court. According to the researchers, about 40% used alcohol. Saunders (1992) examined 182 men during their assessment for admission to a treatment program for men who had assaulted a spouse or partner and had already attended a treatment program. Saunders found that approximately 70% of these men were court-ordered to attend. Most of the others were referred by county organizations. The average age of this sample was 30.6 years (Saunders, 1992). Almost 60% had not been educated beyond high school. About 76% were white, and 18.1% were African-American. Saunders did not describe the ethnicity of the remaining six percent. It is unknown whether there were significant differences between these demographic groups. In summary, most of studies of the characteristics of partner assailants have based their research on men who participated in treatment programs. The findings of these studies reflect a variety in sample size, from 46 to 283 male assailants. Most of the studies found that assailants “4“. . t“; 2‘3”“ were a' of the percen 18% to same v presen assail 0—3 Prflqra settin iteric from t (n 30 were about 30 years old. Unemployment was a factor in 34% of the samples. Most of them were caucasian 74 to 97 percent. Alcohol and psychoactive drug use was reported in 18% to 59% of the samples. Not all studies reported the same variables in a standard format. The next section presents a critique of the findings reported by studies on assailants. t e t' The studies of assailant characteristics in treatment programs present that most of the individuals in these settings have education, employment, and of European American descent. These characteristics were very different from the studies of police arrests. Studies of characteristics of assailants at treatment programs cannot be considered as a part of the continuum that follows the police interventions. Reports of police intervention indicated that the communities where most domestic assaults were reported were poor and racially mixed, and that identified assailants were mostly African American, and unemployed males. Most of these individuals were selected for police interventions, including arrests and alternatives to arrest (Sherman 8 Berk, 1984; Sherman, 1992). The treatment programs in the same city where the police arrests studies were conducted, however, reported that the majority of their clients were white, and with completed high school degrees (Edleson 8 Grusznski, 1988; Edleson 8 Syers, 1990, 1991). Therefore, after the police arrest meeplc D lived : the sai the tr: assign. It is 1 58.“. t0 31 arrest and the assignment for treatment, uneducated, unemployed black males were not in treatment. Furthermore, men from other ethnic minority groups who lived in neighborhoods with low-income housing did not have the same ethnic representation in the police studies as in the treatment programs. The reasons for their not being assigned to treatment are not spelled out in the literature. It is unknown if the probation officers did not send these men to treatment, or if the treatment program rejected them. In any event, it seems that police studies were conducted in different cities than those where the treatment programs were conducted. Another group of studies has examined partner assailants who completed treatment programs and repeated their assaults (DeMaris 8 Jackson, 1987; Edleson 8 Grusznski, 1988; Hamberger 8 Hastings, 1990; Harrell, 1991; Purdy 8 Nickle, 1981). This section will review these studies. These studies are reported in chronological order. Purdy and Nickle (1981) studied 170 male partner assailants who came to a treatment program during a two-year period. About 3% of these men were mandated by the courts for treatment, while 97% accessed the program on a volunteer basis. The researchers contacted the assailants and the victims separately six months after their termination of the program. Purdy and Nickle (1981) found that 75% of these men remained with the same partner. Among these couples, ofc 32 59% of the women had not experienced physical or sexual violence. Although the researchers indicated the percentage of couples who did not experience emotional violence (14%), it was not clear whether this low level of emotional violence was within the group that did not experience any form of violence or within the group that experienced some form of physical violence. DeMaris and Jackson (1987) assessed the number of assaults after assailants had completed the program. The data on assault was obtained from 53 men, one year after they had finished a six-month treatment program in the city of Baltimore, Maryland. The overall recorded recidivism rate of the sample was 35 percent. The researchers did not find any significant differences between individuals who reported repeating their assaults and those who did not in relation to the following variables: court-mandated versus volunteer treatment, living arrangements (living with partner versus separated), current involvement in a relationship, current involvement with the same partner who was abused, drug use (whether in the past or in the present), type of prior offense (violent versus nonviolent), or alcohol problems. In a "T-test" of the mean differences in reduction of violence from before to after treatment, men who attended treatment voluntarily had a significant smaller level of violence after treatment in comparison with court- ordered men. D goesti counse treats use of COHRSE follow progra surrey these The so c)Cva'iia differ. 33 DeMaris and Jackson (1987) sent in the mail a questionnaire to all men who had attended at least one counseling session and were not currently attending treatment group. "Recidivism was defined as the return to use of violence of any kind with a female partner after counseling" (DeMaris 8 Jackson, 1987, p. 460). Edleson and Grusznski (1988) reported three studies of follow-ups for male assailants who attended a treatment program. The first study by Edleson and Grusznski (1988) surveyed 63 men who responded to a follow-up interview. Of these respondents, 32 had completed the treatment program. The survey also included responses from the victims of partner assault, 27 of whom were also interviewed. The group of non-completers included 31 men, and 30 of their partners were interviewed. The summary of findings of the first study indicates that the average age of the individuals interviewed was 29.3 years. About 97% were white, and only 9% were unemployed at the time of the intake. Additionally, 76% were married, 20% were separated, and 4% were single. About 25% reported previous treatment for chemical dependency. In comparing the group who completed the program with the group who did not, Edleson and Grusznski (1988) did not find significant differences in age, race, marital status, occupation, or religion. Furthermore, they did not find differences between the groups in their history of contact with t acts C that i more e group the gr 'mterp ccrple cor-pa: 34 with the judicial system or in the number of their previous acts of violence. Edleson and Grusznski (1988) found in their first study that individuals who completed the program had significantly more education than those who did not. Additionally, the group who completed the program reported less violence than Ithe group who did not. Edleson and Grusznski (1988) interpreted the Chi Square test as indicating that completers were "more often nonviolent at follow-up when compared to non-completers" (p. 10). In their second study, Edleson and Grusznski (1988) assessed 86 male assailants who completed the treatment program in Minneapolis. The researchers interviewed 42 female partners of these men. The number of women interviewed was low because researchers could not found all victims nine months after the men had completed the program. Additionally, in this second study, Edleson and Grusznski (1988) were unable to find individuals who did not complete the treatment program to establish a comparison between the two groups. The average age of the individuals in the program was 32 years. The majority of the assailants were white (86%), only 6% were black, 2% were Native American and 1% Asian. Approximately 5% were racially mixed and were not identified in the study as belonging to any ethnic group. About 76.7% were employed full-time, and 16.3% were unemployed. About 84.3% stated that they had received prior mental health treatment and just over 34.5% reported that they SOLIIC ‘i 621'): :0 4- 1‘6: Pfi‘ 32p6i f .___._..... 35 they had received treatment for chemical dependency. The sources of referrals for these assailants were mostly social service agencies (30%), followed by partners (24.4%), and court mandates (10.7%). In this second study, Edleson and Grusznski (1988) found that both males and females reported that 24% of the assailants were nonthreatening and nonviolent nine months after completion of the treatment. About 43% were reported to have used threats, but not physical violence. The researchers combined these groups and stated that 68% of the sample were not physically violent. About 33% of the partners of these assailants reported suffering at least one more act of physical violence after the treatment. One of the main comparisons in this study was between the assaults reported by victims and the assaults reported by assailants. The comparison between the reports of assailants and survivors did not show a statistically significant difference between these two sources of information. The third study of Edleson and Grusznski (1988) examined 112 men who completed the program and compared them to 47 men who did not. The researchers used telephone interviews to contact 84 female partners of program completers and 37 partners of non-completers between 6 and 7 months after the assailants had completed treatment. Edleson and Grusznski (1988) found that the average age for these assailants was 34 years. About 88% were white, 2.5% Hispanic, 2.5% black, 2.5 Native American, and 4.1% of ELXE urn aau'i'v 563: 36 mixed race. The average education level for this group was almost two years of college. Approximately 90% had received mental health counseling and 42% had received chemical dependency treatment. The majority (29%) were referred by their partners, 26% were referred by a community agency, and 7.4% were referred by the court. Edleson’s and Grusznski’s (1988) findings for this third study compared the men’s violent events by using the victims’ reports. They found that the categorical difference between completers and non-completers was not statistically significant in regard to these reports of violence. Fifty-nine percent of the men who completed the program were nonviolent, and men non-completers were nonviolent in 52% of the cases. Therefore, the difference in recurrence of assaults on females for males who completed a program versus those who did not was not statistically significant. The recurrence rate of violence for men who completed the program was 41 percent, versus 49% for assailants who did not complete the program. Hamberger and Hastings (1990) examined 106 partner assailants one year after they all had completed a cognitive-behavioral intervention. The researchers found that about 30%‘of the assailants had repeated at least one assault on a partner. The demographic characteristics of individuals who repeated their assaults were 78.1% Caucasian, 18.8% black, and 3.1% other. Among non- repeaters, 89.1% were Caucasian, 9.5% black, and 1.4% other. the ma: repeat: were me 15.6: i narriec were d: school, college repeats finishe College He 0f alcc Hillon found t alCoho] assault had nOt denGrist 37 The majority of the men in both groups were employed (75% of repeaters; 82.4% of non-repeaters). Among repeaters, 37.5% were married, 31.3% were single, 15.6% were separated, and 15.6% were divorced. Among non-repeaters, 40.5% were married, 14.8% were single, 27% were separated, and 17.6% were divorced. Among repeaters, 31.3% had not finished high school, 34.3% had finished high school, 28.1% had attended college, and 2.0% held graduate degrees. Among non repeaters, 14.8% had not finished high school, 55.4% had finished high school, 27% had attended college, 1.4% were college graduates, and 1.4 held graduate degrees. Hamberger and Hastings (1990) assessed the involvement of alcohol by using the alcohol and drug section of the Millon Clinical Multiaxial Inventory (MCMI) scales. They found that 56.3% of men who repeated an assault had abused alcohol, but 43.7% had not. Men who did not repeat an assault had used alcohol in 32.4% of the cases, but 67.6% had not used alcohol. The Chi-Square for this relationship demonstrated a significant difference between reported recidivists and non-recidivists in their reported use of alcohol. Repeat assailants also had higher scores on the MCMI drug scale. Pretreatment self-report of alcohol problems was associated with recurrent posttreatment violence. About 56.3% of repeaters had this problem, in comparison with 32.4% of non-repeaters. Furthermore, Hamberger and Hastings (1990) stated that the variables of drug and alcohol successfully discriminated about 71% of the cases. group ‘ repeat Ti a sign. and Ha: court 1 progra: Signif; contini He Vith 3. She ini treatmi L‘lterv; diSPOS; fiISt : the er: Violen< 56792-11 “51am treathe assigns lRtErvj maple: the tre 4‘. live 38 cases. This statistical discrimination distinguishes the group of repeat assailants from the group who did not repeat. The demographic variables reported did not demonstrate a significant difference between the two groups. Hamberger and Hastings (1990) also assessed the difference between court referral and self-referral to the intervention program. The variable of referral did not demonstrate a significant difference between the individuals who reported continuing their assaults and those who did not. Harrell (1991) conducted a quasi-experimental study with 348 partner assailants in Baltimore County, Maryland. She interviewed 171 assailants mandated by a court to attend treatment and 177 not ordered to treatment. The initial interview took place two to eight weeks after the case disposition, and the second interview, six months after this first interview. The purpose of the study was to evaluate the effectiveness of dispositions in cases of partner violence. To do this, Harrell evaluated the incidence and severity of abuse in recurrent partner violence by men mandated to treatment in comparison to men not mandated for treatment. The researcher found that 115 of the individuals assigned to the treatment group completed the initial interview, and 122 of those not assigned to treatment completed the initial interview. About 96 individuals from the treatment group responded to the follow-up interview, while 97 from the group not in treatment completed the 39 follow-up interview. The percentage of individuals who agreed to participate both times was very similar for the group in treatment and the group not in treatment. Harrell (1991) found that 54% of the assailants were between 26 to 35 years old. About 47% had high school educations, and 85% were currently employed. The use of alcohol or drugs was reported in about 64% of the cases. Harrell did not report the ethnicity of the assailants under the two treatment conditions. She found that a higher percentage of men in the treatment group were living with a spouse than in the group not receiving treatment. About 83% of the individuals in the treatment group did not have prior criminal records. This was a high percentage in comparison to the group not assigned to treatment, in which 64% of individuals had prior criminal records. In comparing the group assigned to attend treatment with the group not assigned, Harrell (1991) did not find significant differences between the two groups in severe violence or threats of violence. Fifty-seven percent of the individuals assigned to treatment reported no acts of physical aggression, while 88% of individuals not assigned to treatment reported no acts of physical aggression. Individuals assigned to treatment were significantly more likely to have new domestic violence charges than offenders not assigned to treatment. Harrell stated that, overall, individuals in treatment condition committed more acts of physical violence, required more calls to the police, and o: .e the (193 1‘. \ E‘e 2C3, 40 had more new charges against them than those not assigned to treatment. 59L- 0 1: a, - 7st“ ; on 'e-e.t Pa_ e a—:a’ ..ts In the study of DeMaris and Jackson (1987) the partners of the clients in treatment were not contacted to obtain the number of assaults. Therefore the number of assaults that the male assailants reported was not confirmed with any other source. The records of police visits to the house of the defendants were not examined. DeMaris and Jackson (1987) also cautioned against extensive generalization of their findings because the number of subjects who participated in the study was small. The summary of the three studies by Edleson and Grusznski (1988) was not consistent in comparing assailants who completed the counseling program with those who did not complete the program. Although the intention of the researchers was to compare these two groups, in study number two there was no comparison with non-completers, and no explanation was provided for this. The comparison between study one and three, where groups of non-completers were included, reflects mix findings. In study one, completers were found to be less violent, while in study three completers and'noncompleters were not statistically different in their levels of violence. Thus, based on this review, it appears that men who complete treatment programs are not always less violent than non-completers. In comparing the frequency of violent acts in studies one and HO: vb. 41 three, it was found that completers always reported fewer acts of violence than non-completers. The frequency of threats of violence was not reported in study one. In study three, it appeared that the frequency of the use of such threats was higher for program completers than for non- completers. Edleson and Grusznski (1988) acknowledged that threats of violent behavior continued for participants in the three studies. Furthermore, the report of Edleson and Grusznski (1988) did not present a statistical analysis in the summary of the findings for the three studies. Such an analysis would inform the reader about consistent characteristics and behavior of abusers coming to treatment programs. For instance, the summary indicated an increase in the number of men who had received prior mental health treatment and prior chemical dependency treatment. Prior mental health treatment was reported by 64% of participants in study one, by 84.3% of participants in study two, and 90% of participants in study three. Prior chemical dependency treatment was reported by 25% of participants in study one, 34.5% in study two, and 42% in study three. Although we see a pattern of increase in the prior use of these services, the explanation for this trend is unknown. It would be interesting to know if these services were used in response to prior assault on a partner. Another example of the lack of good statistical comparisons between the three studies was in the matter of court referrals to treatment, which ‘h 544: ‘tll (3 42 showed a variation between the studies. In study one, 9.1% of participants were court referred, 10.7% were referred in study two, and 7.4% were referred in study three. The variation in this number among the studies was not analyzed or discussed. Another limitation of these studies was in their statistical tests. Edleson and Grusznski (1988) did not present the value of Cramer’s V for the strength of the relationship between the variables of completion of treatment and further assaults. Furthermore, these researchers did not use logistic regression to determine if these two groups discriminate on the variable of program completion. Logistic regression would help to differentiate the variables that best predicted the individuals who repeated their assaults. Additionally, although they obtained reports of the assaults from assailants and victims, the researchers failed to compare the groups in the three studies. Only study two compared the reports of violence given by males and females, but it did not find significant differences. Another flaw in the study of Edleson and Grusznski (1988) is the lack of a description and definition of racially mixed individuals in the sample. Furthermore, the reason that these individuals did not define their own ethnic identity is unknown. Perhaps the researchers meant to say that five percent of these assailants were from diverse ethnic backgrounds with no commonality to group them together. c'---_ '. A. o In 43 o ' o e ea a t e ' an s The studies on repetition of partner assault did not show significant differences in many of the variables studied. DeMaris’ and Jackson’s (1987) study of individuals who had completed a treatment program did not find significant differences between those who reported repeating an assault and those who did not. after completing a treatment program. The third study of Edleson and Grusznski (1988) did not find significant differences in the level of violence reported between completers and non-completers of a treatment program. Similarly, the study by Hamberger and Hastings (1990) did not show statistically significant differences between self-identified repeat assailants and non—repeaters in most of the variables studied. The significant differences in the Hamberger and Hastings (1990) study were only in the variables of age, use of alcohol, and use of other substances. These researchers found that self- identified recidivists were younger than nonrecidivists, and tended to abuse alcohol more often than nonrecidivists. Harrell’s (1991) study did not find significant differences in severe violence or threats of violence between offenders in treatment and those not ordered to treatment. In the first study of Edleson and Grusznski (1988), there was a statistically significant difference between individuals who completed the treatment program and those who did not in their level of violence. Completers were nonviolent at follow-up. repeat betvee signii This 6 treats seven resear differ used d test s Gruszn Zackso needed 3638a: not re Amino: likely 44 Thus, not all the studies that compared individuals who repeated partner assault found significant differences between the groups. One study out of seven study found a significant difference between repeaters and not repeaters. This difference was whether or not repeater completed the treatment program. There was little consistency among the seven studies reviewed. A possible reason for this is that researchers did not use similar measures. Each study used different measures to evaluate its sample, and each study used different methodology to evaluate its sample. Although most studies assessed victims and assailants (Edleson 8 Grusznski, 1988; Harrell, 1991), some did not (DeMaris 8 Jackson, 1987). For the most part, the statistical power needed to analyze the design and implementation of the research was not presented or discussed in the studies. The issue of repeating assaults depends on whether or not researchers could identified subsequent assaults. It is unknown if assailants stopped abuse or if they were less likely to be identified as a perpetrator. Summazy of Treatment Programs Like experiments with police intervention, studies from treatment programs lacked investigations of assailants who received all legal sanctions. These studies did not examine the effects of incarceration only, fines only, probation without treatment only, or probation with mandate for treatment. The information regarding legal intervention was limited to the number of referrals from the court to by the eel 45 treatment programs (Edleson 8 Grusznski, 1988; Edleson 8 Tolman, 1992; Grusznski, 1986; Hamberger 8 Hastings, 1990; Rosenfeld, 1992; Saunders, 1992). The results of most of these studies did not distinguish findings about the group referred by the court from findings about the group referred by other sources (Rosenfeld, 1992). Due to this limitation, the characteristics of court-mandated assailants are not well known. Only one study took these differences into account and did not find significant differences between individuals who were court-ordered to treatment and self-referred partner assailants (Edleson 8 Grusznski, 1988: 3). These findings have led to the assumption that the legal experience has a minimal effect, or no effect, on these assailants. Although these studies used samples that were similar to each other, they were very different to police studies. Treatment programs did not accept all self-referred individuals or all those referred by the court. The literature on treatment programs describes the screening process and the individuals rejected from treatment (Rosenfeld, 1992). Rosenfeld stated that treatment programs only selected individuals who were likely to complete the program. In the screening process, individuals who had alcohol and/or drug problems or mental illness, or did not demonstrate a motivation to attend the program were excluded from participation (Gondolf 8 Fisher, 1988). Thus, volunteers in treatment programs do not differ significantly from those under court mandates, because Stu: H " I 5046‘ t.‘“ I V ‘H‘ ‘44". . 3."!- ‘8‘ 46 "difficult" individuals are excluded. The male partner assailants who were referred by the courts to treatment programs, but were then rejected for not meeting criteria, have not been described in the literature. Furthermore, for the most part, the studies reviewed ignored individuals who did not complete treatment programs. However, the studies reported by Edleson and Grusznski (1988) made an effort to compare assailants who completed a treatment program with those who did not. Table 1 presents a summary of the findings of studies of treatment programs. It shows the variety in the characteristics of assailants. The number of men in the samples varied from 46 to 237 assailants. Most of the studies found that assailants were about 30 years old. Unemployment varied from 9 percent to 37 percent. Not all studies reported the same variables in a standard format; this has limited the comparison of findings. Overall, the studies of treatment program showed similar characteristics of participants, who were mostly white, about 30 years old, and employed. Other findings includes the percentage of court referrals to treatment which varied from 3 to 70 percent. Thus, it seems that treatment programs do not attract ethnic minorities, men with low levels of education, or those who are unemployed. Furthermore, the treatment- program studies did not examine all the legal consequences that assailants faced after assaulting a female partner. Li perpetr person it was treatae that di The lit these I signif i Graszns weeks c Di the C111" Offende 47 Lipsey (1992) argued that while more weeks a perpetrator is in treatment, the less likely it is that the person will return to commit another assault. Accordingly, it was expected that the studies that provided more weeks of treatment would provide significant differences than studies that did not provide the same number of weeks of treatment. The literature reviewed in the current study did not find these results. The only study were completers reported significantly less violence, was the study of Edleson and Grusznski (1988:1). This study provided on average of 11 weeks of treatment. Different than the current section, the next section of the current study presents the studies of traditional repeat offenders in criminal justice. Characteristics of Repeat Offenders For All Crimes This section interrupts the analogy presented at the beginning of the current chapter. This section is related to studies that predict characteristics of parolees and probationers who resumed contact with the criminal justice system. Although this section does not continue the steps of the criminal justice system explained in the previous analogy, it is the goal of the current study to conduct a prediction of assault with partner assailants. Therefore this literature ought to be reviewed. , Tahre Stud» Berna Berna '\ gala: Y 1 .acxs 4! $1.85 YES! [0 Fr \¢9° Table 1 - Characteristics Sample stggx gins Bernard and 46 male Bernard (1984) assailants DeMaris and 53 male Jackson (1987) assailants Edleson and 63 male Grusznski assailants (1988:1) 57 female survivors Edleson and 86 male Grusznski assailants (1988:2) 57 female survivors Edleson and 159 male Grusznski assailants (1988:3) 121 female survivors Edleson and 283 male Syers (1990) assailants Fitch and 188 male Papantonio assailants (1983) Hamberger and 106 male Hastings (1990) assailants Harrell (1991) 237 male assailants 237 female survivors Purdy and 170 male Nickle (1981) assailants Saunders (1992) 182 male 48 of Partner Assailants Assailant’s Percent Ethnicity nem NOR. NOR. N.R. N.R. 96.6% white 9.0 86% Cauc. 16.3 6% Afr. Am. 2% Nat. Am. 1% Asian Am. 88% Cauc. 16.0 2.5% Afr. Am. 2.5% Latino 2.5% Nat. Am. 4.1% Mix. Races 74% Cauc. 34.0 11% Afr. Am. 3.8% Nat. Am 2.7% Latino N.R. N.R. 86% Cauc. 20.0 12% Afr. Am. 2% Other NOR. 15.0 NOR. NOR. 76% white N.R. Average ASE 30.8 32 34 32 31.1 26-35 49 Table 1 (cont'd). eredx Bernard and Bernard (1984) DeMaris and Jackson (1987) Edleson and Grusznski (1988:1) Edleson and Grusznski (1988:2) Edleson and Grusznski (1988:3) Edleson and Syers (1990) Fitch and Papantonio (1983) Hamberger and 30.0% Hastings (1990) Harrell (1991) 48.5% Purdy and Nickle (1981) Saunders (1992) Court Time of Beferrel Treee N.R. N.R. 30.0% 12 months 9.1% 12 months 10.7% 9 months 7.4% 6.5 mo. 33.3 % 6 months assailants 31.0% N/A 12 months 6 months 3.0% 6 months 70.0% N/A Length of Treatment 12 weeks 24 weeks 8 weeks 24 weeks 24 weeks 12 weeks N/A 16 weeks 11 weeks 7.8 weeks N/A Repeated Assault N.R. 35% 40% 33% appx. 44% appx. 35-46% N/A 30% 22.5 appx. 41% N/A ref 11$: 0 be. 50 'ct' 0 V' enc The science of the prediction of violence has been used and supported by the field of criminal justice. This is reflected in a number of studies and in court decisions demanding that mental health professionals predict the risk of defendants' engaging in violent behavior. This risk prediction is conducted by means of court referrals, parole R1,. _‘_..- -M ...... supervision, or records from criminal justice organizations (Beze1_g;_ye_ze§e11e, 1983; Burgess, 1928; Geerken & Hayes, 1993; Griffieth, 1985/1987; Hart, 1923; Monahan, 1981; Ohlin, 1951; Petrila, Otto, & Poythress, 1993; Saunders, in press; Sepejak, Menzies, Webster, & Jensen, 1983; Tereeefifi v, Regenge e: Quivegsity e: Califegnia, 1976; Warner, 1923; Webster et al., 1984). Prediction of violence has been based on actuarial ~eerr._ predictions, which have been used in psychiatric hospitals as well as in criminal justice organizations (Kirk, 1989; Marguart, Ekland-Olson, & Sorensen, 1989; Tezesofifi g, BMW. 1976)- In the criminal justice field, actuarial predictions have been made to help parole officers decide whether an individual was more likely to follow or violate parole rules (Dillingham, Montgomery, & Tabor, 1990; Griffieth, 1985/1987; Mannheim & Wilkins, 1955; Ohlin, 1951). Actuarial predictions of violence have used the behavior of persons with similar characteristics to predict the future behavior of a particular individual (Brizer, Ste : 1972) indix Thu , and 5 35565 or rj J“-.- .;.: .c ‘flu ..11’ 4338: 1'18 1 the I :!235 “fig ‘L .“e Fete $5.4. 51 1989; Marguart et al., 1989; Monahan, 1981; Monahan & Steadman, 1994; Ohlin, 1951; Wolfgang, Figlio, & Sellin, 1972). These predictions estimate the likelihood that an individual will act as a similar person acted in the past. Thus, the prediction takes into consideration the situation and similarities in personal characteristics (Morris & Miller, 1985). Monahan and Steadman (1994) noted that the actuarial assessment observes the relationship between specific cues or risk factors and the occurrence of violent behavior. For instance, one study revealed that if a person was arrested four times, the probability was 80% that he or she would be arrested a fifth time. If the person was arrested 10 times, the probability of being arrested again was 90 percent, and the probability was 42% that the offense would be serious (Monahan, 1981). This is an example of how specific variables from the criminal justice system help to predict the likelihood of an offender's return to the system. a ° 1 i ' ns ' C 'mina u A number of studies have examined the characteristics most likely to be associated with individuals who violate conditions of parole (Burgess, 1928; Griffieth, 1985/1987; Hart, 1923; Warner, 1923). The purpose of these studies was to determine predominant characteristics of such individuals as well as to identify characteristics of individuals who were most likely to be parole successes. Most of these studies have been conducted retrospectively with the records ‘\ A 52 of these offenders (Burgess, 1928; Geerken & Hayes, 1993; Griffieth, 1985/1987; Hart, 1923; Ohlin, 1951; Warner, 1923). These studies are discussed in this section in chronological order. Warner (1923) is associated with the first stud on prediction of future behavior of offenders. Warner used 680 records of prisoners in Massachusetts. In this group, 300 were parole violators, 300 were parole successes, and 80 were not paroled but came in front of a parole board. The purpose of Warner's study was to ascertain the criteria that the Parole Board used to determine whether or not to grant parole to an inmate. Warner examined 64 factors that the parole board used to make decisions about parole eligibility. He concluded that none of the factors was a significant criterion of success or failure. Warner also stated that this situation could not improve without a major reorganization in the method and type of information obtained from parolees. Warner's (1923) study had the limitation of not reporting the statistical values of the differences between successful and violator parolees for each of the 64 factors. I Hart (1923) criticized Warner's study because it did not provide statistical significance tests for the data. Hart (1923) proposed that all the elements studied be combined into an outcome score for each inmate. The scoring system would be based on the intercorrelation between the items as ' VJ .7 53 well as the correlation with parole violation. A system for weighting and scoring each element would be developed. Burgess (1928) followed Hart’s (1923) suggestion and applied such a system to parole prediction, particularly to the factors related to the success or failure of parolees. Burgess (1928) studied the significant factors that differentiated parolees who returned to the criminal justice system from those who did not. He analyzed these factors after gathering information from 1,000 records of parolees from the Illinois State Reformatory at Pontiac. He then tested the factors with information from another 1,000 records at the Southern Illinois Penitentiary at Menard. The purpose of the study was to identify, in a scientific manner, those factors in the life of a person that would determine whether or not the parolee would violate parole conditions. Burgess (1928) used statistically significant elements to create a table of the expectancy rates of parole OWM“MW/ violation and non-violation. The success or violation score from parole supervision was assessed for the period during which the offender was under probation supervision. Burgess indicated that the expectancy of success or failure of a parolee was due to a combination of the factors that were _ ...~_-——'_-.‘-_ ... ..- favorable with factors unfavorable to parole success. He . *cn...’ ....—. .. found that b§33999r191t°,21 factors identified 98.5% of parolees who did not violate parole conditions. 54 Burgess (1928) identified the following significant factors in predicting parole success: the type of offense, having parents alive, marital status, type of offender (frequent versus first-time), association with others, community factors, leniency from the judge and prosecutor, previous criminal record, work history, punishment record in the institution, intelligence level, psychiatric personality, and psychiatric prognosis. ‘;;) Ohlin (1951) also tested factors predicting ,individuals' success under parole supervision. He developed the prediction rate from 4,941 paroled cases in the Joliet- Stateville and Menard Divisions of the Illinois State Penitentiary System. These factors were selected from cases paroled from 1940 to 1945. The purpose of Ohlin's study was to test the reliability of information used for parole N r...u-.-»~ .-.. . "rrmm- ‘r—f prediction. Ohlin (1951) selected 12 factors from a list of 27 that the Illinois system used continuously. The other 15 factors originally selected for study did not meet the statistical criteria. Ohlin's major finding was the identification of these 12 factors that best predict parole successes and failures: type of offense, sentence, type of offender (frequent versus first- time), home status, family interest, association with others, work record, community of residence, parole job, number of associates, personality, and psychiatric prognosis. K _ ‘_ 55 fag? Monahan (1981), in his book of prediction of violence, argued that the definition of violence should be a probability figure based on a determined context (e.g., the formula should be based on a specific environment). Monahan . ..y—m-M “mm... 7 . Hr~~-._._ “...—.6" stated that six aspects of the environment provide guidance for the formation of environmental predictors of violence. These aspects are the following: 1) family, 2) peers, 3) job, 4) availability of victims, 5) availability of weapons, and 6) availability of alcohol. Monahan argued that these six different environmental factors correlate with violent behavior and can potentially be used for prediction in the individual case. Monahan (1981) explained that family environment is important because the family plays a critical role in supporting or discouraging violent behavior. If a family encourages robbery as a career, and violence by other members of the family occurs frequently, the probability that an individual will be violent in this environment is greater. If a person has a family that promotes nonviolent modes of interaction and satisfaction of needs, the probability that violence will occur in this setting is less. Monahan also stated that family members are frequently victims of violent behavior. Using prior research as evidence for these statements, Monahan said that in 77% of emergency commitment cases where the person considered violence, the victims were family members (Bard, 56 1969; Discroll, Meyer & Schanie, 1973; Monahan, 1977; Skodol & Karasu, 1978). Monahan (1981) also stated that personality and situation were not independent factors./ Some personalities W look for potentially violent situations. Monahan (1981) m“"~' Wm 'I"~ .-- ....-1- ——. w. H). mm. q» .. VN’WH-H’V' cited Wicker (1972), Endler and Magnusson (1976) who stated that specific situations attract certain personalities. And there might be an interaction between the individual’s behavior, the situations, and the character of these situations. The concluding argument was that there might be a great deal of overlap between the predictor items and the 'persenality;_ For this reason, a predictor instrument for violence most likely will contain one of these two aspects. For instance, being poor and unemployed might be related to committing violent acts. If a person is poor and unemployed, then this person will be considered as having a double potential for violence. According to Monahan, being poor and unemployed correlate with each other. Therefore, he suggests that these two factors should be considered -.VM,” r- -.————-’ w-‘ separately as predictors of violence. (é) «Griffieth (1985/1987) also studied factors related to the success or failure of parolees. She sampled 405 records of women in halfway houses in Michigan for the year following their release to these houses. Griffieth obtained the variables for the research from the files at the central office of the Michigan Department of Corrections. The purpose of the study was to design a statistical model for 57 predicting whether or not an incarcerated woman would succeed in a community correction program. Using factor analysis, Griffieth (1985/1987) identified 17 factors from the women’s pre-prison experience and 19 factors from their post-prison experience. The author found two factors from the pre-prison variables that significantly differentiated women who returned to prison from those who did not: prior criminal history with a serious instant offense, and not having custody of children. From the post- prison variables, Griffieth found seven significant factors that differentiated these women: not having custody of children, prior criminal history and a serious instant offense, being an older woman with alcohol problems, unstable employment history, minor prison misconduct, short prison stay without prison program participation, and juvenile arrest history. i?) Geerken and Hayes (1993) argued that if rearrest was used as a measure of the failure of probation and parole the probation and parole systems were complete failures. They found that between one-third and two-thirds of all probationers were rearrested, and from one-quarter to one- half of all parolees were also rearrested. Geerken and Hayes (1993) examined offenders on probation and parole who were charged with burglary and armed robbery between 1974 and 1986. They found that only 8% of all the arrests for burglary and armed robbery involved adults on probation. They also found that less than 2% of all arrests for these 58 offenses involved adults on parole. Geerken and Hayes stated that these percentages were surprisingly low and contradicted findings in other literature. They assumed that the probability of arrest for a crime was identical for individuals under probation or parole supervision as for those who were not. One limitation of this study is that the risk of future arrest was not analyzed carefully. The risk was obtained based on the frequencies of arrests, but it was not based on a statistical analysis of the data. The authors could have used logistic regression or discriminant analysis to identify the strongest factors of risk for arrest on burglary or armed robbery charges. Furthermore, this study did not examine the personal characteristics of these offenders on probation or parole. ar ct 'st' 0 at In the last section we have looked at studies on prediction of assault in criminal justice in general. Most of the studies about the characteristics of repeat offenders have resulted from pressure by the criminal justice system, which has the goal of identifying those who are most likely to repeat an offense if they are not incarcerated. However, members of parole boards often use subjective criteria to make their decisions about granting parole. Various actuarial studies have examined the characteristics of the individuals under parole supervision who have returned to prison. 5 nside H: obtains and nus Stuii v1- hOd‘G aSSau just: In Co prCSE 552:0 59 prison. These studies recommended that parole boards consider different criteria when making their decisions. Most of the studies of actuarial prediction have obtained significant differences using the variables of kind and number of previous offenses, employment status, and community of residence (Burgess, 1928). These studies selected their variables from the most common variables in the records of offenders (Burgess, 1928; Griffieth, 1985/1987). The majority of actuarial prediction studies used felonious assault instead of misdemeanors (Geerken & Hayes, 1993; Griffieth, 1985/1987). Another problem of actuarial prediction is the measure of recidivism. While some authors used rearrests or return to the institution as a measure of recidivism, others recommended using only arrest for the same offense (Geerken & Hayes, 1993; Griffieth, 1985/1987). Therefore, misdemeanor offenses are not consistently considered as recidivism. Most of these measures of recidivism of probationers and parolees did not consider partner assault. As partner assaults are often found in the misdemeanor category, the studies of prediction of individuals on parole and probation could easily ignore them. One reason cases of partner assault come to court as misdemeanors is that the criminal justice system is reluctant to prosecute partner assailants. In contrast, the criminal justice system is very likely to prosecute assaults on strangers (Field, 1993; Fields, 1994; Hammond, 1977). Thus, the offense of partner assault has not be predic F recidi to the before ffenc argoec easilj 21985, :ente: the r cons; :: Guer- Varia anal: Provj the 1 pTES 60 not been examined as a separate component in the studies of prediction of parolees' and probationers' recidivism. Furthermore, actuarial prediction used the term recidivism for offenses instead of likelihood of returning to the institution. Some offenders repeat their assaults before they are arrested by the police; and many repeat offenders are not rearrested. Geerken and Hayes (1995) also argued that probationers can avoid reconviction even more easily than they can avoid rearrest. Moreover, Griffieth (1985/1987) found that most women in community correction centers were returned to prison as result of violation of the rules of the center, such as tardiness and alcohol consumption. Therefore measuring the repetition of the same offense is inadequate because such repetition is infrequent. The next section presents a model to organize the variables in the current research. Again this is out of the analogy presented at the beginning of the chapter, but it provides a consistent framework to analyze the variables in the study. Ecological Assessment of Repeat Offenders The current study is more complex than the analogy presented at the beginning of the chapter. Before explaining the current study is important to introduce a model that will help organize the variables and its analysis in the current study. The purpose of the current section is to expll ecologi Th assess assault develop life of of indi ecologi charact element then. levels (Earls: 61 to explain the components of this model known as the ecological model. “—0—...«4 ...“, ”Hap-0‘ - -~ The current study will use an ecological approagh to assess the characteristics of men who are rearrested for assault and’SSEEééy offenses. The ecological model developed by CarlsonflL1984) encompasses many dynamics of the life of individuals. This model also presents the behavior of individuals in a sociopolitical context. The use of the ecological model takes into consideration the personal My.-.“ feere_. characteristics of partner assailants as well as the elements that play a role in the court procedures against them. The current ecological analysis corresponds to four levels identified in the ecological development theory (Carlson, 1984). These are the ontogenetic level, the microsystem, the exosystem, and the macrosystem (Carlson, ~Hm. . —._m¢-~ 1_, ..__- “...-””4- 1984; Dutton, 1985, 1988; Edleson & Tolman, 1992). In this section, each level of the ecological model will be defined. ii) The first level is called ontogenetic and encompasses the personal developmental history of the individual (Carlson, 1984; Dutton, 1985, 1988). This includes education, employment, income, age, and ethnicity. Other components include use of alcohol, self-esteem, verbal skills, experience of stress, learning experiences, exposure to violent role models, and options for handling conflicts (Carlson, 1984; Dutton, 1985, 1988). One of the components of the current study is the analysis of the personal characteristics of the assailants. Because these charac will u intera level sale-f Howeve envirc social itarls The c< .. H‘V' n1 ‘otéb 5835). tale is pa the f the c bet'u'e 56tee diffs 62 characteristics are on the ontogenetic level, this study will utilize the ecological theory for its analysis. ,/7 a The second level, the microsystem,¥cprresponds to the ( «1 ...-..— WM interaction between individuals in a social context. This level includes ”the family" (Carlson, 1984, p. 574) or the male-female interaction in an intimate relationship. However, the microsystem could also refer to work environments, neighborhoods, churches, sport teams, or social clubs where individuals are active participants (Carlson, 1984; Dutton, 1985, 1988; Edleson & Tolman, 1992). The common domain in this level is an individual's set of direct interactions with other individuals. The current study examines the assaults of men on their female partners. This indicates an interaction between a male and a female in an intimate environment, and therefore is part of the microsystem of ecological development. In this study, components for analysis at this level include the following: the assaultive incident, the type of weapon, the characteristics of the victim, the conjugal relationship between victim and assailant, the living arrangements between the victim and the assailant, and the race differences between the victim and the assailant. 4. V) ~v) The third level has been labeled the exosystem43nd corresponds to the social and structural conditions in the community. It includes law enforcement and criminal justice practices, distribution of social resources, work groups, support groups, and others (Carlson, 1984; Edleson & Tolman, 1992). profes office deal i coordi envirc offic refer: do int VVHHI assai eh: Hus :ozfo treat Afl.lg ‘y‘r-y 63 1992). An example of the exosystem is the coordination of professionals (i.e., police, prosecutors, judges, probation officers, social workers, and battered women's advocates) to deal with partner assailants (Edleson & Tolman, 1992). This coordination at the societal level influences the social environment in which the assailant participates. The current study examines the referrals from probation officers to partner assault-abatement treatment. Although referrals do not reflect the coordination of services, they do indirectly measure probation officers' interactions with community treatment programs. For instance, if more assailants are referred to alcohol use-abatement treatment, this could mean that probation officers feel more comfortable with treatment for alcohol problems than with treatment for partner assault. In the current study, components for analysis at this level include the following: time in jail, restitution to victim, money paid to court, probation supervision, and mandate for domestic violence- abatement treatment. The current study examines court practices involving offenders who have assaulted their partners. These practices are at the exosystem level. Unlike other studies, the current study will not aggregate court practices. The analysis of court practices will include each component of the judicial system and an evaluation of its influence on individuals who are rearrested for assault. These practices include sanctic Tl identii It focr systems include (<9 64 include type of sentence, time of court procedure, and sanctions imposed by the court. The fourth level of the ecological theory has been identified as the macrosystem,(or the sociocultural level.h .r. fl...- “v—.—H~M~“_flrah.c “...—.- It focuses on societal norms, cultural values, and belief systems (Carlson, 1984; Edleson & Tolman, 1992). This level includes the social rules that create the consistencies in culture, ethnicity, or social class (Edleson & Tolman, 1992). Some of the sociocultural components of this level are sexism, sex-role stereotypes, acceptance of violence in a broad sense, and the norms of the family (Carlson, 1984; Edleson & Tolman, 1992). For example, Edleson and Tolman (1992) have observed that many middle-class families in the United States closely resemble each other. Their similarities are reflected in their attitudes and beliefs, and in their patriarchal standards for male-female behavior (Dutton, 1985, 1988). Carlson (1984) stated that sexism is an example of the norms governing at the macrosystem level, and it is manifested in the criminal justice system when women are not taken seriously despite their being in daily danger. Other elements that Carlson described at this level are the general societal beliefs about sex roles. Males are e“Deemed to be independent and aggressive, and females are expected to be dependent and passive. These sex role SteIt‘eotypes are reflected in the criminal justice system, particularly in the low number of arrests of partner assaila receive it aentior ecoloq; of cri for a: instic {Efiec Corie} indiv; harsh Frog“ inc31v or no and h ‘. e‘93s: 65 assailants and in the low number of sanctions these men receive. Two sociocultural elements of the macrosystem not mentioned by previous researchers who have used the ecological model of partner assault (Carlson, 1984; Dutton, 1985, 1987, 1988; Edleson & Tolman, 1992) are racism and homophobia. In general, society fosters strongly divisive attitudes toward people who are different (Gross, Green, Storck, & Vanyur, 1980; Henley & Pincus, 1978). Examples of this are the genocide of Native Americans and the exploitation and isolation of, and discrimination against, Africans throughout the American continents and the Caribbean. Today this racism is manifested in the policies of criminal justice organizations. Minorities are targeted for arrest and prosecution; and the decisions that criminal justice organizations make and the policies they implement reflect institutional racism (Reiman, 1990; Wordes, Bynum, & Carley, 1994) . For instance, unemployed African American individuals with low levels of education, are often sent to harsher punishment like jail, rather than sent to treatment Programs. Lack of information on the punishment received by individuals arrested makes it difficult to determine whether or not the judicial system is bias free of racism, sexism, and homophobia. The current study does not examine directly the elelllents of the macrosystem level. The macrosystem level is “mucnl less clearly involved in the current study. The result of tht of op; irpor that at di Itn 66 results of the current study can be explained in the context of the macrosystem level. But, the current study is victim of oppressive conditions of the macrosystem level. An important contribution that the ecological theory makes is that it forces its users to understand a level of analysis at different levels. not, . . ugi . ._ - Sl‘ . i-o-a_ 0 -.:e m In summary, the analysis of the characteristics of I repeat offenders includes their personal characteristics, L t ' ' the characterietics of the incident, the court practices lwmq‘ _ _ g ‘ inyolving partner assailants, and the ideology that fosters the behavior of thesemindiyiduals. Thus, the levels of analysis in these four sections correspond to the four levels of the ecological development theory, namely, the individual, the microsystem, the exosystem, and the macrosystem. The analysis of the macrosystem level reminds us to be cautious in interpreting the results of this study due to the strong limitations of the criminal justice system. The Current Research Most of the studies that examined the characteristics of men who assaulted a female partner have been contaminated by a pre-sample selection. Experiments with police interventions screened the cases before providing an intervention. Studies of repeat assaults to partners conducted for treatment centers did not include all partner assai ref le selec evalt crini not I bias just by i stud part fro: 67 assailants who came to court. Thus, these studies also reflect biases in sample selection. To avoid this biased selection, scholars in the area of partner assault should evaluate the procedures used in handling these assailants at criminal justice organizations. Unfortunately, this does not eliminate all the biases, and perhaps is replacing one bias for another. One of the biases of cases at criminal justice organizations is that the cases have been screened by individuals who handled the case previously. One way of studying the details of the legal consequences faced by partner assailants is to follow the cases as they proceed from emergency calls to police intervention, to police reports, to police departments, to prosecutor's offices, and to courts. However, information on these assailants, which was gathered at various criminal justice organizations, is kept in presentence investigation reports at probation departments. Thus, researchers could use the information contained in these records as empirical evidence. The probation department is a criminal justice organization which, because of its legal and routine procedures has a great deal of information regarding partner assailants. Records at probation departments include personal information from the defendant, decisions from the judge and probation officer, treatment referral, violation of conditions of probation, and information about whether or not the assailant completed probation (Canales-Portalatin, 1994). The information contained in the probation records is pr parti indiv more crgar could treat cons: depa: {TEE aha‘ 68 is provided freely by the assailants, victims, and other parties to help determine an appropriate sentence for these individuals. In fact, records from the probation department provide more information than any other criminal justice organization. The use of the information in these records could close the gap between police research and studies of treatment programs. Additionally, it is important to consider that some assailants who come to probation departments are not assigned to partner assault abatement treatment. Probation supervision might include alcohol- abatement treatment, partner abuse abatement treatment, or no treatment (Canales-Portalatin, 1994). Information regarding the treatment referral, or lack of it, is maintained in the probation record. Therefore, the records from the probation department provide an extensive number of cases that received diverse community assignments, not only assignments to treatment groups for partner assault abatement. -The model of the current study is partially based on criminal justice research that has developed a method for predicting which individuals are most likely to repeat offenses. These predictions have often been based on records from criminal justice organizations. Identifying future offenders on the basis of personal characteristics is one of the strengths of the field of criminal justice, and it has a long history. Since 1923, studies have attempted to pre Chara: recore Griff Ohlin st‘di tiara Chara 69 to predict the success of parolees by examining personal characteristics that can be derived from the individuals' records at criminal justice organizations (Burgess, 1928; Griffieth, 1985/1987; Geerken & Hayes, 1993; Hart, 1923; Ohlin, 1951; Warner, 1923). The purpose of most of these studies has been to compare, for a given time period, the characteristics of successful parolees with the characteristics of violators of parole conditions. Therefore, records from probation departments are potentially useful for a study of repeat offenders of partner assault. Such a study may identify factors in the criminal justice system (jail time, fines, type of offense, previous criminal records, etc.) that influence repetition of this offense. The second model for the current study was a series of studies of treatment programs for men who have assaulted a partner. Some of these studies examined the characteristics of such men with the purpose of predicting who among them would commit another assault against a partner (Hamberger & Hastings, 1990; Saunders, in press; Straus, 1993; Tolman & Bennett, 1990). The current research proposes to link the studies in the criminal justice system with those studies of treatment groups for men. Most prediction studies with data from criminal justice settings have not examined repeat offenders who assaulted a partner. For the most part, these studies have examined felonious offenders (Bradshaw, 1987; Burgess, 1928; Gluec are c studi not e Exce; offer fast repea ident inter for r 1993) 70 1928; Griffieth, 1985/1987; Geerken & Hayes, 1993; Glueck & Glueck, 1930; Hart, 1923; Mannheim & Wilkins, 1955; Ohlin, 1951). They have not considered misdemeanor offenders, who are often partner assailants. Conversely, most of the studies about treatment groups for partner assailants have not examined information from probation departments records. Except for noting whether or not assailants had previous offenses, studies of treatment programs have not examined factors in the criminal justice system that might influence repeated partner assault. The treatment-group studies identified the characteristics of repeat assailants from interviews with individuals who contacted treatment programs for men (Edleson & Tolman, 1992; Hamberger & Hastings, 1990). The current study attempts to predict the characteristics of individuals who are likely to commit multiple assaults. This prediction will be based on data from records of men who have assaulted a partner and have been referred to the probation department in a Michigan city. The records at probation departments in the county will provide information about repeated assaults on a female partner. However, the methodology of the current study also uses reports of arrests and convictions recorded at the police organizations in the state of Michigan. This electronic system does not identify the sex of a victim and his or her relationship to an assailant. Conversely, offender with drug or property offenses, or drinking and 71 driving, will not be considered as repeat offenders of assault and battery. As partner assault was not considered a specific crime in Michigan Statutes until the Spring of 1994, it is not possible to restrict the current study to repeat partner-assault offenders. The use of the electronic system, however, allows the current researcher to examine assailants who committed assault and battery in any county in the state of Michigan. This system is a form of communication between criminal justice officials that enables them to obtain information about an individual's prior arrests and punishment and/or pending warrant arrests in Michigan courts. Therefore, this study measures with certainty the number of individuals who returned contact with the criminal justice system after committing assault and battery, but it cannot identify all the assailants who have repeated a partner assault. The electronic system from the State Police does not indicate either the name or the gender of the victim. This information is recorded in police reports at local jurisdiction, but not in state records. Another difficulty with studying partner assault is that much of it occurs without being officially recorded. Because the current methodology does not include direct contact with victims or assailants, it is impossible to measure definitively the recurrence of partner assault. A common feature of studies of treatment groups for partner assailants and actuarial studies in criminal justice 72 is their interest in the recurrence of a particular social offense. Perhaps these studies are based on the concept that the recurrence of an offense pertains to the outcome evaluation of a particular intervention. This study focused on the common goal of developing a prediction for men who commit repeat assault and who have committed an assault against a female partner. Most studies of characteristics of partner assault have obtained their samples from treatment groups. Using these samples from a large number of individuals mandated by the court to attend treatment has not always resulted in successful studies. Many individuals who have assaulted a partner have not had such mandates (Canales-Portalatin, 1994; Sherman, 1992). Furthermore, the effect of the court experience on these assailants has been measured from a limited perspective, since researchers have not examined criminal justice data in detail, and they have overlooked information from third parties. Conversely, actuarial predictions about individuals who return to the criminal justice system have not specifically considered partner assailants. These studies have, however, considered some of the experiences of parolees with the criminal justice system, including previous offenses, number of offenses, and types of offenses (often property offenses). The methodology of these studies is worth emulating, but with a different sample population. Therefore, the current prediction used a common method of risk assessment employed in criminal justice settings. 73 This process included inspecting the criminal records, for a limited time period, of parolees or probationers. The method also examined variables that correlate with reoffense. The sample in the current study, however, was one that has not been traditionally used for studies of prediction in criminal justice or in treatment programs. The sample was selected from records of men who have assaulted a female partner and have been referred to a probation department. The purpose of the current research was to utilize records from a probation department in a midwestern city in the United States. The records examined included closed files of men who assaulted a female spouse or partner or former spouse or partner. This selection of cases was intended to provide an assessment of risk of those individuals who were likely to have multiple referrals to the probation department for committing acts of violence. In these cases, the probation department has intervened at least once because of assault on a partner. Multiple referrals could result from subsequent assaults on a new partner, former partner, or other individuals. In comparison to samples of men who have been diverted into treatment groups by the judicial system, samples from the records of probation departments represent a relatively broad sample of men who have interacted with the criminal justice system (Canales-Portalatin, 1994). ”-— -_..... --—u-._—- .—..a ......— char ret‘. of ‘ ret' hav 5? so: the 332 5X; ES 74 Furthermore, these records contained a sufficient number of details about the partner assaults to measure the characteristics of individuals who were at risk of returning. Traditional risk assessments examined a number of variables to identify those who were most likely to return to the criminal justice system. Risk assessments have also been conducted to predict who is likely to repeat a partner assault (Edleson & Grusznski, 1988; Hamberger & Hastings, 1990). The traditional procedure is to examine socioeconomic variables, called significant elements, from the criminal justice records of individuals in the community. One of the purposes of the current study was to help to educate probation officers and other criminal justice professionals about the differences in treatment that assailants receive in the system. Through observing the issues of sexism and racism, the current study served as an instrument for educating and training to criminal justice professionals. This material from the current study may lead to intervention with various levels of criminal justice organizations. t es Traditionally, studies of repeat offenders have examined the personal characteristics of these individuals as important factors in prediction studies. However, studies using different sample pools have identified different characteristics. For instance, police arrest a 75 disproportionate number of African American men, and men who are poorly educated, unmarried, unemployed and have low incomes. Treatment programs, however, intervene with a disproportionate number of Caucasian, better educated, married men with relatively high incomes. Therefore, the following was a major hypothesis of the current study: The ingegyention eeleeted by a judge (e.g., jail, substance b e m a tne - buse trea ent o a r. 2"?!! '1 ,- 9e :- . cha_; - '— ' s o ,- ee=='.— epeeifiieelly Ieee, mari§e1 s§atus, eeucatien eng empleyment. i}u{ The literature concerning rearrest has indicated that f E being young and having had previous contact with the criminal justice system are factors related to further criminal justice involvement later in adult life (Glueck & Glueck, 1950; Mannheim & Wilkins, 1955; Sepejak et al., 1983). The literature from treatment groups for men has also suggested, in general, that the assailant's previous contact with the criminal justice system is related to partner assault. Some of the specifics of assailants' history with the criminal justice system (e.g., number and type of prior offenses and instant offense), have produced a profile of an unsuccessful parolee. However, this is not true for studies from treatment groups, because they have only superficially examined assailants' criminal histories. Limited examinations of the history of contact with criminal justice have produced nonsignificant results. Hence, the following 76 was a second hypothesis of the current study: A pagtner .s=;' :1 W10 .2 lad no e . r ‘m-‘_'- e? I u- - 'k:. 0 a sa t as e su ed b ol'ce r est. The literature concerning men who have abused a partner indicates that the use of substances is a key factor differentiating men who commit subsequent physical assaults (Gondolf, 1988; Hamberger & Hastings, 1990). Accordingly, the following was a third hypothesis of this research: The .7 -u- t . 7_-s - ces '1 th- Ose..t5-na -e f,-. f'_st a s u a es 't e ' e t e s ' ant w' e e s me 5 ed b ° rres . Additionally, the literature on risk prediction for individuals who return to the criminal justice system has neglected to observe characteristics and experiences considered under the microsystem and exosystem levels of the ecological model. Therefore, the following was a fourth hypothesis of this study: IncTusion of yeriables considegeg u e e 'cal t o wi ov'de be e e 'cto s o e c a c e is ' div'dua 8 who e ea ssa t. w r ° c ud' variab es nder t e mi s te os ste c acte ’st cs wh‘ch are ra e “no uded ' u 'e ' ove the red' to wer t e CHAPTER II METHOD Subjects During phase one of this study, the cases of assault and battery were selected from referrals to a Probation Department in Michigan in the years 1988, 1989, and 1990. In these cases, men were perpetrators of assault on a female partner or ex-partner. During these three years, the judges referred 465 cases of all types of assault and battery to the probation department. Using an operational definition of partner assault, the current study used a sample of 182 male assailants. This represents 39% of all the defendants referred to the probation department for assault and battery during the years indicated. During follow-up the multiple referrals of these 182 male assailants were examined. Seeting The Probation Department works with referrals from judges of this court. Of the three probation departments in the same county, one was selected which had the largest number of cases of assault and battery. The probation department maintains closed records of individuals referred from the court. These records included copies of confidential information, such as the presentence 77 78 investigation report, police report, crime history report, and presentence questionnaire. The other two district court probation departments in this county retained similar documents in their closed records. The records from the other two probation departments were used to follow the assailants identified in phase one of the study. This information is considered confidential. Therefore, the data collection process protected the confidentiality of the records of these defendants. These records also contained other information, such as letters from victims to the judges or probation officers in charge of the particular case. The current study used this information to answer questions from an instrument for data collection. Apparatus The construction of the data collection instrument was based on the conventional method of conducting base expectancy rates of parolees (Burgess, 1928; Glueck & Glueck, 1930; Griffieth, 1985/1987; Ohlin, 1951; Warner, 1923). This method included a selection of the most common variables in the records of the defendants. To find these variables in the records, three preliminary studies selected random files of defendants at the probation department. The studies quantified the number of variables that were present in the randomly selected files. 79 'e ns me t ctio The first pilot study selected 120 random cases from the total cases of assault and battery for the year 1992. The information collected included the following: age, date of birth, sex, ethnicity, education, employment status, gender of victim, relationship between the victim and defendant, fine, court costs, number of days spent in jail, number of days under probation supervision, restitution, protection under the Spouse Abuse Act, placement for treatment, completion of treatment, completion of probation, type of trial, repetition of same assault to same victim, and length of time after the first assault occurred (Canales-Portalatin, 1994). Appendix A contains a copy of the form used to gather information from the probation department records from 1992 for the first preliminary study. The second pilot study used files from the years 1988 to 1990 to make an inventory of the variables that can be collected from the closed files. Another purpose of this pilot study was to evaluate the overall consistency in obtaining these variables. Griffieth (1985/1987) sampled 15 records from a total of 405 files for a similar procedure. Therefore, after the current study identified 465 cases of assault and battery, the second pilot study used 15 randomly selected cases to generate an inventory of variables. Appendix B contains the variables that were present in 15 records. 80 For the purpose of testing and refining the instrument from the inventory study, a third preliminary study selected 51 files at random. Griffieth (1985/1987) used a similar procedure to evaluate the frequency of information in the records. The third pilot study used the instrument in Appendix B to quantify the frequency of information in the records that were selected at random. Appendix C contains the instrument that was used to obtain data for phases one and two of the current study. The instrument contains the following 3 main categories of the ecological model for organizing the data: ontogenetic level, microsystem, and exosystem. W1 This section of the instrument described the variables regarding a defendant’s personal characteristics. This section collected information on the defendant’s age, marital status, education, employment, income, and ethnicity. These variables have been defined by several researchers as indicators of personal stability (Mannheim & Wilkins, 1955; Glaser & O'Leary, 1968; Glaser, 1969; Griffieth, 1985/1987; Reeds & Woods, 1971). Some of these variables have also been identified as common characteristics of partner assailants and indicators of conformity with society's conventional rules (Fitch & Papantonio, 1983; Hamberger & Hastings, 1990; Rosen, 1993; Sherman, 1992; Sherman et al., 1991; Sherman et al., 1992). r _ 81 ime ' 'al Of e se Age at time of initial offense referred to the assailant's age at the time of the initial partner assault. Calculating the difference between the date of birth and the date of the assault provided the age of the defendant at the time of the assault. The age variable was organized in a ratio scale. Mazitel Status Marital status referred to the defendant's self- classification of his legal marital status at the time of interview at the probation department. The following were the choices for this category: married, separated, or single. The marital status does not indicate the relationship between the victim and the assailant. Each category of the marital status was assigned a nominal value. W The education variable referred to the number of years of formal education that the defendant had completed at the time of the interview with a probation officer. The number recorded was the number that an assailant self-reported in the presentence questionnaire. In a few cases there was no response, and the educational level was acquired from the presentence investigation report. Educational level was organized in a ratio scale. 82 t a 5 Employment status referred to whether or not a defendant was employed at the time of the interview at the probation department. The possible choices for this category were yes or no, and they were organized in a nominal scale. The value of zero was assigned to those who were not employed and one for those who were employed. Fitch and Papantonio (1983) cited studies that evaluated the level of unemployment among men who assaulted a partner. IDEQEQ Income referred to the defendant’s earnings from legal employment. This variable included the amount of the defendant's weekly earnings from his current employment. The defendant estimated his weekly earnings and this amount was recorded in a ratio scale. Wham The defendant's ethnicity referred to his self- classification in the presentence investigation questionnaire or to the classification in the police report. The choices for this category were as follows: Latin American/Chicano, Caucasian/white, or African- American/black. In exceptional cases the race information was obtained from the police report. Each ethnic category was assigned a nominal value for comparison purposes. The category of chicano received a value of zero, white received a value of one, and black received a value of two. 2‘3 ' c» V13. lRVC V 561% H: x A .obi \ 9: mi N N, N. 310 83 Miczosyetee Level This section of the instrument described the variables regarding a defendant’s relationship with a female partner. This section collected information on whether or not the victim and the assailant were living together at the time of the assault, relationship between assailant and victim, victim’s ethnicity, victim’s age, assaultive incident, involvement of substances in incident, and relationship between the ethnicity of the assailant and the ethnicity of the victim. V'c d s ' t ivin o ethe The defendant's living arrangement constituted an important category in the current research. Previous research has indicated that unmarried male-female couples who lived together were at higher risk of getting involved in an abusive situation than were married couples (Edleson 8 Tolman, 1992; Rosen, 1993; Sherman, 1992; Sherman et al., 1992; Pate 8 Hamilton, 1992). Other work stated that married couples were more likely to repeat an assault (Bernard 8 Bernard, 1984). This variable explored whether or not victims and assailants were living together at the time of the assault. The answer yes or no was given a nominal value in this category. The value was zero for not living together and one for those who were living together. 84 W This variable referred to what was the intimate relationship between the assailant and the victim. Previous research have assumed that the marital status reflects the victim of the assault. However this is not always the case, specially for separated men who assault a living girlfriend. This issue has not been explored before in the literature. This variable used a nominal classification. The categories were the following: ex-wife/ex-girlfriend, current wife, and current girlfriend. The category of ex-wife/ex- girlfriend had a nominal value of one, current wife a value of two, and the category of current girlfriend a value of three. V , . This variable referred to the ethnicity of the victim recorded in the police report. It used a nominal classification for the ethnicity of the victim. The categories were the following: Latin American/Chicana, Caucasian/white, or African-American/black. The values for these categories were the following: zero for Chicana, one for white, and two for black. V' ’ A Victim's age referred to the age of the victim at the time of the incident. The current study used the difference between the date of birth and the date of the incident to calculate the victim’s age at the time of the assault. The victims’ ages were organized into a ratio scale. EX '29} 85 've c' e The incident referred to the actions in the assault on a female partner, including the following: threatened her with a weapon, threw an object at her, held her down, pushed/shoved her, grabbed her suddenly, pulled her hair, bit her, slapped her, hit her with an object, punched her, kicked her, choked her, beat her up, or physically forced her to have sex. Each of these possible actions received a yes or no code in a nominal scale. Marshall (1992) developed a scale of severity of violence against women examining female college students, and female nonstudents. Female nonstudents rated physical violence acts more extremely than students. Thus, Marshall's weight scale of _.A‘,__~.~‘_ --.... "M physical harm was used for each act of violence identified “hum...“ ..— -—v in the current study. _Table 2presents the weight assigned to each violent act. The current study used in this variable the total weight for the assaultive incidents. W This category refers to whether or not victims and/or assailants were under the influence of alcohol and/or illegal substances at the time of the assault. The category received a yes or no answer for a nominal scale. Fitch and Papantonio (1983) found that, of men who abused a female partner, 59% reported abusing alcohol and 18% abused drugs. The answer or no received a nominal value of zero, while the affirmative answer received a value of one. assa; varia poss rino whit nine ESSE rer of on 86 3:91am This category pertained to the race/ethnicity of the assailant and the race/ethnicity of the victim. This variable was in a nominal scale to distinguish three possible race exchanges in the assaults. One category was minority assailant on minority victim and white assailant on white victim. The second category was white assailant on minority victim, and the third category was minority assailant on white victim. The first group in the category of same race/ethnic received a categorical value of one, while white assaulting a minority victim received a value of two. The third group of minority assaulting a white women received a value of three. Exos ev This section of the instrument contained variables regarding sanctions from the criminal justice system to offenses. Some of these sanctions pertained to the assault on a female partner by the subjects in the study. The main purpose of this section was to collect information on the following: number of prior arrests, type of last prior offense, number of days in jail for current assault on a partner, current payment of restitution to the partner victim, current payment of money to the court, number of days under probation supervision for current offense, and mandate to receive partner assault-abatement treatment. 87 Table 2 - Weight of Assaults According to Marshall Event Weight for Physical Harm Threatened her with a weapon 0.566 Threw an object at her 0.837 Held her down, pinning her in place 0.695 Pushed/shoved her 0.706 Grabbed her suddenly 0.718 Pulled her hair 0.765 Bit her 0.855 Slapped her 0.767 Hit her with an object 0.919 Punched her 0.936 Kicked her 0.939 Choked her 0.945 Beat her up 0.983 Physically forced her to have sex 0.818 88 N sts This section of the instrument examined the prior adult arrests of a defendant. The main goal of this section was to obtain information about the number of prior arrests. The variable of prior arrest has been considered in several studies of repetition of partner assault (Gondolf 8 Fisher, 1988; Hamberger 8 Hastings, 1990; Saunders, 1992; Sherman, 1992; Sherman 8 Berk, 1984; Sherman et al., 1992). Prior arrest history has been considered an important factor in studies of actuarial prediction with parolees (Burgess, 1928; Griffieth, 1985/1987; Warner, 1923). Furthermore, studies have found evidence that prior assaults were the best predictor of future assaults (Monahan, 1981). Thus, the number of prior arrests refers to those reported in the record of the defendant. This number reflected the findings from the probation officer in the Law Enforcement Information Network (LEIN) throughout a computer system. The current study used a ratio measure for the actual number of prior adult arrests before the assault for the current study. Type 9; Lag; Offense This variable referred to the type of offense at last arrest. The main goal of this section was to obtain information about the type of offense at last arrest. A nominal measure identified the types of prior offenses that the defendant had committed at his last arrest before the assault on female partner. Previous offenses were organized 89 into three main categories of alcohol/drugs, offenses against another person, and property offenses. Non prior offenses received a categorical value of zero, value one was assigned to alcohol/drugs, value two was assigned to offenses against another person, and value three was assigned to property offenses. E . J 'J This referred to the number of days that a defendant spent in jail for the instant offense. The data did not allow for specification if the number of days in jail included or excluded those after sentencing. Certainly the number of days in jail before the sentencing day were credited to the jail sentence. Thus, technically this was the number of days in jail that a defendant spent for the instant offense. Furthermore, it was not possible to separate the days in jail prior to sentence. The number of days in jail for the current analysis was in a ratio scale. e ' 'on ' 'm Restitution to the victim referred to whether or not the assailant paid money to the victim as restitution for destroyed property or for medical expenses as a consequence of the assault.' The variable was in a nominal category of paid restitution yes or no. The answers of those who did not pay restitution received a value of zero, while of those who paid received a value of one. 90 W Money paid to court pertained to the total amount of money that the defendant paid in fines, court costs, and state justice fees. The instrument recorded the amount that defendants paid in these three categories. This amount was in a ratio scale from zero to an infinite number of dollars. Qeys e: firebation Supervisien Probation supervision referred to the length of time defendants remained under probation supervision. The difference between date of initial probation and the date of last probation supervision provided the number of days under probation supervision. The number of probation supervision was in a ratio scale of zero to infinite number of days under supervision. This variable referred to the assignment of the defendant to a domestic violence prevention program. The coding for this variable was in a nominal scale of yes or no to such treatment assignment. The cases not assigned to treatment received a value of zero, while those assigned received a value of one. Sueseggege Ceniecg gith ggimigel Justice §ys§em f9; an M This section concerned a defendant's subsequent return ‘Contact with a criminal justice system for an assault f whether or not these defendants were rearrested for an assault. ow a s' Power analysis was calculated for correlation <=<>efficient. Power was calculated for multiple regression ‘“'i:th.k independent variables, with fewer than 24 independent 97 variables. Power was calculated assuming a small effect size (f=.30). Thus, for a multiple regression and correlation analysis the effect size will be 184 at a power of .30 at af=.05 (Cohen 8 Cohen, 1975). CHAPTER I I I RESULTS With the purpose of maintaining consistency throughout the current document, the/first set of results will be organized under the levelsof the ecological model discussed in chapter one. These findings will present descriptive statistics about the variables based on the ecological model. The main components of the ecological model are the ontogenetic or individual level, the microsystem or family level, the exosystem or social structural level, and the macrosystem or sociocultural level. The/second set of findings will present the results of / ., logisticwregression analyses according to the research _ . ...fiw-__, - __ r..__, -_"__v (n‘ hypotheses of the current study. Descriptive Statistics As result of the limited knowledge regarding the .EOD mmnm. Neda. Home. «mom. svmo.n ««««.m5m .089 name menu. omma. ~mmo.u omao. owao. uuzoo on name mono: mace. mnea.u >6«H.n mead. mend. awuofi> ou «*«.mmm Hmo~.n msvo.u oboe. nmno.u come. dawn as name «moo. semo.u «moo. soma.u maeo.u suuomoum mmcmuuo puma smna.u mmoo.u mmno. sonfl.u mmmo. comumm mmcmuuo puma ammo. uaaa.u mnmo.u memo. mono. HoeooHn mmcmuuo pawn n4na.: smeo.u «mad. mmeo.u ummn. ummuum gowns umnssz mafia. mood. mmso. memo. homo.u mocmumuuwo census m~oo.u mmoa.u maao.u mmoo.a moms. moccaumnsm mmoo.n snaa.n mnoo. no~6.u meaa.u promos acmofiocH ammo.n mmaa. «mac. Homm. mums. mom m.anuofi> memo.u smem. ammo. Hena.u ammo.u amuufl> xomHm smbo. mmva.u unmo.u «nod. omoa. sfiuofi> muses oeeo.u mHHo.u mafia. mmmc.u some. «4.886 a .uofi> .Hum -~o.u 6666.: oaon. naaa.u mmea. umeummou mcfi>fiq mmwa. maao.u omoo. «HmH. «Hoe. «.umo cmfimnoamo mono.u nose. shoe. omma.u mmno.n «.umo somam «man. «mud. mmma.u bmsu. mnmo. «soocn oooo.H mmma. emco.n «How. mmso. acmsmoaesm oooo.a memo.u mood. ommo. responses oooo.a ~5n4.u mono.u cassava oooo.H mmmm. mumumdmm oooo.a «.umo mo mm< «.umo ucofihonsm cofiuuoscm cognac: muouonom no 004 mammaun< mafiafibuwamm u m manna 111 chho.l H¢hO.l mHHo. moao. mmHo. «¢t¢%.H# .Ofl> .EOD mmmo. mmno. mmom. mmmo.u mooo. «444.com .oum mama oomo. omoo.u Hmmm. mmmm.u momm. unsoo ou ammo mono: omoo.u mmoo. memo. oomo.u ommo. sfiuom> 0» «««.mmm mmao.u omom. momo.u omoo. omom.u ammo cm mama oomo.u momm.u mmmo. ommo. ommm.u muuomoud mmcmuuo puma mmmo. ommo. ommm.u momm. mmmm.n common mmcmuuo puma mmmo. oomm. moom. moom.u oomm. accommn mmcmuuo puma mmmo.u oooo. oomo.u mooo. momm.u mummuua named umessz mmmo. ammo.u mmmm. mmmo.n oomo. mocmummumo cannon oooo. mmmm. mmmo. mmom.n mmmm.u mocuumnsm mmoo. ommm. ommo. mono. omom. promos unmomocm omoo. omom. momm. ommo.n oomo.u amuom> no «me oomo. omoo. mmom.u mmmm. momm.u amu6m> roman mmom.u mmmo.u memo. ammo.n mmmo. amuom> mums: oooo.m oomo. memo.n ommo. mmmo.n 44.noo a .p0a> .mmm oooo.m mmmm.n «mom. mooo.u umeuomou mcm>mq oooo.m ommm.u omom. «.mmo cummmosno oooo.m mmmm.u «.umo momma oooo.m msoocm «..umo a umeummou «.uoo ..omo .u0a> .Hom osm>mq commmosuu xomam oaoucH .«o.m=oo» m m~nms 112 memo. mmmo. meH.l maho.l HHmo. «istc.uu .Ofl> .EOD omoo.n ommo. mmmo. mmmm.u mmmm. 4«««.e5m .oue mama omom.u mmoo.u mmmm. mmmo.u momo. unsoo cu ammo mono: mmoo. mmmo. ommo.u mmom. ommo.a amuom> 0» ¢««.mmm Homo. mmmm. ommo.u oomo. ammo. HmMm cm mama oooo.m oomo.u mmmm.u mmmo. mama. muummoum omcmumo puma oooo.m mmmm.u oomm. momo.u common anemone puma oooo.H mmmo. ommo.u moeooam mmcmmuo puma oooo.H «moo. mummuuo “Oman umnasz oooo.m muemumuumo omceum auuomoum common Honooau oncouuo omcmuuo oncomuo mumouum mucououuwc anon anon puma momma amass: Cancun oomo.: momo.- mooo. ommo. mmoo.n «««««.us .om> .aoo mmmm.n memo.u mooo.n mmoo. omoo.u «444.com .ouo mama mmmo.u mmoo. oomo.n mmoo.u mmoo. assoc cu ommm mono: ooom. omom. memo. mmom.u mmmo. amuom> op «««.mom ommm. oomo.n mmmm. mono. omoo.u amen cm ammo mmoo. mmmo.u mmmo.u ommo. mmoo.u muuocoud mmcmumo yuan ommo.n mmoo.n mmoo. mmmm. mmom.u common mmcmuuo ammo oomm. mmmo. mmmo. omam.u mmom. moeoomm mmcmuuo ammo ommm. momo.: mmom. mono. mono.u mummuum momma nonssz oomo.u ommo. mono. momm. mmoo.u oocmumuumo umeeum oooo.m memo. ommm. mmmm.u mmmo. manna mocmumnsm oooo.m mmoa.n ommo. momo.n promos ucoomocu oooo.m mmmo.u mmoo. amuom> mo om< oooo.m mmmm.u smuofi> momma oooo.~ Emuoa> oumn3 moans Davao; amu0m> amu0«> amu0m> mocnumncm ucocwosH MO 054 Roman mums: ..o.ueooc m manna 113 oooc.d ««¢c«.Hu .oH> .500 #Cflfiflflflhfl QOCTHOH> OHHGQSOD fl ¢¥%#%.HU .OH> .EOD cammw>uomsm coauwnoum hops: muse u «see .msm .ohm mama cowuflumumom u «¢«.m0m unaccommo can amuOm> coo3uon mmnmCOHHMHom n o«.uoc a .uom> .Hom Nwd hmwn. ONnH. NHmo. oooo.H ONhn. mmeo.l oooo.H mbvo. oooo.H {tfltomfim “H500 OH EfiUUfl> OH .oum mama swam >mco: «oa.m0m unmvcwumo fl t.uwa n mmuoo no z «who.l «««««.Hu .OH> .EOD mwnm.l ««*¢.m5m .OHQ whoa omoa.| unsoo on name xoco: «Aha. Equ0w> 0» 84*.m0m oooo.H Hash ca name amen :m mama ..o.u:ooo m omens 114 Discriminant Analysis The current researcher conducted a discriminant analysis in order to determine which variable would ‘ -M~—.“_, differentiate the court sanctions of assailants. Generally, discriminant analysis is used toassign an observation to one of several different groups. This process identifies the variable, or combination of variables, that are good predictors of the observation. This procedure is preferred when the dependent variable has three or more categories. Additionally, the Wilks' lambda was used as part of this m procedure to determine the measure of power of all the '- variables to discriminate between groups. Wilks' lambda is an inverse measure; the larger it is, the less discriminating information it provides. The statistically significant level of the Wilks' lambda is obtained by its transformation on a chi-square test. Table 6 presents the results of the means of the variables used in the discriminant analysis. Thus, the dependent variable was the court sanction, and the independent variables were those related to the characteristics of the defendant. These independent variables were: defendant’s age, marital status, relationship between victim and defendant, education, ethnicity, employment status, income, total weight in current assault, number of prior arrests, and type of last offense. 115 Table 6 - Group Means for Court Sanctions Serious- ness of Defendant's eentence age Separate narried Education Nothing 30.64 .43 .36 .57 CJ only 30.42 .37 .20 .64 CJ 8 treat 31.60 .48 .20 .48 Total 30.86 .41 .21 .58 Serious- ness of sentence we Ireeme ____Caucasian glee}; nothing .79 212.71 .29 .64 Cj only .55 136.54 .31 .57 Cj 8 treat .80 161.32 .45 .37 Total .66 151.25 .36 .51 SERIOUS- REL. BET. NUMBER OF LAST NESS OF VICT. 8 PRIOR OFFENSE em SS ** W LESTS 239m NOTHING .64 2.00 3.14 .21 CJ ONLY .77 1.51 3.47 .34 CJ 8 TREAT .74 1.54 2.49 .32 Total .75 1.56 3.09 .32 SERIOUS- LAST LAST NESS OF OFFENSE OFFENSE fiEflIEEQE EEB§QH ALQQEQL NOTHING .36 .14 CJ ONLY .26 .23 CJ 8 TREAT .26 .25 Total .27 .23 116 Table 6 (cont'd). Group standard deviations SERIOUS- NESS OF Defendant's §ENTENQE AQE SEPARATE MARRIED EDUCATION NOTHING 8.76 .51 .50 .51 CJ ONLY 8.58 .49 .41 .48 CJ 8 TREAT 7.85 .50 .40 .50 Total 8.31 .49 .41 .50 SERIOUS- NESS OF SENTENCE EMPLOYMENT INCOME CAUCASIAN BLACK NOTHING .43 206.90 .47 .50 CJ ONLY .50 186.28 .47 .50 CJ 8 TREAT .40 210.38 .50 .49 Total .48 196.81 .48 .50 SERIOUS- REL. BET. NUMBER LAST NESS OF VICT. 8 PRIOR OFFENSE SENTENCE SS LANT** INCIDENT ARRESTS PROPERTY NOTHING .50 .78 3.06 .43 CJ ONLY .43 .64 3.86 .48 CJ 8 TREAT .44 .82 2.53 .47 Total .44 .72 3.39 .47 SERIOUS- LAST LAST NESS OF OFFENSE OFFENSE S CE (253593 ALCOHOL NOTHING .50 .36 CJ ONLY .44 .43 CJ 8 TREAT .44 .43 Total .45 .42 117 Table 7 presents the Wilks' lambda F ratio. The court sanction was not very predictable. The only two predictors in a univariate sense were 1) being African American and 2) being employed. African Americans had the highest mean in the category of no criminal justice sanction. This was different from the variable of employment, which had the highest mean in the category of criminal justice sanction and treatment. These two variables produced a small degree of discrimination, as is indicated in the Wilks’ lambda . (.9574) for blacks and (.9348) for employment. Thus observing these Wilks’ lambda, it is fair to state that the variables did not provide sufficient variation between the three groups to differentiate them. The high values of Wilks' lambda suggest that there is no discriminatory power in these variables. Thus, the three groups do not differ significantly from each other, based on these variables. 118 Table 7 - Wilks Lambda Statistics Wilks’ Lambda (U-statistic) and univariate F-ratio with 2 and 179 degrees of freedom Variable Wilks’ Lambda F Significance Defendant's AGE 1.00 .41 .67 SEPARATE .99 .96 .38 MARRIED .99 .92 .40 EDUCATION .98 2.21 .11 EMPLOYMENT .94 6.24 .00 INCOME .99 1.06 .35 CAUCASIAN .98 1.77 .17 BLACK .96 3.98 .02 REL. BET. ASSAIL.** .99 .52 .60 INCIDENT .97 2.83 .06 NUMBER PRIOR ARRESTS .98 1.65 .19 L. O. PROPERTY*** 1.00 .44 .65 L. 0. PERSON*** 1.00 .29 .75 L. O. ALCOHOL*** 1.00 .35 .71 119 Table 7 (cont'd). Discriminant Analysis On groups defined by SERSENTN 0=Nothing, 1=CJ Only, 2=CJ + Treat Analysis number 1 Direct method: all variables passing the tolerance test are entered. Minimum tolerance level.................. .00100 Canonical Discriminant Functions Maximum number of functions.............. 2 Minimum cumulative percent of variance... 100.00 Maximum significance of Wilks' Lambda.... 1.0000 Prior probability for each group is .33333 Canonical Discriminant Functions Pct of Cum Canonical After Wilks' Fcn Eigenvalue Variance Pct Corr Fcn Lambda Chi-square df Sig : 0 .774765 44.021 28 .0277 1* .1876 68.36 68.36 .3975 : 1 .920113 14.362 13 .3488 2* .0868 31.64 100.00 .2826 : . * Marks the 2 canonical discriminant functions remaining 111 the analysis. 120 Table 7 (cont'd). Standardized canonical discriminant function coefficients Func 1 Func 2 Defendant’s AGE .25 -.11 SEPARATE .14 .22 MARRIED .15 .56 EDUCATION -.46 .00 EMPLOYMENT .72 .05 INCOME -.13 .30 CAUCASIAN -.10 .00 BLACK -.48 .45 REL. BET. ASSAIL.** -.17 -.35 INCIDENT .13 .57 NUMBER PRIOR ARRESTS -.48 .20 L. O.*** PROPERTY .12 -.43 L. O.*** PERSON .38 -.19 L. O.*** ALCOHOL .02 -.34 Structure matrix: Pooled within-groups correlations between discriminating variables and canonical discriminant functions (Variables ordered by size of correlation within function) Func 1 Func 2 EMPLOYMENT .60* .12 BLACK -.41* .38 EDUCATION -.36* .08 NUMBER PRIOR ARRESTS -.31* .09 CAUCASIAN .29* -.22 SEPARATE .24* -.03 Defendant’s AGE .15* -.06 INCIDENT .11 .58* MARRIED .03 .34* INCOME .17 .27* L. O.*** PROPERTY -.06 -.22* REL. BET. ASSAIL.** -.10 -.22* L. O.*** ALCOHOL .01 -.21* L. O.*** PERSON .02 .19* * denotes largest absolute correlation between each variable and any discriminant function. ** denotes relationship between victim and assailant *** denotes last offense 121 Table 7 (cont'd). Unstandardized canonical discriminant function coefficients Func 1 Func 2 Defendant's AGE .03 -.01 SEPARATE .29 .44 MARRIED .37 1.37 EDUCATION -.93 5.54E-03 EMPLOYMENT 1.55 .10 INCOME -6.33E-04 1.50E-03 CAUCASIAN -.22 3.17E-03 BLACK -.98 .91 REL. BET. ASSAIL.** -.39 -.79 INCIDENT .18 .79 NUMBER PRIOR ARRESTS -.14 .06 L. O.***PROPERTY .26 -.91 L. O.***PERSON .84 -.43 L. O.***ALCOHOL .04 -.80 (Constant) -.83 -1.07 Canonical discriminant functions evaluated at group means (group centroids) Group no 1 Enne___; NOTHING .24867 .99804 CJ ONLY -.37089 -.04279 CJ 8 TREAT .53416 -.14715 Test of Equality of Group Covariance Matrices Using Box’s M The ranks and natural logarithms of determinants printed are those of the group covariance matrices. M _e_l$__L_cr_L_n_R n o etermina t NOTHING <14 (Too few cases to be non-singular) CJ ONLY‘ 14 -2.997379 CJ 8 TREAT 14 -3.467532 Pooled within-groups covariance matrix 14 -2.228201 Since some covariance matrices are singular, the usual procedure will not work. The non-singular groups will be tested against their own pooled within-groups covariance matrix. The log of its determinant is -1.17263 122 Table 7 (cont'd). Box's M Approximate F Degrees of freedom Significance 332.99833 2.86966 105, 58736.7 .0000 Classification results - No. of Predicted Group Membership Actual Group Cases 0 1 2 Group: NOTHING 14 9 3 2 64.3% 21.4% 14.3% Group: CJ ONLY 103 22 54 27 21.4% 52.4% 26.2% Group: CJ 8 TREAT 65 18 13 34 27.7% 20.0% 52.3% Percent of "grouped" cases correctly classified: 53.30% Classification processing summary 182 (Unweighted) cases were processed. 0 cases were excluded for missing or out—of-range group codes. 0 cases had at least one missing discriminating variable. 182 (Unweighted) cases were used for printed output. 123 Logistic Regression Analysis One of the purposes of using logistic regression is to determine the probability of whether or not an event will occur. The second purpose for using this kind of statistical analysis is to identify the independent variables that best predict an event. Logistic regression was preferable to multiple regression or discriminant analysis for this study because the dependent variable contained only two values (0 or 1), and the independent variables had different forms such as dichotomous and continuous. The mathematical formula for the logistic function is called f(z), and is expressed by 1 over 1 plus e to the minus 2 (Hosmer 8 Lemeshow, 1989; Norusis, 1992). This function is represented in the following expression: 1 f(z)= 1+e" / Where e represents the base of the natural logarithms, ”*’ -.-.-..._ ......»- approximately 2.718 (Norusis, 1992). Here 2 is a linear combination adding X, times beta (6) , plus alpha (a) . This is represented in the following formula, in which the X,’s represent the independent variables of interest and alpha (a) and beta (B) are constant terms that represent unknown parameters: z=a+fllxl+flzxa+. . . 3'3ka 124 In logistic regression, the unknown parameters are replaced by estimates based on the data collected, using the maximum- (—§“‘h —71 m “...... likelihood method. This means that the coefficients that make the observed results most likely are utilized. Therefore, after an estimate of the parameters is made, the values of the X,’s are substituted for the observed values to obtain the probability of the particular dependent variable, in this case, the probability that an event will occur. The predictive technique of logistic regression was used to test the hypotheses identified in chapter one. The analysis for each hypothesis will be presented next. The analysis of the first hypothesis is labeled "Interventions," the analysis of the second is labeled "Court experiences," the analysis of the third is labeled "Involvement of Substances," and the analysis of the fourth is labeled "Ecological Model." Mm Hypothesis one stated that the intervention selected by a judge (e.g., jail, substance abuse treatment, partner- abuse treatment, and probation) depends on the personal characteristics of the assailant, specifically age, marital status, education, employment, weekly income, ethnicity, and relationship with the victim. Thus, the current researcher argued that personal characteristics were the best predictors of the intervention selected for these assailants. To evaluate this hypothesis, four dependent 125 variables related to the sentence were examined using logistic regression: sentenced to jail, mandated to a substance- abuse program, mandated to partner abuse- abatement treatment, and sentenced to probation. Each dependent variable was then examined using the same nine independent variables: defendant's age, marital status (i.e., the actual variables were: separated, married), education, employment status, weekly income, ethnicity (i.e., the actual variables were: black, white), and relationship with the victim. The variables used as controls were number prior arrests, last offense related to property, last offense related to a person, last offense related to alcohol/drugs, and seriousness of current offense. The independent variables were entered in blocks with the control variables first. m Table 8 presents a summary of the logistic regression of the dependent variable, sentenced to jail, and the independent variables. The current model's performance can be evaluated on the basis of how well it classified the observed data (Norusis, 1992). This is evaluated in the likelihood of the observed results. The likelihood is a small number; thus conventionally it would be stated as -2 times the Log of the Likelihood (-2LL).( The ideal model a”... .... ea..- .1-.._.__._..... 7.. would have a very small value for 72LL} In the current model the -2LL was 145.88; this was a smaller number than a model which removed non significant variables. However, 126 this value was still too high, thus indicating that the model did not classify the data very well. The goodness-of- fit statistic with all variables is 206.49; which is displayed under the -2LL value. The model chi-square for the current model was 37.92 (df=14) p<0.0005. The classification table indicates how well the model fits. This table indicates that the numbers in the row of observed cases with an answer of "no jail" totaled 37 cases. From this number, 9 defendants (24%) who had not spent time in jail were correctly identified, based on their personal characteristics, as assailants who would not receive jail sentences. The numbers in the row of observed cases with an answer of yes totaled 145 cases. Among these, 141 defendants (97%) who had spent time in jail were correctly identified, based on their personal characteristics, as assailants who would receive jail sentences. The off- diagonal numbers indicate that 32 cases were incorrectly classified, 28 defendants who did not receive jail sentences, and four defendants who did receive jail sentences. Overall, 82% of the cases were correctly classified. Although this is a high percentage of cases classified accurately, ideal results would have presented fewer cases in the cells of misclassification and more cases in the cells of correct classification. This did not happen for the defendants who were predicted to receive a jail sentence. 127 The lower section of Table 8 indicates the statistics of the variables in the equation. The first column identifies the variables used in the current logistic regression, and the second column indicates the values of beta (B) for each variable in the prediction. Beta is-the estimated coefficient value from the logistic regression model that predicts a jail sentence for each particular variable that is used in the probability equation. Beta indicates the amount of log odds in the dependent variable for a one-unit change in the independent variable if the values of the other independent variables do not change. The next column is the coefficient to standard error ratio. This is one of the tools used to test the statistical significance of the estimated parameters in the equation. The standard error is used as the denominator of the estimated coefficient to calculate the Wald statistic. The Wald column is the maximum likelihood estimation of HM chi-square_§tatistics. The Wald statistic examines a coefficient with a value of zero. The following column indicates the degrees of freedom, which is k-l, where k is the number of estimators in the model. The probability (P) column displays the significance level of the Wald statistics different from zero. The column of (R) statistic presents the partial correlation between the dependent variable and each independent variable. In other words "R" provides the contribution of each independent variable predicting the dependent variable. The value of ”R" ranges 128 from -1 to +1. A value closer to either -1 or +1 represents a better fitting model, and a value closer to zero indicates a poor fit. The Exp (B) column represents the odds of increase when the value for the independent variable changes from zero to one. Thus, the Wald significant values at 0.05 for the coefficient of the variables indicated only one variable with a coefficient different from zero. Employed defendants were less likely than unemployed defendants to receive a jail sentence (Odds Ratio (OR) = 0.07). These odds ratio were in the 95% confidence interval (95% CI) of .01 to 0.34. Furthermore, the values of the ”R" column range from - .22 to +.08. These values do not approach -1 or +1. Most variables had zero correlation, to four decimal places, with the dependent variable. These values indicate that the model did not fit the data very well. Norusis (1992) discussed the prevention of Type II error in logistic regression models. Type II error refers to the failure to reject a null hypothesis when it is false. Thus, the model from Table 8 was compared to a model without high coefficient variables. The findings in the latter model provided a chi-square significant at P6:°°9$,idf=9)r predicted a similar number of positive cases (N=141), but it increased the number of misclassifications (N=36). This second model produced an overall prediction of 80 percent, a smaller percentage than that achieved by the first model. 129 Table 8 - Logistic Regression for Sentenced to Jail -2 Log Likelihood 145-879, Goodness of Fit (206.493 Chi-Square df Significance Model Chi-Square 37.919 14 ..0005 Predicted Observed No Yes Percent correct NO 9 28 24.32% Yes 4 141 97.24% Overall 82.42% Variables in the equation Variable Beta S.E. wald df P R Exp(B) Prior arrests .08 .09 .81 1 .37 .00 1.08 L.O. property .16 .33 .24 1 .63 .oo 1.18 L.O. person .14 .36 .15 1 .70 .00 1.15 L.O. alcohol .46 .36 1.61 l .20 .00 1.58 Incident wt .36 .31 1.38 1 .24 .00 1.44 Age -.05 .03 2.13 1 .14 -.03 .95 Separated .54 .30 3.13 1 .08 .08 1.71 Married -.03 .29 .01 1 .92 .00 .97 Education -.79 .49 2.58 l .11 -.O6 .46 Employment -2.66 .81 10.82 1 .00 -.22 .07 Income -.00 .00 .16 1 .69 .00 1.00 Black 1.33 .72 3.36 l .07 .09 3.76 Caucasian .20 .35 .31 1 .58 .00 1.22 Rel. w/ vic. -.68 .54 1.58 1 .21 .00 .51 Constant. 5.23 1.62 10.48 :1 .00 130 Furthermore, the -2LL of this model was 150.32, a higher value than in the first model. The second model did not contain the variables with low "R" values. When these variables were removed from the model, it did not provide higher explanatory value, and the number of cases correctly classified decreased, while the number of misclassifications increased. Additionally, the - 2LL values for the models from which these variables were removed were higher than the one obtained in the original model, thus indicating that they were less adequate. - ......r....--.4-.. —~—..»—--.~..~..- .m-«fl‘ Furthermore, the comparison of -2LL indicated that the models were not significantly different at alpha .05 (-4.44; df=5). Therefore the null hypothesis is not rejected. In summary, the overall correct classification using this model was 82 percent. It presented several problems, including a high -2LL value. Furthermore, only one variable out of 14 had a significant value in the Wald statistics although the values of the "R" statistics were close to zero. However, the first model performed better than the second to which it was compared. The control variables might have contributed to creating a better model, but none of these variables indicated a relationship with the sentence to jail for assault on a partner. M e S a e us a ment Table 9 presents a summary of the logistic regression of the dependent variable indicated that the assailant was ordered to attend substance-abuse treatment. The -2LL value 131 of this model was 160.131, suggesting that the model did not classify the data well. The goodness-of-fit statistic with all variables was 172.196, which was displayed under the - 2LL value. The model chi-square was 38.88 (df=15) p<0.0007. The classification table indicates that the sum of the row of observed cases with no substance-abuse treatment was 139. Among these cases, 130 (94%) were correctly predicted as not receiving a sentence for substance-abuse treatment. These predictions were based on their personal characteristics, whether or not drug abuse was involved in the incident, and prior criminal history. The sum of the cells of observed cases of substance-abuse treatment was 43. Among these cases, 15 (35%) were correctly predicted, based on their personal characteristics, as receiving a sentence for substance-abuse treatment. The off-diagonal numbers indicate that 37 cases, 9 men who did not receive substance- abuse treatment and 28 men who did, were misclassified. Overall, 80% of the cases were correctly classified. Although this is a high percentage, the number of cases mandated for substance-abuse treatment that could not be predicted was higher than it would be in an ideal model. In an ideal model the number of positive predictions with positive outcomes should be higher than those with negative prediction with negative outcomes. Two independent variables had Wald significant values less than 0.05 for the coefficients that appeared to be different from zero. The first was whether or not the 132 assailant was employed, and the second was whether or not substances were involved in the assaultive incident. Table 9 shows that the partial correlation of each of these two variables with the dependent variable "R" ranges from -.09 to .24. These values indicate that the model did not fit the data very well. Employed assailants were four times more likely to receive a mandate for substance-abuse treatment than unemployed assailants (OR=3.8; 95% CI:1.34, 10.66). In cases where either the assailant or the victim were using substances, assailants were twice as likely to receive a sentence for substance-abuse treatment (OR=2.3; 95% CI:1.48, 3.62). In order to avoid Type II error, the current analysis was conducted removing high coefficient variables that were not significant. The findings in this latter model indicated a chi-square significant at p<.00001 (df=7), but had a smaller number of positive predictions (N=13). This second model also achieved an overall prediction of 80 percent; however the -2LL was 165.655, a higher value than in the first model. The comparison of -2LL indicated that the models were not significantly different at alpha .05 (- 5.5; df=8). Therefore the null hypothesis is not rejected. In summary, the assailants more likely to receive a mandate for substance-abuse treatment were those who were employed, and those who were involved with substances at the time of the assault. The overall correct classification achieved by the model was 80 percent. However, the model 133 Table 9 - Logistic Regression for Mandated to Substance- Abuse Treatment -2 Log Likelihood Goodness of Fit Model Chi-Square Observed No Yes Variables in the equation Variable Prior arrests L.O. property L.O. person L.O. alcohol Incident wt Age Separated Married Education Employment Income Black Caucasian Rel. w/ vic. Substances Constant 160.131 172.196 Chi-Square df Significance 38.880 15 .0007 Predicted No Yes Percent correct 130 9 93.53% 28 15 34.88% Overall 79.67% Beta S.E. Wald ‘ df P R Exp(B) -.16 .08 3.56 1 .06 —.09 .85 -.13 .33 .14 1 .71 .00 .88 .32 .35 .83 1 .36 .00 1.38 .12 .38 .13 1 .72 .00 1.13 -.18 .28 .40 1 .53 .00 .84 .01 .03 .20 1 .65 .00 1.01 .14 .27 .28 1 .59 .00 1.15 .18 .29 .40 1 .53 .00 1.20 -.57 .43 1.75 l .19 .00 .57 1.33 .53 6.30 1 .01 .15 3.78 -.00 .00 .32 l .57 .00 1.00 -087 062 1095 1 .16 .00 042 -.02 .31 .01 l .94 .00 .98 .43 .52 .69 1 .41 .00 1.53 .84 .23 13.64 1 .00 .24 2.32 -l.16 1.27 .83 l .36 134 had several limitations: it had a high -2LL value, only two variables of 15 had significant value for the Wald statistics, and the values of "R" statistics were close to zero. The model that included thecontrol variables related to crime history was better than the second model. However, none of the control variables demonstrated a significant relationship with the dependent variable. Thus the effect seemed to be in the correlation between the variables. t fo As ault Ab t me t e The model to predict defendants receiving an assignment to partner assault-abatement treatment did not produce significant results. The model's -2LL value was 135.459. This high value suggested that it did not classify the data well. The model's chi-square was 17.36 (df=14) p=.24 indicating that the use of the independent variables did not produce an exceptional model. genfeneeg no finenetion Table 10 presents a summary of the logistic regression of the dependent variable, sentenced to probation. Thus in the current model the -2LL was 182.871, a smaller value than in the model that excluded variables with small "R" values. However, this number was high enough to suggest that the model did not classify the data well. The goodness-of-fit statistic with all variables was 166.46. This is displayed under the -2LL value. The current model’s chi—square was 50.62 (df=14) p<0.00001. 135 The classification table indicates that the sum of the numbers in the row of observed cases with an answer of no to the question concerning sentence to probation was 62. Of this number, 28 men (45%) who were not sentenced to probation were correctly predicted, on the basis of their personal characteristics, not to receive probation. The sum of the numbers in the row of observed cases with an answer of yes to this question was 120. Among these cases, a sentence of probation was correctly predicted for 105 men (88%) based on their personal characteristics. The off- diagonal numbers indicate that 49 cases were misclassified: probation was predicted for 34 men who were not sentenced to probation, and 15 men for whom a probation sentence was not predicted were sentenced to probation. Overall, 73% of the cases were correctly classified. Although this is a high percentage of cases classified accurately, ideal results would have presented fewer cases in the cells of misclassification and more cases in the cells of correct classification. For instance, the 55% of defendants predicted to be under probation supervision but were not, represents a great distance from the ideal model. The strongest predictors for individuals with probation sentence were the number of prior arrest, whether or not the last offense was property related, whether or not the last offense was related to a person, whether or not the defendant was employed, and whether or not was black. Defendants with fewer prior arrests were more likely than 136 defendants with numerous prior arrests to be under probation supervision (OR=0.70; 95% CI:0.59, 0.82). Defendants whose last offense was property related were two times more likely than their counterpart to be under probation supervision (OR=1.93; 95% CI:1.04, 3.56). Defendants whose last offense was related to a person were two times more likely than their counterpart to be under probation supervision (OR=2.27; 95% CI:1.14, 4.52). Employed defendants were three times more likely than unemployed defendants to be under probation supervision (OR=3.2, 95% CI:1.44, 7.17). African American defendants were less likely than their counterpart to be under probation supervision (OR=0.27, 95% CI:0.08, 0.93). In order to avoid Type II error, another analysis was conducted removing high coefficient variables that were not significant. The findings in this latter model indicated a chi-square significant at p<.00001 (df=12), but had a smaller number of negative predictions (N=26). This second model achieved a smaller overall prediction with 72 percent; furthermore the -2LL was 183.929, a higher value than in the first model. In summary, the assailants more likely to receive probation supervision sentences were those who had fewer prior arrests, last offense was property related, last offense was related to a person, employed, and not of African American ethnicity. The overall correct classification resulting from the use of this model was 73 137 percent. However, the model has a number of problems, such as a high value for -2LL. A low percentage of correct classification. Only five variables of 14 had significant values in the Wald statistics, and the "R" statistics, ranging from -.27 to +.16, were very close to zero. This model was not much different to one with small number of low "R" values. Thus, the models were not significantly different at alpha .05 (-1.1; df=2). Therefore the null hypothesis is not rejected. We All three models of court sentences presented high values for -2LL, indicating that they did not classify the. datgwyerylwell. Although the three models presented a better -2LL than the models without high coefficient variables, their values remained distant from zero for over a hundred units. For the most part, the models correctly classified 73% to 82% of the sample thereby indicating that they were good models. However, in all the models, the "R" statistics were close to zero or zero to four decimal places. This indicated that the individual parameters did not have a relationship with the dependent variables and they had a poor distribution. In summary, the data did not fit the models selected, and the null hypothesis_should not be rejected. Given the values of the -2LL, it is safer to do not reject the null hypothesis. Thus, it can be concluded that the sentence selected by a judge did not depend on the personal 138 Table 10 - Logistic Regression for Probation Supervision -2 Log Likelihood Goodness of Fit Model Chi-Square Observed No Yes Variables in the equation Variable Prior arrests L.O. property L.O. person L.O. alcohol Incident wt Age Separated Married Education Employment Income Black Caucasian Rel. w/ vic. Constant 182.871 166.460 Chi-Square df Significance 50.624 14 .0000 Predicted No Yes Percent correct 28 34 45.16% 15 105 87.50% Overall 73.08% Beta S . E . Wald df P R Exp (B) -.36 .08 18.72 1 .00 -.27 .70 .66 .31 4.39 1 .04 .10 1.93 .82 .35 5.42 1 .02 .12 2.27 .33 .32 1.06 1 .30 .00 1.40 -.04 .26 .03 l .88 .00 .96 -.00 .03 .02 1 .89 .00 1.00 .29 .25 1.34 1 .25 .00 1.33 .16 .25 .42 1 .52 .00 1.18 .32 .40 .65 1 .42 .00 1.38 1.17 .41. 8.08 1 .00 .16 3.21 -.00 .00 .20 1 .65 .00 1.00 -1.32 .64 4.30 1 .04 -.10 .27 -.12 .32 .14 1 .71 .00 .89 .65 .46 2.00 1 .16 .00 1.91 2.26 1.21 3.50 1 .06 139 characteristics of the assailant. Prior criminal history and the intensity of the current assault did not predict the intervention selected by a judge. Therefore, none of these variables predicted the court intervention for men who had assaulted a female partner. Courf Experiences Hypothesis two stated that a partner assailant who has had more court experiences is more likely to repeat an assault as measured by police arrest. To evaluate this hypothesis, the dependent variable of rearrest for an assault within 36 months was examined with a logistic regression statistical technique. The independent variables included the assailant's number of prior arrests, and whether or not the assailant had committed property offense. This model did not produce significant results. The -2LL was 209.268 with very high standard errors for each independent variable. The chi-square of the -2LL for the current model was 0.7 (df=2) p=.70. Thus the model does not support the idea of rejecting the null hypothesis. Invglvemenf of Subsfances Hypothesis three stated that the involvement of substances in the first partner assault makes it more likely that the assailant will repeat an assault as measured by police arrest. To evaluate this hypothesis, the dependent variable, rearrest for an assault within 36 months, was examined with a logistic regression statistical technique. The independent variable was whether or not victims and/or 140 assailants were under the influence of alcohol and/or illegal substances at the time of the assault. This model did not fit the data very well. The -2LL was 208.835 and the model could not identify any observed case based on the predictive variable. The chi-square of the -2LL for the current model was 1.167 (df=3) p=.76. Thus the model did not support rejecting the null hypothesis. Therefore, there was no relationship between the abuse of substances in the first assaultive offense for which the police arrested an assailant and the second assault for which he was arrested. Ecologica; Model [/HyfiOthesis‘fOUr-stated that the inclusion of variables _..1—’ -‘ “ca“. considered under the ecological theory will provide better predictors of the characteristics of individuals who repeat an assault, as defined by police arrest, than other models. The dependent variable evaluated if within a 36-month period, were assailants rearrested for an assault. The independent variables were organized into three main groups. Group one consisted of characteristics before the event, group two consisted of characteristics of the event, and group three consisted of court sentence and disposition. This model did not fit the data very well. The -2 times the log of the likelihood (-2LL), was 179.01. The chi-square value for the model was 31 (df=26; p=.23), proving to be not significant. Therefore the model did not support rejecting the null hypothesis. Thus there was no relationship between 141 the ecological model and the second assault for which the defendant was arrested. §pnnepy pf Logiefic Regfession The current study used logistic regression to evaluate .____--— recidivism. The prediction of court intervention evaluated (.5 l the prediction of court interventions and prediction of the first hypothesis in the study, which stated that interventions selected by judges were dependent upon the personal characteristics of the assailants. The possible interventions included: sentenced to jail, mandated to a substance-abuse program, mandated to partner abuse-abatement treatment, and sentenced to probation. Nine of the independent variables were personal characteristics. Additionally, the analysis included five variables related to the criminal history of the defendants as control variables. All these variables were included in all the analyses. As this first hypothesis implied, judges have different interventions for partner assailants. Thus, the analysis of the first hypothesis included four analyses, based on the particular intervention. The first analysis concerned jail intervention. The current analysis revealed that the variables used contributed to an 80% classification of the cases. However, only one independent variable had a significant value--employment. The second analysis related to substance-abuse treatment. The variables used in this second analysis also contributed to an 80% classification of 142 the cases. Similar to the previous analysis, only two independent variables had significant value--employment and the substances involved in the incident. The third analysis relates to partner assault-abatement treatment. None of the variables used in this third analysis contributed to significant results. The fourth analysis pertained to sentence to probation. All the variables used in this model contributed to the correct classification of 73% of the cases. However, only five independent variables had significant values as individual predictors. These variables were: number of prior arrests, whether or not the last offense was property related, whether or not the last offense was related to a person, whether or not the defendant was employed, and whether or not he was black. Although the classification tables of the models for the first hypothesis showed that the models were able to classify between 73% to 82% of the cases, the overall -2 times the log of the likelihood (-2LL) values were too high. This was an indication that the models did not classify the data well. A good indication that the model classified the data well is when the -2LL value is close to zero. When the original models were analyzed without high coefficient variables that were not significant, the -2LL values were higher than that of the originals. The comparison between the first models and the alternative models suggested that the null hypothesis should not be rejected. 143 The second, third, and fourth hypotheses were related to the prediction of recidivism. None of the variables used in these analyses contributed to significant results. CHAPTER IV DISCUSSION As explained in chapter one, there are evident limitations in a study using the data recorded in a probation department to make predictions about partner assailants. Some of the literature on partner assailants described events before the cases came to court (e.g., police intervention), and other studies reported the characteristics of assailants in treatment programs. Both of these approaches failed to examine assailants who came to court for assault. Traditional studies of the prediction of recidivism in criminal justice organizations have ignored partner assailants. Thus, the current study is the first to study the prediction of recidivism of partner assailants using a sample obtained from a probation department. General Findings The general findings of this study indicated that the sample of assailants at the selected probation department tended to be young (about 30 years old). They tended to be employed, but most had not completed high school. Their average income was $151 a week, and most were of non- 144 145 European descent. These demographics suggest that men of non-European descent living in poor socioeconomic conditions are easy targets for police arrest for partner assault. It has been argued that partner assault originates in male socialization (Binder 8 Meeker, 1992). Therefore males in all socioeconomic and ethnic groups have the potential for partner assault (Dobash 8 Dobash, 1979). In the city selected for this study, the majority of the population was of European descent. However, the current research shows that those who were of non-European descent and living in poor socioeconomic conditions were more often confronted by the criminal justice system. Buzawa and Buzawa (1990) explained that African Americans, Latinos and those in lower socioeconomic groups tend to call the police more often than men of European descent and those who are more affluent. Findings related to the assailants' interactions with their victims indicated that these assailants were living with partners at time of the assault. Most of them assaulted partners of European descent and their victims were young (about 29 years old). Nineteen percent of victims were assaulted by assailants of non-European descent. Involvement of alcohol in the assaults was fairly frequent. These findings thus reflect cases in which assailants were somewhat similar in age, in which alcohol was a problem, and in which ethnicity differences were not extremely marked. 146 Findings related to the criminal justice contact revealed that these assailants had previous arrests (a mean of 3.1), and most of their last offenses were assault and battery or involved property. Assailants had spent an average of 14 days in jail for the current offense of partner assault. They were not required to pay restitution to the victim, and the average court fine was $117. Each assailant spent about ten months under probation supervision. Most of the assailants were not sent to treatment for domestic violence. The number of prior arrests indicated that these assailants were not strangers to the criminal justice system, and the type of offenses continued to reflect poverty. The sanctions also indicated that they were not punished with high fines, but were sentenced to more time in jail. While a description of a sample of partner assailants from a probation department was obtained for the current study, the results should be interpreted with caution. First, the current study examined records from only one probation department in a medium-sized city in Michigan; therefore the results may not apply to all probation departments in the United States. Table 11 presents a comparison of the current study with studies of police intervention and treatment programs. This comparison is disadvantageous because not all studies reported the same variables, nor did they use the same method of data collection, which could account for the 147 differences. However, the current study used a unique sample that resembled the findings in studies of police intervention more than the findings in studies of treatment programs. For instance, the ethnicity of the sample in the current study is similar to that of samples for three studies. The current study reported that an over representation of African American men came to the court system (51%). The studies of Dunford (1990), Pate and Hamilton (1992), and Sherman et al. (1992) also found this overrepresentation, with percentages closed or over the percent of caucasian men in their samples. The comparison with treatment programs did not reveal the same findings. Most of the studies of treatment programs reported that over 75% of their cases were caucasian. A second comparison of the current study with other studies concerns the percentage of cases involving the unemployed. Police studies reported that between 29 to 60 percent of their cases involved unemployed persons. However, the majority of the studies of treatment programs reported less than 20% of individuals unemployed. Only one study, Edleson and Syers (1990), reported 34% of unemployed clients aligning their findings with the current study and the police studies previously mentioned. A third comparison of the current study with other studies concerns average age. For the most part, all the studies agreed on the age of the assailants. In terms of education, most of treatment programs reported high 148 educational levels, while only the study of Hamberger and Hastings (1990) reported the lowest percentage of individuals with a high school education. This percentage was similar to the current study and closer to the studies of police intervention (Sherman 8 Berk, 1984, and Sherman et al., 1992). A significant comparison involves the percent of repeated assaults. In the current study, the percentage of assailants rearrested for an assault was similar to the percentage of individuals who repeated an assault in one police study (Sherman 8 Berk, 1984) and in two treatment- program studies (Hamberger 8 Hastings, 1990; Harrell, 1991). Most of the studies about treatment programs reported higher percentages of recidivism than police studies. For instance, Edleson and Grusznski (1988:1) reported that approximately 33% of the cases assessed repeated assault, while other studies reported higher percentages. The study of Edleson and Syers (1990) reported the highest percentage of repeated assault, indicating that 46% of the individuals in the sample repeated their assault. Thus, the sample of cases at a probation department in Michigan was similar to cases in studies of police intervention in other locations in the United States. Furthermore, the results for some variables showed similarities to results in studies of treatment programs conducted in Baltimore, Milwaukee, Minneapolis, and other cities. Therefore, although the current researcher rejects Table 11 Findings Comparison Studies of Treatmenr_£regl .Sire Ethnicity Bernard and Bernard (1984) DeMaris and Jackson (1987) Edleson and Grusznski (1988:1) Edleson and Grusznski (l988:2) Edleson and Grusznski (l988:3) Edleson and Syers (1990) Fitch and Papantonio (1933) Hamberger and Hastings (1990) Harrell (1991) Purdy and Nickle (1981) Saunders (1992) 149 Sample Assailant's Percent Average Queenie Age 46 male N.R. N.R. 31 assailants 53 male N.R. N.R. N.R. assailants 63 male 97% Cauc. 9.0 29 assailants 57 female survivors 86 male 86% Cauc. 16 32 assailants 6% Afr. Am. 57 female 2% Nat. Am. survivors 1% Asian Am. 159 male 88% Cauc. 16 34 assailants 3% Afr. Am. 121 female 3% Latino survivors 3% Nat. Am. 4% Mix. Races 283 male 74% Cauc. 34 32 assailants 11% Afr. Am. 4% Nat. Am 3% Latino 188 male N.R. N.R. N.R. assailants 106 male 86% Cauc. 20.0 31.1 assailants 12% Afr. Am. 2% Other 237 male N.R. 15.0 26-35 assailants 237 female survivors 170 male N.R. N.R. N.R. assailants 182 male 76% Cauc. N.R. 30.6 Table 11 Findings Comparison Studies of II§§§m§n§_22991 Bernard and Bernard (1984) DeMaris and Jackson (1987) Edleson and Grusznski (1988:1) Edleson and Grusznski (1988:2) Edleson and Grusznski (1988:3) Edleson and Syers (1990) Fitch and Papantonio (1983) Hamberger and Hastings (1990) Harrell (1991) Purdy and Nickle (1981) Saunders (1992) 149 Sample Assailant's Percent Average Size Fraulein Enema]... Age 46 male N.R. N.R. 31 assailants 53 male N.R. N.R. N.R. assailants 63 male 97% Cauc. 9.0 29 assailants 57 female survivors 86 male 86% Cauc. 16 32 assailants 6% Afr. Am. 57 female 2% Nat. Am. survivors 1% Asian Am. 159 male 88% Cauc. 16 34 assailants 3% Afr. Am. 121 female 3% Latino survivors 3% Nat. Am. 4% Mix. Races 283 male 74% Cauc. 34 32 assailants 11% Afr. Am. 4% Nat. Am 3% Latino 188 male N.R. N.R. N.R. assailants 106 male 86% Cauc. 20.0 31.1 assailants 12% Afr. Am. 2% Other 237 male N.R. 15.0 26-35 assailants 237 female survivors 170 male N.R. N.R. N.R. assailants 182 male 76% Cauc. N.R. 30.6 Table 11 (cont'd). Studies of Sample 291192.1nigll 5113 Berk et. al. 1658 male (1992) Dunford 247 male (1990) Dunford et. 330 male al. (1990) Pate 8 907 male Hamilton (1992) Sherman and 205 male Berk (1984) Sherman et 1092 male al. (1992) 150 53% 31% 14% 45% 43% 50% 43% 4% 3% 36% 42% 22% 45% 36% 16% 3% 76% Assailant's Ethnieirx Cauc. Afr. Am. Latino Cauc. Afr. Am. Cauc. Afr. Am. Latino Nat. Am. Cauc. Afr. Am. Latino Cauc. Afr. Am. Nat. Am. Other Afr. Am. Percent [1116le g 11 35 31 29 60 55 Average Age 30 29 31 18-25 26-35 36-50 50+ 15%: 44%: 34%: 7%: 32 32 151 Table 11 (cont'd). Study of Sample Assailant's Percent Average merrier Size Ethnieir! Maire Canales-Porta- 182 male 51% Afr. Am. 34 31 latin (1996)* 36% Cauc. 14% Latinos * Result obtained from Chapter 3 of this dissertation .-..... Table 11 (cont'd). Studies of W Bernard and Bernard (1984) DeMaris and Jackson (1987) Edleson and Grusznski (1988:1) Edleson and Grusznski (1988:2) Edleson and Grusznski (1988:3) Edleson and Syers (1990) Fitch and Papantonio (1983) Hamberger and Hastings (1990) Harrell (1991) Purdy and Nickle (1981) Saunders (1992) NOR. Average H.S. Average H.S. Average 2 years college Average 1.5 years college Average 1 yr coll. N.R. 20% N.H.S. 49% H.S. N.R. N.R. N.R. 152 Time of Relationship Edeeetien .Treee wz Vietin NOR. N.R. 12 mo. 86% Husb. 12 mo. N.R. 9 mo. N.R. 7 mo. N.R. 6 mo. N.R. N/A N/A 12 mo. N.R. 6 mo. N.R. 6 mo. N.R. N.R. N.R. Repeated Aeeault N.R. 35% 40% 33% apx. 44% apx. 35-46% N/A 30% 23% apx. 41% 4 an”. ‘ ~ 0 ' '1' ‘ Table 11 (cont’d). Studies of Beliee_Inrer1 Edneerien Berk (1992) N.R. Dunford 41% H.S. (1990) 17% Post H.S. Dunford et. 50% H.S. a1. (1990) 31% Post H.S 19% Some HOS. Pate 8 Hamilton N.R. (1992) Sherman and 31% H.S. Berk (1984) Sherman et 31% H.S. al. (1992) 153 Time of IIQQQ 6 mo. 19 mo. 18 mo. 23 mo. 6 mo. 6 mo. Relationship Repeated rL_Yierim Aeeeelr 64% Husb.] 15% Living Toge. 28% Lover 4% Husb.]Not Living Together 40% Husband N.R. 29% Lover] Boyfriend 3% Ex-husb. 1% Wife] Girlfriend] Ex-Girlfriend 42% Husband N.R. 39% Lover] Boyfriend 9% Ex-Lover/ Boyfriend 5% Other 4% Wife] Girlfriend] Ex-Girlfriend 1% Divorce] Separated 79% Husband N.R. 16% Boyfriend 3% Separated 2% Divorced 35% Husband 26% Otr. 45% Lover 13%Ar. 3% Ex-Hus. 30% Husband 4%Arrest 63% Lover 7% Other 154 Table 11 (cont'd). Studies of Time of Relationship Repeated menu WWW m1; Canales-Porta- 42% N.H.S. 36 mo. 34% Husband 26% latin (1996)* 48% H.S. 41% Boyfriend 25% Ex-partner * Result obtained from Chapter 3 of this dissertation 155 the generalizations of this study regarding other locations, its findings provide the potential for using its sample as representative of other probation departments in the United States. Hypetnesis T The first hypothesis of the current study was that court sanctions (e.g., jail, substance-abuse treatment program, partner abuse-abatement treatment, probation) were dependent on the personal characteristics of the assailants. .' . .1. inn-...”? 4‘ To test this hypothesis, the variables of personal characteristics, previous arrest history, and current incident were included in the prediction model. These variables classified correctly between 73 and 82 percent of the sample. The range in percentage was due to each court sanction receiving an independent test. The analysis of the four classification models suggested not rejecting the null hypothesis, because the strength of the relationship between the variables and the models was weak. 5.4mm The prediction model relating to the individuals who are more likely to be sentenced to jail was conducted with only one predictor. This predictor was whether or not the defendants were employed. The analysis revealed that the employed were less likely than the unemployed to be sent to jail. This is similar to the literature reviewed earlier. For instance, Reiman (1990) found that there were fewer people with high paying jobs and white collar criminals in 156 prisons, than the unemployed and poor, who were overrepresented in the prison system. There is not a study that specifically reported high levels of unemployed partner assailants sentenced to jail. The closest studies were the ones by Sherman and Berk (1984) and Sherman et al. (1992), who randomly assigned individuals to arrest or not arrest. As part of the arrest condition, the assailants spent a number of hours in a detention facility. These studies found that over 55% of the total cases in the experiments consisted of unemployed individuals. Thus, from these latest studies, it could be argued that if unemployment is a predominant characteristic in cases of partner assault, it is logical that a high proportion of unemployed individuals would be committed to jail. However, in the current study, only one third of the sample was unemployed, yet they were more likely than employed to receive a category of jail sentence. Thus, the likelihood of unemployed receiving a jail sentence could be attributed to the pre-conceived idea that unemployed have more free time and can therefore spend more time in jail than employed individuals, who are supposed to be busy at their job. Returning to the prediction model, the employment variable alone cannot be responsible for classifying 100% of the cases. WW Another example of a weak relationship between the predictors and the model was the prediction of who would attend a substance-abuse treatment program. This model had 157 only two variables with significant interactions with the dependent variable. These variables were the following: whether or not addictive substances were involved in the current incident of assault, and whether or not the defendants were employed. In cases were either the assailant or the victim were using a substance, assailants were two times more likely to receive a sentence for , substance abuse-abatement treatment. This finding seems to _-...-»w fit with the literature on substance abuse and partner assault. This literature is controversial and inconclusive. Some researchers have not been able to prove that controlled substances are the causes for the abuse, but others have stated that substances are the cause (Collins 8 Schlenger, 1988; Flanzer, 1993; Gelles, 1993; Island 8 Letellier, 1991; Rosenberg, Stark, 8 Zahn, 1986). Consequently, some scholars think that treating the alcohol problem does not solve the problem of assault, while others think that it does (Island 8 Letellier, 1991). The attitude of the judges concerning an involvement of substances in the assault, has been to refer the cases to substance abuse-abatement treatment. Contradictory enough, the partner abuse treatment programs traditionally have rejected individuals who were addicted to alcohol (Edleson et al., 1985; Hamberger 8 Hastings, 1990; Rosenfeld, 1992). The second variable that demonstrated a significant interaction with substance abuse-abatement treatment was whether or not the defendant was employed. Employed 158 defendants were almost four times more likely than non- employed to be sent to substance abuse treatment programs. There is no study that specifically reported the percent of employed partner assailants involved in substance abuse treatment. The closest reports of employed partner assailants are in studies of treatment programs. Most of these studies found that over 80% of their patrons were employed. Only one study found that 66% of the clients were employed (Edleson 8 Syers, 1990). These findings, although not conclusive, suggest that employed assailants were able to remain in the community and receive a sentence for community treatment. ego: -- 7., -. T ‘::a , iJa -ue1t _ea ue t The model predicting which assailants would be ordered to receive partner abuse-abatement treatment was weaker than the previous models. The classification model was not statistically significant. This meant that none of the variables used were good predictors for partner abuse- abatement treatment. Therefore, individuals who received a sentence to a partner abuse-abatement program, and those who did not receive this sentence, were very similar in the variables utilized. A direct comparison of these findings with the literature has not been possible because there is no research directly related to this area. A number of studies about treatment outcomes compared individuals who were ordered by the court and those who came voluntarily (Edleson 8 Grusznski, 1988; Grusznski, 1986). The -.. J fi-J‘ITP“? 159 researchers did not find differences between individuals referred by the court and those referred by others. W The model predicting whether or not the assailants would receive probation supervision also had a weak relationship with the predictor variable. The model had five predictors with significant interactions with the ’ dependent variable. These predictors were the following: I number of prior arrests, whether or not the last offense was 1 property related, whether or not the last offense was person related, whether or not the defendants were employed, and whether or not the defendants were black. Although this model had the highest number of variables that accounted for the prediction, the percentage of correct classifications was very low--only 73 percent. The predictor of the number of prior arrests indicated that individuals with a low number of prior arrests were more likely to be sentenced to probation than those with a high number of prior arrests. This prediction is consistent with the literature on probation. This literature explains that probation has been traditionally assigned to individuals who are first offenders, and with a minimal number of prior arrests (Champion, 1996; Dillingham et al., 1990). The Comptroller General of the United States has found this characteristic of no prior arrests an important predictor of probation success (Dillingham et al., 1990). Moreover, the judges in the district court, where the 160 current study was conducted, used the characteristic of prior arrest to sentence individuals to probation. Another predictor that determined whether the individual would be sentenced to probation was whether or not the last offense was property related. Individuals with a prior offense that was property related were two times more likely than individuals who did not have a property related offense to receive a probation sentence. This finding seems contradictory to what the literature suggests about probation success. The literature has indicated that individuals with no prior offenses were more likely to have good behavior and to comply with the laws in the non- incarcerated community (Dillingham et al., 1990). However, this literature does not indicate the likelihood of the success of individuals who committed property offenses. Thus, a possible interpretation of this phenomenon is that property offenses have no impact on the success of these individuals. 7 Similarly, a predictor that determined whether the individual would be sentenced to probation was whether or not the last offense was related to a person (i.e., disorderly conduct, indecent exposure, criminal sexual conduct). Individuals with a prior offense related to a person were two times more likely than individuals who did not have an offense related to a person to receive a probation sentence. Again, this finding seems contradictory to what the literature has suggested about probation success 161 and individuals with no prior offenses (Dillingham et al., 1990). However, the literature on the prediction of probation has not indicated the likelihood of the success of individuals with offenses related to a person. Thus, a possible interpretation of this phenomenon is that this offense has no impact on the success of these individuals. Another variable that takes part in the prediction of I who is sentenced to probation is whether or not the assailants were employed. Employed assailants were three times more likely than non employed to receive a sentence to probation. This also forms part of the literature on probation (Champion, 1996). When probation officers interview defendants, they take into consideration the employment status of the individual in their sentence recommendation. Therefore inferring from this literature, the employment status for partner assailants in probation was not by chance but as part of the standard procedure for probationers. The race variable also plays a role in the prediction of who is to be sentenced to probation. Black assailants were less likely than their counterpart to receive a sentence for probation. The literature on probation has not been able to present arguments on the issue of race. Race was not discussed, and it did not form part of the prediction of the success that the Comptroller General established for probationers (Dillingham et al., 1990). Perhaps the variable of race was not included because it has 162 been prohibited for being used as a differential measure in a public facility, according to the United States Civil Rights Act of 1964. However, Rector, Bagby and Nicholson (1994) affirmed that race/ethnicity influences verdicts in the courtrooms of the United States. Thus, it could be interpreted that the skin color and ethnicity of the defendants might influence sentencing. W t. In summary, the variables used to predict court sanctions correctly classified between 73 and 82 percent of the sample. Although these percentages included over fifty percent of the sample, the goal in a prediction study is to obtain the closest percentage to exact prediction (i.e., 100%). However, the analysis of each court sanction suggests to rejecting the null hypothesis, as result of the tenuous relationship between the variables and the prediction model. Only a small number of variables were responsible for the correct classification of individuals. Furthermore, it was not always possible to predict the sanctions that the defendants received. Thus, it is safer to conclude that in the current research, the demographic characteristics, previous arrest history, and the variable of current incident were unrelated to the interventions selected by judges. These findings suggest the possibility that other relevant variables not considered in the models could be more relevant to the prediction of the intervention selected 163 by judges. The current researcher had assumed that the demographic variables used for this study were the ones that influenced court sanctions, because they were always present in the records of the defendants. Furthermore, as the majority of the people in the sample were a non-European descend, this led the current researcher to believe that these demographic characteristics were the primary ones used for court sanctions. However, this proved to be incorrect. Perhaps variables that could provide better prediction would include the interactions between the assailants and the judges or probation officers, the personality of the judge, the personal policy of the judge, the facilities available for assailants, or the daily operation of the court. Perhaps assailants were sentenced to interventions according to the availability of facilities that deal with the specific criminal conduct, and their sentences were not related to either their personal characteristics or their criminal histories. Perhaps assailants were sentenced according to the recommendation from the probation officers, who might have based their decision on his or her interaction with the defendants. Perhaps because the policy for sentencing partner assailants has so many exceptions, the judges’ consistency is only in allowing exceptions. Another component of this inconsistency may be the lack of specific training for judges dealing with cases of partner assault. 164 In 1992, the Michigan Judicial Institute conducted a special seminar on domestic violence. The curriculum of the seminar focused, for the most-part, on the prevalence of partner assault. It also assessed the laws that have been established since 1979 to aid the criminal justice system, including judges, in handling cases of partner assault. The seminar failed to include information about sentencing for these assailants. Perhaps this information was not included for lack of research on the sentencing of partner I 13‘ WW” assailants. Hypothesis TI The second hypothesis of the current study states that a partner assailant with more court experiences is more likely to repeat an assault. The current researcher assumed that experiences with the court system would demystify it and diminish its effect on people, allowing individuals to learn how to use the system once they are acquainted with it. This proved to be untrue in the current study. The group who repeated assaults included individuals with a low number of prior criminal offenses, as well as individuals with a high number. No statistically significant specific pattern was found that differentiated one group from another. A possible explanation for this is that all accurate prediction of future assault must include other variables, perhaps unrelated to court experiences. 165 WI The third hypothesis of the current study stated that the involvement of substances in partner assault makes it more likely that the assailant will repeat an assault. Previous studies have shown that there is a relationship between substance abuse and assault. This was the rationale for testing the variables of substance abuse and repeating an assault. However, this relationship was not established v Item I: _ in the current study. One of the possible explanations for n‘.‘-_-r . .. ‘ i ,1 i this is that the involvement of substance abuse in the assault was reported not only for assailants, but for victims. This made the evidence in support of the hypothesis less strong. Moreover, this hypothesis relied on only one variable for its support. Given the limited current data, it was difficult to obtain other correlational variables to test the hypothesis. Hypotneeis IV The fourth hypothesis of the current study stated that the ecological model would provide better predictors for future assaults than those used in traditional risk assessment. Traditional risk assessment did not include these variables and did not consider partner assailants. Thus, the current researcher assumed that these variables would be good predictors with this group of offenders. However, the findings of this study did not provide evidence that this assumption was correct. The null hypothesis could not be rejected, because the ecological model did not 166 predict repeated assault. The chi-square value was not significant. One reason for this may be that the variables used were not the best variables for predicting recidivism; perhaps other variables would be better predictors. For instance, the personal pride for African Americans and Latinos, birth order, sex and the number of siblings may be factors in recidivism (Horton 8 Medley, 1978; Horton 8 Whitesell, 1979). A second reason may be that the dependent variable, ‘ l“ ‘r'7izu— - .4 repetition of assault, was too broad, and that only repetition of partner assault should have been considered. A third reason may be that human nature and circumstances are so complex and varied that it is not possible to predict when an individual will commit another act of assault. A fourth reason may be that individuals learn how to avoid rearrest. In the study of Geerken and Hayes (1993), most of the ”repeat offenders" committed a different offense than the one they were accused of in the first place. Thus, these offenders had learned not to commit the first crime but have not learned how to avoid future contact with the criminal justice system. Implications of the Results The overall purpose of the current research was to develop an empirically supported risk-prediction model for men who assaulted female partners. The rationale for the models was to predict court sanctions and repetition of an 167 assault, using variables from the ecological model. The ecological model has been used to understand partner assault (Carlson, 1984; Edleson 8 Tolman, 1992). The use of the ecological model provided a framework for organizing the variables in the current study. The areas used from the ecological model were the following: individual level, the microsystem level, and the exosystem level. Additionally, the current study argued that these predictions could be conducted using records from a probation department. The findings of the current study suggest that an empirically based risk-prediction model could be conducted for men who assaulted female partners. In the current study, a small number of variables were responsible for predicting correctly between 73 and 82 percent of the cases in the study. One way of observing these findings is that they support the possibility of conducting prediction studies. Another way of evaluating these findings is that the use of the ecological model to organize the variables in the current study was a good direction to take. The ecological model provided a conceptual organization for a number of variables that did not necessarily stand on their own. In general, the variables used in the current study were good. The data derived from the routine process that probation officers use to analyze their cases. These variables are among those that one expects the court system to use to pass judgement on assailants, since they were 168 consistent in all the cases. Furthermore, some of these variables were responsible for correctly classifying between 73 to 82 percent of the cases, depending on the prediction model. It was remarkable, however, that few variables were able to classify this high number of cases. However, other variables were not as effective and could not help the model to predict a 100% of the cases. Additionally, the variables that predicted some cases could not be used to predict all the models in the study. E Furthermore, the method of data collection was good. ;1 Evidence of this can be found in inter-rater and intra-rater reliability. The inter-rater reliability was 95%, while the intra-rater reliability was 94 percent. These high percentages in reliability levels implied that the information collected was consistent in the records and that it could be obtained in a consistent manner. Thus, different coders could obtain similar data from these records at the probation department, as evidenced in the inter-rater reliability. Additionally, the high intra-rater reliability implied that coders were consistent in obtaining the data. However, despite the guidance of researchers who believe that behavior can be predicted, and despite the use of accurate measures and techniques, the current study failed to predict the behavior of partner assailants (Gottfredson 8 Gottfredson, 1988; Menzies, Webster, 8 Sepejak, 1985; Monahan, 1981; Ohlin, 1951). Taking the 169 stand of researchers who believe that behavior can be predicted, the correct statement to use regarding this type of failure is that perhaps there were some other variables that the current researcher did not use which could have provided predictive results. Analyzing the current findings from the point of view of researchers who are opposed to the prediction of behavior, current evidence would support the idea that one cannot predict future behavior from a set of past or present behaviors (Menzies, Chunn, 8 Webster, 1992; Morris, 1995; Pepinsky, 1980; Wenk 8 Emrich, 1972). This is due to the fact that the current study had difficulties in predicting behavior in terms of sentencing and characteristics of repeat assailants. Characteristics of individuals could not be relied on predicting which assailants would repeat an assault, or which assailants would be sentenced to various options (i.e., jail, substance abuse treatment, partner abuse-abatement treatment, or probation). The modest findings of the current study regarding the prediction of the sentences of partner assailants reflect the lack of consistency in the criminal justice system. In the current study, it was observed that partner assailants were similar in terms of many personal characteristics, but were not similar in the sentence that they received for the offense of assault. Thus it was not possible to predict the characteristics of the assailants sentenced to jail, the 170 characteristics of the assailants sentenced to substance- abuse treatment program, the characteristics of the assailants sentenced to partner abuse-abatement treatment, or the characteristics of assailants sentenced to probation. The first inference that stems from these findings is that the criminal justice system is not consistent into sentencing process, and that there seems to be little justice operating in the system. The second inference is that the criminal justice system is chaotic and cannot be predicted. Thus, the main policy recommendation is for consistency in the sentencing process. If these assailants are treated consistently in terms of their sentencing, this suggests that the system is fair. Furthermore, the consistency in sentencing could deliver a message to the community, and to partner assailants, that the system is serious about addressing the problem of partner assault. Key players in developing a consistent process of sentencing are probation officers. Providing education to them concerning the effective manner of sentencing partner assailants could be an answer to the issue of the prediction of court sanctions. The training for district-court probation officers currently includes, among other things, information on screening clients who are substance abusers. However, it does not include training in laws, assessment, or treatment referrals of partner assailants. On the other hand, the training for judges about partner assailants 171 includes various aspects of the law, except sentencing. Thus, the information needed for educating probation officers and judges must address the lack of consistency in working with partner assailants, and the need to provide consistency in the sentencing of these cases. Thus, it is not only that there is a lack of education in sentencing partner assailants, but there is also a lack in the policies of the probation department in terms of how to handle cases of partner assault. Some policies of police organizations have been developed, based on the studies of police arrests. Other policies for treating partner assailants in community programs have been developed from the comparison of individuals who attended programs, the modality of treatment, and the length of treatment. However, this has not happened for cases that come to court. Furthermore, it would be inappropriate to create policies for probation departments, using samples from police or from treatment programs but not from probation departments. Such policies risk being inadequate for the reality of the organization. Limitations of the Study The current study examined a different criminal offense and a different setting than those used for other studies of prediction of criminal behavior. However, the demographic characteristics of the study’s sample did not prove to be an accurate predictor model for sentencing and recidivism. One 172 of the reasons for this may be that the variables examined are not predictors of sentencing or recidivism. Other variables might be more meaningful. Perhaps another reason that the model could not predict sentencing or repeat assault was that the data were unreliable. The data sources in the current study were documents in a district court probation department. The most important document used in the current study was the Presentence Investigation Report. This report, developed by probation officers prior to sentencing, provided details of the assailant’s background, reported the details of the assault, and provided recommendations to the judges. Additionally, the probation file contained the sentence for the assault, violations of court orders, and other information. The reason for using this document was that the information about the defendants was extensive and was kept in one location. It was the belief of the current researcher that these data were used by judges to carry out sentence dispositions; thus, "the most important documents for the court" were used for the current study. These findings are therefore disappointing and possible explanations for the results are contradictory. A second source of information for the current study was state police computer data, which provided information on arrests, prosecutions, and convictions for assaults. The statewide recidivism data were limited to information in the police reports in the computer, which included only the 173 arrest date and jurisdiction. This technical element limited the findings. It was unknown if individuals who recidivated had assaulted the same or another female victim. In order to increase knowledge about recidivism, guidelines should be developed to include specific identification of the victim in the official records of state police. However, information about victims might affect the seriousness of the effort to resolve the case. Moreover, it might violate victims’ rights and perhaps result in blaming the victim. Court clerks were not required to enter cases of partner assault in the state police computer records if they were considered misdemeanor assaults. The current researcher perhaps could have found more recidivism data if this data had been consistently entered in the computer. Additionally, more cases of repeated assault could have been identified if the police always arrested partner assailants. However, police have their personal policies and use discretion in cases of partner assault. This results in inconsistent information about who is arrested for subsequent offenses. In the spring of 1994, the Michigan legislature ordered that cases of partner assault be entered in the state police computer system. Thus, further studies examining recidivism of partner assailants could be carried out more easily, at least in Michigan. The current researcher was limited to studying recidivism in Michigan rather than at the national level. 174 The reason for this was that the Federal Bureau of Investigations (FBI) did not grant permission to conduct research with their data. Thus, it was not known if individuals repeated assaults within jurisdictions outside Michigan. Future Research The current study is the first quantitative study of assailants referred to a probation department. Other researchers in this area conducted telephone surveys with probation officers, or developed reflective essays without providing quantitative data (Davis, 1984; Hofford, 1991). However, the probation department is a good place to access these cases, because many assailants who come to court are sentenced to treatment or to jail, or the case is dismissed immediately. Thus, in order to obtain more information about men who have assaulted a female partner, other people involved with them should be interviewed. This includes probation officers, judges, and victims. In addition, a more detailed account of the social norms, values, and beliefs of men referred to probation departments could be obtained by conducting research at this level. Unlike the sample in studies of treatment programs, the current sample from a probation department resembles the characteristics of samples in studies involving arrested partner assailants (Sherman, 1992; Sherman 8 Berk, 1984; Sherman et al., 1991; Sherman et al., 1992). Thus, a probation department is an ' A“ II. I 175 excellent place to conduct research. Future research should also use experimental models at the court level to determine if court and probation interventions are effective in preventing future partner assaults. Conducting research at the court perhaps excludes a population of men who are self-referred to treatment programs or are directly referred by social service organizations, lawyers, or partners. According to the literature, this population tends to be educated, white, and i1: i‘t‘rb" . middle or upper-class. Referrals to these programs vary by location. In the city of the current study, the vast majority of men attending treatment programs were referred by the court. Thus, observing partner assailants who have been involved with the court system provides a sample screened by police and the prosecutor department, but it however provides a more varied pool than that at treatment centers. Other variables not considered in the current study that might improve the models for predicting recidivism include the following: the number of prior assaults to other individuals, including acquaintances, the history of being violent when angered, a high score on a test measuring desire for domination and control, acceptance of violence as a means of problem solving, and depression. These variables have been studied by other researchers and provide different typologies of abusive men (Gondolf, 1988; Saunders, 1992). 176 Perhaps these variables would be better predictors of men who repeat an assault than those used in this study. Furthermore, qualitative research might help in understanding the circumstances that influence men in avoiding further arrest for assault. Quantitative research has not been able to provide this information. Prediction studies have not been able to find a sufficient number of cases of men who repeated assaults. However, qualitative research could perhaps provide information about the motivation for avoiding further arrests. This question implies that assault on a partner continues despite arrest. However, if assaults on a partner do not continue after the first arrest, then perhaps the arrest has been effective. Conclusion The main accomplishment of the current study was in extending the knowledge about the characteristics of partner assailants who are referred to a probation department. As noted in the literature reviewed, this population has not been addressed by previous researchers. Thus, the current research brought to light information about a group of people often ignored by researchers in the area of partner assault. The current study had a second objective of predicting the behavior of individuals involved in cases of partner assault. The results of the current study indicate some of the potential problems connected with our ability to predict 177 the behavior of individuals. One prediction philosophy states that prediction could be achieved if better variables are considered. Thus, the current study could support that argument because the variables demonstrated a potential for predicting court sanctions--which indirectly are predicting the behavior of the judges sentencing partner assailants. The other side of the prediction philosophy-~composed of people who support the anti-predictions side of the argument--states that prediction cannot be achieved no matter what variables are used or the sample or the technique used. Thus, the current study could support that argument because the models studied were not able to predict with any degree of certainty the court sanctions nor repeated assaultive behavior. Therefore, the current researcher concludes this study in an awkward position, not being able to elect one position of the prediction debate over the other because the findings were not strong enough to support one or the other side. Perhaps failing to reject the null hypotheses in the current study leaves the current researcher with the option of supporting the anti-predictions philosophy. However, the support for this philosophy does not mean that we should stop conducting prediction studies, since one of the primary concerns of science is to explore new knowledge. Furthermore, researchers should be able to continue exploring the science of prediction in order to support the 178 particular philosophy of prediction chosen, whether it be for or against prediction. , . ( 1,:l:'3¢ APPENDIX A APPENDIX A SEARCH ON PROBATIONERS RECORDS Community Intervention Against Domestic Abuse (CIADA) Probate court Age of defendant Date of birth Gender of the defendant a. Male b. Female Ethnicity: a. Caucasian/White (non-Hispanic) b. African-American/Black (non-Hispanic) c. Native American d. Latin American/Chicano e. Asian-American/Asian-Pacific Education a. Some schooling but no high school degree b. High school graduate/GED diploma c. Some college d. Completed college, specify degree e. Trade School f. Some Graduate School, please specify 9. Graduate degree, please specify Employment status a. Employed b. Unemployed Offense for current case a. Aggravated assault b. Assault c. Assault and battery d. others Who was the primary victim: a. spouse (living together) b. boyfriend/girlfriend (living together) c. spouse (not living together) d. ex-boyfriend/ex-girlfriend (not living together) e. acquaintance - not living together 1. other f. child(ren) 179 :‘____.__V___ 11 .- A...L ”wt... 180 g. neighbor h. mother/father i. extended relatives 1. brother-in-law 2. sister-in-law 3. mother-in-law 4. father-in-law j. other SANCTIONB 10. Fine (amount): 11. Court costs (amount): 12. Jail (number of days): 13. Probation (dates): to 14. Restitution (CVRF): 15. Domestic abuse statute 16. Other 17. Placement for treatment a. Life skills b. Cristo Rey c. Community Mental Health d. Substance Abuse Treatment (Specified): e. Dimension of life f. No treatment 9. Other 18. Completed treatment a. Yes b. No c. Not applicable 19. Completed probation a. Yes b. No c. Not applicable 20. Type of trial a. Plead guilty on arraignment b. Judge trial c. Jury trial d. Does not indicate e. Other 21. Recidivated - for same offense a. Yes b. No 22. Number of months after first same offense a. Date of first offense b. Date of second offense F rise-rs APPENDIX B _ --__._..___ Fur . . 3 -, .. j I 0' A‘- _.. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. APPENDIX B PRELIMINARY DATA FORM Coding Form Probation Department Data Subject number: District Court: Sentence date: Probation term (months): Fine (amount): Court costs (amount): SJF: Restitution: CVRF: Jail (number of days): Special conditions: a. No contact with victim b. Finish GED c. Attend substance abuse program d. Attend domestic violence program e. Attend counseling (mental health) f. No assaultive behavior g. Credit for jail time h. Other: Appearance date in court: Plea of guilty (Yes/No): Date of current offense: Bond (amount): Bond’s date: Lawyer representation (Yes/No): How many adult arrests: Prior arrest history (Yes/No): a. Place: b. Offense: c. Date: d. Sanction (punishment): Prior juvenile cases (Yes/No: How many): 181 H 'F: ... .u 4 .TS'J‘mrn-Jw V 19. 20. 21. 22. 23. 24. 25. 25. 27. 28. 29. 182 Who was the primary victim of current incident 8 age: a. Spouse (living together) (Age): b. Girlfriend (living together) (Age): c. Spouse (not living together) (Age): d. Ex-girlfriend (not living together)Age: e. Acquaintance - not living together 1. Whom, relationship (Age): f. Child(ren) (Age): g. Neighbor (Age): h. Sibling (Age): 1. Mother (Age): j. Father (Age): k. Other (Age): Victim’s race: “1 Victim’s injury incapacitation (Y/N): r i t . Ej£{"h.‘2 2K Incident: a. Threw something at victim (Yes/No): r (Describe): b. Shoved or pushed victim (Yes/No): c. Punched victim (Yes/No): d. Pulled victim’s hair (Yes/No): e. Bit victim (Yes/No): f. Pulled or dragged victim (Yes/No): g. Raped or other criminal sexual conduct (Yes/No):___ h. Choked victim (Yes/No): 1. Other: Weapon utilized: a. Fist (Yes/No): b. Knife (Yes/No): c. Gun (Yes/No): d. Kitchen Appliances (Yes/No): e. Other Home Appliances (Yes/No): g. Construction Materials (Yes/No): h. Teeth (Yes/No): i. Other (Explain) Location of incident a. b. c. d. e. Town of incident: Residence (Town): Wife employed (Yes/No): Wife income (Amount): Defendant understand seriousness of the case (Y/N): ___ Home (both-assailant and victim’s home) Victim’s home only Defendant’s home only Street (includes inside of car) Other (specify): 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 183 Defendant remorseful for assault (Y/N): Defendant embarrassed for assault (Y/N): Marital status now (divorce etc.): Number of marriages: Age of spouse: Date of marriage: Age of spouse: Date of marriage: Number of children: Pays child support (amount): With whom defendant lives: Age of defendant: Date of birth: Education (years of completed ed.): Gender of the defendant a. Male b. Female Ethnicity: a. Caucasian/White (non-Hispanic) b. African-American/Black (non-Hispanic) c. Latin American/Chicano d. Native American e. Asian-American/Asian-Pacific Military experience a. Enlisted (Yes/No) Date: Branch: b. Inducted Date: Branch: c. Rank at time of discharge: d. Type of discharge: e. Date of discharge: f. Disciplinary actions: Employment status now: a. Employed b. Unemployed c. Weekly Earnings (Amount): d. Dates of employment: From: To: Past employment status: a. Employed b. Unemployed c. Weekly Earnings (Amount): d. Dates of employment: From: To: Other source of income: Confirmed relationship with victim: Spouse abuse act (769.4a) (Yes/No): 184 48. Probation (dates): To: 49. Completed probation (Yes/No): 50. Use of substances involved in incident (Yes/No): 51. Original charges a. Aggravated assault b. Assault c. Assault and battery d. other 52. Charges prosecuted under: a. Aggravated assault b. Assault c. Assault and battery d. others 53. Placement for treatment a. No treatment b. Cristo Rey Domestic Violence Program c. Life skills domestic violence program d. Community Mental Health e. Substance Abuse Treatment (Specified): f. Other 54. Completed treatment a. Yes b. No c. Not applicable 55. Victim was contacted by the probation officer (Y/N) 56. Date of contact: 57. Victim returned victim’s impact statement (Y/N): 58. Victim’s statement on the victim’s impact statement: 59. Victim’s statement on police report regarding contact with a5sailant: 60. Assailant states that he keeps contact with the victim: 61. Recidivated - for same offense a. Yes b. No 62. Date of second offense 63. Number of months after first same offense: 64. Reviewer number: 65. Date of review: 66. Data entered date: Jo gm v . " . “WV I APPENDIX C -' ii 1.5m“? .'.l 1. 2. 3. 4. 5. 6. 7. 8. 10. 11. 12. 13. 14. 15. 16. 17. 18. APPENDIX C LAST CODING FORM PROBATION DEPARTMENT DATA Subject number: Probate court: Sentence date: Probation term (months): Fine (amount): Court costs (amount): SJF: CVRF: Other restitutions (amount): Jail (number of days): Special conditions: [ ] No Yes a. No contact with victim b. Attend education/training program c. Attend substance abuse program d. Attend domestic violence program e. Attend counseling (mental health) f. No assaultive behavior 9. Credit for jail time h. Other: Appearance date in court: Date of current offense: Bond (amount): Bond’s date: Lawyer representation (Yes/No): Prior arrest history (Yes/No): How many adult arrests: Last adult arrest: a. Place: b. Offense: c. Date: d. Sanction (punishment): Who was the primary victim of current incident: a. Ex-girlfriend b. Ex-spouse c. Spouse d. Girlfriend 185 ‘F... .4' 2 .K‘p‘nflr ‘ 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. Victim’s Age: 186 Victim’s Date of Birth: Victim and assailant living together (at incident) (Yes/No): Victim’s race (According to police report): a. b. Asian-American/Asian-Pacific Native American 2n 4. . 1’2‘?M:, c. Latin American/Chicano d. Caucasian/White (non-Hispanic) e. African-American/Black (non-Hispanic) Incident: a. Threw something at victim (Yes/No): (Describe): b. Shoved or pushed victim (Yes/No): c. Punched victim (Yes/No): d. Pulled victim’s hair (Yes/No): ; e. Bit victim (Yes/No): ‘ f. Pulled or dragged victim (Yes/No): g. Raped or other criminal sexual conduct (Yes/No):___ h. Choked victim (Yes/No): 1. Other: Weapon utilized: a. Fist (Yes/No): b. Knife (Yes/No): c. Gun (Yes/No): d. Kitchen Appliances (Yes/No): e. Other Home Appliances (Yes/No): g. Construction Materials (Yes/No): h. Teeth (Yes/No): 1. Other (Explain) Use of substances involved in incident (Yes/No): Marital status now (Circle appropriate) a. b. c. d. e. Number of marriages: Widowed Divorced Still married Separated Single Number of children: Pays child support (Yes/No): Age of defendant: Defendant’s date of birth: Defendant’s education (years completed): 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 187 Ethnicity (Circle appropriate): a. Asian-American/Asian-Pacific b. Native American c. Latin American/Chicano d. Caucasian/White (non-Hispanic) e. African-American/Black (non-Hispanic) Currently Employed (Yes/No): Current job title: Current weekly Earnings (Amount): Start date of current employment: End date of current employment: Previously employed (Yes/No): Previous job title: Previous weekly Earnings (Amount): Start date of previous employment: End date of previous employment: Date probation began: Date probation ended: Completed probation (Yes/No): Mandated for treatment (Yes/No): Type pf Tfeatnent a. Substance abuse treatment (Yes/No): b. Domestic violence treatment (Yes/No): c. Mental health counseling (Yes/No): d. Other (Specify): P a ement eat e a. Cristo Rey Domestic Violence Program (Yes/No): b. Life skills domestic violence program (Yes/No): c. Community Mental Health (Yes/No): d. Cristo Rey Substance Abuse Treatment (Yes/No): e. Other (Specify) Completed treatment (Yes/No): Data collector code: Date of data collection: Code of data entry: Date of data entered: sun-i? ',‘ "4"!!! ' T.- 5321- -- ‘1—‘_ ‘5“ I APPENDIX D APPENDIX D SUMMARY OF VALUE OF FACTOR ANALYSIS ROTATED MATRIX Variables Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Age .32385 .02961 .66918 .05393 -.15145 Separated .04797 .12492 .78217 .00959 .01302 Education .00842 -.10041 .05553 -.20878 .74721 Employed -.25991 .02782 .40900 .23898 .37945 Income -.22972 .16502 .50324 -.03140 .18457 Pine paid -.09575 .17479 -.05145 .30998 .73934 Caucacian .03255 .91455 .15216 .02546 .07938 Black .09879 -.90275 -.05405 .00510 .02456 Num. Prior Arrests .78507 -.01694 .07354 -.07780 .00568 Last Offense .54277 .09064 .24903 .37631 .16205 Days in Jail .71824 .00137 -.06432 -.20974 -.O7203 Days on Prob. -.24898 p .10352 .01219 .69873 .27213 Treatment .01062 -.04373 .05766 .79787 -.07413 Variables Factor 6 Factor 7 Factor 8 Married .75730 .18651 .07411 Living togeth .75692 .23140 .22362 Substances .15959 .57058 .03018 Ethnic Diff. -.05399 .58190 -.27690 Incident scale .11529 .05547 .86869 188 APPENDIX E APPENDIX E APPROVAL TO OBTAIN INFORMATION FROM HUMANS MICHIGAN STATE UNIVERSITY March 27, 1995 TO: David Canalas-Portalatin RE: IRBI: 93-282 TITLE: COMMUNITY INTERVENTION AGAINST DOMESTIC ASSAULT, A FEASIBILITY STUDY REVISION REQUESTED: 03 01/95 CATEGORY: - APPROVAL DATE: 07/06/94 Tha Univarsity Committaa on Rasaarch Involving Human Sub acts'(UCRIHS) it; review of this project is complata. I am plaasad to adv so that the rights and waltara of tha human subjacts appear to be adequately protected and mathods to obtain informad consent are appropriate. lhagzgore, tha UCRIHS approvad this project including any revision 1.. lbOVO . RENEWAL: UCRIHS approval is valid for one calendar year, beginning with the approval data shown above. Investigators planning to continua a projact ha and one year must use the groan ranawal form (anclosad with t a original a roval latter or whan a pro act is ranawad) to sack u at certification. Thara is a max of four such axpadlt ranawals ssibla. Invastlgators wishin to continua a projact bayond tha tima naad to submit it again or complata rav . REVISIONS: UCRIHS lust raviaw any changas in rocaduras involving human subjacts, rior to in tiation of t a changa. If this is dons at tha tima o ranawal, plaasa usa the green ranawal form. To ravisa an a rovad protocol at an 0 bar time during tha yaar sand your wr ttan raquast to tha CRIHS Chair, requasting ravisad approval and rafarancin tha project's IRB I and titla. Includa in our raquast a dascr ption of tha change and any ravi ins rumants, consant forms or advertisements that ara applicabla. PROBLEMS CHANGES:I Should aithar of tha followin arlsa during the coursa of tha work, lnvasti ators must noti UCRIHS romptly: (I) groblams (unaxpactad s da affacts comp aints, a c.) involv ng uman subjects or 2 changas in the rasaarch anvironmant or naw information n icating graatar risk to the human sub acts than axistad whan tha protocol was pravlously raviawad an approvad. OFHCEOF If wa can ha of an futura hal laasa do not haaitata to contact us “saga: at (517)355-2190 oi sax (511)335- 171. GRADUATE Sincaraly, STUDIES lhhuflucumumnos . Runuuumnhhg avid E. Wri ht WWW ucams Chairg ' (uuuwn MW State Unwersoty D“: pjm mmmstmoaemlding cc: William S. Davidson ENmememmn «ma4ow mumsmm FAx:517/432-1171 "2 “€99!an 5U! UWVPISrIy IDEA IS MSIIMWI DM‘ISIIy. Exam In Amen MSU :5 an Mom-3am mlwmry mmm 189 LI ST OF REFERENCES LIST OF REFERENCES Barefoot v. Estelles, 103 S. CT. 3383 (1983). Bell, C. C., & Chance-Hill, G. (1991). Treatment of violent families. ournal t e Na 'on edical Assogiatignl 83, 203-208. Berk, R. A., & Sherman, L. W. (1988). Police responses to family violence incidents: An analysis of an experimental design with incomplete randomization. Journa of t e Angrignn ggngisgiggl Assoc1ation, 83 (401), 70-76. Bernard, J. L., & Bernard, M. L. (1984). The abusive male seeking treatment: Jekill and Hyde. £nn11y_3gln§ign§L 33 543- 547. Binder, A., & Meeker, J. (1992). The development of social attitudes toward spousal abuse. In E. S. Buzawa & C. G. Buzawa (Eds. ), mest' vio ence: The c a 'n ' 'nal iQ_Li§§_:_§an§g (pp. 3-19). Westport, CT: Auburn House. Bonsignore v. City of New York, No. 81-7733, U.S. Court of Appeals, Second Circuit, June 15, 1982. Bradshaw, R. A. (1987). Multivariate actuarial prediction of felonious recidivism of male parolees: Development and cross-validation of a series of risk assessment modelts using stepwise logistic regression. (Doctoral dissertation, Michigan State University, 1987). Dissentntign ansgzngts Ingennational, 48, 751A. Burgess, E. W. (1928). Can scientific methods be applied to parole administration? In A. A. Bruce, E. W. Burgess, & A. J. Harno, (Eds. ), or e ate-se tenc la and t e are e s ste 'n lino's (pp. 246-249). Chicago, IL: State of Illinois. Buzawa, E. S., & Buzawa, C. G. (1990). Domestic violgnce; Ina gnimina; justice response. Newbury Park, CA: Sage. 190 b-7flf‘?.—*av 191 Canales-Portalatin, D. (1994, March). Befg;:nls_§g_§ng e art e t: e w ve ass te spgn§g1 Paper presented at the 1994 Annual Meeting of the Academy of Criminal Justice Sciences, Chicago, IL. Carlson, B. (1984). Causes and maintenance of domestic violence: An ecological analysis. c' S ' e v' w 8 569-586. Champion. D- J- (1996) W £2EEBE1§¥.29£I§§L12E§1 Upper Saddle River, NJ: Prentice Hall. Cohen. J” 8 Cohen. P- (1975)- W c e av' s 'e c s Hillsdale, NJ: Lawrence Erlbaum Associates. Collins, J. J., & Schlenger, W. E. (1988). Acute and chronic effects of alcohol use on violence. gnnnnnl_gf Stng1§§ on AlconolI 49l 516-521. Davis, L. V. (1984). Beliefs of service providers about abused women and abusing men. §oc1n1 Wgrk. 22. 243-250. DeMaris, A., & Jackson, J. K. (1987). Batterers’ reports of recidivism after counseling. oc'a C e or 68 458-465. Dillingham, S. 0., Montgomery, R. H., a Tabor, R. W. (1990). 2:9 bn§1gn nn ng parole 1n prnct1ge. Cincinnati, OH: Anderson. Dobash, R. E., & Dobash, R. (1979). V191enge §g§1nst w1yg§3 A gasg agn1nst pagz1angny. New York: Free Press. Dunford, F. W. (1990). System-initiated warrants for suspects of misdemeanor domestic assault: A pilot study. W 631-653- Dunford, F. W. (1992). The measurement of recidivism in cases of spouse assault. Jgunnnl Q; Cnim1nnl an & _Iimian_Q¥i_§li 120 136- Dutton, D. G. (1985). A systems approach to the prediction of wife assault: Relevance to criminal justice system policy. Ina C. D. Webster, M. H. Ben-Aron, 5 S. J. Hucker, (Eds. ), e u 88' P ba t d r ' ' ' ' ' (pp. 103-114). New York: Cambridge University. Dutton, D. G. (1987). The criminal justice response to wife assault. MW 189-206- 192 Dutton. D- G- (1988) W W Boston: Allyn and Bacon. Edleson, J. L. (1991L Coordinated community resposes. In M. Steinman (Ed. ), Wonan bgtgen1ng; £911ccy Iggpgnge § (pp. 203-219). Cincinnati, OH: Anderson. Edleson, J. L., & Grusznski, R. J. (1988). Treating men who batter: Four years of outcome data from the Domestic Abuse Project. Journal 9f Sgc1a1 Sezx1ge ResearcnI 12. 3-22. Edleson, J. L., Miller, D. M., Stone, G. W., & Chapman, 0. G. (1985). Group treatment for men who batter. §g§1n1 Wgzk Reseangn nnd nbstzncts, 21, 18-21. Edleson, J. L., & Syers, M. (1990). Relative effectiveness of group treatments for men who batter. §Qc1al W 10-17- Edleson, J. L., & Syers, M. (1991). The effects of group treatment for men who batter: An 18-month follow-up study- WW 227-243. Edleson, J. L., & Tolman, R. M. (1992). 1ntenvention f2r_men_xh2__atLer1_An_e_glggigal_annrgaghi Newbury Park CA: Sage. Emery, R. E. (1989). Family violence. Angn1gnn W411 321-328. Endler, N., & Magnusson, D. (1976). Toward an interactional psychology of personality. £5ycholog1ga1 W 956-974- Fagan, J. (1989). Cessation of family violence: Deterrence and dissuation. In L. Ohlin & M. Tonry (Eds. ), '.e 1 d '-r c ° : - 'ew -_ _-:-. c ° V--. .u V191ence (pp. 377- 425). Chicago, IL: University of Chicago Press. Fagan, J. A., Stewart, D. K., & Hansen, K. V. (1983). Violent men or violent husbands? Background factors and situational correlates. In D. Finkelhor, R. J. Gelles, G. T. Hotaling, & M. A. Straus (Eds. ), e a s e : WW (1):» 49 67) Beverly Hills. CA: Sage. Feld, S., & Straus, M. A. (1989). Escalation and desistance of wife assault in marriage. gn1ngnnugnnLLJELL 141-157. . 1‘1. 1... 193 Field, J. K. (1993). Domestic violence. In J. A. Curtis, S. Bassett, & L. M. Collins (Eds.), M1cn1gan Eani1y ' ' (pp. 16.1-16.47). Ann Arbor, LaE1_£QBILD_EQ1§1QDI_¥QLL_II MI: The Institute of Continuing Legal Education. Fields, M. D. (1994). Criminal justice response to violence against women. In A. Duff, S. Marshall, R. E. Dobash, & R. P. Dobash (Eds. ), Pena; gneory and prac§1ce: ov t o r' ' a ' s 'c (pp. 199-215). Manchester, UK: Manchester University Press. Fitch, F. J., & Papantonio, A. (1983). Men who batter: Some pertinent characteristics. The gournal of Nazyana ang HQELQL Q1seasa. 171, 190-192. Flanzer, J. P. (1993). Alcohol and other drugs are key causal agents of violence. In R. J. Gelles & D. R. Loseke (Eds.), co ve s es f i v'o c (pp. 171- 181). Newbury Park, CA: Sage. Ganley, A. L. (1987). Perpetrators of domestic violence: An overview of counseling the court-mandated client. In D. J. Sonkin (Ed.), Damest1c v1o1ence on finial (pp. 155-173). New York: Springer. Garner, J., Fagan, J., & Maxwell, C. (1995). Published findings from the spouse assault replication program: A critical review. _2urna1__f_QuanL1_aii___gri_1nglegxi_111 3- 28. Geerken, M. R., & Hayes, H. D. (1993). Probation and parole: Public risk and the future of incarceration alternatives. Qn1n1n919g21_;11 549-564. Gelles, R. J. (1993). Alcohol and other drugs are associated with violence--They are not its cause. In R. J. Gelles & D. R. Loseke (Eds.), gnzrent contraver§1es an fan11y violence (pp. 182-196). Newbury Park, CA: Sage. Glueck, S., & Glueck, E. (1930). 500 Cr1mina1 Caneens. New York: Knopf. Gondolf, E. W. (1985). Men who batter: An 1ntegnated apnnaach for agapp1ng w1fe abuse. Holmes Beach, FL: Learning Publications. Gondolf, E. W. (1988). Who are those guys? Toward a behavioral typology of batterers. V101enca ang V1gt1nsI 3I 187-203. Gondolf, E. W., & Fisher, E. (1988). Intervention with batterers. ' o ' An te 'v o 194 tnaan1ng leannag nelp1essness (pp. 59-75). Lexington, MA: Lexington Books. Gottfredson, D. M., & Gottfredson, S. D. (1988). Stakes and risks in the prediction of violent criminal behavior. Wm 247- 262- Griffieth, W. E. (1987). Risk prediction models for female offenders (Doctoral dissertation, Michigan State University, 1985). Disserta ion st cts te t'o al 47 3656-8. Grusznski, R. J. (1986). Court-ordered treatment of men who batter. D1ssertatign Abstract 1ntennat1gnall 46l 3594B. (University Microfilms No. 85-26, 416) Grusznski, R. J., & Carrillo, T. P. (1988). Who completes batterer’s treatment groups? An empirical investigation. Journal of Fanily Violencel 3, 141-150. E. Hamberger, L. K., & Hastings, J. E. (1988). Personality characteristics of spouse abusers: A controlled comparison. 42W 331-48- Hamberger, L. K., & Hastings, J. E. (1989). Counseling male spouse abusers: Characteristics of treatment completers and dropouts. V191enge ang V1ct1nsI 5, 275-286. Hamberger, L. K., & Hastings, J. E. (1990). Recidivism following spouse abuse abatement counseling: Treatment program implications. V1o1ance and V1ct1n§II 5. 157-170. Hammond, N. (1977). ngestic assau1t; A Ignaz; an W Lansing. MI: Michigan Women' 8 Commission. Harlow, C- W- (1991). W Washington, DC: U.S. Department of Justice. Harrell, A. (1991). u ’o -o de e t e e t ' ' ce er ° n e ort. Washington, DC: The Urban Institute. Hart, H. (1923). Predicting parole success. Joannal of crifliflél Law ana CgiminOLOQX, 14, 405-413. Hofford, M. (1991). Family violence: Challenging cases for probation officers. Eaaana1_£;ana§1an1_§§1 12-17. Horton, A. M., Jr., & Medley, D. M. (1978). Prediction of recidivism in a criminal population by birth order and family size. W 27- 30 195 Horton, A. M., Jr., & Whitesell, L. J. (1979). Prediction of recidivism by sex and number of siblings. angnolgg1ca1 RaportsI 45l 98. Hosmer, D. W., & Lemeshow, S. (1989). Appl1ed 1ogistic ragga§§1an1 New York: John Wiley 8 Sons. Island, D. & Letellier, P. (1991). Man_gna_naa§_;na_nan ov e ° a e e a men and domest'c v'o ence. Binghamton, NY: Harrington Park. Kirk, A. (1989). The predication of violent behavior during short-term civil commitment. §u11et1n 9f tne Anar1aan Academy of Payghiatny and the LawI 17I 345-353. Langan, P. A., & Innes, C. A. (1986). Preventing domestic violence against women. Augean a; gn§§1ce S§a§15t1cs: Sneg1a1 Renongs (Report No. NCJ-102037). Washington, DC: U.S. Department of Justice. Lipsey, M. W. (1992, November). ID..§£I§§£§_Q£ atment v 1 e en esults et - analys1s. Paper presented at the NIMH Meeting for Potential Applicants for Research to Prevent Youth Violence, Bethesda, MD. Mannheim, H., & Wilkins, L. T. (1955). £I§Qi§£12fl megnad a 1n nelat1gn £9 nonstal aza1n1ng. London: Her Majesty' s Stationery Office. Marguart, J. W., Ekland-Olson, S., & Sorensen, J. R. (1989). Gazing into the crystal ball: Can jurors accurately predict dangerousness in capital cases? Law ang So Q1at y w 449-468- Marshall, L. L. (1992). Development of the severity of violence against women scales. u V' nce 11 103-121. Menzies, R. J., Webster, C. D., & Sepajak, D. S. (1985). The dimensions of dangerousness: Evaluating the accuracy of psychometric predictions of violence among forensic patients. MW 49-70- Menzies, R. J., Chunn, D. E., & Webster, C. D. (1985). Risky business: The classification of dangerous people in the Canadian carceral enterprise. In K. McCormick & L. A. Visano (Eds.), Canaa1an_nann1agy (pp. 61-94). Toronto: Canadian Scholars' Press. Meyers, M. (1994). News of battering. Joannal of QQEEBD19§§12n1_iiI 47' 53- 196 Michigan Domestic Violence and Treatment Board (1993). Annna1_3angrn1 Michigan Department of Social Services. Michigan Judicial Institute (1992). Mag1§nzaga WWW Lansing. MI: Author. Monahan, J. (1981). ' ' o e v c ' .Beverly Hills, CA: Sage. Monahan, J., 8 Steadman, H. J. (1994). Toward the rejuvenation of risk assessment research. In J. Monahan 8 H. J. Steadman (Eds. ),V Violence and mental gisorden: Qayalgnnants 1n r185 assessment (pp. 1-17). Chicago, IL: University of Chicago Press. Morgan v. District Columbia, No. 79-588, District Columbia Courts or Appeals, August 31, 1982. Morris, R. (1995). Pena a ' io : h t'c ang1aa1 Toronto: Canadian Scholars' Press. Morris, N., 8 Miller, M. (1985L Prediction of dangerousness. In M. Tonry 8 N. Morris (Eds. ), Q;1na_an_ jns s§1ce: An annnal rev1ew gr neaearcn: Vol, é (pp. 1- 50). Chicago, IL: University of Chicago Press. Norusis, M. J. (1992). SP C+ dvance at' 'cs. Chicago, IL: SPSS. Ohlin, L. (1951). Select1gn no; paro1a. New York: Sage. Pate, A., 8 Hamilton, E. E. (1992). Formal and informal deterrents to domestic violence: The Dade county spouse assault experiment. Anerican Soc1o1ogical Aevieg, 57, 691- 697. Pepinsky, H. E. (1980). C ' e t t 'es: innnaauc§1gn SQ tne stnay of cn1ne. New York: Oxford University Press Petrila, J., Otto, R. K., 8 Poythress, N. G. (1994). Violence, mental disorder and the law. In M. Costanzo 8 S. Oskamp (Eds). W (1111- 161-180)- Newbury Park, CA: Sage. Pirog-Good, M., 8 Stets-Kealey, J. (1985). Male battererse and battering prevention programs: A nationale survey. Vcti ati n of W e §(3), 8-12. Police Foundation- (1976) W W Washington. DC: Author. 197 Purdy, F., 8 Nickle, N. (1981). Practice principles for working with groups of men who batter. §oc1a1 Wonk w1th Gzogpg, 5, 111-122. Rector, N. A., Bagby, R. M., 8 Nicholson, R. (1994). The effect of prejudice and judicial ambiguity on defendant guilt ratings. Ina Joanna; of Socia1 Esycno1ggy, 133, 651- 659. Reiman, J. (1990). Ine p1gn get :1cha: and pne pgor get pn1agn1 New York: Macmillan. Reiss, A. J. (1985). Some failures in designing data collection that distort results. In L. Burstein, H. E. Freeman. 8 P- H- Rossi (EdS-). Q21lasting_exaluat1on_datai Problem§_and_solution§ (pp- 161-177). Beverly Hills. CA: Sage. Rosen, R. (1993). e 'v' 0 et e ° e wgnen and nen ahould know. Far Hills, NJ: New Horizon Press. Rosenberg, M., Stark, E., 8 Zahn, M. (1986). Interpersonal violence: Homicide and spouse abuse. In J. M. Last (Ed-). Pub1ic_hea1th_and_2rexentixe_medicine (12th ed-. pp. 153-217). Norwalk, CT: Appleton-Century-Crofts. Rosenfeld, B. D. (1992). Court-ordered treatment of spouse abuse. Q11n1ca1 Egycnglggy Eey1ew, 12, 205-226. Saunders, D. G. (1988). Issues in conducting treatment research with men who batter. In G. T. Hotaling, D. Finkelhor, J. T. Kirkpatrick, M. A. Straus (Eds.), Coping w1tn fan11y v1olange: gasaarch ang po11gy pepspecpiyes (pp. 145-156). Newbury Park, CA: Sage. Saunders, D. G. (1992). A typology of men who batter: Three types derived from cluster analysis. AE§£1§QE_£22£E§1 Qt Ortngpsxgniagzx, §2, 264-275. Saunders, D. G. (in press). Prediction of wife assault. In J. C. Campbell (Ed.), Asses1ng tne p1sk of gangepousness. Newbury Park, CA: Sage. Sepejak, D., Menzies, R. J., Webster, C. D., Jensen, F. A. S. (1983). Clinical predictions of dangerousness: Two- year follow-up of 408 pre-trial forensic cases. Bn11etin of the Anepigan Agageny pf Psycn1atry and the LawI 11, 171-181. Shepard, M. (1987, July). 1npepvent1on w1pn nen wnp r: e a ua ' 0 me t' se 0 Paper presented at the Third National Family Violence Conference for Researchers. Durham, NH. 198 Sherman, L. W. (1992). Bo11c1ng dgnest1c v101enge: Exeeriments_and_dilemmas_ New York: Free Press- Sherman, L. W., 8 Berk, R. A. (1984). The specific deterrent effects of arrest for domestic assault. Anapican Sociglogi__l_Bexi__1_A2i.261- -272- Sherman, L. W., Schmidt, J. D., Rogan, D. P., Gartin, P. R., Cohn, E.G., Collins, D. J., 8 Bacich, A. R. (1991). From inicial deterrence to long-term escalation: Short- custody arrest for povery ghetto domestic violence. Criminologx1_22i 821-850- Sherman, L. W., Smith, D. A., Schmidt, J. D., 8 Rogan, D. P. (1992). Crime, punishment, and stake in conformity: Legal 8 informal control of domestic violence. Anap1gan Spg1plgg1ga1 Beviaw, 57, 680- 690. Straus, M. A. (1978). Wife beating: How common and why? V1ap1mglogy, 2, 443-458. Straus, M. (1993). Identifying offenders in criminal justice research on domestic assault. Anap1aan_§anay1gpa1 Sgigntist, 36ll 587-600. Straus, M. A., 8 Gelles, R. J. (1988). How violent are American families? Estimates from the National Family Violence Resurvey and other studies. In G. T. Hotaling, D. Finkelhor, J. T. Kirkpatrick, 8 M. A. Straus (Eds.), Eamily abuse_and_i_§_conseguences (pp- 14- 36) Newbury Park. CA: Sage. Straus, M. A., Gelles, R. .J" 8 Steinmetz, S. K. (1980) s d : V' ce t e e ' ' Garden City, New York: Anchor Books. Syers, M., 8 Edleson, F. L. (1992). The combined effect of coordinated criminal justice intervention in woman abuse. Jgnrnal of Intenpersonal Violence, 7, 490-502. Tarasoff v. Regents of University of California, 17 Cal. 3d 425, 131 Cal. Rptr. 14, 551 P. 2d 334 (1976). Thurman v. City of Torrington, 595 F. Supp. 1521 (D. Conn. 1984). Tolman, R. M., 8 Bennett, L. W. (1990). A review of quantitative research on men who batter. Jo o Inteppersgnal V191ance, 5, 87-118. Warner, S. (1923). Factors determining parole from Massachusetts reformatory. Joanna; of Qr1nina1 Law ang Cp1m1nology. 14, 172-207. 199 Webster, C. D., Sepejak, D. S., Menzies, R. J., Slomen, D. J., Jensen, F. A. S., 8 Butler, B. T. (1984). The reliability and validity of dangerous behavior predictions. t o t a em Ps c 'at d e Law 121 41- 50. Weis, J. G. (1989). Family violence research methodology and design. In L. Ohlin 8 M. Tonry (Eds.), Cp1ne and ' t’c : ev' w f es a c ° V0 11 m' v' lence (pp. 117-162). Chicago, IL: University of Chicago Press. Wenk, E. A., 8 Emrich, R. L. (1972). Assaultive youth: An exploratory study of the assaultive experience and assaultive potential of California youth authority wards. Jgurnal of Beseapcn 1n Qr1ma 8 Qelingnency, 2, 171-196. L.- Wilson, W. J. (1991). Studying inner-city social ') dislocations: The challenge of public agenda research. <‘ Amep1can Sogiglog1ca1 gev1ew, 56, 1-14. 81‘ Wolfgang, M., Figlio, R., 8 Sellin, T. (1972L W ChicagO. IL: University of Chicago Press. Wordes, M., Bynum, T. S., 8 Corley, C. J. (1994). Locking up youth: The impact of race on detention decisions. WWW 149-1-65-