Richard R. Johnson
Department of Criminal Justice, The University of Toledo
Factors Associated with Recidivism on Probation
Method
Findings
Discussion and Conclusion
DOMESTIC VIOLENCE is one of the most prevalent forms of violent crime to come to the attention of the criminal justice system. In 2004, more than 1,144,900 incidents of domestic assault were reported to the police in the U.S., resulting in over 948,000 arrests (FBI, 2005). Among the arrestees who were convicted, most typically receive a sentence to complete a batterers treatment program and to serve a period of community supervision on probation (Belknap, Graham, Hartman, Lippen, Allen, & Southerland, 2000; Buzawa, Hoteling, & Klein, 1998; Ferraro & Boychuk, 1992; Hofford, 1991). As a result, the research on identifying recidivism risk factors for domestic batterers should concentrate most on probationers and probation programs.
Prior research has suggested that domestic batterers frequently differ from the generalized population of violent offenders in their motivations, target selection, and thinking processes (Holtzworth-Munroe & Stuart, 1994; Saunders, 1995). These differences in criminal cognitive processes and behaviors suggest unique differences may exist for domestic batterers with regard to other factors as well, such as those influencing their chances of recidivism. While research findings currently exist on the recidivism risk factors of probationers in general (Morgan, 1993; 1994; Sims & Jones, 1997), little is known about what specific recidivism risk factors exist for domestic violence offenders. Furthermore, even less is known about the recidivism risk factors for domestic batterers convicted of felony offenses.
The present study sought to identify what risk factors were correlated with recidivism for a sample of felony domestic batterers serving a sentence of 24 months probation. Specifically, this study attempted to identify which offender characteristics were correlated with re-arrest for a new violent offense for a sample of 273 male offenders from one suburban county in Illinois, who were serving sentences of probation for a felony offense committed as part of an act of domestic violence. Successfully identifying such characteristics would help policy makers determine offender risk and decide upon the most appropriate sentence for felony domestic batterers. This information could also be useful for probation departments in developing rehabilitative case plans for felony domestic batterers.
Factors Associated with Recidivism on Probation
Several studies have identified factors associated with general offender recidivism while on probation. Morgan (1993) reviewed 24 studies published prior to 1990 and found that 9 offender characteristics were consistently associated with committing new criminal offenses while on probation. First, male offenders generally were more likely to be re-arrested than female offenders. Second, younger offenders were more likely to re-offend than older offenders. Third, unmarried offenders were re-arrested more frequently than married offenders. Fourth, education level was negatively correlated with re-offending, as each year of formal education reduced the likelihood of a new offense.
Fifth, members of racial minority groups were more likely to be re-arrested while on probation than Whites. Sixth, employment instability was associated with failure, as the more job changes the probationer experienced the more likely he/she was to re-offend. Seventh, the more prior criminal offense convictions the offender had collected, the higher the likelihood the offender would be re-arrested. Eighth, violent offenders were more likely to be re-arrested than were property offenders. Finally, the offender's sentence length was correlated with failure as the longer the period of supervision, the more likely the offender was to experience a new arrest (Morgan, 1993).
The more recent research literature has continued to support the significance of most of these nine predictors of recidivism among general probationers. Morgan (1994) studied a sample of 266 felons on probation in Tennessee and found that five of the nine predictors were significantly correlated with whether or not the offender's probation was revoked for a new offense. Sex, marital status, employment instability, type of offense, and prior criminal history were all significantly associated with recidivism. Age, education, race, and length of sentence, however, were not found to be significant influences in this study.
Sims and Jones (1997) examined factors associated with sentence revocation among 2,850 felony offenders on probation in North Carolina. Their findings were consistent with the previous research in that age, marital status, education level, race, employment instability, prior criminal record, offense type, and length of sentence were all correlated with re-arrest while on probation. This study also revealed three other factors that were significant: address instability (more frequent address changes increasing the odds of arrest), age at first offense (younger first offenses increasing the odds of arrest), and history of substance abuse (more frequent use of drugs or alcohol increasing the odds of arrest).
Only a few studies, however, have evaluated the offender characteristics associated with re-offending specifically among probationers convicted of domestic violence offenses. Olson and Stalans (2001) evaluated the likelihood of re-offending among 411 violent offenders on probation in Illinois, and compared domestic batterers with other violent offenders. They found that although both types of violent offenders were re-arrested for any type of criminal offense at similar rates, the domestic violence offenders were more likely to commit a new violent act, and commit it against the victim in their original offense. Although these researchers only included a limited number of potential recidivism correlates in their model (age, race, education, criminal history, substance abuse, and sentence length), they did find that four of these factors B age, education, substance abuse, and sentence length B were significantly correlated with arrest for a new violent offense among domestic batterers, and these correlations were in the same directions as those found with other types of probationers. Race and criminal history were not significant predictors of re-arrest for the domestic batterers in their sample (Olson & Stalans, 2001).
Hanson and Wallace-Capretta (2004) examined recidivism risk factors in a Canadian sample of 320 male domestic batterers who were sentenced to attend a domestic batterer counseling program, most of whom (but not all) were also serving a sentence of probation supervision. They found that re-arrest for a new violent offense was significantly correlated with a number of offender characteristics. Offenders re-arrested for violence within five years of sentencing were more likely to have experienced residential instability, employment instability, and financial instability, and were more likely to have criminal peers. Offenders with longer histories of substance abuse, prior arrests, and prior violent offense charges were also more likely to be arrested for a new violent act than those with shorter histories.
While previous studies have identified a number of factors related to the re-arrest among probationers generally, and only a few have studied domestic batterers on probation, none to date have specifically considered the correlates of re-arrest for domestic violence felony offenders on probation. Detecting such correlates would be useful in determining risk factors that could identify specific offenders who should not be considered for a sentence of probation due to their higher risk of re-offending. Identifying such risk factors could also assist probation officers in determining which offenders require closer supervision and case management due to their risks to public safety. The present study involved an examination of a sample of male offenders convicted of a felony level domestic battery offense, serving sentences of 24 months of probation and counseling in a suburban county in Illinois. This study sought to explore what offender characteristics were known to the court at the time of sentencing and were significantly predictive of being re-arrested for a new violent offense.
Method
The present study examined a sample of male offenders sentenced to probation for a domestic violence-related felony offense in a suburban county in the Chicago metropolitan area. This data was collected as part of a program evaluation of a domestic violence court program (Johnson, 2001). Automated court records were reviewed for all felony offenders sentenced by the county court between August 1, 1992 and July 31, 1999 to identify all offenders who had been charged with a count of domestic battery. This produced a sampling frame of 328 felony domestic violence offenders. Approximately 91 percent of these offenders were male and all but 26 received probation supervision as part of their sentence. In reviewing the sentences for these offenders it became apparent that in this court system the standard sentence, or Agoing rate,@ for a felony domestic violence-related charge had been 24 months of probation supervision and a requirement to successfully complete a court-approved 26-week batterers counseling program.
In order to control for differences in sentence conditions (and possibly the severity of the instanter offenses), only offenders who received this standard 24-month probation and counseling sentence were included in the sample. Also, because the number of female offenders in the sampling frame was so small, only male offenders were included in this analysis. This produced a final sample of 273 male offenders who had been sentenced for a violent felony charge associated with a domestic battery, received a sentence of 24 months of probation, and had been ordered to attend a 26-week domestic violence counseling program.
The Dependent Variable
The dependent variable in this analysis was whether or not the offender was re-arrested for a new violent offense while still on probation. This was determined by searching the judicial circuit's automated court records system to determine if a new violent charge had been recorded for the offender during the 24 months that immediately followed the offender's sentencing. Limitations exist, however, when measuring the dependent variable this way. Looking only within the circuit may have caused us to miss new arrests outside the three counties of the circuit. Nevertheless, due to the restrictions on the researcher's access to federal and statewide criminal history databases, this was the most prudent option available for measuring re-arrest.
Evidence also exists that the police frequently do not detect or are reluctant to initiate arrests in incidents of domestic violence (Buzawa & Buzawa, 2003), creating the possibility that some of the offenders in the sample who were not re-arrested for new acts of violence were simply not caught. Some of the offenders may have committed new acts of violence but the victim failed to report the incident. Some of the offenders may have re-offended and had the police intervene, yet the officers declined to make an arrest. Because of issues of victim privacy the author was unable to obtain permission to contact and interview the victims to determine if they had experienced new incidents of abuse by the offender. Therefore, reliance on an official record of re-arrest within the judicial circuit was the only option available.
Of the 273 offenders in the sample, 112 (41 percent) were found to have been re-arrested within the judicial circuit for a new violent offense before the end of their 24-month sentences. This percentage of recidivism is slightly higher than revealed by the previous literature on domestic batterers on probation. In Olson and Stalans' (2001) sample of misdemeanant and felon domestic batterers, 32.3 percent were re-arrested for a violent offense before completing their probation supervision. Hanson and Wallace-Capretta's (2004) sample of (convicted and non-convicted) batterers attending counseling revealed that 25.6 percent of the offenders committed a new violent act within five years of sentencing. Neither of these two previous studies, however, had focused only on felony offenders, which may explain the higher recidivism found here.
The Independent Variables
The independent variables selected for analysis were all of the descriptive variables identified in the previous literature that were known to the court at the time of the offender's sentencing. These variables were the offender's race, age, education level, employment stability, address stability, shared residence status with the victim, history of substance abuse, prior completion of batterer's counseling, and the offender's prior criminal record.
As Table 1 reveals, the sample involved 97 White (non-Hispanic) offenders, 97 African-American offenders, and 79 Hispanic offenders. Because some of the prior literature has suggested that probationers who are members of ethnic minority groups are more likely to be re-arrested than Whites (Morgan, 1993; Olson & Stalans, 2001; Sims & Jones, 1997), this variable was collapsed into a dichotomous dummy variable (White = 1, non-White = 0). It was predicted that being White would be negatively related to re-arrest, with Whites being less likely than non-Whites to be arrested again for violence while on probation.
Data was available for the offender's age in years at the time of sentencing; therefore this ratio level of measurement was used for the age variable. The ages of the offenders in the sample ranged from 18 to 50, with a mean of 32 years old. Because previous findings suggested that age was negatively correlated with re-offending (Morgan, 1993; Olson & Stalans, 2001; Sims & Jones, 1997), it was hypothesized that the younger the offender the more likely he would be to experience re-arrest while on probation.
Approximately 68 percent of the offenders in the sample had at least a high school diploma or GED certificate and 11 percent had at least a two-year college degree. This sample was generally more educated than most typical samples of felony offenders (Lochner & Moretti, 2001), which was likely a result of the suburban county environment, with a higher-than-average level of income and education. Because only about a tenth of the offenders had a college degree and most of the prior studies on the influence of education on probation outcome focused on whether or not the offender was a high school graduate (Morgan, 1993; Olson & Stalans, 2001; Sims & Jones, 1997), the measure used in the present study was a dichotomous variable assessing whether the offender had completed high school or an equivalency certificate. Consistent with the previous literature, it was hypothesized that offenders who did not possess a high school diploma or GED were more likely to be re-arrested for a new offense than more educated offenders.
Employment stability had been identified as a significant risk factor by the previous literature, suggesting that the more job changes an offender experienced while on probation the more likely the offender was to be re-arrested. In the present study data was available on the number of jobs the offender had held in the 36 months immediately prior to sentencing. As Table 1 reveals, the offenders' jobs ranged from one to six, with a mean of approximately two jobs held during the three years prior to sentencing. Address stability had also been suggested as a risk factor in failing probation and re-offending (Hanson & Wallace-Capretta, 2004; Sims & Jones, 1997), and the number of official addresses for the offenders in the sample was available for the 36 months prior to sentencing. The number of different addresses held by the offenders ranged from one to five with a mean of approximately two different residence locations during the three years prior to sentencing. As was the case with employment stability, we hypothesized that offenders who had more address changes were more likely to re-offend than offenders with fewer address changes.
Dealing with a sample of primarily misdemeanor domestic batterers, Hanson and Wallace-Capretta (2004) found that whether or not the offender continued to reside with the victim of the domestic violence offense significantly predicted re-arrest and failure to complete counseling. In their study, domestic batterers who continued to reside with their victim (potentially recreating the situational circumstances that led to the initial offense) were more likely to re-offend and fail counseling. Of the sample for the present study, 55 percent of the offenders reported residing with their victim at the time of their sentencing. We hypothesized that the offenders who were still residing with their victims would be more likely to re-offend than the offenders who were not.
Hanson and Wallace-Capretta (2004) and Olson and Stalans (2001) found that substance abuse issues were related to re-offending among probationers. Offenders who abused alcohol or illegal drugs were more likely to re-offend. The court records on the offenders in this sample did not specifically detail whether or not any had disclosed having a drug or alcohol addiction. Even if asked, the accuracy of the offenders' statements may be unreliable due to the denial so deeply associated with alcohol and drug addictions, or a desire to appear socially responsible to the court. A proxy measure was established, however, whereby all offenders who had a prior conviction for a drug- or alcohol-related offense were counted as having a substance abuse issue. Therefore it was hypothesized that the offenders who had a previous conviction for a drug- or alcohol-related offense were more likely to be re-arrested than the offenders who lacked such a conviction on their record.
The existing research on the effectiveness of domestic batterer treatment or counseling programs at reducing recidivism has not been encouraging (Buzawa & Buzawa, 2003; Gondolf, 2004). These programs have generally been found to have little or no effect on recidivism, yet most courts continue to order domestic batterers to such counseling programs with hopes that they will help the offender rehabilitate. Information on whether or not the offender had previously successfully completed a court-recognized 26-week domestic batterer counseling programs was in the court record and included in the analysis. As Table 1 reveals, approximately half of the sample had already completed a 26-week domestic batterer counseling program at least once prior to sentencing for the present offense. It was hypothesized that if such programs were effective, then those who had already completed this counseling requirement would be less likely to re-offend than those who had not completed this type of a program.
Finally, the length of the offender's prior criminal record has been reported as a strong correlate with re-offending (Morgan, 1993; Olson & Stalans, 2001; Sims & Jones, 1997); therefore, the number of prior criminal convictions the offender had on his record was included as an overall measure of past criminal activity. The number of prior criminal convictions ranged from 1 to 29, with a mean of 7.68 prior criminal offense convictions. A second measure was also used, counting the number of prior convictions the offender had for violent offenses. The number of prior violent offense convictions ranged from 0 to 20, with a mean of 4.45 prior violence convictions. It was hypothesized that offenders with more prior convictions would be more likely to re-offend than offenders with fewer prior convictions. It was also hypothesized that offenders with more violent offense convictions would be more likely to be re-arrested than offenders with fewer prior violence convictions.
The dependent variable, whether or not the offender was re-arrested on a violent charge at any time during his 24-month sentence of probation, was then regressed by the 10 independent variables described above. Because the dependent variable was dichotomous, a binary logistic regression test was utilized to determine the influence of each predictor on the odds that the offender would be re-arrested for a new act of violence. The results of this test are presented in Table 2.
Findings
As can be seen in Table 2, the model chi-square value of the logistic equation was highly significant (p < .000), indicating that the independent variables in the model increased the goodness of fit of the chance of the model. Two model pseudo R-squares were also calculated. The standard Nagelkerke pseudo R-square measure suggested that the model explained approximately 36 percent of the chance of re-arrest, and the more conservative Cox and Snell pseudo R-square suggested the model explained 26 percent. This finding was consistent with the models utilized in the previous studies of the recidivism of probationers (Morgan, 1993; Olson & Stalans, 2001; Sims & Jones, 1997).
Seven of the independent variables in the model were significant to the logistic equation. These were age, employment instability, residential instability, residing with the victim, substance abuse, prior criminal convictions, and prior violence convictions. The directions of all of these significant relationships were in the predicted directions. Younger offenders were more likely to be re-arrested than older offenders. The more job changes or address changes the offender had experienced, the more likely the offender was to be re-arrested. Offenders who were still residing with their victims were more likely to be rearrested than offenders who were not. Offenders with a prior alcohol or drug offense conviction were more likely to be re-arrested than offenders who had not; and the longer the offender's criminal record (both total offenses and violent offenses) the more likely he was to be re-arrested. An eighth variable, education, approached significance with a probability of .071, and displayed a negative relationship as predicted. Finally, two of the independent variables, race and prior batterer counseling, did not approach significance.
An examination of the odds ratios for those variables that were significant predictors of re-arrest revealed that prior arrest for a drug or alcohol offense was the strongest predictor of offender outcome. Having a prior conviction for a drug and/or alcohol offense almost doubled an offender's odds of re-arrest for a violent offense while on probation. While 27 percent of the offenders who had not been convicted of a drug or alcohol offense were re-arrested for violence while on probation, approximately 54 percent of the offenders with a past substance abuse related offense conviction were re-arrested.
The second most influential predictor of re-arrest was residence instability, followed by employment instability. For every address change the offender experienced during the three years prior to sentencing, the offender's odds of re-arrest increased approximately 1.5 times. The mean number of addresses held by those who were re-arrested was 1.9, while those who were not re-arrested had a mean of 1.3 addresses. Similarly, for every job change the offender had experienced, his odds of re-arrest increased approximately 1.4 times. While the mean number of jobs held by those who were not re-arrested was 1.87, those who were re-arrested held an average of 2.86 different jobs during the three years prior to being sentenced.
The number of prior criminal convictions and, to a lesser extent, prior violent offense convictions also increased the offender's odds of re-arrest for violence while on probation. Each prior criminal offense conviction on the offender's record increased the offender's odds of re-arrest by approximately 1.2 times, and each prior violent offense conviction raised the odds of re-arrest by .76. The offenders in the sample who were re-arrested for violence averaged 10 prior criminal convictions and 6 prior violent offense convictions. By contrast the offenders who were not re-arrested only averaged 6 prior criminal convictions and 4 prior convictions for violence.
Age was significantly and negatively related to re-arrest for violence. Younger offenders were more likely to be re-arrested than were older offenders. While the mean age of the offenders who were re-arrested was 30.5 years old, the mean age of the offenders who were not re-arrested was 33.1 years old. Finally, residing with one's victim increased the odds of re-arrest for the offenders in the sample. More than 60 percent of the offenders in the sample who were re-arrested for violence had been residing with their original victim when they began probation, while only about 39 percent of those not living with their original victim were re-arrested while on probation.
Discussion and Conclusion
Caution should be taken when inferring policy recommendations directly from these findings without waiting for replication by other researchers. This study was limited to only male offenders in one suburban county in a Midwestern state. Differences may exist with female offenders, offenders in heavily urban or rural communities, or offenders from different parts of the nation or world. Furthermore, the outcome measure relied upon in this study was re-arrest for a new violent offense, not conviction or victim reporting. It is possible that some of these re-arrested offenders were never convicted of their new alleged offense. It is also possible that some of the offenders who were not re-arrested actually did re-offend, yet the crime was never reported to the police or an arrest made. It is not known how these possibilities could have affected the present findings.
Nevertheless, the findings here did identify several characteristics of felony domestic batterers that were predictive of re-arrest for a violent offense while on probation for this sample. These findings were also consistent with the previous research studies regarding the factors associated with recidivism for offenders of various types on probation. The fact that similar results were found by others is encouraging and increases confidence in the findings here.
The prior research on the influence of substance abuse on the offender's potential for getting into further trouble supports the results of the present study. Studying felons on probation for a variety of offenses, Sims and Jones (1997) found that substance abuse issues were a significant predictor that the probationer's sentence would be revoked. Olson and Stalans (2001) found that substance abuse issues were a significant predictor of both re-arrest for domestic batterers and non-domestic violent offenders on probation. Hanson and Wallace-Capretta (2004) evaluated the recidivism of domestic batterers undergoing court-ordered domestic batterer counseling, finding that substance abuse was also a significant predictor of re-arrest for these primarily misdemeanor offenders. These consistent findings suggest that having a prior history of substance abuse issues does increase a domestic violence probationer's odds of re-arrest. Illegal drug or alcohol abuse can cloud the mind, making attempts at cognitive change more difficult. It can lower inhibitions that would normally prevent an offender from re-offending. Finally, it can add stress to the relationship between the batterer and his domestic partner.
Sims and Jones (1997) found that residential instability increased a probationer's odds of recidivism. Hanson and Wallace-Capretta (2004) found that residential instability was correlated to re-arrest specifically with domestic batterers. The present study continued to support residential instability as a predictor of re-arrest among domestic batterer felons. Offenders who frequently moved might have experienced increased stress in their lives and a decreased sense of social stability. This increased pressure on the offender and lack of social stability could have resulted in an inability to control one's impulses toward violent behavior. The same could have been true for employment instability, which not only was correlated to re-arrest in the present study, but was also a factor correlated to probation recidivism in most of the prior literature (Hanson & Wallace-Capretta, 2004; Morgan, 1993, 1994; Sims & Jones, 1997).
Age was consistently another offender characteristic that has aided in the prediction of probationer recidivism (Morgan, 1993; Sims & Jones, 1997), and was again substantiated in the present study, as younger offenders were more likely to be re-arrested than older offenders. Another predictive characteristic well supported by the prior literature involved the offender's criminal record. In prior studies the more prior offenses that were found on the offender's record the higher the probability that the offender would recidivate (Morgan, 1993, 1994; Sims & Jones, 1997). The present study found that more prior criminal convictions, and also more prior violent offense convictions, predicted re-arrest.
Finally, whether or not the offender had successfully completed a domestic batterer counseling program prior to his current sentencing was not significantly associated with re-arrest. The inability of such batterer intervention programs to measurably impact offender behavior was not inconsistent with the existing literature regarding the effectiveness of domestic batterer counseling programs, as the numerous empirical evaluations of these programs have produced inconclusive findings (Gondolf, 2004; Healey, Smith, O'Sullivan, 1998).
Some exceptions to the prior literature, however, were found in the present study. While Morgan (1993, 1994) and Sims and Jones (1997) found that married offenders were less likely to recidivate on probation than unmarried offenders, and Hanson and Wallace-Capretta (2004) found that whether or not the batterer was residing with his former victim was unrelated to recidivism, the present study found that residing with one's previous victim did increase the odds or re-arrest. This inconsistency with the prior research may be a difference unique to felony-level domestic batterers; however, replication must occur before this determination can be made. Furthermore, while race and education have often been correlated with probation success or failure, these characteristics were not significant here. These inconsistencies also need further investigation.
Community-based supervision on probation continues to be used extensively as a sentencing option for domestic batterers, including felony offenders. The importance of identifying true factors predictive of re-offending, therefore, is clear. Judges, when determining the most appropriate sentences for domestic batterers, need to have a clear understanding about the characteristics of defendants that suggest elevated levels of risk of re-offending. Likewise, probation officers need to be able to identify the characteristics of offenders that increase their odds of re-offending so that case planning can address these risk markers, and offenders at elevated risk can receive closer supervision. Such information may help increase the efficiency of the legal responses to domestic violence.