Likelihood Pretrial Re-Arrest 06

CJA

NEW YORK CITY CRIMINAL JUSTICE AGENCY, INC.

Jerome E. McElroy Executive Director

PREDICTING THE LIKELIHOOD OF PRETRIAL RE-ARREST FOR VIOLENT FELONY OFFENSES AND EXAMINING THE RISK OF PRETRIAL FAILURE AMONG NEW YORK CITY DEFENDANTS: AN ANALYSIS OF THE 2001 DATASET

Qudsia Siddiqi, Ph.D. Project Director

FINAL REPORT

November 2006

52 Duane Street, New York, NY 10007

(646) 213-2500

PREDICTING THE LIKELIHOOD OF PRETRIAL RE-ARREST FOR VIOLENT FELONY OFFENSES AND EXAMINING THE RISK OF PRETRIAL FAILURE AMONG NEW YORK CITY DEFENDANTS: AN ANALYSIS OF THE 2001 DATASET

Qudsia Siddiqi, Ph.D. Project Director

Research Assistance Justin P. Bernstein Senior Research Assistant Raymond P. Caligiure Graphics and Production Specialist

Information Systems Programming Barbara Geller Diaz Associate Director, Information Systems Wayne Nehwadowich Senior Programmer/Analyst

Administrative Support Annie Su Administrative Associate

November 2006

This report can be downloaded from http://www.cjareports.org Š 2006 NYC Criminal Justice Agency, Inc.

When citing this report, please include the following elements, adapted to your citation style: Siddiqi, Qudsia. 2006. Predicting the Likelihood of Pretrial Re-Arrest for Violent Felony Offenses and Examining the Risk of Pretrial Failure Among New York City Defendants: An Analysis of 2001 Dataset. New York: New York City Criminal Justice Agency, Inc.

ACKNOWLEDGEMENTS The author would like to thank Jerome E. McElroy, Executive Director of CJA, and Dr. Richard R. Peterson, Director of Research at CJA, for their timely review and suggestions on the final draft. The author extends special thanks to Barbara Geller Diaz, Associate Director, Information Systems, who did the programming to extract the re-arrest data from the CJA database.

The author appreciates the assistance of Wayne Nehwadowich for extracting the First Quarter 2001 Dataset, used in this report. The author extends special thanks to Justin P. Bernstein, who worked tirelessly on his contributions to the report, and Raymond P. Caligiure, for his able assistance in constructing and refining the tables for the report. Thanks to Annie Su for her administrative support.

Finally, the author would like to thank the New York State Division of Criminal Justice Services for providing supplemental criminal history data.

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

TABLE OF CONTENTS

LIST OF TABLES.............................................................................................................................ii INTRODUCTION .............................................................................................................................1 SECTION ONE: REVIEW OF LITERATURE ................................................................................4 A. Literature Predicting Re-Offending For Violent Offenses ....................................................4 B. Literature on Pretrial Failure..................................................................................................7 C. Factors Predicting Pretrial Failure to Appear ........................................................................16 SECTION TWO: METHODOLOGY ...............................................................................................28 A. Sampling and Data Sources ..................................................................................................28 B. Dependent Variables..............................................................................................................33 C. Independent Variables ...........................................................................................................36 D. Statistical Methods.................................................................................................................37 SECTION THREE: PHASE I ANALYSIS - VIOLENT FELONY RE-ARRESTS ........................40 A. Sample Characteristics...........................................................................................................40 B. Charge Related Information for Defendants Re-Arrested for Violent Felony Offenses .......49 C. Multivariate Analysis of Re-Arrest for Violent Felony Offenses..........................................52 D. Did the Same Factors Predict Pretrial Re-Arrest in General and Pretrial Re-Arrest for a Violent Felony Offense? .....................................................................................................55 E. Summary and Discussion.......................................................................................................59 SECTION FOUR: PHASE II ANALYSIS - PRETRIAL FAILURE ...............................................62 A. Sample Characteristics...........................................................................................................62 B. Multivariate Analysis of Pretrial Failure ...............................................................................68 C. Constructing a Point Scale .....................................................................................................71 D. Comparing the Point Scale Predicting Pretrial Failure with the Point Scale Predicting Pretrial FTA..........................................................................................................76 E. Distribution of the Point Scale Scores by Pretrial Failure .....................................................79 F. Developing a Risk Classification System for Pretrial Failure ...............................................81 G. Comparing the Combined with the ROR Risk Classification System...................................85 SECTION IV: SUMMARY AND CONCLUSIONS ........................................................................93 REFERENCES ..................................................................................................................................98 APPENDIX A....................................................................................................................................103 APPENDIX B ....................................................................................................................................104 APPENDIX C ....................................................................................................................................106 APPENDIX D....................................................................................................................................110

-i-

LIST OF TABLES Table 1:

Arraignment Outcome First Quarter 2001 Dataset, Defendant-based...............................................................30

Table 2:

Released Status at Arraignment First Quarter 2001 Dataset, Defendant-based...............................................................31

Table 3:

Distribution of Pretrial Re-Arrest in the 2001 At-Risk Sample First Quarter 2001 Dataset, At-Risk Sample ................................................................32

Table 4:

Re-Arrest Charge in the 2001 Re-Arrest Sample First Quarter 2001 Dataset, Re-Arrest Sample .............................................................34

Table 5:

Distribution of Pretrial Failure 2001 Re-Arrest Sample First Quarter 2001 Dataset, Re-Arrest Sample .............................................................35

Table 6:

Demographic Attributes and Case-Processing Characteristics for Defendants Re-Arrested Pretrial First Quarter 2001 Dataset, Re-Arrest Sample .............................................................41

Table 7:

Community Ties Items for Defendants Re-Arrested Pretrial First Quarter 2001 Dataset, Re-Arrest Sample .............................................................44

Table 8:

Criminal History for Defendants Re-Arrested Pretrial First Quarter 2001 Dataset, Re-Arrest Sample .............................................................45

Table 9:

Top Arrest Charge at Initial Arrest for Defendants Re-Arrested Pretrial First Quarter 2001 Dataset, Re-Arrest Sample .............................................................46

Table 10:

Top Re-Arrest Charge for Defendants Re-Arrested Pretrial First Quarter 2001 Dataset, Re-Arrest Sample .............................................................48

Table 11:

Top Charge at Initial Arrest for Defendants Re-Arrested for Violent Felony Offenses First Quarter 2001 Dataset, Re-Arrest Sample, Violent Felony, Re-Arrest Subsample ..........................................................................50

Table 12:

Top Re-Arrest Charge for Defendants Re-Arrested for Violent Felony Offenses First Quarter 2001 Dataset, Re-Arrest Sample, Violent Felony Re-Arrest Subsample ...........................................................................51

Table 13:

Logistic Regression Model Predicting Violent Felony Re-Arrests First Quarter 2001 Dataset, Re-Arrest Sample .............................................................54

Table 14:

Multiple Logistic Regression Model Predicting Pretrial Re-Arrest First Quarter 2001 Dataset, At-Risk Sample ................................................................56

Table 15:

Demographic and Case Processing Characteristics First Quarter 2001 Dataset, At-Risk Sample ................................................................63 -ii-

Table 16:

Community Ties First Quarter 2001 Dataset, At-Risk Sample ................................................................64

Table 17:

Criminal History First Quarter 2001 Dataset, At-Risk Sample ................................................................66

Table 18:

Top Charge at Initial Arrest First Quarter 2001 Dataset, At-Risk Sample ................................................................67

Table 19:

Multiple Logistic Regression Model Predicting Pretrial Failure First Quarter 2001 Dataset, At-Risk Sample ................................................................69

Table 20:

Multiple Logistic Regression Model Used to Develop a Point Scale for Pretrial Failure First Quarter 2001 Dataset At-Risk Sample .................................................................73

Table 21:

Point Scale Predicting Pretrial Failure First Quarter 2001 Dataset, At-Risk Sample ................................................................75

Table 22:

A Comparison of the Point Scale Predicting Pretrial Failure with the Point Scale Predicting to Appear First Quarter 2001 Dataset, At-Risk Sample ................................................................77

Table 23:

Distribution of Point Scale Scores by Pretrial Failure ..................................................80

Table 23A: CJA’s New ROR Recommendation System.................................................................83 Table 24:

Combined Risk Classification System by Pretrial Failure First Quarter 2001 Dataset, At-Risk Sample ................................................................84

Table 25:

New ROR Recommendation System by Pretrial FTA First Quarter 2001 Dataset, At-Risk Sample ................................................................86

Table 26:

Risk Classification Systems by Pretrial FTA First Quarter 2001 Dataset, At-Risk Sample ................................................................89

Table 27:

Risk Classification Systems by Pretrial Re-Arrest First Quarter 2001 Dataset, At-Risk Sample ................................................................90

Table 28:

Risk Classification Systems by Pretrial Failure First Quarter 2001 Dataset, At-Risk Sample ................................................................91

-iii-

PREDICTING THE LIKELIHOOD OF PRETRIAL RE-ARREST FOR VIOLENT FELONY OFFENSES AND EXAMINING THE RISK OF PRETRIAL FAILURE AMONG NEW YORK CITY DEFENDANTS: AN ANALYSIS OF THE 2001 DATASET

INTRODUCTION Pretrial release programs, as an alternative to the traditional bail system, have their roots in the bail reform movement of the early 1960s. The "Manhattan Bail Project," set up by the Vera Foundation in October 1961, was among the first experimental pretrial-release projects in the country. In an attempt to assist indigent defendants by establishing an alternative to the moneybail system, the project used a community-ties model to determine defendant eligibility for pretrial release on their own recognizance (ROR). Upon application, defendants who were released on recognizance were found to have low failure-to-appear (FTA) rates. Consequently, the Manhattan Bail Project was considered a great success (Ares et al. 1963). By 1965, 48 jurisdictions had instituted pretrial-release programs modeled after the Vera Project (Thomas 1976).

As more jurisdictions began to release defendants on their own recognizance, concerns about public safety began to grow. It was generally believed that the bail practices were putting the public’s safety at-risk by releasing dangerous defendants back to the streets. In response to those concerns, in 1971 the first federal preventive detention statute was passed. The statute led the District of Columbia's Pretrial Services Agency to adopt a new policy, which allowed for consideration of public safety risk, as well as risk of flight. Currently, almost all of the states and the federal system consider public safety when making pretrial release decisions, and permit more restrictive pretrial release conditions, including preventive detention, where the risk is seen as great. However, the New York State Criminal Procedure Law (CPL) does not permit the explicit consideration of “dangerousness” in the setting of pretrial release conditions. In New York City, the pretrial release recommendations are based on a "risk of flight" model, and are made by the New York City Criminal Justice Agency, Inc. (CJA).

The concern for the effects of recidivism upon public safety has also resulted in many studies of recidivism among convicted offenders. However, knowledge about pretrial recidivism is quite limited, and the findings are either inconsistent or not clear. It is also not known whether the findings from the general literature containing post-conviction recidivism would hold for those released pending disposition of the charge. To shed light on those issues, CJA began a series of research projects. Using a random sample of defendants arrested in 1989, the Agency developed statistical models that would identify significant predictors of pretrial rearrest. The analysis was then repeated on more recent samples of defendants, which included data collected in the third quarter of 1998 and first quarter of 2001 (Siddiqi March 2003; June 2003).

Recently, the Re-arrest Project shifted its focus to two other research issues. The first issue focused on pretrial re-arrests made for violent felony offenses. The second issue was related to the prediction of pretrial failure, which included both pretrial FTA and re-arrest.

Beginning with the first issue, defendants re-arrested for violent offenses present a serious threat to the safety of the community. However, our knowledge about re-offending for such offenses among adult defendants is quite limited. The existing research mostly concentrates on recidivism in general. Furthermore, the focus is generally on post-conviction recidivism.

With respect to the second issue, several studies have examined the likelihood of combined failure while on pretrial release (Cuvelier and Potts 1993; Goldkamp et al. 1981; VanNostrand 2003). However, it is not known whether the findings from those studies would apply to the defendant population in New York City. Therefore, although the New York statute does not permit consideration of public safety in making pretrial release decisions, the issue is worthy of inquiry. The research presented in this report explores both issues. The analysis is based on data collected from the first quarter of 2001 and conducted in two phases. In Phase I, we examine the likelihood of pretrial re-arrest for violent felony offenses. Our research focuses on the following questions.

-2-

1. What proportion of defendants is re-arrested for a violent felony offense before adjudication of the initial charge?

2. Do the defendants re-arrested for a violent felony offense initially commit the same offense?

3. What factors predict pretrial re-arrest for a violent felony offense?

4. Do the same factors predict pretrial re-arrest in general and pretrial re-arrest for a violent felony offense?

In the second analytical phase, we assess the risk of pretrial failure (FTA and re-arrest).

Our research addresses the following questions.

1. What factors predict pretrial failure?

2. Does the point scale predicting the risk of pretrial failure differ from the point scale predicting the risk of pretrial FTA? 3. How does the combined risk classification system compare with CJA’s new ROR recommendation system?

The report is organized into five sections. The first section summarizes literature on pretrial re-arrest for violent felony offenses and pretrial failure. The second section describes the research methodology. The third and fourth sections present research findings. The last section summarizes the research findings and offers conclusions.

-3-

Section One Review of Literature This section summarizes the review of literature on recidivism for violent offenses, and pretrial failure. The review pertaining to pretrial failure summarizes findings from prior research on pretrial FTA, pretrial re-arrest, and pretrial FTA and re-arrest. The findings are presented below.

A. Literature Predicting Re-Offending For Violent Offenses A number of studies have been conducted to identify predictors which would differentiate violent re-offenders from nonviolent re-offenders. However, most of those studies focus on violence among juvenile and youthful offenders. This is primarily in response to the problem of violence among youth in urban areas. In some jurisdictions, actuarial models are used to assess risk of future violence in making post-conviction release decisions. Research on violent reoffending during the pretrial period among adult defendants is quite limited. Furthermore, we do not know whether the findings from research on youthful offenders and post-conviction recidivism will hold for pretrial recidivism for violent offenses. Nonetheless, findings from the literature on youthful offenders, along with those derived from the examination of adult defendants are presented below.

Predictors of Recidivism for Violent Offenses Offense Type A study conducted by the Bureau of Justice Statistics (1987) on a sample of young parolees (between the ages of 17 and 22), who were released from state prisons in 1978, found similarities between offense type at re-arrest and offense type at incarceration among serious offenders.

In the study on young parolees, offense type at prior arrest was also related to the offense type at re-arrest during the follow up period (Bureau of Justice Statistics 1987). Specifically, defendants with a prior arrest for a violent offense were more likely to be re-arrested for a -4-

similar offense. Another study indicated that a prior arrest record for a violent offense slightly explained future re-arrests for similar offenses (Weiner 1990). In contrast to these findings, Toborg et al. (1984) found that defendants charged with murder or robbery were less likely to be re-arrested for dangerous or violent offenses, whereas defendants charged with larceny or drugs were more likely to be re-arrested for such offenses.1 Criminal History Several researchers have examined the effect of criminal history variables on the likelihood of violent recidivism, and the findings have been consistent. Lattimore et al. (1995) examined recidivism for violent offenses on a sample of male youths paroled from the California Youth Authority between July 1, 1981, and June 30, 1982. The offenders were followed for a period of 3 years. The research revealed that criminal history variables had a significant effect on the likelihood of recidivism for violent offenses. It was also found that violent behavior during institutionalization was related to risk of violent recidivism immediately after release.

Furthermore, parolees with a larger number of prior violent arrests were more likely to be rearrested for the same offense than parolees with fewer prior violent re-arrests. However, the effect disappeared after 18 months of release. By two years after release, parolees with more prior violent re-arrests were less likely to re-offend for such offenses.

With regard to violent re-offending prior to case disposition, Goldkamp et al. (1981) found that pending charges and recent arrests were significantly related to the likelihood of re-arrests for serious offenses. Serious re-arrests in his study included serious personal offenses, robbery, aggravated assault, burglary, and the manufacture or sale of drugs.

Toborg et al. (1984)

indicated that prior convictions were significantly related with the probability of re-arrest for dangerous or violent crimes.

1

Dangerous or violent crimes included robbery, burglary, rape, assault with a dangerous weapon, and sale of narcotics.

-5-

Age Age was found to be a significant predictor of re-arrest for dangerous or violent offenses (Goldkamp et al. 1981; Toborg et al. 1984). For example, Goldkamp et al. (1981) found that defendants who were over 25 years of age at the time of initial arrest were less likely to be rearrested pretrial for serious offenses than defendants who were 25 years old or younger.

Family Background/Employment/Education A few studies have examined the effect of family background on re-offending for violent offenses. Those studies suggested that family violence and parental neglect significantly increased the likelihood of recidivism for violent offenses (Lattimore et al. 1995; Widom 1989).

Employment was found to be a significant predictor of pretrial re-arrest for dangerous or violent offenses (Goldkamp et al. 1981; Toborg et al. 1984). With respect to educational attainment, one study suggested that parolees who quit school were more likely to re-offend for violent offenses (Lattimore 1995).

Other Research Correction Service Canada uses several actuarial risk-assessment instruments to predict general recidivism, and recidivism for violent offenses in making parole release decisions (Glover et al. 2002). An evaluation study found two instruments performing better than the other instruments in identifying high-risk violent re-offenders. Those instruments included the General Statistical Information on Recidivism-Revised (GSIR-R) and the Violence Risk Appraisal Guide (VRAG). The GSIR-R is a 15-item scale used to assess general recidivism in making parole decisions. It contains information on previous criminal justice involvement, age at first admission to the correctional facility, current age, current offense, employment status and number of dependents. The VRAG, originally designed to assess the likelihood of violent recidivism among offenders in a Canadian maximum security forensic psychiatric institution, is based on a sum of 12 items which include information on antisocial personality traits, antisocial lifestyle, alcohol problems, victim injury, nonviolent offense history, and separation from biological parents before age 16.

-6-

Summary To summarize, the findings from our review of prior research on violent re-offenders suggest that criminal history, and prior arrests or convictions for violent offenses, are useful factors in predicting future violence among youthful offenders. In one study, criminal history has also been found to predict pretrial recidivism for violent offenses among adult offenders. Prior research also points out that the significance of some factors may shift with an increase in the follow up time after release. We used those findings to guide the selection of variables for our current analysis of pretrial re-arrest for violent felony offenses.

B. Literature on Pretrial Failure There is an extensive body of literature on pretrial FTA and re-arrest. Most of those studies have examined FTA and re-arrest separately. Nonetheless, some studies have examined factors, which would explain both types of failure (FTA and re-arrest). Our review of literature summarizes the findings by the type of failure under scrutiny.

Factors Predicting Recidivism A number of studies have been conducted to identify factors which are predictive of recidivism. However, most of those studies focus on the likelihood of recidivism after a sentence has been imposed. There are very few studies of recidivism during the pretrial period.

Nonetheless, findings from the general literature, along with those derived from the pretrial examinations, are summarized below.

Before presenting these findings, it is important to note the difficulty involved in trying to make generalized statements. One reason for this is the inconsistent operationalization of the dependent variable, recidivism. Included among the various measures of recidivism are the following: re-arrest (Bureau of Justice Statistics 1987; 1989; 1991; November 1992; 1994; Clarke and Harrison 1992; Goldkamp et al. 1981; Goodman 1992; Jones 1991; Landes 1974; Rhodes 1996; Siddiqi March 2003; June 2003), reconviction (Jones 1991), prison return (Bureau of Justice Statistics 1985), petition to revoke probation (Hepburn and Albonetti 1994), probation -7-

revocation (Hepburn and Albonetti 1994) re-arrest, reconviction, resentence to prison and return to prison with or without a new sentence (Bureau of Justice Statistics 2002).

In addition, different jurisdictions, time periods, and defendant populations have been analyzed. With regard to jurisdictional differences, the focus has often been limited to specific cities (e.g., New York City in Belenko et al. 1994; Siddiqi March 2003; June 2003) or states (e.g., North Carolina in Clarke and Harrison 1992; Kansas in Jones 1991). At the same time, there are some studies that have included several jurisdictions (e.g., 11 states in Bureau of Justice Statistics 1989; 22 states in Bureau of Justice Statistics 1987; 32 counties across 17 states in Bureau of Justice Statistics February 1992).

Differences in the time period under

investigation and defendant population utilized are also apparent. For example, Jones (1991) examined re-arrests and re-conviction among offenders sentenced between July 1983 and June 1984, to community corrections, probation, or prison. In contrast, a study conducted by the Bureau of Justice Statistics (1987) relied upon a sample of young parolees (between the ages of 17 and 22), who were released from state prisons in 1978. As another example, an evaluation of Kansas' Community Corrections Act looked at nonviolent offenders convicted of Class D or E felony offenses, with no more than one prior felony conviction (Jones 1991). Belenko et al.

(1994) used a random sample of arrests made in 1989 to determine the likelihood of recidivism for offenders adjudicated in New York City's drug courts, and for those processed through standard means.

Generalization is further complicated when the point of case processing and the follow-up period are considered. As previously noted, most studies have analyzed recidivism after conviction and sentencing. However, because all convicted defendants do not receive the same sentence, some researchers have examined the likelihood of re-offending after release from prison (e.g., Bureau of Justice Statistics 1989; 2002), while others have looked at those on probation (Bureau of Justice Statistics February 1992). At the same time, examinations exist where both groups of sentenced defendants (i.e., those on probation or parole) were

-8-

simultaneously considered (Clarke and Harrison 1992). Also noteworthy is the relative lack of research conducted during the pretrial period.

Turning to the follow-up period under investigation, defendants have been tracked for as little as a year (Bureau of Justice Statistics 1989) to as long as five (Rhodes 1989) or more years (Bureau of Justice Statistics 1987). It should be noted however, that most research contains follow-up periods falling within that range (e.g., Belenko et al. 1994; Bureau of Justice Statistics 1989: 1991; 2002; Clarke and Harrison 1992; Jones 1991). This is of concern, because longer time periods are likely to show higher rates of recidivism. In comparison, pretrial release periods are not of long duration. Despite those difficulties involved, a discussion of findings from past research follows.

Type of Offense The type of offense committed has been found to influence the likelihood of recidivism.

However, a definitive conclusion cannot be offered because the impact of specific types of offenses varies from one study to the next.

One study found recidivism rates among

probationers to be higher for those charged with drug offenses (Bureau of Justice Statistics February 1992). Other studies regarding the recidivism of probationers and prison releasees disagree with that finding (Bureau of Justice Statistics 1987; 1989; Liberton et al. 1992; Whitehead 1991). For example, in a study conducted on young parolees, property offenders had higher recidivism rates than those paroled for violent or drug offenses (Bureau of Justice Statistics 1987). Similar findings were obtained from a study of prisoners released in 1983 (Bureau of Justice Statistics 1989).

With regard to pretrial recidivism, a pattern similar to that observed among probationers was suggested. The findings from a survey of felony cases filed in February 1988 in the 75 largest United States counties showed that 19% of the felony drug defendants, who were released before the disposition of their case, were re-arrested for a felony during the pretrial period (Bureau of Justice Statistics 1991). Of the other released defendants, about the same percentage of property offenders (18%) were re-arrested pretrial for a felony offense. The felony re-arrest rates for -9-

defendants charged with a violent offense and defendants charged with a public order offense were 16% and 12%, respectively. Similar findings were obtained from another survey of felony cases filed in May 1990 in the 75 largest United States counties (Bureau of Justice Statistics November 1992). To be specific, defendants charged with property and drug offenses in the 1990 sample had similar re-arrest rates, 21% and 20%, respectively, whereas defendants charged with public order offenses had the lowest re-arrest rate.

The above findings from state felony defendants differ from those reported in a study on federal felony defendants who were released in 1990 (Bureau of Justice Statistics 1994). The study showed that of those released pretrial, defendants charged with a violent offense had the highest rate (6%) of pretrial re-arrest for a new offense.2 This was a larger pretrial re-arrest rate than was found for those charged with a drug offense (4%) and those charged with a property offense or a public order offense (2% for each). It should be noted that differences among various categories of defendants were minor. In another study, the probability of pretrial re-arrest was higher among defendants arrested for gambling and prostitution offenses and lower among defendants arrested for serious personal offenses (Goldkamp et al. 1981).

The findings from CJA research on pretrial recidivism among New York City defendants suggested that having a drug charge at initial arrest significantly increased the likelihood of pretrial re-arrest (Siddiqi March 2003; June 2003). In contrast, having a violent offense decreased the probability of re-arrest.

Kirby (1977) in his study on pretrial releasees in

Tennessee found higher re-arrest rates among defendants initially charged with burglary and lower re-arrest rates for defendants charged with assault. Toborg et al. (1984) found certain charges associated with the likelihood of pretrial re-arrest. They included burglary, drugs, possession of the implements of crime, larceny, robbery, stolen property, fraud, prostitution, forgery or automobile theft.

Winterfield et al. (2003) found higher re-arrest rates among

defendants arrested for violent crimes. In their analysis of pretrial failure, Visher and Linster 2

The overall re-arrest rate for a new offense for Federal felony defendants released pretrial in 1990 was much lower than the overall felony re-arrest rate for State felony defendants released in 1988, 3% versus 18% (Bureau of Justice Statistics 1994).

-10-

(1990) found that the relationship between offense type and re-arrest varied as the time at risk increased. They found that in the early stages of pretrial release, relative to a misdemeanor assault charge, having a violent felony offense or an “other� offense was associated with a higher risk of re-arrest. However, the discriminating power of those charges diminished as the time at risk continued. In contrast, the predictive ability of being charged with a drug or larceny offense increased with an increase in time at risk.

The type of offense at the time of re-arrest Another question worthy of inquiry is the type of offense for which one is likely to be rearrested. Do defendants recommit the same types of offenses? Or, alternatively, does the type of re-arrest offense differ from the original?

A study of felons placed on probation from 1986-89 sheds some light on these questions.

The findings indicated that the probationers were most likely to be re-arrested for a similar offense as the one for which probation had been imposed (Bureau of Justice Statistics, February 1992). Although the likelihood of being charged with a similar offense was evident, the majority of the probationers were not re-arrested for the same offense for which they were serving probation. Using the defendants charged with murder to illustrate this point, 20.8% of such offenders were re-arrested -- 4.9% for a new murder and 15.9% for a different offense.

A similar conclusion was reached by a Bureau of Justice Statistics' (1987) investigation of young parolees released in 1978. The analysis showed that 41% of paroled burglars were rearrested for burglary within six years of release. Despite the tendency to be re-arrested for the same type of offense, the parolees were often arrested for other crimes as well. This was consistent with what was noted for probationers. As an example, while 59% of the paroled property offenders were re-arrested for a property offense, 35% were re-arrested for a violent offense (Bureau of Justice Statistics, 1987).

Similar findings were reported from a study of

prisoners released in 11 states in 1983 (Bureau of Justice Statistics, 1989). For every type of offender, the likelihood of re-arrest for a similar crime was high. However, those released from prison were also re-arrested for other offenses.

-11-

The findings from another study which focused both on probationers and parolees indicated that the likelihood of re-arrest for a similar offense held only for certain categories of defendants; offenders with felony or misdemeanor drug charges had an increased probability of re-arrest for another drug offense (Clarke and Harrison 1992). This was also true when the focus was shifted to probationers only. The study also suggested that probationers with a violent misdemeanor charge had a greater likelihood of re-arrest for a violent offense. Among other probationers, however, the pattern was less consistent. To elaborate, probationers charged with property felony or “other” (charges not categorized as violent, property or drug offenses) felony offenses had a greater probability of re-arrest for a violent offense. In comparison, probationers charged for “other” misdemeanor offenses had an increased probability of re-arrest for drug charges. Thus, specialization may be reserved for a specific category of probationers -- in this case, drug offenders.

With regard to pretrial behavior, one study focused on felony defendants who were released pretrial from the 75 most populous counties in the United States (Bureau of Justice Statistics 1991). The results showed that defendants were often re-arrested for the same type of felony as the one for which they were initially brought into the system. For example, among drug defendants who were re-arrested for a felony, 57% were re-arrested for a drug offense before their cases were disposed. Similar findings were obtained from CJA research on pretrial re-arrest where more than half of the defendants initially arrested for a drug offense and who were rearrested pretrial were re-arrested for the same type of offense (Siddiqi March 2003; June 2003).

The research also showed some overlap for other offense types but the findings were not consistent across different samples. For example, in the 1989 sample, one-half of the re-arrested defendants who were initially arrested for property offenses were re-arrested for the same offense type. The proportion went down in the 1998 and 2001 samples where one-third of the property offenders were re-arrested for the same offense. The research showed that in addition to re-committing the same offense, defendants in each of the offense types re-committed entirely new offenses.

-12-

Altogether, the above pieces of research present a picture that suggests that defendants are likely to recommit the same offenses as those for which they were originally charged. At the same time, this does not entail offense specialization, because many commit entirely new offenses.

Prior Record A number of studies have examined the effect of prior record on recidivism. The findings suggest that defendants with prior records have consistently been found to be more likely to recidivate than those not having a criminal history (Bureau of Justice Statistics 1985; 1991). The literature also suggests that the likelihood of recidivism increases as one's prior record grows more extensive (Bureau of Justice Statistics 1987; Clarke and Harrison 1992; Hepburn and Albonetti 1994; Whitehead 1991). The effects of specific types of prior arrests were also noted (Bureau of Justice Statistics 1989). The results showed that those with prior violent arrests were more likely to be re-arrested than those not having such arrests. Re-arrest rates were also found to be higher among those released with prior drug arrests than those without a drug arrest.

With regard to re-arrest before case disposition, an analysis of federal felony defendants released pretrial in 1990 found criminal history to be among the most determinant of the factors examined (Bureau of Justice Statistics 1994). However, variation was not large. Other studies also found criminal history factors to be related with pretrial re-arrest (Goldkamp et al. 1981; Landes 1974; Rhodes et al. 1996; Roth and Wice 1980; Toborg et al. 1984; Winterfield et al.

2003). The findings from CJA research showed that defendants with a criminal history were more likely to be re-arrested pretrial than defendants without a criminal history (Siddiqi March 2003; June 2003). Several studies revealed that those with a history of pretrial failure to appear were more likely to be re-arrested than those who made all of their appearances on previous arrests (Goldkamp et al. 1981; Siddiqi March 2003; June 2003).

-13-

Type of Release Only a few studies explicitly examined the effect of type of release (bail vs. ROR). Reaves (1994) found little difference in rates of pretrial re-arrest between defendants released on bail and those released on recognizance. CJA research showed that defendants released on ROR were more likely to be re-arrested during the pretrial period than those released on bail in the 1989 and 1998 samples (Siddiqi March 2003). However, a subsequent analysis found that type of release had no effect on the likelihood of pretrial re-arrest in the 2001 sample (Siddiqi June 2003).

Drug Use Some studies showed higher re-arrest rates among defendants who were tested positive for drug use (Bureau of Justice Statistics 1992; Visher and Linster 1990, Winterfield et al. 2003).

Socio-Demographic Variables AGE AT PROBATION, PAROLE, PRISON, RELEASE, OR ARREST The age of the defendant has consistently been found to be related to the likelihood of recidivism; the probability of re-arrest or prison return declined as age increased (Bureau of Justice Statistics 1985; 1987; 1989; 1994; 2002; Clarke and Harrison 1992). Similar findings were obtained for recidivism during the pretrial period (Belenko et al. 1994; Goldkamp et al.

1981; Goodman 1992; Siddiqi March 2003; June 2003; Toborg et al. 1984; Winterfield et al.

2003).

RACE/ETHNICITY Findings pertaining to the effect of race/ethnicity on recidivism have been inconsistent.

Several studies indicated that black defendants had higher re-arrest rates than white defendants (Bureau of Justice Statistics 1987; 1989; Clarke and Harrison 1992; Hepburn and Albonetti 1994; Siddiqi March 2003; June 2003; Toborg et al. 1984; Whitehead 1991). In some studies, the rate of re-arrest among Hispanics was also higher than that of whites, although less than that observed among blacks (Bureau of Justice Statistics 1987; Siddiqi June 2003). Other studies

-14-

failed to find a relationship between race and recidivism (Bureau of Justice Statistics 1985; Jones 1991).

GENDER Unlike race/ethnicity, findings regarding the effect of gender on recidivism have been consistent; men were more likely to be re-arrested than women (Clarke and Harrison 1992; Bureau of Justice Statistics 1987; 1989). Similar findings were obtained from CJA research on pretrial recidivism (Siddiqi March 2003; June 2003). However, in another study of pretrial recidivism, a difference of only one percentage point was noted between the sexes, with 3% of the sampled males having a re-arrest (Bureau of Justice Statistics 1994).

EMPLOYMENT AND EDUCATION The effects of employment and educational attainment on recidivism are not clear, because only a handful of researchers have examined these variables. Beginning with a defendant's level of educational attainment, a study on young parolees indicated that individuals who had graduated from high school were less likely to be re-arrested than those who did not complete high school (Bureau of Justice Statistics 1987). This relationship was also observed in a separate investigation of prisoners of all ages, who were released in 11 states in 1983 (Bureau of Justice Statistics 1989). However, the level of educational attainment did not predict the likelihood that one would recidivate in a study of re-offending among persons convicted of D and E felonies (Jones 1991), or in another study of prison return (Bureau of Justice Statistics 1985).

With regard to the influence of employment, some researchers have found that not being employed at the time of the offense increased the likelihood of recidivism (Bureau of Justice Statistics 1985; Clarke and Harrison 1992; Liberton et al. 1992; Toborg et al. 1984; Whitehead 1991). In contrast, Jones (1991) did not find a relationship between employment and recidivism.

With respect to the pretrial period, employment decreased the likelihood of re-arrest (Goldkamp et al. 1981; Goodman 1992; Roth and Wice 1980). In another study, the probability of pretrial re-arrest was lower among defendants who were employed, in school, or in a training program full time at the time of initial arrest (Siddiqi March 2003; June 2003).

-15-

To summarize the findings, the review of the literature suggested that certain variables were found to be more consistent predictors of recidivism than others. They included a defendant’s criminal history, gender, and age. The variables with inconsistent effects included charge at initial arrest, ethnicity, employment and educational attainment.

C. Factors Predicting Pretrial Failure to Appear A large number of studies have examined pretrial failure to appear among defendants who were released prior to trial. However, it is difficult to compare these studies and formulate conclusions for a number of reasons. First, the studies differ in their definition and calculation of failure-to-appear rates. Some of these studies use appearance-based FTA rates (Eskridge 1979), while others use defendant-based FTA rates (Center for Governmental Research, Inc. 1983; Schaffer 1970; Siddiqi 1999; Thomas 1976; Toborg 1981).3 Beyond the differences in the use of defendant- and appearance-based measures, some researchers have also distinguished between willful and nonwillful failure-to-appear rates (Goldkamp et al. 1981; Schaffer 1970; Toborg 1981), while others have not (Eskridge 1979). In general, if a defendant intentionally fails to appear for a scheduled court hearing, without legal justification for that absence, it is considered a willful FTA. However, if the defendant has a legal or verifiable reason for failing to appear, such as hospitalization or incarceration, then his or her absence in court is classified as a nonwillful FTA. If every missed court appearance is considered a failure, regardless of whether the defendant's absence in court is willful or nonwillful, then the FTA rate would be greater than if only willful nonappearances are considered.

In addition to the above inconsistencies in the manner in which FTA (the dependent variable) is measured, there are also differences regarding which factors predicting FTA (the independent 3

Appearance-based FTA rates take into consideration all missed court appearances. They are usually derived by dividing the number of missed appearances (i.e., the number of appearances in which a bench warrant is issued) by the total number of scheduled appearances. In contrast, defendant-based FTA rates are generally calculated by dividing the number of defendants who failed to appear at least once for a scheduled court date by the total number of defendants included in the sample.

-16-

variables) are utilized. For instance, some researchers include a wide variety of factors in their studies (Austin et al. 1985; Goodman 1992; Petee 1994; Siddiqi 1999; 2000), while others examine only a few variables (Britt et al. 1992; Kirby 1977; Wheeler and Wheeler 1981).

Additionally, each jurisdiction has a unique population of defendants. Thus, factors which may successfully predict FTA in one jurisdiction may not be successful in another due to differences in each jurisdiction's defendant population. For these reasons, it is not surprising that findings are inconsistent regarding the factors that significantly predict FTA. Beyond methodological inconsistencies, there is another reason to cautiously interpret the findings from previous studies. Most of the research on pretrial release was conducted in the late 1970s. As a result, the findings reported may not be generalizable to more recent defendant populations. Given the problems mentioned here, the discussion of the literature that follows requires careful interpretation.

Type of Release When comparing the FTA rates for defendants released on their own recognizance with those released on monetary bail, many studies have drawn different conclusions. Furthermore, the findings vary in accordance with the city or pretrial program under examination. Several studies have reported higher FTA rates for defendants released on their own recognizance as compared with those released on money bail (Obert, 1973; Alameda Regional Criminal Justice Planning Board, 1976). These studies can be contrasted with evaluations of pretrial release programs in Columbus, Ohio and Des Moines, Iowa, which showed that individuals released on their own recognizance had lower FTA rates than those released on money bail (Bell et al., 1974; National Institute of Law Enforcement and Criminal Justice, 1976). Still, other studies (Conklin and Meagher 1973; Helland and Tabbarok 2004; Maxwell 1999; U.S. Department of Justice 1991; Wheeler and Wheeler 1981) have concluded that FTA rates do not significantly differ for the two types of release. For example, according to findings reported by the U.S. Department of Justice (1991), defendants released on a full cash or deposit bond failed to appear at the same rate as those released on their own recognizance or on citation.

-17-

Consistent with these findings, Thomas (1976) indicated that, in most of the 20 American cities considered in his study, no significant difference existed in the FTA rates for the two groups. However, he noted that in Chicago and San Francisco, the FTA rates for defendants released on their own recognizance were lower than those found for bailed defendants, regardless of whether the charge was a misdemeanor or felony. On the other hand, Thomas found that in Los Angeles, Boston, and Detroit, the FTA rates were substantially higher for cases in which defendants were released on their own recognizance. A 1973 survey conducted by the United States Office of Economic Opportunity yielded similar results (U.S Office of Economic Opportunity 1973). The study examined FTA rates for 16 different pretrial programs across the nation. Specifically addressed was whether the FTA rates differed for individuals released on their own recognizance versus those released on monetary bail. Each of the pretrial programs used a similar formula to calculate FTA rates, thus allowing for comparisons to be made from one jurisdiction to the next. Although the survey results showed that the type of release was related to the likelihood of failure to appear, the direction of the relationship varied with the program. For ten programs, the FTA rate was lower for defendants released on their own recognizance than for defendants who were released on bail. For five other programs, the reverse was true, and for one program the FTA rate for each group of defendants was similar.

Similar findings were obtained from another research on felony defendants in state courts (Kennedy and Henry 1996).

CJA research has found that defendants released on recognizance in New York City were more likely to fail to appear than defendants released on bail (Siddiqi 1999).

Community Ties Studies that have examined the relationship between a defendant's community ties and pretrial failure are inconsistent with regard to the significance of the relationship. A few studies indicate that strong ties to the community decrease the probability of failure. For example, FTA rates have been found to be significantly lower among those with a stable residence and close family ties (Belenko et al. 1992; Venezia 1973). Lower FTA rates have also been associated -18-

with a defendant’s access to a telephone (Wilson 1975). Research from CJA has consistently shown that having a telephone, residing at a New York City area address, and expecting someone at arraignment were associated with lower FTA rates (Siddiqi 1999; 2000; 2002).

Toborg (1981) also noted that defendants who failed to appear for scheduled court hearings had weaker community ties.

Other researchers have failed to find significant relationships between community ties and FTA (Center for Governmental Research, Inc. 1983; Eskridge 1979; Feeley and McNaughton 1974; Roth and Wice 1980). Roth and Wice, for instance, found that defendants who had a local residence were neither more nor less likely to FTA than those who did not. It is important to note, however, that due to the limitation of the data, the Roth and Wice study did not examine such factors as a defendant's length of time in the community, length of time at current address, or family ties.

The effects of employment and education on pretrial failure have also been assessed, with mixed results. Cuvelier and Potts (1993), Venezia (1973), and Wilson (1975) found a significant relationship between employment and lower rates of failure. Similarly, unemployment was a significant predictor of failure to appear in several other studies (Boudouris et al. 1977; Lazarsfeld 1974; Rhodes et al. 1996). In the 1983 study of pretrial release services in New York State, the Center for Governmental Research, Inc. (1983) concluded that length of employment and level of educational attainment were inversely related to FTA. More recently, several studies have found lower FTA rates among those who were employed, in school, or receiving government benefits (Goodman 1992). Similar findings were obtained in more recent research

by CJA (Siddiqi 1999; 2000; 2002).

Although a number of studies have indicated that various socioeconomic factors affect failure, others have failed to find a relationship between these variables (Feeley and McNaughton 1974; Roth and Wice 1980). For example, Roth and Wice showed that income was unrelated to FTA. Demographic Variables -19-

Some studies have also included demographic variables, other than those related to income and employment, in order to ascertain their relationship to FTA. As with community ties and socioeconomic variables, some researchers found that these variables affected FTA, while others did not. For example, Roth and Wice (1980), in their study of pretrial releasees in the District of Columbia, noted that age and race were unrelated to FTA. Maxwell (1999) also did not find a relationship between age and FTA. In contrast, several studies have found that age and race were both predictive of pretrial failure. Younger defendants were more likely to FTA than older defendants, and black defendants were more likely to FTA than white defendants (Illinois Criminal Justice Information Authority 1992; Siddiqi 1999). Goldkamp et al. (1981) also found higher FTA rates among younger defendants.

Criminal History A number of researchers have also addressed the question of whether criminal history is related to FTA. It is difficult to compare each of the studies in this area, because the operationalization of prior record varies from one study to the next. Prior arrests, prior felony convictions, prior misdemeanor convictions, past bench warrants, length of prior record and history of pretrial failure have all been used as indicators of criminal history.

Several researchers have noted that defendants with criminal histories are more likely to fail pretrial (Austin et al. 1985; Cuvelier and Potts 1993). With regard to prior arrests, Toborg (1981) reported that defendants who failed had more serious arrest histories.4 This finding is consistent with research conducted by the Center for Governmental Research, Inc. (1983), which reported higher FTA rates among defendants with prior violent felony arrests and prior nonfelony arrests.

Similarly, in studies of pretrial release in Philadelphia (Goldkamp et al. 1981) and in Cook County, Illinois (Illinois Criminal Justice Information Authority 1992), defendants with prior arrests also had higher failure rates. Prior convictions were also found to be associated with pretrial FTA (Bureau of Justice Statistics 1991; Goodmann 1992; Maxwell 1999; Winterfield et

4

“More serious" is used by Toborg to indicate a greater average number of prior arrests and convictions.

-20-

al. 2003). The Bureau of Justice Statistics (1991) reported that as the number of prior felony convictions increased, so did the rate of FTA. Eskridge however, has cautioned that prior misdemeanor convictions and prior felony convictions may have the opposite effect upon FTA. He reported that defendants with prior felony personal convictions had lower failure-to-appear rates than those with prior misdemeanor property convictions (1979). In addition to prior arrests and convictions, pending cases were also found to increase the likelihood of FTA (Siddiqi 1999; 2000; 2002). Prior research also shows that having a history of pretrial failure increases the likelihood of FTA (Belenko et al. 1992; Center for Governmental Research, Inc. 1983; Cuvelier and Potts 1993; Eskridge 1979; Goodman 1992 Siddiqi 1999; 2000; 2002; Straus and Golbin 1980; Wilson 1979; Winterfield et al. 2003). In contrast to these findings, Maxwell (1999) found no effect of prior failure to appear on current FTA. While a number of studies have indicated that criminal history is related to FTA, it is important to add that other analyses have failed to find a relationship (Landes 1974).

Charge Severity and Type Several researchers have indicated that an individual's probability of failure decreases as the severity of the charge increases (Center for Governmental Research, Inc. 1983; Eskridge 1979; Rhodes et al. 1996; Straus and Golbin 1980). For example, the findings of the statewide study conducted by the Center of Governmental Research, Inc. (1983) indicated that defendants charged with violent felony offenses were less likely to miss court appearances. Siddiqi (1999) found that the likelihood of FTA was higher among defendants arrested for A misdemeanors.

Other studies have failed to find a statistically significant relationship between charge severity and FTA (Landes 1974).

The arrest charge type has also been considered as a predictor of FTA (Austin et al. 1985; Bureau of Justice Statistics 1991; Goldkamp et al. 1981; Goodman 1992; Kirby 1977; Mawell 1999; Siddiqi 1999; 2000). However, findings regarding the effect of specific types are not consistent.

For example, few studies indicated that defendants charged with burglary and

property crimes had higher FTA rates than defendants charged with other offenses (Maxwell -21-

1999; Kirby 1977). However, Kirby noted that his findings should be considered cautiously because of the small number of cases included in the study. In contrast, another study showed that the FTA rate among defendants charged with drug offenses was double the rate found among those charged with public-order offenses (Bureaue of Justice Statistics 1991). Goldkamp et al. (1981) found lower FTA rates associated with gambling and serious personal offenses.

CJA research showed that defendants arrested for property or drug offenses were more likely to FTA than those with the average across the charge variable (Siddiqi 1999; 2000). The research also showed that defendants arrested for gambling or driving while under the influence of alcohol or drugs had lower FTA rates. The findings further revealed that defendants arrested for violent offenses significantly predicted FTA in one study only where they were more likely to fail than the overall effect of that variable (Siddiqi 1999).

Time at Risk In studying pretrial release, another factor upon which many studies have focused is the relationship between FTA and the length of time defendants were on release. The period of time during which a defendant is released from custody has also been referred to as time at risk, because the defendant has an opportunity to fail to appear. Problems arise in attempting to compare the findings of various studies since time at risk is not uniformly defined. In some studies, it has been calculated as the number of days the defendant is free from pretrial detention (Center for Governmental Research, Inc. 1983; Bureau of Justice Statistics 1991). Others have defined time at risk to include the time from the first court appearance until final disposition or sentencing (Goodman 1992). Some studies have counted the time from the date of release until (1) the case was disposed, or (2) the defendant failed to appear for a scheduled court appearance, whichever occurred first (Siddiqi 1999; Toborg 1981)5. Still others regard only failures to appear occurring in the four month period following the defendant's release, irrespective of whether the case has been resolved (Goldkamp et al. 1981).

5

Toborg’s calculation of time at risk also included the time to re-arrest for a new charge.

-22-

Although the measure of time at risk has not been standardized, the findings of various studies are consistent in showing a significant relationship between this variable and FTA. However, differences existed in the type of relationship reported. The Center for Governmental Research, Inc. (1983) reported that the rate of FTA increased in direct proportion to the amount of time the defendant was out on release. Conversely, other researchers have reported an inverse relationship between FTA and time at risk, with more defendants failing to appear earlier on in their release (Austin et al. 1985; Goldkamp et al. 1981; Siddiqi 1999; 2000; Toborg 1981; Bureau of Justice Statistics 1991).

Other Variables In addition to the aforementioned variables, researchers have also examined the effect of current criminal justice involvement on FTA. The Center for Governmental Research, Inc. (1983) and Sorin et al. (1979) reported higher FTA rates among defendants with current criminal justice involvement. Both Landes (1974) and Rhodes et al. (1996), however, found that being on probation or parole was not related to FTA. Another factor considered by some researchers when examining pretrial failure is the use of drugs by defendants. Higher FTA rates have been found among drug users (Roth and Wice 1980; Smith et al. 1989; Winterfield et al. 2003). However, several studies that test the ability of drug testing to predict FTA have found contrary results (Belenko et al. 1992; Britt et al. 1992).

In short, the literature shows mixed support for many of the independent variables discussed.

This is not surprising given the methodological inconsistencies from one study to the next.

However, seven tentative conclusions can be made. The effect of a defendant's community ties on FTA seems to be contingent upon the type and number of community-ties variables examined (1). With regard to socioeconomic variables, employment may be more important than income in predicting FTA (2). Demographic variables, such as age and race, display less predictive ability than variables that are not related to demographic characteristics (3). Criminal history appears to be important regardless of how it is measured (4). Defendants charged with less serious offenses appear to have higher FTA rates (5). The majority of studies find a significant relationship -23-

between charge severity and FTA. Although the operationalization of time at risk has not been standardized, various studies have consistently shown a relationship between this variable and FTA (6). And finally, other less often examined variables, such as drug use, may predict FTA (7).

Factors Predicting Pretrial FTA and Re-Arrest We found a few published studies that examined pretrial failure (Boudouris et al. 1977; Clarke et al. 1976; Cuvelier and Potts 1993; Goldkamp et al. 1981; VanNostrand 2003). In these studies, pretrial failure included both failure to appear and/or re-arrest. Boudouris et al. (1977) focused on two samples of defendants drawn from Polk and Linn counties in Iowa. The study examined the effectiveness of the point scales in predicting pretrial failure. The study conducted

by Clarke et al. (1976) focused on a random sample of defendants arrested in Charlotte, North Carolina in the first quarter of 1973. One of the principal objectives of the study was to examine the relative importance of various factors related to the combined risk of failure.

The study conducted by Goldkamp et al. (1981) focused on 4,800 defendants who were arraigned in the Municipal Court of Philadelphia between the summers of 1977 and 1979. The study had two major analytical objectives: (1) prediction of ROR decisions at arraignment, and (2) prediction of pretrial failure. Utilizing a number of variables, multivariate analyses were performed to achieve those purposes. Regarding pretrial failure, a number of models were developed. The final model was used to develop a risk assessment instrument.

The study conducted by Cuvelier and Potts (1992) initially focused on defendants who were arrested in Harris County, Texas in 1990. The study aimed at assessing the performance of the existing bail classification instrument and developing new instruments predicting pretrial failure.

Applying multivariate analysis, several models of pretrial failure were developed. The findings derived from the analysis were used to develop a new risk assessment instrument.

VanNostrand (2003) designed a study to assess pretrial risk among the defendant population in Virginia. The analysis focused on a sample of defendants who were arrested in seven Virginia locations between July 1, 1998, and June 30, 1999. Multivariate analysis was performed to -24-

identify significant predictors of pretrial failure, which were translated into a risk assessment instrument. The major findings from those studies are summarized as follows.

Community Ties Having a telephone was significantly related to lower failure (Cuvelier and Potts 1993; Goldkamp et al. 1981).

In her study, VanNostrand (2003) found a significant association

between pretrial failure and length of time at current residence—higher failure rates were observed among defendants who had lived at their current residence for less than one year.

Criminal History Boudouris et al. found that among those released on recognizance, having a record of commitment to a juvenile institution, having a prior jail or prison commitment, and having a prior adult conviction increased the likelihood of failure. Clarke et al. (1976) showed that criminal history was significantly related to pretrial failure--defendants with two or more prior arrests were significantly more likely to fail than defendants with one or no prior arrests.

Goldkamp et al. (1981) reported that defendants with one or more prior failures to appear, pending charges and recent arrests were more likely to fail than defendants without those characteristics. Cuvelier and Potts (1993) found that having prior misdemeanor convictions, prior felony convictions, and prior failures to appear significantly increased the risk of failure.

Similar findings were obtained by VanNostrand (2003). She found that the risk of failure was higher among defendants with prior misdemeanor convictions, prior felony convictions, prior violent convictions, prior failure to appear convictions, pending charges, and outstanding bench warrants.

Charge Goldkamp et al. (1981) found lower failure rates associated with certain charges. They included serious, personal, sex, non-narcotic drug offenses and miscellaneous offenses. He also pointed out that the probability of failure increased if a defendant had recent arrests and was -25-

charged with serious, personal, sex, and non-narcotic drug offenses. In another study, defendants charged with a felony were more likely to fail than defendants charged with a misdemeanor (VanNostrand (2003). In contrast, Clarke et al. (1976) found no significant relationship between the type of offense and pretrial failure.

Age Older defendants were less likely to fail than younger defendants (Boudouris et al. 1977; Cuvelier and Potts 1993; Goldkamp et al. 1981).

Other Variables Boudouris et al. (1977) reported higher failure rates among unemployed defendants. Higher failure rates were also observed among defendants in unskilled jobs. Cuvelier and Potts (1993) found lower failure rates among defendants who were either employed, in school, or homemakers. Similarly, VanNostrand (2003) found that defendants who were not employed or primary child caregivers were more likely to fail pretrial. Cuvelier and Potts (1993) also found that having an automobile decreased the risk of failure. Clarke et al. (1976) found that employment and income had no significant effect on pretrial failure.

To summarize, the research on pretrial failure suggested that criminal history variables were consistent predictors of pretrial failure across studies. The studies identified a few communityties variables which were related to pretrial failure--having a telephone, and being employed or a primary child caregiver were significant in two of three studies reviewed. The findings related to charge at initial arrest were inconsistent.

Summary The review of literature suggests that criminal history variables were consistent predictors of pretrial failure, irrespective of how it was measured (pretrial FTA, pretrial re-arrest, or both). We were not able to draw any conclusions about the effect of community-ties variables. This was mainly because the studies differed with respect to the variables examined. However, employment seemed to have a significant effect on both types of failure, whether examined -26-

separately or jointly. The literature also pointed to age as a consistent predictor of failure.

Finally, the charge at initial arrest and time at risk also aided prediction of failure. However, findings regarding their effects were not consistent across studies. The similarities in factors associated with pretrial FTA and re-arrest suggested that the two types of failure could be examined jointly, as was the case in several studies cited above. Consequently, one risk assessment instrument could be developed to predict both types of failure. These findings were used as a guide to develop research for our current analysis of pretrial failure among New York City defendants.

-27-

Section Two Methodology This section presents the research methodology used in this report. Sampling and data sources are discussed, and a description of the dependent and independent variables is provided.

In addition, statistical methods used in the analysis are described.

A. Sampling and Data Sources Data for the present analysis were drawn from a cohort of arrests made between January 1, 2001 and March 31, 2001, in which the defendants were prosecuted on new charges (as opposed to being re-arrested on a bench warrant, for example). The dataset excluded cases that were not docketed in the CJA database (UDIIS), unless there was an indication that they were prosecuted as “A” dockets in Manhattan, or as direct indictment6 (cases for which prosecution information is not available in CJA’s database).

The dataset contained 91,728 docketed arrests.7 A desk appearance ticket (DAT) was issued to 6% of the defendants, and the remaining 94% were held for arraignment in Criminal Court (summary arrests). Defendants issued desk appearance tickets were excluded from the study sample.

The primary data source was the CJA database.8 The Criminal Court data were tracked through November 30, 2001.9 By that time, 90% of the cases had reached a disposition in

6CJA’s

database does not contain court data for dockets with the same docket number Thus, court data for “A” dockets in Manhattan (the designation is used in Manhattan to distinguish between two court cases with the same docket number, one of which receives a suffix “A”) were not available for analysis. Felony prosecution in the Supreme Court as the result of a direct indictment by a grand jury is also unavailable. Arrest information is available for both these types of records, and defendant information may be available for arrests receiving “A” dockets. To the extent that these records could be distinguished from other types of non-docketed arrests, they were retained in the dataset, to maintain a complete cohort of prosecuted arrests.

7This number excludes cases transferred to Family Court prior to arraignment, and voided arrests.

8 Information about the arrest is provided by an on-line feed from the New York City Police Department.

9 If a case had multiple dockets, the Criminal Court information, including warrants, was pulled from the docket having the most severe arraignment charge (Penal Law severity).

-28-

Criminal Court. The cutoff date for Supreme Court data was January 31, 2002. Approximately 88% of the cases had reached final outcomes by that date. The criminal history information was supplemented with data from New York State Division of Criminal Justice Services (DCJS).10 In the first quarter 2001 dataset, 14% of the defendants had multiple arrests. To examine defendant behavior, the arrest-based file was converted into a defendant-based file, in which only a defendant’s first arrest during the sampling period was utilized. This file contained 67,848 defendants. Their arraignment dispositions are presented in Table 1.

As shown by the table, in the first quarter of 2001, 16% of the defendants had their cases dismissed at arraignment. Defendants who pled guilty comprised one-third of cases, and one-half had their cases adjourned for further appearances.

The analyses presented in this report focused on defendants whose cases were not completed at Criminal Court arraignment, and who were at risk of pretrial failure to appear and/or re-arrest (i.e., released on recognizance or bail prior to the disposition of all charges in Criminal or Supreme Court). Table 2 presents the release status for defendants whose cases were adjourned at Criminal Court arraignment. As shown by the table, 62% (21,379) of the defendants were released at arraignment; 57% were released on recognizance and 5% made bail. Another 16% (6,251) were released post-arraignment prior to the disposition of their case in Criminal or Supreme Court (table not shown). Less than 1% of the defendants were classified as juveniles (16 years or under) by CJA. Since our study focused on adult defendants only, they were excluded from the study sample. After excluding juvenile defendants, the 2001 sample contained 27,630 defendants who were released pretrial in Criminal or Supreme Court.

The present analysis was conducted in two phases. In Phase I, we examined the likelihood of re-arrest for a violent felony offense. We focused on a sample of defendants who were released and re-arrested pretrial. Table 3 presents the distribution of re-arrest for defendants who were

10

DCJS did not provide data for sealed cases. The New York City Police Department, DCJS, or any agency providing data bear no responsibility for the methods of analysis used in this report or its conclusions.

-29-

Table 1 Arraignment Outcome First Quarter of 2001 Dataset (Defendant-based) (N=67,848)

Arraignment Outcome

N

%

Non-Disposed

34626

51

Pled Guilty

22062

33

Dismissed

11125

16

35

0

67848

100

Other1 Total 1

Other includes transfer to other borough and family court.

-30-

Table 2 Release Status at Arraignment First Quarter of 2001 Dataset (Defendant-based)

Release Status

N

%

497

2

Bail Set, Not Made

12388

36

Bail Made

1709

5

ROR

19670

57

Total

34264

100

Remand

-31-

Table 3 Distribution of Pretrial Re-Arrest in the 2001 At-Risk Sample First Quarter of 2001 Dataset At-Risk Sample Pretrial Re-Arrest

N

%

Yes

4827

17%

No

22803

83%

Total

27630

100%

-32-

released pretrial in the 2001 at-risk sample. As shown by the table, in the 2001 at-risk sample, 17% (N=4,827) of the 27,630 released defendants were re-arrested pretrial. Our Phase I analysis focused on those 4,827 defendants.

In Phase II, we examined the likelihood of pretrial FTA or re-arrest (combined failure) among defendants who were released pretrial in the 2001 dataset. Of those released pretrial, close to 3% were missing a CJA release recommendation. They were excluded from the study sample. To be specific, the Phase II analysis focused on 26,820 defendants who were at-risk of combined failure in Criminal or Supreme Court.

B. Dependent Variables In the 2001 re-arrest sample, 10% of the re-arrested defendants were charged with a violent felony offense, whereas 8% were re-arrested for violent offenses of misdemeanor or lesser severity (Table 4). The remaining 82% were re-arrested for nonviolent felony or misdemeanor (or lesser) offenses. For the Phase I analysis, defendants re-arrested for nonviolent offenses and violent misdemeanor or lesser offenses were collapsed into one category. To be specific, when examining violent felony re-arrests, the dependent variable was dichotomized into “yes” and “no” categories. The “yes” indicated the presence of violent felony re-arrests and “no” indicated the absence of such re-arrests.

In the second phase of the analysis, we shifted our focus to a measure of combined failure.

This measure included both pretrial FTA and pretrial re-arrest. The pretrial FTA measured the issuance of a bench warrant prior to the disposition of a case in Criminal or Supreme Court. The pretrial re-arrest measured any re-arrest prior to the disposition of the initial case in Criminal or Supreme Court. The dependent variable was dichotomized into any combined failure (pretrial FTA or re-arrest) and no combined failure. Table 5 presents the distribution of pretrial FTA and/or re-arrest in the 2001 at-risk sample. As shown by the table, in the 2001 at-risk sample, 10% of the defendants failed to appear in Criminal or Supreme Court (FTA only). A slightly

-33-

Table 4 Re-Arrest Charge in the 2001 Re-Arrest Sample First Quarter of 2001 Dataset Re-Arrest Sample

N

%

% All At Risk

Re-Arrest for Violent Felony Offenses

497

10%

2%

Re-Arrest for Violent Misdemeanor Offenses*

365

8%

1%

Re-Arrest for Non-Violent Offenses

3965

82%

14%

Total

4827

100%

27630

* Includes violations and infractions

-34-

Table 5 Distribution of Pretrial Failure 2001 Re-Arrest Sample First Quarter of 2001 Dataset At-Risk Sample N

%

Pretrial FTA Only

2750

10%

Pretrial Re-Arrest Only

3194

12%

Pretrial FTA and Re-arrest

1473

6%

No Pretrial Failure

19403

72%

Total

26820

100%

-35-

higher proportion of defendants (12%) were re-arrested pretrial (re-arrest only). Six % of the defendants had both types of failure. Taken together, the combined failure rate for the 2001 atrisk sample was 28%.

C. Independent Variables In our analysis, we examined a number of independent variables. They included communityties items, criminal-history indicators, top initial arrest charge, demographic attributes and caseprocessing characteristics. Prior research and a review of correlations with each dependent variable aided the selection of the independent variables.

The community-ties items contained information on whether the defendants had a working telephone in their residence or had a cellular phone, the length of time at their current address, whether they had a New York City area address, whether they expected someone at their Criminal Court arraignment, and whether they were employed, in school, or in a training program full time at the time of their initial arrest. The criminal-history variables provided data on a defendant’s prior arrests, prior convictions, pending cases and prior FTA. The top charge at initial arrest considered both the type and severity of the offense. The offense "type" was based on its Uniform Crime Reports' (UCR) category. The offenses were categorized into (1) violent, (2) property, (3) drug, (4), public order offenses, and (5) other (see Appendix A for classification of offenses). These categories were similar to those used by the Bureau of Justice Statistics in its various reports on recidivism (Bureau of Justice Statistics 2002). The severity of the top arrest charge was derived from its New York State Penal Law offense class. The hierarchy from most to least serious severity level was: A felony, B felony, C felony, D felony, E felony, A misdemeanor, B misdemeanor, unclassified misdemeanor (U misdemeanor), violation and infraction. For our analysis, we used the type and severity of the arrest charge to compute a new charge variable, labeled as graded offense type. The graded offense type classified all offense types into felony and misdemeanor (or lower) level offenses. Consequently, we had felony level

-36-

violent, property, drug, public order and “other” offenses. Likewise, we had misdemeanor (or lower) level violent, property, drug, public order and “other” offenses.

The demographic variables provided information about a defendant’s sex, ethnicity, and age.

The case-processing variables included information on borough of initial arrest, borough of first pretrial re-arrest, time from arraignment to disposition on the initial arrest (case-processing time), type of first release, and court of disposition. The type of first release variable indicated whether a defendant was initially released on own recognizance or by the posting of bail. The court of disposition variable accounted for whether a case was disposed in Criminal Court or was transferred to Supreme Court. Included in the borough of arrest were the five boroughs comprising the City of New York: Brooklyn, Manhattan, Queens, the Bronx, and Staten Island.

D. Statistical Methods In the first phase, we examined the likelihood of pretrial re-arrest on violent felony offenses.

In the second phase, we shifted our focus to a measure of combined failure—pretrial FTA and/or re-arrest. Since dependent variables in both analytical phases were dichotomous (pretrial rearrest for a violent felony offense versus pretrial re-arrest for other offenses, pretrial FTA or rearrest versus no pretrial FTA and no re-arrest), logistic regression analysis was used to develop multivariate models (see Appendix B for coding of variables).

Multiple logistic regression analysis is a statistical technique that is used to test the individual effect of a number of independent variables on a dichotomous dependent variable, while controlling for the other variables in the model. A logistic regression procedure predicts the log-odds (the logit coefficients) of an observation being in one category of the dependent variable versus another. When reporting the results from a logistic regression model, one may also wish to transform the log-odds into odds ratios. This is accomplished by taking the antilog of the logit coefficient. The result is then interpreted as how much the odds of an outcome change, given a specific category of an independent variable. An odds ratio greater than one indicates an increase in the likelihood of an event occurring, and an odds ratio of less than one -37-

indicates a decrease in the likelihood of an event occurring. An odds ratio of one indicates the odds remain unchanged (no association between the independent and dependent variable). If the independent variable is continuous, such as age, the odds ratio measures the change in the odds of an outcome given one unit change in the independent variable. For dichotomous independent variables, such as gender, the odds ratio tells us how much the odds of an outcome change when cases are in one category versus another category. If a categorical independent variable has more than two categories, such as borough of initial arrest, the odds ratio measures the effect of being in each category of the independent variable versus a specified reference category. In the present analysis, the effect for each category was compared to the overall effect of that variable (deviation contrast technique). The last category was specified as the excluded category.

As an example, assume that a dichotomized independent variable is coded "1" if a defendant has a history of failure to appear, and "0" otherwise (no prior FTA). Also assume that the dependent variable, indicating re-arrest for a violent felony offense, is coded "1" if a defendant is re-arrested, and "0" if a defendant is not re-arrested. Estimating a univariate logistic regression model with prior FTA as the only independent variable produces a logit coefficient (log-odds) of .344. This suggests that when the variable of prior FTA changes from 0 to 1, there is an associated increase of .344 in the log-odds of re-arrest. Taking the antilog of the logit coefficient gives an odds ratio of 1.410. This indicates the odds of being re-arrested for a violent felony for defendants with prior failures to appear are about 1.4 times greater than that for defendants who do not have a history of failure to appear.

In the present analysis, a .05 level (or less) was used to ascertain whether an observation had a statistically significant effect on the dependent variable. A .05 level of significance means that the observation could have occurred by chance alone five times in 100. The overall ability of all the independent variables in a logistic regression model to predict the outcome variable was measured by examining Nagelkerke R2 (SPSS, Inc., 1999). This statistic indicates what proportion of the variation in the dependent variable is explained by all the independent variables in the model. Its values range from 0 to 1, with 0 indicating no variation in the -38-

dependent variable and 1 suggesting that all the variation in the dependent variable was explained by the independent variables in the model.

In both analytical phases, we developed several models predicting the outcome variable of interest. Variables were added or dropped depending upon their contribution to the dependent variable. The final models from each phase are presented in the current report.

In Phase II of the research, the final model was used as a guide to develop a point scale that would assess combined failure. For policy and practical concerns, a defendant’s demographics, case-processing characteristics, and top charge at initial arrest were dropped, and the model was re-estimated.

Points were assigned to each of the independent variables based on the logit

coefficients and significance levels. The effect for the excluded category was obtained by choosing an alternative reference category. Because the effect for each category was compared with the average across all categories of that variable, changing the reference category did not alter the effects of the other categories. For the purpose of standardization, the statistically significant logit coefficients were divided by .15 and were then rounded to the nearest whole number. The decision to divide by .15 was arbitrary, although consistent with several previous studies (Goldkamp et al., 1981; Goodman, 1992). If the coefficient was negative and statistically significant, a negative value was given, indicating that a defendant was less likely to FTA or be re-arrested pretrial. Likewise, positive values were given for positive significant coefficients, meaning that the likelihood of combined failure increased. A value of zero was given to categories that did not produce a statistically significant effect on combined failure. The total score for each defendant was obtained by summing those points. The point scale was used to develop a risk-classification system. The cutoff scores from the current ROR recommendation system were used to classify defendants into low, moderate and high-risk categories. The two systems were compared with respect to their ability to predict failure.

-39-

SECTION THREE PHASE I ANALYSIS VIOLENT FELONY RE-ARRESTS This section presents results from the analysis of pretrial violent felony re-arrests. Both descriptive and multivariate findings are described. The section concludes with a summary and discussion of findings.

A. Sample Characteristics When examining violent felony re-arrests, we focused on a sample of defendants who were released and re-arrested pretrial in the first quarter of 2001. The sample characteristics are presented below.

Demographic Attributes and Case-Processing Characteristics Table 6 presents the demographic and case-processing characteristics for defendants who were re-arrested pretrial. As shown by the table, a majority of the re-arrested defendants were male. Slightly more than half of the defendants were black, one third were Hispanic, and the remainder were white or other ethnicity. The median age was 28 years.

The median case-processing time from arraignment to case disposition for the sample’s initial arrest was 119 days. Less than 10% of the re-arrested defendants had their initial cases disposed in Supreme Court.

In the re-arrest sample, one third of the defendants were initially arrested in Manhattan, followed by Brooklyn (30%), the Bronx (20%), Queens (13%), and Staten Island (4%). Similar distributions were observed for the borough of re-arrest. To be specific, Manhattan had the highest proportion of defendants re-arrested pretrial (30%), followed by Brooklyn (29%), the Bronx (22%), Queens (14%), and Staten Island (5%).

-40-

Table 6 Demographic Attributes and Case-Processing Characteristics for Defendants Re-Arrested Pretrial First Quarter of 2001 Dataset Re-Arrest Sample Defendant Characteristics

N

%

Gender Male Female Total

4231 593 4824

88 12 100

Ethinicity Black Hispanic White 1 Other Total

2511 1617 471 155 4754

53 34 10 3 100

Age at Initial Arrest 18 and under 19-20 years 21-24 years 25-29 years 30-34 years 35-39 years 40 and higher Total

660 465 821 651 680 674 875 4826

14 10 17 13 14 14 18 100

Median Age (Years)

Median: 28 years

Borough of Initial Arrest Brooklyn Manhattan Queens Staten Island Bronx Total

1467 1572 626 207 955 4827

30 33 13 4 20 100

Type of Release Bail Made ROR Total

948 3699 4647

20 80 100

Court of Disposition Criminal Court Supreme Court Total

4442 385 4827

92 8 100

1

OTHER includes Asian, American Indian, and others.

Page 1 of 2

-41-

TABLE 6 (continued) Demographic Attributes and Case-Processing Characteristics for Defendants Re-Arrested Pretrial First Quarter of 2001 Dataset Re-Arrest Sample Defendant Characteristics

N

%

Borough of Re-Arrest Brooklyn Manhattan Queens Staten Island Bronx Total

1374 1466 693 217 1077 4827

29 30 14 5 22 100

Median Case Processing Time on Initial Arrest (Days)

119

Page 2 of 2

-42-

Community Ties Items Table 7 presents the community-ties variables for defendants who were re-arrested pretrial in the 2001 at-risk sample. As shown by Table 7, an overwhelming majority of the defendants in the 2001 re-arrest sample reported living at a New York City area address. Defendants that reported living with someone at the time of their initial arrest comprised three fifths of defendants re-arrested pretrial. Additionally, two thirds of the re-arrested defendants reported having a working telephone in their residence, or a cellular phone, and living at their current address for 18 months or longer. Less than one half reported being employed, in school, or in a training program full time. Only two fifths expected a relative or friend at Criminal Court arraignment.

Criminal History Table 8 provides a defendant’s criminal history recorded at the time of the sample’s initial arrest. Of those re-arrested pretrial, three fourths had been arrested prior to the sample’s initial arrest. Slightly more than two fifths had been convicted previously on misdemeanor charges, and one third had a prior felony conviction. Defendants with (a) prior violent felony conviction(s) comprised 12% of re-arrests. Those with a history of FTA comprised two fifths of re-arrests.

Defendants who had one or more cases open at the time of the sample’s initial arrest constituted one third of pretrial re-arrestees, and 14% had a bench warrant attached to their RAP sheet.

Charge at Initial Arrest Table 9 presents information related to the top charge at initial arrest. As shown by the table, nearly one half of the defendants in the 2001 re-arrest sample were initially arrested for felony charges, primarily B and D felonies. Slightly more than one fourth of the initial arrests were made for violent offenses. Combining type and severity (graded offense type), 15% of the initial arrests were made for violent felony offenses. The same proportion applied to violent offenses of misdemeanor or lesser severity. With respect to drug charges, 15% of the defendants were initially arrested for a felony drug offense, whereas 12% were arrested for misdemeanor or lesser drug charges.

-43-

Table 7 Community Ties Items for Defendants Re-Arrested Pretrial First Quarter of 2001 Dataset Re-Arrest Sample Defendant Characteristics

N

%

Verified Telephone Yes Unverified Yes Verified No Univerfied No Verified Unresolved Conflict Total

1979 1083 1248 61 150 4521

44 24 28 1 3 100

Verified Full Time Employment/ School/ Training Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

1462 554 1889 498 107 4510

33 12 42 11 2 100

Verified Length of residence of at least 18 months Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

2168 895 1129 214 119 4525

48 20 25 5 2 100

Expects Someone at Arraignment Yes No Total

1734 2760 4494

39 61 100

Verified NYC Area Address Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

3100 1118 199 18 85 4520

69 25 4 0 2 100

1731 899 1564 233 92

38 20 35 5 2

4519

100

Family Ties within the Residence Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

-44-

Table 8 Criminal History for Defendants Re-Arrested Pretrial First Quarter of 2001 Dataset Re-Arrest Sample Defendant Characteristics

N

%

First Arrest Yes No Total

1051 3390 4441

24 76 100

Prior Misdemeanor Convictions Yes No Total

1955 2624 4579

43 57 100

Prior Felony Convictions Yes No Total

1511 3068 4579

33 67 100

Prior Violent Felony Convictions Yes No Total

577 4250 4827

12 88 100

Open Cases Yes No Total

1621 2958 4579

35 65 100

Prior FTA Yes No Total

1990 2837 4827

41 59 100

Bench Warrant Attached to Rap Yes No Total

632 3956 4588

14 86 100

-45-

Table 9 Top Arrest Charge at Initial Arrest for Defendants Re-Arrested Pretrial First Quarter of 2001 Dataset Re-Arrest Sample N

%

Charge Severity A Felony B Felony C Felony D Felony E Felony A Misdemeanor B Misdemeanor or Lesser Charges Total

28 743 263 785 502 1953 511 4785

1 16 5 16 10 41 11 100

Offense Type Violent Property Drug Public Order Other Total

1404 697 1271 663 792 4827

29 14 26 14 16 100

Felony-Level Charges Violent Property Drug Public Order Other

699 477 701 264 180

15 10 15 6 4

Misdemeanor or Lesser Charges Violent Property Drug Public Order Other Total

705 220 570 398 571 4785

15 5 12 8 12 100

Graded Offense Type

-46-

Charge at Re-Arrest Table 10 provides information related to the top re-arrest charge. In the 2001 re-arrest sample, two fifths of the defendants were re-arrested for a felony, which was considerably lower than the proportion initially arrested for a felony (41% versus 48%). Slightly less than one fifth of the re-arrests were made for violent offenses. The variable reflecting graded offense type at re-arrest indicated that among defendants who were re-arrested for a felony, drug offenses were the most common charges (13%), followed by violent offenses (11%). Among those re-arrested for misdemeanor or lower charges, drugs were the most common re-arrest charges (19).

Summary The data in this section indicated that the majority of defendants in the 2001 re-arrest sample were male with a median age of 28 years. Slightly more than half of the defendants were black. Less than one half reported being engaged in a full-time activity at the time of initial arrest.

Manhattan was the borough where one third of the defendants were arrested. Less than one fifth were initially arrested for violent felony offenses. For the majority of defendants, the type of first release was ROR. Summarizing the criminal history information, three quarters of the defendants in the rearrest sample had been arrested previously. Slightly more than two fifths had been convicted previously of a misdemeanor offense. Defendants with prior felony convictions comprised one third of the re-arrest sample. Defendants with pending cases also comprised one third of rearrests. Those with prior failures to appear constituted two fifths of the study sample. The criminal history information indicated that the proportion of defendants with previous criminal justice involvement in the 2001 re-arrest sample was much higher than the proportion observed for the 2001 at-risk sample (see Table 17 in Phase II of the analysis). This was not unexpected, as our previous research has shown a significant relationship between criminal-history variables and pretrial re-arrest (Siddiqi June 2003).

-47-

Table 10 Top Re-Arrest Charge for Defendants Re-Arrested Pretrial First Quarter 2001 Dataset Re-Arrest Sample1 N

%

Re-Arrest Charge Severity A Felony B Felony C Felony D Felony E Felony A Misdemeanor 2 B Misdemeanor/Other Total

41 659 185 508 549 1864 887 4693

1 14 4 11 12 40 19 100

Offense Type at Re-Arrest Violent Property Drug Public-Order Other Total

861 720 1492 624 1130 4827

18 15 31 13 23 100

Felony Level Charges Violent Property Drug Public-Order Other

497 347 582 178 338

11 7 13 4 7

Misdemeanor or Lesser Charges Violent Property Drug Public-Order Other Total

364 373 910 429 675 4693

8 8 19 9 14 100

Graded Offense Type

1

Percentages do not add to 100 due to rounding.

2

Other includes: Unclassified Misdemeanors, Violations, Infractions, and charges outside the N.Y. State Penal Law and Vehicle and Traffic Law (e.g. Administrative and Public Health Codes).

-48-

Approximately one half of the defendants in the 2001 re-arrest sample were initially arrested for felony charges, primarily B and D felonies. The proportion charged with a felony declined at the time of re-arrest. In the re-arrest sample, 15% of the initial arrests were made for violent felony offenses. Drug offenses were the most common charge at the time of re-arrest.

B. Charge Related Information for Defendants Re-Arrested for Violent Felony Offenses In the 2001 re-arrest sample, 10% of the defendants were re-arrested for a violent felony offense. We examined the top charge for those defendants both at the initial arrest and re-arrest.

For the charge at initial arrest, we examined the severity, type, and graded offense type. For the re-arrest charge, we focused on the severity and Penal Law Article. The findings are presented below.

Charge at Initial Arrest Table 11 provides information on the top charge type and severity at the time of initial arrest for defendants who were re-arrested for violent felony offenses. The table shows that of those rearrested for a violent felony offense in the 2001 re-arrest sample, three fifths were initially arrested for a felony. Of those with violent felony re-arrests, one third were initially arrested for an A Misdemeanor. Defendants re-arrested for a violent felony that were initially arrested for violent offenses comprised two fifths of re-arrests. The remaining defendants were initially arrested for a variety of offenses, ranging from 18% for drug offenses to 12% for “other� offenses.

The graded offense type showed that more than one quarter of the defendants re-arrested for a violent felony offense were initially arrested for such an offense.

Top Re-Arrest Charge Table 12 presents the information on the severity and Penal Law Article of the top re-arrest charge for defendants re-arrested for a violent felony offense. As shown by the table, a D felony was the most common charge among those re-arrested for a violent felony offense. It accounted

-49-

Table 11 Top Charge at Initial Arrest for Defendants Re-Arrested for Violent Felony Offenses First Quarter 2001 Dataset Re-Arrest Sample Violent Felony Re-Arrest Subsample N

%

Severity at Initial Arrest A Felony B Felony C Felony D Felony E Felony A Misdemeanor B Misdemeanor 1 Other Total

3 73 57 123 39 160 30 10 495

1 15 11 25 8 32 6 2 100

Offense Type at Initial Arrest Violent Property Drug Public Order Offenses 2 Other Total

202 82 88 64 60 496

41 16 18 13 12 100

Felony-Level Charges Violent Property Drug Public Order Other

139 56 49 37 15

28 11 10 8 3

Misdemeanor or Lesser Charges Violent Property Drug Public Order Other Total

64 26 39 27 44 496

13 5 8 5 9 100

Graded Offense Type

1

2

Other includes: Unclassified Misdemeanors, Violations, Infractions, and charges outside the N.Y. State Penal Law and Vehicle and Traffic Law (e.g. Administrative and Public Health Codes).

Other includes charges not classified as violent, property, drug, or public disorder offenses.

-50-

Table 12 Top Re-Arrest Charge for Defendants Re-Arrested for Violent Felony Offenses First Quarter 2001 Dataset Re-Arrest Sample Violent Felony Re-Arrest Subsample1 N

%

Re-Arrest Charge Severity A Felony B Felony C Felony D Felony E Felony Total

12 139 85 243 18 497

2 28 17 49 4 100

Re-Arrest Penal Law Article Assault (PL 120) Robbery (PL 160) Homicide (PL 125) Sex Offenses (PL 130) Total

256 197 25 18 496

52 40 5 4 100

1

Percentages do not add to 100 due to rounding.

-51-

for almost one half of all the violent felony re-arrests. Re-arrest for a B felony (28%) was the second most common charge, followed by a C felony (17%), an E felony (4%) and an A felony (2%).

Table 12 also shows that assault was the most common charge at the time of re-arrest among defendants with violent felony re-arrests. It accounted for one half (52%) of all such re-arrests.

Robbery accounted for another 40% of violent felony re-arrests. The remaining 9% of the violent felony re-arrests were made for homicide (5%) and sex offenses (4%).

These findings suggest that defendants with violent felony re-arrests were more likely to have been initially arrested for felony offenses than defendants for all other re-arrests. The offense types at the time of initial arrest and re-arrest also showed some overlap. However, defendants were also charged for other offenses at the initial arrest. The graded offense type also showed some consistency between initial arrest and re-arrest. Most of the violent re-arrests were made for D felonies. Assaults accounted for half of the re-arrests for violent offenses.

C. Multivariate Analysis of Re-Arrest for Violent Felony Offenses Using the 2001 re-arrest sample, we ran logistic regression analysis to identify statistically significant predictors of violent felony re-arrests. Based on prior research, we selected a number of independent variables. Before examining them through logistic regression analysis, we examined their bivariate distributions, and correlations with the outcome variable. Appendix C showed that most of the variables showed little or no variation with respect to violent felony rearrests (yes or no).

The variables that showed slight variation by violent felony re-arrests included prior misdemeanor convictions, prior felony convictions, prior violent felony convictions, type of release, and ethnicity. The court of disposition, age, full-time activity, and graded offense type at initial arrest were significantly associated with re-arrests for violent felony offenses. Beginning with the graded offense type, defendants initially arrested for a violent felony offense had the highest proportion of defendants re-arrested for a violent felony. The proportion of defendants -52-

with a violent felony re-arrest decreased with age—defendants who were 18 years old or younger had the highest proportion re-arrested for a violent felony offense (18%), whereas defendants who were 40 years or older had the lowest proportion with such re-arrests (5%). The re-arrest rates for violent felony offenses were higher among defendants whose cases were disposed in Supreme Court, and who were not engaged in a full-time activity. The correlation of those variables with the outcome variable yielded the same results (data not shown).

The findings from the bivariate analysis were used to develop a multiple logistic regression model. The model contained all the variables that were significantly correlated with violent felony re-arrests. In addition, the model controlled for prior misdemeanor convictions and prior felony convictions. Defendants with prior misdemeanor convictions and prior felony convictions had slightly higher proportions re-arrested for violent felony offenses (11% versus 9%, for each).

In most of the prior research, prior convictions were found to be significant predictors of rearrest in general. Therefore, despite their insignificant association with the outcome variable in the current sample, they were added to the model to study their effect when the effects of all other variables in the model were controlled for.

The findings from the multivariate analysis of re-arrest for violent offenses are presented in Table 13. The table shows that when controlling for the effects of other variables in the model, only two variables attained statistical significance—age at the initial arrest, and graded offense type at initial arrest charge. A defendant’s age was negatively related to violent felony re-arrests, suggesting that older defendants were less likely to be re-arrested for a violent felony offense than younger defendants. With respect to the graded offense type at initial arrest, the odds of being re-arrested for a violent felony offense were higher among defendants who were initially arrested for such an offense. In contrast, defendants initially arrested for drugs were less likely to be re-arrested for a violent felony offense. This was true both for felony and misdemeanor (or lower) drug charges.

-53-

Table 13 Logistic Regression Model Predicting Violent Felony Re-Arrests First Quarter of 2001 Dataset Re-Arrest Sample (N=4325) Logit Coefficient

Significance Level

Odds Ratio

EMPL/SCHOOL/TRAINING: Excluded Category = Unresolved Conflict Yes Yes Verified No No Verified

-0.190 -0.167 -0.447 -0.522

0.505 0.580 0.119 0.096

0.827 0.846 0.640 0.593

AGE

-0.049

0.000

0.952

Black Hispanic White

0.164 -0.211 -0.045

0.112 0.198 0.687

1.178 0.81 0.956

PRIOR FELONY CONVICTION

0.049

0.702

1.050

PRIOR MISDEMEANOR CONVICTION

0.184

0.137

1.202

TYPE OF RELEASE

-0.155

0.212

0.856

Felony-Level Charges Violent Property Drug Public Order Other

0.656 0.233 -0.521 0.309 -0.181

0.000 0.107 0.001 0.086 0.516

1.928 1.262 0.594 1.362 0.835

Misdemeanor or Lesser Charges Violent Property Drug Public Order

-0.02 0.357 -0.404 -0.349

0.887 0.086 0.02 0.077

0.981 1.429 0.667 0.705

COURT OF DISPOSITION

0.347

0.056

1.415

Variable

ETHNICITY: Excluded Category = Other

GRADED OFFENSE TYPE AT INITIAL ARREST (excluded category = Misdemeanor other)

2

Nagelkerke R for the Model = 8%

-54-

An examination of logit coefficients for the significant variables suggested that the initial arrest for a violent felony offense had the strongest impact on the likelihood of re-arrest for a violent felony offense (B=.66). Being charged with a felony drug offense at initial arrest also had a strong negative effect on the likelihood of re-arrest for a violent felony offense (B=-.52). The odds of re-arrest for a violent felony offense were moderately lower among defendants who were initially arrested for drug offenses of misdemeanor or lesser severity (B=-.40).

Table 13 also provides the odds ratios for the variables in the model. In this model, the odds ratio for a variable reflected a change in the likelihood of being re-arrested pretrial for a violent felony offense after controlling for the effects of all of the other variables in the model. When compared with the odds of defendants with the mean effect of the graded offense type at initial arrest, the odds of being re-arrested pretrial for a violent felony were 1.93 times higher for defendants who had been initially arrested for such an offense. In contrast, the odds of being re-arrested for a violent felony offense were .594 times lower for felony drug offenders and .667 times lower for misdemeanor (or lower) drug offenders.

As shown by Nagelkerke R Square, the total amount of variance explained by the model was 8%. This suggested that our model was not strong in explaining the variation in the dependent variable.

D. Did the Same Factors Predict Pretrial Re-Arrest in General and Pretrial Re-Arrest for a Violent Felony Offense?

We compared the logistic regression model for violent felony re-arrests with the model that we had developed previously using the 2001 at-risk sample to predict pretrial re-arrest in general (Siddiqi June 2003). The objective was to determine whether the same variables would predict both types of recidivism.

As shown by Table 14, the model predicting re-arrest irrespective of the charge at initial arrest controlled for the community-ties variables, criminal-history indicators, charge type and

-55-

Table 14 Multiple Logistic Regression Model Predicting Pretrial Re-Arrest First Quarter 2001 Dataset At-Risk Sample (N=25021) Logit Coefficient

Significance Level

Odds Ratio

TELEPHONE Excluded Category: Unresolved Conflict Yes Yes Verified No, No Verified

-0.209 -0.044 0.152

0.000 0.522 0.001

0..811 0.957 1.165

EMPL/SCHOOL/TRAINING Excluded Category: Unresolved Conflict Yes Yes Verified No No Verified

-0.060 -0.315 0.341 0.042

0.340 0.000 0.000 0.493

0.942 0.730 1.406 1.043

BOROUGH OF ARREST Excluded Category: Bronx Brooklyn Manhattan Queens Staten Island

0.142 -0.043 -0.056 0.072

0.000 0.221 0.209 0.350

1.153 0.958 0.946 1.075

SEX (Male)

0.362

0.000

1.436

AGE

-0.035

0.000

0.965

ETHNICITY Excluded Category: Other White Black Hispanic

-0.095 0.198 0.096

0.059 0.000 0.010

0.910 1.219 1.101

COURT OF DISPOSITION (Supreme Court)

-1.407

0.000

0.245

Case Processing Time

0.007

0.000

1.007

PRIOR FTA

0.440

0.000

1.553

OPEN CASES

0.364

0.000

1.439 Page 1 of 2

-56-

TABLE 14 (contd.)

Logit Coefficient

Significance Level

Odds Ratio

A Felony B Felony C Felony D Felony E Felony A Misdemnor

-0.694 0.008 0.144 0.052 0.237 0.229

0.000 0.894 0.055 0.338 0.000 0.000

0.500 1.008 1.155 1.053 1.268 1.257

Prior Misdemeanor Conviction

0.430

0.000

1.538

Prior Felony Conviction

0.111

0.019

1.117

Top Arrest Chart Type Excluded Category: B Misd/Other Violent Property Drug Public Order

-0.137 -0.099 0.192 -0.032

0.000 0.022 0.000 0.455

0.872 0.905 1.212 0.969

TOP ARREST CHARGE SEVERITY Excluded Category: B Misd/Other

2

Nagelkerke R for the Model = 17%

Page 2 of 2

-57-

severity, the defendant’s demographics, and case-processing characteristics. Beginning with the community-ties variables, having a telephone in the defendant’s residence or having a cellular phone, and being employed, in school, or in a training program full time significantly predicted the likelihood of pretrial re-arrest. Defendants with a “yes” response to the telephone variable and defendants with a “yes verified” response to the full time activity variable were less likely to be re-arrested than defendants with the mean effect of the variable. In comparison, the model examining violent felony re-arrests only controlled for the full-time employment, in school, or in a training program variable. It did not attain statistical significance when the effects of other variables in the model were controlled for.

With respect to the criminal history variables, defendants having prior misdemeanor convictions, prior felony convictions, open cases, and a history of FTA at the time of initial arrest were more likely to be re-arrested pretrial than defendants who did not have such a history.

The model predicting violent felony re-arrest controlled for prior misdemeanor convictions, prior felony convictions, and open cases. None of these variables proved to be significant when all the other variables in the model were controlled for.

In both models, the charge at initial arrest had a statistically significant effect on the outcome variable. However, differences were observed in the interpretation of individual categories. In the model predicting pretrial re-arrest, drug offenders were more likely to be re-arrested than defendants with the mean effect of that variable, whereas defendants charged with a violent or property offense at the time of initial arrest were less likely to be re-arrested. Furthermore, defendants initially arrested for E felonies and A misdemeanor were more likely to be re-arrested pretrial than defendants with the mean effect of that variable. In comparison, in the model predicting violent felony re-arrests, defendants initially arrested for drug offenses were less likely to be re-arrested for a violent felony offense. This finding applied to both felony and misdemeanor or lower drug charges. In contrast, defendants initially arrested for violent felony offenses were more likely to be re-arrested for the same offenses.

-58-

In the model predicting pretrial re-arrest, all else being equal, the probability of re-arrest was higher for male, black, and Hispanic defendants. Because of their insignificance, these variables were not included in the model predicting violent felony re-arrest. The likelihood of being rearrested pretrial or being re-arrested on a violent felony offense decreased with age.

Turning to variation in case-processing characteristics, borough of initial arrest, court of disposition, and case processing time significantly predicted the likelihood pretrial re-arrest.

None of these variables had a statistically significant effect on the likelihood of re-arrest for a violent felony.

E. Summary and Discussion In the 2001 at-risk sample, 17% of the defendants were re-arrested pretrial. Of those who were re-arrested, 10% were charged with a violent felony at the time of re-arrest. Most of the violent felony re-arrests were made for assaults, the plurality of which were D felonies. An examination of the top initial arrest charge for defendants with violent felony re-arrests suggested some overlap. However, the defendants were also charged with a variety of other offenses at the time of initial arrest.

We developed a logistic regression model to identify significant predictors of pretrial rearrest for a violent felony offense. When running that model, only two variables proved to be statistically significant predictors of violent felony re-arrest, graded offense type at initial arrest, and age at initial arrest. The likelihood of pretrial re-arrest for a violent felony offense was higher among defendants who were initially arrested for a violent felony charge. In contrast, the odds of being re-arrested for a violent felony offense were lower among defendants initially arrested on drug charges. The finding related to the effect of graded offense type is consistent with prior research, which suggests similarities between the type of initial offense and the type of offense at the time of re-arrest (Bureau of Justice Statistics 1987). Regarding age at initial arrest, older defendants were less likely to be re-arrested for a violent felony offense than younger

-59-

defendants. This is also consistent with prior research on recidivism for violent offenses (Goldkamp, et al. 1981).

We compared the findings from the current research with the findings derived from our previous research on pretrial re-arrest in general. The comparison indicated that in both models, top initial arrest charge and age at initial arrest had a significant effect on the outcome variable.

When examining re-arrest irrespective of the re-arrest charge, top initial arrest charge was a significant, but weak, predictor of pretrial re-arrest; defendants initially arrested for violent or property offenses were less likely to be re-arrested, whereas defendants initially arrested for drug charges were more likely to be re-arrested. The present research shows that although violent offenders were less likely to be re-arrested, violent felony offenders had a tendency to be rearrested for similar offenses. Since our focus in the present research was re-arrest for a violent felony offense, we do not know whether this finding will hold for re-arrests for violent offenses of misdemeanor or lesser severity. In the present research, defendants initially arrested for drug offenses were less likely to be re-arrested for violent felony offenses.

The comparison between the two models (re-arrest and re-arrest for a violent felony offense) further showed that a defendant’s community ties (telephone in the defendant’s residence/cellular phone, full time activity), criminal history (prior misdemeanor convictions, prior felony convictions, open cases, prior failures to appear), demographic attributes (gender, ethnicity, age), initial arrest charge, and case processing characteristics (case processing time, borough of initial arrest, court of disposition) were significant predictors of pretrial re-arrest only. When examining re-arrests for violent felony offenses, only two variables significantly predicted the likelihood of violent felony re-arrests—age and graded offense type. This was not surprising, as most of the sample characteristics did not show any discrimination by the outcome variable. This may be attributed to one common factor in our sample—all the defendants in our sample were re-arrested pretrial. Furthermore, only 10% of them were re-arrested for a violent felony offense, which made the task of prediction quite difficult. More specifically, we were trying to predict an event which was quite rare in our sample. This may very well be the issue -60-

with previous studies too, which have identified only a handful of variables as the significant predictors of recidivism for violent offenses (Goldkamp et al. 1981). In sum, what predicts pretrial re-arrest in general may not predict pretrial re-arrest for violent felony offenses.

Research should examine the two types of re-arrest separately.

-61-

SECTION FOUR PHASE II ANALYSIS: PRETRIAL FAILURE This section presents results from the analysis of pretrial failure, which takes into consideration both pretrial FTA and pretrial re-arrest. The sample characteristics and findings derived from the multivariate analysis of pretrial failure are described. Based on the multivariate model, a new risk classification system to assess pretrial failure is developed. Comparisons are made with the ROR recommendation system.

A. Sample Characteristics The analysis focused on a sample of defendants who were released pretrial in the First Quarter 2001 Dataset and were at risk for pretrial failure. Their characteristics are described below.

Demographic and Case-Processing Characteristics Table 15 displays demographic and case-processing characteristics for defendants released pretrial. As shown by the table, an overwhelming majority of the defendants were male. Slightly less than one half of the defendants were black, about one third were Hispanic, and the remainder were white or an “other� ethnicity. The median age was 30 years.

In the 2001 at-risk sample, Brooklyn and Manhattan had the highest proportion of defendants arrested (30% in each of these boroughs). Bronx arrestees comprised one fifth of the defendants, and 16% were arrested in Queens. Staten Island had the lowest number of arrests (4%). Slightly more than one tenth of the defendants had their cases disposed in Supreme Court. The majority was released on recognizance.

Community Ties Table 16 presents community-ties variables for defendants released pretrial. As shown by the table, an overwhelming majority of the defendants in the 2001 at-risk sample reported living in the New York City area. Approximately three fourths reported having a telephone in their -62-

Table 15 Demographic and Case Processing Characteristics First Quarter 2001 Dataset At-Risk Sample Defendant Characteristics

N

%

DEMOGRAPHIC ATTRIBUTES Sex Male Female Total

22455 4354 26809

84 16 100

Ethnicity Black Hispanic White 1 Other Total

12343 9114 3476 1484 26417

47 34 13 6 100

Age at Arrest 18 and under 19-20 years 21-24 years 25-29 years 30-34 years 35- 39 years 40 and older Total

3063 2244 4132 3818 3851 3643 6070 26821

12 8 15 14 14 14 23 100

Median Age (Years)

30

CASE-PROCESSING CHARACTERISTICS Borough of Arrest Brooklyn Manhattan Queens Staten Island Bronx Total

7938 7905 4397 1078 5503 26821

30 30 16 4 20 100

Type of Court Criminal Court Supreme Court Total

23616 3205 26821

88 12 100

Type of First Release ROR Bail Total

21081 5600 26681

79 21 100

-63-

Table 16 Community Ties First Quarter 2001 Dataset At-Risk Sample1 N

%

COMMUNITY-TIES ITEMS Verified NYC Area Address Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

16902 7291 1289 156 314 25952

65 28 5 1 1 100

Verified Length of residence of at least 18 months Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

12696 5811 5515 1432 532 25986

49 22 21 6 2 100

Verified Family Ties Within Residence Yes Unverified Yes Verified No, Unverified No Verified Unresolved Conflict Total

9834 5712 8337 1700 386 25969

38 22 32 6 2 100

Expects Someone at Arraignment Yes No Total

9866 16013 25879

38 62 100

Verified Telephone Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

12472 7115 5380 324 668 25959

48 27 21 1 3 100

Verified Full Time Employment/ School/ Training Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

9784 4233 8657 2704 563 25941

38 16 33 10 2 100

1

Percentages do not add to 100 due to rounding.

-64-

residence, or having a cellular phone, and living at their current address for 18 months or longer. Defendants reportedly living with someone at the time of initial arrest comprised three-fifths of the sample. Slightly more than one half of the defendants reported being employed, in school, or in a training program full time and about two fifths expected a relative or friend at Criminal Court arraignment.

Criminal History Table 17 provides criminal history information for defendants who were released pretrial in the 2001 at-risk sample. The table shows that almost three fifths of the defendants had been arrested previously. Slightly more than one fourth had been convicted previously on misdemeanor charges, and one fifth had (a) prior felony conviction(s). Prior violent convictions accounted for 8% of the sample. Defendants with one or more cases at the time of the sample arrest comprised nearly one fourth of the sample, and almost one tenth had a bench warrant attached to their RAP sheet.

In the 2001 at-risk sample, 16% of the defendants failed to appear pretrial in Criminal or Supreme Court. In comparison, 17% were re-arrested pretrial. As mentioned previously, the combined failure rate was 28% (see Table 5).

Top Charge Information Defendants initially arrested for felony charges, primarily B and D felonies, comprised slightly more then one half of the sample (Table 18). One third of the arrests were made for violent offenses, and one fourth for drug offenses. Combining both type and severity, 16% of the arrests were made for violent felony offenses. The same percentage applied to violent arrests made for misdemeanor or lesser severity. The proportions of defendants initially arrested for felony level property or drug offenses were considerably higher than that for misdemeanor (or lesser) level property or drug offenses (11% versus 4% for the former, 16% versus 8% for the latter). The proportion of defendants arrested for “other” felony offenses was much lower than that for “other” misdemeanor offenses (3% versus 12%).

-65-

Table 17 Criminal History First Quarter 2001 Dataset At-Risk Sample N

%

10953 15284 26237

42 58 100

2031 24790 26821

8 92 100

5950 20287 26237

23 77 100

7505 18732 26237

29 71 100

Yes No Total

6221 20016 26237

24 76 100

Type of Warrant Attached to Rap Sheet Bench Warrant No Bench Warrant Total

2413 23945 26358

9 91 100

Yes No Total

6924 19897 26821

26 74 100

Pretrial FTA on the Sample Arrest Yes No Total

4223 22598 26821

16 84 100

Pretrial Re-Arrest on the Sample Arrest Yes No Total

4667 22154 26821

17 83 100

CRIMINAL HISTORY First Arrest Yes No Total Prior Violent Felony Convictions Yes No Total Prior Felony Convictions Yes No Total Prior Misdemeanor Convictions Yes No Total Open Cases

Prior FTA

-66-

Table 18 Top Charge at Initial Arrest First Quarter 2001 Dataset At-Risk Sample N

%

Top Arrest Charge Severity A Felony B Felony C Felony D Felony E Felony A Misdemeanor B Misdemeanor 1 Other Total

335 4275 1430 5114 2618 9956 1503 1441 26672

1 16 5 19 10 37 6 6 100

Top Arrest Charge Type Violent Property Drug Weapon Gambling DUI (alcohol or drugs) Criminal Mischief VTL (excluding DUI) Other Total

8663 3380 6221 1090 219 870 870 915 4446 26674

33 13 23 4 1 3 3 3 17 100

Felony-Level Charges Violent Property Drug Public Order Other

4209 2881 4164 1737 781

16 11 16 6 3

Misdemeanor or Lesser Charges Violent Property Drug Public Order Other Total

4454 1066 2057 2109 3214 26672

16 4 8 8 12 100

Graded Offense Type

1

OTHER includes unclassified misdemeanors, violations, infractions, and charges outside the New York State Penal Law and Vehicle & Traffic Law (e.g. Administrative and Public Health Codes).

-67-

B. Multivariate Analysis of Pretrial Failure Based on prior research, we selected a number of variables for our analysis of combined failure. Most of those variables were examined previously in our analyses of pretrial FTA and rearrest. Those variables included community-ties factors, criminal history variables, charge related information, demographic attributes, and case-processing characteristics. Before entering them into a multiple logistic regression model, their correlation with combined failure was examined (data not shown) (also see bivariate relationship in Appendix D). The correlation matrix revealed that expecting someone at arraignment and the court of disposition were weakly related with pretrial failure. This was not surprising, as our previous research showed that expecting someone at arraignment was a significant predictor of pretrial FTA only (Siddiqi June 2003). In contrast, the variable reflecting the court of disposition was a significant predictor of pretrial re-arrest only. These two variables lost statistical significance when we used a combined measure of pretrial failure, and so they were not included in our model.

Table 19 displays the findings from the logistic regression model of combined pretrial failure. Beginning with the community-ties variables, living at a New York City area address, having a telephone in their residence, and being employed, in school, or in a training program full time significantly predicted a lower likelihood of a defendant exhibiting pretrial failure. This was true regardless of verification. In contrast, the odds of failing to appear or being re-arrested pretrial were higher among defendants with “no” or “no verified” responses to those variables.

Furthermore, when examining pretrial failure, the effect of the “unresolved conflict” category of the New York City area address variable became significant—defendants categorized as “unresolved conflict” were more likely to fail than the mean effect of that variable. It should be noted, however, that this category was comprised of less than 1% of the defendants.

Regarding the criminal history variables, defendants having prior misdemeanor convictions, open cases, and a history of FTA were more likely to FTA or be re-arrested than defendants who did not have such a history. Having (a) prior felony conviction(s) did not attain statistical significance when pretrial FTA and re-arrest were examined together.

-68-

Table 19 Multiple Logistic Regression Model Predicting Pretrial Failure First Quarter 2001 Dataset At-Risk Sample (N=24926) Logit Coefficient

Significance Level

Odds Ratio

TELEPHONE Excluded Category: Unresolved Conflict Yes, Yes Verified No, No Verified

-0.185 0.275

0.000 0.000

0.831 1.316

EMPL/SCHOOL/TRAINING Excluded Category: Unresolved Conflict Yes Yes Verified No, No Verified

-0.137 -0.215 0.199

0.001 0.000 0.000

0.872 0.807 1.220

NYC AREA RESIDENCE Excluded Category: Unresolved Conflict Yes Yes Verified No, No Verified

-0.103 -0.297 0.193

0.039 0.000 0.003

0.902 0.743 1.213

BOROUGH OF ARREST Excluded Category: Bronx Brooklyn Manhattan Queens Staten Island

0.037 -0.040 -0.120 0.239

0.237 0.187 0.001 0.000

1.037 0.961 0.887 1.269

SEX (FEMALE)

-0.191

0.000

0.826

AGE

-0.027

0.000

0.973

ETHNICITY Exclueded Category: Other White Black Hispanic

-0.129 0.199 0.092

0.002 0.000 0.003

0.879 1.220 1.096

CASE PROCESSING TIME

0.006

0.000

1.006

PRIOR FTA

0.621

0.000

1.860

OPEN CASES

0.285

0.000

1.330 Page 1 of 2

-69-

TABLE 19 (contd.)

PRIOR MISDEMEANOR CONVICTION PRIOR FELONY CONVICTION

Logit Coefficient 0.223

Significance Level 0.000

Odds Ratio 1.250

0.037

0.376

1.038

-0.339 -0.202 -0.241 -0.227 0.202

0.000 0.000 0.000 0.000 0.011

0.713 0.817 0.786 0.797 1.224

-0.027 0.247 0.318 0.210

0.479 0.000 0.000 0.000

0.973 1.280 1.374 1.234

GRADED OFFENSE TYPE AT INITIAL ARREST EXCLUDED CATEGORY: MISDEMEANOR, OTHER Felony-Level Offenses Violent Property Drug Public Order Other Misdemeanor or Lesser Charges Violent Property Drug Public Order 2

Nagelkerke R for the Model = 17%

Page 2 of 2

-70-

The top charge at initial arrest had a statistically significant effect on the likelihood of combined failure. The probability of failure was lower among defendants initially arrested for all types of felony offenses, with the exception of offenses categorized as “other” felonies.

Defendants initially arrested for “other” felonies were more likely to fail than defendants with the mean effect of that variable. The odds of pretrial FTA or re-arrest were higher among defendants initially arrested for all types of misdemeanor or lesser offenses, with the exception of those arrested for violent offenses. Defendants having been arrested for misdemeanor level violent offenses had no significant effect on the likelihood of combined failure.

An examination of demographic variables indicated that, all else being equal, the probability of FTA or re-arrest was higher among male, black, Hispanic and younger defendants. The likelihood of failure decreased with age. With respect to case-processing characteristics, the likelihood of failure increased the more time it took to reach a disposition. The borough where the initial arrest occurred also proved relevant: the likelihood of pretrial failure was lower among defendants initially arrested in Queens. In contrast, defendants initially arrested in Staten Island were more likely to exhibit failure pretrial than the citywide average.

As shown by Nagelkerke R Square, the total amount of variance explained by the model was 17%.

C. Constructing a Point Scale Model 1, presented in Table 19, was used to guide the development of a point scale that would measure both pretrial FTA and re-arrest. However, we needed to re-estimate the model to address a number of policy and practical issues. To be specific, the model presented in Table 19 suggested that the likelihood of pretrial failure was higher among male, black, Hispanic, younger defendants, and defendants arrested in Staten Island. As a policy, CJA does not discriminate against defendants on the basis of demographics or borough of arrest. Therefore, these variables were excluded from the final model. The model in Table 19 also indicated that case-processing

-71-

time was significantly related with pretrial failure. Since this information is not available at Criminal Court arraignment, it was also dropped from the model.

Another issue was related to the charge variable. This issue surfaced a few years ago during our research on pretrial FTA. Our research suggested that the severity and the type of top initial arrest charge were significantly related to pretrial FTA (Siddiqi 1999; 2000). To be specific, when controlling for the other variables in the model, the severity of the top arrest charge was found to be a significant, but weak predictor of pretrial FTA. Furthermore, its interpretation was not consistent across different samples. In the 1989 sample, defendants arrested for an A misdemeanor were more likely o FTA than the mean effect of that variable (Siddiqi 1999). In the 1998 sample, the likelihood of FTA was lower among defendants who were arrested for A or B felonies (Siddiqi 2000). For those reasons, it was excluded from the final model of FTA. Regarding the type of top arrest charge, all else being equal, defendants arrested for property, drug, criminal mischief, and VTL (Violation of Traffic Law, excluding DUI) offenses were more likely to FTA than those with the mean effect of that variable, and would score negative points on the point scale (Siddiqi 2000). In contrast, defendants arrested for gambling or driving while under the influence of alcohol or drugs were less likely to fail to appear, and would score positive points. When those findings were presented to the CJA staff, concern was expressed over assigning positive points to certain charge categories. After discussing a number of alternatives, it was agreed that CJA, as a policy, would not assign positive points to any category comprising the type of initial arrest charge. As such, those defendants would score zero points on this point scale item. Later, at the suggestion of Criminal Court judges, and due to the difficulties in operationalizing, this variable was excluded from the new ROR recommendation system.

For the current analysis, we re-estimated the model with and without the charge variable.

However, in order to be consistent with the current ROR recommendation system, and due to time constraints, we will only report the model that excludes the charge variable. The model is presented in Table 20.

-72-

Table 20 Multiple Logistic Regression Model Used to Develop a Point Scale for Pretrial Failure First Quarter 2001 Dataset At-Risk Sample (N=25399) Logit Coefficient

Significance Level

Odds Ratio

TELEPHONE Yes, Yes Verified No, No Verified Unresolved Conflict

-0.173 0.234 -0.061

0.000 0.000 0.425

0.841 1.263 0.941

EMPL/SCHOOL/TRAINING Yes Yes Verified No, No Verified Unresolved Conflict

-0.127 -0.198 0.162 0.163

0.001 0.000 0.000 0.032

0.881 0.821 1.175 1.177

NYC Area Residence Yes Yes Verified No, No Verified Unresolved Conflict

-0.107 -0.256 0.108 0.255

0.023 0.000 0.080 0.035

0.898 0.774 1.114 1.291

PRIOR FTA

0.636

0.000

1.889

OPEN CASES

0.347

0.000

1.415

PRIOR MISDEMEANOR CONVICTIONS

0.147

0.000

1.158

Variable

-73-

As shown by Table 20, the re-estimated model included a defendant’s community ties and criminal history. Excluded from the model were the charge variable, case-processing characteristics, and demographic attributes. All the variables that were significant in the original model continued to be significant in the re-estimated model, with the exception of a slight shift in some of the categories of the New York City area address variable. In the original model, defendants with a negative response to the New York City area address were more likely to exhibit failure than defendants with the mean effect of that variable. When we re-estimated the model, defendants with “no” or “no verified” responses had no significant effect on the likelihood of combined failure.

The re-estimated model was used to develop a point scale for defendants at risk of pretrial failure to appear or re-arrest in Criminal or Supreme Court. Points were assigned to each of the independent variables based on their estimated coefficients and significance levels (Table 21).

For purposes of standardization, the logit coefficients were divided by a constant (.15) and were then rounded to the nearest whole number. If the coefficient was negative and significant, a negative value was given, indicating that a defendant was less likely to fail pretrial. Likewise, positive values were given for significant positive coefficients, meaning that the likelihood of pretrial failure increased. The insignificant coefficients were assigned a value of zero. The signs for the logit coefficients were reversed when the values were translated into a point scale. To be specific, negative points indicated higher probability of failure, whereas positive points showed lower probability of failure. As an example, the logit coefficient for having prior FTA was positive .636 (Table 20), indicating that a defendant was more likely to fail pretrial. When it was divided by .15, and rounded to the nearest whole number, a value of positive 4 points was obtained. Reversing the sign for that coefficient yielded a value of negative 4 points. Thus, that defendant would lose 4 points on the point scale.

Beginning with the community-ties variables, defendants with affirmative responses (“yes” and “yes verified”) to having a telephone in their residence would score one point on the new scale. In contrast, defendants with negative responses (“no” and “no verified”) would score -74-

Table 21 Point Scale Predicting Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample

Variable

Points

TELEPHONE Yes, Yes Verified No, No Verified Unresolved Conflict

1 -2 0

EMPL/SCHOOL/TRAINING Yes, Yes Verified No, No Verified Unresolved Conflict

1 -1 -1

NYC AREA ADDRESS Yes Yes Verified No, No Verified Unresolved Conflict

1 2 0 0

PRIOR FTA Yes No

-4 4

OPEN CASES Yes No

-1 1

PRIOR MISDEMEANOR CONVICTIONS Yes No

-1 1

-75-

negative two points. Due to statistical insignificance, no points would be assigned to defendants who were categorized as “unresolved conflict” on the telephone variable. With regard to being engaged in a full-time activity, defendants with affirmative responses (“yes” and “yes verified”) would receive one point. Defendants recorded as “no” or “no verified” responses would be assigned negative one point. Defendants categorized as “unresolved conflict” would also score negative one point. Living at a New York City area address was also a significant predictor of pretrial failure. Defendants with a “yes, unverified” response would score positive one point.

Defendants categorized as “yes verified” would have two points added to their point scale score.

Defendants with a negative response (“no” or “no verified”) to that variable had no significant effect on the likelihood of pretrial FTA or re-arrest, and therefore would score zero points on the scale. Due to the small proportion of defendants comprising the “unresolved conflict” category, no points would be assigned to that category.

Of the criminal-history variables, the prior bench warrant item contributed the most to a defendant’s total score; defendants with a prior bench warrant would lose four points on the point scale, whereas defendants with no history of failure to appear would receive four points.

Defendants who have prior misdemeanor convictions and/or open cases would lose one point on each item.11 Defendants who do not have a prior misdemeanor conviction or an open case would score one point on each item. A defendant’s total score would range from -9 to +10 points.

D. Comparing the Point Scale Predicting Pretrial Failure with the Point Scale Predicting Pretrial FTA A comparison of the point scale predicting pretrial FTA or re-arrest with the point scale predicting pretrial FTA (Table 22) revealed that with the exception of slight differences, the two

11

In the model that controlled for all the variables, having a prior felony conviction was not a significant predictor of pretrial FTA or re-arrest. It became statistically significant when top initial arrest charge, demographic attributes, and case processing characteristics were excluded from the model (data not shown here). Having a prior misdemeanor conviction was significantly related to pretrial failure regardless of which variables were included in the model. Hence we decided to include that variable in the final model.

-76-

Table 22 A Comparison of the Point Scale Predicting Pretrial Failure with the Point Scale Predicting Failure To Appear First Quarter of 2001 Dataset At-Risk Sample Failure Points

FTA Points

TELEPHONE Yes, Yes Verified No, No Verified Unresolved Conflict

1 -2 0

1 -2 0

EXPECTS AT ARRAIGNMENT Yes No

-----

1 -1

EMPL/SCHOOL/TRAINING Yes, Yes Verified No, No Verified Unresolved Conflict

1 -1 -1

1 -1 -2

NYC AREA ADDRESS Yes Yes Verified No, No Verified Unresolved Conflict

1 2 0 0

0 3 -2 0

PRIOR FTA Yes No

-4 4

-5 5

OPEN CASES Yes No

-1 1

-1 1

PRIOR MISDEMEANOR CONVICTIONS Yes No

-1 1

-----

Variable

-77-

scales were quite similar with respect to the predictor variables. However, the strength for some of the variables was altered. The findings are presented as follows.

With the exception of expecting someone at arraignment, all of the community ties variables predicted both pretrial FTA and combined failure. These variables included having a telephone in the defendant’s residence, residing in the New York City area, and being employed, in school, or in a training program full time. With the exception of having a telephone or having a cellular phone, when analyzing each variable individually, differences in interpretation were found.

Wh e n assessing a defendant’s risk of flight, defendants with verified affirmative responses to living at a New York City area address were less likely to FTA than defendants with the mean effect of that variable, and would score three points on the scale. In contrast, defendants with no or no verified responses were more likely to FTA, and would lose two points. Due to their insignificance, no points were assigned to defendants who were recorded as yes unverified or unresolved conflict on that variable. When assessing the risk of combined failure, defendants with an unverified affirmative response and defendants with a verified affirmative response earned one and two points, respectively. No points were assigned to defendants with a negative response (no or no verified), or to defendants whose response was in conflict with that of a verifier. In sum, the effect of living at a New York City area address was slightly weaker in the point scale measuring combined failure. This was not surprising, as our prior research showed that living at a New York City area address was a significant predictor of pretrial FTA only (Siddiqi June 2003). It had no significant effect on the likelihood of pretrial re-arrest in the 2001 at-risk sample (Siddiqi June 2003). The variable became significant when two types of failure were examined together. That significance was mainly due to its relationship with pretrial FTA. However, as described above, the strength for some of the categories was altered.

Continuing with the community-ties variables, the two point scales behaved similarly with respect to the “engaged in a full-time activity” variable, with the exception of the “unresolved conflict” category. When examining pretrial FTA, defendants recorded as unresolved conflict lost two points on the scale. In comparison, when examining combined failure, those defendants -78-

lost one point. No differences were observed with respect to the points assigned to having a telephone in the defendant’s residence or having a cellular phone.

More criminal history variables predicted combined failure than pretrial FTA. When predicting combined failure, having prior misdemeanor convictions, open cases, and prior FTA increased the chances that a defendant would either FTA or be re-arrested pretrial. When predicting pretrial FTA, only open cases and prior FTA proved to be relevant.

In both scales, prior FTA was the strongest contributor to a defendant’s total score.

Nevertheless, the effect was slightly stronger for the scale measuring a defendant’s risk of flight than for the scale measuring risk of flight and/or re-arrest (five points versus four points). This was not surprising, as prior FTA was a stronger predictor of pretrial FTA than pretrial re-arrest.

When examining both pretrial FTA and/or re-arrest, the strength of the variable decreased slightly. In both point scales, defendants with a pending case lost one point.

To summarize, the two point scales were quite similar with respect to most of the predictor variables. However, the strength for some of the variables was altered, depending upon the type of failure under scrutiny.

E. Distribution of the Point Scale Scores by Pretrial Failure Table 23 displays the distribution of the combined measure of failure by point-scale scores.

In general, defendants scoring lower points had higher failure rates. As shown by the table, a few defendants in the 2001 at-risk sample scored -9 points. Approximately three fifths of them failed to appear for at least one scheduled court appearance and/or were re-arrested pretrial. This score would be given to a defendant who: did not have a telephone or cellular phone (-2 points), was not employed, in school, or in a job training program full time, or had an unresolved conflict on this item,(-1), did not reside at a New York City area address (0), had prior bench warrants (-4), had at least one prior misdemeanor conviction (-1), and had one or more pending cases (-1).

-79-

Table 23 Distribution of Point Scale Scores by Pretrial Failure

Points -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Total

Total Defendants N % 79 0 429 2 140 1 755 3 656 3 720 3 1231 5 777 3 914 4 594 2 347 1 598 2 465 2 1638 6 1305 5 1870 7 3598 14 2554 10 4183 17 2545 10

Pretrial Failure N % 47 60 223 52 70 50 386 51 285 43 293 41 544 44 305 39 336 37 198 33 119 34 202 34 155 33 548 34 404 31 492 26 820 23 531 21 697 17 359 14

25398

7014

100

* Percentages may not add up to 100 due to rounding

-80-

28

The highest score in the sample under study was 10 points. Defendants with such a score comprised 10% of the at-risk sample. Their failure rate was 14%. These defendants had a telephone or a cellular phone (1 point), were employed, in school, or in a job training program full time (1), were verified as living at a New York City area address (2), had no prior bench warrant (4), had no prior misdemeanor conviction (1), and had no open case (1).

F. Developing a Risk Classification System for Pretrial Failure At the next step of the research, we were interested in developing a risk classification system that would classify defendants on the basis of their risk of failure on the new point scale.

However, in order to create different risk groups, we needed to determine cutoff points. For the current report, we applied the cutoff points from the ROR recommendation system to the point scale predicting combined failure. However, it should be noted that the risk classification system based on those cutoff points was designed for research purposes only, so that we could compare its performance with the ROR recommendation system. Should the Statute allow consideration of public safety in making pretrial release decisions, the appropriate indicator for public safety risk, as well as the cutoff points will need to be redefined, based upon a determination of an acceptable level of failure, by criminal justice professionals, for the lowrisk group.

CJA’s New ROR Recommendation System The cutoff scores for the CJA release on recognizance recommendation system were chosen by considering a number of criteria. The first criterion was to create subgroups of defendants which would have different average FTA rates, and as such represent different risk levels. It was agreed that the FTA rates for individual points comprising the low-risk group should not exceed the average FTA rate for the recommended category of the previous recommendation system.

The cutoff score for the moderate-risk category was guided by the criterion that its FTA rate should be considerably higher than the FTA rate for the low-risk category. The cutoff score for the high-risk category was selected to include defendants with the highest FTA rate.

-81-

The second criterion was setting a target proportion of defendants to be recommended for ROR. It was agreed that the proportion of defendants in the low-risk category of the new ROR recommendation system should be higher than that under the previous system, without increasing the FTA rate for that category. The classification of defendants under the new ROR recommendation system and their corresponding points are displayed in Table 23A. As shown by the table, CJA’s ROR recommendation system classifies defendants into four categories. The first two categories, A and B, and the fourth subcategory of category C are based on a defendant’s total score on the point scale. In order to be categorized as “Recommended for ROR,” a defendant needs seven or more points. This is CJA’s highest positive rating for ROR recommendation. Defendants with scores ranging from three to six points receive the second highest recommendation, “Moderate Risk for ROR.” The third and fourth categories of the recommendation system reflect policy exclusions, with the exception of C-4. The “Not Recommended for ROR” category includes defendants with a bench warrant attached to their RAP sheet (C-1), defendants charged with bail jumping (C-2), defendants with conflicting residence information (C-3) and defendants considered at high risk of FTA due to receiving a total point scale score lower than “3” (C-4). No recommendation is offered to defendants with a missing NYSID (D-1), defendants charged with a homicide (D-2) or those with an incomplete interview (D-3). To ease interpretation for the present report, the third and fourth categories were collapsed into one category, and labeled “high risk”.

Combined Risk Classification System Table 24 presents the distribution of defendants and their corresponding combined failure rates, when the cutoff points from the ROR recommendation system were applied to the 2001 atrisk sample of defendants, with point scale scores ranging from -9 through 10 points. As shown by the table, under the combined risk classification system, 48% of the defendants would be considered low risk for failure, 19% moderate risk and 33% high risk. Their combined failure rate would be 18%, 29% and 40%, respectively.

-82-

Table 23A CJA's New ROR Reommendation System

Recommendation Category

Points

A. Recommended for ROR

+7 Points to +12 Points

B. Moderate Risk For ROR

+3 Points to +6 Points

C. Not Recommended for ROR Bench Warrant Attached to NYSID Bail-Jumping Charge Conflicting Residence Information High Risk for FTA D. No Recommendation No NYSID Available For Information Only Interview Incomplete

-83-

+2 Points to -12 Points

Table 24 Combined Risk Classification System by Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample

Risk Category

Defendants Classified N %

Defendants With Failure N %

Low Risk (7 to 10 Points)

12172

48

2187

18

Moderate Risk (3 to 6 Points)

4814

19

1423

29

High Risk (2 to -9 Points)1

8416

33

3404

40

Total

25402

100

7014

28

1

This category also includes policy exclusions.

-84-

G. Comparing the Combined with the ROR Risk Classification System Prediction of Failure We compared the combined risk classification system presented in Table 24 with the CJA ROR recommendation system (Table 25) in terms of its ability to classify more defendants as low risk without increasing their failure rate. Since the two systems had different base rates, their failure rates could not be directly compared. To illustrate, when applied to the 2001 at-risk sample, the failure rate for the low-risk defendants under the combined risk classification system would be 18% (Table 24). In comparison, the failure rate (FTA rate) for the low-risk category of the ROR recommendation system would be 9% (Table 25). When comparing the two systems, this would suggest that the ROR recommendation system did a better job in terms of identifying low-risk defendants. However, that is not the case here. Under the combined risk classification system, the failure rate for low-risk defendants would decrease from the base rate of 28% to 18% (a 10 percentage point decrease), whereas for the ROR recommendation system, the failure rate would decrease from the base rate of 16% to 9% (7 percentage point decrease). This suggested that the combined risk classification system would perform slightly better than the ROR recommendation system with respect to classifying defendants on the basis of risk of failure. For this reason, we compared the failure rates for individual risk categories with the base rate for the system as a whole.

As shown by Table 25, when applying the ROR recommendation system to the 2001 at-risk sample, 42% of the defendants would be categorized as low risk. Their FTA rate would be 9%, which would be seven percentage points lower than the average FTA rate for the sample as a whole (16%). In comparison, as shown by Table 24, the combined risk classification system

-85-

Table 25 New ROR Recommendation System by Pretrial FTA First Quarter of 2001 Dataset At-Risk Sample

Risk Category

Defendants Classified N %

Defendants With FTA N %

Low Risk (7 to 12 Points)

10687

42

945

9

Moderate Risk (3 to 6 Points)

4985

20

691

14

High Risk (2 to -12 Points)1

9624

38

2339

24

Total

25296

100

3975

16

1

This category also includes policy exclusions.

-86-

would classify a somewhat higher proportion of defendants as low risk, 48% versus 42%.12 Under both systems, one-fifth of the defendants would be considered moderate risk (19% for the combined system and 20% for the ROR system), with slight differences in their failure rate.

To be specific, under the ROR system, 14% of the moderate-risk defendants would fail to appear pretrial, which was 2 percentage points lower than the average FTA rate for the at-risk sample.

Under the combined system, the combined failure rate for the moderate-risk group would be 1 percentage point higher than the average combined failure rate (29% versus 28%). Finally, the ROR system would classify 38% of the defendants as high risk. Their FTA rate would be 8 percentage points higher than the average FTA rate for the at-risk sample (24% versus 16%). In comparison, the combined risk classification system would classify a slightly lower percentage (33%) of defendants as high risk. It would increase the combined failure rate by 12 percentage points over the average failure rate (40% versus 28%).

To summarize, relative to the ROR recommendation system, the combined risk classification system would classify more defendants as low risk, while decreasing their failure rate over the average failure rate, for the sample as a whole. It would classify the same proportion of defendants as moderate risk while slightly increasing the failure rate over the sample’s average failure rate. Finally, when compared with the ROR recommendation system, the combined risk

12This

was mainly due to a shift in points assigned to some of the variables. Those variables included living at a New York City area address and having a history of prior FTA. The variable reflecting prior misdemeanor conviction would also account for the increase in the proportion of low-risk defendants under the combined system.

That variable was only included in the combined system—defendants with a prior misdemeanor conviction would lose one point, whereas defendants with no such conviction would earn one point. Expecting someone at arraignment was included in the ROR recommendation system only—one point would be deducted if the defendant did not expect a relative or friend at arraignment. Alternatively, defendants expecting someone at arraignment would score one point on the scale. One may argue that since the two variables shared the same points, their presence or absence would not affect a defendant’s total score on the point scale. Alternatively, defendants would earn or lose one point irrespective of which variable was included in the model. Although that was the case, the proportion of defendants with no prior misdemeanor conviction was somewhat higher than the proportion of defendants who expected someone at arraignment in the 2001 at-risk sample (71% versus 62%). Therefore, the proportion of defendants earning one point for not having a prior misdemeanor conviction would be higher under the combined system than the proportion of defendants scoring one point for expecting someone at arraignment under the ROR system.

-87-

classification system would classify fewer defendants as high risk, and increase their failure rate considerably.

-88-

Prediction of FTA The two risk classification systems were compared with respect to their ability to predict pretrial FTA. Table 26 presents FTA rates for both systems. As can be seen, the two systems are quite similar with respect to their FTA rates, with the exception of the moderate risk group.

Under both systems, 9% of the low-risk defendants and 24% of high-risk defendants would fail to appear for at least one scheduled court appearance. For the moderate-risk group, the FTA rate would increase from 14% under the ROR recommendation system to 17% under the combined risk-classification system. In sum, the proportion of defendants failing to appear is about the same when the two types of failure are combined.

Prediction of Pretrial Re-Arrest We also compared the two risk classification systems with respect to their ability to predict the likelihood of pretrial re-arrest. As mentioned previously, CJA’s ROR recommendation system was derived from a model that did not take into consideration a defendant’s likelihood of pretrial re-arrest. Table 27 presents the re-arrest rates for both systems.

As shown by the table, when applied to the 2001 at-risk sample, the re-arrest rates for the low-, moderate- and high-risk categories of the ROR recommendation system would be 11%, 15%, and 24, respectively. These percentages are quite similar to those projected for the combined risk-classification system. To be specific, under the combined system, 11% of low-risk defendants would be re-arrested pretrial. The re-arrest rates for the moderate- and high-risk defendants would be 18% and 26%, respectively.

Prediction of Combined Failure (FTA or Re-Arrest) Table 28 presents the two systems by pretrial failure (FTA or re-arrest). As shown by the table, the two systems were virtually similar with respect to their failure rates for the low risk and high risk categories. The failure rate for the moderate risk category of the ROR recommendation system was slightly lower than that for the combined risk-classification system (25% versus 29%).

-89-

Table 26 Risk Classification Systems by Pretrial FTA First Quarter of 2001 Dataset At-Risk Sample

Risk Category

Combined Risk System N with FTA % with FTA

New ROR System N with FTA % with FTA

Low Risk

1133

9

945

9

Moderate Risk

802

17

691

14

High Risk1

2056

24

2339

24

1

This category also includes policy exclusions.

-90-

Table 27 Risk Classification Systems by Pretrial Re-Arrest First Quarter of 2001 Dataset At-Risk Sample

Risk Category

Combined Risk System Re-Arrest Re-Arrest N %

ROR System Re-Arrest Re-Arrest N %

Low Risk

1337

11

1215

11

Moderate Risk

865

18

753

15

High Risk1

2218

26

2426

24

1

This category also includes policy exclusions.

-91-

Table 28 Risk Classification Systems by Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample

Risk Category

Combined Risk System N with Failure % with Failure

New ROR System N with Failure % with Failure

Low Risk

2187

18

1910

18

Moderate Risk

1824

29

1252

25

8416

40

3817

40

1

High Risk

1

This category also includes policy exclusions.

-92-

Summary The analysis presented in this section showed that community-ties variables, criminal-history indicators, graded offense type at initial arrest, demographic attributes, and case-processing characteristics were significantly related to pretrial failure. When developing a point scale, demographic attributes, case-processing characteristics, and graded offense type were dropped from the model for policy reasons. Points were assigned to all the significant variables. The resulting point scale ranged from -9 points to 10 points. Defendants scoring higher points were less likely to FTA or be re-arrested pretrial, whereas defendants scoring fewer points were more likely to fail. The point scale was used to classify defendants into different risk categories.

The combined risk classification system was compared with the ROR recommendation system with respect to the proportion of defendants in various risk categories, and their corresponding failure rates (combined failure rate for the former and FTA rate for the latter). The comparison suggested that relative to the ROR recommendation system, the combined risk classification system would classify more defendants as low risk, while decreasing their failure rate, over the average failure rate, for the sample as a whole. It would classify the same proportion of defendants as moderate risk while slightly increasing the FTA rate over the sample’s average failure rate. Finally, when compared with the ROR recommendation system, the combined risk classification system would classify fewer defendants as high risk, and increase their failure rate considerably.

Our analysis showed that the new ROR recommendation system, when applied to the 2001 at-risk sample, would also be capable of predicting pretrial re-arrest. This is not surprising, as the combined risk classification system and the ROR recommendation system shared similar variables. Consequently, the ROR system, although only designed to predict risk of pretrial failure to appear, would identify both types of failure.

-93-

SECTION IV SUMMARY AND CONCLUSIONS The analysis presented in the current report was a continuation of our previous research on pretrial recidivism among New York City defendants. The current analysis was performed in two phases. In Phase I, we examined the likelihood of pretrial re-arrest for violent felony offenses. In Phase II, we developed models that would predict pretrial FTA and/or re-arrest. The findings are summarized as follows: When examining violent felony re-arrests, we focused on a sample of defendants who were re-arrested pretrial after an initial arrest during the first quarter of 2001. In that sample, 10% of the defendants were re-arrested for a violent felony offense. Re-arrests made for a D felony comprised about one half of the violent felony re-arrests. Assaults also comprised about half of violent felony re-arrests. Slightly more than one quarter of the defendants re-arrested for a violent felony offense were initially arrested for the same offense.

We developed a multivariate model to identify significant predictors of violent felony rearrests. The model showed that graded offense type at initial arrest and a defendant’s age were significantly related to the outcome variable. Defendants initially arrested for a violent felony offense were more likely to be re-arrested for the same offense than defendants with the mean affect of that variable. In contrast, defendants initially arrested for drug offenses were less likely to be re-arrested for violent felony offenses. These findings were consistent with those derived from prior research, which suggest that defendants are likely to re-commit the same offenses as those for which they were originally charged. Our analysis further showed that younger defendants were more likely to be re-arrested for violent felony offenses. As indicated by previous studies, the current research showed that the likelihood of re-arrest for a violent felony offense decreased with age. These findings could be instrumental to researchers and policy makers in conducting research and/or constructing predictive instruments assessing potential risk of committing violent crimes while on pretrial release.

-94-

The model predicting violent felony re-arrests was compared with the model that predicted pretrial re-arrest regardless of offense type and severity. The findings indicated that in the 2001 at-risk sample, having a telephone in the defendant’s residence, being engaged in a full-time activity, having criminal history, the type and severity of the initial arrest charge, and a number of demographic and case-processing characteristics were significantly related to pretrial re-arrest in general. None of these variables proved to be statistically significant predictors of violent felony re-arrests, with the exception of graded offense type and age at initial arrest. This suggests that re-arrest in general and re-arrest for violent felony offenses are empirically distinct events and need to be addressed separately.

The findings from our study raise other research questions. For example, our analysis suggested that defendants initially arrested for a violent felony offense and who were re-arrested were more likely to re-commit the same offense. However, we do not know whether this applies to re-arrests made for violent offenses of misdemeanor or lesser severity. In other words, are the defendants initially arrested for violent misdemeanor offenses more likely to re-commit the same offenses? We would also like to know whether there is a relationship between violent felony offenses at initial arrest and re-arrest for violent misdemeanors. Alternatively, do the defendants initially arrested for a violent felony have the tendency to re-commit violent offenses of lesser severity? We recommend using the same dataset to explore those issues.

The second question relates to drug offenders. Our research showed that defendants initially charged with drug charges were less likely than other defendants to be re-arrested for violent felony offenses. However, we don’t know what offenses they were re-arrested for. We recommend further research to explore whether drug offenders are more likely to re-commit the same offense. We recommend use of the same dataset, so that the findings may be compared, and conclusions regarding the relationship between the charge at initial arrest and charge at re-arrest could be offered.

In the second phase of the analysis, the 2001 at-risk sample was used to examine combined failure—pretrial FTA or re-arrest. The combined failure rate for that sample was 28%. We -95-

developed a number of models that would predict both FTA and re-arrest. The final model contained community-ties variables, criminal-history indicators, charge related information, demographic attributes, and case-processing characteristics.

The model showed that when controlling for the effects of the other variables in the model, the probability of combined failure was lower among defendants who lived at a New York City area address, had a phone in their residence, and were engaged in a full-time activity. Defendants with no criminal history were less likely to FTA or be re-arrested than defendants with a criminal history. The graded offense type at initial arrest also proved to be a significant predictor of combined failure. The likelihood of failure was lower among defendants initially arrested for all types of felony offenses, with the exception of offenses categorized as “other” felonies.

Defendants initially arrested for “other” felonies were more likely to fail than defendants with the mean effect of that variable. The probability of pretrial FTA or re-arrest was higher for defendants initially arrested for all types of misdemeanor or lesser offenses, with the exception of those arrested for violent offenses. Defendants having been arrested for misdemeanor level violent offenses had no significant effect on the likelihood of combined failure. The probability of failure was higher among Hispanic, black and younger defendants. Finally, the likelihood of combined failure increased with case-processing time. The model was used to guide the development of a point scale that would predict combined failure.

When developing the point scale, we dropped a number of variables for practical and policy reasons. They included the top charge at initial arrest, demographic attributes, and caseprocessing characteristics. Points were assigned to all the significant variables. The point scale ranged from -9 points to 10 points. Defendants scoring higher points were less likely to FTA or be re-arrested pretrial, whereas defendants scoring fewer points were more likely to fail. The point scale was used to classify defendants into different risk categories. For comparison purposes, the cutoff scores were similar to those applied under CJA’s current ROR recommendation system. Those scores might change if the Statue permits consideration of public safety in making pretrial release/detention decisions. When applied to the 2001 at-risk sample, -96-

the combined risk classification system would classify 48% of the defendants as low risk for failure, 19% as moderate risk and 33% as high risk. Their combined failure rate would be 18%, 29% and 40%, respectively.

The combined risk classification system was compared with the ROR recommendation system with respect to the proportion of defendants in various risk categories, and their corresponding failure rates (combined failure rate for the former and FTA rate for the latter). For each system, the average failure rate for the system as a whole was used to determine increases and decreases in the failure rate for each risk category. The comparison suggested that relative to the ROR recommendation system, the combined risk classification system would classify more defendants as low risk, while decreasing their failure rate, over the average failure rate, for the sample as a whole. It would classify the same proportion of defendants as moderate risk while increasing the FTA rate over the sample’s average failure rate by one percentage point. Finally, when compared with the ROR recommendation system, the combined risk classification system would classify fewer defendants as high risk, and increase their failure rate considerably.

The two risk classification systems were compared with respect to their ability to predict pretrial FTA. The findings showed that the two systems were quite similar with respect to their FTA rates, with the exception of the moderate-risk group. When applied to the 2001 at-risk sample, the FTA rate for the moderate-risk category of the combined risk classification system would increase slightly. In sum, the proportion of defendants failing to appear did not alter when the two types of failure were combined.

The two risk classification systems were also compared with respect to their ability to predict pretrial re-arrest. The comparison showed that when applied to the 2001 at-risk sample, the rearrest rates corresponding to the low-, moderate- and high-risk categories of both systems would be quite similar. This was not surprising, as the two systems shared similar variables. More specifically, CJA’s new ROR recommendation takes into consideration six variables. These variables include living at a New York City area address, having a telephone in the defendant’s residence or having a cellular phone, expecting someone at arraignment, being engaged in a full-97-

time activity, pending cases, and prior FTA. Our previous analysis of pretrial re-arrest showed that four of those six variables significantly predicted pretrial re-arrest. Those variables included having a telephone in a defendant’s residence or having a cellular phone, being engaged in a fulltime activity, open cases, and prior FTA. Consequently, the new ROR system, although only designed to predict risk of pretrial failure to appear, would take into consideration risk of pretrial re-arrest.

To conclude, our findings suggest that although the current New York State statute does not permit the consideration of public safety in making pretrial release or decision decisions, CJA’s new ROR recommendation system would predict both pretrial FTA and pretrial re-arrest. We expect that defendants considered low risk for FTA by the ROR system would also be low risk for pretrial re-arrest. It should be noted that our findings are based on data that were collected prior to the implementation of the new ROR recommendation system. We recommend validating those findings on the post-implementation data.

-98-

REFERENCES

Ares, C., Rankin, A., and Sturz, H. 1963. “The Manhattan Bail Project: An Interim Report on the Use of Pre-trial Parole.” New York University Law Review 39: 67-95.

Austin, James, Barry Krisberg, and Paul Litsky. 1985. "The Effectiveness of Supervised Pretrial Release." Crime and Delinquency 31(4):519-537.

Belenko, Steven, Iona Mara-Drita, and Jerome E. Mcelroy. 1992. "Drug Tests and the Prediction of Pretrial Misconduct: Findings and Policy Issues." Crime and Delinquency 38(4):557-582.

Belenko, Steven, Jeffrey A. Fagan, and Tamara Dumanovsky. 1994. “The Effect of Legal Sanctions on Recidivism in Special Drug Courts.” The Justice System Journal 17(1):53-81.

Boudouris, James, Richard Thomasgard, and Teresa Lacsina. 1977. Pre-trial Decision Making in Iowa: Comparisions of Pre-trial Programs in Polk County and Linn County. State of Iowa Bureau of Correctional Evaluation.

Britt, Chester L. III, Michael R. Gottfredson, and John S. Goldkamp. 1992. "Drug Testing and Pretrial Misconduct: An Experiment on the Specific Deterrent Effects of Drug Monitoring Defendants on Pretrial Release." Journal of Research in Crime and Delinquency 29(1): 62-78.

Bureau of Justice Statistics. 1985. Examining Recidivism. Washington, D.C.: U. S. Department of Justice, Special Report, NCJ-96501.

--1987. Recidivism of Young Parolees. Washington, D.C.: U.S. Department of Justice, Special Report, NCJ-10496.

--1989. Recidivism of Prisoners Released in 1983. Washington, D.C.: U.S. Department of Justice Special Report, NCJ-116261.

--1991. Pretrial Release of Felony Defendants, 1988. Washington, D.C.: U.S. Department of Justice, Special Report, NCJ-127202.

--1992. Pretrial Release of Felony Defendants, 1990. Washington D.C.: U.S. Department of Justice. NCJ-139560.

--1993. Survey of State Prison Inmates, 1991. Washington, D.C.: U.S. Department of Justice, Special Report, NCJ-136949.

-99-

--1994. Pretrial Release of Federal Felony Defendants. Washington D.C.: U.S. Department of Justice, Special Report, NCJ-145322.

--2002. Recidivism of Prisoners Released in 1994. Washington D.C.: U.S. Department of Justice, Special Report, NCJ-193427.

Center for Governmental Research, Inc. 1983. An Empirical and Policy Examination of the Future of Pretrial Release Services in New York State. Rochester, New York: Center for Governmental Research, Inc. Clarke, Stevens H. and Anita Harrison. 1992. Recidivism of Criminal Offenders Assigned to Community Service Correctional Programs or Released from Prison in North Carolina in 1989. University of North Carolina at Chapel Hill: Institute of Government.

Clarke, Stevens. H., Jean Freeman, and Gary Koch. 1976. "Bail Risk: A Multivariate Analysis."

Journal of Legal Studies 5:341-85.

Cuvelier, Steven J. and Dennis W. Potts. 1993. Bail Classification Profiles Project Harris County, Texas. VA: State Justice Institute.

Eskridge, Chris W. 1979. "An Empirical Study of Failure to Appear among Accused Offenders: Construction and Validation of a Prediction Scale." Annual Journal of Pretrial Services 2:105-17.

Feeley, Malcolm M. and John McNaughton. 1974. A Pre-trial Process in the Sixth Circuit: A Quantative and Legal Analysis. New Haven: Yale University.

Glover, Anthony J.J., et.al. 2002. “A Comparison of Predictors of General and Violent Recidivism Among High Risk Federal Offenders.” Criminal Justice and Behavior 29(3):235-249.

Goldkamp, John S., Michael R Gottfredson, and Susan Mitchell Herzfeld. 1981. Bail Decisionmaking: A Study of Policy Guidelines. Washington, D.C.: U.S. Department of Justice, National Institute of Corrections.

Goodman, Rebecca. 1992. Hennepin County Bureau of Community Corrections Pretrial Release Study. Minneapolis, Minnesota: Planning and Evaluation Unit.

Hepburn John R. and Celesta A. Albonetti. 1994. “Recidivism Among Drug Offenders: A Survival Analysis of the Effects of Offender Characteristics, Type of Offense, and Two Types of Intervention.” Journal of Quantitative Criminology 10(2):159-179 Illinois Criminal Justice Information Authority. 1992. Pretrial Release and Crime in Cook County. Chicago, Illinois: Illinois Criminal Justice Information Authority.

Jones, Peter R. 1991. “The Risk of Recidivism: Evaluating the Public-Safety Implications of a

-100-

Community Corrections Program.” Journal of Criminal Justice 19:45-66.

Kirby, Michael P. 1977. The Effectiveness of the Point Scale. Washington, D.C.: Pretrial Services Resource Center.

Landes, William M. 1974. "Legality and Reality: Some Evidence on the Criminal Procedure." Journal of Legal Studies 3:287-300.

Lattimore, Pamela K., Christy A. Visher, and Richard L. Linster. 1995. “Predicting Rearrest for Violence Among Serious Youthful Offenders.” Journal of Research in Crime and Delinquency 32(1):54-83.

Lazarsfeld, Paul F. 1974. An Evaluation of the Pretrial Services. New York: Agency of the Vera Institute of Justice.

Liberton, Michael, Mitchell Silverman, and William R. Blount. 1992. “Predicting Probation Success for the First-Time Offender.” International Journal of Offender Therapy and Comparative Criminology 36(4):335-347.

Maxwell, Sheila Royo. 1999. “Examining the Congruence between Predictors of ROR and Failures to Appear.” Journal of Criminal Justice 27:127-141.

Petee, Thomas A. 1994. "Recommended for Release on Recognizance: Factors Affecting Pretrial Release Recommendations." The Journal of Social Psychology 134(3):375-382.

Rhodes, William. 1989. “The Career Criminal: Estimates of the Duration and Frequency of Crime Commission.” Journal of Quantitative Criminology 5(1):3-32.

Rhodes, William M., Raymond Hyatt, and Paul Scheiman. 1996. "Predicting Pretrial Misconduct with Drug Tests of Arrestees: Evidence from Eight Settings," Journal of Quantitative Criminology 12(3):315-348.

Roth, Jeffrey A. and Paul B. Wice. 1980. Pretrial Release and Misconduct in the District of Columbia. Washington, D.C.: Institute for Law and Social Research (INSLAW).

Schaffer, Stephen A. 1970. Bail and Parole Jumping in Manhattan in 1967. New York: Vera Institute of Justice.

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

NewYork: New York City Criminal Justice Agency.

--2000. Prediction of Pretrial Failure to Appear and Alternative Pretrial Release Risk

-101-

Classification Schemes in New York City: A Validation Study. NewYork: New York City Criminal Justice Agency.

--2002. Prediction of Pretrial Failure to Appear and an Alternative Pretrial Release Risk-Classification Scheme in New York City: A Reassessment Study. New York, NY: NYC Criminal Justice Agency.

--March 2003. Predicting the Likelihood of Pretrial Re-Arrest: An Examination of New York City Defendants. New York: New York City Criminal Justice Agency.

--June 2003. Predicting the Likelihood of Pretrial Re-Arrest Among New York City Defendants: An Analysis of the 2001 Dataset. New York: New York City Criminal Justice Agency.

Smith, Douglas A., Eric D. Wish, and G. Roger Jarjoura. 1989. "Drug Use and Pretrial Misconduct in New York City." Journal of Quantitative Criminology 5(2):101-126.

Sorin, Martin D., Mary A. Toborg, and David A. Pyne. 1979. "The Outcomes of Pretrial Release: Preliminary Findings of the Phase II National Evaluation." Annual Journal of Pretrial Services 2:141-57.

SPSS, Inc. 1999. SPSS Advanced Statistics User's Guide. Chicago, Illinois: SPSS, Inc. Straus, F. and J.J. Golbin. 1980. Suffolk Probation's Pre-trial Services Program: Report #3.

Suffolk County, NY: Suffolk County Probation Department.

Thomas, Wayne. 1976. Bail Reform in America. Berkeley, CA: University of California Press.

Toborg, Mary A., Martin D. Sorin, and Raymond H. Milkman. 1981. Pretrial Release: An Evaluation of Defendant Outcomes and Program Impact. Volume 1: Release Practices and Outcomes: An Analysis of Eight Sites. Washington D.C. The Lazar Institute.

Toborg, Mary A., Anthony M. J. Yezer, Philip Tseng, and B. Lynn Carpenter. 1984. Pretrial Release Assessment of Danger and Flight: Method Makes a Difference. VA: Lazar Management Group, Inc. VanNostrand, Marie. 2003. Assessing Risk Among Pretrial Defendants in Virginia. Virginia: Virginia Department of Criminal Justice Services.

Venezia, Peter S. 1973. Pre-trial Release with Supportive Services for High Risk Defendants: Three-year Evaluation of the Polk County Department of Court Services Community Corrections Project. Davis, CA: National Council on Crime and Delinquency.

Visher, Christy A. and Richard L. Linster. 1990. “A Survival Model of Pretrial Failure.�

Journal of Quantitative Criminology 6(2):153-184.

-102-

Weiner, N. A. 1990. At Risk of Rearrest for a Violent Crime: Predicting High-Stakes, High-Speed Recidivism and Developing Prediction Models in Two Birth Cohorts.

Washington D.C.: National Institute of Justice, Final Report, 86-IJ-CX-0052.

Wheeler, Gerald R. and Carol L. Wheeler. 1981. "Two Faces of Bail Reform: An Analysis of the Impact of Pretrial Status on Disposition, Pretrial Flight, and Crime in Houston."

Review of Policy Research (1):168-182.

Whitehead, John T. 1991. “The Effectiveness of Felony Probation: Results From an Eastern State.” Justice Quarterly 8(4):525-543.

Widom, C. S. 1989. “Child Abuse, Neglect, and Violent Criminal Behavior.” Criminology 27: 251-271.

Wilson, Robert A. 1975. A Practical Procedure for Developing and Updating Release on Recognizance Criteria. Newark, DE: University of Delaware, Division of Urban Affairs.

Wilson, Robert A. 1979. Evaluation of the Chester County Bail Agency. Pennsylvania: Commission on Crime and Delinquency.

Winterfield, Laura, Mark Coggeshall, and Adele Harrell. 2003. Development of an EmpiricallyBased Risk Assessment Instrument. Washington, D.C.: Urban Institute.

-103-

Appendix A Classification of Offenses Violent offenses include murder, non-negligent murder, negligent murder, forcible rape, robbery, aggravated assault, simple assault and kidnapping.

Property offenses include burglary, larceny-theft, arson, motor vehicle theft, possession of burglar’s tools, embezzlement, fraud, forgery and counterfeiting, unauthorized use of vehicle and stolen property.

Drug offenses include: A) sale/manufacture of controlled substance including opium, cocaine or derivatives, marijuana, synthetic narcotics and other dangerous drugs, and B) use/possession of controlled substance including opium, cocaine or derivatives, marijuana, synthetic narcotics and other dangerous drugs.

Public-order offenses include coercion, criminal mischief, fraud, gambling, offenses against public order, bribery, promoting prostitution, prostitution, patronizing prostitutes, offense against family, unauthorized use of vehicle, disorderly conduct, liquor-law violation, public narcotic intoxication, sex offenses (excluding forcible rape and prostitution) and use/possession of dangerous weapons.

The other category consists of all other offenses not included in the aforementioned categories but which are included in the UCR codes.

-104-

Appendix B Coding of Variables for Regression Models Variable

Codes

PRETRIAL RE-ARREST No Yes

0 1

PRETRIAL FTA No Yes

0 1

EMPL/SCHOOL/TRAINING Yes Yes Verified No No Verified Unresolved Conflict NYC AREA ADDRESS Yes Yes Verified No, No Verified Unresolved Conflict

Variable BOROUGH OF ARREST Brooklyn Manhattan Queens Staten Island Bronx TIME FROM ARRAIGNMENT TO DISPOSITION

1 2 3 4 5

1 2 3 4

SEX Female Male

0 1

ETHNICITY Black White Hispanic Other

1 2 3 4

COURT OF DISPOSITION Criminal Court Supreme Court

0 1

OPEN CASES No Yes

0 1

PRIOR FTA No Yes

0 1

-105-

Codes

1 2 3 4 5 Days

TYPE OF FIRST RELEASE Bail ROR

0 1

PRIOR MISDEMEANOR CONVICTIONS No Yes

0 1

PRIOR FELONY CONVICTIONS No Yes

0 1

AGE

Years

GRADED ARREST CHARGE FELONY-LEVEL CHARGES Violent Property Drug Public Order Other

1 2 3 4 5

MISDEMEANOR OR LESSER CHARGES Violent Property Drug Public Order Other

6 7 8 9 10

PRETRIAL FAILURE No Yes

0 1

Appendix B (continued) Coding of Variables for Regression Models Variable ARREST CHARGE SEVERITY A Felony B Felony C Felony D Felony E Felony A Misdemeanor B Misdemeanor/Other EXPECTING SOMEONE AT ARRAIGNMENT No Yes

Codes

1 2 3 4 5 6 7

0 1

-106-

Variable

Codes

ARREST CHARGE TYPE Violent Property Drug Public Order Other

1 2 3 4 5

TELEPHONE/CELLULAR PHONE Yes, Yes Verified No, No Verified Unresolved Conflict

1 2 3

Appendix C Defendant Characteristics by Violent Felony Re-Arrest First Quarter of 2001 Dataset Re-Arrest Sample Pretrial Re-Arrest Yes1 Defendant Characteristics

No2

Total

N

%

N

%

N

%

Gender Male Female Total

444 53 497

10 9

3787 540 4327

90 91

4231 593 4824

100 100

Ethnicity Black Hispanic White 3 Other Total

282 154 37 16 489

11 10 8 10

2229 1463 434 139 4265

89 90 92 90

2511 1617 471 155 4754

100 100 100 100

Age at Initial Arrest* 18 and under 19-20 years 21-24 years 25-29 years 30-34 years 35-39 years 40 and higher Total

118 69 96 71 52 47 43 496

18 15 12 11 8 7 5

542 396 725 580 628 627 832 4330

82 85 88 89 92 93 95

660 465 821 651 680 674 875 4826

100 100 100 100 100 100 100

166 154 57 24 96 497

11 10 9 12 10

1301 1418 569 183 859 4330

89 90 91 88 90

1467 1572 626 207 955 4827

100 100 100 100 100

118 363 481

12 10

830 3336 4166

88 90

948 3699 4647

100 100

445 52 497

10 14

3997 333 4330

90 86

4442 385 4827

100 100

Demographic Attributes

Case Processing Characteristics Borough of Initial Arrest Brooklyn Manhattan Queens Staten Island Bronx Total Type of First Release Bail ROR

Court of Disposition* Criminal Court Supreme Court Total

1 2 3

Yes = Violent Felony Re-Arrest No = All Other Re-Arrests OTHER includes Asian, American Indian, and others.

Page 1 of 4

-107-

Appendix C (continued) Defendant Characteristics by Violent Felony Re-Arrest First Quarter of 2001 Dataset Re-Arrest Sample

Pretrial Re-Arrest Yes Defendant Characteristics

No

Total

N

%

N

%

N

%

Verified Telephone Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

209 125 109 11 18 472

11 12 9 18 12

1771 958 1139 50 132 4050

89 88 91 82 88

1980 1083 1248 61 150 4522

100 100 100 100 100

Verified Full Time Employment/ School/ Training* Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

171 75 157 47 18 468

12 14 8 9 17

1291 479 1732 451 89 4042

88 86 92 91 83

1462 554 1889 498 107 4510

100 100 100 100 100

Verified Length of Residence of at Least 18 Months Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

222 108 105 21 15 471

10 12 9 10 13

1946 787 1024 193 104 4054

90 88 91 90 87

2168 895 1129 214 119 4525

100 100 100 100 100

Expects Someone at Arraignment* Yes No Total

208 257 465

12 9

1526 2503 4029

88 91

1734 2760 4494

100 100

Verified NYC Area Address Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

310 131 15 0 14 470

10 12 8 0 16

2790 987 184 18 71 4050

90 88 92 100 84

3100 1118 199 18 85 4520

100 100 100 100 100

Family Ties Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

184 107 142 25 13 471

11 12 9 11 14

1547 792 1422 208 79 4048

89 88 91 89 86

1731 899 1564 233 92 4519

100 100 100 100 100

Community-Ties Items

Page 2 of 4

-108-

Appendix C (continued) Defendant Characteristics by Violent Felony Re-Arrest First Quarter of 2001 Dataset Re-Arrest Sample Pretrial Re-Arrest Yes Defendant Characteristics

No

Total

N

%

N

%

N

%

First Arrest Yes No Total

112 360 472

11 11

939 3030 3969

89 89

1051 3390 4441

100 100

Prior Misdemeanor Convictions Yes No Total

183 292 475

9 11

1772 2332 4104

91 89

1955 2624 4579

100 100

Prior Felony Convictions Yes No Total

138 337 475

9 11

1373 2731 4104

91 89

1511 3068 4579

100 100

Prior Violent Felony Convictions Yes No Total

69 428 497

12 10

485 3711 4196

88 90

554 4139 4693

100 100

Open Cases Yes No Total

173 302 475

11 10

1448 2656 4104

89 90

1621 2958 4579

100 100

Prior FTA Yes No Total

191 306 497

10 11

1799 2531 4330

90 89

1990 2837 4827

100 100

Bench Warrant Attached to RAP Sheet Yes No Total

61 411 472

10 10

571 3545 4116

90 90

632 3956 4588

100 100

Criminal History

Page 3 of 4

-109-

Appendix C (continued) Defendant Characteristics by Violent Felony Re-Arrest First Quarter of 2001 Dataset Re-Arrest Sample Pretrial Re-Arrest Yes Defendant Characteristics

No

Total

N

%

N

%

N

%

Felony-Level Charges Violent Property Drug Public Order Other Total

139 56 49 37 15 296

20 12 7 14 8

560 421 652 227 165 2025

80 88 93 86 92

699 477 701 264 180 2321

100 100 100 100 100

Misdemeanor or Lesser Charges Violent Property Drug Public Order Other Total

64 26 39 27 44 200

9 12 7 7 8

641 194 531 371 527 2264

91 88 93 93 92

705 220 570 398 571 2464

100 100 100 100 100

Graded Offense Type*

* statistically significant at P=.05 or less Page 4 of 4

-110-

Appendix D Defendant Characteristics by Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample Pretrial Failure Yes Defendant Characteristics1

No

Total

N

%

N

%

N

%

Gender Male Female Total

6413 1001 7414

29 23

16042 3353 19395

71 77

22455 4354 26809

100 100

Ethnicity Black Hispanic White 2 Other Total

3774 2531 741 269 7315

31 28 21 18

8569 6583 2735 1215 19102

69 72 79 82

12343 9114 3476 1484 26417

100 100 100 100

Age at Initial Arrest 18 and under 19-20 years 21-24 years 25-29 years 30-34 years 35-39 years 40 and higher Total

928 696 1246 1047 1059 1048 1394 7418

30 31 30 27 27 29 23

2135 1548 2886 2771 2792 2595 4676 19403

70 69 70 73 73 71 77

3063 2244 4132 3818 3851 3643 6070 26821

100 100 100 100 100 100 100

2159 2463 976 329 1491 7418

27 31 22 31 27

5779 5442 3421 749 4012 19403

73 69 78 69 73

7938 7905 4397 1078 5503 26821

100 100 100 100 100

1461 5906 7367

26 28

4139 15175 19314

74 72

5600 21081 26681

100 100

6531 886 7417

28 28

17084 2319 19403

72 72

23615 3205 26820

100 100

Demographic Attributes

Case Processing Characteristics Borough of Initial Arrest Brooklyn Manhattan Queens Staten Island Bronx Total Type of First Release Bail ROR

Court of Disposition Criminal Court Supreme Court Total

1

All the variables were statistically significant at .05 (or less) level of significance, with the exception of: type of release, court of disposition, and expecting someone at arraignment.

2

OTHER includes Asian, American Indian, and others.

Page 1 of 4

-111-

Appendix D (continued) Defendant Characteristics by Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample

Pretrial Failure Yes Defendant Characteristics

No

Total

N

%

N

%

N

%

Verified Telephone Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

3197 1596 2049 95 220 7157

26 22 38 29 33

9275 5519 3331 229 448 18802

74 78 62 71 67

12472 7115 5380 324 668 25959

100 100 100 100 100

Verified Full Time Employment/ School/ Training Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

2438 825 2991 732 163 7149

25 19 35 27 29

7346 3408 5666 1972 400 18792

75 81 65 73 71

9784 4233 8657 2704 563 25941

100 100 100 100 100

Verified Length of Residence of at Least 18 Months Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

3446 1298 1900 345 177 7166

27 22 34 24 33

9250 4513 3615 1087 355 18820

73 78 66 76 67

12696 5811 5515 1432 532 25986

100 100 100 100 100

Expects Someone at Arraignment Yes No Total

2596 4529 7125

26 28

7270 11484 18754

74 72

9866 16013 25879

100 100

Verified NYC Area Address Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

4913 1640 438 45 124 7160

29 22 34 29 39

11989 5651 851 111 190 18792

71 78 66 71 61

16902 7291 1289 156 314 25952

100 100 100 100 100

Family Ties Yes Unverified Yes Verified No Unverified No Verified Unresolved Conflict Total

2711 1287 2628 401 132 7159

28 23 32 24 34

7123 4425 5709 1299 254 18810

72 77 68 76 66

9834 5712 8337 1700 386 25969

100 100 100 100 100

Community-Ties Items

Page 2 of 4

-112-

Appendix D (continued) Defendant Characteristics by Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample Pretrial Failure Yes Defendant Characteristics

No

Total

N

%

N

%

N

%

First Arrest Yes No Total

2009 5237 7246

18 34

8944 10047 18991

82 66

10953 15284 26237

100 100

Prior Misdemeanor Convictions Yes No Total

2847 4421 7268

38 24

4658 14311 18969

62 76

7505 18732 26237

100 100

Prior Felony Convictions Yes No Total

2251 5017 7268

38 25

3699 15270 18969

62 75

5950 20287 26237

100 100

Prior Violent Felony Convictions Yes No Total

800 6618 7418

39 27

1231 18172 19403

61 73

2031 24790 26821

100 100

Open Cases Yes No Total

2380 4888 7268

38 24

3841 15128 18969

62 76

6221 20016 26237

100 100

Prior FTA Yes No Total

2927 4491 7418

42 23

3997 15406 19403

58 77

6924 19897 26821

100 100

Bench Warrant Attached to Rap Yes No Total

1008 6283 7291

42 26

1405 17696 19101

58 74

2413 23979 26392

100 100

Criminal History

Page 3 of 4

-113-

Appendix D (continued) Defendant Characteristics by Pretrial Failure First Quarter of 2001 Dataset At-Risk Sample Pretrial Failure Yes Defendant Characteristics

No

Total

N

%

N

%

N

%

Felony-Level Charges Violent Property Drug Public Order Other Total

1001 746 1236 438 245 3666

24 26 30 25 31

3208 2135 2928 1299 536 10106

76 74 70 75 69

4209 2881 4164 1737 781 13772

100 100 100 100 100

Misdemeanor or Lesser Charges Violent Property Drug Public Order Other Total

1073 362 801 627 839 3702

24 34 39 30 26

3381 704 1256 1482 2375 9198

76 66 61 70 74

4454 1066 2057 2109 3214 12900

100 100 100 100 100

Graded Offense Type

Page 4 of 4

-114-