DV Re-Arrests 03

CJA

NEW YORK CITY CRIMINAL JUSTICE AGENCY

Jerome E. McElroy Executive Director

THE IMPACT OF CASE PROCESSING ON RE-ARRESTS AMONG DOMESTIC VIOLENCE OFFENDERS IN NEW YORK CITY

Richard R. Peterson, Ph.D. Project Director and Director, Research Department

FINAL REPORT March 2003

52 Duane Street, New York, NY 10007

(646) 213-2500

THE IMPACT OF CASE PROCESSING ON RE-ARRESTS AMONG DOMESTIC VIOLENCE OFFENDERS IN NEW YORK CITY

Richard R. Peterson, Ph.D. Project Director and Director, Research Department Barbara Geller Diaz Associate Director, Information Systems Elizabeth Walton Senior Research Assistant Raymond Caligiure Graphics and Production Specialist Taehyon Kim Junior Research Programmer Bernice Linen-Reed Administrative Assistant March 2003

This report can be downloaded from www.nycja.org\research\research.htm

達 2003 NYC Criminal Justice Agency

ACKNOWLEDGEMENTS

This report could not have been completed without the assistance of colleagues at CJA, Judges, members of the District Attorneys’ offices and others.

The author extends special thanks to Barbara Geller Diaz who did the programming to extract re-arrest data from the CJA database. Her suggestions for methods of accurately identifying re-arrests, her willingness to try additional methods suggested by the author, and her patience were all essential to the success of the project. The author also appreciates the assistance of Elizabeth Walton, who spent countless hours checking the accuracy of the matching process for thousands of re-arrests, writing memos to record how the matching was done, and adding new measures to the dataset. The author also acknowledges other colleagues at CJA who provided advice, information and editorial suggestions: Jerome E. McElroy, Executive Director of CJA, and Mari Curbelo, Marian Gewirtz, Peter Kiers, Mary T. Phillips, Jukka Savolainen, Frank Sergi, Qudsia Siddiqi and Freda F. Solomon.

The author would also like to thank Deirdre Bialo-Padin, Esq., Abena Darkeh, Esq., Robert C. Davis, Hon. Matthew J. D’Emic, Prof. Jo Dixon, Peter Glick, Esq., Darlene Haywood, Scott Kessler, Esq., Karen Kleinberg, Esq., Elisa Koenderman, Esq., Melissa Labriola, Sharon Lastique, Hon. John M. Leventhal, Wanda Lucibello, Esq., Audrey Moore, Esq., Hon. Esther M. Morgenstern, Joseph Muroff, Esq., Lieut. James Ranelli, Michael Rempel, Terri Roman, C.S.W., Deborah Tuerkheimer, Esq. and Amanda Voytek for their assistance with the project.

Finally, thanks to the New York State Division of Criminal Justice Services (DCJS) and the New York City Police Department (NYPD) for providing supplemental data.

The methodology, findings and conclusions of the study, as well as any errors, omissions and misinterpretations are the responsibility of the author.

TABLE OF CONTENTS I. Introduction...............................................................................................................1 A. CJA’s Research Agenda on Domestic Violence ...................................................1 B. Review of Prior Research .....................................................................................2 1. The Impact of Case Outcomes and Criminal Sanctions on Recidivism Among DV Offenders...................................................................................................2 2. The Impact of Prosecutorial Screening Policies on Recidivism Among DV Offenders .........................................................................................................6 C. Research Questions................................................................................................7 II. Methodology .............................................................................................................9 A. Overview of the CJA Database and the Third Quarter 1998 Dataset ...................9 B. Identifying Domestic Violence Cases.................................................................11 C. Victim-Defendant Relationship ..........................................................................13 D. Selection of the Crimes against Persons and Property Subsample .....................14 E. Analytic Issues ....................................................................................................15 1. Using a Defendant-based Data File................................................................15 2. Cross-Borough Comparisons of DV Cases.....................................................16 3. Plan of Analysis..............................................................................................16 III. Measuring Recidivism ..........................................................................................17 A. Relative Merits of Re-arrest as a Measure of Recidivism ..................................17 B. Selecting a Re-arrest Measure ............................................................................19 C. Time Period for Collecting Data on Re-arrests...................................................20 D. Advantages and Disadvantages of the Re-arrest Measure ..................................22 IV. Effects of Case Outcomes and Criminal Sanctions on Re-Arrest Rates for DV Offenses...................................................................................................................25 A. Measuring Case Outcomes and Criminal Sanctions...........................................25 B. Re-arrest Rates by Case Outcomes and Criminal Sanctions ..............................26 C. Logistic Regression Model Examining the Effect of Case Outcomes and Criminal Sanctions on the Re-arrest Rate for DV Offenses ...............................30 D. Summary and Discussion of Findings ................................................................38 V. The Effect of Case Screening Policies on Re-Arrest Rates .................................41 A. Measuring the Impact of Case Screening Policies..............................................41 B. Re-arrest Rates by Case Outcomes and Criminal Sanctions in the Bronx ........43 C. Logistic Regression Model Examining the Effect of Case Screening on the Re-arrest Rate for DV Offenses ..........................................................................44 D. Summary and Discussion of Findings ................................................................49

i

TABLE OF CONTENTS, CONTINUED VI. Conclusion ...............................................................................................................51 A. Major Findings....................................................................................................51 B. Discussion and Conclusions ...............................................................................52 C. Policy Implications .............................................................................................53 D. Future Research...................................................................................................54 VII. References .............................................................................................................57 Appendix A: Procedures Used for Identifying Re-Arrests ......................................61 A. NYSID Numbers.................................................................................................61 B. Procedures for Identifying Re-arrests .................................................................62 Appendix B: Statistical Methods ................................................................................65 Appendix C: Coding of Variables for Regression Models .......................................69

ii

LIST OF TABLES Table 4-1

Re-Arrest Rates by Case Outcomes and Criminal Sanctions in DV Cases...............27

Table 4-2

Re-Arrest Rates by Case Outcomes and Criminal Sanctions in Non-DV Cases .....30

Table 4-3

Logistic Regression Model Predicting Likelihood of Re-arrest for a DV Offense: DV Offenders Citywide ............................................................................................32

Table 5-1

Re-Arrest Rates for DV Offenses by Arrest Outcomes, Case Outcomes and Criminal Sanctions in DV Cases in the Bronx .........................................................45

Table 5-2

Logistic Regression Model Predicting Likelihood of Re-Arrest for a DV Offense: DV Offenders in the Bronx ...................................................................... 47

iii

THE IMPACT OF CASE PROCESSING ON RE-ARRESTS AMONG DOMESTIC VIOLENCE OFFENDERS IN NEW YORK CITY I. INTRODUCTION The criminal justice system’s response to domestic violence has changed dramatically over the last decade in many jurisdictions throughout the U.S. Domestic violence incidents are now being taken more seriously as a result of more frequent arrests, vigorous prosecution and harsher sentencing laws. Federal legislation, notably the Violence Against Women Act (VAWA) of 1994, has encouraged new efforts to combat domestic violence. Under this Act, the Department of Justice has provided funding and technical assistance to state and local law enforcement agencies.

The national trends are reflected in changes in New York State and New York City.

Recent state legislation includes mandatory arrest requirements for certain family offenses, enhanced penalties for violating an order of protection, a “primary physical aggressor” statute, and anti-stalking legislation. In conjunction with the new legislation, New York City promoted changes in the criminal justice response to domestic violence. Specialized police units, prosecution bureaus, and court parts were created to focus more attention on domestic violence cases, and to develop a consistent approach to handling these cases. New York City is now in the forefront of efforts to combat domestic violence.

These changes in New York City’s approach to domestic violence raise significant research questions. How are domestic violence cases processed in the courts? Are there borough differences in court processing? How often are offenders re-arrested on domestic violence charges? To address these questions, the New York City Criminal Justice Agency (CJA) has developed a research agenda on domestic violence.

A. CJA’s Research Agenda on Domestic Violence The current report is the third of a series of CJA reports on domestic violence. The first report compared the case processing of domestic violence (DV) and non-domestic violence (Non-DV) cases in New York City’s Criminal Courts (Peterson 2001). That report concluded that the conviction rate in Criminal Court was lower for DV cases than for comparable Non-DV cases. The lower conviction rate in DV cases was primarily due to a lack of victim cooperation in those cases. The report also found that the Criminal Courts relied primarily on batterer intervention programs rather than jail sentences for convicted DV offenders. This reflected the courts’ emphasis on monitoring defendant behavior through reports on program compliance.

CJA’s second report on domestic violence examined borough differences in the processing of domestic violence cases in New York City (Peterson 2002). The report found that prosecution policies for screening DV cases had a significant impact on the likelihood of prosecution and conviction. When a DA’s office relied primarily on first-party complaints, where the victim signs the complaint, many DV cases were declined for prosecution but the conviction rate was very high for the cases that were prosecuted. When a DA’s office used a

2 no-drop prosecution policy, cases were rarely declined for prosecution but the conviction rate was low. The report also found that boroughs that used specialized domestic violence court parts monitored convicted DV offenders through batterer intervention programs, whereas convicted DV offenders processed in all-purpose parts were more likely to be sentenced to jail. This difference reflected the emphasis of the specialized parts on monitoring DV offenders as a means of controlling their behavior.

In the current report, we examine rates of re-arrest for defendants in domestic violence cases. We determine whether case outcomes and criminal sanctions, such as conviction and jail sentences, affect re-arrest rates. We also examine how prosecutorial case screening policies affect re-arrest rates.1 B. Review of Prior Research In this section of the report, we discuss previous research conducted in the U.S. on recidivism in domestic violence cases and we develop research questions to be addressed in our study of New York City. Most of the research studies cited below define DV offenses to include acts that cause physical, psychological or financial harm to a victim who is an intimate partner or family member of the offender. Many of the studies focus specifically on cases of physical violence by men against women, the most common type of DV case. Non-DV offenses are usually defined as comparable acts against a victim who is not an intimate partner or family member of the offender, including acts against strangers, neighbors, friends, co-workers, etc.

The definition of domestic violence used in our analyses will be discussed in Section II of this report.

1. The Impact of Case Outcomes and Criminal Sanctions on Recidivism Among DV Offenders Most research on recidivism has focused on specific deterrence: identifying interventions that deter the offender from future acts of domestic violence.2 In the mid-1980’s, the “Minneapolis Arrest Study” (Sherman and Berk 1984), reported that arresting perpetrators of domestic violence reduced recidivism. The findings of this study provided support for dramatic changes in policing domestic violence that were already taking place. During the 1980’s, many police departments adopted a pro-arrest policy for domestic incidents, and many state legislatures passed mandatory arrest statutes. Mandatory arrest policies and statutes are considered an essential tool in dealing with domestic incidents.

1

We had also hoped to examine how the use of specialized court parts affect re-arrest rates, but were unable to do so for reasons that will be explained below in Section II-E.2.

2 Lerman (1992:225) argues that “specific deterrence of a particular offender is not the only goal.” She argues that general deterrence (deterring other potential offenders) and establishing community standards against domestic violence are also important goals. The current study does not address these goals because they are beyond the scope of the available data (Polsby 1992).

3 Interestingly, research has focused primarily on the impact of arrest on recidivism rates.

Few studies have examined the impact of pre-trial detention, prosecution, conviction, or jail sentences on recidivism (Lerman 1992). Yet there are significant policy questions about the impact of the criminal justice outcomes that follow arrest. Some have argued that arrest followed by conviction and a sanction may deter future domestic violence, while arrest followed by the dismissal of the case might not deter recidivism. Others have expressed concern about “escalation” effects: prosecuting DV offenders may increase the likelihood of recidivism. More severe criminal justice sanctions increase recidivism by isolating offenders, labeling them as criminals, and increasing their opportunity to associate with and learn from other offenders (see Sherman and Berk 1984 for a brief overview of this perspective). In DV cases in particular, some have expressed concern that more severe sanctions may lead the offender to become angry and retaliate against the victim (e.g., Ptacek 1999).

To address these issues, several studies have examined the impact of case outcomes and criminal sanctions on rates of recidivism. Each of the studies also examined other factors that might affect recidivism. Because our analyses extend and improve on previous research, we review the key elements of each of these studies.

Fagan et al.’s (1984) National Family Violence Evaluation Study interviewed a sample of battered women who sought assistance from federally funded programs in Florida, North Carolina, Ohio and Vermont in 1980. They found no statistically significant difference in recidivism over a 6-month period between men charged with spouse abuse who were convicted and those who were prosecuted and not convicted. However, the study did find that defendants who had a history of prior violence were more likely to re-offend.

Ford and Regoli (1992, 1993) used an experimental design to study recidivism rates of men charged with misdemeanor assault against their female partner in Indianapolis in 1986-87.

Cases pursued under a “no-drop” prosecution policy, where charges could not be dropped at the victim’s request, were assigned to one of three prosecutorial tracks: 1) diversion with counseling, 2) conviction with probation and counseling, and 3) conviction with traditional sentencing and without counseling. Recidivism rates during the 6 months following case settlement were virtually the same for the three prosecutorial tracks, suggesting that case outcomes and criminal sanctions had no effect on recidivism. The study did find two variables that influenced recidivism rates. Defendants who committed acts of domestic violence while their case was pending (i.e., between arrest and case disposition) were more likely to re-offend after case disposition. Also, defendants who cohabited with the victim after case disposition were more likely to be re-arrested.

Tolman and Weisz (1995) examined recidivism rates for domestic violence in DuPage County Illinois among men arrested in 1992 for violence against a wife, ex-wife, or cohabiting partner. Men who were convicted had lower rates of police contact, and of re-arrest, for domestic violence incidents over an 18-month period when compared to those who had their cases dismissed or who were found not guilty. However, these differences were not statistically significant. Police contact and re-arrest rates were significantly higher for men who had previous contact with the police or a previous arrest than for those who did not.

4

Davis et al. (1998) examined re-arrest rates of domestic violence misdemeanor offenders in Milwaukee in 1994-95. Specifically, they compared cases that were prosecuted and later dismissed to cases that resulted in conviction with a sentence of probation, and cases that resulted in conviction with a jail sentence. They found that case outcomes and criminal sanctions had no impact on re-arrest rates over a 6-month period. The only variables that were statistically significant predictors of re-arrest were the number of prior misdemeanor convictions and the number of prior battery arrests that did not result in conviction.

Murphy et al. (1998) studied re-arrest rates of male domestic violence offenders in Baltimore in 1994. Their measure of re-arrests included only re-arrests for battery or violating an order of protection, both of which they believed were likely to be for new DV offenses. Murphy et al. found that case outcomes and criminal sanctions had no impact on re-arrest rates over a 12 to 18 month follow-up period. The only variable that was a statistically significant predictor of re-arrest was age: older defendants were less likely to be re-arrested.

Gross et al. (2000) examined re-arrests for DV offenses among men convicted of DV offenses in Chesterfield County, Virginia (which includes the city of Richmond) in 1997. Over an 18 to 24 month follow-up period, they found that case outcomes and criminal sanctions had no impact on re-arrest rates.

Preliminary results from a more recent study also indicate that case outcomes and criminal sanctions for misdemeanor domestic violence offenders had no effect on subsequent abuse over a 6-month period, as measured through victim interviews (Sullivan et al. 2000). This study, conducted in three jurisdictions in the U.S., found that the strongest predictor of subsequent abuse was severity of prior abuse.

The results of these seven studies examining the effect of case outcomes and criminal sanctions on recidivism in domestic violence cases can be summarized easily: none of the studies cited has found that case outcomes and criminal sanctions have an impact on recidivism.

However, these studies have methodological problems that may have prevented them from finding an effect of case outcomes and criminal sanctions on recidivism rates. Sullivan et al.

(2000) relied on victim reports of case outcomes and criminal sanctions, which may be unreliable. In Ford and Regoli’s (1992) experimental design, 25% of the cases did not receive the assigned treatment (Davis et al. 1998), making it more difficult to detect a treatment effect.

The sample sizes in Fagan et al.’s (1984) study (N = 74), Tolman and Weisz’s (1995) study (N = 175), Murphy et al.’s (1998) study (N = 235), Gross et al.’s (2000) study (N = 177) and Sullivan et al.’s (2000) study (N = 160) were small, which limited the power of their statistical tests to detect an effect of conviction. The sample sizes in Davis et al.’s (1998) study (N = 574) and Ford and Regoli’s (1992, 1993) study (N = 538) were larger, but the number of cases in each of the case outcome and criminal sanction categories was still relatively small (between 154 and 220 cases). Davis et al.’s (1998) measure of recidivism included any re-arrest, whether or not it was for a DV offense, and Murphy et al.’s (1998) study included re-arrests for battery or violating an order of protection, whether or not these were DV offenses. Finally, four of the seven studies used a short (6-month) follow-up period for tracking recidivism. These limitations

5 of prior research suggest that it is premature to conclude that case outcomes and criminal sanctions have no effect on recidivism. Further research with larger sample sizes and a longer follow-up period is needed.

There is one additional study that stands out from those already cited, because it overcame many of the methodological limitations of previous studies. Wooldredge and Thistlethwaite (2002)3 used a much larger sample, limited their recidivism measure to re-arrests for DV assaults, and used a 2 year follow-up period. Their sample included 3,110 domestic violence misdemeanor offenders in Hamilton County, Ohio (which includes the city of Cincinnati) in 1993 and 1998. They examined four categories of case outcomes and criminal sanctions: cases that ended in dismissal or acquittal, cases where the defendant was sentenced only to a counseling program, cases where the defendant was sentenced to jail and/or probation with a counseling program, and cases where the defendant was sentenced to jail and/or probation without a counseling program. After controlling for the effects of demographic and case characteristics, there were no differences in re-arrest rates among the 4 categories of case outcomes and criminal sanctions. This finding, based on a larger sample, is consistent with the findings of previous studies.

Wooldredge and Thistlethwaite (2002) also found that prior criminal history was associated with the likelihood of re-arrest. Defendants with a larger number of prior convictions for violent crimes were more likely to be re-arrested for a DV offense. Additionally, men, younger defendants, those who lived with their partner at the time of the arrest, and those with pending charges at the time of arrest were all more likely to be re-arrested.

Interestingly, Wooldredge and Thistlethwaite (2002) also examined whether the defendant’s “stake in conformity” (also referred to as “community ties”) affected the likelihood of re-arrest for DV offenses. According to this hypothesis, “individuals with a greater stake in conformity are more likely to be deterred through formal criminal justice sanctioning because they have more to lose than their counterparts” (Thistlethwaite et al. 1998, p. 391). Wooldredge and Thistlethwaite (2002) found that defendants who lived at the same residence for 5 years or more were significantly less likely to be re-arrested than those who lived at their residences for less than 5 years. This difference was statistically significant after case outcomes, criminal sanctions, prior record and other variables were taken into account.

While the Wooldredge and Thistlethwaite (2002) study is suggestive, further research is needed to explore their findings. Because they overcame many of the limitations of previous studies, their findings have greater credibility. However, their findings about the impact of case outcomes and criminal sanctions need to be replicated with data from other jurisdictions.

Furthermore, their finding on the effect of community ties needs to be explored more carefully, and with a wider range of measures.

3

A preliminary report on this study, based on a smaller sample, reported some findings inconsistent with those reported here (see Thistlethwaite et al. 1998). These differences appear to be due to measurement problems and the smaller sample in the earlier report. Consequently, results reported here are based on the later study (Wooldredge and Thistlethwaite 2002).

6

In summary, prior research provides little support for the hypothesis that more severe case outcomes and criminal sanctions in domestic violence cases reduce recidivism. No study has found that convicted defendants are less likely to re-offend than those who are not convicted.

Nor has any study found that more severe sentences for convicted defendants reduced recidivism.

Although most of the studies had methodological problems, the one study that overcame these problems also confirmed that case outcomes and criminal sanctions have little impact on re-arrest rates. If case outcomes and criminal sanctions don’t affect re-arrest rates, what factors do have an impact? All the studies examined the impact of criminal history, and with the exception of Murphy et al. (1998), concluded that a history of prior arrests or prior convictions is associated with higher re-arrest rates. Only one study examined the impact of community ties on re-arrest rates, reporting that defendants who were residentially stable were less likely to be re-arrested.

Studies with a strong research design are needed in other jurisdictions to corroborate the findings on the impact of case outcomes and criminal sanctions, and to explore further the impact of community ties.

2. The Impact of Prosecutorial Screening Policies on Recidivism Among DV Offenders Jurisdictions throughout the U.S. use a variety of prosecutorial screening policies for deciding which DV cases will be prosecuted and which will not. In some jurisdictions, prosecutors rely on “first-party complaints,” which require the victims of domestic violence to sign the complaint. However, because many victims are reluctant to press charges, prosecutors in many jurisdictions have developed alternative policies and practices. In many district attorneys’ offices prosecutors or the police sign the complaint. This reduces the risk that the victim will become the target of retaliation or pressure from the defendant to drop the charges (Lerman 1986). To encourage victims to cooperate after charges are filed, DA’s often use victim advocacy programs to provide counseling and support services (Cahn 1992). These programs provide information about how the criminal justice system works, and address victims’ concerns about housing, financial support, fear of retaliation, etc. This approach can reduce the victim’s isolation and inform the victim about the potential benefits of prosecution (McGuire 2000).

However, even when prosecutors file the complaints and make services available to the victims, many victims in DV cases refuse to cooperate with the prosecution.

Some jurisdictions have a “drop-permitted” policy, and dismiss the case when efforts to encourage and maintain victim cooperation fail. However, in many jurisdictions, more aggressive efforts are made to pursue these cases. Some district attorneys’ offices have established “no-drop” or “mandatory” prosecution policies that require prosecutors to go forward with most DV cases even when the victim refuses to cooperate. This action makes it clear to the defendant that the state, and not the victim, is responsible for prosecuting the case (Cahn and Lerman 1991, Herrell and Hofford 1990). Nevertheless, many victims cannot be reached by the DA’s office or refuse to cooperate or to testify (McGuire 2000). Under no-drop prosecution policies, “evidence-based prosecution” of DV cases is often used when victims are not cooperating. This practice uses “hearsay exceptions” (such as “excited utterances” on 911 tapes), photographs, police testimony, medical reports and physical evidence to build a case without the cooperation of the victim (Stone 2000). Of course, in many cases such evidence is unavailable or

7 insufficient to make a viable case. If evidence-based prosecution is not possible and the victim is unwilling to testify if the case goes to trial, the case is likely to be dismissed.

Studies of the impact of prosecutorial screening policies on recidivism are rare. A few of the studies described above included data on cases that were not prosecuted. Fagan et al.’s (1984) study included an additional 196 defendants whose cases were declined for prosecution.

They found no difference in victim reports of violence over a 6-month period between offenders who were prosecuted and those who were not. Davis et al.’s (1998) sample (also described above) included an additional 464 defendants whose DV cases were declined for prosecution.

Their re-arrest rates over a 6-month period were similar to the re-arrest rates of defendants whose cases were prosecuted. Murphy et al.’s (1998) study included 69 defendants whose DV cases were deferred; their re-arrest rate was similar to the rate for defendants who were prosecuted.

However, Wooldredge and Thistlethwaite (2002) did find a statistically significant difference in re-arrest rates between 2,892 offenders whose cases were prosecuted and 218 offenders in cases where no charges were filed. The re-arrest rate was 24% when no charges were filed, compared to 15% for cases that were prosecuted. The difference remained statistically significant even after taking into account the impact of other variables. Their findings suggest that once arrested, prosecuting an offender reduces the likelihood of re-arrest for domestic violence.

As noted earlier, however, the conclusions of most of these studies were affected by methodological problems. The sample sizes were fairly small in the three studies that found no effect of prosecution on re-arrest rates. Wooldredge and Thistlethwaite (2002), with a larger sample, did find that prosecution reduced the re-arrest rate. Further research is needed to examine the impact of prosecution in other jurisdictions.

C. Research Questions In summary, very little research has been done to evaluate the impact of case outcomes, criminal sanctions or prosecutorial screening policies on recidivism, either nationally or in New York City. The current study provides information about re-arrest rates among those arrested for DV offenses in New York City in the third quarter of 1998. Because our sample sizes are larger than those of any previous study, our statistical tests are powerful enough to detect effects of case outcomes and criminal sanctions on re-arrest. Unlike many of the studies cited above, we track re-arrests over an 18-month period rather than a 6-month or 12-month period. Our statistical models improve on the models used in previous research because we include better measures and a variety of control variables, enabling us to isolate the effects of case outcomes, criminal sanctions and prosecutorial screening policies on re-arrest rates. Our study addresses the shortcomings of prior research and provides valuable information about the re-arrest rates of DV offenders in New York City. Specifically, the current report addresses the following questions: 1) What are the effects of case outcomes and criminal sanctions on re-arrest rates for DV offenders?

2) What are the effects of prosecutorial case screening policies on re-arrest rates for DV offenders?

8

The hypotheses to be tested are: 1) Defendants in DV cases with more severe case outcomes and criminal sanctions have lower rates of re-arrest for DV offenses.

2) Arrestees whose DV cases are declined for prosecution have higher rates of re-arrest for DV offenses than those whose cases are prosecuted.

9 II. METHODOLOGY A. Overview of the CJA Database and the Third Quarter 1998 Dataset The data for this study were drawn primarily from the CJA database. This database contains information about the arrest, case processing and case outcomes of most New York City arrestees. The CJA database includes data from three sources: CJA’s pre-arraignment interview,4 the New York City Police Department’s Online Booking System (OLBS) Database, and the New York State Office of Court Administration (OCA). Information concerning demographic characteristics and the community ties of the offenders is taken from the CJA pre-arraignment interview. Information about the arrests is based on the OLBS data. Detailed Criminal Court and Supreme Court processing and outcome data on each of the arrests are drawn from the OCA data.

This report is based on analyses of the Third Quarter 1998 Dataset, which consists of data collected on a three-month cohort of arrests made from July 1, 1998 through September 30, 1998 (Eckert and Curbelo 2000). The dataset includes information on 89,524 arrests where the district attorney elected to bring charges and where a docket number was assigned.5 The third quarter 1998 dataset was supplemented with information on re-arrests that occurred within 18 months of the disposition of the third quarter 1998 case. The details of our construction of the re-arrest data file are described in more detail in Section III of this report.

In addition to information in the CJA database, the Third Quarter 1998 dataset also includes information provided by the New York State Division of Criminal Justice Services (DCJS).6 DCJS data were used to supplement and check the reliability of criminal history information that was routinely collected by CJA interviewers.7 4

CJA conducts pre-arraignment interviews to measure the defendant’s community ties and to serve as the basis for making a recommendation as to whether or not the defendant should be released on recognizance at his or her first court appearance. Defendants who are arrested on a bench warrant, given a Desk Appearance Ticket (DAT), or who are held for arraignment on prostitution charges in the downtown Manhattan Criminal Court, are not interviewed by CJA. CJA collects police arrest and Criminal Court information for all arrestees, and they were included in the Third Quarter 1998 Dataset whether or not they were interviewed (Siddiqi 1999).

5 For a more detailed discussion of how the sample was drawn, see Eckert and Curbelo 2000, Appendix A. The sample includes both Summary Arrests and DAT’s. Arrests that had no docket number were retained in the sample if they appeared to be either “A” docket cases in Manhattan (the designation used in Manhattan for a court case that has a docket number with the suffix “A” to distinguish it from a court case that has the same docket number without the suffix “A”) or direct indictments. Arrest information for these two types of cases was supplemented with defendant information and court processing information when available.

6 DCJS, OCA, and the NYPD are not responsible for the methods or conclusions of this report.

7 CJA provided fingerprint identification numbers (known as New York State Identification, or “NYSID,” numbers) to DCJS for all printable offenses in the third quarter of 1998—approximately 81,000 of 89,524 arrests. NYSID numbers for the remainder of the arrests were not available in CJA’s database, since the arrests were for nonprintable offenses. DCJS returned information on the defendant’s

10

For cases that had multiple dockets, case-processing information in this study is based on the docket that had the most severe arraignment charge (based on Penal Law severity) in Criminal Court.8 When the most severe arraignment charges on 2 or more dockets are of equal Penal Law severity, the top charge is determined according to guidelines developed by OCA. These guidelines provide a consistent set of rules for determining which of two arraignment charges of equal severity will be identified as the top arraignment charge.

In New York State’s two-tiered court system for handling criminal cases, the Criminal Courts only have trial jurisdiction over cases having a most serious charge of misdemeanor or lesser severity. Most defendants charged with felonies are first arraigned in Criminal Court.

Cases sustained at the felony level must be brought for prosecution in Supreme Court. In felony cases where the DA decides not to prosecute the case in Supreme Court (or the Grand Jury fails to return an indictment), the case may be disposed in Criminal Court by dismissal or by a plea to a reduced charge less severe than a felony, or by a transfer to another court’s jurisdiction (e.g., Family Court).

The cases selected for inclusion in the analyses in this report include only those third quarter 1998 cases that reached a final disposition in Criminal Court. Our research was originally designed to examine both Supreme Court and Criminal Court cases.9 However, the overwhelming majority (about 98%) of domestic violence cases citywide are disposed in Criminal Court. The Third Quarter 1998 Dataset includes case processing information in Criminal Court through final disposition (and sentencing, if there was a conviction), or until August 6, 1999, if the case was not yet disposed. Information about the final disposition in Criminal Court beyond this cutoff date was not included in the dataset.

To inform our discussion of the statistical results, we also conducted field research in New York City criminal courts and interviews with representatives of the District Attorney’s offices in Brooklyn, Manhattan, Queens and the Bronx. We observed the operation of specialized DV Criminal Court parts in Brooklyn and the Bronx to develop an understanding of the problems posed by domestic violence cases and the strategies used by DA’s and judges to overcome these problems. We also observed the operation of Brooklyn’s two specialized DV Supreme Court parts. We consulted with judges as well as other key court personnel, including assistant district attorneys, defense attorneys, domestic violence resource coordinators, and criminal history for 46,000 of the 81,000 cases. In the remaining cases, the defendant had no prior convictions on his/her record, or the court had sealed the current case.

8 New York State Penal Law categorizes most offenses according to their severity. The most serious crimes are A felonies, followed by felonies classified as being of severity B through E. Misdemeanors are less severe than felonies, and are classified as A or B misdemeanors (A misdemeanors are more severe). Violations are less severe than misdemeanors, and are not considered crimes, although they can result in jail sentences. No distinctions of severity are made within the category of violations.

9 The Family Courts have concurrent jurisdiction over certain domestic violence cases (Aldrich and Domonkos 2000). Some DV cases are heard only in Criminal Court, some are heard in both Criminal Court and Family Court, and others are heard only in Family Court. We do not have access to data on DV cases that are heard only in Family Court, and our report draws no conclusions about these cases.

11 probation officers. Our field research and interviews provided valuable information about the special features and problems of domestic violence cases, and the prosecution and adjudication of these cases.

B. Identifying Domestic Violence Cases Social scientific and legal definitions of domestic violence have changed over the last 30 years. We reviewed the history of these changes in our first report (Peterson 2001, p. 11). In the current report, we review how New York City courts identified DV cases in 1998, and the differences in how DV cases were handled in each borough.

In New York State the statutory definition of domestic violence approximates what has come to be known in the social scientific literature as “family violence.” Under New York State’s Criminal Procedure Law (CPL) §530.11 (as amended by the 1994 Family Protection and Domestic Violence Intervention Act), family offenses are defined as offenses committed against a member of the same family or household, where “family or household” are defined as: (1) persons related by consanguinity or affinity; (2) persons legally married to each other; (3) persons who were formerly married, and; (4) persons who have a child in common, whether or not they have ever been married or lived together.

New York State’s statutory definition of domestic violence excludes unmarried partners, unless they have a child in common. However, the New York City Police Department (NYPD) operates with an expanded definition of domestic violence that includes individuals who are not married, but who are cohabiting or have previously lived together. This NYPD definition of “family” expands on New York State law by including “common-law” marriages, same-sex couples, and registered New York City domestic partners (NYPD 2000). By citywide agreement, the DA’s offices and the Criminal Courts in all five boroughs also use this expanded definition to identify DV cases, whether or not the relationship between the victim and defendant meets the New York State statutory requirements contained in CPL §530.11.

To identify domestic violence cases, Assistant District Attorneys (ADA’s) use information collected by the police about the relationship between the victim and the defendant, if any. ADA’s also often ask victims about their relationship with the defendant. When this information indicates that the victim-defendant relationship meets the NYPD expanded definition of domestic violence, the case is identified as a DV case. DV case files in all 5 boroughs are then given beige “backs” (special color-coded back sheets) to distinguish them from other case files. At Criminal Court arraignment, Court Clerks assign an arraignment hearing type of “DV” to domestic violence cases, and this designation is entered in OCA’s computerized court records.

At the time the defendants in the Third Quarter 1998 Dataset were arrested, cases identified at arraignment as DV cases were processed in different ways depending on the borough. In Manhattan, all cases with a DV hearing type were sent to all-purpose parts for post-arraignment appearances. (Manhattan now has specialized DV court parts, but these parts were not yet in operation in 1998.) In all boroughs except Manhattan, most cases with a DV

12 hearing type were sent to a specialized Criminal Court domestic violence part for post-arraignment appearances. However, there were exceptions. In Brooklyn and the Bronx, cases with a DV hearing type that involved physical or sexual abuse of children or other types of non-intimate partner violence (e.g., violence between siblings) were not sent to the specialized domestic violence parts. Finally, some cases that did not have a DV hearing type at arraignment were also sent to the specialized DV parts, presumably because information that these cases involved domestic violence became available only after arraignment.

In this study, we identified “domestic violence” cases by relying on the court’s identification of these cases. We used information about both hearing type and court part, since not all DV cases were assigned a DV hearing type. We identified cases as domestic violence cases if the Office of Court Administration reported that: (1) the case had a domestic violence hearing type at Criminal Court arraignment, and/or, (2) the case had one or more appearances in a specialized domestic violence Criminal Court part.10 Using these criteria, we identified 7,591 domestic violence cases in the Third Quarter 1998 Dataset, about 8.5% of the total sample of 89,524 cases. Of the 7,591 cases identified by the courts as involving domestic violence, 4,192 cases (55% of 7,591) had both a domestic violence hearing type at arraignment and at least one appearance in a specialized domestic violence court part. An additional 2,036 cases (27%) had a domestic violence hearing type at arraignment, but no appearance in a specialized domestic violence part. These included cases in Manhattan where there was no specialized DV part (N = 1,147), DV cases in other boroughs that were disposed at arraignment (N = 198), as well as DV cases in the other boroughs that were sent to Non-DV parts (N = 691). There were also 1,363 cases (18%) that had at least one appearance in a specialized domestic violence part, but did not have a domestic violence hearing type at Criminal Court arraignment.

Our method of identifying domestic violence cases is appropriate for the purposes of this study. The cases that we identified as DV cases were clearly known to the courts as DV cases.

Since this report examines how the processing of DV cases affected re-arrest rates, it is important that we examine only cases where the DA’s, judges and other key personnel were aware that the case was a DV case.

As we noted in our first report, the measure identifying DV cases in the third quarter of 1998 does have some limitations. First, there may be instances where a DV case was not identified as such in court records (i.e., it did not receive a DV hearing type at Criminal Court arraignment and did not appear in a specialized domestic violence part). The measure did not identify these as DV cases but instead categorized them as Non-DV cases. In current study, this limitation affected our analyses in two ways. First, it reduced the sample size of DV cases on 10

In the third quarter of 1998, the specialized domestic violence Criminal Court parts were AP-12 and AP-15 in Brooklyn, AP-10 and TAP-2 in the Bronx, AP-4 in Queens, and AP2-DV in Staten Island.

Although AP2-DV in Staten Island was identified as a separate court part in our data, it was actually a specialized DV calendar. DV cases on this calendar were heard in an all-purpose part two days a week.

We identified cases as DV cases if they had one or more appearances on this calendar.

13 which we report. Nevertheless, we have an adequate sample size for our analyses, and were able to draw valid conclusions about DV cases from our sample. Second, we slightly overestimated the number of Non-DV cases in the sample. However, the number of DV cases misidentified as Non-DV cases is likely to be a very small proportion of the total number of Non-DV cases.

Overall, we believe that the possible misidentification of DV cases is likely to be minimal, and that these misidentified cases had little impact on the results of our analyses.

A second problem with our definition of domestic violence is that it relies, in part, on identifying cases that appeared in specialized domestic violence courts. At the time of this study, Manhattan had no specialized domestic violence parts while each of the other boroughs had at least one specialized Criminal Court DV part and Brooklyn also had two specialized DV Supreme Court parts. In our Criminal Court sample, we identified 990 DV cases in Manhattan, based solely on DV hearing type at arraignment. Since not every domestic violence case receives a DV hearing type at arraignment, we are almost certainly failing to identify some domestic violence cases in Manhattan. It is difficult to gauge the magnitude of the problem, but it is possible to generate a rough estimate. Specifically, by examining DV cases that we identified in Brooklyn, the Bronx, Queens and Staten Island, we can determine what proportion of these cases had one or more appearances in a specialized domestic violence part but did not have a DV hearing type at arraignment. About 21% of the DV cases in Brooklyn, the Bronx, Queens and Staten Island were identified as having an appearance in a specialized DV part, but did not have a DV hearing type at arraignment. Assuming that the same pattern holds for Manhattan, about 21% of DV cases in Manhattan were not identified by our measure. Stated another way, we have identified about 79% of the DV cases in Manhattan during the third quarter of 1998.

C. Victim-Defendant Relationship To provide additional information about the DV cases, we also used data from the NYPD about the nature of the relationship between the defendant and the victim. Unfortunately, information about the relationship was missing for about 25% of all DV cases. At the time of the arrest, the police may not have been aware that there was a relationship between the parties. In 61% of the cases, the relationship was categorized as an “intimate partner” relationship (boyfriend-girlfriend, “common-law” spouse, married). In the remaining cases (14%), the relationship was a family relationship that did not involve intimate partners (parent-child, sibling, uncle-niece, etc.). Because we did not have information about the age of the victim, we were unable to clearly identify cases of child abuse or elder abuse using the NYPD data.11 Nevertheless, we were able to distinguish between “intimate partner” violence and other types of domestic violence, a distinction that has been important in the social science literature.

For our analyses of cases that were declined for prosecution, we used the information about the victim-defendant relationship to determine whether or not the case was a DV case.

Since these cases were never arraigned, and never had any court appearances, we could not rely 11

For example, when the defendant is identified as the parent of the victim, we do not know if the child was a minor under the age of 18, or an adult. Similarly, when the defendant is identified as the son or daughter of the victim, we do not know if the victim was elderly or not.

14 on DV hearing type at arraignment or court appearances in a DV part to determine which were DV cases. The NYPD information about the victim-defendant relationship was the only information available to determine whether cases declined for prosecution were DV cases. To be consistent with the NYPD expanded definition of domestic violence described above, we categorized both intimate partner and other family relationships as DV cases.

D. Selection of the Crimes Against Persons and Property Subsample The current study used data from the Crimes Against Persons and Property (CAPP) Subsample described in a previous report (Peterson 2001).12 This subsample was selected so that we could examine domestic violence cases across a wide range of charges. We began by selecting cases for this subsample where there was an alleged attempt to cause injury or where an overt threat of injury was made (Weis 1989). We used the most severe arraignment charge (based on Penal Law severity) to determine the nature of the offense, since this charge determines how the case is handled in the court system. We did not use the most severe arrest charge, which reflects charging decisions made by the police. We initially selected all cases that had a top (i.e., most severe) arraignment charge from any of the following New York State Penal Law articles: PL 120 (Assault), PL 130 (Sex Offense), PL 160 (Robbery), PL 260 (Crimes Against Children), or PL 265 (Weapons).13 Unfortunately, we were not able to include cases disposed in Criminal Court that had top arraignment charges from PL 125 (Homicide), PL 150 (Arson), and PL 135 (Kidnapping).14 Cases charged with offenses in these Penal Law articles were excluded since there were less than 50 domestic violence cases with top arraignment charges in each of these Penal Law articles—too few cases for reliable multivariate analysis.

Recognizing that domestic violence often includes offenses that result in financial and psychological harm, rather than just physical harm, we also selected cases if they had a top arraignment charge from one of the following Penal Law articles: PL 140 (Burglary), PL 145 (Criminal Mischief), PL 155 (Larceny), PL 205 (Escape and Resisting Arrest), PL 215 (Criminal Contempt),15 and PL 240 (Public Order Offenses).

Within each Penal Law article, we selected only those cases that had charges that could plausibly include elements comparable to those found in domestic violence cases. For example, cases with a top arraignment charge of Assault in the Third Degree (PL §120.00) were included 12

In our first report, we also discussed results for an Assaults Subsample, which included all cases where the top arraignment charge was assault (PL 120). We used the Assaults Subsample to provide a more focused comparison of DV cases to similar Non-DV cases. In this study, we focus primarily on DV cases, and make only a few comparisons of DV to Non-DV cases. We therefore use only the Crimes Against Persons and Property Subsample for our analyses. This subsample includes information about the full range of DV offenses, including assaults as well as violations of orders of protection, crimes against children, etc.

13 Because we are interested in court processes and the decisions of prosecutors and judges, we selected cases based on the most severe arraignment charge (based on Penal Law severity). We did not use arrest charges because these reflect charging decisions made by the police.

14 Most, but not all, of these cases were sustained as felonies and disposed in Supreme Court.

15 Penal Law article 215 includes violations of orders of protection.

15 in the subsample. However cases of Gang Assault in the First Degree (PL §120.07) or Gang Assault in the Second Degree (PL §120.06) were excluded, since it is unlikely that a domestic violence case would include a gang assault charge. Similarly, in PL 240, prostitution charges (PL §240.37) were excluded. These restrictions yielded a subsample of 32,299 cases, including 7,383 domestic violence cases.

We then narrowed the subsample further to identify an appropriate group of cases for the analysis. First, we limited the sample to Summary Arrests (i.e., cases in which the defendant was held in custody pending Criminal Court arraignment), excluding cases where the defendant was issued a Desk Appearance Ticket (DAT) and released by the arresting officer. DAT’s are rarely issued in DV cases. We also excluded cases with juvenile defendants (under age 16), cases that did not reach a final disposition by the cutoff date, and cases that were missing data on defendant’s criminal history or sex. We also excluded cases that were disposed in Criminal Court on Vehicle and Traffic Law (VTL) or Administrative Code (AC) charges. After these exclusions, the Crimes Against Persons and Property Subsample included 28,110 cases, of which 6,989 (about 25%) were domestic violence cases.16 E. Analytic Issues 1. Using a Defendant-based Data File The dataset analyzed in this study is a defendant-based data file that includes information on all defendants who were arrested in the third quarter of 1998. While the vast majority of defendants had only one case in the case-based data file,17 some defendants were arrested two or more times during the third quarter of 1998. For most defendants who had multiple arrests, we included information in the defendant-based data file about the first arrest, i.e., the arrest that occurred earliest during the third quarter of 1998. However, for defendants whose first arrest was not for a DV case but who had a second or subsequent DV arrest, we selected the first DV arrest, not the first arrest. This procedure enabled us to identify all the defendants who had at least one DV arrest during the third quarter of 1998.

When we supplemented the defendant-based data file with information about re-arrests, we began by determining whether each defendant had a qualifying re-arrest (see discussion in 16

Following are the numbers of cases excluded by each restriction (number of DV cases shown in parentheses). Beginning with 32,299 cases (7,383), we excluded 1,861 DAT’s (39), 198 juveniles (7), 1,781 cases that did not reach a disposition by the cutoff date (323), 134 cases missing data on criminal history (12), 15 cases missing data on sex (5), and 200 cases that were disposed on VTL or AC charges in Criminal Court (8). Some cases were excluded on multiple grounds (e.g., missing data on criminal history and disposed on VTL charges). These cases are counted only once in this tally.

17 After excluding Supreme Court cases and cases not disposed by August 6, 1999, and adding cases declined for prosecution, the Crimes Against Persons and Property Subsample includes information about 26,631 cases for 23,322 defendants. About 90% of these defendants have only one case in the data file, 8% had two cases initiated during the third quarter of 1998, and 2% had 3 or more cases (the maximum was 9 cases).

16 Section III below) during the third quarter of 1998. We also extracted information from the CJA database about re-arrests (if any) after the third quarter of 1998.

2. Cross-borough Comparisons of DV cases We had hoped to examine the impact of the use of specialized DV court parts in the current study. Our plan was to compare re-arrest rates in Brooklyn and the Bronx, which had specialized DV court parts, to re-arrest rates in Manhattan, which did not. That strategy was used successfully in an earlier report examining the impact of specialized DV court parts on case outcomes in DV cases (Peterson 2002). However, that strategy is not feasible for the current study, which focuses on re-arrest rates. Re-arrests for a DV offense were less likely to be classified correctly as DV re-arrests in Manhattan than in other boroughs. (Section III-B, below, explains how DV re-arrests were identified). As a result, the rate of re-arrests for DV offenses in Manhattan was considerably lower (about 1/3 lower) than in Brooklyn or the Bronx. While this lower rate could reflect a real difference between Manhattan and the other boroughs, we are not confident enough in this conclusion to use the data. Our reluctance to use the data is based on our finding that the overall rate of re-arrests (for any offense, DV or Non-DV) was virtually the same in Manhattan as in Brooklyn and the Bronx. Since the overall rates of re-arrest were the same across boroughs, it seems likely that the lower re-arrest rate for DV offenses in Manhattan was due to weakness in the measure of DV re-arrests. The alternative explanation for this pattern of results, that the rate of DV re-arrests is lower in Manhattan than in Brooklyn or the Bronx, while the rate of Non-DV re-arrests is higher in Manhattan than in Brooklyn or the Bronx, seems implausible. It is more likely that many of the re-arrests in Manhattan that were classified as Non-DV re-arrests were actually DV re-arrests. Therefore, in the current study we have decided not to assess the impact of specialized DV court parts on re-arrest rates.

3. Plan of Analysis In the next section (Section III) we discuss how to measure recidivism and describe the re-arrest measure used in the current study. We then report re-arrest rates by case outcomes and criminal sanctions, and we present statistical models to assess the influence of these outcomes and sanctions on likelihood of re-arrest (see discussion in Section IV below). Next, we examine the impact of prosecutorial case screening policies by comparing cases that were declined for prosecution to cases that were prosecuted (see discussion in Section V below). We then present statistical models to assess the influence of case screening policies on the likelihood of re-arrest.

The examination of re-arrests focuses on the two major questions posed in the Introduction to this report: 1) What are the effects of case outcomes and criminal sanctions on re-arrest rates for DV offenders?

2) What are the effects of prosecutorial case screening policies on re-arrest rates for DV offenders?

17 III. MEASURING RECIDIVISM We measured recidivism in this study by examining re-arrests among defendants in the Third Quarter 1998 dataset. In this section, we discuss several issues regarding the use of re-arrest as a measure of recidivism: 1) What are the relative merits of re-arrest versus other measures of recidivism? 2) What types of re-arrests should qualify as incidents of recidivism?

3) Over how long a time period should we collect data on re-arrests? 4) What are the advantages and disadvantages of our measure of recidivism?

A. Relative Merits of Re-arrest as a Measure of Recidivism Prior research on domestic violence offenders has used three different measures of recidivism: reports of new DV incidents based on interviews with victims, calls to the police for a domestic incident, and re-arrest. Each of these measures has advantages and disadvantages, and there is an extensive literature discussing their relative merits.

Many researchers believe that interviews with victims provide the most complete assessment of recidivism (Ford and Regoli 1992). Interviewers can ask about incidents that did not result in calls to the police or in re-arrest of the defendant. Since many incidents are not reported to the police, measures based on victim interviews generally indicate higher rates of recidivism than measures based on calls to the police or on re-arrests (Hirschel and Hutchison 1996). Perhaps more importantly, calls to the police and re-arrests are unlikely to be a representative sample of all domestic incidents. Police involvement is more likely when injuries are more severe, when the parties are unmarried, and when the victim believes that police intervention will be helpful rather than harmful. In some situations, victims may believe that police intervention will result in an escalation of the violence, and may therefore be reluctant to seek help.

Victim interviews, however, have at least two serious weaknesses. First, most studies have found it very difficult to reach victims and to complete interviews with them (Binder and Meeker 1988, Smith et al. 2000, Tolman and Weisz 1995). Victims often move, change their telephone numbers, or refuse to speak to interviewers as a result of fear that the batterer will learn about the interview and retaliate (Hirschel and Hutchison 1996). While it is often possible to obtain comprehensive information from victims who agree to an interview, no information is available from the many victims who are not interviewed. It is very unlikely that interviewed victims are representative of all victims. Victims who are in the greatest danger, victims whose contact with outsiders is controlled by the batterer, and victims who have left the batterer are less likely to be interviewed than other victims.

A second problem with victim interviews is that they ignore the possibility that the defendant has re-offended with a new victim. Studies based on victim interviews are limited to information about new offenses by the same defendant against the same victim (McCord 1992).

However, most recidivism studies define recidivism to include any new domestic violence offenses by the defendant, whether these are committed against the same victim or a new victim.

Some DV offenders cease contact with the original victim and start a new relationship, which

18 may lead to the battering of a new victim. Studies based on police contact and re-arrest, which track the defendants rather than the victims, have the possibility of measuring recidivism in this situation. Studies based on interviews with the original victim are unable to measure this recidivism.

Calls to the police are another commonly used measure of recidivism. Studies using this measure generally identify all incidents of “domestic disturbances” where the police were called to the scene. Information in police reports is used to identify offenders and to learn about the nature of the alleged offense. The studies that use police calls as a measure of recidivism generally report higher rates of recidivism than studies that use re-arrest. Calls to the police may provide a more comprehensive measure of recidivism than re-arrest, since many calls do not result in an arrest. Moreover, the decision to arrest is often affected by factors unrelated to the nature or severity of the incident, e.g., whether the defendant was present at the scene when the police arrived, whether the victim was willing to sign a complaint, whether children were present when the incident occurred, whether the offender was belligerent, etc. (Elliott 1989, Feder 1998, Frisch et al. 2001).

Calls to the police have two major weaknesses as a measure of recidivism; these weaknesses have opposite effects on estimates of the amount of recidivism. On the one hand, calls to the police may overestimate recidivism because they include calls that prove to be erroneous or unfounded (Elliott 1989). For example, neighbors hearing a loud argument may call police, but on arrival at the scene, police may find no indication that an offense has been committed. On the other hand, calls to the police may underestimate recidivism because victims and neighbors may be reluctant to call the police, even when an offense has occurred. Fear of retaliation, concern that the courts may punish the offender too harshly, and concern about lost income if an offender is arrested and incarcerated are among the reasons that domestic violence victims may be reluctant to call the police (Rennison and Welchans 2000).

Many studies use re-arrest data to measure recidivism. Re-arrest measures have two main advantages over other measures. First, re-arrest data are potentially available for all offenders.

Re-arrest information is available for offenders whose original victims cannot be reached by interviewers because they have moved, changed their telephone numbers, etc. Hirschel and Hutchison (1992) found that re-arrests occurred in 13.5% of cases where victims could not be interviewed. Re-arrest information is also available for offenders whose new offenses have been committed against a different victim. Second, using re-arrest as a measure of recidivism avoids one of the problems associated with calls to the police: overestimating the rate of recidivism.

Unlike a call to the police about a “domestic disturbance,” a re-arrest indicates that the police found sufficient evidence that an offense had been committed. Unfounded calls are therefore excluded.

Re-arrest, however, does have a major weakness: it is likely to underestimate the occurrence of criminal offenses because it depends on two events occurring once an offense has been committed: the police must be called, and the police must make an arrest. As noted above, victims may be reluctant to call the police for a variety of reasons, including fear of retaliation.

Data from the 1998 National Crime Victim Survey indicate that 43% of victims of intimate

19 partner violence did not report their victimization to the police (Rennison and Welchans 2000).

Furthermore, even when the police are called and a criminal offense has occurred, they may not make an arrest. One study found that among victims who reported recidivism in an interview, only 30% of the offenders had been re-arrested (Hirschel and Hutchison 1992).

Because of the limitations associated with each of the measures of recidivism, many studies have chosen to collect recidivism data from two sources. Data from victim interviews are often combined with official records of re-arrests or official records of calls to the police. This approach can compensate for the weakness of one of the measures by using the other as a source of information (Fagan 1996). However, it also may create new problems. For example, in some studies the findings using one of the measures of recidivism are not confirmed when the other measure is used (Berk et al. 1992, Sherman 1992). As a result, it can be difficult to draw definitive conclusions.

In our study, we propose to use official records of re-arrests, as recorded in the CJA database, as our measure of recidivism. This decision was made on practical grounds. Re-arrest data are easily accessible, and CJA obtains comprehensive information about virtually all arrests in New York City. We have no access to the information needed to contact the victims, and it would be difficult to reach victims even if we obtained contact information. Official records of calls to the police for domestic incidents were also inaccessible to us. “Domestic Incident Reports” (DIR’s) in New York City have only recently been electronically stored in a centralized database. During most of the period of our study, only precinct-based paper or electronic records of DIR’s were maintained. These records of DIR’s were widely dispersed, since they were filed in the precinct in which the incident occurred. While we may in the future attempt to conduct studies of recidivism based on DIR’s or victim interviews, we believe an important first step is a study based on an analysis of re-arrest data. Re-arrest is commonly used as a measure of recidivism in domestic violence studies (Davis et al., 1998).

B. Selecting a Re-arrest Measure Having decided to use re-arrest as a measure of recidivism, we must address several questions about how to count re-arrests. Should we include only re-arrests that involve crimes against the same victim? Should we include only DV re-arrests, i.e., those that involve offenses against intimate partners or family members? Or should we use all types of re-arrests as evidence of recidivism?

Because we lacked information about the identity of the victim, we were unable to limit our study to re-arrests for crimes against the same victim. We do not view this as a serious limitation of our study. In fact, one of the advantages of measuring recidivism through re-arrest data, rather than through victim interviews, is that re-arrest data enable us to identify cases of recidivism against new victims. One study of domestic incidents in New York State found that limiting a measure of recidivism to new incidents with the same victim reduced estimates of recidivism by 15 to 20 percent (Frisch et al. 2001). This suggests that a significant number of domestic violence offenders move on to new victims. Our measure of re-arrest will not be limited to new offenses against the same victim.

20

Our study presents data on two re-arrest measures: re-arrests for any offense (whether DV or not), and re-arrests for DV offenses. The general re-arrest measure enables us to examine overall rates of re-offending. However, since our study is concerned primarily with understanding deterrence of domestic violence, not crime in general, we focus most of our attention on re-arrests for new DV offenses. To create our re-arrest measures, we collected data on all re-arrests and then determined whether each re-arrest was for a DV case or a Non-DV case.

This determination was made using the criteria described in Section II above. That is, we counted a re-arrest as a domestic violence case if either: 1) the case had a DV hearing type at arraignment, or 2) the case had any appearances in a specialized DV part. For re-arrests that were declined for prosecution, we used NYPD’s information about the nature of the defendant-victim relationship to identify DV cases. Our DV re-arrest measure enables us to address questions about the deterrent effects of case outcomes, criminal sanctions and prosecutorial screening policies on domestic violence offenders. Appendix A provides further information on the methods used to identify re-arrests.

C. Time Period for Collecting Data on Re-arrests We collected data on re-arrests during the 18-month period following the case outcome of each defendant’s first case during the third quarter of 1998 (hereafter referred to as the original case). The 18-month “at-risk” period for recidivism began on the date of the case outcome in the original case. For original cases that were declined for prosecution, the beginning of the at-risk period was the date of arrest.18 For original cases that ended in dismissal or an adjournment in contemplation of dismissal (ACD), the date of the dismissal or ACD was used as the beginning of the at-risk period. For original cases that resulted in conviction (by plea or trial), the date of disposition was used as the beginning of the at-risk period. The Third Quarter 1998 Dataset included data on processing of original cases through August 6, 1999; original cases disposed after that date were excluded from the sample. The at-risk period for recidivism therefore began no later than August 6th of 1999, and ended no later than February 6th of 2001.

Case processing data on each re-arrest occurring during the 18-month at-risk period was collected for at least 12 months. This 12-month tracking period enabled us to follow the case through arraignment and several court appearances. As discussed earlier, we determined whether the re-arrest was for a DV case or Non-DV case on the basis of hearing type at arraignment and/or whether the case had appearances in a specialized DV court part. All of the re-arrests that were prosecuted had an opportunity to be arraigned and/or to appear in a specialized DV part within 12 months. The twelve-month tracking period began on the date of the re-arrest. Since the latest re-arrest recorded for the study occurred no later than February 6th of 2001, the twelve-month tracking period ended no later than February 6th of 2002.

18

We would have preferred to use the date that the decision to decline prosecution was made; however this information was unavailable in our data. Since defendants whose cases are declined for prosecution are never arraigned in court, their at-risk period for re-arrest usually begins within one day of the date of arrest.

21 Our study excluded from our post-disposition re-arrest measure any re-arrests that occurred while the original case was pending. Although these re-arrests indicate that recidivism has occurred, it would not be appropriate to include them in our post-disposition re-arrest measure. Our study is designed to assess the impact of case outcomes and criminal sanctions on subsequent re-arrests. Although we excluded re-arrests that occurred during the pendency of the original case, we collected data on them. Pre-disposition re-arrests were used as predictors of the likelihood of post-disposition re-arrests. Specifically, we expect that defendants re-arrested during the pendency of the original case were more likely to be re-arrested after that case was disposed.

It is customary in re-arrest studies to exclude jail time from the at-risk period for recidivism. In this report, we have not excluded jail time from the at-risk period for re-arrest.

First, it is possible in New York State to be re-arrested for domestic violence offenses committed while in jail. Virtually all DV defendants in New York City are subject to an order of protection, the violation of which can lead to re-arrest. Convicted defendants sentenced to jail usually are subject to a full, “no contact” order, forbidding them from having any contact by telephone, mail or any other means with the victim in the case. Convicted defendants serving time in jail who call or write to the victim are subject to re-arrest. These defendants may also enlist the aid of friends and relatives who are not in jail to intimidate or attack the victim; this is also a violation of the order of protection and may subject the defendant to re-arrest. Excluding jail time from the at-risk period might therefore result in the exclusion of re-arrests for some defendants.

Second, very few defendants convicted in DV cases receive post-disposition jail time. The number of defendants whose time at risk would be affected by this exclusion is very small.

Finally, we conducted analyses of re-arrests after excluding post-disposition jail time from the at-risk period and found no change in our results. This confirms our expectation that the at-risk period is not significantly affected by jail time.

It is also customary in re-arrest studies to ignore information about how the re-arrest case is ultimately disposed (e.g., declined for prosecution, dismissed, convicted). We intend to follow the customary practice, but believe it is worthwhile to explain why. Critics of this practice might object that our criminal justice system is based on the principle “innocent until proven guilty.”

From this point of view, a measure of recidivism that includes arrests that did not lead to conviction is mistakenly counting “innocent” (or at least “not guilty”) defendants as recidivists.

While in principle this argument is valid, there are several reasons for using re-arrest, rather than re-conviction, as a measure of recidivism. First, re-arrest is likely to underestimate rather than overestimate recidivism. Many crimes (both DV and Non-DV) never come to the attention of the police. Even when crimes come to the attention of the police, they often do not result in arrest. For example, among DV incidents that met the statutory requirements for mandatory arrest, Frisch et al. (2001) found that arrest rates varied from 44% to 82% in various jurisdictions across New York State. These findings suggest that re-arrest rates underestimate recidivism.

Second, since the defendant is well known to the victim in DV cases, there is rarely a case of “mistaken identity.” When a crime has occurred and the defendant and victim know each other,

22 it is extremely unlikely that the person arrested has been mistakenly identified.19 Third, conviction on a re-arrest for a DV case is strongly influenced by whether the victim is cooperating with the prosecution (see e.g., Carlson and Nidey 1995). A measure of recidivism that relied on conviction for a re-arrest would therefore primarily measure whether the victim cooperated with the prosecution of the case, not whether the defendant had actually committed an offense. For these reasons, we believe it is appropriate to follow the standard practice of using re-arrest, not conviction on a re-arrest, as our measure of recidivism.

D. Advantages and Disadvantages of the Re-arrest Measure We conclude this section with a summary of the strengths and limitations of our re-arrest measure. On balance, we believe our measure is quite strong, and will help us to advance understanding of recidivism among DV offenders.

Our measure of re-arrest has several important advantages as a measure of recidivism.

First, our measure of DV re-arrests provides us with valuable information about the rates at which DV offenders were re-arrested for new DV offenses. This measure improves on the measure used in Davis et al. (1998), which included all re-arrests. It also improves on the measure used in Murphy et al. (1998), which included all re-arrests for battery or violations of orders of protection, whether or not these were DV offenses.

Second, our re-arrest measure includes re-arrests for DV offenses against new victims as well as against the same victim. Since prior research suggests that significant numbers of DV offenders recidivate against new victims, our study is capable of including these offenses.

Third, we identified all DV re-arrests that occurred within 18 months of the date of the final disposition in the original case. Many previous studies have been limited because they used time periods as short as 6 months. Our study uses a longer time period, which enables us to identify recidivism among defendants who may be able to avoid being re-arrested for 6 months to a year, but who eventually do re-offend.

Our measure of recidivism has two disadvantages. First, and most importantly, re-arrest is likely to underestimate recidivism because new DV offenses may not lead to re-arrest. As discussed at length earlier, many victims do not call the police when a new offense occurs. Even when an arrestable offense has been committed and the police are called, an arrest may not be made. This underestimation of recidivism rates is a common problem for studies that use re-arrest measures. We compensate for this problem in part by using a long at-risk period. Since it is likely that recidivists are likely to re-offend multiple times, using a longer at-risk period increases the chances that at least one of the re-offenses will lead to a re-arrest. Of course even with an 18-month at-risk period, we are likely to underestimate the number of defendants who re-offend.

19

Of course, a person can be wrongly accused of committing a domestic violence offense. Our argument here is that once a person commits a domestic violence offense, it is extremely unlikely that the victim will mistakenly identify another person as the offender.

23

A second limitation of our re-arrest measure is that it may underestimate re-arrests for DV offenses because we were only able to identify DV re-arrests that occurred in New York City.

This geographic limitation is common in most studies of re-arrest (e.g., Dunford 1992), and we do not believe it is a serious limitation. There is no reason to believe that DV offenders are highly mobile. Our measure of re-arrests within New York City has probably identified most of the DV re-arrests for defendants in our sample.

24

[This page intentionally left blank]

25 IV. EFFECTS OF CASE OUTCOMES AND CRIMINAL SANCTIONS ON RE-ARREST RATES FOR DV OFFENSES The first major question to be addressed in the current study is whether re-arrest rates for DV offenses vary by case outcomes and criminal sanctions. In this section of the report, we first describe how we measure case outcomes and criminal sanctions, and then examine how re-arrest rates vary by these outcomes and sanctions. We conclude by presenting a model that shows how case outcomes and criminal sanctions affect the likelihood of re-arrest after controlling for demographic characteristics, criminal history, community ties, and case processing characteristics.

A. Measuring Case Outcomes and Criminal Sanctions Many prosecuted DV cases result in dismissal in New York City (Peterson 2001). Lack of victim cooperation is probably the most important reason for dismissal. Although DA’s try to pursue “evidence-based” prosecutions, using “hearsay exceptions,” photographs, physical evidence, and medical reports, such evidence is frequently unavailable, or insufficient to obtain a conviction in a misdemeanor DV case.

Many DV cases are also disposed as “adjourned in contemplation of dismissal” (ACD).

ACD’s are offered primarily to first-time offenders. Although ACD’s are not convictions, they sometimes have conditions attached (e.g., that the defendant successfully complete a program, such as a batterer intervention program). When these conditions are not fulfilled, or if the defendant is re-arrested within 6 months (12 months in the case of a family offense), the case can be re-opened and restored to the calendar for another, possibly more severe, disposition. The use of ACD’s is one of a variety of strategies employed to achieve the goal of monitoring defendants in DV cases. ACD’s in DV cases appear to be used as a disposition similar to, but more lenient than, a conviction with a sentence of conditional discharge.

In this report, convictions are defined to include pleas of guilty and findings of guilty after trial, including pleas or findings of guilty to a violation. (Although violations are not considered crimes under New York State Penal Law, they can result in a jail sentence.) DV cases that end with a conviction can result in a variety of types of sentences: conditional discharge, fine, probation, jail or some combination of these. In practice, most convictions in DV cases in New York City result in a sentence of a conditional discharge, with attendance at a batterer’s intervention program as a condition of the sentence. However, an important subgroup of convicted defendants receives a jail sentence; in the third quarter of 1998, their sentences averaged over 90 days (Peterson 2001). In our analyses of case outcomes and criminal sanctions, we classified convicted defendants in DV cases in two categories: those who received a jail sentence, and those who did not. Most of those who did not receive a jail sentence were sentenced to a conditional discharge.

The hypothesis to be tested in this section of the report is that defendants in DV cases with more severe case outcomes and criminal sanctions had lower rates of re-arrest. Dismissal is

26 considered to be the least severe case outcome, followed, in order of increasing severity, by ACD, conviction without a jail sentence, and conviction with a jail sentence.

B. Re-arrest Rates by Case Outcomes and Criminal Sanctions Table 4-1 presents the re-arrest rates for each of the categories of case outcome. The re-arrest rate for DV offenses indicates the percentage of defendants who were re-arrested at least once for a DV offense within 18 months of case disposition. The re-arrest rate for any offense indicates the percentage of defendants who were re-arrested for any new offense, whether it was a DV offense or not. As discussed in Section II-E.1 above, each case identified in this table represents the first DV arrest during the third quarter of 1998, if any, for each defendant arrested in the quarter.

As shown in Table 4-1, about 17% of DV offenders were re-arrested for a DV offense within 18 months of the disposition of the third quarter 1998 case (the original case). However, the re-arrest rate varied depending on the outcome of the original case. About 16% of DV offenders whose original case was dismissed were re-arrested for a DV offense within 18 months of the dismissal. The re-arrest rate for DV offenders who received an ACD was slightly lower, about 14%. However, re-arrest rates were higher for DV offenders who were convicted. About 19% of convicted DV offenders who were not sentenced to jail were re-arrested, and 26% of those who were convicted and sentenced to jail were re-arrested. These differences in re-arrest rates by case outcomes and criminal sanctions were statistically significant.20 Table 4-1 also shows that about 38% of DV offenders in the third quarter of 1998 were re-arrested for any offense (whether DV or not) within 18 months of the original case. This is more than double the re-arrest rate for DV offenses only. About 36% of defendants whose original case was dismissed were re-arrested, compared with 29% whose original case ended in an ACD. The re-arrest rate for any offense was 40% for defendants who were convicted but did not receive a jail sentence, and 65% for defendants who were convicted and sentenced to jail.

This pattern is similar to the pattern for re-arrests for DV offenses.

These results are surprising because, with one exception, they appear to contradict our hypothesis. We had expected to find that more severe dispositions would reduce re-arrest rates.

The results in Table 4-1 show that the defendants who received the most severe dispositions had the highest re-arrest rates. Defendants who were convicted and sentenced to jail had higher re-arrest rates than defendants who were convicted and not sentenced to jail. In turn, defendants who were convicted and not sentenced to jail had a higher re-arrest rate than those who received an ACD. While these results contradict our hypothesis, it is important to remember that we have 20

Statistical significance tests assess the probability that the percentage differences by case outcome that were observed in the sample could have occurred by chance alone. The tests take into account the size of the sample and the magnitude of the differences observed. Larger percentage differences and percentage differences based on larger samples are more likely to be statistically significant. In this report, following standard convention, significance levels of .05 or less were treated as statistically significant.

27 TABLE 4-1: RE-ARREST RATES BY CASE OUTCOMES AND CRIMINAL SANCTIONS IN DV CASES1 Crimes Against Persons and Property Subsample Third Quarter 1998 Dataset CASE OUTCOMES AND CRIMINAL SANCTIONS

Re-Arrest Rate For DV Offenses2

Re-Arrest Rate For Any Offense3

Dismissed (N of cases)

16% (3,022)

36% (3,022)

Adjourned in Contemplation of Dismissal (N of cases)

14% (1,235)

29% (1,235)

Convicted, no Jail Sentence (N of cases)

19% (1,849)

40% (1,849)

26% (383)

65% (383)

17% (6,489)

38% (6,489)

Convicted, with Jail Sentence (N of cases) Total, All DV Cases (N of cases)

1

Each case represents the first DV arrest during the third quarter of 1998, if any, for each defendant arrested in the quarter. Defendants whose first DV arrest during the third quarter of 1998 resulted in a case disposed in Supreme Court (N = 158) are excluded from the numbers reported here. Also excluded are defendants whose first DV case during the third quarter of 1998 did not reach a final disposition in Criminal Court by August 6, 1999 (N = 304).

2 Chi-square = 36.93, indicating that differences by case outcome and criminal sanction are statistically significant at p < .001 3 Chi-square = 168.34, indicating that differences by case outcome are statistically significant at p < .001

not taken into account differences among the defendants. In particular, the results presented in Table 4-1 do not take into account differences in the prior records of defendants who received various dispositions. ACDâ&#x20AC;&#x2122;s are generally given to first-time offenders, while jail sentences are more likely to be imposed on defendants who have multiple prior convictions. If more severe case outcomes and criminal sanctions are given to chronic offenders, variation in re-arrest rates by case outcomes and criminal sanctions may reflect differences in the type of offender. This explanation is addressed below, in our discussion of models that control for the influence of prior record on re-arrest rates.

28

There is one finding in Table 4-1 that supports our hypothesis. Defendants who received an ACD had a lower re-arrest rate than those whose cases were dismissed (29% vs. 36% for any re-arrest, 14% vs. 16% for a DV re-arrest21). Defendants who received an ACD were almost always subject to an order of protection; some were also required to complete a batterer intervention program. An ACD therefore imposed some constraints on defendants, subjecting them to potential monitoring by the court. These constraints, and the potential for court monitoring, were absent when the defendant’s case was dismissed. The findings may suggest that the constraints imposed by an ACD slightly reduced the probability of re-arrest. However, differences in criminal history may provide an alternative explanation for the lower re-arrest rate of defendants who received an ACD vs. a dismissal. Defendants who received an ACD were usually first-time offenders, whereas more of those whose cases were dismissed were chronic offenders (data not shown). Whether criminal history accounts for the difference in re-arrest rates will be addressed in Section IV-C below.

The current study found that 17% of defendants who were arrested for a DV offense in the third quarter of 1998 were re-arrested for a DV offense within 18 months following the disposition of their case. Three studies that used re-arrest as a measure of recidivism reported rearrest rates similar to the rate reported here. 22 These studies reported re-arrest rates of 15% (Wooldredge and Thistlethwaite 2002; based on a two-year follow-up period), 16% (Murphy et al. 1998; based on a 12 to 18 month follow-up period) and 24% (Gross et al. 2000; based on an 18 to 24 month follow-up period). Taking into account differences in the length of the follow-up period, these rates are similar to the 17% rate we found. The similarity of our re-arrest rate to those found in previous studies strengthens our confidence in the accuracy of our measure of the re-arrest rate for DV offenses.

As discussed earlier, re-arrest rates generally underestimate recidivism (see discussion in Section III of this report). Studies based on victim interviews generally found higher recidivism rates, even for shorter periods, than the 17% rate reported here. Three studies based on victim interviews report recidivism rates of one-quarter to one-third over a 6-month follow-up period (see Sullivan et al. 2000 (26%), Fagan et al. 1984 (29%), and Ford and Regoli 1993 (34%)).

Tolman and Weisz (1995), using domestic incident reports to the police as a measure of recidivism, found that 30% of DV offenders had new incidents of abuse during an 18-month follow-up period. These comparisons confirm that the re-arrest rate provides a lower estimate of recidivism than estimates based on victim interviews or police reports.

The Third Quarter 1998 Dataset includes a comparison sample of Non-DV cases.

Although the primary focus of the current study is on re-arrests in DV cases, we present a limited set of results for Non-DV cases. These results for Non-DV cases provide baseline information on re-arrest rates for comparison to the DV cases. They also enable us to assess the likelihood 21

The difference in re-arrest rates between the “Dismissed” and “ACD” categories was statistically significant for any re-arrest (p < .05), but was not statistically significant for DV re-arrests.

22 One study reviewed in Section I of this report (Davis et al. 1998) did not report the re-arrest rate for DV offenders, and is not discussed here.

29 that someone arrested for a Non-DV offense was re-arrested for a DV offense. This allows us to assess the degree of overlap in the types of offenses committed by a given defendant. Table 4-2 presents the re-arrest rates by case outcomes and criminal sanctions for Non-DV cases. Each case outcome and criminal sanction identified in this table represents the case outcome and criminal sanction for the first Non-DV arrest during the third quarter of 1998 for each defendant who had no DV arrests during the quarter.

The findings in Table 4-2 show that the re-arrest rate for DV offenses among those whose original arrest was for a Non-DV offense was about 5%. This rate was considerably lower than the re-arrest rate of 17% reported in Table 4-1 for those whose original arrest was for a DV offense. These findings suggest that DV offenders were more likely to be re-arrested for new DV offenses than Non-DV offenders. Table 4-2 also shows that the re-arrest rate for DV offenses did not vary by the case outcome or criminal sanction of the original arrest for a Non-DV offense.

The re-arrest rates varied from 4% to 5%, and the differences by case outcomes and criminal sanctions were not statistically significant. This pattern of findings suggests that case outcomes and criminal sanctions in Non-DV cases had no impact on the likelihood that the defendant would be re-arrested for a DV offense within 18 months.

Table 4-2 also shows that among those whose original arrest was for a Non-DV offense the re-arrest rate for any offense was 44%. This rate was higher than the overall re-arrest rate of 38% for DV offenders reported in Table 4-1. The pattern of re-arrests for any offense by case outcome and criminal sanction was similar to the patterns reported in Table 4-1. Re-arrests were highest among those convicted and sentenced to jail (67%), somewhat lower among those convicted and not sentenced to jail (44%), lowest for those who received an ACD (32%) and slightly higher for defendants whose cases were dismissed (36%). It is interesting that among Non-DV offenders the pattern of re-arrests for DV offenses was not affected by case outcome, but the pattern of re-arrests for any offense was affected by case outcome. This suggests that outcomes in Non-DV cases are useful in predicting future rates of committing Non-DV offenses, but that outcomes in Non-DV cases are not related to future rates of committing DV offenses.

The remainder of the analyses discussed in this report focus on re-arrests among offenders whose original arrest in the third quarter of 1998 was for a DV offense. This reflects the primary focus of our study, which examines recidivism among DV offenders. Furthermore, our findings for those arrested for Non-DV offenses show that case outcomes and criminal sanctions in Non-DV cases did not influence re-arrest rates for DV offenses (see discussion of Table 4-2). For these reasons, we did not examine in further detail the re-arrest rates in Non-DV cases.

30 TABLE 4-2: RE-ARREST RATES BY CASE OUTCOMES AND CRIMINAL SANCTIONS IN NON-DV CASES 1 Crimes Against Persons and Property Subsample Third Quarter 1998 Dataset CASE OUTCOMES AND CRIMINAL SANCTIONS

Re-Arrest Rate For DV Offenses2

Re-Arrest Rate For Any Offense3

Dismissed (N of cases)

4% (3,737)

36% (3,737)

Adjourned in Contemplation of Dismissal (N of cases)

4% (3,682)

32% (3,682)

Convicted, no Jail Sentence (N of cases)

5% (5,080)

44% (5,080)

Convicted, with Jail Sentence (N of cases)

5% (3,497)

67% (3,497)

Total, All Non-DV Cases (N of cases)

5% (15,996)

44% (15,996)

1

Each case represents the first Non-DV arrest during the third quarter of 1998 for each defendant in the dataset who had no DV arrests during the quarter. Defendants whose first Non-DV arrest during the third quarter 1998 resulted in a case disposed in Supreme Court (N = 2,129) are excluded from the number reported here. Also excluded are defendants whose first Non-DV case during the third quarter of 1998 did not reach a final disposition in Criminal Court

by August 6, 1999 (N = 1,295).

2 Chi-square = 3.07, indicating that differences by case outcome and criminal sanction are not statistically significant at p < .05 3 Chi-square = 1,064.12, indicating that differences by case outcome and criminal sanction are statistically significant at p < .001

C. Logistic Regression Model Examining the Effect of Case Outcomes and Criminal Sanctions on the Re-arrest Rate for DV Offenses Our logistic regression model assesses whether the type of case outcome and criminal sanction in a DV case affected the likelihood that a defendant was re-arrested for a DV offense in the subsequent 18 months. As described earlier, our dependent variable is whether or not the defendant was ever re-arrested for a DV offense. The hypothesis to be tested is whether cases

31 with more severe case outcomes and criminal sanctions have lower re-arrest rates. Dismissal is the least severe case outcome, followed, in order of increasing severity, by ACD, conviction without a jail sentence, and conviction with a jail sentence. To test the hypothesis, we used logistic regression models where the categories of case outcome and criminal sanction were represented by dummy variables. The logistic regression technique is described in further detail in Appendix B. The statistical models include control variables to address the problem discussed earlier, that case outcomes and criminal sanctions may be influenced by characteristics of the defendants.

As Davis et al. 1998 (p. 439) point out, cases are not randomly assigned to the various case outcomes and criminal sanctions. A variety of factors may affect case outcomes and criminal sanctions, such as strength of evidence, defendantâ&#x20AC;&#x2122;s demographic characteristics, the nature of the charges, criminal history, etc. If these factors also affected the likelihood that the defendant was re-arrested, then the relationship between case outcomes and criminal sanctions and re-arrest would appear to be stronger than it really was. To address this problem, our models control for defendant and case characteristics measured at the time of the original arrest. These variables include: type of charges, borough, demographic characteristics (sex, ethnicity, and age), type of defendant-victim relationship, defendantâ&#x20AC;&#x2122;s criminal history, community ties and case processing characteristics.23 These control variables are described in Appendix C. The inclusion of these control variables in the model gives us more confidence that the relationship between case outcomes and criminal sanctions and re-arrest rates reflects an independent effect of these outcomes and sanctions. Furthermore, the inclusion of the community ties measures enables us to re-examine whether community ties influence the likelihood of re-arrest for domestic violence offenses. As reviewed in Section I of this report, only one other study that examined the impact of case outcomes and criminal sanctions on re-arrests also examined the impact of community ties. That study found that residential stability reduced re-arrest rates. Our study examines additional measures of community ties to more fully assess their impact on re-arrest rates.

The results of the logistic regression model are presented in Table 4-3. These results show that odds ratios for the case outcome and criminal sanction variables are all relatively close to 1, and their effects are not statistically significant. This suggests that case outcomes and criminal sanctions had no impact on the likelihood of re-arrest after controlling for defendant and case characteristics. Earlier, we reported in Table 4-1 that re-arrest rates varied by case outcomes and criminal sanctions: defendants who were convicted, especially those who were sentenced to jail, had higher re-arrest rates than defendants who were not convicted. We speculated that convicted defendants, particularly those sentenced to jail, were likely to be chronic offenders. The results presented in Table 4-3 show that this speculation was correct.

Criminal history does affect the likelihood of re-arrest. Defendants who had a prior adult arrest record and who had a larger number of prior misdemeanor convictions were significantly more likely to be re-arrested than those who had never been arrested or who had fewer misdemeanor

23

Measures of strength of evidence were not available in this dataset.

32

TABLE 4-3 LOGISTIC REGRESSION MODEL PREDICTING LIKELIHOOD OF REARREST FOR A DV OFFENSE: DV OFFENDERS CITYWIDE CRIMES AGAINST PERSONS AND PROPERTY SUBSAMPLE 1 Third Quarter 1998 Dataset 2

INDEPENDENT VARIABLES CASE OUTCOME Reference Category: Dismissed Adjourned in Contemplation of Dismissal Convicted, no Jail Sentence Convicted, with Jail Sentence

Standardized b

Odds

0.00 0.07 -0.01

1.01 1.12 0.97

0.14 ** 0.01 0.00 -0.03

1.37 1.04 0.98 0.93

0.10 -0.13 * 0.04 0.01

1.17 0.75 1.07 1.02

-0.40 ***

0.44

0.04 -0.12 * -0.10

1.10 0.82 0.73

-0.18 * -0.15 -0.28 ***

0.75 0.79 0.61

0.06 0.10 -0.07 0.01

1.12 1.19 0.87 1.02

Ratio

CONTROL VARIABLES ARRAIGNMENT CHARGE PENAL LAW ARTICLE:

Reference Category: Assault (PL 120) Criminal Contempt (PL 215) Harassment (PL 240) Crimes Against Children (PL 260) Other BOROUGH Reference Category: Bronx Brooklyn Manhattan Queens Staten Island DEFENDANT'S DEMOGRAPHIC CHARACTERISTICS SEX (Female) ETHNICITY:

Reference Category: Black White Hispanic Other AGE:

Reference Category: Age 16-20 Age 21-29 Age 30-39 Age 40 and over DEFENDANT-VICTIM RELATIONSHIP Reference Category: Married Boyfriend-Girlfriend Common-Law Marriage Other Relationship Missing

Table Continues on Next Page

33

TABLE 4-3 (continued)

2

INDEPENDENT VARIABLES

Standardized b

Odds Ratio

DEFENDANT'S CRIMINAL HISTORY ANY PRIOR ARRESTS NUMBER OF PRIOR MISDEMEANOR CONVICTIONS NUMBER OF PRIOR FELONY CONVICTIONS ANY ARRESTS FOR A DV OFFENSE PRIOR TO CASE DISPOSITION ANY ARRESTS FOR A NON-DV OFFENSE PRIOR TO CASE DISPOSITION

0.33 *** 0.09 * 0.07

1.64 1.02 1.07

0.31 ***

2.43

0.12 **

1.39

DEFENDANT'S COMMUNITY TIES UNEMPLOYED AT CURRENT ADDRESS 1 YEAR OR LESS LIVES WITH SOMEONE DOES NOT EXPECT ANYONE AT ARRAIGNMENT HAS NO TELEPHONE LIVES OUTSIDE NYC AREA

0.18 0.09 0.13 0.11 0.10 -0.02

*** * * *

1.31 1.15 1.22 1.18 1.18 0.92

ARREST AND ARRAIGNMENT CHARGE CHARACTERISTICS NUMBER OF ARREST CHARGES ARRAIGNMENT CHARGE IS A FELONY CHANGE IN CHARGE SEVERITY FROM ARREST TO ARRAIGNMENT:

Reference Category: No Change Charge Severity Reduced from Arrest to Arraignment Charge Severity Increased from Arrest to Arraignment

0.08 0.05

1.07 1.09

-0.04 0.09 *

0.94 1.38

-0.10 -0.20 ***

0.81 0.99

CASE PROCESSING CHARACTERISTICS DEFENDANT EVER RELEASED NUMBER OF WEEKS FROM ARRAIGNMENT TO DISPOSITION

Nagelkerke R2 (N of cases)

.11 *** (6,489)

NOTES 1 2

See text for a description of the dataset and the subsample.

See Appendix C for information about the measurement and coding of the variables.

* Statistically significant at p < .05 ** Statistically significant at p < .01 *** Statistically significant at p < .001

34 convictions. Furthermore, defendants who were re-arrested, whether for a DV or Non-DV offense, while the original case was pending were also more likely to be re-arrested after their case was disposed. These findings confirm our speculation that jail sentences did not increase re-arrest rates. Instead, jail sentences were imposed more often on chronic offenders, who were more likely to be re-arrested regardless of the outcome of their original case.

Similarly, the results in Table 4-3 show no difference in re-arrest rates between those who received an ACD and those whose cases were dismissed. More of those whose cases were dismissed were chronic offenders, whereas most of those who received an ACD were first-time offenders. Once criminal history was taken into account, there was no difference in re-arrest rates between those who received an ACD and those whose cases were dismissed.

It is interesting to compare the results for the model in Table 4-3 to the results for a model that includes case outcomes and criminal sanctions as the only predictors of likelihood of re-arrest (model not shown). These results contrast the net effect of case outcomes and criminal sanctions (i.e., after controlling for all the other variables in the model), and the total effect of case outcomes and criminal sanctions (i.e., using case outcomes and criminal sanctions as the only predictors in the model). When case outcomes and criminal sanctions were the only predictors in the model, the odds ratios were .85 for defendants who received an ACD, 1.21 for defendants convicted without a jail sentence, and 1.84 for defendants convicted with a jail sentence. The latter two odds ratios were statistically significant, while the first was not. These results indicate that the odds of re-arrest were 1.84 times higher for defendants who were convicted and sentenced to jail than for defendants whose cases were dismissed. They also indicate that the odds of re-arrest were 1.21 times higher for defendants who were convicted and not sentenced to jail than for defendants whose cases were dismissed. These total effects can be contrasted with the net effects reported in Table 4-3, where the odds ratios were 1.01 for an ACD, 1.12 for defendants convicted without a jail sentence, and .97 for defendants convicted with a jail sentence. None of these odds ratios were statistically significant, indicating that controlling for defendant and case characteristics accounted for the initial differences in re-arrest rates by case outcomes and criminal sanctions.

Which variables in Table 4-3 were primarily responsible for accounting for differences in re-arrest rates by case outcomes and criminal sanctions? An examination of results (not shown) indicates that criminal history and whether the defendant was ever released while the original case was pending are the variables primarily responsible for explaining away the differences in re-arrest rates by case outcomes and criminal sanctions. As discussed above, chronic offenders were more likely to be convicted, and particularly to be convicted and given a jail sentence, than other offenders. DV offenders who were released pending the outcome of their original case were more likely to have their cases dismissed. Since chronic offenders were more likely to be re-arrested for DV offenses, while those who were released pending the case outcome were less likely to be re-arrested for DV offenses, controlling for these variables reduced the differences in re-arrest rates by case outcomes and criminal sanctions.

Our test of the hypothesis that the severity of case outcomes and criminal sanctions reduces the likelihood of re-arrest has two important limitations. First, we could not include all

35 possible control variables in our model. To the extent that variables not included in the model might have affected both the case outcome/criminal sanction and the likelihood of re-arrest, our results may have understated the effect of case outcomes and criminal sanctions on re-arrest.

This risk is inherent in all studies of the effects of case outcomes and criminal sanctions. Even an experimental design cannot randomly assign case outcomes and criminal sanctions to avoid this problem (see Ford and Regoli 1992, 1993 for a discussion of an experimental design).24 The second limitation on our ability to test the hypothesis is that the case outcomes and criminal sanctions may affect the willingness of victims to call the police when new incidents occur. Earlier, when we discussed the limitations of using re-arrest as a measure of recidivism, we noted that re-arrest depends on two steps being taken: a call to the police, and an arrest by the police. Because not all new incidents of domestic violence resulted in a call to the police, or an arrest by the police, our re-arrest measure is likely to be an underestimate of the actual rate of recidivism. As long as this underestimation was roughly the same across the various categories of case outcome, it should not have affected our conclusions. However, some previous studies have suggested that the willingness of victims to call the police may be related to the case outcome or criminal sanction of the original case. Specifically, victims who were unhappy with the case outcome or criminal sanction in the original case may have been reluctant to call the police when subsequent offenses occurred. This could pose a problem for our analysis. For example, suppose that victims who cooperated with the prosecution to get a conviction were the most satisfied with the case outcome, whereas victims who did not cooperate with the prosecution, leading to a dismissal of the case, were the least satisfied with the case outcome. If these assumptions are correct, victims in cases where the defendantâ&#x20AC;&#x2122;s case was dismissed were the least satisfied with the outcome, while those in cases where the defendant was convicted and sentenced to jail, or to a conditional discharge with a batterer intervention program, were the most satisfied. If victims who were more satisfied with the outcome were more likely to call the police and to have the defendant re-arrested when a new DV offense occurred, then victims who were satisfied that the defendant was convicted were more likely to call the police again and to have the defendant re-arrested. As a result, re-arrest rates would have been higher for convicted defendants and lower for defendants whose cases were dismissed. This pattern of differences in the propensity to call the police could have produced the pattern of re-arrest rates reported in Table 4-1.

Two features of our study enable us to compensate for this problem. First, the long 18-month at-risk period for re-arrest provided an opportunity for the offender to re-offend multiple times. While victims who were dissatisfied with the case outcome or criminal sanction in the original case may have been reluctant to call the police again, the chances that they called 24

Ford and Regoli (1992, 1993) randomly assigned DV cases to one of several case outcomes (declined prosecution, prosecution with pre-trial diversion to a batterer intervention program, prosecution with conviction and a batterer intervention program, or prosecution with conviction and sentence (fine, probation and/or jail). These random assignments were recommendations, not mandatory outcomes.

Prosecutors, victims, defendants and judges could take actions that disrupted the randomly assigned case outcome. Approximately 25% of defendants did not receive the case outcome assigned (Davis et al.

1998).

36 the police at least once during the 18-month period probably increased with each new offense after the disposition of the original case (Hutchison 1998). Second, in cases where the victim and the offender ended their relationship after the original incident, the 18-month period allowed sufficient time for the offender to start a new relationship. If the offender re-offended in a new relationship, the prior case outcome or criminal sanction was unlikely to have influenced the new victim's willingness to call the police.

In addition to our findings on the impact of case outcomes and criminal sanctions, there are several other noteworthy findings in our model. First, many of the community ties variables had a statistically significant impact on the likelihood of re-arrest. DV offenders who were unemployed at the time of their arrest in the third quarter of 1998, defendants who did not expect someone to come to their arraignment, and defendants who did not report having a telephone were all more likely to be re-arrested for a DV offense. DV offenders who had lived at their current address for one year or less in the third quarter of 1998 were also more likely to be re-arrested for a DV offense, although this effect did not quite reach statistical significance (p = .055). Taken together, these findings suggest that defendants who lacked strong community ties were more likely to be re-arrested for DV offenses.

One measure of community ties had no impact on the likelihood of re-arrest. Defendants who did not live in the New York City area in the third quarter of 1998 were no more likely to be re-arrested for a DV offense than those who did. This finding may indicate that the variable is not a relevant factor in predicting re-arrests. It may also suggest that DV offenders who lived outside the New York City area were more likely to be re-arrested in jurisdictions outside New York City. Information about re-arrests outside New York City was not available in our study.

Another measure of community ties had a statistically significant impact on re-arrest rates in a direction opposite that proposed by the community ties hypothesis. Defendants who lived with someone in the third quarter of 1998 were more likely to be re-arrested for a DV offense.

Although our measure does not indicate with whom the defendant lived, it may be that many of the DV offenders were living with the victim. If so, and if these defendants returned to live with the victim after their case was disposed, then this variable actually measured whether the defendant was likely to have further close contact with the victim.25 Continuing to live with the victim provides greater opportunities for re-offending and for re-arrest (Wooldredge and Thistlethwaite 2002). If this speculation is correct, then living with someone is not a useful measure of community ties for DV offenders. Instead, it is more likely that living with the victim increases the risk of re-offending. This explanation is consistent with the findings reported here.

A second noteworthy finding from our model is that chronic offenders were more likely to be re-arrested for a DV offense. Among the strongest predictors of re-arrest (as measured by 25

Because “no contact” orders of protection are issued in certain DV cases, there are some defendants who were not likely to be living with the victim after their case was disposed. “No contact” orders would have been most common among those convicted and sentenced to jail, and also very common among those convicted and not sentenced to jail. Defendants whose cases were dismissed were unlikely to be subject to “no contact” orders, unless they were issued in a subsequent case.

37 standardized beta) were whether the defendant was re-arrested for a DV offense while the original case was pending and whether the defendant had any prior adult arrests (both of which increased the likelihood of a post-disposition re-arrest). Defendants who were re-arrested for a Non-DV offense while the original case was pending, and defendants who had a larger number of misdemeanor convictions were also more likely to be re-arrested for a DV offense. These findings are consistent with those of many prior studies.

Third, our study found no statistically significant effect of the type of relationship between the defendant and the victim on the likelihood of re-arrest. Defendants who were in boyfriend-girlfriend, common-law or other family relationships were no more or less likely to be re-arrested for a DV offense than married defendants. This finding must be viewed as tentative, since the defendant-victim relationship was listed as â&#x20AC;&#x153;missingâ&#x20AC;? in 24% of the cases in this sample. More complete data would provide a better test of the impact of defendant-victim relationship. Furthermore, we have no data on whether the relationship was a current or former relationship at the time of the arrest in the third quarter of 1998.26 Fourth, we found that the likelihood of re-arrest was significantly lower for defendants whose cases were disposed in Manhattan than in other boroughs. While this might indicate a real difference among the boroughs, we believe it is more likely that the lower re-arrest rate for DV offenses in Manhattan is due to weakness in the measure of DV re-arrests in this borough. As discussed in Section II-E.2 above, we believe that many DV re-arrests in Manhattan were incorrectly classified as Non-DV re-arrests. We included the borough variables in the model to statistically control for this problem. As a result, the estimates of the effects of other variables in the model are not confounded with the measurement problem associated with identifying DV re-arrests in Manhattan.

Finally, the model explained only 11% of the variation in likelihood of re-arrest (see Nagelkerke R2 in Table 4-3). Stated another way, 89% of the variation in likelihood of re-arrest cannot be accounted for by the variables included in our model. This indicates that the model is relatively ineffective in identifying the factors that influence the likelihood of re-arrest for a DV offense. While we have discussed the influence of many variables that had a statistically significant impact on re-arrest, the combined effects of these variables were relatively small.

Interestingly, none of the 8 studies of case outcomes and criminal sanctions discussed in Section I reported a Nagelkerke R2 (or similar measure) for their recidivism models.

Wooldredge and Thistlethwaite (2002) report that their best models explained 14% of the variation within census tracts and 19% of the variation between census tracts. Our results are not directly comparable, but appear to be somewhat weaker. None of the other studies identified more than a few variables that had a statistically significant effect on recidivism, so it seems likely that their models were relatively weak. The predictive power of the model presented here

26

Previous research comparing those who were current vs. former intimate partners at the time of arrest found no impact of defendant-victim relationship on the likelihood of re-arrest (Davis et al. 1998).

However, this study was limited by its small sample size, as discussed in Section I above.

38 is probably at least as strong as, or stronger than, the predictive power of models in those seven studies.

Why is it so difficult to predict recidivism among DV offenders? First, re-arrest for a DV offense is relatively rare, occurring for only 1 out of 6 DV offenders. It may be more difficult to develop a strong predictive model for a rare event. Second, the measures of recidivism may be unreliable, and subject to significant measurement error. It is also possible that the predictors of re-arrest included in the models are poorly measured. Measurement error reduces the statistical power of predictive models. More refined measures of criminal history, community ties or the defendant-victim relationship might improve the predictive power of the model. Third, factors not currently included in the model may play a larger role in predicting recidivism. For example, psychological factors (e.g., anger, anti-social personality), being physically, sexually or emotionally abused as a child, or other factors may influence the likelihood of recidivism.

Finally, random events or unique situational factors may trigger new DV offenses.

D. Summary and Discussion of Findings Our analyses of the effect of case outcomes and criminal sanctions on re-arrest rates for DV offenses have produced several important findings.

First, our hypothesis that DV offenders whose case outcomes and criminal sanctions were more severe would be less likely to be re-arrested for DV offenses was not confirmed. (Recall that case outcomes and criminal sanctions were ranked from least to most severe in the following order: dismissal, adjournment in contemplation of dismissal, conviction without a jail sentence, and conviction with a jail sentence). When we initially examined re-arrest rates by case outcome, the results appeared to be in the direction opposite that predicted in our hypothesis.

Defendants who were convicted, particularly those who were convicted and sentenced to jail, were more likely to be re-arrested than those whose cases were dismissed or ACDâ&#x20AC;&#x2122;d. A closer examination revealed that there were no differences in re-arrest rates by case outcome or criminal sanction after controlling for defendant and case characteristics. The initially higher re-arrest rates for convicted defendants reflected their prior criminal histories. Once the effect of criminal history on re-arrest rates was taken into account, the re-arrest rate for defendants who were convicted was similar to that of defendants who were not convicted. Although it was re-assuring to find that more severe case outcomes and criminal sanctions did not increase re-arrest rates, these findings nevertheless fail to support our hypothesis.

Second, our analyses showed that weak community ties increase the likelihood that a DV offender will be re-arrested for a DV offense. Three of our six measures had a statistically significant impact on re-arrest rates, and the fourth was nearly statistically significant.

Defendants who were unemployed, who did not expect someone to be at their arraignment, who did not report having a telephone, and who lived at their current address one year or less were all more likely to be re-arrested for a DV offense. A previous study that examined the impact of community ties on re-arrest rates for DV offenses found that length of time at current address, but not employment status, affected re-arrest rates (Wooldredge and Thistlethwaite 2002).

However, that study had a smaller sample size. Our sample size was twice as large, and our

39 community ties measures have been validated in several studies (Siddiqi 1999). Indeed, a composite measure of our community ties variables is used to make release recommendations to the court at arraignment.

Third, we found that measures of criminal history were among the strongest predictors of the likelihood of re-arrest for a DV offense. Defendants who were arrested for a DV offense while their original case was pending were more likely to be re-arrested for a DV offense after the disposition of that case. Defendants who had prior adult arrests for any type of offense were more likely than those with no prior adult arrests to be re-arrested for a DV offense. Statistically significant increases in the likelihood of re-arrest were also found for defendants who were arrested for a Non-DV offense while their original case was pending and for those with prior misdemeanor convictions. These findings are consistent with the findings from previous studies reviewed above.

Finally, our statistical model was relatively unsuccessful at predicting the likelihood that a DV offender would be re-arrested for a DV offense. We accounted for 11% of the variation in likelihood of re-arrest, leaving 89% unexplained by any of the variables included in the model.

This finding appears to be typical of studies of recidivism among DV offenders. Considerable work remains to be done to measure and identify the factors that predict re-arrest for DV offenses.

40

[This page intentionally left blank]

41 V. THE EFFECT OF CASE SCREENING POLICIES ON RE-ARREST RATES The second major question to be addressed in this study is whether screening of DV cases affected re-arrest rates for offenders arrested on DV charges. We first describe case screening policies used in three boroughs: the Bronx, Brooklyn and Manhattan. We then present data showing the impact of case screening on re-arrest rates in the Bronx, which is the only borough that screens out (declines to prosecute) any significant number of DV cases. Finally, we present a model that shows how case screening affects the likelihood of re-arrest after controlling for demographic characteristics, criminal history and community ties.

A. Measuring the Impact of Case Screening Policies Not every arrest on domestic violence charges results in a criminal prosecution. Some arrests are referred to Family Court, others are “voided” by the police, while in other cases the DA’s office declines to prosecute (an outcome referred to as a “DP”). Referrals to Family Court usually occur where the DA believes that a case can be better handled as a dispute rather than a crime.27 Voided arrests and DP’s usually occur because the evidence in the case cannot sustain bringing any charges against the defendant. Referrals to Family Court and voided arrests were rare among DV cases in the third quarter of 1998.28 DP’s were also rare, except in the Bronx, where about 19% of DV cases resulted in a decision by the DA to decline to prosecute.

While DP’s may seem similar to dismissals in their consequences for the defendant, they differ in two important ways from dismissals. First, in cases that are declined for prosecution, there is no opportunity for the criminal court to issue a temporary order of protection at arraignment. In a case that is prosecuted and later dismissed, a temporary order of protection is issued at arraignment and remains in effect until the case is dismissed. This order forbids or restricts the defendant’s contact with the victim. Second, cases that result in dismissal keep the defendant under the jurisdiction of the court for a period of time (up to 30 to 90 days, depending on the severity of the charge) pending disposition. During this time period, the defendant is under scrutiny by the court for violations of the order of protection and for failure to appear in court. These differences suggest that in dismissed cases, there is more judicial monitoring of the defendant than there is in cases that are declined for prosecution. If judicial monitoring deters recidivism, we would expect to find higher rates of re-arrest in cases declined for prosecution than in dismissed cases.

The high rate of declining to prosecute DV cases in the Bronx reflected the DA’s policy to focus primarily on “first-party complaints.” When an arrest was made as a result of a domestic violence incident in the Bronx, the DA would generally not pursue the case unless the victim (the “first party”) signed the complaint. The Bronx DV Bureau emphasized the importance of speaking to the victim to learn the history of the relationship and details about the facts of the 27

Cases referred to Family Court by the DA’s office are distinct from cases where the victim elects to bring the case in Family Court. As noted earlier, we have no data on the latter type of case.

28 For these analyses of arrests that were never arraigned in Criminal Court, we identified DV cases on the basis of the NYPD data on defendant-victim relationship.

42 case that were not included in police reports. Because defendants in New York must generally be arraigned within 24 hours of arrest, DV victims in the Bronx usually had less than 24 hours to decide whether or not to sign the complaint. If the victim refused to speak to the DA’s office or to sign the complaint, the case would be difficult to prosecute. Without victim cooperation, the DV Bureau almost always declined to prosecute DV cases.

In contrast to the Bronx, Brooklyn and Manhattan actively prosecuted cases without victim cooperation. During our discussions with representatives from the Brooklyn and Manhattan DA’s offices, several reasons for prosecuting these cases were mentioned. First, the DA’s were sometimes able to secure the cooperation of the victim even after an initial refusal to cooperate. Second, the DA’s office may have tried to develop other evidence in order to prosecute the case without the victim’s cooperation. Sometimes, for example, the victim’s signed statement on a “Domestic Incident Report” is sufficient corroborating evidence to sustain the charges. They could also pursue an “evidence-based” prosecution without the victim’s cooperation, using “hearsay exceptions” (e.g., statements made by victims and recorded on 911 tapes), photographs, police testimony, medical evidence, etc. Third, the DA’s office was able to keep an order of protection in effect while the case against the defendant was pending. This may have prevented further violence, or may have enabled a subsequent prosecution if there was a violation of the order of protection during the pendency of the case. Finally, the DA’s office sometimes worked with uncooperative victims to provide services (counseling, housing assistance, etc.), which improved victim safety even if the case was subsequently dismissed.

The two different policies for screening cases reflect different philosophies for prosecuting DV cases. Focusing on first-party complaints enabled the Bronx DA’s office to focus its efforts on viable cases where the victim agreed to file a complaint. This policy may have increased the chances of winning a conviction, since the victim indicated an initial willingness to cooperate with the prosecution. The alternative approach, prosecuting virtually all cases that came to their attention, enabled the Brooklyn and Manhattan DA’s offices to keep more cases active. During this time, the ADA’s tried to develop additional evidence, to encourage the victim to cooperate, and to provide services to the victim. In Brooklyn, even though many weak cases ended in dismissal, they were kept active as long as possible in order to allow the court and the DA’s office to monitor the defendants. In Manhattan, the DA’s office was not as concerned with monitoring defendants’ behavior prior to disposition. Although their goals were somewhat different, the DA’s offices’ in both Brooklyn and Manhattan prosecuted many DV cases where evidence-based prosecution was not possible and the victim never cooperated with the prosecution. Most of these cases were presumably likely to end in dismissal of the charges.

Based on these descriptions of the policies of the DA’s offices, the Brooklyn and Manhattan offices had what is generally described in the literature as a no-drop policy in DV cases. Virtually all DV cases in Brooklyn and Manhattan were prosecuted. In both Brooklyn and Manhattan, charges were not dropped at the victim’s request except in rare cases. The Bronx relied on a first-party complaint policy in DV cases. The DA’s office did not usually pursue cases where the victim refused to sign the complaint. These different policies in Brooklyn, Manhattan and the Bronx indicate that the goals in each borough were different. In Brooklyn, the

43 emphasis was on monitoring the behavior of as many defendants as possible for as long as possible, even when conviction was unlikely. In Manhattan, the emphasis was on developing additional evidence to obtain a conviction. In the Bronx, resources were concentrated on cases where the victim indicated an initial willingness to go forward with the case in order to obtain convictions in those cases. Victims rarely had an influence over the prosecutor’s decision to file a complaint in Brooklyn or Manhattan, while victims were primarily responsible for deciding whether a prosecutable case would be pursued in the Bronx.

The case screening policies described above were generally applied in most DV cases.

However, in Brooklyn and Manhattan, as in most jurisdictions with a no-drop policy, the policy was not absolute. Cases were sometimes declined for prosecution, and cases were also occasionally dropped at the victim’s request if the DA’s office was satisfied that dropping the case would not endanger the victim. Smith et al. (2000) have suggested that “no-drop” is more accurately described as a philosophy than as a strict policy. While the general policy is to prosecute every case, exceptions are made on occasion. Similarly, there were also exceptions to the first-party complaint policy in the Bronx. Prosecutors sometimes signed the complaint (e.g., if the victim was in the hospital) and later tried to obtain a supporting deposition from the victim.

Although there were exceptions to the no-drop policies in Brooklyn and Manhattan, and to the first-party complaint policy in the Bronx, the vast majority of cases in each borough in the third quarter of 1998 were screened according to the policies described above.29 As a result of the policy requiring that victims sign the complaint, the volume of DV cases pursued in the Bronx was lower than it otherwise would have been. The screening policy in the Bronx not only reduced the volume of DV cases that were docketed, but also prevented many weak DV cases from being docketed. DV cases where the victim refuses to cooperate are among the most difficult to prosecute, and often result in a dismissal of the charges. In the Bronx, cases were usually declined for prosecution if the victim refused to sign the complaint, whereas in Brooklyn and Manhattan cases where the victim did not wish the prosecution to go forward were usually prosecuted.

B. Re-arrest Rates by Case Outcomes and Criminal Sanctions in the Bronx To assess the impact of case screening policies on re-arrest rates, we conducted analyses limited to the Bronx, comparing re-arrest rates for cases that were DP’d to cases that were prosecuted. The number of DP’d cases in the other boroughs is too small for reliable statistical comparisons of DP’d cases to prosecuted cases. Information about the re-arrest rates for DP’d cases and prosecuted cases in the Bronx is provided in Table 5-1.

Our test of the impact of case screening focuses on the difference in re-arrest rates between cases that were declined for prosecution and cases that were dismissed. In the Bronx, 29

The use of the term “policy” in this report to describe how the DA’s offices screen DV cases does not necessarily mean that there were written procedures stating how these cases were to be screened. The term “policy” is used here to describe the dominant practices used to screen cases in each of the three boroughs.

44 cases were declined for prosecution when the victim refused to sign the complaint. When the victim signed the complaint, cases in the Bronx were dismissed if the victim’s testimony and/or other evidence in the case was insufficient to obtain a conviction. Some victims who initially signed the complaint later refused to cooperate with the prosecution; many of these cases also ended in dismissal. Dismissal would have been the most likely outcome if cases that were DP’d had been prosecuted, since all victims in DP’d cases were initially uncooperative. The re-arrest rate for cases that were declined for prosecution in the Bronx was 18%, compared to 14% for cases that were dismissed (see Table 5-1). This difference in re-arrest rates was in the direction we expected, and suggests the possibility that victims in cases that were declined for prosecution were in greater danger than victims in cases that were dismissed.

Although the results appear to support our hypothesis, these data are not sufficient to support this conclusion, for two reasons. First, although the re-arrest rate for cases that were DP’d is higher than for cases that were dismissed, the difference between these two percentages was not statistically significant (results not shown). This means that the higher re-arrest rate for DP’d cases may have been due to chance alone, and not to any real difference in re-arrest rates.

Second, even if the difference in re-arrest rates was statistically significant, we cannot be sure that re-arrest rates would have been lower if the Bronx DA had pursued the DP’d cases without victim cooperation. There may have been other differences between defendants whose cases were DP’d and defendants whose cases were dismissed. For example, defendants whose cases were declined for prosecution may have had a more extensive criminal history, which may have accounted for their higher re-arrest rate. The logistic regression model presented below reports results that take these differences into account.

Although the difference between DP’d and dismissed cases was not statistically significant, the overall differences in re-arrest rate by case outcomes and criminal sanctions in the Bronx were statistically significant. As we found in the citywide data (see discussion of Table 4-1), defendants who were convicted, particularly those who were sentenced to jail, were more likely to be re-arrested. The pattern of findings for the Bronx is quite similar to the citywide pattern: the re-arrest rate is highest for convicted defendants who were sentenced to jail, somewhat lower for convicted defendants who did not receive a jail sentence, and lowest for dismissed cases. (The Bronx was the only borough that did not dispose of DV cases with an ACD, so no comparable data for ACD’s are available).

C. Logistic Regression Model Examining the Effect of Case Screening on the Re-arrest Rate for DV Offenses Our logistic regression model assesses whether the case screening policy in the Bronx affected the likelihood that a defendant was re-arrested in the subsequent 18 months. The hypothesis to be tested is whether cases that were declined for prosecution had higher re-arrest rates than cases that were prosecuted. The test of the hypothesis was limited to cases from the

45

TABLE 5-1: RE-ARREST RATES FOR DV OFFENSES BY ARREST OUTCOMES, CASE OUTCOMES AND CRIMINAL SANCTIONS IN DV CASES IN THE BRONX1 Crimes Against Persons and Property Subsample Third Quarter 1998 Dataset ARREST OUTCOME, CASE OUTCOMES AND CRIMINAL SANCTIONS

Re-Arrest Rate For DV Offenses

Declined for Prosecution (N of cases)

18% (268)

Dismissed (N of cases)

14% (417)

Convicted, no Jail Sentence (N of cases)

18% (587)

Convicted, with Jail Sentence (N of cases)

29% (163)

Total, All DV Cases (N of cases)

18% (1,435)

Chi-square = 16.64, statistically significant at p < .001 1

Each case represents the first DV arrest during the third quarter of 1998, if any, for each defendant arrested in the quarter. Defendants whose first DV arrest during the third quarter of 1998 resulted in an ACD (N = 7) or in a case disposed in Supreme Court (N = 21) are excluded from the numbers reported here. Also excluded are defendants whose first DV case during the third quarter of 1998 did not reach a final disposition in Criminal Court by August 6, 1999 (N = 113).

Bronx, which was the only borough that had a sizeable number of cases that were declined for prosecution. To test the hypothesis, the logistic regression model used dummy variables to compare case outcomes and criminal sanctions for prosecuted cases to cases that were declined for prosecution. The logistic regression technique is described in further detail in Appendix B. Our statistical model generally paralleled the model used in the analysis of case outcomes and criminal sanctions, as described in Section IV. The model included a variety of control variables, since cases were not randomly assigned to be prosecuted or declined for prosecution.

If the factors that affected the decision to prosecute also affected the likelihood that the defendant was re-arrested, then the relationship between the decision to prosecute and re-arrest would

46 appear to be stronger than it really was. A smaller number of control variables were available for the comparison of DP’d cases to prosecuted cases, since DP’d cases, by definition, did not have case processing data available. We used the following control variables in this model: penal law article of most severe arrest charge, sex, age, ethnicity, defendant-victim relationship, criminal history, community ties and number of arrest charges. The inclusion of these control variables gave us more confidence that any relationship between case screening outcomes and re-arrest rates reflected their independent effects on re-arrest rates.

The logistic regression model presented in Table 5-2 examines the effect of case outcomes and criminal sanctions on re-arrest rates in the Bronx. Cases that ended in dismissal or conviction with or without a jail sentence were compared to cases that were declined for prosecution. The results show that there were no statistically significant differences in re-arrest rates by case outcome. Those whose cases were dismissed, and those who were convicted, whether or not they were sentenced to jail, had re-arrest rates similar to those whose cases were declined for prosecution.

To understand these findings, we compared the results for the model in Table 5-2 to the results for a model that included case outcomes and criminal sanctions as the only predictors of likelihood of conviction (model not shown). These results contrast the net effect of case outcomes and criminal sanctions (i.e., after controlling for all the other variables in the model), and the total effect of case outcomes and criminal sanctions (i.e., using case outcomes and criminal sanctions as the only predictors). When case outcomes and criminal sanctions were the only predictors in the model, there were no statistically significant differences between defendants whose cases were dismissed, those who were convicted and did not receive a jail sentence, and those whose cases were declined for prosecution. However, when case outcomes and criminal sanctions were the only predictors in the model, the odds ratio for defendants sentenced to jail was statistically significant. The odds ratio of 1.86 indicates that the odds of re-arrest for these defendants were 1.86 times greater than for a defendant whose case was declined for prosecution. This total effect can be contrasted with the net effect reported in Table 5-2, where the odds ratio for a defendant sentenced to jail was a statistically insignificant 1.03.

This indicates that the odds of re-arrest were about the same for a defendant sentenced to jail and a defendant whose case was DP’d. In other words, once control variables were included in the model, the difference in re-arrest rates between defendants sentenced to jail and defendants whose cases were DP’d essentially disappeared. Which variables in Table 5-2 were primarily responsible for reducing this gap? An examination of results (not shown) indicates that the number of prior misdemeanor convictions was the main variable that narrowed the gap.

Defendants who were convicted and sentenced to jail had more prior misdemeanor convictions than defendants whose cases were DP’d.

Although the difference was not statistically significant, defendants whose cases were dismissed had a re-arrest rate that was lower than defendants whose cases were DP’d. The odds ratio for dismissed cases was .72, indicating that the odds of re-arrest for a dismissed case were only .72 times as large as the odds for a DP’d case. Stated another way, the odds of re-arrest

47

TABLE 5-2 LOGISTIC REGRESSION MODEL PREDICTING LIKELIHOOD OF REARREST FOR A DV OFFENSE: DV OFFENDERS IN THE BRONX CRIMES AGAINST PERSONS AND PROPERTY SUBSAMPLE 1 Third Quarter 1998 Dataset

2

INDEPENDENT VARIABLES CASE OUTCOME Reference Category: Declined for Prosecution Dismissed Convicted, no Jail Sentence Convicted, with Jail Sentence

Standardized b

Odds

-0.21 -0.03 0.02

0.72 0.95 1.03

0.29 ** 0.18 -0.06 0.00

1.83 1.88 0.82 1.00

-0.43 **

0.44

0.01 -0.03 0.05

1.04 0.97 1.27

-0.05 -0.13 -0.08

0.93 0.83 0.86

0.06 0.11 -0.05 -0.02

1.12 1.18 0.91 0.97

Ratio

CONTROL VARIABLE ARREST CHARGE PENAL LAW ARTICLE:

Reference Category: Assault (PL 120) Criminal Contempt (PL 215) Harassment (PL 240) Crimes Against Children (PL 260) Other DEFENDANT'S DEMOGRAPHIC CHARACTERISTICS SEX (Female) ETHNICITY:

Reference Category: Black White Hispanic Other AGE:

Reference Category: Age 16-20 Age 21-29 Age 30-39 Age 40 and over DEFENDANT-VICTIM RELATIONSHIP Reference Category: Married Boyfriend-Girlfriend Common-Law Marriage Other Relationship Missing

Table Continues on Next Page

48

TABLE 5-2 (continued)

2

INDEPENDENT VARIABLES

Standardized b

Odds Ratio

DEFENDANT'S CRIMINAL HISTORY ANY PRIOR ARRESTS NUMBER OF PRIOR MISDEMEANOR CONVICTIONS NUMBER OF PRIOR FELONY CONVICTIONS ANY ARRESTS FOR A DV OFFENSE PRIOR TO CASE DISPOSITION ANY ARRESTS FOR A NON-DV OFFENSE PRIOR TO CASE DISPOSITION

0.12 0.22 * 0.04

1.19 1.04 1.04

0.29 ***

2.36

0.02

1.06

0.18 0.10 0.11 -0.02 0.24 * 0.03

1.30 1.17 1.16 0.98 1.43 1.11

0.24 *

1.20

DEFENDANT'S COMMUNITY TIES UNEMPLOYED AT CURRENT ADDRESS 1 YEAR OR LESS LIVES WITH SOMEONE DOES NOT EXPECT ANYONE AT ARRAIGNMENT HAS NO TELEPHONE LIVES OUTSIDE NYC AREA

ARREST AND ARRAIGNMENT CHARGE CHARACTERISTICS NUMBER OF ARREST CHARGES

Nagelkerke R2 (N of cases)

.10 *** (1,435)

NOTES 1 2

See text for a description of the dataset and the subsample.

See Appendix C for information about the measurement and coding of the variables.

* Statistically significant at p < .05 ** Statistically significant at p < .01 *** Statistically significant at p < .001

49 were 1.39 (1 / .72) times greater for a case that was DPâ&#x20AC;&#x2122;d than for a case that was dismissed.

Although not statistically significant (p = .18), this difference is in the direction predicted by the hypothesis. It is possible that in a larger sample this difference would be statistically significant.

A review of the effects of the other independent variables in the model indicates that the strongest predictors of re-arrest (as measured by standardized beta) were whether the defendant was re-arrested for a DV offense before the disposition of the original case, whether the defendant was arrested for a violation of an order of protection (PL 215), the number of arrest charges and the number of prior misdemeanor convictions (all of which increased the odds of re-arrest) and whether the defendant was female (which decreased the odds of re-arrest). These findings were generally consistent with the citywide findings reported in Table 4-3. The model for the Bronx explained about 10% of the variation in the likelihood of re-arrest, which is similar to the 11% of variation explained citywide (see Nagelkerke R2 in Tables 5-2 and 4-3).

At first glance, the findings on the effect of community ties on re-arrest rates in the Bronx appear to be weaker than in the citywide model. Only one of the variables (â&#x20AC;&#x153;has no telephoneâ&#x20AC;?) had a statistically significant effect on the likelihood of re-arrest. However, our finding that the effects of the other community ties variables in the Bronx were not statistically significant may be due to the smaller sample size (1,435 for the Bronx vs. 6,489 for the citywide model).

Statistically significant effects are more difficult to detect in smaller samples. A closer examination of the model for the Bronx supports this explanation. The effect of being unemployed was nearly statistically significant (p = .088). Furthermore, the magnitude of the effect of three of the variables (being unemployed, living at current address less than one year, and living with someone) was about the same in this model as it was in the citywide model.

These findings suggest that with a larger sample size, the effects of the community ties measures in the Bronx were large enough to be statistically significant predictors of the likelihood of re-arrest for a DV offense.

The effects of criminal history were also not as strong in the Bronx as in the citywide findings. The effects of prior arrests and of re-arrests for Non-DV offenses while the original case was pending were not statistically significant in the Bronx as they were in the citywide model. Moreover, the effects of these variables, as measured by the odds ratios, were weaker.

However, two findings were consistent with the citywide model. Defendants who were re-arrested for a DV offense while the original case was pending and who had a larger number of prior misdemeanor convictions were more likely to be re-arrested for a DV offense.

Finally, in the Bronx, as in the citywide model, the defendant-victim relationship did not have a statistically significant impact on re-arrest rates.

D. Summary and Discussion of Findings Our hypothesis that DV offenders whose cases were declined for prosecution would be more likely to be re-arrested for DV offenses than those whose cases were prosecuted was not confirmed. Our initial examination of re-arrest rates by case outcome and criminal sanction in the Bronx appeared to support the hypothesis. About 18% of defendants whose cases were

50 declined for prosecution were re-arrested, compared to only 14% of those whose cases were dismissed. (We compared the DPâ&#x20AC;&#x2122;d cases to the dismissed cases because if a DPâ&#x20AC;&#x2122;d case had instead been prosecuted, the most likely outcome would have been a dismissal.) However, this difference was not statistically significant. Furthermore, a closer examination revealed that there were no differences between DPâ&#x20AC;&#x2122;d and dismissed cases after controlling for defendant and case characteristics. Although re-arrest rates were higher for cases declined for prosecution than for dismissed cases, the difference was not statistically significant and our findings failed to support our hypothesis.

These findings are consistent with two previous studies of the impact of case screening (Fagan et al. 1984, Davis et al. 1998), neither of which found that defendants whose cases were declined for prosecution had a higher re-arrest rate than those whose cases were prosecuted.

However, our findings are inconsistent with those of Wooldredge and Thistlethwaite (2002), who found that Cincinnati defendants whose cases were declined for prosecution had a statistically significant higher re-arrest rate. One reason for the inconsistent findings may be that all the studies, including the current one, are limited by relatively small sample sizes. Fagan et al.

(1984) compared 196 cases declined for prosecution to 74 prosecuted cases. Wooldredge and Thistlethwaite (2002) compared 218 cases declined for prosecution to 2,892 prosecuted cases.

Davis et al. (1998) had a larger sample of cases declined for prosecution (N = 464) than the current study (N = 268), but their sample of prosecuted cases was smaller (N = 574) than the sample in the current study (N = 1,167). The hypothesis should be tested in future research with larger samples before any definitive conclusions are drawn.

Fewer variables were statistically significant in the model predicting the likelihood of re-arrest in the Bronx than in the citywide model. The finding that many of the community ties variables did not have statistically significant effects is probably due to the small sample size in the Bronx. However, the finding that the effects of two of the criminal history variables were not statistically significant may indicate that these variables were less influential in the Bronx than in other boroughs.

51 VII. CONCLUSION A. Major Findings The current study has examined re-arrests for a DV offense among a sample of offenders arrested in the third quarter of 1998 for crimes against persons and property. Re-arrests were tracked over an 18-month period following the disposition of the third quarter 1998 case, and classified as DV re-arrests or Non-DV re-arrests. The study tested two hypotheses: 1) Defendants in DV cases with more severe case outcomes and criminal sanctions have lower rates of re-arrest for DV offenses.

2) Arrestees whose DV cases are declined for prosecution have higher rates of re-arrest for DV offenses than those whose cases are prosecuted.

There are four major findings from the study.

First, we found no support for the first hypothesis, that more severe case outcomes and criminal sanctions are associated with lower re-arrest rates. Our analysis of citywide data showed that re-arrest rates were actually higher for defendants whose cases had more severe outcomes. (Outcomes were ranked from least to most severe in the following order: dismissal, adjournment in contemplation of dismissal (ACD), conviction without a jail sentence, and conviction with a jail sentence.) Re-arrest rates were higher for defendants who were convicted, particularly those who were convicted and sentenced to jail, than for defendants whose cases were dismissed or ACDâ&#x20AC;&#x2122;d. However, when we used a statistical model to take into account defendantâ&#x20AC;&#x2122;s criminal history, these differences in re-arrest rates disappeared. The higher re-arrest rates for defendants who were convicted actually reflected the continuation of their prior records as chronic offenders. Our statistical model showed that the severity of case outcomes and criminal sanctions was not related to the likelihood of re-arrest. After taking criminal history into account, there was no evidence that conviction and jail deterred future acts of domestic violence. In short, our statistical model failed to provide support for the first hypothesis.

Second, we found little support for our second hypothesis, that arrestees whose DV cases are declined for prosecution have higher rates of re-arrest than those whose cases are prosecuted.

Our analysis was limited to cases in the Bronx, the only borough to decline prosecution of significant numbers of cases. We found that the re-arrest rate for cases declined for prosecution was 18%, compared to 14% for cases that were dismissed. Although this suggests that declining to prosecute a case may increase the likelihood of re-arrest, the difference was not statistically significant. Furthermore, our logistic regression model that included a variety of control variables also found that the difference was not statistically significant. We concluded that our analyses did not support our second hypothesis.

Third, we found several factors that do affect re-arrest rates for DV offenses. We found that weak community ties are associated with higher rates of re-arrest. This hypothesis had been examined in only one of the previous studies of the impact of case outcomes and criminal

52 sanctions (Wooldredge and Thistlethwaite 2002). That study found only one measure of community ties that affected recidivism: residential stability. Our study is the first to find consistent effects of several community ties measures on re-arrest rates for DV offenses. As most previous studies have reported, we also found that defendants with more serious criminal histories were more likely to be re-arrested. In our study, defendants who were re-arrested for DV offenses while their case was pending, defendants who had any prior arrests, and defendants with more prior misdemeanor convictions were all more likely to be re-arrested for a DV offense.

Finally, although we identified variables that had statistically significant effects on re-arrest rates for DV offenses, the explanatory power of our statistical models was relatively weak. Our citywide model explained 11%, and our Bronx model 10%, of the variation in the likelihood of re-arrest for a DV offense. This suggests that additional factors, unmeasured in our statistical models, are likely to play a strong role in predicting re-arrest rates. Unfortunately, our results appear to be typical of research in this area.

B. Discussion and Conclusions Our finding that more severe case outcomes and criminal sanctions did not reduce re-arrest rates for DV offenses is consistent with all the previous studies that tested this hypothesis. Although it is consistent with these studies, this finding is nevertheless an important contribution. Only one of the previous studies (Wooldredge and Thistlethwaite 2002) used a strong research design and a sufficiently large sample to draw definitive conclusions. Our study, with a strong design and an even larger sample, confirms Wooldredge and Thistlethwaiteâ&#x20AC;&#x2122;s (2002) findings. Ours is also the first study to examine data from New York City, where the volume of DV cases is high and significant resources have been devoted to combating domestic violence.

Our finding that declining to prosecute DV cases did not have a statistically significant effect on re-arrest rates is consistent with the findings of three of the four previous studies.

Again, our study has improved on prior research. Our sample size is larger than the Fagan et al.

(1984) and the Murphy et al. (1998) studies. Our measure of recidivism identifies re-arrests for a DV offense, unlike the Davis et al. (1998) study, which examined re-arrests for any offense.

Only Wooldredge and Thistlethwaiteâ&#x20AC;&#x2122;s (2002) Cincinnati study found higher re-arrest rates when charges were not filed against a DV offender. Our study contradicts their findings. This may reflect differences between Cincinnati and New York in the way decisions about declining prosecution are made. At the very least, it suggests that research in other jurisdictions is needed to determine whether declining to prosecute DV offenders increases the likelihood of re-arrest for a DV offense.

Of course the magnitude of the difference between cases declined for prosecution and those dismissed was relatively small in the current studyâ&#x20AC;&#x201D;about 4%. If a future study with a larger sample size found a statistically significant difference of this size, some might argue that the finding supports using a no-drop policy to prosecute virtually all DV cases. Such a policy would presumably reduce the likelihood of re-arrest in cases that were formerly declined for prosecution. On the other hand, it could also be argued that the small (4%) decrease in the

53 re-arrest rate would not justify the commitment of resources to additional cases that are likely to result in dismissal. In this view, the first-party complaint policy produces a more efficient allocation of prosecutorial resources to cases that have a reasonable chance of conviction.

Although we did not find support for either of our hypotheses, it is worth noting that we also did not find any indication that there are escalation effects. Prosecuting, convicting, and jailing DV offenders did not increase the risk of re-arrest, as some have speculated (Fagan 1996, Ptacek 1999). Our initial finding that DV offenders who were convicted, particularly those sentenced to jail, were more likely to be re-arrested raised the possibility of an escalation effect.

However, our statistical model demonstrated that the higher re-arrest rate was due to the criminal history of the offender. Once criminal history was taken into account, there was no evidence of an escalation effect of conviction with or without jail.

Furthermore, the use of judicial monitoring and treatment programs rather than jail does not appear to increase the risk of re-arrest. We compared convicted defendants who received a non-jail sentence (primarily a conditional discharge with a requirement for completing a drug, alcohol or batterer intervention program) to convicted defendants who were sentenced to jail.

After taking differences in their criminal histories into account, the re-arrest rate was the same for those who received a non-jail sentence and those who received a jail sentence. This suggests that the use of treatment programs is appropriately targeted and does not increase the risk of re-arrest.

The judicial monitoring that routinely accompanies these programs is as effective as jail in preventing future DV offenses.

C. Policy Implications The finding that case outcomes and criminal sanctions (prosecution, conviction, and jail sentences) did not deter re-arrest raises the question of whether and how domestic violence can be prevented or controlled. The literature on deterring domestic violence indicates that little is known about how to prevent DV offenders from re-offending (Fagan 1989, 1996). Why is domestic violence so difficult to control? Fagan (1996) suggests several reasons. Unlike other types of violent crime, domestic violence usually involves victims and offenders who are in daily contact and who have emotional ties to each other. Domestic violence frequently occurs in the privacy of the home, where it is more difficult for the police or others to intervene. The DV offenderâ&#x20AC;&#x2122;s opportunity for violence and the relationship that may trigger it are often ongoing.

Furthermore, many interventions assume that the DV offender will act rationally, weighing costs and benefits. However many offenders may batter as a result of psychological dysfunction or a mental disorder. Finally, many offenders may lack community ties that provide informal social control, and many live in neighborhoods characterized by social disorganization.

Our analyses showed that prosecuting, convicting and sentencing DV offenders to jail did not deter re-arrest. Once the criminal history of DV offenders was taken into account, their rearrest rates were essentially the same regardless of the outcomes of their cases. At first glance, this may seem disheartening, since it suggests that â&#x20AC;&#x153;getting toughâ&#x20AC;? with DV offenders does not deter future violence. However, the failure of the sanctions to reduce future violence is not a failure of the courts, but instead is a reflection of the inherent difficulties of preventing domestic

54 violence. A closer examination of the findings provides support for some of the key elements of current criminal justice interventions, suggests some promising ideas for reducing recidivism, and draws attention to the importance of goals other than deterrence. The results of our research on recidivism have four important implications for policy and practice.

First, our research indicates that criminal sanctions are applied in DV cases in ways that do not increase the risk of future violence. The control approach used in the specialized DV parts, which is based on judicial monitoring and treatment programs, was as effective at deterring re-arrest as sentencing DV offenders to jail. Although monitoring and treatment may not seem as “tough” as jail, our research suggests that they are appropriate sentences for DV offenders.

Furthermore, we found no evidence that violence escalates when the criminal justice system does “get tough” with certain DV offenders. Prosecuting, convicting and jailing DV offenders did not increase the likelihood of future violence.

Second, the findings in the current study suggest some ways to tailor different criminal justice interventions for different types of offenders. The criminal justice response may need to be more severe for those with weak community ties, for example, by requiring a greater level of monitoring by the court. A lower level of monitoring may be adequate for defendants with stronger community ties. Similarly, our research supports the current practice of predicating the severity of the criminal justice response on the defendant’s criminal history. DV offenders who had more prior convictions were more likely to be sentenced to jail. First offenders were more likely to receive a conditional discharge or an ACD. Third, the finding that case outcomes and criminal sanctions do not influence re-arrest rates does not necessarily mean that criminal justice interventions are ineffective. Arrest, regardless of the subsequent case outcome, may be a deterrent to domestic violence. The current study did not compare defendants who committed offenses but were not arrested to those who were arrested. However, previous research suggests that arrest may deter future violence (Frisch et al. 2001, Maxwell et al. 2002).

Finally, goals other than deterrence may be appropriate in DV cases. Criminal sanctions and judicial monitoring may reduce the severity of future acts of violence, even if they do not deter the commission of violence. Furthermore, some DA’s offices provide services (e.g., counseling, referrals to shelters) to victims in DV cases and offer victims additional means of staying safe (e.g., developing a safety plan, providing emergency 911 cell phones). Finally, and perhaps most importantly, prosecuting and sentencing domestic violence offenders sends a message to the batterer, the victim, and the community that the criminal justice system is prepared to intervene to stop the violence.

D. Future Research What should be the priorities for future research on re-arrests among DV offenders?

First, more research is needed to test the hypothesis that declining to prosecute DV cases increases re-arrest rates. Our study found that the re-arrest rate was higher for cases declined for

55 prosecution than for dismissed cases; this difference was not statistically significant. A previous study, using data from Cincinnati, did find a higher re-arrest rate when charges were not filed.

We believe further study in other jurisdictions and with larger samples is warranted to explore this possibility more fully.

Second, we believe further study is needed to examine the impact of community ties on re-arrest rates. Our finding that weak community ties increase the re-arrest rate suggests that this is a fruitful area for further research. Using several measures of community ties, it might be possible to develop a profile of high-risk and low-risk offenders. Such a profile might be useful to district attorneys in deciding on which cases to concentrate their efforts, or to the court in deciding on conditions of release while the case is pending, levels of judicial monitoring, and appropriate sentences for DV offenders. It also would be interesting to know if the impact of community ties depends on the case outcome, criminal sanction or criminal history. Are community ties stronger predictors of re-arrest rates for defendants whose cases were dismissed or ACDâ&#x20AC;&#x2122;d vs. those who were convicted? Are community ties stronger predictors of re-arrest rates for defendants who have been arrested for the first time vs. repeat offenders? We hope to conduct additional research to explore the impact of community ties on re-arrest rates.

Finally, to provide more detailed information about deterrence, additional dependent variables should be examined. What factors affect the severity of the injuries inflicted in DV rearrests? Can we identify offenders who are at greatest risk of committing assaults that result in significant physical injury or death? Addressing these questions would require data not available in the official records to which we have access.

The risks faced by the victims of domestic violence are significant, and there is a great need for more research and new policies and practices to address the problem. We hope that the current study will contribute to the development of new approaches for policy, practice and research.

56

[This page intentionally left blank]

57 VII. REFERENCES Aldrich, Liberty and Julie A. Domonkos. 2000. “Navigating the NYPD: A Guide to New York City Police Department Policies and Procedures for Family Court Attorneys Handling Order of Protection Cases.” Pp. 279-292 in Julie A. Domonkos and Jill Laurie Goodman, Lawyer’s Manual on Domestic Violence: Representing the Victim. New York: Supreme Court of the State of New York, Appellate Division, First Department.

Berk, Richard A., Alec Campbell, Ruth Klap and Bruce Western. 1992. “A Bayesian Analysis of the Colorado Springs Spouse Abuse Experiment.” The Journal of Criminal Law and Criminology 83:170-200.

Binder, Arnold and James W. Meeker. 1988. “Experiments as Reforms.” Journal of Criminal Justice 16:347-358.

Cahn, Naomi R. (ed.). 1992. “Innovative Approaches to the Prosecution of Domestic Violence Crimes: An Overview.” Pp. 161-180 in Eve S. Buzawa and Carl G. Buzawa (eds.), Domestic Violence: The Changing Criminal Justice Response. Westport, CT: Greenwood.

Cahn, Naomi R. and Lisa G. Lerman. 1991. “Prosecuting Woman Abuse.” Pp. 95-112 in Michael Steinman (ed.), Woman Battering: Policy Responses. Cincinnati, OH: Anderson Publishing.

Carlson, Christopher and Frank J. Nidey. 1995. “Mandatory Penalties, Victim Cooperation, and the Judicial Processing of Domestic Abuse Assault Cases.” Crime and Delinquency 41:132-149.

Davis, Robert C., Barbara E. Smith and Laura B. Nickles. 1998. “The Deterrent Effect of Prosecuting Domestic Violence Misdemeanors.” Crime and Delinquency 44:434-442.

Dunford, Franklyn W. 1992. “The Measurement of Recidivism in Cases of Spouse Assault.”

The Journal of Criminal Law and Criminology 83:120-136.

Eckert, Mary A., and Mari Curbelo. 2000. Alternative to Incarceration Information Services: First Half Fiscal Year 2000. New York: NYC Criminal Justice Agency.

Elliott, Delbert S. 1989. “Criminal Justice Procedures in Family Violence Crimes.” Pp. 427480 in Lloyd Ohlin and Michael Tonry (eds.), Crime and Justice, A Review of Research: Family Violence (Volume 11). Chicago: University of Chicago.

Fagan, Jeffrey. 1989. “Cessation of Family Violence: Deterrence and Dissuasion.” Pp. 377-425 in Lloyd Ohlin and Michael Tonry (eds.), Crime and Justice: A Review of Research, Volume 11: Family Violence. Chicago: University of Chicago.

58 Fagan, Jeffrey. 1996. The Criminalization of Domestic Violence: Promises and Limits.

National Institute of Justice: Washington, D.C. Fagan, Jeffrey, Elizabeth Friedman, Sandra Wexler, and Virginia Lewis. 1984. National Family Violence Evaluation Final Report. Washington, D.C.: U.S. Department of Justice.

Feder, Lynette. 1998. “Police Handling of Domestic and Nondomestic Assault Calls: Is There a Case for Discrimination?” Crime and Delinquency 44:335-349.

Ford, David A. and Mary Jean Regoli. 1993. “The Criminal Prosecution of Wife Assaulters: Process, Problems, and Effects.” Pp. 127-164 in N. Zoe Hilton (ed.) Legal Responses to Wife Assault: Current Trends and Evaluation. Newbury Park, CA: Sage.

Ford, David A. and Mary Jean Regoli. 1992. “The Preventive Impacts of Policies for Prosecuting Wife Batterers.” Pp. 181-208 in Eve S. Buzawa and Carl G. Buzawa (eds.), Domestic Violence: The Changing Criminal Justice Response. Westport, CT: Auburn House.

Frisch, Lisa A., Melissa I. Mackey, Donna Hall, and Alissa Pollitz Worden. 2001. Family Protection and Domestic Violence Intervention Act of 1994: Evaluation of the Mandatory Arrest Provisions (Final Report to the Governor and Legislature). Albany, NY: Division of Criminal Justice Services.

Gross, Melissa, Elizabeth P. Cramer, Janett Forte, Jill A. Gordon, Tara Kunkel and Laura J. Moriarty. 2000. “The Impact of Sentencing Options on Recidivism Among Domestic Violence Offenders: A Case Study.” American Journal of Criminal Justice 24:301-312.

Herrell, Stephen B. and Meredith Hofford. 1990. “Family Violence: Improving Court Practice.”

Reno, NV: The National Council of Juvenile and Family Court Judges.

Hirschel, J. David and Ira W. Hutchison. 1992. “Female Spouse Abuse and the Police Response: The Charlotte, North Carolina Experiment.” The Journal of Criminal Law and Criminology 83:73-119.

Hirschel, J. David and Ira W. Hutchison. 1996. “Realities and Implications of the Charlotte Spousal Abuse Experiment.” Pp. 54-82 in Eve S. Buzawa and Carl G. Buzawa (eds.) Do Arrests and Restraining Orders Work? Thousand Oaks, CA: Sage.

Hutchison, Ira W. 1998. “Substance Use and Abused Women’s Utilization of the Police.”

Paper presented at the Annual Meetings of the American Society of Criminology, Washington, D.C. Lerman, Lisa G. 1986. “Prosecution of Wife Beaters: Institutional Obstacles and Innovations.”

Pp. 250-295 in M. Lystad (ed.), Violence in the Home: Interdisciplinary Perspectives.

New York: Brunner/Mazel.

59

Lerman, Lisa G. 1992. “The Decontextualization of Domestic Violence.” The Journal of Criminal Law and Criminology 83:217-240.

Maxwell, Christopher D., Joel H. Garner and Jeffrey A. Fagan. 2002. “The Preventive Effects of Arrest on Intimate Partner Violence: Research, Policy and Theory.” Criminology and Public Policy 2:51-80.

McCord, Joan. 1992. “Deterrence of Domestic Violence: A Critical View of Research.”

Journal of Research in Crime and Delinquency 29:229-239.

McGuire, Linda A. 2000. Criminal Prosecution of Domestic Violence. Washington, DC: U.S. Department Of Justice.

Menard, Scott. 1995. Applied Logistic Regression Analysis. Sage Publications: Thousand Oaks, CA. Mohr, Lawrence B. 1990. Understanding Significance Testing. Sage Publications: Thousand Oaks, CA. Murphy, Christopher M., Peter M. Musser, and Kenneth I. Maton. 1998. “Coordinated Community Intervention for Domestic Abusers: Intervention System Involvement and Criminal Recidivism.” Journal of Family Violence 13:263-284.

NYPD. 2000. Patrol Guide. New York: New York City Police Department.

Peterson, Richard R. 2001. Comparing the Processing of Domestic Violence Cases to Non-Domestic Violence Cases in New York City Criminal Courts: Final Report. New York: NYC Criminal Justice Agency.

Peterson, Richard R. 2002. Cross-Borough Differences in the Processing of Domestic Violence Cases in New York City Criminal Courts: Final Report. New York: NYC Criminal Justice Agency.

Polsby, Daniel D. 1992. “Suppressing Domestic Violence with Law Reforms.” The Journal of Criminal Law and Criminology 83:250-253.

Ptacek, James. 1999. Battered Women in the Courtroom: The Power of Judicial Responses.

Boston: Northeastern University Press.

Rennison, Callie M. and Sarah Welchans. 2000. Intimate Partner Violence. Washington, D.C.: Bureau of Justice Statistics.

60 Sherman, Lawrence W. 1992. “The Influence of Criminology on Criminal Law: Evaluating Arrests for Misdemeanor Domestic Violence.” The Journal of Criminal Law and Criminology 83:1-45.

Sherman, Lawrence W. and Richard A. Berk. 1984. “The Specific Deterrent Effects of Arrest for Domestic Assault.” American Sociological Review 49:261-272.

Siddiqi, Qudsia. 1999. Assessing Risk of Pretrial Failure to Appear in New York City: A Research Summary and Implications for Developing Release-Recommendation Schemes.

New York, NY: NYC Criminal Justice Agency.

Smith, Barbara E., Rob Davis, Laura B. Nickles and Heather J. Davies. 2000. “An Evaluation of Efforts to Implement No-Drop Policies: Two Central Values in Conflict.” Chicago, IL: American Bar Association.

Stone, Audrey. 2000. “Gathering Evidence with an Eye Towards Winning the Case: A Prosecutor’s Guide to Winning Domestic Violence Cases.” Pp. 293-307 in Julie A. Domonkos and Jill Laurie Goodman, Lawyer’s Manual on Domestic Violence: Representing the Victim. New York: Supreme Court of the State of New York, Appellate Division, First Department.

Sullivan, Cris M., Joanne Belknap, Ruth Fleury, Amy Leisenring, Heather Melton, and Meghan Chandek. 2000. “The Relationship between Domestic Violence Case Disposition and Victimization Six Months Later.” Paper presented at the Annual Meeting of the American Society of Criminology, San Francisco, CA, November.

Thistlethwaite, Amy, John Wooldredge and David Gibbs. 1998. “Severity of Dispositions and Domestic Violence Recidivism.” Crime and Delinquency 44:388-398.

Tolman, Richard M. and Arlene Weisz. 1995. “Coordinated Community Intervention for Domestic Violence: The Effects of Arrest and Prosecution on Recidivism of Woman Abuse Perpetrators.” Crime and Delinquency 41:481-495.

Weis, Joseph G. 1989. “Family Violence Research Methodology and Design.” Pp. 117-162 in Lloyd Ohlin and Michael Tonry (eds.), Crime and Justice: A Review of Research, Volume 11: Family Violence. Chicago: University of Chicago.

Wooldredge, John and Amy Thistlethwaite. 2002. “Reconsidering Domestic Violence Recidivism: Conditioned Effects of Legal Controls by Individual and Aggregate Levels of Stake in Conformity.” Journal of Quantitative Criminology 18:45-70.

61 APPENDIX A: PROCEDURES USED FOR IDENTIFYING RE-ARRESTS Our sample of re-arrests was drawn from the CJA database, which includes information about most persons arrested in New York City. We identified re-arrests of the defendants in our sample using two methods to match re-arrests to the original (third quarter 1998) arrest: 1) the defendant’s fingerprint identification number, known as a New York State Identification, or NYSID, number, and 2) the defendant’s name and date of birth. For most defendants in our sample, we used both methods to identify re-arrests. For a small number of defendants, only name and date of birth were used, since no NYSID number was available. In this Appendix, we provide background information about NYSID numbers, and we then describe the procedures that were used to identify re-arrests.

A. NYSID Numbers When an individual is arrested for the first time as an adult (16 years of age or older), a unique NYSID number is assigned to the arrestee’s fingerprints by the New York State Department of Criminal Justice Services (DCJS). If a first-time adult arrestee is ultimately convicted of a misdemeanor or a felony offense, the NYSID number will be permanently assigned to the defendant. The next time the defendant is arrested and fingerprinted, the fingerprints are checked by computer against the database of existing fingerprints. If the new fingerprints are correctly matched to the old fingerprints, the same NYSID number will be assigned to the defendant. If the new fingerprints are not correctly matched to the old fingerprints, a new NYSID number may be mistakenly assigned to the defendant. If this error is subsequently discovered by DCJS, the full criminal history of the defendant is retained using one of the NYSID numbers as a consolidated NYSID number. Under these circumstances, a defendant may have two or more NYSID numbers in the CJA database, although DCJS records information for that defendant using only the consolidated NYSID number.

If a first-time adult arrestee is convicted of a violation or infraction, convicted of a misdemeanor or felony but given Youthful Offender (YO) status, given an Adjournment in Contemplation of Dismissal (ACD), or has his/her case dismissed, the NYSID number assigned at the time of the arrest may eventually be expunged. The NYSID number will not be used again for that arrestee unless he/she is re-arrested within a certain period of time (usually 6 or 12 months) after the case is disposed. If the adult is re-arrested after the 6- or 12-month period has expired, the arrestee is treated as a first-time adult arrestee and a new NYSID number is assigned.

Depending on the disposition of the case for the new arrest, the NYSID number will either become permanent or will be expunged according to the rules described above. It is possible for an adult to be arrested several times and to have his/her record expunged each time. Under these circumstances, a defendant may have multiple NYSID numbers, and each arrest for the same defendant is linked to a different NYSID number.

In some cases, individuals are arrested for “non-printable” offenses, i.e., offenses for which New York State law does not require the taking of fingerprints. These are usually minor offenses in the New York State Penal Law or Vehicle and Traffic Law, or local offenses in the New York City Administrative Code. In our third quarter 1998 Crimes Against Persons and

62 Property sample, most offenses were printable offenses. The major exceptions were disorderly conduct (PL §240.20), harassment in the second degree (PL §240.26) and trespass (PL §140.05).

Arrestees charged with these offenses were not fingerprinted, and the CJA database did not have a NYSID number associated with these arrests.

B. Procedures for Identifying Re-arrests To identify re-arrests for our study, we started with a data file extracted from the CJA database of arrests in New York City for the third quarter of 1998. The data file included 25,457 cases from the Crimes Against Persons and Property Sample (see Peterson 2001, which describes how the sample was selected). This was a case-based file that included information on all eligible prosecuted arrests that were held for arraignment. Some defendants were arrested two or more times during the third quarter of 1998, and information about each of their cases was included in the file. Because the re-arrest study was also designed to examine the re-arrest rate for arrestees whose cases were declined for prosecution (DP’d), we added 905 DP'd cases that met the qualifications for inclusion in the study. This created a file of 26,362 cases. The file included the NYSID number assigned at the time of arrest, and for some arrestees, we had information from DCJS about a consolidated NYSID number for the arrestee. We took two steps to prepare this file for the re-arrest study.

1. Finding NYSID numbers for cases where they were missing. Many of the DP’d cases did not have a NYSID number. Other cases were missing NYSID numbers because the offense was a “non-printable” offense, or because of data entry errors. The total number of cases without a NYSID number was 748. We used two methods to find a NYSID number for the arrestee. First, we checked the CJA database for other arrests under the same or a similar name.

Using information about birth dates, addresses and criminal record, we were often able to establish that the arrestee had a previous or subsequent arrest for which a NYSID number had been assigned. Second, we consulted CJA interview forms when available, since the interviewers sometimes recorded NYSID numbers for cases that were ultimately declined for prosecution or cases where the NYSID number was missing from the police arrest record. These efforts enabled us to find 595 NYSID numbers, leaving 153 cases without a NYSID number.

2. Creating a defendant-based file. Once we had found as many NYSID numbers as possible, we created a defendant-based file that included 23,322 of the 26,631 cases. Each case in the file was either a) the first DV arrest for the defendant in the third quarter of 1998 or, if the defendant had no DV arrests in the third quarter of 1998, b) the first Non-DV arrest for the defendant in the third quarter of 1998. Of these 23,322 defendants, 153 did not have a NYSID number, and 23,169 did have a NYSID number. The defendant-based file excluded arrests that were second or later arrests for some of the defendants (these re-arrests were identified using the procedures described below) as well as arrests that were not eligible for the study (e.g., because the case was disposed in Supreme Court).

Once the defendant-based file was prepared, we used two procedures to identify re-arrests in the CJA database:

63 1. Identifying re-arrests using NYSID matching. To identify re-arrests, we began by searching the CJA database for qualified re-arrests that matched on the defendant's NYSID number for the 23,169 cases that had a NYSID number. To qualify, a re-arrest had to occur either a) between the data of arrest and the date of final disposition of the case in Criminal Court, or b) between the date of final disposition of the case in Criminal Court and 18 months after that date. This process yielded 25,347 re-arrests that matched on NYSID number. We checked this list for possible errors, i.e., re-arrests where the NYSID number on the re-arrest matched the NYSID number on the original arrest, but the defendant on the re-arrest appeared to be a different defendant. This error can occur because of incorrect computer matching or paperwork mistakes.

We decided to exclude 69 re-arrests (0.27% of the total number of re-arrests) because it appeared that the defendant had been assigned the wrong NYSID number. As we did name-searching for these erroneous matches, we identified an additional 21 re-arrests that should have been included in the re-arrest file. After these exclusions and additions, the re-arrest file contained information about 25,299 re-arrests.

2. Identify re-arrests using name matching. We then searched the CJA database using name matching in conjunction with information about date of birth for the 153 cases that had no NYSID number. We found 23 re-arrests that appeared to be valid matches. We also used name and date of birth matching to search for re-arrests for the 23,169 defendants who had a NYSID number associated with their third quarter 1998 arrest. We found 2,668 re-arrests that had not already matched on NYSID number but that appeared to be valid matches.

To summarize our procedures for identifying re-arrests, we began with a file of eligible arrests from the third quarter of 1998 that included 26,631 arrests. We identified NYSID numbers for as many of these arrestees as possible. We then selected the first eligible arrest for each arrestee in the file, yielding a file of information about 23,322 defendants. We searched for qualified re-arrests that matched on NYSID number and found 25,347 re-arrests. We excluded 69 of these re-arrests because the NYSID number match appeared to be incorrect, and added 21 re-arrests found while checking for erroneous matches, leaving us with 25,299 re-arrests that matched on NYSID number. We then used name and date of birth to search for re-arrests that had not already matched on NYSID number. We found 23 re-arrests for those who had no NYSID number, and 2,668 re-arrests for those who did have a NYSID number. This brought the total number of re-arrests to 27,990 (25,299 + 23 + 2,668).

Virtually all (99.7%) of the re-arrests that were identified by matching NYSID numbers appeared to be accurate matches. However, using name and date-of-birth matching added a large number of re-arrests that were not found by NYSID number matching. About 9.6% of the total number of 27,990 re-arrests were found by name and date-of-birth matching, and would not have been found by NYSID number matching alone.

64

[This page intentionally left blank]

65 APPENDIX B: STATISTICAL METHODS The predictive models used to test our hypotheses required the use of logistic regression analysis to predict likelihood of re-arrest. Logistic regression analysis is a statistical technique that is used when the outcome to be explained (i.e., the dependent variable) has two categories.

In our analyses, all cases were coded on our dependent variable in one of two categories: not re-arrested (coded 0) and re-arrested (coded 1). The models predict the likelihood that defendants were re-arrested for a DV offense during the 18 months following the outcome of the prosecution. These predictions were made on the basis of information we have about a variety of defendant and case characteristics (i.e., the independent variables). Logistic regression techniques provide several ways of evaluating the effect of these independent variables on the likelihood of re-arrest.

The current study examined three statistical measures to evaluate the effect of the independent variables. First, we report the statistical significance of each independent variable.

Statistical significance takes into account the size of the sample as well as the magnitude of the effect of the independent variable. Taking this information into account, statistical significance assesses the probability that the effect observed in the sample could have occurred by chance alone. In this report, following standard convention, significance levels of .05 or less were treated as statistically significant. In other words, when an effect has a 5% or less probability of having occurred by chance, we conclude that the independent variable is a statistically significant predictor of the likelihood of re-arrest. One weakness of using statistical significance to measure the effect of an independent variable is that when sample sizes are large (e.g., more than several thousand cases), many independent variables have statistically significant effects even when the magnitude of the effects is small. For example, in a large sample, we may be able to say that having a prior arrest has a statistically significant effect on the likelihood of re-arrest, even though it increases the likelihood of re-arrest only from 17% to 19%. In this hypothetical example, we can say that this difference in re-arrest rates is unlikely to be due to chance.

However, it is also clear that knowing whether a defendant had a prior arrest or not does not explain much of the variation in likelihood of re-arrest.

Our second statistical measure used to evaluate the effect of the independent variables is the odds ratio. The odds ratio supplements information about statistical significance by evaluating the magnitude of the effect of the independent variable. Specifically, it tells us how much the odds of an outcome (e.g., re-arrest) change, for each one unit increase in the independent variable. If an independent variable is coded in two categories (e.g., 0 and 1), then the odds ratio tells us how the odds of the outcome change when cases are coded 1 on the independent variable (vs. cases coded 0). An odds ratio greater than one indicates an increase in the likelihood of the outcome occurring, while an odds ratio less than one indicates a decrease in the likelihood of the outcome occurring. An odds ratio of 1 indicates that the odds of an outcome occurring are not affected by the independent variable.

To return to our previous example, if the odds ratio for the effect of having a prior arrest on the likelihood of re-arrest was 1.12, this would mean that in cases where the defendant had a prior arrest the odds of re-arrest are 1.12 times greater than in cases where the defendant did not

66 have a prior arrest. In contrast, if we examined the impact of whether the defendant was ever released from custody while the case was pending, we might find an odds ratio less than 1. For example, if the odds ratio was .83, this would mean that in cases where the defendant was released from custody the odds of re-arrest are only .83 times as large as the odds when the defendant was not released from custody. To simplify interpretation of odds ratios less than 1, it is common to examine the inverse of the odds ratio (1 divided by the odds ratio). When this is done, the interpretation of the effect of the independent variable is reversed. For example, if the odds ratio for being released from custody is .83, we can take the inverse of the odds ratio, 1.20 (1 divided by .83), and say that in cases where the defendant was not released from custody, the odds of re-arrest were 1.20 times greater than in cases where the defendant was released. Finally, if the odds ratio was 1.00, this would mean that whether or not the defendant was released from custody had no impact on the odds of re-arrest. (These examples are hypothetical and do not necessarily reflect our expectations about the findings).

In the analyses presented in this report, results are presented for independent variables coded in three different waysâ&#x20AC;&#x201D;categorical variables that have two categories, categorical variables that have more than two categories, and continuous variables that measure the quantity of a defendant or case characteristic (e.g., the number of prior felony convictions for the defendant). When a categorical independent variable has two categories, the odds ratio measures the change in the odds when cases are in one category vs. the other (e.g., defendant had a prior arrest vs. did not have a prior arrest). When a categorical independent variable has more than two categories, one of the categories is chosen as a reference category, and the odds ratios measure the effect of being in each of the other categories vs. being in the reference category (e.g., cases in the Brooklyn, Manhattan, Queens and Staten Island are compared to cases in the Bronx, which is used as the reference category). Finally, when the independent variable is continuous, the odds ratio measures the change in the odds associated with an increase of one unit on the scale of the independent variable (e.g., for number of prior misdemeanor convictions, the odds ratio measures the effect of having one additional misdemeanor conviction).

Our third statistical measure used to assess the effect of the independent variables is the standardized beta coefficient (Menard, 1995). Standardized betas take into account not only the change in the likelihood of the outcome associated with a change in the independent variable, but also the distribution of the cases among the categories of the independent variable. Being in one category of an independent variable may have a large effect on the likelihood of an outcome (and therefore the variable may have a large odds ratio), but if there are relatively few cases in that category, the variable will not help to explain much of the variation in the likelihood of the outcome. For example, a defendant charged with a sex offense might have a high probability of re-arrest, and this variable would have a high odds ratio. However, if only a small number of defendants in the sample are charged with a sex offense, this variable would not be able to explain much of the variation in likelihood of re-arrest. Standardized betas measure this overall effect of the independent variable on the dependent variable. Standardized betas vary from -1 to +1; values closer to zero indicate that the effect of the independent variable is relatively small, while values closer to +1 or -1 indicate that the effect of the independent variable is relatively strong.

67 In the current study, we used all three of the measures discussed above. We used the statistical significance level to distinguish those independent variables that had a detectable30 effect on the dependent variable from those that did not. We used the odds ratio to evaluate the size of the effect of the independent variable, and we used the standardized beta to evaluate the ability of the independent variable to account for variation in the dependent variable.

The models we discuss include a large number of predictors of the dependent variable. In these models, the measures of the effect of each independent variable (statistical significance, odds ratio, and standardized beta) evaluate the effect of that independent variable after controlling for the effects of all the other independent variables in the model. These effects represent the net effect of a given independent variable after the effect of all the other independent variables have been taken into account. This net effect differs from the total effect of the independent variable, which is the effect of the independent variable when it is used as the only predictor of the dependent variable. Our tests of the hypotheses were based on the net effect of each variable. For example, we examined the net effect of case outcomes and criminal sanctions on the likelihood of re-arrest. This insured that the effects of any of the independent variables that affect both case outcomes and criminal sanctions and re-arrest were not included in our estimate of the effect of case outcomes and criminal sanctions.

30

Due to sampling error, and limitations of logistic regression techniques, it is possible that some independent variables that do affect the dependent variable are found to be statistically insignificant in our particular sample of cases. See Mohr (1990) for a further discussion of these issues.

68

[This page intentionally left blank]

69 APPENDIX C: CODING OF VARIABLES FOR REGRESSION MODELS

Third Quarter 1998 Dataset 1

CODING SCHEME

VARIABLES DEPENDENT VARIABLE

REARREST FOR A DV OFFENSE WITHIN 18 MONTHS OF CASE DISPOSITION

Re-arrested = 1, Not re-arrested = 0

CASE OUTCOME (Citywide Model, Table 4-3) Reference Category: Dismissed Adjourned in Contemplation of Dismissal

Dismissed: Reference Category Adjourned in Contemplation of Dismissal = 1, All other categories = 0 Convicted, no Jail Sentence = 1, All other categories = 0 Convicted, with Jail Sentence = 1, All other categories = 0

Convicted, no Jail Sentence Convicted, with Jail Sentence CASE OUTCOME (Bronx Model, Table 5-2) Reference Category: Declined for Prosecution Dismissed Adjourned in Contemplation of Dismissal

Declined for Prosecution: Reference Category Dismissed = 1, All other categories = 0 Adjourned in Contemplation of Dismissal = 1, All other categories = 0 Convicted, no Jail Sentence = 1, All other categories = 0 Convicted, with Jail Sentence = 1, All other categories = 0

Convicted, no Jail Sentence Convicted, with Jail Sentence CONTROL VARIABLES ARRAIGNMENT CHARGE PENAL LAW ARTICLE: Reference Category: Assault (PL 120) Criminal Contempt (PL 215) Harassment (PL 240) Crimes Against Children (PL 260) Other

Assault: Reference Category Criminal Contempt = 1, All other categories = 0 Harassment = 1, All other categories = 0 Crimes Against Children = 1, All other categories = 0 Other = 1, All other categories = 0

BOROUGH Reference Category: Bronx Brooklyn Manhattan Queens Staten Island

Bronx: Reference Category Brooklyn = 1, All other categories = 0 Manhattan = 1, All other categories = 0 Queens = 1, All other categories = 0 Staten Island = 1, All other categories = 0

DEFENDANT'S DEMOGRAPHIC CHARACTERISTICS SEX (Female)

Female = 1, Male = 0

ETHNICITY: Reference Category: Black White Hispanic Other

Black: Reference Category White = 1, All other categories = 0 Hispanic = 1, All other categories = 0 Other = 1, All other categories = 0

AGE: Reference Category: Age 16-20 Age 21-29 Age 30-39 Age 40 and over

Age 16-20: Reference Category Age 21-29 = 1, All other categories = 0 Age 30-39 = 1, All other categories = 0 Age 40 and over = 1, All other categories = 0

Table Continues on Next Page

70

APPENDIX C: CODING OF VARIABLES FOR REGRESSION MODELS (continued)

1

VARIABLES

CODING SCHEME

DEFENDANT-VICTIM RELATIONSHIP Reference Category: Married Boyfriend-Girlfriend Common-Law Marriage Other Relationship Missing

Married: Reference Category Boyfriend-Girlfriend = 1, All other categories = 0 Common-Law Marriage = 1, All other categories = 0 Other Relationship = 1, All other categories = 0 Missing = 1, All other categories = 0

DEFENDANT'S CRIMINAL HISTORY ANY PRIOR ARRESTS

Any Prior Arrests = 1, All other categories = 0

NUMBER OF PRIOR MISDEMEANOR CONVICTIONS

Number of convictions, ranges from 0 to 67

NUMBER OF PRIOR FELONY CONVICTIONS

Number of convictions, ranges from 0 to 7

ANY ARRESTS FOR A DV OFFENSE PRIOR TO CASE DISPOSITION

Any Arrests for a DV Offense Prior To Case Disposition = 1, All other categories = 0

ANY ARRESTS FOR A NON-DV OFFENSE PRIOR TO CASE DISPOSITION

Any Arrests for a Non-DV Offense Prior To Case Disposition = 1, All other categories = 0

DEFENDANT'S COMMUNITY TIES UNEMPLOYED

Unemployed = 1, All other categories = 0

AT CURRENT ADDRESS 1 YEAR OR LESS

At Current Address 1 Year or Less = 1, All other categories = 0

LIVES WITH SOMEONE

Lives with Someone = 1, All other categories = 0

DOES NOT EXPECT ANYONE AT ARRAIGNMENT

Does Not Expect Anyone At Arraignment = 1, All other categories = 0

HAS NO TELEPHONE

Has No Telephone = 1, All other categories = 0

LIVES OUTSIDE NYC AREA

Lives Outside NYC Area = 1, All other categories = 0

ARREST AND ARRAIGNMENT CHARGE CHARACTERISTICS NUMBER OF ARREST CHARGES

Number of charges, ranges from 1 to 4

ARRAIGNMENT CHARGE IS A FELONY

Charged as a Felony = 1, All other categories = 0

CHANGE IN CHARGE SEVERITY FROM ARREST TO ARRAIGNMENT: Reference Category: No Change Charge Severity Reduced from Arrest to Arraignment Charge Severity Increased from Arrest to Arraignment

No Change: Reference Category Charge Severity Reduced = 1, All other categories = 0 Charge Severity Increased = 1, All other categories = 0

CASE PROCESSING CHARACTERISTICS DEFENDANT EVER RELEASED

Defendant ever released = 1, Defendant never released or case was disposed at arraignment = 0

NUMBER OF WEEKS FROM ARRAIGNMENT TO DISPOSITION

Number of weeks, ranges from 0 to 56

NOTE 1

See text for a description of the variables in the models.