Journal of Criminal Justice, Vol, 13, pp.217-226 (1985)
(X)47-2352/85 $3:110 + ,(~) Copyrigh! f~, It;85 Pergamon Press Ltd.
Pergamon Press. Printed in U.S.A,
E V A L U A T I O N OF A H A L F W A Y H O U S E FOR W O M E N
D A V I D A . D O W E L L , CECELIA K L E I N , AND C H E R Y L KRICHMAR
Community Psychology Program California State University Long Beach, California 90840
ABSTRACT A halfway house for female offenders in Long Beach, California was evaluated by comparing the number and severity of offenses of graduates of the house to those of a matched group of women released in the same area and period. The groups had comparable criminal histories. A U.S. Bureau of Prisons index was used to classify the severity of offenses. Results indicated that the halfway house reduced both the number and severity of offenses. This conclusion withstood a review of the threats to validity. The policy implications of the finding are discussed.
Halfway houses have gained wide acceptance in the corrections community as a rehabilitation method (Abt Associates, 1979: Ericson, 1978; Neithercutt, 1972: Brodsky, 1974; Iacouetta, 1975; Lyles, 1976: Landolf, 1976: Nelson, 1976: Seiter et al., 1977; Latessa and Allen, 1982). There appear to be good reasons for this acceptance. Halfway houses have been shown to be cost effective compared to institutional placement (Abt Associates, 1979). They do not appear to lead to an increase in crime in the neighborhoods where they are located (Seiter et al., 1977), and they do not cause a decline in local property values (Seiter et al., 1977). Halfway houses also seem to provide for the maintenance of the offenders" community ties (Abt Associates, 1979: Ericson, 1978), and they may be especially well suited to meeting the needs of female offenders (Brodsky, 1974). There are, in addition, important theo-
retical reasons for expecting halfway houses to be effective. It is consistent with sociocultural views of criminal behavior (Merton, 1969; Cohen, 1966; Cioward and Ohlin, 1960; Miller, 1966; Sutherland, 1966; Shaw and McKay, 1942) to expect that recidivism is more likely if the offender returns to a culture that encourages or accepts crime. Similarly, it is likely that economic desperation, prompted by the lack of a place to stay and food to eat, leads to recidivism. Halfway houses provide a social and economic haven that may allow at least some offenders to gain skills necessary to "go straight" and to reduce the impact of factors prompting crime. Wide acceptance in the corrections community and good theoretical reasons do not ensure the efficacy of a rehabilitation approach, a good example being juvenileoffender buddy programs, which have also gained wide acceptance in the corrections
217
218
DAVID A. DOWELL et al.
community. What could make better theoretical sense than inserting a positive role model into the environment of a young, impressionable first offender.'? In fact, however, the evidence indicates that such buddy programs do not significantly impact juvenile recidivism (Dowell, 1978). It is, or should be, incumbent upon innovative social programs to evaluate their actual outcomes for clients and impacts upon society. With respect to halfway houses, there is modest evidence that they are effective in impacting recidivism. Mclvor, H o r n e r , and Boittiaux (1979) reported results from a community-release center in Canada and found evidence that two-thirds of the offenders "succeeded" after release. The Michigan D e p a r t m e n t of Corrections (1980) reported that their extensive halfway house system is working, although the State Corrections Director attributed outcomes to the selection of offenders for the houses. Selecting offenders and rehabilitating them are, of course, quite different matters. In what is apparently the most comprehensive review of (largely unpublished) evaluations of halfway houses, Seiter and his colleagues (Seiter, Carlson, Bowman, Grandfield, and Beran, 1977) found thirtyfive reports that included some examination of recidivism data. Of these, seventeen employed quasi-experimental designs, two used true experimental designs, including randomization, and sixteen used nonexperimental designs (i.e., reports of follow-up figures without any comparison). Of the nineteen experimental and quasi-experimental designs, only three found that recidivism for the halfway house groups was statistically significantly lower; none of these three were true experimental designs. However, of the nineteen studies, only one found that the halfway house group recidivated more. These results are curious because, on the one h.and, three out of nineteen studies provide weak evidence for the effectiveness of halfway houses. Yet, if the facilities had no true effect, we would expect to observe that halfway house subjects recidivated more in approximately nine or ten of the
nineteen studies. A possible interpretation of the existing pattern of findings is that the impact of halfway houses upon recidivism is modest and is often obscured by "'noise" in the data arising from individual differences, local economic conditions, and changes in programs over time. The only study we could find that evaluates the recidivism of women placed in a halfway house was conducted by Williams (1980), who studied the recidivism rates of women released from the Charlotte House, a halfway house for women in Massachusetts. He found that women who went through the house were less likely to be incarcerated within one year of release than matched women who did not receive a halfway house placement. Recidivism rates for the two groups were 7 and 23 percent, respectively. The impact of facilities upon recidivism may be a complex outcome affected by many factors, including, perhaps, the local economy and local sociocultural conditions. In addition, halfway houses are not all alike. The degree of structure provided to clients varies from facility to facility. The rehabilitation opportunities available vary, as do staff characteristics. Another important variable is the selection of offenders for halfway house placement. Although some may benefit more than others from the placement, virtually no systematic research has been done on the matching of offenders to placements. These complexities make the task of evaluating halfway houses formidable. The literature on halfway houses leads to several conclusions. First, halfway houses are cost effective in terms of the public expenditures needed to maintain offenders under supervision as compared to institutional placement. Second, halfway houses seem to achieve some of their program objectives, including maintaining offenders' community ties and making community resources available to offenders. Third, there are compelling theoretical reasons to expect halfway houses to reduce crime. Fourth, the evidence regarding the effectiveness of halfway houses in reducing crime is mixed and does not clearly point to
Evaluation of a Halfway House for Women effectiveness. Finally, little evidence exists that specifically addresses the question of the effectiveness of halfway houses for women. With these conclusions in mind, and despite the formidable difficulties of evaluating corrections p r o g r a m s , we proposed to evaluate the H o f f m a n H o u s e , a halfway house for w o m e n located in Long Beach, California that has been open since 1972. The house provides bed and board to six to twelve w o m e n at a t i m e - - a l l recent discharges from a federal correctional institute. The facility o p e r a t e s on the basis of a r e i m b u r s e m e n t contract with the federal government. In terms of rehabilitation offered within the facility, there was little consistency over the period we studied. Due to a frequent turnover of directors and staff, in some years counseling and " r a p groups" were offered, while in other years little was available. The p r o g r a m seems to have operated without a consistent focus. We a t t e m p t e d to answer the following evaluative question about this facility: Is there evidence that the halfway house helps reduce recidivism or the severity of crimes committed?
METHODS
Subjects T r e a t m e n t subjects were 60 w o m e n who resided in the H o f f m a n H o u s e between 1972 and 1977. These 60 w o m e n comprised about 43 percent of the w o m e n admitted to the house during this period. The 60 w o m e n were chosen from all residents only because they had c o m m i t t e d at least one federal offense and, therefore, had federal records available to us through the U.S. Bureau of Prisons. The m e a n age of these w o m e n at the time they entered the house was 31.5 years. Sixty-three percent were white, 13 percent were Chicana, and 20 percent were black. One individual was half A m e r i c a n Indian and half white. A c o m p u t e r tape containing information
219
about a comparison group of 134 w o m e n was obtained from the U.S. Bureau of Prisons. These comparison w o m e n (1) had been released into the Southern California area between 1972 and 1977; (2) had not resided in a halfway house; and (3) had a federal record of offenses. Because of the way that background data were coded on the tape, we could not fully establish the ethnic composition of the comparison sample.J We were able to determine that 29 percent of the comparison subjects were black, a percentage similar to that for the treatment subjects. One and one-half percent of the comparison subjects were American Indian. We could not determine the percentage Hispanic. The mean age of the comparison subjects was 4.5 years older than that of the treatment subjects, a significant difference (z = 9.82, p < .0001). The comparability of the two groups in terms of criminal histories was examined; the groups were not significantly different on this dimension.
Measures Arrest records were used as the best available indicators of criminal activity, for several reasons. These reasons included: (1) while arrest records underrepresent crime because they reflect only a p p r e h e n d e d criminal activity, conviction and sentencing records underrepresent crime even m o r e and (2) the reporting of arrests is m o r e complete in the Federal Bureau of Investigation data base used in this study than the reporting of sentencing or convictions. It is true that from a civil liberties viewpoint, arrest data should not be p r e s u m e d indicative of actual criminal activity. H o w e v e r , for research purposes, arrest data provides the best estimate of criminal activity. W o m e n in the treatment and comparison groups had as m a n y as forty-five arrests. Three variables were constructed to s u m m a rize their arrest data. The first was a simple count of the n u m b e r of arrests. The second variable was the mean severity of the crimes committed by each woman. The severity variable associated with each arrest was
220
D A V I D A. D O W E L L e t
derived from the "'Crime Scoring Index" d e v e l o p e d by the U.S. Bureau of Prisons for their Community Treatment Center Field Study. T h e third variable was the product of the n u m b e r of arrests times the mean severity values of the crimes for each w o m a n . Each variable is of interest because each reflects a slightly different dimension of criminal activity. T h e n u m b e r of arrests reflects the simple frequency of criminal activity; the m o r e often crimes are committed, the greater the likelihood of arrest, all else being equal. The m e a n severity variable reflects the seriousness of crimes a w o m a n may be likely to commit. The third variable, n u m b e r of arrests times severity, combines the concepts of frequency of criminal activity and severity to provide a measure perhaps m o r e sensitive than either alone, as well as a s u m m a r y of the criminal activity of each w o m a n .
Design The design was a pretest, post-test, nonequivalent c o m p a r i s o n group design. Background-crime records were c o m p a r e d to establish comparability. Follow-up crime records were c o m p a r e d to test the effectiveness of the halfway house. The strategy we initially expected to use to analyse the data involved multivariate techniques for controlling for differences between the t r e a t m e n t and control subjects. We especially expected to find it necessary to control for differences in criminal histories, H o w e v e r , we found that there were no significant differences between the groups on the relevant history variables. In this situation, introducing controls was not desirable, and we p r o c e e d e d to m a k e the simpler follow-up c o m p a r i s o n s without statistical controls. The design was complicated by the issue of follow-up-data censoring. For the treatment group, there was no ambiguity about when the follow-up period began; it started upon release from the halfway house, which occurred at an u n a m b i g u o u s point for each w o m a n between 1972 and 1977. For the comparison group, there was ambiguity,
al.
since most comparison-group women, like most treatment-group women, had long arrest histories. It was necessary to develop a way of comparing the two groups so that the follow-up periods would be the same. One alternative would have been one-toone matching of the treatment and comparison women. This approach, however, would have been somewhat arbitrary and inexact, as well as inefficient, since there were more than twice as many comparisofi w o m e n on whom data were available. It was decided to compute the follow-up crime scores for the entire comparison group using each potential year between 1972-77 as a starting point for the follow-up period. That is, for the entire comparison group, follow-up crime scores were first c o m p u t e d using 1972-1977 as the time frame, then using 1973-1977 as the time frame, and so on until the final, one-year time frame of" 1977. For each follow-up crime variable c o m p u t e d , means and standard deviations were weighted using the corresponding proportion of treatment-group women for each time frame. The aggregated weighted means and standard deviations then became the values against which the treated group's scores were tested.-' This approach controls for the follow-up-time censoring p r o b l e m , since both groups have the same time period available for rearrest. It also uses the comparison-group data efficiently, since all w o m e n are used for all comparisons. A similar procedure was used to compute background-crime scores. RESULTS
Comparabilitv of the Groups Table 1 shows the means and standard deviations of each of the three crime variables for the treatment and comparison groups in the background period. The z-statistic was used to c o m p a r e the means since the groups represent a large proportion of the populations rather than small samples. None of these comparisons approached statistical significance, suggesting that the groups were comparable.
221
Evaluation of a Halfway House for Women
TABLE 1 COMPARABILITY OF THE TREATMENT AND COMPARISON GROUPS
Treatment Variable
N of crimes A v e r a g e severity N × severity
Comparison"
R
SD
X~
SD,,
9.3 4.0 37.4
9.1 1.2 36.3
8.2 3.8 30.7
6.6 1.6 27.7
z
P
0.84 0.96 1.27
NS NS NS
z
P
2.77 2.85 3.92
0.01 0.01 0.001
~' Weighted and aggregated TABLE 2 TREATMENT OUTCOMES
Treatment
Comparison"
Variable
R
SD
X,,.
SD,,.
N of crimes A v e r a g e severity N × severity
0.9 1.4 3.6
1.6 2.0 6.3
1.8 2.3 8.1
2.9 2.1 9.4
~' Weighted and aggregated TABLE 3 ETHNIC GROUP COMPARISONS FOR BACKGROUND VARIABLES
White (N = 38)
Black (N = 12)
Chicana (N = 8)
White vs. White vs. Black Chicana
Variable
X
SD
X
SD
X
SD
z-value
z-value
N of crimes N x severity Average severity
9.8 38.6 4.1
8.5 31.7 1.1
10.1 40.9 3.8
13.6 56.8 1.3
6.9 31.3 4.0
4.2 21.4 0.5
0.07 0.13 0.72
1.43 0.80 0.40
Treatment Outcomes
Table 3 shows the m e a n s and standard deviations of each of the three crime variables for the t r e a t m e n t and comparison groups in the follow-up period. The .--statistic was again used to c o m p a r e means. Each of these statistical comparisons is highly significant. The average n u m b e r of crimes is half in the t r e a t m e n t group what it is
in the comparison group (.9 versus 1.8). The average severity of the crimes for which treatment-group w o m e n were arrested in the follow-up period is less than two-thirds the average severity of the crimes for which control-group w o m e n were arrested (1.4 versus 2.3). The value of the variable based on the product of n u m b e r of crimes times severity is half in the treatment group what it is in the comparison group (3.6 versus 1.8).
222
DAVID A. DOWELL et al.
TABLE 4 ETHNIC GROUP COMPARISONS FOR FOLLOW-UP VARIABLES
White ( N = 38)
Black ( N = 12)
Chicana ( N = 8)
White vs. White vs. Black Chicaml
Variable
X
SD
X
SD
X
SD
z-value
z-value
N of crimes N × severity Average severity
0.71 2.90 1.30
1.3 5.2 2.0
1.4 5.7 1.3
2.6 10.0 2.0
1.1 4.1 2.0
1.4 4.8 2.2
0.89 0.93 0.00
0.72 0.63 0.83
Ethnic Group and Age Comparisons The potential differential effectiveness of the treatment was assessed by comparing, for the treatment group, each of the minority ethnic groups with the white majority group. Table 3 gives the means and standard deviations of the background variables for each ethnic group and for the majority-minority comparisons. Z-values were used because the groups represent large proportions of the populations of each ethnic grouping that underwent treatment, and no assumptions about generalizability were made. None of these differences are significant. Table 4 gives means, standard deviations, and ethnic-group comparisons for the followup variables. None of these differences are significant, either. In sum, there is no evidence for the differential effectiveness of the treatment facility as a function ofethnicity of clients. The potential impact of the age of clients upon effectiveness of treatment was also examined. Age is a well-known correlate of criminal activity (Glaser, 1974), as was borne out by correlations between the age of entering the halfway house and each of the three b a c k g r o u n d variables. There was a positive correlation between the age of entry and two b a c k g r o u n d variables: number of crimes (r = .42, p < .0005) and N of crimes times severity (r = .39, p < .001). The correlation between the average severity of crimes and age at entry was not
significant ( - .06, p -- .3). These results are not surprising, since women come to the halfway house after an incarceration. The rate of incarcerations is correlated with age: hence, an indirect correlation exists between age of entry into the halfway house and indicators of criminal activity. Of more interest are potential correlations between follow-up variables and age of entry into the house. No significant correlations were identified. N u m b e r of crimes correlated at r = .05 09 > .3): number times severity correlated at r = .03 (p > .4). Average severity correlated at r = -.007 (p > .4). These results suggest that there is no differential effectiveness of treatment that is a function of age. This conclusion is strengthened by the fact that the mean range and standard deviation of ages ('~ = 31.5 years, range = 18-56 years, SD = 9.5 years) seem to be an adequate test of any potential differential effectiveness. The exception to this conclusion is, of course, juveniles.
CONCLUSIONS AND DISCUSSION These findings seem to indicate that this halfway house succeeded in reducing the frequency and severity of crime among the women who resided there as an aftercareplacement. Because a quasi-experimental design was used to examine the data, there are six important rival hypotheses to this conclusion:
Evaluation of a Halfway House for Women (1) age differences between groups may account for results; (2) ethnic-group differences may account for results; (3) some other group difference(s) may account for results; (4) differential incarceration rates may account for results: (5) differences in the post-test observation periods may account for results; (6) a treatment times subject interaction may limit the external validity of results. Each of these rival hypotheses is discussed in turn. As noted above, the treatment group was younger, with a mean age of about 28 years in 1972, the first possible follow-up observation year. The older, comparison group had a mean age of about 32.5 in 1972. Since research has consistently shown that age is a correlate of criminal activity (Glaser, 1974) and that crime peaks around ages 18-24, the age difference between treatment and comparison groups should have operated to produce fewer crimes in the comparison g r o u p - - a conservative bias. The observation that treatment subjects had a lower crime rate despite this bias is evidence supportive of the hypothesis that the treatment was effective. Ethnic-group composition data for the two groups were poor and do not allow unambiguous determination of the comparison-group composition. However, if ethnicgroup differences were a factor in producing differential crime rates in the research populations, we would expect to observe such differences in the background period as well as in the follow-up period. Since no such differences were observed in the background period, the emergence of ethnic differences that could create the pattern of results we saw in the follow-up period seems unlikely. This brings us to the third rival hypothesis: some other group difference(s) might account for the results. Particularly troublesome is the possibility of selection into the house on the basis of a motivation to "'go straight." This possibility cannot be abso-
223
lutely ruled out, but it is made less credible by several things. First, the mean age of women admitted to the house was younger than that of the comparison group, suggesting a higher crime risk and less motivation to "'go straight" for the treatment group. Second, motivational differences did not appear as background-crime differences, since the groups were similar in the background period. Without other evidence suggesting that selection was an operative factor, it is simplest to infer that the treatment produced results. Next we consider a particularly troublesome rival hypothesis: the possibility of differential incarceration. The control group had slightly, though statistically insignificantly, more, and more serious, crimes in the background period. If arrests for these crimes resulted in more and/or longer incarcerations, that would reduce the control group's opportunity to commit crime in the follow-up period. The differences between groups in the background period are slight. In order for these slight differences to produce the substantial differences observed in the follow-up rates, the difference in mean levels of background crime would have to be somehow amplified in the followup period, perhalSs by a ~critical value" of criminal activity that prompts the criminal justice system to respond with differential sentencing practices. That is, above the "critical value," sentencing would be quite severe, while below it, sentencing would be moderate. In more analytic terms, this rival hypothesis is based upon the assumption of a nonlinear relationship between crime severity and sentencing, plus the assumption that the criminal histories of the two groups are in a region on the distribution of the crime variable where this nonlinearity becomes pronounced. This notion of critical value did not seem implausible. This rival hypothesis requires that a negative correlation exist between level of criminal activity in the background period and level in the follow-up period, so that rearrest for individuals with extensive criminal histories is prevented (by incarceration). We tested this hypothesis by autocorrelating the
224
DAVID A. DOWELL et al.
three crime variables across the backgroundto-follow-up periods within the treatment group. N o n e of the three correlations were negative. In fact, two were strong positive correlations (.46 and .49, both significant at p = .001). T h e third was near zero ( - .05, n.s.). We also e x a m i n e d autocorrelations in the c o m p a r i s o n group. H e r e there were eighteen correlations to examine (six followup starting points, 1972-1977, and three crime variables). T h e r e were no significant negative correlations. If differential incarceration were an explanation, then it would have to a p p e a r as a negative correlation between background and follow-up arrest records. The complete lack of such a correlation m a k e s this rival hypothesis untenable. A n o t h e r rival hypothesis concerns the comparability of the post-test observation period. As described earlier, the study a d o p t e d a p r o c e d u r e that examined six follow-up observation periods derived from the c o m p a r i s o n - g r o u p data set and that c o m b i n e d these observation periods using a weighting strategy. T h e data bases for the two groups were identical (FBI records): thus these p r o c e d u r e s insured that the a m o u n t of time in which rearrests might occur was the same in the two groups. This rival hypothesis is, therefore, untenable. Finally, a t r e a t m e n t by subject interaction might have occurred such that the results o b s e r v e d for the t r e a t m e n t group were not indicative of what might have h a p p e n e d with c o m p a r i s o n - g r o u p subjects, had they gone through the t r e a t m e n t , because treatment results d e p e n d upon subject characteristics (say ethnic composition). This is not a threat to internal validity but could [~e a threat to external validity. The age and ethnic-group analyses reduce the plausibility of the notion that results d e p e n d e d on either of these characteristics. The results are appropriately viewed as generalizable only to similar subject populations and programs. T h e r e are no other important, tenable rival hypotheses. The results and arguments presented a b o v e lead us to conclude that within the limits of methodological accuracy, the present study indicates that the
halfway house was an effective intervention that reduced crime by a statistically significant amount for women who participated in it. The results are appropriately generalizable only to similar programs with similar subject populations. It is appropriate to have more confidence in the statistical significance of these findings than in the p a r a m e t e r estimates that findings provide. However, it is interesting to examine the magnitude of the crime estimates developed in this study. The n u m b e r of crimes committed bv treatmentgroup women is half that c o m m i t t e d by comparison women. The average severity of crimes is less by about one-third in the treatment group. These results are not only indicative of a statistically significant reduction in crime: they also indicate a socially important impact for this program. Finally, the ethnic and age analyses suggest that there is no interaction between these variables and the treatment in a way that influences results. This conclusion is tentative due to the small sample sizes. We are aware that the d e p e n d e n t variables used in this evaluation are limited criteria for determining the success or failure of a rehabilitation program and, likely, not the best criteria. Completely ignored by these criteria are issues concerning the quality of adjustment to c o m m u n i t y living made by the female clients of the program. Results suggest, but do not prove, that the quality of adjustment is higher in the treatment group than in the comparison group, if we assume that crime is one response to poor adjustment. The study certainly suggests that more evaluations of halfway houses for women focusing on quality of adjustment are warranted. Why should a halfway house for w o m e n succeed? The particulars of the p r o g r a m studied seem unlikely to be the explanation, since the program evolved over the years under several directors and has not had a consistent focus. More likely is the possibility that female offenders have few economic or social opportunities on discharge: the halfway house provides a tangible "'edge" in the form of social support, shelter, encourage-
Evaluation of a Halfway House for Women
ment, and opportunity. A different, but not i n c o m p a t i b l e , e x p l a n a t i o n is t h e p o s s i b i l i t y that female offenders merely feel economically and socially powerless; the house may p r o v i d e a p s y c h o l o g i c a l ' ~ e d g e " t h a t is p e r h a p s less c r u c i a l f o r m e n . T h e s e e x p l a n a t i o n s are merely speculative~ but they deserve a m p l i f i c a t i o n in f u r t h e r r e s e a r c h . In sum, this study provides a favorable evaluation for a halfway house for women and offers no evidence that results depend upon (at least two) subject characteristics. Explanations for the results are speculative, b u t t h e p o l i c y i m p l i c a t i o n s a r e c l e a r : (1) more controlled evaluations of halfway h o u s e s s h o u l d b e c o n d u c t e d a n d (2) u n l e s s contradictory results emerge from other studies, similar programs should be funded widely.
ACKNOWLEDGMENTS This work was completed under contract with the U.S. Bureau of Prisons. Contents have not received official approval or endorsement from the Bureau. Reprint requests can be sent to the first author at the Department of Psychology, California State University, Long Beach. CA 9(184(I.
NOTES
225
Enforcement Assistance Administration Report. Washington, D.C.: Superintendent of Documents. Brodsky, A.M. (1974). Planning for the female offender. Crim Just and B 1: 392-400. CIoward, L.E., and Ohlin, R.D. (1960). Delinquen o, and opportunity. New York: Free Press. Cohen, A. (1966). The delinquency subculture. In Juvenile delinqueno,, ed. R. Giallombondo. New York: John Wiley. Dowell, D. (19781. Volunteers in probation: A research note on evaluation. J Grim Just 6: 357-61. Ericson, D. (19781. New hope ,(or women exo[(enders--Project Esperanza. Department of Labor Manpower Administration Report. Washington, D.C.: Superintendent of Documents. Glaser, D. (1974). Strategic" criminal justice planning. Washington, D.C.: U.S. Government Printing Office. lacouetta. R.G. (19751. Corrections and the female offender. National Criminal Justice Reference Service. Microfiche. LandolL J. (1976). Charlotte House pre-release center .(or women--A profile o f participants and a recidivism /ollow-up. Boston, MA: Massachusetts Department of Corrections. Manuscript. Latessa, E., and Allen, E.E. (1982). Halfway houses and parole: A national assessment. J Crim J 10: 153-63. Lyles, B.D. (1976). Criminal justice system--Innovations, failures, trends, and alternatives. Washington, D.C.: Howard University Institute for Urban Affairs. Manuscript.
Two race variables were used on the tape. The first encoded Hispanic persons as white. The second was designed to distinguish Hispanic persons, but data were missing from our tape.
Mertom R. (1969). Social structure and anomie. In Delinquency, crime, and social process, eds. D.R. Cressey and D.A. Woods. New York: Harper and Row.
: Given here are the numbers of treatment-group women associated with each follow-up time frame and the weights applied to aggregate control-grotip means and standard deviations. The weighting factors are based upon the distribution of treatmentgrottp women across time frames.
Michigan Department of Corrections. (1980). Michigan study concludes halfway houses might help. Crime Control Digest 14: 5.
Time Frame
N
Weighting Factor
1972-77 1973-77 1974-77 1975-77 1976-77 1977
1 6 13 15 19 6
0.0167 0.1 0.2167 0.25 0.3167 0.1
Total 60
1.001
REFERENCES Abt Associates, Inc, (19791. Women's community center exemplary project validation report. Law
McIvor, D.: Homer, B.: and Boittiaux, R. (1979). Preliminary results from a community release centre. Can J Crim 21: 340-45. Miller, W.B. (1966). Lower class culture as a generating milieu of gang delinquency. In Juvenile delinqueno', ed. R. Giallombondo. New York: John Wiley. Neithercutt, M.G. (19721. Parole violation patterns and commitment offense. J Res Crime and Delinq 9: 87-98. Nelson G. (1976). Female offender--After release where? Paper presented at the American Correctional Association. College Park, Maryland. Seiter. R.P.: Carlson, E.W.: Bowman, H.H.; Grandfield, J.: and Beran, N.J. (1977). National evaluation program: Phase I summa O, report: Halfway houses. Washington, D.C.: U.S. Government Printing Office.
226
D A V I D A. D O W E L L et al.
Sutherland, E. (1966). T h e o r y of differential association. In Juvenile delinquency, ed. R. Giallombondo. New York: John Wiley. Shaw, C., and M c K a y , H. (1942). Juvenile delinquency and urban areas. Chicago: University of Chicago Press.
Williams, L.T. (1980). Analysis of recidivism of w o m e n released from the Charlotte House, 1977 and 1978 releases. Boston. MA: Massachusetts Department of Corrections. Manuscript.