It 101.5: 111,1 1111111IIII ,1,” H,11'1 IWII III11 1111131'1'1111I‘1‘11111IINIIII ---- III1IIII 1,1111 1111I111111113111'.11,11 III11. 1,1,1. 11111 111111, "I 1"“ 11I1I1I1I11,1 'II11,11II1‘"11I “1. ,,,,,1111 ,1: .11 ,11 111, 11 ,,11I1,111,.1111,,.,,, I I IIIII11 11 111 «11,11 1111111 11 111. I1I111111 I111 1 1111,,11111111 II 111I111,1,I,"1III 111.1, £11,;1II11I1' ,1,111 1111} III ‘ III I: I" .I 1’.II~ :- 1 11111”, 1111: 111111111111 I 11I11I11I1I’IIIi I 1 11 I 111111,- 111, ,1“ 1,1111 ,1111H11I111LII I I111 1111 ,1 ,1111 111%? (m. 9 I111! I1 I11 ’ :wm.a V. .,.__'— j;"—_: " r... #9:: mo . x 5, put W. w 111 v - _ . - 2.2.4 .~.-.-a.-v—~—-—-- . ~ - .A_ -g. 7— , i 1 III: 11.1.“. 1, ,_ ‘ 1111111111,,111,1111,1,111,11-1,,21,1,,.,I,,:1 1 ..I 1 1, 16'11 “1‘ 1 11:1. 11:11:sz1 f; 11 ”1‘1, 111 11‘, 1,1111, , 1.: ,- :1: III-I 1: I . 1211‘s": 1-11111“'1' 111 1 1,1I1I11.111.311111111151111 1.111111: 11* 1 1111111 :11 1‘ 1111111111 I“1 III 11 111.1131 III I 11 , 111 1,1111§:.11,,11111,1I11.1-I, VII I1, 1111 x,,1’,1,,,,I1I,,I,I '1; ,1. ...'. ,' 1,1 11,111,, 1 11111111111 , ,1 '1. 2 11:, 11,11. '1: ""1: 111' 1,141 I11? 1 '; 11" 111 I 11 .§:-2,: 1,1,1 II, 3111 111 1 11 II . 1111,,,:,: ,,11: III11 I111I1I11IIIII1II1I II ’1’”in III; 1 1 1111’11'111'1IL- I" II III:I1I1111IIT‘111II.IIIIIE‘I’I’ILIIIIIIIIII II II IIII 1%.. 1111‘“ III": s . 4 1); . .2} ’ I n '31' III 1‘ ,1“ -. 1.1. -. I III? 1 1 11111111: 131111111 ' . ' 1 1 1. l j.” ’ I 111. 1‘11 , t .1 1 115 11 1 1. 4111 I 'I I I 1 x ‘ 1 . 1 W '._.:——- _.. ' --«_."""'7-‘._.." , , “fix-i: 11439»? W W! / MW III/W 3 1293 00658 ”WW L i'“ "L a F‘%! 52:13 I . .V, _ .4 . 5‘. 3‘ “Vb-’9‘ .-,‘ ‘ i-i’ OJ-.- This is to certify that the dissertation entitled Risk Pred1ct1on Models for Female Offenders presented by Winnie Ruth Griffieth has been accepted towards fulfillment of the requirements for Ph. D. degreein Psychology / flajor prolessor 7 August 8, 1985 MS U is an Affirmative Aca‘on/Equal Opportunity Institution 0-12771 1 MSU BEIURNING MATERIALS: Place in book drop to remove this checkout from LIBRARIES -—:—- your record. FINES will be charged if book is returned after the date stamped below. 2 , ‘55? 2 5 ”’1', I it Ir [1:6 '0 twat 0‘9. .1. ,. _._~ ; . . ‘ ".,-3 ' Q... ~(' .2 i“) J“ ‘7‘ -. ' .0 ‘ r r v: u “at; T . hat. RISK PREDICTION HODELS FOR FEMALE OFFENDERS By Hinnia Ruth Griffiath A DISSERTATION Submitted to Michigan State Univ-rsity in partial fulfill-ant of the raquirenants for thl dagrec of DOCTOR OF PHILOSOPHY Department of Psychology 1985 l“. -.f a ( ~ l, .l n . ll . p . Ni ,,‘ V o l . In p .n . 'v A . on f . . o. .l I); p q! 0—. ‘ ABSTRACT RISK PREDICTION MODELS FOR FENALE OFFENDERS By Binnie Ruth Griffieth A number of models have been developed to evaluate the liklihood of an offender returning to prison once released into the_ community. This research attempted to develop risk prediction models for female offenders released from prison into community correction centers. There were two basic models developed. The first model looked only at those variables (indicators) available prior to the time of release from prison into the community correction centers. The second model utilized information available from the community correction centers (the first data base was also included). As measures of recidivism, a number of techniques were employed to assess whether the offender was unsuccessful in her community adjustment. The results of the study indicated that the first model (pre prison variables) were poor at predicting ones return to prison and the post prison generated models were better at predicting recidivism. Neither groups of models explained a reasonable amount of the variance. An overall conclusion was that more time and effort needed to spent assessing information once the offender in released into the community. In addition the research results have broad implications for correctional programming and policy making. To my mother with love ii TABLE OF CONTENTS LISTWTMESIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII INTROMTIMIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII Literature Review................................ Narner...................................... Hart........................................ Mg“.IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII Bluxk.IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII Other Studies-Prior to 1960................. Gottfredson and Colleagues.................. Elaser...................................... Comparison of Statistical Methods of Prediction.............................. Halfway House/Community Correction Center........ Historical Development...................... Community Correction Programs and the Roman 0“”deIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII Community Correction Programs and Recidivism.............................. Critique of Research on Recidivism............... Thm.tic.l I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I “thMOIWic‘l I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I Rationale for Present Research................... WDIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII Subjects......................................... Design........................................... Instrument Construction.......................... Data Collection.................................. REst-IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII Descriptive Statistics........................... Data Reduction................................... Cluster Analysis............................ Principal Components Analysis............... 12 12 13 14 lb 21 23 41 41 42 45 53 54 55 67 67 83 85 Discriminant Function Analysis................... 97 Pre Prison Release Variables................ 103 Post Prison Release Variables............... 120 Multiple Regression.............................. 139 Pre Prison Release Variables................ 141 Post Prison Release Variables............... 145 Summary.......................................... 157 DISCUSSIDNIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII 178 Validity of Sample............................... 178 Model Predictability............................. 187 Policy Implications.............................. 189 Decision Making Model....................... 189 Programming................................. 192 Implications for Future Research................. 194 REFMSIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII 196 iv LIST OF TABLES Table Page «*1 Female Population, Labor Force , and Persons Arrested, 1930-1980........................... 5 2 Female Arrests Michigan 1975-1979............. 9 3 Female Commitments and Incarcerations Michigan 1975-1979............................ 10 ,4 Criminal Offenses and Statuatory Sentence Lengths....................................... :3&E¥K “’5 Frequencies of Personal Characteristics....... b9 6 Frequencies of Criminal History............... 73 “~7 Frequencies of the Most Serious Instant Offense....................................... 76 8 Frequencies of Behavior in Prison............. 79 9 Freqencies of Behavior in Community Correction Center........................................ 81 10 Frequencies of Post Prison Violations by Type. 82 11 Frequencies of Post Prison Dispositions by Type of Disposition................................ 84 12 Cluster Analysis and Cluster Loadings......... 86 13 Correlation of All Independent Variables Hith Outcome Variable.............................. 90 14 Eigenvalues and Percent Variance Accounted For By the 17 Empirical Pre-Prison Release Components.................................... 93 15 Principal Components Results of the Pre-Prison Release Variables Hith Accompanying Component Names and Component Loadings.................. 94 16 Eigenvalues and Percent Variance Accounted For By 19 Empirical Post-Prison Release Components.................................... 98 17 Principal Components Results of the Post-Prison Release Variables With Accompanying Component Nmaes and Component Loadings.................. 99 18 Discriminat Function Analysis for Pre Prison Release Components with Recidivism I as Dependent Variable............................ 104 19 Classification Results - Pre Prison Release and indiVis. IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII 105 20 Discriminant Function Analysis for Pre Prison Release Components Hith Recidivism II as Dmendmt v‘riabIEIIIIIIIIIIIIIIIIIIIIIIIIIIII 107 21 Classification Results - Pre Prison Release and R‘CidiVi“ IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII 110 22 Discriminant Function Analysis for Pre Prison Release Components With Recidivism III as 23 24 25 26 27 28 30 31 32 81 id 35 36 37 3B 39 41 42 Dependent Variable............................ Classification Results Pre Prison Release and Recidivism III............................ Discriminant Function Analysis for Pre Prison Release Components Hith Recidivism IV as Dependent Variable............................ Classification Results Pre Prison Release and Recidivism IV................................. Discriminant Function Analysis for Post Prison Release Components With Recidivism I as Dependent Variable............................ Classification Results Post Prison Release and Recidivism I.............................. Discriminat Function Analysis for Post Prison Release Components with Recidivism II as Dependent Variable............................ Classification Results Post Prison Release and Recidivism II............................. Discriminant Function Analysis for Post Prison Release Components with Recidivism III as Dependent Variable............................ Classification Results Post Prison Release and Recidivism III............................ Discriminant Function Analysis for Post Prison Release Components Uith Recidivism IV as Dependent Variable............................ Classification Results Post Prison Release and Recidivism IV............................. Stepwise Multiple Regression Analysis for Pre- Prison Release Components with Recidivism V as Dependent Variable......................... Stepwise Multiple Regression Analysis for Pre- Prison Release Components flith Recidivism VI as Dependent Variable......................... Stepwise Multiple Regression Analysis for Post Prison Release Components with Recidivism V as Dependent Variable......................... Stepwise Multiple Regression Analysis for Post Prison Release Components With Recidivism VI as Dependent Variable......................... Intercorrelations of the Dependent Variables.. T-Values for Pre Prison Release Components By Recidivism Types.............................. T-Values for Post Prison Release Components By Recidivism Types.............................. Correlations Between Predicted Scores and Criteria of Pre and Post Prison Release Components.................................... Comparison of Predictor Equations Generated From Pre Prison and Post Prison Factors....... vi 112 114 116 119 121 124 125 128 130 133 135 138 142 146 149 153 156 159 163 168 169 43 44 45 Significant Pre Prison Release Components By Recidivism Types for the Four Research Hypothesis.................................... 171 Significant Post Prison Release Components By Recidivism Types for the Four Research Hypothesis.................................... 172 Significant Components by Pre and Post Prison Release Equations Across Recidivism Types..... 176 vii INTRODUCTION One of the most important decisions encountered by the criminal justice system is that of releasing individuals from incarceration. This decision affects not only the criminal justice staff, but persons incarcerated in the correctional facilities as well. Much record on improving decision making in this area, both empirical and theoretical, continues to be presented in the literature. However, given this attention to improving the efficiency of prison release decision making, recidivism rates continue to be high across the country. In the past ten years, 1969-1979, the recidivism rates (usually in terms of parole failure) ranged from a low of 132 to a high of 38% (U.S. Department of Justice, LEAA, 1980). This study concentrates on two important areas in the criminal justice system. These two areas are female offenders and prison release decision making; specifically, release from prison into community correction centers. The first area, female offenders, is of importance because of the lack of research on this population, as well as the overall increase in the arrest and incarceration rates for women. A breakdown of these rates by year and offense will be given. First, however a brief overview of the etiology of female crime will be presented in hopes of providing insight on the population being studied. In addition a theoretical basis for data collection will be determined by this overview. Most writers have based their female crime theories on their definitions of femininity or femaleness. Freud’s contention was that “women who are not passive, and who are not content with their roles as mothers and wives are maladjusted. Their sources of maladjustment is penis envy“ (Freud, 1933). According to Freud, all women experience penis envy to some extent, the difference being that the “adjusted“ woman was able to compensate for the lack of a penis through the sex act and through motherhood (Freud, 1938). Female criminality was according to Freudian theory seen as a deviant way of compensating for the lack of a penis. Other writers continued in supporting this biological/psychological thesis. Lombroso described female criminality “as an inherent tendency produced in individuals that could be regarded as 'biological atavisms’: a survival of primitive traits in individuals, particularly those of the female and the nonwhite race" (Lombroso, 1903; Klein, 1973). within this framework of ”biological atavisms” Lombroso characterized the ”female offender as masculine and the noncriminal woman as feminine“ (Lombroso, 1903). This conclusion was generated from his study on the skulls of female criminals. Lombroso’s results, even though later discredited, were that ”the physiognomy and brain capacity of female offenders more closely approximated that of the male skull than that of the noncriminal female skull” (Lombroso, 1903). Later writers continued to base their theories of the female criminal on biological/psychological foundations. Some writers related female criminality to such things as physical size (Cowie, Cowie, & Slater, 1968); lack of control of sexual impulses (Elueck 8: Elueck, 1934); and menstrual cycles (Ellis and Austin, 1971). An important deviation from the biological/psychological orientation began with Pollak’s “The Criminality of "omen“ (1950). His theory "challenged the basic assumptions concerning the extent and quality of women’s involvement in criminal behavior“ (Simon, 1975). He felt that the etiology of female crime was not much different from that of male crime; however, because of the types of crimes women commit they are less likely to be detected, reported, and prosecuted; women appear to commit less crimes than that of their male counterparts. In terms of the etiology of female crime, Pollak felt that economic factors played a role in the explanation of female crime, even though a minor one. Though Pollak broadened his understanding of female crime to include the possibility of its having an economic basis, he continued to give some 'credence to the sexual/psychological explanations of crime. Pollak further thought that, even with this additional economic factor, crimes with economic motives were ”masculine” and those crimes of sexual activity were "feminine.“ I More recently a few writers have considered other factors in the explanation of female crime. The profiles of women arrested and/or incarcerated for crime have been used as a theoretical underpinning in understanding female crime. “omen arrested, convicted, and/or incarcerated tended to be poor, under-educated, unskilled, and more often than not of a minority race (Slick & Neto, 1977). Most of their crimes tended to be an expression of these characteristics: shoplifting, larcenies, drugs, murder (usually a lover, husband, or friend) (Adler, 1975; Simon, 1975). Nith the apparent end to racism, classism, and sexism, the profiles of those women arrested, convicted, and/or incarcerated should reflect a wider distribution of the female population. For instance, it was suggested that with the onset of the women’s liberation movement, female crime, arrest, conviction, and incarceration rates would increase (Adler, 1975; Simon, 1975). Also the crimes for «muse H. «magma voucdmumos. races «exam. was emsmoam >xxmmnma. Homo-Hmmo mmsm_m races “exam mesmmm >1smmum voucdmnmoz wwwmm ”wwwMMMa mmmwdwmwmmuwmmwoz «Hymmmxmmwwdwmmoz E . E E . A: E Home Ham.auw.ooo a~.~mw.o¢o H.mwu.ooo um.» w.~w emsnmze szoammmm am us How we as» memo moa.uom.ooo uo.umm.coo mau.ooo ~m.m o.mo umwnmzn msowmmmm pm we Hum mo moo Home mo.mo~.coo -.o~o.oo aom.ooo Na.m o.am emsnmzndmznxmmmm No um one Hm mmo Homo um.mma.ooo Ha.mmm.ooo Nu.ooo mm.m o.~o umxnman msnxmmmm .Hm NV am Ho am Home om.mom.ooo Hu.oc~.ooo m~.ooo Hm.m 6.65 emxnman sansmmmm m mm Hus Hm mum some mo.oum.coo Ho.muw.ooo Ho.oocm Hu.m o.ow mocxnm" newness. Home. mmmnmswama we mmscdm mxnxmsodmasoz om man‘smnmm mos Hmwmuaeuo. which women committed would reflect their new role in society. It was thought that women would commit more white-collar crime, i.e., embezzlement, fraud, because of their emergence into the labor force and white collar jobs (Adler, 1975; Simon, 1975; Smart, 1977). Unfortunately,' this scenario was (is) not supported (Chapman, 1980). The female population and their participation in the labor force has moderately increased over the past six decades. Of interest however is the ratio of women in the labor force to the female population. There has been a slight increase over the past few decades in this labor force/population ratio (see Table 1). However in examining the ratio of female arrests to population, for this same time period it appears to have skyrocketed. The arrest ratio does not seem to mirror the ratio of women’s participation in the labor force. Therefore Adler’s and Simon’s contention that as women become more involved in the work world they will proportionally be more involved in crime (white-collar crime) is not supported by the data. According to the Uniform Crime Reports (UCR), 1,571,140 women were arrested for crimes committed during the year 1981. Uniform Crime Reports classifies crime into two types; Type I and II. Type I, or serious crimes, includes ‘the offense of murder, forcible rape, robbery, aggravated assault, burglary, larceny-theft, motor vehicle theft, and arson. In 1981, one out of 5.2 persons arrested for a serious crime was a woman. The largest percent increase between 1977 and 1981 in Type I crimes was for violent crimes (+14.4%); with property crime percent increase for this same period, +6.02. The Type 11 crime category accounted for, in 1981, 73.42 of all women arrested (Type II offenses-other assaults, forgery and counterfeiting, fraud, embezzlement, stolen property, vandalism, weapons, prostitution, and commercialized vice, sex offenses, drug abuse violations, gambling, offenses against children and family, driving under the influence, liquor laws, drunkenness, disorderly conduct, vagrancy, and other offenses, except traffic). The largest percent increase between the years 1977 and 1981 was for embezzlement (+53.5X), driving under the influence (+50.1Z), and fraud (+45.2%). The increase in fraud was, for the most part, a result of welfare fraud (Chapman, 1980). Likewise, most of the arrests for embezzlement were a result of a lower level employee (i.e., bank teller) having access to corporate funds (Chapman, 1980). It appears from the Uniform Crime Report (UCR) data that there was an overall increase in female crime. However, much of this increase was not only a result of women committing more crimes, but also a function of more women being arrested, the change in UCR crime category definitions, and the actual number of agencies reporting data to the Uniform Crime Reports (i.e., 1970 - 5,270 agencies reporting data to UCR compared to 12,042 agencies in 1980). The second area to be emphasized is the decision-making process of a woman’s release from prison into community correction centers. The national arrest statistics showed an increase in the number of women arrested; therefore, one could assume an increase in both prison commitments and community correction center placements. However, the State of Michigan (the State supplying data for this project) showed a somewhat different trend. The number of arrests between and including the years 1975—1979 showed a decrease for women. There was an approximately 30% decrease in the number of women arrested in 1975 and 1979 (Table 2). However, the number of women on the prison "rolls" increased; 242 women in 1975 and 457 women in 1979. Likewise, the number of women placed in community correction centers increased by 162% between the years 1975 and 1979 (Table 3). It appears that women during this period were receiving more severe sentences for their crimes; therefore, leading to a higher number of women released from prison into community correction centers. This apparent increase in commitment may have implications for predicting risk of prison releasees. Table 2: Female Arrests - Michigan 1975 - 1979 Female Population Number of Year 2 15 years Arrests Reporting Agencies 1979 3, 634, 343 49, 859 570 1978 3,583,557 48,610 576 1977 3,533,872 53,623 x 1976 3,486,418 62,464 x 1975 3,450,987 62,664 510 Source; Michigan State Police, Uniform Crime Reports Michigan Department of Management and Budget, Office of Vital Statistics t Number of reporting agencies not available Table 3: Female Commitments and Incarcerations Michigan 1975 - 1979 New Total Prison Count Year Commitments Commitments Start End Center Count 1979 206 359 472 457 377 1978 268 451 433 472 339 1977 244 408 347 433 196 1976 196 332 248 347 150 1975 228 295 202 248 143 Source: Michigan Department of Corrections 11 To summarize, the purpose of this study will be to develop a prediction model for community correction center placement for female offenders. The literature review will first present an overview of the theory and research in the area of recidivism prediction. Although corrections literature of recidivism has primarily focused on the return on the prison parolees, most of the prediction theories and models were not relevant to the present study. One reason was that the prior prediction theories and accompanying models were based on the outcome of parole success. In addition, for the most part, these models were based on a theoretical understanding of criminality in terms of a male sample and not a female study sample. Finally, although the degree to which sex differences play a part in the understanding of crime and the prediction of recidivism is not thoroughly understood, it must not be overlooked. The second section of the review will present a brief overview of the literature of community corrections; specifically, the history of community correction centers. Also included is an overview of the research on recidivism .and community correction centers. The final section of this chapter will present the rationale of the study and the suggested hypotheses. 12 LITERATURE REVIEW WWW This section will first review the literature on paroles prediction research; starting with the initial effort of Warner to the more recent studies of Gottfredson and his colleagues. Next, various statistical methods for the development of prediction models will be compared. mm: 8.8. Warner’s (1923) study onparole outcome for the Massachusetts Department of Corrections is usually regarded as the 'first' parole prediction research. Warner selected 680 prisoners of the Massachusetts State Reformatory; of which 300 were parole successes, another 300 were parole violators and the remaining 80 were not paroled. The entire sample had appeared before the Parole Board between the years of 1912 and 1920. Warner collected information under twelve general categories: family background, race and nationality, residence, education, habits, circumstances at time of crime, the crime, prior criminal record, physical examination in reformatory, report of alienate, life in reformatory and parole behavior (Warner, 1923). 13 Warner compared the three groups in terms of pre, during, and post-prison behavior. His general conclusion was that, for most of the criteria used by the Parole Board, there appeared not to be any differences between the parole violators and nonviolators (Mannheim & Wilkins, 1955). The two most distinguishable factors in terms of parole success/failure were sex offense commitment and property offense commitments (larceny and breaking and entering). Approximately two-thirds of paroled sex offenders were successes as compared to the overall rate of 502. Prisoners committed for larceny and breaking and entering offenses had a violation rate of 572 (Warner, 1923). Elect Immediately following the publication by Warner’s results, his conclusions were strongly criticized by Hornell Hart (1923). The primary criticism of Hart was that Warner's findings, or lack of them, were due to his failure to apply statistical significance tests in analyzing his data. Hart suggested that the Parole Board “could greatly improve its parole results by proper utilization of the information already at its disposal" (Hart, 1923). Hart (1925) suggested that all of the significant -flmctors in Warner’s study should have been combined into an outcome score for each prisoner. 14 To derive such a scoring system, the intercorrelation between the various items tabulated by Professor Warner as well as their correlation with parole violating would have to be studied so as to work out the best possible weighting system for scoring the pertinent facts (Hart, 1923). The overall weakness of Hart’s study (analyzed Warner’s data) was ”that he accepted the prisoner’s own story when it suited him, but rejected it when it seemed to conflict with previous criminological findings” (Mannheim & Wilkins, 1955). Hart discussed devising a weighting scoring system, but did not present one. We The importance of Ernest Burgess’ work rested on a suggestion made by Hornell Hart, that of applying a weighted scoring system to parole prediction data. aurgggg studied the factors related to the success or failure of parolees and from this developed the first table of prediction rates for parole outcome. His study sample comprised three groups of 1,000 paroled men. These 3,000 men came from the Illinois State Reformatory at Pontiac, the Illinois Penitentiary at Joliet, and the South Illinois Penitentiary at Menard (Burgess, 1928). When the overall rate of parole violators were compared to the rates for each item for which data was collected, it was found that for certain items, these rates were high and for others lower than the overall rate. The 15 overall rate of parole violations was 28.4%, however for offenders of fraud/forgery offenses the rate was 42.42; offense of murder, 92; offenders with no previous work history, 44.42 ; offenders with regular work history, 12.2% (Mannheim & Wilkins, 1955; Burgess, 1928). In order to develop a decision-making model that would be used by the Parole Board for the release of individuals from prison, Burgess assigned an arbitrary weight of one point to each of the 12 discriminating factors. Therefore 'a paroles whose violation rate or, say 12 factors, was found to be below the (overall) rate for his institution, was given 12 favorable points, whereas the factors for which his violation rate was above the (overall) rate of violation was ignored“ (Mannheim & Wilkins, 1955; Burgess, 1928). A table was derived from this method, entitled I'Expectancy Rates of Parole Violations and Non-Violations." This table illustrated the expectancy rates for violation of men in various score classes; for example, one having 6 to 21 favorable points had an expectancy violation rate of 1.52, with 2 to 4 points an expectancy violation rate of 76% (Mannheim & Wilkins, 1955; Burgess, 1928). Burgess’ method, while being a foundation to Base Expectancy Rating, has been both methodologically and theoretically criticized. The criticisms were that Burgess 16 used only data contained in the official files of the paroles and used only that data showing the prisoners’ conduct during the official period of the parole. Another criticism was that his categories (variables) were both overlapping and subjective and he did not take measures to assure the reliability and validity of the data and instruments (Mannheim & Wilkins, 1955; Vold, 1931). A final criticism,- that of assigning equal weights of one point to each of the 21 discriminating factors, was questionable, or more strongly inappropriate. The assumption of equal weighting assumes that each factor is equal to all others in discriminating between those who violate parole and those that do .not violate. Not all variables (factors) were equally correlated with the criterion. 51.23515: Sheldon and Eleanor Elueck (1930) were the first to conduct longitudinal research on parole recidivism. The first of their series, "500 Criminal Careers,” on parole recidivism was based ”upon a careful investigation into the life histories of all prisoners, 510 men, released from the Massachusetts Reformatory“ (Eluecks, 1930). A period of five years was used as the follow-up. 17 The study data were comprised of the following; institutional files; fingerprint files (from the Corrections Department) used to verify the identity of the subject, and interview data (in approximately 74% of the cases, the ex-prisoners and/or their relatives were interviewed in order to evaluate and supplement the institutional data) (Bluecks, 1930). Of the sample used for the study, more specifically, of the 422 men whose post-parole conduct could be collected, 21.1% were parole “successes“, 16.8% were “partial successes“, and 61.1% were total "failures” on parole (Glueck and Blueck, 1930). All of the factors, pre-reformatory, reformatory, parole period, and postparole periods were analyzed in relation to the post-parole criminality of the paroles. These factors were analyzed using the statistical technique of the mean square contingency coefficient. The results led to a classification of the factors in three categories; (a) those having a coefficient of correlation under .20, therefore, only slightly or not at all related to the criticism of success/failure; (b) those having a coefficeint of from .20 to .40, therefore appreciably related to the criterion; and (c) those having a coefficient of .40 to .60 considerably related to the criterion (Glueck & Elueck, 1930). From the above correlations, the Gluecks constructed prognostic tables using the six most important pro-reformatory factors. The six factors and their 18 correlation with the criterion, success, were pre-reformatory work habits (C 8 .42), seriousness and frequency of prereformatory crime (C =- .36), arrest for crime preceding the offense for which sentenced to reformatory (C a .29), penal experience preceding reformatory incarceration (C a .29), econOmic responsibility preceding sentence to reformatory (C 8 .27), and mental abnormality on entrance to reformatory (C 8 .26). The total score, based on the combination of these factors, yielded a correlation of .45 with the criterion. The scoring classes were developed by ”adding the highest and lowest percentages of total failures which an offender could have and by adding two more score classes between these two extremes“ (Mannheim & Wilkins, 1955). The second study conducted by the Sluecks, "Later Criminal Careers“ was a continuation study to “500 Criminal Careers.” The Bluecks followed the parolees in their first study for an additional period of five years. A total of 63 factors were examined. Of the 26 factors showing a considerable relationship to the criteria of success/failure, the highest five correlations ranged from .43 (mental condition) to .23 (age at first delinquency). The remaining factors were, on the whole, the same as those found in the Sluecks first study; however, the correlation between the criterion and the ”work habits“ was somewhat 19 lower in the second five year followup and the correlation with “mental condition“ was slightly higher than it was in the first study. The top five factors determining criminal conduct over the 10-year time period were used to develop a prognostic table. The third study in this series, “Criminal Careers in Retrospect“ followed the same group of parolees over an additional five-year time period. The. Bluecks conducted a second prediction series, the first of which was entitled ”One Thousand Juvenile Delinquents“ (1934). In this study the authors examined the procedure of release conducted by the Borstal Juvenile Court and the Child Guidance Clinic of the Judge Baker. The authors looked at the relationship of approximately 60 factors to the criterion of success/failure. The data were analyzed by the use of two different statistical techniques, the ”factor method“ and the "failure score method“ (Bluecks & Elueck, 1934). A prediction table was constructed by using the top six factors having the highest correlation with success. The factors and their correlations were: discipline of juvenile by father (C I .233), school misconduct (C 8 .206), age at first-known behavior disorder (C I .187), discipline of juvenile by mother (C a .161), length of time between onset of delinquency and examination of child by the clinic (C a .161), and school retardation (C - .151). The correlation 20 of all of the six factors to the criterion was a mere .28 (Glueck & Glueck, 1934). In the second study of this series, ”Juvenile Delinquent Grown Up“ (1940), the previous study sample was followed for a 10-year time period. A prediction table was constructed using the 13 factors that showed the “likelihood of success, failure, early failure and later success; or erratic behavior during various forms of treatment such as probation, production under suspended sentence, reformatory, prison and also (conduct) in the Army and Navy“ (Glusck & Blueck, 1940)» The practical application of these tables to a small number of cases was demonstrated, but no general validation was attempted (Mannheim & Wilkins, 1955). A study conducted by the Eluecks in the late 1940’s, attempted to predict delinquency prior to its onset. ”Unraveling Juvenile Delinquency” (1950) was different from their prior studies in that the authors attempted to predict delinquency using a sample of both delinquents and nondelinquents. This sample consisted of 500 delinquent boys committed to State Correctional Schools in Massachusetts. The data consisted of the following: ”(a) traits of character structure derived from the Rorschach test, (b) personality traits derived from psychiatric interviews, and (c) social background” (Mannheim & Wilkins, 21 1955; Glueck & Glueck, 1950). From the data three tables of five factors each were constructed. The tables were further developed by comparing the incidence of three subclasses by each factor of the delinquent and the nondelinquent groups, and by comparing score classes. Comparisons of the results of the three tables demonstrated that there was agreement between the social and the psychiatric tables in 67.9% of the cases, and between the psychiatric and Rorschach tables in 69.8% of the cases. In only 49% of the cases did all three tables place the subject in the accurate predictive category (Glueck & Glueck, 1950). WWW George Vold evaluated a set of factors and their relation to parole outcome (Vold, 1931). Using the coefficient of mean square contingency, the highest correlation between a factor and parole outcome was .28 for the variable previous work record (Vold, 1931). An additional component of the Vold research was the comparison of the efficacy of the Burgess technique in using all available factors without weighting them, to the Glueck technique of using only the most significant factors (Mannheim & Wilkins, 1955). A negligible difference was found between the results of the two techniques. Another feature of the Vold study was his attempt to validate his 22 findings by dividing his sample into two randomly selected groups. Similar results were obtained by Clark Tibbet (1931) in articles published during this same time period. Tibbet used essentially the same procedure as that used by Burgess. The highest correlation (tetrachoric 'r') between predictive factors and the criterion of parole success/failure was .179 for “type of offender” (Tibbet, 1931). In 1951 Lloyd Ohlin constructed experience tables for male parolees from Illinois. Ohlin initially selected 29 factors of which he retained 12 for the development of his experience tables. Each of the 12 factors were divided into subclasses; violation rates were developed for each subclass by dividing the number of violators by the total number of persons in the subclass (Ohlin, 1951). The subclasses were also defined as either being “favorable,“ neutral,” or "unfavorable" in terms of their predictive quality. Each subject was given one point for every favorable subclass, one point for every unfavorable subclass, and zero for every neutral subclass in which he appeared. His total score was derived by subtracting the number of unfavorable points from the number of favorable ones (neutral sub-classification was ignored). From these calculations, violation rates ranged from 3% for subjects 23 with 5 to 10 favorable points to 75% for those subjects with 5 to 6 unfavorable points (Ohlin, 1951). Some obvious criticisms of Ohlin’s technique were that he relied on subjective and sometimes overlapping factors, assumed that all factors were equally significant to the criterion, and did not employ statistical tests of significance. §9§tfiggdggg_§gg leleagges Don M. Gottfrsdson was perhaps one of the more. important researchers in the area of criminal justice prediction and recidivism. Gottfrsdson’s major accomplishment was the development of the Bass Expectancy Scoring System. A base expectancy “is a statement of the expected parole violation rate for a given group; and this statement is made on the basis of past experience with such groups” (Gottfrsdson, Ballard & Bonds, 1962). The technique for the development of the Base Expectancy Scoring System was to: (1) select a representative sample for study; (2) divide the sample into groups on the basis of a criterion of parole outcome; (3) collect information for each person; (4) measure the relationship (correlation) of each item to every other item, including the parole outcome criterion; (5) select a small number of items which, appropriately weighted, do most of the work of prediction; and, (6) test the resulting 24 scale by applying it to a different sample of person, to see how well it works in predicting parole outcome (Gottfrsdson, et al., 1962). In 1962 Gottfrsdson and colleagues utilized a Base Expectancy Scoring System for women released from the California Institution for Women. The sample included 719 women released between the years 1955 and 1958. The criterion variable was returned to prison (or not returned to prison) within two years after release from prison onto parole. A number of items were seen to be related to parole outcome. The following predictors were items effecting parole outcome: The proportion of non-returnees went up with increased length of the longest marriage; the proportion of non-returnees decreased with any previous arrest history; more of the non-returness had worked a longer period on any one job, had fewer aliasis, fewer prior arrests, fewer jail commitments, and had been incarcerated fewer times (Gottfrsdson, et al., 1962). The best single predictor found, related to known history of heroin use--fifty-five percent with heroin history were non-returnees compared to 82% without heroin history; while a little over one-fourth of the sample had heroin history, these cases made up 472 of all those returned to prison before two years (Gottfrsdson, et al., 1962). These relationships permitted the calculation of a multiple regression equation to provide for a Base Expectancy Scale. The general finding was that the proportion of nonreturnees decreased with decreasing bass expectancy scores. A validity check was conducted in order to test 25 the scoring method on the second sample, again the proportion of nonreturnees decreased with decreasing scores. Other studies conducted by Gottfrsdson and colleagues supported their Base Expectancy technique as a viable tool in predicting recidivism of offenders released from prison (see Gottfrsdson, 1965; Gottfrsdson, Ballard, Manning, and Babst, 1965). Elise: In 1964 Daniel Glaser published “The Effectiveness of a Prison and Parole System.” In this work Glaser utilized a method of combining predictor information, a method not involving the use of predictor scores. The method employed was 'Configural Analysis.” Glaser’s configuration table was developed using a cross-section of all adult male federal prisoners released during the year 1956; a total sample of 1015 men. The criterion was post-prison release success. “Success“ was defined as ”not being returned to prison or receiving an non-prison sentence for a felony-like offense within three years from release" (Glaser, 1964). Of the 1015 men released, 65% were successful. The prediction problem as defined by Glaser was to "determine how to utilize information available when (the) men were in prison (in order) to divide them into 26 categories with success rates markedly above or below 65% (Glaser, 1964). In order to do this Glaser collected information, on each subject, from their prison “admissions summaries“ and “progress reports.“ The information was then classified into 63 items or predictors. The data consisted of 12 sets of information on the subject’s criminal record; 8 classifications on his family and home background; classification of his work history and drug/alcohol use; and information on his marital history, education, and military record. In addition to the pre-prison data, information reflecting the subject’s behavior and experience while incarcerated was also collected. These variables included: prison assignments, work reports, disciplinary reports, self-improvement activities; communication with outsiders, release plans/arrangements, and psychological, psychiatric diagnoses and test results (Glaser, 1964). The next step in the formation of the "configuration table“ was to establish the extreme boundaries. Glaser set the boundaries at 76.62 and above for the upper boundary and 43.4% and below for the lower boundary. The upper boundary was defined as the favorable risk group and the lower boundary was the unfavorable risk group (Glaser, 1964). 27 The first step in the actual development of the configuration table was to determine which of the 63 predictors ”would place the largest percentage of the sample into a category or categories with success rates outside 43.3 and 76.6 percent ranges“ (Glaser, 1964). The predictor item which differentiated the most between risk groups was, ”Prior Penal Institution Commitment.“ Approximately “one-third of the sample who had no prior institutional commitment had a 77% success rate; the remaining two-thirds had a 59% success rate" (Glaser, 1964). The second step, and the remaining steps in developing the configuration table was to classify each of the two groups differentiated in the previous stages by the remaining predictors. The overall task was to continue to distinguish more extreme risk groups if possible or to place a larger percentage of the sample into the previous extreme risk range (Glaser, 1964) . Qggggciggg_gfi §tgtis§ica1 Methods of Pregiggigg This section will briefly summarize a few studies designed to evaluate the prediction methods outlined above. Babst, Gottfrsdson, and Ballard (1968) compared the statistical techniques of multiple regression and Configural Analysis in developing Base Expectancy Tables for parole prediction. The authors found that: 28 despite the different statistical methods used, the bass expectancy tables which resulted were quite similar. The violation rates for the configural method varied from 20 to 70 percent, while those based on the regression model varied from 20 to 60 percent (Babst, et al., 1968). In 1971 Simon did a comparison of several multivariate methods for combining variables into a prediction instrument. The techniques compared were point scores (Glueck and Burgess method), multiple linear regression (Gottfrsdson method), and hierarchical configurations (Glaser method). Simon’s conclusions suggested that “in practice (the different methods) all worked equally well" (Simon, 1971). .HsliesngQeeenimgsccsstiQLantecs This section will discuss the halfway house in terms of its historical development, theory on recidivism and women offenders. Due to the scarcity of research on the prediction of recidivism from halfway houses, only a small portion of this review will focus on the critique of those studies. fiisteuss...‘ 1 Dexeleeesnt The first documented halfway house, Temporary Asylum for Discharged Female Prisoners," in the United States opened its doors in 1864. "By the early 1920’s, houses widely known as 'Hope Halls’ existed in the states of 29 Louisiana, Ohio, Iowa, California, Florida, and Texas“ (LEAA, 1979). Between the 1930’s and 1950’s, the operation and establishment of halfway houses ceased as a result of "the depression, the expansion of parole, and the pre-release plan requirement that the offender have a job prior to release“ (LEAA, 1979). The National Halfway House Movement was initiated during the 1950’s and increased its popularity in the 1960’s under President Kennedy’s administration. Also during this time period, the International Halfway House Association" was established and delineated guidelines for halfway houses (LEAA, 1979). The impetus behind the halfway house movement was due to a general dissatisfaction with prisons; more specifically with the overcrowding of prisons, shortage of staff, lack of prison programs, and the cost of maintaining an individual in prison. An additional reason for the development of halfway houses can be linked to changes in the correctional theory as it pertained to the treatment of offenders. The correctional model associated with the traditional prison-parole cycle (was) rehabilitation. This model defined correctional workers as therapist and emphasized their ability to "cure” the offender (LEAA, 1979). The community corrections program had as its foundation a different model, that of reintegration not rehabilitation. This model: 30 focuses on the “destructive“ affect of isolating the offender from the community and on the need for transitional programs between the institution and the community. The contention was that it was unrealistic to expect an offender to return directly to the community after a period of incarceration and "make it" (LEAA, 1979). Qgggggitx_gggggction Erggrggg ang the_!gm§n Djfgggg; Prisons, as well as community correction programs (halfway houses), are traditionally designed for male offenders and the- rationale for their existence have been extended to female offenders. However, contrary to this practice, women are known to have needs and concerns that are unique to them as a group as well as some that are similar to their male counterparts (see Miller. 1975; Griffieth, 1980). One major special problem area is the children of (women offenders). when dealing ‘with the woman offender, more is at stake then just the individual herself...Another special problem relates again to the double standard. A man, in release, is an ex-convict, which is a handicap serious enough to establishing a normal adjustment to society. But a female ex-convict is also a “fallen woman" and is therefore, doubly unacceptable in polite society. This negative image is one of the worst that society can bestow and has direct implications for employment processes, possibility of marriage and other avenues to successful reintegration (Miller, 1975). 31 "One of the greatest failings of contemporary corrections pertaining to women is the lack of understanding that on release the typical women offender must perform a dual role, at least initially, of homemaker and breadwinner" (miller, 1975). Qw....x.....it Corrsssismfiggcssngssw ivis Recidivism rates have been used as primary indicators in evaluating the effectiveness of community correction programs. These rates have seldom, as shown in reviewing the literature, been used to develop prediction models for offender release into community correction programs. Therefore, all but one of the studies to be reviewed had the evaluation of correctional programs as their goals; the rationale being that post-release information, as well as pre-release information were collected in building prediction models for this study. There have been two studies which have examined the evaluation of halfway house effectiveness and thereby recidivism. The studies were conducted by the Law Enforcement Assistance Administration (1975) and the researchers, Sullivan, Seigel and Clear (1974). An overview of the Law Enforcement Assistance Administration (LEAA) studies are presented. 32 Of the studies reviewed by LEAA, 35 of them focused on the post-release outcome of residents of halfway houses. The authors separated the 35 studies into three main categories: Quasi-Experimental Design, True Experimental Design, and Non-Experimental Design. Of the 17 studies which used quasi-experimental design, in comparing post-program recidivism rates on the halfway house residents and comparison groups, 11 studies reported that the recidivism rates or criminal behavior assessments of ex-residents were leg; than those of the comparison group (LEAA, 1975). The comparison groups for most of these studies were institutional parolees. Of these 17 studies, three of them indicated that the difference was statistically significant and five studies concluded that there was no statistically significant difference in recidivism rates between groups. For the true experimental designed studies, LEAA found only two halfway house evaluation studies. Both studies found no significant difference in recidivism or failure rates between the experimental and control groups. The remaining studies fell into the category of non-experimental design. The recidivism rates ranged from a low of zero to a high of 43%. The authors noted that these rates should be reviewed with caution given the lack 11f consistent operationally defined outcome criterion of recidivism (LEAA, 1975). 33 The LEAA in their overview did not report the variable for which there was a significant difference in recidivism rates among the community correction residents and the control group (compared group). The reasoning was that few studies reported this type of information. Seiter et al. (1976) also conducted an extensive survey of the literature on inmate aftercare evaluations. The authors found there to be “few valid conclusions capable of being drawn from the data he surveyed“ (Seiter et al,1976). However, the authors did note that education, intelligence, sex, age, -history of drug or alcohol abuse, employment skills, community ties, length of time at halfway house, history of psychiatric treatment, age at beginning of criminal career, number of prior incarceration, and type and length of criminal records were significant predictors of success in residential home programs (Seiter et al, 1976). Overall, "there has been little empirical data to discriminate among the type of offenders who will or will not benefit from halfway house treatment“ (Soldfarb & Singer, 1973). The following studies are representations of those studies measuring community correction program recidivism. The purpose of the study conducted by Hclvor and his colleagues (1979) as outlined by the authors, was to present the results of one year of experience with 211 34 residents that were housed in community release centers located in Manitoba, Canada. Of the total population there was 144 successes (68.2%) and 67 failures (31.8%) (Hclvor, 1979). The authors contrasted a variety of characteristics of those residents that were determined successes against those defined as failures. There were no distinctions in success for the characteristics of age, education, place of residence, offense and prior incarceration. The authors found that drug related offenders were “somewhat more likely to be successful than were all other offenders, while all residents with a prior record of escape or unlawfully at large charges were failures“ (Hclvor, 1979). Additional variables that seemed to differentiate those that succeeded and failed were marital status; 73% of those married, even common law relationships, were successful, whereas 57% of those separated or divorced were failures. Unstable employment history and drug dependence were also associated with failure (Hclvor, 1979). One of the two major criticisms of Hclvor’s study was that there were not any statistical tests of significance used to determine whether the difference between the success and failure groups were indeed significant. Also to be be able to do any valid comparisons between groups, a sample size needed to be much larger than that reported. 35 Vasoli and Fahey (1975) conducted a study on individuals released from a reformatory to a halfway house. The halfway house ”grew out of a collaborative effort involving a university, a large steel corporation, and federal, state, and local agencies“ (Vasoli & Fahey, 1975). The subjects for the study were 77 state reformatory male releases between the ages of 18 and 25. The reformatory residents had to meet certain prerequisites prior to being admitted into the halfway house; those requirements were that he be in good health, have an ID of 90 or above, tenth grade education, and with a criminal background without felonious assaults, sexual offenses, arson, or narcotics. Among the 77 subjects, 16 (212) recidivated (returned to prison) compared to 16% of those released from the reformatory directly onto parole (Vasoli & Fahey, 1975). The authors theorized that the reasons for the recidivism rate being higher for those in halfway houses than for those released directly onto parole had to do with the characteristics of the study sample (Vasoli & Fahey, 1975); thereby indirectly concluding that age, offense, and possibly education were in some way related to one’s outcome. However, there was not any attempt on the part of the authors to evaluate this claim. 36 Moran et al. (1977) studied the factors or characteristics of clients that were associated with success in a private community-based correctional program. The study sample comprised a total of 205 high-risk offenders, of which 111 were men and 94 were women. The variables chosen to be correlated with the outcome criteria were: current age, history of drug abuse, longest stay at a single job, psychiatric history, age at first arrest, average arrests per year, number of months incarcerated, legal status, a test measure of psychological deviance, highest grade completed in school, and IO. The outcome criteria were the highest program level attained, number of consecutive weeks worked at one job/training (or currently employed or in training at the time of termination), and a Judge’s criterion rating of success-failure (no impact, some impact, success) (Moran et al., 1977). The authors used Stepwise Multiple Regression to correlate the predictor variables with the outcome criteria. According to the regression equation, the strongest predictors were the longest stay at a single jOb prior to incarceration, and the highest grade completed in school. There were “secondary predictors" which varied according to race and sex. For males, the important secondary predictors were current age and IQ. Age when first arrested, length of incarceration, age of admission, 37 and psychiatric history were the most important "secondary predictors" for women in the sample (Moran et al., 1977). Two authors, Lambert and Madden (1976) studied the relationship between pre-, intra-, and post-institutional factors in terms of their influence on recidivism (and other measures of community adjustment). The subjects for their study were 338 women sentenced to up to two years at the Vanier Centre for women in Ontario, Canada. The "pre' data included personal and family background information obtained through interviews at time of incarceration and criminal history information from criminal records. Institutional data were provided by both staff and residents; it measured both institutional adjustment and social climate. The ”post“ data or follow-up data consisted of one year follow-up interviews and an examination of criminal records. Only one-half of the sample were interviewed during the post period. The recidivism results indicated that during the first year follow-up, 22% of the sample (n 8 74) were reconvicted (for a new crime) and 2% were reincarcerated for a parole violation. During the second year follow-up, 13% (n I 44) had reconvictions (Lambert & Madden, 1976). It was not evident whether the second year reconvictions were the same women in the previous year reconvictions or a different group of women. 38 The "pre" factors that differentiated between the group that recidivated and those that did not were prior criminality (particularly juvenile), early family problems (instability, criminality, drug/alcohol abuser), serious personal problems (drug/alcohol abuse), unstable history of employment, and race (Indian vs. Hhite) (Lambert & Madden, 1976). This differentiation was determined by the chi-square statistic and p < .05 significance levels. The 'intra-institutional' factors that discriminated between those that recidivated and those that did not were cottage assignment (women in certain cottages did better than others), time spent in the Centre (very short periods at the Centre showed the fastest return rates), and indications of serious misbehavior at Centre (Lambert & Madden, 1976). The cottage assignment results were questionable in that women were not randomly assigned to cottages, but were assigned to cottages according to established criteria. The post-institutional factors that Lambert and Madden (1976) stated differentiated on the outcome variable recidivism were employment situation, financial adequacy, physical health, emotional health, family relationships, and residential adequacy. The authors’ chi-square results did not fully support this conclusion. The chi-square statistic was only significant for the factors of 39 employment situation, family relationships, and emotional health. It was also not clear how these factors were defined. The authors’ final conclusion was that: Of all the experiences during the first year following return to the community, the employment pattern had the most strength in terms of intervening in predicted recidivism 'based on earlier patterns of prior criminality. women with prior criminality were shown to have an overall recidivism .rate of 46 percent, compared to 14 percent among those with no prior criminality. However, among those with prior criminality, who also had stable employment patterns, the rate dropped to 15 percent recidivism (Lambert & Madden, 1976). This result must be viewed with caution: only 32 subjects had very ”stable“ post institutional employment situations. Once the factor of prior criminality was considered the number of subjects falling into the various categories becomes very small. Moczydlowski (1980) evaluated the variables associated with halfway house success. The objective of the study, as outline by Moczydlowski: (was) to determine through a multivariate analysis the variables which might be relevant to the resident’s success or failure...A causal model based on predictive significance (were) constructed to indicate the direct and indirect effects of resident characteristics on program adjustment and performance (Moczydlowski, 1980). 40 Moczydlowski’s study was conducted at a private nonprofit community based correctional agency in Durham, North Carolina. The client referred to the halfway house had an average age of 27 years, a 10.6 average educational level, mostly unemployed, single, and had a past history of drug abuse. Ethnic group representation was approximately equal, with Blacks comprising 52% and Whites 48% of the halfway house population. The average length of time served prior to their halfway house stay was 15 months. In terms of criminal histories, the halfway house residents had an average of 5.6 arrests, first offenders comprised 242, majority were property offenders, and 462 had their first arrest at the age of 16 years. The independent variables chosen were from the offender’s personal, criminal, institutional histories, and halfway house evaluations. Personal history variables were age, employment record, drug/alcohol use, IO, educational level, community ties and family life. Variables included under the criminal history section were number of prior arrests, sentence (present offense), severity of crime committed, history of assaultive behavior, age at first arrest, referral source, institutional adjustment, and longest time served. Halfway house variables were overall evaluations of program behavior: initial, program adjustment and release behavior, length of stay at house, 41 percent time employed, community treatment, and satisfaction with employment (Moczydlowski, 1980). The author found that the variables, alcohol use, severity of crime, age, prior arrests, prior employment, and percent time employed were were significantly related to program ”success.” From these results, alcohol use, prior arrests, and percent time employed were significant predictors of initial adjustment at the .01 level while age was significant at .05 level. At the program level, the initial adjustment, alcohol use, severity of crime, and age were significant at the .01 level. (The) release evaluation was found to be predicted by the degree of conflict in the program, prior arrests, percent time employed, and prior employment. No variable was a significant predictor at all three program levels (Moczydlowski, 1980). gritiggg 9f Research on Rggidixigm An overall assessment of the previous and continual development of recidivism models is its inability to predict. The problem of predicting recidivism can be evaluated in terms of two major categories: theoretical and methodological. Ihggggtigglj The apparent lack of predictive power attributed to prediction models can clearly be traced to a lack of theoretical support. Most researchers fail to report the theoretical basis for selecting their predictor variables. “Without theoretical guidelines, such studies are less likely to select most relevant variables to 42 produce significant results and to contribute to the accumulation of a body of empirically supported knowledge“ (Dean, 1968). A study conducted of Buikhuisen (1973) on the factors related to recidivism showed that ”many factors supposedly related to recidivism in fact (have) little or no predictive power" (Buikhuisen, 1973). In addition to the. lack of theoretical support for study variables is also the failure to integrate, or make use of, items which are linked to the criminological theories related to recidivism. There has been no reference made to ”the labeling process well described in the criminological literature or the ‘impact of more sociologically-oriented theories like differential association, differential opportunity,’ one’s environment, inequity, racism, or classism” (Buikhuisen (1973). aginggglggigglg Perhaps the three most outstanding criticisms of previous recidivism research were the data, outcome measurement, and the statistical analysis of the data. Most of the research conducted in this area has exclusively relied on data compiled by various staff members working with the incarcerated and released offender. Therefore, the quality of the data may vary from one individual or agency/institution to another. Using 43 file data exclusively may not provide the necessary data to construct a prediction model. Most data collected on individuals incarcerated may be questionable in terms of its accuracy. Therefore, “no considerable improvement is possible in this area without a complete change both in the methods of obtaining information...and in the nature of the information obtained" (Warner, 1923). The measurement most widely used as the criterion for research on prediction was (is) the success/failure dichotomy. Success being referred to as “not being returned to prison,“ and failure being ”the return to prison." There usually was not any distinction made between being returned to prison as a technical violator or as having committed a new offense. A judicial or an administrative decision regarding an individual’s return to prison may not be reflective of his/her actual behavior. Therefore, actual behavior instead of the criminal justice staff decisions should be incorporated in the outcome-criterion measurement. A third and final methodological criticism of prediction studies had to do with the use of statistical analysis. Most of the initial studied on recidivism failed to apply statistical tests of significance to their findings. When used, the majority of the researchers had resorted to univariate analysis. Information about the 44 amount of variance explained by the predictors were rarely reported (Buikhuisen, 1973). Even when fairly sophisticated techniques were used, their appropriateness also were in question. whenever a statistical technique is used, it is essential that the researcher have a clear understanding of the assumptions upon which the technique is used....If a technique is applied to a situation for which it was not designed, inferences derived from available data may be misleading and policy decisions based on these inferences may differ from those based on a more appropriate method (Palmer & Carlson, 1976). 45 525.1222;st Paw The theoretical and methodological development of statistical prediction techniques has been, and continues to be, a central concern to criminal justice researchers and theorists. Prediction techniques in criminal justice have not only maintained an important role in theory testing and reformation, program development and evaluation, but also in the area of public policy. In the area of public policy, “statistical prediction techniques have application for (offender) risk assessment“ (VanAlstyne & Gottfredson, 1978). This study will attempt to design a statistical model to predict whether, given a set of characteristics and circumstances, an incarcerated woman offender will “succeed“ in a community correction program. The facets of the research are women offenders, community correction programs, and measurement of recidivism . Most criminal justice researchers and persons working directly with offenders would agree that recidivism rates are disproportionately high during the time period immediately following release from prison. More specifically, some researchers have narrowed this time period down to the first 60 to 90 days after release (Miller, 1975). The community correction program is the correctional agency that is involved with the inmate during 46 this time period, thereby making it an important institution within the criminal justice system. These programs must address themselves to those problems encountered by the offender on her release from the correctional institution. These areas of concern, such as obtaining employment and housing, play an important role in determining whether an offender will make a ”successful” adjustment into her community. Another important reason for the focus on community correction programs is the contention that inmate after-care programs can provide help to the offenders in terms of their reintegration needs at a cost less than that of incarceration. For the State of Michigan, incarcerating a woman offender is $45.00 a day (note: This figure does not include medical costs.), compared to $22.75 a day for her stay in a correction center (Michigan Department of Corrections, 1980). However, some researchers feel that the comparison should be made between the cost of her stay at a halfway house and that of supervising her on parole. This contention is based on the fact that if she were not released into a halfway house, she would be released directly onto parole. For the State of Michigan, the woman offender would be maintained in the prison until her earliest release date; eligibility for community correction center placement is not synonymous to eligibility for parole. 47 The reason for the emphasis on the woman offender is two-fold. The first is that there has been an overall lack of research, particularly quality research conducted using a sample of women offenders. The second is, given the apparent increase in the number of women incarcerated and thereby released from prison in a community correction program, there is an immediate need to attempt to construct a model predicting the risk of this release. The third aspect of this study is its focus on the problem plaguing many correctional researchers and other personnel, that of recidivism. In reviewing the recidivism literature, it becomes evident that there is a failure in models being able to predict recidivism. The models aren’t working due to theoretical, methodological, and/or analytical problems. This study will address each of the problematic areas as they relate to predicting recidivism. In terms of the theoretical problem, the author will provide the theoretical support for each category for which data is collected. The following section presents the theoretical support for the categories of variables included in this study. One of the most critical methodological problems encountered in developing prediction instruments lies in the accuracy of the data source. The questionable validity of archival data cannot not be over-emphasized. However, 48 using alternative data sources (i.e., interviewing family about family background, prison staff about prison behavior, community correction center staff about center behavior, police about arrest situations, as well as talking to the woman offender) was not a feasible option. One way this issue was addressed was through the use of continuous inter- and intra-rater reliability checks to insure the reliability of data coded across and within raters. However, one must bear in mind that, despite the accuracy (or lack of) of the data source, this ggmg information is used by criminal justice employees to make decisions about inmates within the system. There were two predictive designs used to evaluate the relationship between the independent variables and the outcome criteria. The first method, multiple linear regression, considers both the “evaluation and measurement of overall dependents of a variable on a set of other variables” (SPSS, 1975) and examines the effects of a particular variable to the criterion within a multivariate context: an important feature being that the criterion is an interval variable. The second method of analysis used was discriminant function analysis. The equation generated was based on a dichotomous, not on interval measured criterion. This method allowed the researcher to, based on a set of predictor variables, differentiate those women 49 that recidivated from those who did not recidivate. The reason these two methods were employed was because it was not clear, based on prior research, which one would generate the most predictive equation. The question of whether the criterion of recidivism was best represented by an interval scale was addressed. The data collected was divided into four general categories: Pre-Institutional-personal background, criminal history - (juvenile and adult); Institutional-behavior in the institution; and Post-Institutional-behavior at community correction center. The independent variables were those variables in the categories personal background, criminal history, behavior in the institution, and behavior at the community correction center. The independent variables were those variables in the categories personal background, criminal history, behavior in the institution, and behavior at the community correction center. The violations occurring during the twelve-month follow-up period served as the outcome criterion. P r Bac nd The literature has shown personal background factors to be related to post-institutional success/failure. The data indicate that the age of the offender is related to post-institutional adjustment. The younger the offender, 50 the more likely she will recidivate (Burgess, 192B; Tibbets, 1931; Vasoli & Fahey, 1975; Lambert & Madden, 1976; Moczydlowski, 1980). The educational level attained by the offender has been shown to be directly related to post-institutional adjustment. The lower the educational achievement, the more likely the offender will not adjust “successfully“ to the community (Berecochea & Spencer, 1971; Moran, 1977). Also following is the assumption that employment history will affect the post-institutional adjustment of the offender. Offenders that did not have any prior employment history or limited employment histories have tended to have difficulty adjusting to community life (Glueck & Glueck, 1930; Berecochea & Spencer, 1971; Lambert & Madden, 1976; Moczydlowski, 1980). Single offenders and offenders coming from ”broken homes“ tend not to adjust to post-institutional life (Berecochea & Spencer, 1971; Buikhuisen & Hoekstra, 1974; Moran, 1977). A common indicator of post-institutional success has been the offender’s past drug involvement. Extensive, habitual, and/or continual use of drugs/alcohol has had a negative effect on the post-institutional adjustment of the offender (Gottfredson, et al., 1965; Berecochea & Spencer, 1971; Moran, 1977). 51 flxngghgsis 1: The less stable a woman’s personal background (i.e., unemployment, low education, single, young, family criminal involvement, ”broken home,” drug involvement, the more likely she will be to recidivate. This category covers both juvenile and adult criminal behavior. Juvenile involvement has been shown to be indicative of later adult criminal involvement (Glueck & Slueck, 1950; Glaser & O’Leary, 1966; Mannheim & Wilkins, 1955; Gottfredson, et al., 1964). The specifics of one’s criminal history, age of first arrest, length of time incarcerated, number and type of prior offenses, and instant offense have produced the following profile of an "unsuccessful“ post-institutional adjustment: first arrest at an early age, many prior arrests (specifically property offenses), currently serving a sentence for a property offense (Warner, 1923; Burgess, 1928; Glueck 8: Glueck, 1930; Moran, 1977). flxpgghggis 2: The more criminal involvement a woman has had, the more likely she will be to return to prison on her release. WEI-ISM There appears to be little empirical data addressing the issue of the affect institutional behavior has on post- instititutional adjustment. Negative prison adjustment, lack of program participation, and prison misconducts are 52 associated with poor post-institutional adjustment (Lambert & Madden, 1976; Moczydlowski, 1980). Hypothesis-_§: The more negative a woman’s prison behavior accompanied with the lack of prison program involvement, the more likely she will be to recidivate. WMGLEELBAQ: Few researchers have attempted to analyze post- institutional adjustment utilizing those factors/events following release from prison into community correction programs. Factors that have been linked to continued criminality and poor post-institutional adjustment were the inability to locate employment or maintain employment (Gottfredson, 1965; Moran at al, 1977; Fairweather, 1980; Moczydlowski, 1980) and drug and alcohol involvements (Gottfredson, 1965; Miller, 1975; LEAA, 1978). This evidence, along with the overall lack of data in this area, suggests that the impact of post-institutional factors (via community correction programs) on recidivism must not be overlooked. flypgthggig 4: The more negative the offenders’ post- institutional behavior (via community correction programs) (i.e., no community correction center programming), the more likely one will be to recidivate. CHAPTER 2 METHOD §LIQJ£S$§ The research subjects were a 402 sample of all female offenders released from prison to community correction centers in the State of Michigan between and including the years 1975 to 1979. The .40 rate was calculated for each month within a year. A total of 1,202 women were released from prison into correctional centers during this time period. If the offender had more than one community correction center release only one release date was selected (randomly) for inclusion in the study. The total available population became 1,013; thereby resulting in a sample size of 405 subjects. The central office files maintained by the Michigan Department of Corrections were reviewed and used as the data base for the sample. These files consisted of the following types of information/documents: the presentence reports, which contained demographic information on the inmate (i.e., family background, employment history, educational background, and juvenile/adult criminal history); the State Police Record which outlined both the 53 54 arrest and conviction history of the inmate; the Reception and Guidance Center recommendations for the inmate during incarceration; Parole Eligibility Reports, which evaluated the progress of the inmates during incarceration--this progress is evaluated in terms of prison program participation and prison behavior; Disciplinary-board reports; Progress and Violation Reports from both community correction center staff and parole agents. Design There were essentially two predictive designs utilized for this study. The first design examined prison/correction center recidivism in terms of a defined set of pre-prison release variables. The second employed an additional set of post-prison release variables in evaluating recidivism. The predictor variables were divided into the following categories: 1) Personal Characteristics (non-criminal) 2) Criminal History 3) Behavior in Prison, and -prison program participation -prison misconducts 4) Behavior at Community Correction Center -program participation The dependent variable was based on the illegal behavior of the female offender once released to the correction center. The procedures for determining illegal behavior is' addressed under Index of Recidivism. 55 19231:..th Eilet.§tsgx_sng.lust:2esnt.§snstcsstign The variables included in the data collection instrument were determined by an exhaustive review of the literature on prison and community correction center recidivism. In addition, variables were also included and/or eliminated based upon the results of the pilot studies. 5 Two pilot studies were conducted prior to the onset of the study. The first was conducted prior to the development of the instrument. This initial pilot study was used to 1) identify those variables that could be collected from the prisoner’s institutional files and 2) to determine the overall consistency in locating these variables. Approximately 15 files were randomly selected by the principal researcher in order to complete this pilot study. A preliminary list of variables was compiled after this initial review of the files. From this list, a preliminary draft of the data collection instrument was completed. A second pilot study was conducted using the preliminary draft of the instrument, for the purposes of refining the instrument. Data was collected from 50 files chosen at random utilizing the instrument. A final draft of the instrument was completed subsequent to this second pilot study (see Appendix A). 56 Escaensl...sbacsstscistiss 9* SD2_-Qi£§QQSEA This category includes the noncriminal aspects of the offender’s life prior to her incarceration in prison. The emphasis here is on the noncriminal factors of the offender’s life. Distinguishable characteristics such as age, race, marital status, number and custody of children, history of drug addiction, marital status of parents, family criminal background, prior employment and education histories were evaluated. These personal background variables evaluated at the timed of the current (sampled) offense were seen as indicators of social stability (Mannheim &.Wilkins, 1955; Zalba, 1964, 1964; Glaser & O’Leary, 1968; Glaser, 1969; Reeds & Woods, 1971). Occupation at the time of the current offense, number of jobs, and longest length of time of any one job were indicators of the offender’s economic stability (Mannheim & Wilkins, 1955; Berecochea & Spencer, 1972; ‘Irvin, 1974; Vasoli & Fahey, 1975; Moran, 1977). The variables included in this category were: Rac_ - Two codes for race either Black or non-Black. e__at Time of_ Instant Offgggg - The age at the time the subject committed the offense (the sampled offense). The subject’s date of birth and the date the subject committed the offense was used to calculate age. Eggggtigg - The number of full years of education completed at the time of the instant offense. 57 Qggggggign - The type of job held by the subject at the time of the instant offense. The occupation lists/codes developed by Hollingshead and Redlich (1958) was used to code the subject’s occupation. Their occupation lists grouped jobs into seven major categories ranging from higher professionals to unskilled workers. An eighth category was added to account for the unemployed. uggggr Qf Jggs figlg - The total number of jobs held prior to the instant offense. This number included both full- and part-time employment. L299s2s_.Lsngsn_9i.Iiss.§snlsxsg - The longest length of time employed on any gag job prior to instant offense. Baiggg - Who raised the subject, majority of time, prior to adulthood (0-17 years.). The options were: both parents (includes step), one parent, relatives, friends, or institution. figgitgl Staggg_gf_Pa;egt§ - The parents’ marital status during most of subject’s upbringing. the possible codes were: never married, separated/divorced/widowed, married, or common law married. Sibling; - Total number of brothers/sisters (including step). Mggitgl Staggg__gfi__§ggjggg - The subject’s marital status at the time of the instant offense. The codes were: never married (single), common law married, separated/divorced/widowed, or married. thlgggg - Whether the subject had children. Children did not have to be in her custody. nggggx__gi__ggilgggg (1) - Whether the subject had full custody of her children at the time she committed the instant offense. Qggtggy of Chilgcgg_(2) - Whether the subject had full custody of her children at the time she was released from prison. Drug _figgi§gigg - Any history of drug addiction. This addiction was based on the following: heroin, pills (barbituates, uppers, downers, etc.), cocaine, or alcohol. 58 Eggi;x_ Criminal Higtgcx - Any immediate family member (parents and/or siblings) having adult criminal histories. Qcimingl nigtggx. Juvenile crimes and their dispositions as well as the commission of adult crimes as a juvenile were indicative of later criminal activity and the seriousness of later criminal behavior (Glueck & Glueck, 1934; Mannheim & Wilkins, 1955; Gottfredson & Ballard, 1965; Glaser & O’Leary, 1966; Gottfredson, et al., 1974). Adult criminal history was measured in terms of the age of the first adult arrest, number of prior misdemeanor and felony arrests, number of prior non-prison and prison sentences, and the current (sampled) offense. These variables were used to measure the length and seriousness of the offender’s adult criminal involvement (Gottfredson & Ballard, 1965; Glaser & O’Leary, 1966; Berecochea & Spencer, 1972; Gottfredson, et al., 1974; Moran,1977; Moczydlowski, 1980). Variables evaluated were: Ingtgnt Qfifignge - The offense for which the subject was incarcerated. A list of offenses compiled by Michigan Department of Corrections ranked in terms of statutory sentence lengths was used to code the instant offense (Table 4). If the subject had more than two instant offenses, the the two most serious offenses were coded. Lgngth gf Inggggggatigg - The actual number of months spent in prison for instant offenses. Two dates were used: the corrected date (adjusted for time spent in jail prior to court conviction and sentencing) and the date released from prison into the community correction center. 59 Table 4: Criminal Offenses and Statuatory Sentence Lengths Cod. Offense ' Lfigfigzm 3::fzzgf" Explanation of Offenses 1 Murder. 1st degree Life Premeditated. intentional killing 2 Murder. 2nd degree Life or any Murder not premeditated. e.g.. tenm of years bar-room brawls. less than life 3 Attempted murder Assault with intent to do great bodily harm. 4 Assault with intent to commite murder 5 Robbery armed Life or any 6 Assault to rob, armed term of years 7 Rapa “ 8 Kidnapping ' 9 Conspiracy " 10 Bank safe or vault robbery ll Narcotics. unlawful sale. distrib., manufacturing 13.3 yrs. 20 years 11 Burning a dwelling house 13.3 yrs. 20 yaars Threatening a person with injury in order to obtain property. 11 Extortion 13.3 yrs. 20 years 11 Accept earnings of a prostitute pandering 13.3 yrs. 20 years Pimping 12 Robbery. unarmed 10 yrs. 15 years 12 Assault to'rob 10 yrs. 15 years 12 hanslaughtar 10 yrs. 15 years Killing but offender was provoked. Retaliation. 12 Breaking and entering an accupied dwelling 10 yrs. 15 years 12 Sodomy 10 yrs. 15 years Sexual assault (not violent). 12 Perjury 10 yrs. 15 yrs. Lying in a situation when you're under oath to tell the truth. 12 Placa explosive by property with intent to discharge 10 yrs. 15 years 12 Firearm. cause death w/o malice 10 yrs. 15 years e.g.. gun goes off by mistake and someone is killed. 13 Uttaring and publishing 9.3 yrs. 14 years Passing a bad check. 13 Forgary of records 9.3 yrs. 14 years 14 Breaking and entering 6.66 yrs. 10 years 14 Possession of burglary tools 6.6 yrs. 10 years 14 Larceny from a person 6.6 yrs. 10 years Stealing from a person, 2.9.. purse snatching. Bargained down robbery. 14 Assault less than murder 6.6 yrs. 10 years 14 Assault coumitting rape, sodomy. or gross indecency 6.6 yrs. 10 years 14 Assault to commit a felony 6.6 yrs. 10 years Assault with a dangerous weapon. without intent to comit murder, and without intent to inflict great bodily harm. i.e.. less than murder. 14 False pretense to defraud 6.6 yrs. 10 years Falsely obtaining money. goods. or services from an individual. No theft because given articles voluntarily Table 4 -- Continued 60 Statuatory Code Offense :::?;;_ "a‘imu_ Explanation of Offenses l4 Indecent liberties with child 6.6 yrs. 10 years 14 Burning other real property 6.6 yrs. 10 years 14 Drunk driving-third offense 6.6 yrs. 10 years 14 Possession of a stolen auto 6.6 yrs. 10 years 14 lncent 6.6 yrs. 10 years 15 Non-narcotic drug. illegal sale. distribution 4.66 yrs. 7 years 15 Hallucinogens. sales. distri.. and manufacturing 4.66 yrs. 7 years 16 Escape from prison 3.33 yrs. 5 years 16 Carrying a concealed weapon 3.3 yrs. 5 years 16 Receiving stolen property 3.3 yrs. 5 years Commonly called attempted BBE. Usually bargained down from ass. 16 Unlawful driing away auto 3.3 yrs 5 yrs. 16 Larceny over $100 3.3 yrs. 5 years 16 Larcenv from motor vehicle 3.3 yrs. 5 vrs. 16 Larceny by conversion over $100 5 years Receiving money. goods. or other property and wrongfully applying it to a purpose other than that for which it was delivered to him. e.g. defendant given funds to buy stock for someone but uses money to buy himself a car. 16 Attempted gross indency between male and female 3.3 yrs. 5 years 16 Carrying weapon w/unlawful intent 3.3 yrs. 5 years 16 Possession of forged notes 3.3 yrs. 5 years 16 Transport drugs into prison 3.3 yrs 5 years 16 Mfg. or poss. illegal weapon 3.3 yrs. 5 years 16 Possession of bomb 3.3 yrs. 5 years 16 Common law offense 3.3 yrs. 5 years 16 Gross indecency between females 3.3 yrs. 5 years 17 Larceny from a building 2.6 yrs. 4 years 17 Felonious assault 2.6 yrs. 4 years Hitting a person 17 Nargotic drugs. possession of 2.6 yrs. 4 years 17 Intent to sell or use credit cards 2.6 yrs. 4 years 17 Marijuana. illeg. sale. distr.. mfg. 2.6 yrs. 4 years 17 Mel. dest. property over $100 2.6 yrs. 4 years 17 Burning of personal property 2.6 yrs. 4 years 17 Prepare to burn property over $50 2.6 yrs. 4 years 17 Sale or use of credit cards 2.6 yrs. 4 years 17 Cruelty to children 2.6 yrs. 4 years 17 Mel. dest. heuse. barn. other bldg. 2.6 yrs 4 years 17 False statement to obtain relief over $500 2.6 yrs 4 years 17 Larceny of livestock 2.6 yrs 4 years 17 Theft of credit cards 2.6 yrs. 4 years 17 Obscounding or forfeiting bond 2.6 yrs. 4 years 18 U.D.A.A. w/o intent to steal 1.3 yrs 2 years Joy-riding. 18 Checks w/o account or suff. funds 1.3 yrs 2 years Checks that bounce. 18 Non-narcotic drug possession 1.3 yrs 2 years 18 Resisting or obstructing officer 1.3 yrs 2 years 18 Negligent homicide 1.3 yrs 2 years Death due to reckless driving 18 Careless use of firearms 1.3 yrs 2 years 18 Larceny of rented motor vehicle under $100 1.3 yrs. 2 years 18 Felonious driving 1.3 yrs. 2 years 19 Misdemeanor .66 yrs 1 year SOURCE: Michigan Department of Corrections. 61 flughgr of Stgtg§__giigggg - Number of status offenses committed (as juvenile) (i.e., runaway, truancy, incorrigibility, curfew). ene_.et First iexsnils..Eslgnx-eccses - Age of subject when arrested for first felony offense as a juvenile. Buses: of Juvenile..Eelan..ecce2§s - Actual number of arrests for a felony offense as a juvenile. 593$..Ssci99:..9235993..9£ a .ésxsoils..ficcss§ - Host serious disposition to any juvenile arrest. The possible codes were: warned/released, probation, or institutional placement. ens..es_Eics$_eeuls.ecceet - Ace when first arrested for any crime as an adult (17 years of age or older). 8:19;..egsls..fliseeseennc..eccesta - "both-r subject had been arrested for a misdemeanor offense. Misdemeanor offenses were defined by the type of offense and the length of sentence, less than one year in jail. this variable was delineated by type of offense: property (i.e., shoplifting), person (i.e., simple assault), drug (i.e., drug use), and other (i.e., soliciting). flushes..e£_.Eciec._e§uls._EeL9nx..flccss§e - The actual number of prior adult felony arrests. Felonies were defined by the length of sentence, one or more years incarcerated and/or on probation. This variable was delineated by type of offenses property (i.e., breaking and entering), person (i.e., armed robbery), drug (i.e., sale), and other (i.e., arson). flgghgc__of Ngg:ficiggn__5engggcgg - The actual number of non-prison sentences included any of the following dispositions: jail, probation, treatment program, restitution, or any disposition short of prison incarceration. 39593; of Prigg Prison Sggtgnggg - The actual number of ------- --C: times incarcerated in prison for a new offense. uggbg;__gfi__gnigg Michiggg Prison Sentences - The actual number of times incarcerated in a Michigan prison for a new offense. 62 Bebaxigc..dscino.-insecssce§isn- The emphasis here was on the two aspects of prison life: prison program involvement and institutional adjustment. Program involvement was measured by the type of program participation encourages successful community adjustment (Glaser, 1964). Number of minor/major prison misconducts and the dispositions to these violations (i.e., toplocks and involuntary segregations) have been shown to be measures of institutional adjustment and later community adjustment (Lambert & Madden, 1976). The category variables were: Eniggn 85995;; Pgrtigingtigg - whether the subject participated in any type of prison program. The type of programs were: educational, vocational, counseling, other (i.e., church choir, jaycettes), and daily routine work. Eciggn Migggnggggg - Actual number of misconducts subject received while incarcerated. The misconducts were of the following type: major (i.e., assaults) or minor (i.e., contraband, two in a room). giggggitiggg to thg Migggggggtg Prison dispositions included toplocks (room detainment), involuntary segregations (detained in a segregation unit for disciplinary reasons) and voluntary segregations (detained in a segregation unit for protective reasons). fishi!iQE..éS.SQQEQQiS¥_EQ£ESESi99-929;SE; This category includes not only those variables used as predictors but also the variables included in the Index of Recidivism. A few researchers have studied indirectly, the affect 63 behavior during correction center placement has on one’s ability to “make it” (see E. Eugene Miller, 1975; Vasoli & Fahey, 1975: Moran, 1977; LEAA, 1978; Moczydlowski, 1980). This category evaluated three variables: Egggunity Qgggggtigg_gggtgg - The name of the community correction center was recorded. Soccestisn..Qente:..£:99:es_.lnxglxeoent - The type of program involvement while residents at the correction center. The program types included the following: job, on-the-job training, educational, and counseling. Dgys _g;__§gggggigx Cgcregtion Center - Total number of days at the community correction center. The dates used to calculate this variable were the date released from prison into the community correction center and the date released from the community correction center. Inggx 9f regigivigg. The Index of Recidivism measures the seriousness of post-prison release criminal involvement. The violations were coded for a period of twelve months. from the time the offender was released into the correction center. Therefore, in addition to a violation being coded while the offender was a resident at the community correction center, a violation could also be coded while the offender was on parole or in prison, as long as the violation occurred within this twelve month time period. The first five violations were coded. If more than five violations occurred during this follow-up period, the most serious violation was coded, the Michigan Department of Correction Offense codes were used. These 64 offenses were ranked in terms of seriousness according to the minimum and maximum sentences (see Table 4). Qgtg_-Qi__giglgtigg. The actual date the violation occurred was recorded. giglggigg - The type of violation was recorded. The violation was recorded in any of the following categories: (1) no illegal activity (no violations for the 12-month period): (2) traffic violations (i.e., parking tickets); (3) center rule violations (i.e., alcohol use, tardiness); (4) prison rule violations-minor (only if subject was returned to prison during the 12 month follow-up period) (i.e., two in a room, insolence); (5) prison rule violation-major (only if subject was returned to prison during the 12 month follow-up period) (i.e., assault); (6) misdemeanors - property, prison, drugs, other; (7) felony - property, prison, drugs, other; (8) absconder (can’t locate) Qiggggitiogg - The disposition to each violation was recorded. The possible dispositions are listed below: (1) no violation; (2) dismissed, no official action taken (i.e., handled by correction center staff): (3) dismissed, official action taken (e.g., law enforcement officials) (4) toplock - prison misconduct; (5) segregation - prison misconducts; (b) probation: (7) jail confinement; (8) returned to prison; (9) absconder, can’t locate. Numerical codes were identified for coding purposes only; seriouness of violation or disposition was not implied by the above codes. 65 DQSLQQLLSEELQ'LEEQESQQESS the instrument carefully over a period of two days. At the end of this two-day period, raters were introduced to the files from which the data were to be collected. By the end of this training period each rater had coded one file together with the primary researcher and an additional file on their own. Problems regarding the coding and the variables were discussed at this time. The final phase of the training included the coding of five files each by all raters. Comparisons of results from each rater was compared. Discrepancies as well as agreements were discussed. A simple percent agreement reliability check was computed for the five files across raters. An inter-rater reliability of 90% agreement was accomplished prior to the collection of the study data. Throughout the data collection both random inter-rater and intra-rater reliabilities were conducted. Once a week students were required to code at least one case previously coded by another student. The inter-rater reliability checks range from a low of .75 to a reliability of 1.00. The average inter-rater reliability was .95 across all raters. The intra-rater reliability was determined by having each student, twice a month, recode a case that they had previously coded. These intra-rater reliabilities ranged from .85 to 1.00; with a mean overall intra- 66 reliability being .98. All discrepancies in coding, as depicted by the actual reliability score, were discussed and resolved in group sessions. CHAPTER 3 RESULTS To facililtate an accurate and meaningful representation of the data, the results will be presented in the format that the analyses were approached. An overview of the descriptive statistics will be presented in terms of the female offender’s personal characteristics, criminal history, prison and community correction center behavior. The second half of this section will introduce the data reduction techniques utilized, BCTRY Cluster Analysis (Tryon and Bailey, 1965) and Principal Components Analysis (SPSS, 1975), followed by both multivariate and discriminant function analyses. Beanies—LEM“ 5 i is: Due to the relative lack of data on females incarcerated in both prison and community correction centers, a generous portion of this section will be devoted to the presentation of the descriptive findings. To facilitate the presentation of the data , all variables have been assigned to five categories: Demographic Characteristics, Criminal History, Instant Offense, Behavior in Prison, and Behavior at Community Correction Center. 67 68 Eggggcgghic ghggggterisgigs - flgn Qgiming; From a sample of 405 female offenders incarcerated in prison and released into community correction centers definite patterns about their background emerged (see Table 5). The majority of the group sampled tended to be Black (69.4%) and young (x age I 25.75). Most women sampled were not high school graduates, the average attained educational level was 10.1 grades. Employment history was evaluated by three variables: occupational status at the time of the offense, total number of jobs held, and longest length of time employed at any one job. It appeared from the data that most women (84.2%) in the sample were unemployed at the time of the offense and had limited prior job experience. The average number of jobs held was less than two and the longest length of time employed on any one job was less than one year. The marital status represented most by the sample was single (47.9%). However, at the time of the offense most women (87.2%) were either single or not living with a spouse. Most women had children (79.5%) with approximately one-third of these women not having custody of their children at the time of incarceration: an additional 42 (13%) lost custody of their child(ren) while incarcerated in prison. In terms of the women offender’s upbringing most (92.1%) were raised by one or both parents and had approximately four other siblings 69 Table 5: Frequencies of Personal Characteristics VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY Age (at time of instant offense) 17-19 49 12.1 20-24 158 39.9 25-29 117 28.9 30-34 37 9.1 35-39 23 5.7 40-44 11 2.7 45-49 7 1.7 50-54 2 .5 55-59 1 .2 Race Black 281 69.4 Nonblack 124 30.6 Educational Level 0-6 18 4.4 7-9 129 31.9 10-12 235 58.0 College 23 5.7 Occupation (at time of instant offense) Professional 0 0 Semi- Professional 1 .2 Clerical 11 2.7 Skilled 6 1.5 Semi-Skilled 26 6.4 Unskilled 20 4.9 Unemployed 341 84.2 Number of Jobs Held 0 78 19.3 1 117 28.9 2 89 22.0 3 or more 121 29.9 Longest Length of Time on Any One Job 1 year 248 61.2 1 year 78 19.3 2 years 36 8.9 3 years 43 10.6 Who Raised? Both Parents 242 59.8 Single Parent 119 29.4 Relatives 37 9.1 Friends 4 1.0 Institution 3 .7 70 Table 5 (con’t): Frequencies of Personal Characteristics VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY Marital Status of Parents Married 257 63.5 Separated! Divorced] Widowed 116 28.6 Never Married 26 6.4 Other 6 1.4 Number of Siblings 0 29 7.2 1 , 49 12.1 2 50 12.3 3 or more ' 277 68.4 Subject’s Marital Status Married 51 12.6 Separated! Divorced/ Widowed 151 37.3 Never Married 194 47.9 Other 9 2.2 Children Yes 322 79.5 No 83 20.5 Custody of Children (at time of instant offense) Yes 224 55.3 No 98 24.2 Not Applicable 83 20.5 Custody of Children (at time of prison release) Yes 184 45.4 No 140 34.6 Not Applicable 81 20.0 History of Heroin Addiction Yes 234 57.8 No 171 42.2 71 Table 5 (can’t): Frequencies of Personal Characteristics VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY Hi story of Pi l l Addiction Yes 51 12.6 No 354 87.4 History of Cocaine Addiction Yes 11 2.7 No 394 87.4 History of Alcohol Addiction Yes 43 10.6 No 362 89.4 Family Criminal History Yes 145 35.8 No 260 64.2 72 in the household. Of these family members approximately one-third had contact with the criminal justice system. In the incidence of drug usage, more than half of the women sampled reported a history of heroin addiction. Much less than one-third reported addiction to alcohol and/or non-opiate (pills/cocaine) drugs. chxiggg Qcigingl Higgggx Criminal history prior to the incarceration for the instant offense was evaluated both in terms of juvenile and adult criminal justice involvement (Table 6). Few women (30.9%) had reported juvenile criminal justice contacts. Only 24.7% of those sampled were charged with status offenses and approximately 10% were charged with felony offenses. Of those with juvenile delinquency contacts, approximately 221 were adjudicated. In terms of adult arrests, women in the sample tended not to be arrested at an early age. The average age of first adult arrest was 21 years. The arrest pattern of the women included, for misdemeanor offenses, a high incidence of property offenses (51.62) (i.e., shoplifting, larceny under $100, petty larceny) followed by a fairly high (39.3%) representation of other misdemeanor offenses (i.e., disorderly conduct, prostitution). Few women were charged with either misdemeanor person (i.e., assaults) or drug offenses. For Tal iii M “9: Orr: 73 Table 6 : Frequencies of Criminal History VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY Age at First Juvenile Arrest 5-9 1 .2 10-14 11 2.7 15,16 27 6.7 No Arrest 366 90.4 Number of Juvenile Felony Arrest 0 366 90.4 1 22 5.4 2 8 2.0 3 or more 9 2.1 Number of Status Offenses 0 305 75.3 1 52 12.8 2 17 4.2 3 or more 31 7.7 Most Serious Outcome of Juvenile Arrest No Arrest 280 69.1 Warned/Released 35 ‘ 8.6 Probation 34 8.4 Institution/ Out of Home Placement 56 13.8 Age at First Adult Arrest 17-19 204 50.4 20-24 130 32.1 25-29 41 10.1 30-34 12 3.0 35-39 11 2.7 40-44 3 .7 45-49 3 .7 50-54 1 .2 Prior Misdemeanor Arrests -Property Yes 209 51.6 No 196 48.4 ~Person Yes 43 10.6 No 362 89.4 74 Table 6 (con’t): Frequencies of Criminal History VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY -Drug Yes 39 9.6 No 366 90.4 -Other Yes 159 39.3 No 246 60.7 Prior Felony Arrests -Property 0 196 48.4 1 85 21.0 2 54 13.3 3 or more 70 17.3 -Person 0 353 87.2 1 35 8.6 2 10 2.5 3 or more 7 1.7 -Drug 0 355 87.7 1 36 8.9 2 12 ‘ 3.0 3 or more 2 .5 -Other 0 376 92.8 1 19 4.7 2 9 2.2 3 or more 1 .2 Number of Prior NonPrison Sentences 0 99 24.4 1 68 16.8 2 63 15.6 3 or more 175 43.2 Number of Prior Michigan Prison Sentences 0 339 83.7 1 53 13.1 2 10 2.5 3 3 .7 Number of Prior Prison Sentences 0 327 80.7 1 57 14.7 2 15 3.7 3 or more 6 1.4 75 the more serious felony offenses, the arrest patterns mirrored those of the misdemeanor classification. Again, women in this sample tended to show higher incidences of arrest in felony property offenses (51.6%) (i.e., breaking and entering, larceny). Even though more than half of those women had at least one felony property arrest, their criminal behavior was not extensive: the mean number of felony property arrests was 1.32. Also, the apparent lack of prior criminal involvement was reflected in the low number of non-prison sentences (x 8 2.83) and prior prison sentence convictions (x a 2.74). n f s . The instant offense refers to the offense(s) sampled for the study. Most of the women in the sample were incarcerated for property offenses (Table 7). The offenses for which the highest percentage of women were serving time included: larceny from a building (24%): armed robbery (102): violation of drug law (9%): and uttering and publishing (8%). An interesting finding was that approximately 23% of those women sampled were serving time for more than one offense. Mb V' “mess Once convicted and incarcerated the women in this sample were evaluated in terms of her behavior in prison. The two behaviors of interest were prison program participation and prison misconducts. Approximately 802 of Table 7: Frequencies of the Most Serious Instant Offense 76 ABSOLUTE RELATIVE’ CODE INSTANT OFFENSE FREQUENCY FREQUENCY 01 Murder, lst Degree 0 --- 02 Murder, 2nd Degree 15 3.7 03 Attempted Murder 1 0.2 04 Assault with Intent to Commit Murder 5 1.2 05 Robbery, Armed 39 9.6 06 Assault to Rob, Armed 9 2.2 07 Rape O --- 08 Kidnapping 0 --- 09 Conspiracy l 0.2 10 Bank Safe or Vault Robbery 1 0.2 11 Narcotics,Unlawful Sale, Distribution,Manufacturing 37 9.1 12 Burning a Dwelling House 1 0.2 13 Extortion O --- 14 Accept Earnings of a Prostitute.Pandering O --- 15 Robbery, Unarmed 12 3.0 16 'Assault to Rob 2 0.5 17 Manslaughter 19 4.7 18 Breaking and Entering an Unoccupied Dwelling 1 0.2 19 Sodomy 0 --- 20 Perjury 0 --- 21 Place Explosive by Property with Intent to Discharge 0 --- 22 Firearm,Cause Death without Malice 1 0.2 23 Uttering and Publishing 34 8.4 24 Forgery of Records 3 0.7 25 Breaking and Entering 8 2.0 26 Possession of Burglary Tools 0 --- 27 Larceny From a Person 15 3.7 28 Assault Less Than Murder ‘ 4 0.9 29 Assault Committing Rape, Sodomy,or Gross Indecency O --- 30 Assault to Commit a Felony O --- 31 False Pretense to Defraud, Embezzlement 18 4.4 32 Indecent Liberties with Child O --- 33 Burning Other Real Property 3 0.7 34 Drunk Driving -Third Offense O --- 35 Possession of a Stolen Auto 0 --- 36 Incest O --- 37 Non-narcotic Drug,Illegal Sale, Distribution 0 --- 38 Hallucinogens,Sale,Distribution, and Manufacturing 0 --- 39 Escape From Prison 5 1.2 40 Carrying a Concealed weapon 16 3.9 77 Table 7 (can't): Frequencies of the Most Serious Instant Offense ABSOLUTE RELAT CODE INSTANT OFFENSE FREQUENCY FREQUENCY 41 Receiving Stolen Pr0perty 11 2.7 42 Entering without Breaking 3 0.7 43 Unlawful Driving Away Auto 1 0.2 44 Larceny Over $100 15 3.7 45 Larceny From Motor Vehicle 0 --- 46 Larceny By Conversion Over $100 2 0.5 47 Attempted Gross Indecency Between Male and Female 0 --- 48 Carrying weapon Hith Unlawful Intent 2 0.5 49 Possession of Forged Notes 0 --- 50 Transport Drugs Into Prison 0 --- 51 Manufacturing or Possession Illegal Weapon 0 --- 52 Possession of Bomb 0 --- 53 Common Law Offense O --- 54 Cross Indecency Between Females 0 --- 55 Larceny From a Building 96 23.7 56 Felonious Assault 3 0.7 57 Narcotic Drug,Possession of 4 0.9 58 Intent to Sale or Use Credit Cards 1 0.2 59 Marijuana,lllegal Sale, Distribution,Manufacturing O --- 6O Malicious Destruction 0 --- 61 Burning of Personal Property 0 --- 62 Prepare to Burn Property Over $50 2 0.5 63 Sale or Use of.Credit Cards 2 0.5 64 Cruelty to Children 2 0.5 65 Malicious Destruction House, Barn,0ther Building 0 --- 66 False Statement to Obtain Relief Over $500 0 --- 67 Larceny of Livestock 0 --- 68 Theft of Credit Cards 3 0.7 69 Absconding or Forfeiting Bond 0 --- 7O UDAA without Intent to Steal O --- 71 Checks without Account or Sufficient Funds 7 1.7 72 Non-narcotic Drug Possession O --- 73 Resisting or Obstructing Officer 0 --- 74 Negligent Homicide O --- 75 Careless Use of Firearms 1 0.2 76 Larceny of Rented Motor Vehicle Under $100 0 --- 77 Felonious Driving 0 --- 78 these women participated in some type of prison program (Table 8). Most women participated in educational (59.5%) (i.e., BED, College) and counseling programs (44%) (i.e., SHAR-drug counseling, group counseling). The lowest incidence of program involvement occurred in the vocational area (14.82). More than half of the women in the sample were involved in some type of prison work assignment (55.82). In terms of prison misconduct behavior, less than one-third had one or more minor prison misconducts (i.e., lying to officer, two in a room, contraband) and approximately 162 of those sampled had reported major prison misconducts (i.e., fighting, weapons). Very few women in this sample were involuntarily segregated as a disposition to a misconduct during their stay at prison: approximately 10% were segregated away from general population and 17.8% were restricted to their rooms. Weefixfiem 1 Ben er. Behavior at the community correction center was evaluated by both program involvement and misconduct/violations. The average length of time women in this sample spent at the center was 4.9 months. Surprisingly, most women (692) had at least one job while at the center. Other programs in which women participated included job training programs (23.5%): counseling programs, typically drug/alcohol counseling (31.92), and 79 Table 8 : Frequencies of Behavior in Prison VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY Program Participation -Educational Yes 241 59.5 No 164 40.5 -Vocationa1 Yes 60 14.8 No 345 85.2 -Counseling Yes 177 43.7 No 228 56.3 -Other (i.e. choir) Yes 100 24.7 No 305 75.3 -Routine Work Yes 226 55.8 No 179 44.2 Number of Prison Misconducts -Minor 0 281 69.4 ‘ 1 48 11.8 2 18 4.4 3 or more 58 14.3 -Major 0 342 84.4 1 46 11.4 2 7 1.7 3 or more 10 2.5 -Major (fighting) 0 379 93.6 1 24 5.9 2 1 .2 3 or more 1 .2 Number of Involuntary Segregations 0 363 89.6 1 35 8.6 2 2 .5 3 or more 5 1.2 Number of Toplocks 0 333 82.2 1 41 10.1 2 12 3.0 3 or more 17 4.2 Number of Voluntary Segregations 0 404 99.8 1 1 .2 80 educational programs (16.52). The extent to which the women participated in the various programs was not evaluated. (see Table 9). Within one year of being released into the correction center, approximately two thirds of the women sampled had at least one violation. The average number of days between being released into the correction center and the first violation was 63.5 days. During the one year follow-up period, the most common violations engaged in by the woman offender were community correction center rule violations (i.e., tardiness, alcohol consumption) and misdemeanor offense behavior (includes property, person, drugs, 'other’ misdemeanor offenses) (see Table 10). On the other hand, the least common violations committed during this time period were traffic infractions, prison misconducts (if woman was returned to prison during this 12 month follow-up violations continued to be recorded), and parole rule violation (i.e., unemployment, failure to report) (if woman was on parole during the follow-up period violations continued to be recorded). The low number of prison and parole rule violations was a function of the length of time women spent in these conditions. 81 Table 9 : Frequencies of Behavior in Community Correction Center VARIABLE CODE ABSOLUTE RELATIVE FREQUENCY FREQUENCY Community Correction Center-Detroit Yes 223 55.1 No 182 44.9 Community Correction Center Program Participation -Job Yes 280 69.1 No 125 30.9 -On the job training Yes 95 23.5 No 310 76.5 -Counseling Yes 129 31.9 No 276 68.1 -Educational Yes 67 16.5 No 338 83.5 Number of days at Community Correction Center 30 57 14.1 30 60 57 14.1 60 90 72 17.8 90 120 49 12.1 120 180 35 8.6 180 135 33.3 Table 10 : EIQLQIIQU§ Traffic Center Rule Prison Misconducts Parole Rule Misdemeanor Offense Felony Offense Escapee/ Absconder ;_n/ 1 ( .3) 152 (55.9) 0 ( 0.0) 3 ( 1.1) 70 (25.7) 11 ( 4.0) 35 (12.9) 2 n5; 2 ( .9) 124 (58.2) ( 0.9) ( 0.9) 55 (25.8) 15 ( 7.0) 14 ( 6.6) 272 213 82 Frequencies of Post Prison Violations By Type § "/1 4 "/Z §_e£Z é_n£Z 2 ( 1.2) 98 (56.6) ( 2.3) ( 1.2) 42 (24.3) 12 ( 6.9) 13 ( 7.5) 2 ( 1.5) 72 (55.0) 11 ( 8.4) 1 ( 0.8) 28 (21.4) ( 6.1) q ( 6.9) 173 131 1 ( 1.0) 50 (47.6) 8 ( 7.6) 1 ( 1.0) 30 (28.6) 7 ( 6.7) ( 7.6) 105 0 ( 0.0) 13 (15.9) (30.5) 14 (17.1) 83 In terms of dispositions to these offenses, the most common dispositions given to the violations were either dismissed or returned to prison (see Table 11). Probation and jail confinement were the dispositions used most infrequently. It appears from looking at the tables that the more violations committed by a woman, the less chance that the violation would be dismissed. However, multiple offenses/violations did not interpret to more severe dispositions. Of the first violations approximately 811 were dismissed as compared to 70.52 of those having committed five violations. On the other hand, if there were six or more violations the most serious of these was usually not dismissed but were returned to prison (48.8%). Women having one violation tended to commit more: there was only a 12% decrease in the number of women having one violation and those having six or more violations. Most women (64.9%) were favorably released from the community correction center. Approximately one-third of the sample was either returned to prison, jail, or on absconder status (unable to locate). 921.23.293.45 £99. The next step in the analyses was the use of a data reduction techniques. The basic principles underlying these procedures were (1) the reduction of a set of NIIII!III9‘1.Q:‘ N ..~.s..~ 84 mm Ao.c V A~.m V .w.m.V o. AN.s V A~.~ V Ao.H~V a A~.-V cg A~.~HV N\z a an: J0 sno;uas 150w Bugugema oi uotigsodsga mod Ao.o V Am.~ V A~.m_V es Ae.m V Ao.o V Ae.m V 2m.m V as 28.3cV N\z "0119101A 419 01 HOLIISOdSLG “mg Ao.c V Am.~ V A~.m V N_ As.m V Ac.o V Am.m V Am.~HV my Ae.ch ca N\z UOIIPLOLA 41v 0; uotitsodsga mmfi Ac.c O . 25.2 V Am.eHV mm so.. V A~.s V Am.~ V 2N.m V es A~.~cV a\z “OLIPlOIA pds 01 naiiisodSLa «MN Am.o V Am.o V Am._oV .N Am.m V 2o.o V Am.o V Am.-V em .m.seV a\z "0149101A P02 01 U011150d510 NAN Am.o V H Am.m V m Am.m V AN Am.m V m A~.o V N Ao.o V o A¢.HsV gm Ao.HeV a\z “OllPLOIA 151 01 uogigsodsta =o.»..oam_a oz .gbama msumum concouma< com_ca P_me :o_ucnoca xuopaoe \copummmcmmm cmxmh :o_uu< P.Vu_eeo .em.._e._o :mxmh eoesue _._u_eeo oz .e..m_e._o szVPHmomm_o :oVHVmcampo so was» so m=c_u.moqmpo com_ca «mos so mmsucmzcmeu "me open» 85 variables to a limited number, and (2) an extraction by these factors in such a way as to insure either, that a factor is independent, uncorrelated with all other factors or in the case of the BCTRY Cluster Analysis factors / clusters may be intercorrelated. There were two data reduction techniques used to evaluate the data sets. Principal Components Analysis (SPSS, 1975) with varimax rotation generated orthogonal dimensions or components for both data sets. The BCTRY Cluster Analysis program (Tryon and Bailey, 1965) when applied to both data sets, with the dependent variable included, resulted in non-orthogonal clusters (oblique rotation was employed). The basic principle underlying both processes was the reduction of a set of variables to a limited number. The results of the BCTRY Cluster Analysis are presented first followed by the results of the Principal Components Analyses. The V-Analysis was performed using all variables (pre and post data sets). The results of this V-Analysis are found in Table 12. There were 11 clusters generated. A number of variables did not load on a cluster: in particular the dependent variable was excluded (Note - Only one dependent variable was included in the cluster runs due to the high correlations among the six dependent variables. The dependent variable incorporating the 86 Table 12: Cluster Analysis and Cluster Loadings Cluster 1 - No Custody of Children Mecieele L._Qing Custody of Children at Time of Incarceration .9248 Custody of Children at Time of Prison Release .7585 Children ? .6054 Cluster 2 - Prior Prison Incarcerations and Arrests 21199;: Legging Prior Prison Sentences .8399 Prior Michigan Prison Sentences .7346 Prior Felony Property Arrests .6322 Prior Non-Prison Sentences .6198 Instant Offense .6198 Prior Misdemeanor Property Arrests (delete) Cluster 3 - Juvenile Delinquency Contacts Mecisels ‘ Leasing Most Serious Outcome to Juvenile Arrest .8838 Age at First Arrest -.6079 Number of Status Offenses .5775 Age at Instant Offense (delete) Number of Juvenile Felony Arrests .4438 Cluster 4 - Prison Misconducts and Dispositions Vecieele Leading Prison Misconducts - Major .8222 Prison Involuntary Segregation .7112 Prison Misconducts - Fighting .6210 Prison Misconducts - Minor .5529 Prison Toplocks .5389 Length of Time in Prison (delete) 87 Table 12 (can’t): Cluster Analysis and Cluster Loadings Cluster 5 - Prior Felony Arrests Vecieble Leasing Prior Felony Arrests - ‘Other' .8220 Prior Felony Arrests - Drug .6823 Prior Felony Arrests - Person .6186 Cluster 6 - Parents Not Married Eecisels seeding Marital Status of Parents -.7835 Who Raised ? .6711 Cluster 7 - Prison Program Participation Eecieele Legging Prison Program Participation .7820 Prison Program Participation - Educational .6245 Prison Program Participation - Routine Work .4920 Length of Time in Prison -.4199 Prison Program Participation - Counseling .3919 Prison Program Participation - ‘Other’ .3321 Cluster 8 - Older Women Eecieele .eedi.s Age at Instant Offense .6301 Age at First Arrest (delete) Marital Status .3406 Longest Length of Time Employed .3094 Cluster 9 - Prior Misdemeanor Arrests Variable ngg;gg Prior Non-Prison Sentences (delete) Prior Midemeanor Arrests - Property .4312 Prior Prison Sentences (delete) Prior Michigan Prison Sentences (delete) Prior Misdemeanor Arrests - ‘Other’ .3055 88 Table 12 (con't): Cluster Analysis and Cluster Loadings Cluster 10 - 2951991: Legging Prison Toplocks (delete) Prison Misconducts - Minor (delete) Cluster 11 - Stable Employment History Vecieele Leading Occupation Type Prior to Incarceration -.4812 Longest Length of Time Employed .4524 Number of Jobs .3601 Highest Grade Completed at Time of Instant Offense .3243 89 following outcome was used as the dependent variable: return to prison or return to jail or absconder status. The relationship between the predictors and the dependent variable are found in Table 13. Given the V-Analysis results it is not surprising that the dependent variable does not significantly correlate with any of the 11 clusters. Given these results, O-type Analyses were not performed and the analyses were continued using orthogonal factors with the expectations that these factors would lend themselves to the predictive model development. The predictive models were therefore developed utilizing a set of orthogonal factors generated using the SPSS subprogram Principal Components. The Principle Components data reduction technique results will be presented for two sets of independent variables. The first computer run included the personal characteristics, previous criminal history, instant offense history, and behavior in prison variables. The second computer run included those variables listed above and the 'behavior at the community correction center’ variables. For simplicity, the first run is referred to as the “Pre-Prison" and the second the “Post-Prison.: 90 Table 13 : Correlation of All Independent Variables With the Outcome Variable Independent Outcome Independent Outcome 2.4.91.” ' a e Mecieele 223213.12 Essie-.912 MPRSN -.140 MISCMN .098 RACE .072 MISCMJ .002 AGE .143 MISCFG .010 EDUC .060 SEGINV .113 OCCUP -.186 SEGVOL -.008 JOBS .063 TOPLOCK .120 ETIME .130 DETROIT -.006 RAISED .037 CCJOB -.257 PMARTL .006 CCOJT -.063 SIBS -.014 CCONSL .144 MARTL .068 CCEDUC -.102 CHILD .046 C81 -.139 CC2 -.208 HEROIN .257 NOPIAT -.001 ALCHL -.003 CRMFAM .104 OFF1 -.082 CRIME2 .066 PTIME .096 JUVFEL -.136 JUVOUT -.185 STATUS -.147 FRTARR .217 MPROP .123 MPER .058 MDRUG .029 MOTH .083 FPROP -.177 FPER -.043 FDRUG -.054 FOTH -.084 NPRSN -.147 PRISON -.098 PROGPAR -.036 PPEDUC -.035 PPVOC .027 PPRW -.119 PPCONSL .039 PPOTH .001 91 The factor program from the Statistical Package for Social Science (SPSS, 1975) was used to reduce the initial set of data. The specific factoring technique employed was principal factoring without iteration. Varimax rotation was used to insure orthogonality of the factors (components). Kaiser’s Normalization was used as the criterion for determining the number of factors to be considered (eigenvalue equal to or greater than 1.0000). Once principal factors were generated, a decision had to be made regarding the variables to be used in explaining the resulting factor. So as not to over-complicate this process, the decision was made to include all variables as explaining a given factor having a factor loading of .40 and above. With this .40 criterion most variables were contained in a factor. Factor scores, based on the resulting factoring method, were computed utilizing the following formula: f; - fsc'; z + fscézz + fscs. 23 + ..fscnx 2" “Where fsc1 is the factor-score coefficient for variable j J and factor i, and zj is the case’s standardized value on variable j" (SPSS, 1975). The factors were named according to the content in the initial set of variables. Those variables having the highest loading, and above the .40 cutoff, defined the principal factor. The specific results of each ‘Pre’ and ‘Post’ factoring method are presented in following section. 92 Prgzfiriggg Priggigg; Fgctoring In accordance with the Kaiser Normalization criterion, 17 factors were generated. These factors accounted for approximately 662 of the variance (see Table 14). Of the 46 variables considered for the principal components analysis, all but one of the pre-prison variables met the .40 factor loading criterion for inclusion in one of the 17 principal factors. The excluded variable was prior misdemeanor drug arrests. Factor scores were computed for each case as described above. Each factor was named according to the content of the defining variables (see Table 15). W The difference between the pre-prison factoring and post-prison factoring was that the latter principal factoring included, in addition to the pre-prison variables, ‘behavior at the community correction center’ items. Nineteen principal factors were extracted using the Kaiser normalization criterion. These nineteen factors accounted for 66% of the total variance (Table 16). All variables except heroin addiction, Detroit center placement, and educational programming at the community correction center, had factor loadings of .40 and above. As previously mentioned, factor scores were computed for each case. 93 Table 14: Eigenvalues and Percent Variance Accounted For By The 17 Empirical Pre-Prison Release Components COMPONENT EIGENVALUE PERCENT OF VARIANCE CUMMULATIVE PERCENT 1 4.19112 9.1 9.1 2 3.57871 7.8 16.9 3 2.63347 5.7 22.6 4 2.07435 4.5. 27.1 5 1.98976 4.3 31.5 6 1.91196 4.2 35.6 7 1.76244 3.8 39.4 8 1.61793 3.5 43.0 9 1.52278 3.3 46.3 10 1.38193 3.0 49.3 11 1.28561 2.8 52.1 12 1.23474 * 2.7 54.7 13 1.09817 2.4 57.1 14 1.08061 2.3 59.5 15 1.05058 2.3 61.8 16 1.03295 2.2 64.0 17 1.00024 2.2 66.2 94 Table 15: Principal Components Results of the Pre-Prison Release Variables With Accompanying Component Names and Component Loadings COMPONENT COMPONENT NAME/VARIABLES COMPONENT LOADINGS 1 LLEBLQB-QRINIB.&L HI§IQBX_e!Q §§BIQQ§JU§IB§LQEE§E§§ Prior Non-Prison Sentences -.78414 Prior Misdemeanor Property Arrests .64905 Prior Misdemeanor Person Arrests .53330 Instant Offense -.50926 Prior Misdemeanor 'Other’ Arrests .48351 Prior Felony Property Arrests -.48056 2 §§BIQQ§.E..____RISON 8159011121315 Number of Prison Major Misconducts .83666 Number of Prison Misconducts for Fighting .75953 Number of Involuntary Prison Segregations .73643 3 EBLQB_E§;§QN INCQRCERATIONS Prior Prison Incarcerations .89310 Prior Prison Incarcerations in Michigan .85103 Prior Felony Property Arrests .57724 4 §EQBI£BL§QN_.§IQLQN.Q.!Q.EBL& EBQEBQQ PeBTIQIPATIQfl Prison Programming - Routine Work .66037 Length of Time Spent in Prison -.63522 Prison Programming - Educational .61434 Prison Programming - Counseling .52667 Prison Programming - Other .50116 5 QUEEBLL§ABB§§IJL§IQBX Number of Juvenile Felony Arrest .88419 Juvenile Felony Arrests -.86865 Most Serious Outcome of a Juvenile Arrest .52332 95 Table 15: Principal Components Results of the Pre-Prison Release Variables With Accompanying Component Names and Component Loadings (con’t) COMPONENT COMPONENT NAME/VARIABLES COMPONENT LOADINGS 6 QLQEBJSQNS INQLEJQEEN. Age at Instant Offense .78019 Age When First Arrested .68658 Marital Status at Time of Instant Offense .50736 7 8.61mi“)! PARENTJBELBIIXQ Who Raised ? -.88607 Marital Status of Parents .88436 3 mm W Custody of Children at Time of Prison Incarceration .88688 Custody of Children at Time of Prison Release .86591 9 STASLE EMPLQXQENT HISTQEY Number of Jobs .69408 Occupation at Time of Instant Offense -.67315 Longest Length of Time Employed On Any One Job .58937 Highest Grade Completed at Time of Instant Offense .47973 10 MINQS PRISQN MISCONDUCTS Toplock Segregations .78106 Number of Minor Prison Misconducts .72645 Prison Programming - Vocational .45324 11 QQVENIL§_STATUS OFFEQSSS Number of Status Offenses .77541 Most Serious Outcome of Juvenile Arrest .66403 12 EOWFAELLX£BIMUQLCL§IBQX Criminal Involvement of Immediate Family Members .72956 Number of Siblings -.72654 96 Table 15: Principal Components Results of the Pre-Prison Release Variables With Accompanying Component Names and Component Loadings (con’t) COMPONENT COMPONENT NAME/VARIABLES COMPONENT LOADINGS 1 3 EBIQVVR FELQBXVQEEEHSES Number of Prior Felony ‘Other’ Arrests .68382 Number of Prior Felony Person Arrests .63671 1 4 mmPsLosmxsemeNmo £1.8le Non-Opiate Drug Addiction .68178 Number of Prior Felony Drug Arrests -.46791 Race -.43748 Opiate Drug Addiction .41457 1 5 film 1 NSIeN__..T 0F...FEL1§E§ Serving Time for More Than One Instant Offense .73382 1 6 BLWQIQIIQB Alcohol Addiction .76676 17 SHILDLESS ANQ NOT MARRIEQ Children ? .74087 Marital Status at Time of Instant Offense -.43284 97 Table 17 presents the factors, their names, and the variables included with accompanying factor loadings. The principal factors were labeled according to the content of the variables. There were two predictive techniques utilized in analyzing the data. The one to be presented in this section was discriminant function analysis. The goal of this analysis was to statistically distinguish between two groups in terms of a set of defined discriminators. In terms of this study, the dependent variable, groups, was whether those women incarcerated and released into community correction centers would recidivate. In discriminant analyses the dependent variable was dichotomous: did women recidivate according to the four defined measures of recidivism. This technique had two basic objectives and both were employed. The first aspect was that of analysis: most importantly, statistical tests for evaluating the degree to which the discriminating variables (independent variable) actually discriminated between the two groups (dependent variable). Classification was the second objective of discriminant function analysis. After the discrimination, a set of classification functions were derived to be used to predict unknown group membership. The accuracy of the 98 Table 16: Eigenvalues and Percent Variance Accounted For By The 19 Empirical Post-Prison Release Components COMPONENT EIGENVALUE PERCENT OF VARIANCE CUMMULATIVE PERCENT 1 4.22440 8.3 8.3 2 3.59709 7.1 15.3 3 2.68395 5.3 20.6 4 2.28382 4.5 25.1 5 2.04963 4.0 29.1 6 1.97743 3.9 33.0 7 1.81222 3.6 36.5 8 1.66208 3.3 39.8 9 1.54082 3.0 42.8 10 1.48720 2.9 45.7 11 1.38453 2.7 48.4 12 1.32094 2.6 51.0 13 1.22177 2.4 53.4 14 1.12209 2.2 55.6 15 1.10769 2.2 57.8 16 1.07294 2.1 59.9 17 1.04905 2. 1 62.0 18 1.02638 2.0 64.0 19 1.01655 2.0 66.0 99 Table 17: Principal Components Results of the Post-Prison Release Variables With Accompanying Component Names and Component Loadings COMPONENT COMPONENT NAME/VARIABLES COMPONENT LOADINGS 1 LQ_EBLQB-QBIULBQL.BL§IQBX.&UQ §§BIQQ§-I!§I&§T OFFENSE Prior Non-Prison Sentences -.79342 Prior Misdemeanor Property Arrests .72817 Instant Offense -.57856 Prior Misdemeanor Property Arrests -.52148 Prior Misdemeanor ‘Other’ Arrests .43269 Prior Felony Person Arrests .40969 2 QQVENILS QBBSST HISIQBX Juvenile Felony Arrests -.87163 Number of Juvenile Felony Arrest .81886 Most Serious Outcome of a Juvenile Arrest .76506 Number of Status Offenses .55888 3 §§BIQU....S PBI§QMLS§QUDUCVVLS Number of Prison Major Misconducts .83879 Number of Prison Misconducts for Fighting .75943 Number of Involuntary Prison Segregations .73982 4 PR 195 P81 §QU.IU§&BQ§B&IIQU§ Prior Prison Incarcerations .90406 Prior Prison Incarcerations in Michigan .84991 Prior Felony Property Arrests .53160 5 SHORT PRISON STAY QQD QQ_EBlSQy EBQEBBB PERTICIPATIQ! Prison Programming - Routine Work .69196 Prison Programming - Educational .63846 Length of Time Spent in Prison -.59431 Prison Programming - Other .51988 Prison Programming - Counseling .46901 100 Table 17: Principal Components Results of the Post-Prison Release Variables With Accompanying Component Names and Component Loadings (con’t) COMPONENT COMPONENT NAME/VARIABLES COMPONENT LOADINGS 6 QLQEB._!QU§!_!LIH.6LQ%QL_V_PRQ§L.§E§ Age at Instant Offense .75536 Age When First Arrested .63350 Alcohol Problem -.41256 7 WM N Custody of Children at Time of Prison Release .88093 Custody of Children at Time of Prison Incarceration .87073 3 exegesmw Number of Jobs .76355 Occupation at Time of Instant Offense -.65440 Longest Length of Time Employed On Any One Job .60161 Highest Grade Completed at Time of Instant Offense .42290 9 RAISES BY PfiRSflTS/RELQTIXSS Marital Status of Parents .89541 Who Raised ? -.88070 1 0 fl..._1 NOR£BI§Q!-!L§QQ...NDUQI§ Toplock Segregations .74880 Number of Minor Prison Misconducts .67296 Prison Programming - Vocational .51323 1 1 11......0 PRIQB-QBQ§£BQ§L§H-BBQ.§L&QK Non-Opiate Drug Addiction .70363 Community Correction Center Programming - Counseling .62111 Race -.52146 12 QQMQEIILQQBBEQILQE.QEUIEB-£Q§ EBQGRQUEING WITHOUT TBBLEIUE Community Correction Center Programming - Counseling .73052 Community Correction Center Programming - Job -.65530 101 Table 17: Principal Components Results of the Pre-Prison Release Variables With Accompanying Component Names and Component Loadings (con’t) COMPONENT COMPONENT NAME/VARIABLES COMPONENT LOADINGS 13 !Q_EBUILX_§BIfllfleL_fll§IBQX Number of Siblings -.76257 Criminal Involvement of Immediate Family Members .71644 14 QBLLQL§§§_QUQ.!Q!§lfl§L§ Children ? -.70078 Marital Status at Time of Instant Offense .57067 15 EBIQB.E§LQEX.QEE§B§§§ Number of Prior Felony Person Arrests .71524 Number of Prior Felony ‘Other’ Arrests .55575 16 UQLIIEL§.1E§IBBI.QEE§U§§§ Serving Time for More Than One Instant Offense .80348 17 EBLQB_E§LQUX-QBQ§.BBB§§I§ Number of Prior Felony Drug Arrests -.73625 Alcohol Problems .40635 '18 EBLQB_fll§2§n§eUQB.QBQ§_eBB§§I§ Number of Prior Misdemeanor Drug Arrests -.73549 Number of Status Offenses .42274 19 LQ.§DQQBILQUQL ATTAIUEENT QED.EBI§QE EDQQQILQHQL_EBQ§BQEUIBQ Highest Grade Completed at Time of Instant Offense -.56289 Prison Programming - Educational -.40413 102 discriminant function was then expressed in terms of the probability that the case was correctly classified in the group based on this function. The independent variables used for building the discriminant functions were the empirically defined set of factors. As mentioned earlier, factor scores were generated for each case. Therefore, a subject had a factor score for each principal factor. V For the ‘Pre-Prison’ factoring, 17 principal factors were generated: and 19 factors for ‘Post-Prison’ factoring. A stepwise method was the criterion by which the independent variables were selected for inclusion into the discriminant analysis. The independent variables were entered based upon their ability to discriminate between those women that were recidivists and those that did not recidivate. The specific method by which these two groups were separated was Rao’s V. The independent variables were selected based upon their ability to contribute to the largest increase in ‘V’ when added to variables previously in the equation. The outcome equation resulted in the greatest overall separation of the two groups. In determining which independent variables to include in the (discriminant function, a criterion had to be determined. 111a inclusion criterion was the change in Rao’s V statistic that was statistically significant at the .05 level or Del ow. 103 Given the complexity of the data, in terms of multiple outcome variables and the two different sets of independent variables, the discriminant function analyses were presented by independent variables across outcome. The 'Pre-Prison’ variables will be presented first with each outcome variable then the independent variables defined as 'Post-Prison’ will be presented with each set of outcomes. Wiw As mentioned earlier, 17 factors were empirically derived from the 'pre-prison’ variables. These factors, the independent - variables, were placed into the discriminant analysis program. The outcome, Recidivism I, was defined as a woman being sent back to prison for any violation during the first year subsequent to her release from prison. Of the 17 factors only two significantly discriminated (at p g .05), between those that returned to prison and those that were not returned to prison (Table 18). The discriminating factors were: 1e.eciec.s:ieineL_bis;9:2.eng.eeciese.insteet.e££ee5: (prior non-prison sentences, prior misdemeanor property arrests, prior misdemeanor person arrests, instant offense, prior misdemeanor ‘other’ arrests, prior felony property arrests): and fig ggsgogy_gf chilgreg (custody of children at time of prison release, custody of children at time of incarceration). 104 amuVm Hm" csmnxsasamza meznfiVc: >5mVdnosoV maasnasoa m.oamo ac=aVeme mow new seamen memmmm coauosmaam zen: xmnsasVnozoV maaVnnsoz 26 81361 51:8 neochs man menr chnchm azmamsw osmmammm A.momm 5.m-o a.m~mo H.mmmu ~.moum H.mmmu H.mmmH H .ommaau .oummma .emumam .wmmomo .ommmuo .ommmmm .ommumm .oumm .onN .oomm .oomm .oomm .oomH .oomm a.~mm~ m.ummm -.owmm Ha.moem Hm.mumm Hm.aH~o No.umu~ .owmm .cama .ooum .oomm .ooam .ooao .oomN a.~mmw A.omaw w.mumm ~.mm~w ~.wmam H.mw~H H.m>m~ .owmm .oanw .ommm .Homm .HNNm .moua .Nme amuVm mo Ano=.nVu csmnxssszmsa meannVos >5asemsm so: tam umeoa memmmm aoauoamanm =24: amnsqs5kame was 616 seamen szmmmm nosuosmanm zVa: amneas.no:oV maaVnnsos 6.0HNN u.m-a ac5kamsm «ox cam vamo: memmmm nosuozmzam an: nsmzom a: zoom < msmsssVnwanm m.uumo .Hmwm ~.Hmwo .Haau N.Hw_u .Hsaw ~.~Hms .Nmou mVnansnmznm .coaa 114 Table 23: Classification Results - Pre Prison Release and ‘ Recidivism III Predicted Group Membership 1 2 1 82 63 145 (56.6%) (43.3%) Actual Group Membership 2 90 170 260 (34.6%) (65.4%) 172 233 Percent of Grouped Cases Correctly Classified ~ 62.2% 115 Pre-Prison and Regidivigg_1¥ As a dependent variable, Recidivism IV was generated by taking into consideration both the number of violations and the dispositions to these violations. Different from above, group one, or the successful group, were those women who over a period of one year did not have any violations or had one violation with a disposition of ‘dismissed.’ The unsuccessful women, group two, or those that were 'recidivists’ had multiple violations and/or serious dispositions. Multiple violations were defined as greater than one and serious dispositions included being put on probation, sent to jail, returned to prison or absconding-unable to locate. This measure of recidivism addressed the issue of limiting success/failure to simply being returned to prison. Recidivism IV attempted to explore other degrees of the success/failure dichotomy. Of the 17 pre-prison factors only three discriminated between the two groups at a significance level of p g .05 (Table 24). Similar to the above functions the factors were: 19.8:iec_scieioel_bies9:2.eod_se:iese.ineieet.9£iense (prior non-prison sentences, prior misdemeanor property arrests, prior misdemeanor person arrests, instant offense, prior misdemeanor 'other’ arrests, prior felony property arrests); eldec_nensingle.eeesn (age, marital status, age when first arrested): and 116 33m ma" 3.81.3563 3:526: >3;me 2:. 38 V01 .3... ”1.3mm noanosmsam :25 8331...... 2 mm cmcmsamsn <61Vmcsm m «o manmx 5oVkam mos new vssmoo xoVommm nosooomonm zen: zonVos=oVkmam «on oomn ostoo xonmmm noaooomznm zen: monsos .oomm .oooo o.m-m .oHHo .oooo m.u~ou .oHHo .oooo m.nouo .oHna .oooo o.ummm .oumm 122 :.on No A823“ om coomoooon =oVkmsm nos oomn oxsmoo zoVoomo noaooomonm an: wmnsos=oVemsm no: oomn oxsmoo memomm noaooooonm zen: monVoV55V=oVkmsm now oomn oonoz onoomm nosooomonm zen: zonsos=oVooVoooomoVooooV mooonnooo zoooo ooomoo stnooocnnm mooon ooomoo mnok moo oo ooomoo oooaomo ooononooonooo zo nmaooe nooaoooV oomnooe n no monoo oo xoaoooVemom moo oon ooomoo moooomo nooooooonm zono monsooomVemom ooo oom ooomoo xmommmm nooooomonm zono xmnooo215; 3o oom ooomoo szommo nosooooonm Sn: monoooonoooV mooonnooo .mamwo .moo .mmwum .oowmm .ooomm -.oomum ~.~Hmoo .oHo zo smooVk nooaoooV oomnooe .Hmomm .mmm .mmmmu .omwoo .ooowm .oonmo N.omomo .osm zoooo ooomoo somnoooonnm .ommmw .umo .mmwuw .omoas .ooomw -.omumw H.65owm ,.o~o ooooVemom moo oomn ooomoo meoomo nosooomonm zono monsooooVomom ooo oomn ooomoo noooomo nooooooonm son: wonooooooomom ooo oomn ooomoo memmmm nooooooonm zono zonooookamom moo oomn ooomoo moVoomo nosooomonm zono nonoooooVooVomom woo oomn ooomoo zoomomo oooooomonm zono xmnoooooomn zomnooo oVooo zooumooan zoom: zoomoo no ooooonm\mmVono o.mo o.oo no.om R2 with R5 w.mH o.~H :~.mw -o.o~ o.mm N.w~ no.0w um.mm ~.Ho o.wm TH.m~ o.mu o.mu 160 noon we Aooo.nVu nuVnoooV >ooonnooo nooooVomm moo zon zoooomo 2 3 R R h h “u “n W W mm mm o.~a o.oo -o.~m -o.mm -o.o~ -o.oo -o.o~ -o.mo R1 with R4 R1 with R5 o.m6 o.mm o.mm o.ww zonm . we no so oooonmm omnoooooomn zomnooo oVomo zooumoouom :oamo woommo no woomonm\meonoVnoooV >ooonnooo nooVonmm moo zon :oooomo zonm . we no so omoonom oonooo1smmn zimnosk mmx‘ocm vsfimo: z.mno=acnnm anion vsdmoa ~=nmsnm1mn.¢=m m:c1n vxfimo: mnak man zo vsdmo: usimoa vsoosma vuan.n¢umndo= odams roam: :‘n: >.no:o. vxoadmam zo ncmnoqk on nsfidnam: mnmcdm mandokaman :‘mnosk zm‘mmq ck umwmsnm\xmdmndm H.ou u~.Hu H.mb o.mu >.~m o.um no.Hm 6 3 4 R R R h h h t t “luv .m .m w 1 2 2 R R R m.~a» -~.a_ o.am -~.mm -o.~o -o.Ha -o._~ -_.VH -o.Ho -H.~w -w.»m -o.oa -a.~o m.am -o.am m.aH* -m.pa -o.om -~.oo H.0m o.- ~.~o -~.- o.c~ -~.wm _.oH o.uo N.a¢ o.- -o.mH o.mH a.H» -o.wu a.- c.om o.mm H.mm o.~¢ o.m~ -o.wm R2 with R5 N.Ha um.a~ no.om uH.Ho uw.mu .k m.aa um.mm H.0m um.om H.mH o.mH 164 noodm be Ano=.nv” qu11mmnm no.o~ vodos zémomammooa coco >11mmnm o.w~ no mooomn‘oood >nnoioam=n moo oximoo moonondoood vxonamaaéac o.ou R1 with R3 CO hit-I «kw c.- c.am no.Nm 4 5 R R .n h “n “n W H mm mm -o.~o -_.o. o.oo o.- -o.om -~.oo -o.o~ -o.»a -o.om -o.oo zonm - an no no omoonmm xmo‘o*<.ma nkomm zmo0o01ommn :omuook mooooom ooomoo zomnoooonam ooooo ooomoo Honmonmomnooom :6.moo2.. almoo mnmk moo zo Elmo: 1 O ooomoo oooooma omoaoooomaooo odomo zoaoo to": >onoooo ooooomam zo nomnook oo noodoomo mfimoom moodokamom zomnooz zmommo ck omomomm\zmomnooommam ooooo zomooaomooo coco oooommm .0 ro mooomnooomo >nnmooaoon moo ooomoo m moonmioom.‘ ooonomoaooo R2 with R6 4 5 6 5 R R R R h h h h u... .n o... H. w w w w 3 3 3 4 R R R R nH.w~ uH.oo u~.om no.mu no.o~ e.go o.om o.~o u~.em nH.~o a~.mm u~.oH no.um no.mw uo.- no.wm no.wo u~.- o.om o.om zoom . an no am omooaom smooooomskmomv . .uumw qeom Ho . omnooomo no ooomoo + omo. Acomnooaoomon >omokmomv .woom qkom H- - noncoomo no ooomoo + omoo + oomnooomom Aoomnooaoomon >omskmomv .wwwo deco H< . omokmomv .wooH «so .moo .Nom .muo .Nmm .NmH .mmu 1 on .oum .omm .oom .ooo .owm .moo N 1 Ca .oom .oHo .owm .owH .omm .omm m 3 ON .Hoo .Hmm .Hmo .wwm .Hmo .NHN nooomomnooo oonzmmo «om nooamoom moo oomuooomoo zoommmm noaooomonm nooomomnooo amazon: mom noonmoom moo oomnuooomoo zoommmo noaooomoam 1 ow . roocmnmo zumocmomm woo woo nooaooom woo woo oomnooomoo ”mommmm nosooomonm 1 O N I ooocmnmo znmoomoom moo non noowooom moo mom oomn-ooomoo xmoommm noaooomonm noooodoaooo.oonzomo now oomoonamo « moooom A oom- moo oome-ooomoo noaooomommv r. 169 Table 42: Comparison of Predictor Equations Generated From Pre-Prison and Post-Prison Factors RECIDIVISH TYPE CONFIDENCE LIMITS Type I - Type II - Type III - Type IV - Type V - Type VI - :3 significant Recidl Recid2 Recid3 Success YSuccess Viols at .05 -.346 < - g -.106 x: 753/62 -.323 < - g .091 xx 01 2 -.281 §I‘- g —.053 xx 01 02 -.341 - g -.121 it 01 02 -.338 <74- 3 -.112 n 01 02 -.357 §’._ 3 -.137 it 01 02 170 that g, 3%,, was rejected at the ‘.05 level of significance. The results of lein’s formula together with the amount of variance explained by the equations supports the conclusion that in all six recidivism measures, the ‘Post-Prison’ equations were significantly more predictive than the Pre-Prison ones. Ecegicsimigsotigmggogns The significant factors, for both pre- and post-prison variables are categorized by recidivism types for the four research hypotheses (Tables 43 and 44). The four research hypotheses were that (l) the less stable a woman’s personal background (i.e., unemployment, low education, single, young, family criminal involvement, ‘broken home,’ drug involvement) the more likely she will be to recidivate; (2) the more criminal involvement a woman has had, the more likely she will be to recidivate; (3) the more negative a woman’s prison behavior accompanied with a lack of prison program involvement, the more likely she will be to recidivate; and (4) the more negative the offender’s post-institutional behavior (via community correction center) (i.e., no community correction center programming), the more likely one will be to recidivate. Overall, the four stated hypotheses were not fully supported given the lack of predictability of the equations. However, the content of each equation in terms of the hypothesis categories can be presented (Table 45). 171 zomomono :«oonommmm zmnoco<_mz 4soooos oooosmsm -mooooeaoon zomnook -zo nomnook om nooooooo -ooomn :omnook -ooooo nooaoom. zom- nook moo room nomo mmoooom Homnmon onnoomm -ooooo ooomoo Hoomo- nmomnooom -o:ooomnm zooonoomom «oooou ooomoo woomsnooos ooooomam -zo ncmnook on nooooomo =oommn. Iomnooe -ooooo nnoaoom. zom- nook moo room nomo mmooocm Homnmon cnnoomo -o=ooomn zomnook -ooooo noosooms zom- nooo moo room nomo mooooom oomnmon commomo -o=