CJA
Research Brief
NEW YORK CITY CRIMINAL JUSTICE AGENCY, Inc.
No. 24
September 2010
A series of reports summarizing current research from CJA Executive Director, Jerome E. McElroy Director, Research Dept., Richard R. Peterson, Ph.D. Research Brief Editor & Deputy Director, Research, Mary T. Phillips, Ph.D. Graphics & Production, Raymond P. Caligiure Administrative Associate, Annie Su
CJA is a not-for-profit corporation that provides a variety of criminal justice services under a contract with the City of New York. CJA staff interview defendants arrested in New York City, make recommendations for pretrial release, and notify released defendants of upcoming court dates. Within the Agency, the Research Department conducts studies covering a broad array of criminal justice policy concerns. The Research Brief series summarizes the results of some of these studies. New York City Criminal Justice Agency, Inc. 52 Duane Street New York, NY 10007 PHONE: 646 213-2500 FAX: 646 213-2650 WEB: www.nycja.org © 2010 NYC Criminal Justice Agency, Inc.
PREDICTING POST-SENTENCING RE-ARREST By Qudsia Siddiqi, Ph.D. A defendant’s likelihood of The research summarized here re-arrest after release from a addressed the problem statisUsing jail or prison term is of intically by conducting sepalikelihood terest to Alternative-torate analyses for four of re-arrest Incarceration programs different groups of ofto target cases (ATIs) in targeting fenders, based on cases for interventhe length and for intervention: tion in court. type of senProblems and possibilities Defendants tence each rewith a low risk of re-arrest have the ceived. This approach provided greatest chance of success in pro- separate predictive estimates for grams that replace incarceration defendants facing a wide range with treatment or other activity, of sentences. without posing a threat to the com- The research identified case munity. and defendant characteristics as The difficulty for ATI programs sociated with a lower-than-averis that a potential client’s risk of age risk of re-arrest within each re-arrest is impossible to determine group. precisely, and may vary depending Following a summary of the on the nature of the case as well findings, some policy implicaas the defendant’s criminal history tions for ATI targeting are disand demographic characteristics. cussed. This Research Brief is adapted from Post-Sentencing Re-Arrest Among New York City Released Offenders: An Analysis Of The Fourth Quarter Of 2003 Dataset (2008, Revised 2010) by Qudsia Siddiqi, Ph.D., Project Director The full report is available on CJA’s web site: www.nycja.org/research/research.htm Research Assistance: Elyse J. Revere, Steve Mardenfeld Systems Programming: Barbara Geller Diaz, Aïda Tejaratchi Address comments to the author at qudsia_siddiqi@yahoo.com Please cite as follows, adapted to your citation style: Siddiqi, Qudsia. 2010. “Predicting Post-Sentencing Re-Arrest.” Research Brief series, no. 24. New York: New York City Criminal Justice Agency, Inc.
CJA
Research Brief Cases In The Data File
Data for the analysis were drawn from a cohort of arrests made October 1 through December 31, 2003. The analysis focused on the likelihood of rearrest after a convicted offender was released from jail or prison. The offenders in the sample were tracked from the date on which they finished the incarcerative portion of their sentence to September 30, 2007. The analysis was limited to offenders who were incarcerated for 6 to 18 months and were at risk for re-arrest. The 18-month cutoff point allowed all the incarcerated offenders in the sample to have a minimum of one year at risk. The final sample included 1,575 convicted offenders who were sentenced to 6 to 18 months and released at least a year prior to September 30, 2007. Of this number, 561 offenders received a Criminal Court sentence of 6 to 12 months. The remaining 1,014 were sentenced in Supreme Court.
Sentence Types
There are three types of incarcerative sentence in New York: definite — a single length of time no longer than one year; indeterminate — a range, with a minimum and maximum (for example, 1 to 3 years); and determinate — a single length of time greater than one year. All sentences in Criminal Court are definite, but Supreme Court sentences can be of any type. Cases in the study were divided into Criminal Court sentences of at least 6 months, and three categories of Supreme Court sentences: (1) 6 months jail plus probation; (2) definite sentence between 6 and 12 months, no probation; (3) indeterminate sentence with a minimum between 12 and 18 months or a determinate sentence no longer than 18 months.
Plan Of Analysis
Analyses were carried out separately for each sentence type category. First, selected criminal history and charge charactistics were examined. Then logistic regression analyses were performed that distinguished offenders who were re-arrested from those who were not re-arrested within each sentence type category. Due to the small number of cases in Staten Island, that borough was excluded from multivariate models. For each sentence type, the statistically significant predictors were used to identify subgroups of “low-risk” offenders. Finally, overall re-arrest rates and rates of re-arrest for a violent felony offense (VFO) were examined, comparing the low-risk subgroups with the sentence type category as a whole.
2
Time At Risk
Time at risk is the time the offender spends at large in the community when he or she is at risk for re-arrest. Time spent in jail is excluded from “time at risk.” All offenders in the research sample were at risk — they were released after serving a jail or prison term at least a year before the end of the tracking period. Since defendants who were released early had more time to commit crimes, all the multivariate statistical (logistic regression) analyses controlled for the amount of time at risk.
Interpreting Odds Ratios The results of logistic regression models are pre-
sented graphically by using bars to illustrate the size and direction of odds ratios. The odds ratio is a measure of the increase or decrease in the odds that the outcome will occur (in this case, re-arrest), given a specific characteristic, while controlling for other characteristics. A factor that significantly increases the odds of rearrest is represented by a blue bar above the line, indicating that the odds ratio is greater than 1. For example, Figure 2 shows that being male more than doubled the odds of re-arrest (odds ratio 2.304), compared to the odds for females (controlling for time at risk, criminal history, and charge type). A factor that significantly decreases the odds of rearrest is represented by a black bar below the line, indicating that the odds ratio is less than 1. For example, the odds of re-arrest for offenders convicted on a criminal facilitation charge were lower than the average odds for offenders convicted on all charges combined. In order to represent the odds ratio (0.358) visually so that the length of the bar shows its weight relative to other bars, we took its inverse value (1 / .358 = 2.793) and reversed the sign. Thus the black bar representing criminal facilitation in Figure 2 shows that the odds of no re-arrest were almost triple for offenders with this conviction charge, compared to the average for the group — another way of saying that the odds of re-arrest were lower.
Measures Of Re-Arrest
The re-arrests considered in this analysis were those that occurred after the offender had been released at the completion of a jail or prison sentence during the offender’s at-risk period (date of release to September 30, 2007). Re-arrest information was collected for any re-arrest during this period, as well as for a violent felony offense (VFO) re-arrest. The VFO re-arrests examined in the study included charges defined as a violent felony offense by the New York State Penal Law, as well as the felony charges of murder, arson, and kidnapping.
September 2010
CJA
Research Brief
Criminal Court Offenders With Sentences Of 6 To 12 Months in Jail Figure 1 Descriptive Characteristics
Selected Charactistics Figure 1 presents selected characteristics of the defendants sentenced in Criminal Court. These offenders (Criminal Court Offenders Sentenced To 6 – 12 Months) 99% had extensive criminal histories: almost all had been arrested previously (99%), with a majority having a prior 85% 83% 82% 76% conviction on a misdemeanor (85%) or a felony charge (76%). (The prior arrest measure reflects only unsealed arrests on the rap sheet at the time of the sample arrest.) 43% More than two-fifths (43%) had a pending (open) case, and more than four-fifths (83%) had a prior warrant. 28% The most severe charge at disposition was larceny for 28% of offenders, and this was the most common charge Prior Prior Prior Open Prior Larceny Retype. The re-arrest rate among released offenders was Arrest Felony Misd Case Warrant Disp. Arrested Conv. Conv. Charge quite high—82% were re-arrested at least once after re- N= (541) (543) (543) (543) (536) (561) (561) lease from jail and before the cutoff date.
Figure 2 presents the results of a logistic regression model predicting post-sentence re-arrest. The figure shows only the factors that had a statistically significant effect on re-arrest, although many other variables were also tested. Having a prior misdemeanor conviction, being convicted on a larceny charge, and being male were factors that significantly increased the probability of rearrest, as reflected by the blue bars (see “Interpreting Odds Ratios,” previous page). Offenders convicted on offenses involving criminal facilitation (Penal Law Article 115) were significantly less likely to be re-arrested, (black bar). (The results were obtained after controlling for the effect of the offender’s time at risk.)
Figure 2 Logistic Regression Model Predicting Re-Arrest (Criminal Court Offenders Sentenced To 6 – 12 Months) N=533 Prior misdemeanor Larceny conviction disposition charge 1.856
Odds Ratio
Predictors Of Re-Arrest
Research Brief #24
2.304
1.787 Criminal facilitation disposition charge
1.000
-2.793
Identifying Low-Risk Offenders
Based on the finding that offenders with prior misdemeanor convictions were more likely to be re-arrested, we examined re-arrest information for offenders at lower risk (those with no prior misdemeanor convictions) and compared their re-arrest rates with overall rates for the sentencing group. (The regression model also suggested a lower probability of re-arrest among females and offenders who were convicted on criminal facilitation charges, but there were too few in these subgroups—56 and 38 respectively—to examine separately.) Figure 3 shows that the re-arrest rate for offenders with no prior misdemeanor convictions was considerably lower than the average re-arrest rate for the group as a whole (72% versus 82%). Even so, 72% still represents a high risk of re-arrest. Moreover, the two samples did not differ much with respect to re-arrests on a violent felony charge (20% versus 21%).
Male
Figure 3 Comparing Re-Arrest Rates For “Low-Risk” versus All Offenders (Criminal Court Offenders Sentenced To 6 – 12 Months) No prior misdemeanor conviction (low risk) All offenders in the sentencing group
72%
82%
20%
N = (80) (561) Any Re-Arrest
21%
(80) (561) VFO Re-Arrest
3
CJA
Research Brief
Offenders With Supreme Court Sentences
Figure 4A Selected Charactistics (Offenders With Supreme Court Sentences) Group 1: 6 Months Jail Plus Probation
Supreme Court offenders were grouped into three categories, based on the length and type of sentence: Group 1: Supreme Court sentence of 6 months plus probation (N=156)
73%
Group 2: Supreme Court definite sentence of 6 to 12 months (N=529) Group 3: Supreme Court indeterminate sentence with 12- to 18-month minimum (N=329). This category also includes cases with a determinate sentence greater than 12 months up to 18 months, but there was only one case in the sample fitting this description (a determinate 18-month sentence). (See “Sentence Types,” page 2.)
33%
34%
36%
(148)
Prior Felony Conv.
(149)
Prior Misd Conv.
(149)
Open Case
(149)
Prior ReWarrant Arrested
(141)
(156)
Figure 4B Selected Charactistics (Offenders With Supreme Court Sentences) Group 2: 6 – 12 Months (Definite) 86% 69% 48%
47% 26%
Drug Prior Disp. Arrest Charge
N= (529)
(504)
Prior Felony Conv.
(504)
Prior Misd Conv.
(504)
50% 36%
Open Case
(504)
Prior ReWarrant Arrested
(497)
(529)
Figure 4C Selected Charactistics (Offenders With Supreme Court Sentences) Group 3: 12 – 18 Months Minimum (Indeterminate)
Offenders in Group 1 had a smaller proportion of defendants with each criminal history indicator, but even so, nearly three quarters had a prior arrest (73%, Figure 4A). In addition, 10% had a prior felony conviction, 28% had a prior misdemeanor conviction; 34% had an open case; and 36% had a prior warrant.
4
Drug Prior Disp. Arrest Charge
N= (155)
For all three groups of Supreme Court offenders, a drug charge was the most common charge at disposition. Drug offenses comprised 33% of the top disposition charges among cases in Group 1 (Figure 4A), 47% among Group 2 (Figure 4B), and 40% among Group 3 (Figure 4C).
The highest re-arrest rate was found among offenders in Group 2 (69%, Figure 4B), followed by offenders in Group 1 (65%, Figure 4A), and offenders in Group 3 (57%, Figure 4C). The lower re-arrest rate for offenders sentenced to 12 to 18 months of incarceration could be attributed to the relatively lower amount of time they spent at risk before the end of the tracking period.
28% 10%
Descriptive Characteristics
Offenders in Group 3 had the most extensive criminal histories: 89% had a prior arrest, 58% had a prior misdemeanor conviction, 58% had a prior felony conviction, 45% had a pending case, and 60% had a prior warrant (Figure 4C). Offenders in Group 2 had less extensive criminal histories on all of these measures, especially prior felony convictions: 26% of defendants in Group 2 had a prior felony conviction (Figure 4B), compared to 58% in Group 3 (Figure 4C).
65%
89% 58%
45%
40%
Drug Prior Disp. Arrest Charge
N= (329)
60%
58%
(311)
Prior Felony Conv.
(311)
Prior Misd Conv.
(311)
Open Case
(311)
57%
Prior ReWarrant Arrested
(307)
(329)
September 2010
CJA
Research Brief
Offenders With Supreme Court Sentences (continued)
Figure 5A Logistic Regression Model Predicting Re-Arrest (Offenders With Supreme Court Sentences) Group 1: 6 Months Jail Plus Probation N=138
Predictors Of Re-Arrest
Group 2: The odds of post-sentencing re-arrest were significantly higher among offenders with a prior arrest (compared to those with no prior arrest), among those convicted on burglary and related charges (Penal Law Article 140; compared to the average odds of re-arrest for all charge types), and among offenders initially arrested in the Bronx (compared to the average odds citywide), as illustrated by Figure 5B. Prior arrest and prior warrant were again almost interchangeable as predictors of re-arrest. For Group 2, the stronger predictor was prior arrest, so the model presented here includes prior arrest rather than prior warrant. The likelihood of no re-arrest was higher among offenders convicted on a weapon possession charge (Penal Law Article 265) and among those who reported living at their current address for less than 18 months. Group 3: The only factor that significantly increased the likelihood of re-arrest among Group 3 offenders was conviction on burglary or a related charge, compared to the odds for all charge types combined (Figure 5C). The likelihood of no re-arrest was significantly higher among offenders convicted on assault charges (Penal Law Article 120) and those initially arrested in Queens, compared to all boroughs combined.
3.674
4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.01.000 -0.5 -1.0 -1.5 -2.0 -2.5 -3.0 -3.5 -4.0
Odds Ratio
25 years of age or older
–3.623
Figure 5B Logistic Regression Model Predicting Re-Arrest (Offenders With Supreme Court Sentences) Group 2: 6 – 12 Months (Definite) N=513 Burglary disposition charge 4.811
Prior arrest 3.156
Odds Ratio
Group 1: The odds of post-sentencing re-arrest were significantly higher among offenders who had a history of warrants (compared to those with no warrant history) and lower among older offenders, as illustrated by Figure 5A. For older offenders, the black bar represents the greater likelihood of no re-arrest for offenders who were 25 years of age or older at the time of the sample arrest, compared to the odds for offenders aged 18 and under. Having a prior arrest was almost as good a predictor as having a warrant history. However, the two variables were so closely related that the effect of a prior warrant overshadowed the effect of a prior arrest when entered together in the same model (not shown).
Prior warrant
Bronx Weapon possession disposition charge
Length of residence less than 18 months
1.757
1.000
–1.964
–2.342
Figure 5C Logistic Regression Model Predicting Re-Arrest (Offenders With Supreme Court Sentences) Group 3: 12 – 18 Months Minimum (Indeterminate) N=317 Burglary disposition charge 5.450
Odds Ratio
To identify statistically significant predictors of postsentencing re-arrest, multivariate logistic regression models were developed separately for each group of Supreme Court offenders. All results presented here were obtained after controlling for the effect of the offender’s time at risk (see “Interpreting Odds Ratios,” page 2).
Assault disposition charge
Queens
1.000
–1.848 –3.390
Research Brief #24
5
CJA
Research Brief
Offenders With Supreme Court Sentences (continued) Identifying Low-Risk Offenders Group 1: Based on the model presented in Figure 5A, two subgroups of low-risk offenders were identified: those who had no prior warrants (or had not been arrested previously); and those who were 25 or older at the initial arrest. Figure 6A displays their re-arrest rates, relative to all offenders in Group 1. Those in the low-risk categories had lower overall re-arrest rates (57% and 54% respectively), compared to the rate for Group 1 as a whole (65%). Re-arrest on a violent felony offense (VFO) was not much affected by whether an offender was categorized as “low risk” based on priors: the VFO re-arrest rate for this group was 18%, similar to the overall rate (20%). However, the VFO re-arrest rate for older offenders was relatively small (8%). Group 2: Based on the model presented in Figure 5B, two subgroups of low-risk offenders were identified: those who had not been arrested previously or had no prior warrants and those who were convicted on a weapon charge. The model also suggested lower likelihood of re-arrest among offenders who had lived at their address for less than 18 months, but this finding was somewhat surprising as prior research on pretrial misconduct suggests higher failure rates associated with weaker community ties. Until this finding is verified by further research, this factor will not be considered in our identification of low-risk offenders. Results for the low-risk offenders in Group 2 are presented in Figure 6B. Relative to the overall re-arrest rate for Group 2 (69%), offenders in the low-risk categories were rearrested at a lower rate (61% and 55% respectively). However, the VFO re-arrest rates for the two subgroups (21% and 19% respectively) were quite similar to the overall rate (21%). Group 3: The model presented in Figure 5C suggested that the offense type of the conviction charge and being arrested in Queens affected likelihood of re-arrest, but because of the small number of cases in each low-risk group, these offenders were not investigated further.
6
Figure 6A Comparing Re-Arrest Rates For “Low-Risk” versus All Offenders (Offenders With Supreme Court Sentences) Group 1: 6 Months Jail Plus Probation
No prior warrant or no prior arrest 25 years of age or older
]
“low risk” subgroups
All offenders in Group 1
57%
65% 54%
20%
18% 8% N = (96) (61) (156) Any Re-arrest
(96)
(61) (156) VFO Re-arrest
Figure 6B Comparing Re-Arrest Rates For “Low-Risk” versus All Offenders (Offenders With Supreme Court Sentences) Group 2: 6 – 12 Months (Definite) No prior warrant or no prior arrest Convicted on weapon charge All offenders in Group 2
61%
]
“low risk” subgroups
69% 55%
21% 19%
N = (260) (73) (529) Any Re-arrest
21%
( 260) (73) (529) VFO Re-arrest
Summary: Despite somewhat lower re-arrest rates for “lowrisk” subgroups identified among the Group 1 and Group 2 Supreme Court offenders, all the offenders presented a considerable risk of post-sentencing re-arrest. More than half of the offenders in each “low-risk” subgroup were re-arrested after their release from jail or prison.
September 2010
CJA
Research Brief
Conclusions Our analysis identified prior misdemeanor convictions, prior arrests, and prior warrants as significant predictors of post-sentencing re-arrest. This was consistent with findings from other studies of recidivism, which suggest that offenders with prior criminal justice involvement are more likely to recidivate than offenders without such involvement. Our analysis also pointed to disposition charge as a significant predictor of re-arrest. This is also consistent with prior research on recidivism, which has suggested that the type of charge influences the likelihood of recidivism. Nevertheless, findings regarding specific types of charges are not consistent. For each sentence type, the subgroups of offenders — which were identified based on statistically significant predictors of re-arrest — had lower rearrest rates than the sentencing group’s average rearrest rate. Those rates were still quite high, ranging from 54% for the subgroup of older offenders with a Supreme Court sentence of 6 months plus probation (Group 1) to 72% for the “low-risk” offenders who were sentenced in Criminal Court and had no prior misdemeanor convictions.
Our inability to identify truly low-risk offenders was mainly due to very high re-arrest rates among all offenders in the sample. Therefore, although the re-arrest rates for the “low-risk” subgroups were considerably lower than the average re-arrest rates for all offenders with comparable sentences, they were still too high to categorize offenders in any of the subgroups as having a truly low risk of postsentencing re-arrest. All of the Criminal Court and Supreme Court offenders in our sample had extensive criminal justice involvement at the time of the initial arrest during the study period. The high post-sentencing rearrest rates for each sentencing group, as well as for the subgroups, suggest that the criminal behavior of offenders in the study sample did not stop after release from jail or prison. Alternatively, serving jail or prison time did not deter them from committing more crimes. In fact, more than one fourth of the total re-arrests were for a violent felony offense (not shown). This suggests that some of these offenders presented a serious threat to the safety of the community after their release.
Policy Implications Our research on Criminal Court and Supreme Court offenders identified a few factors as significant predictors of post-sentencing re-arrest. Of the statistically significant variables, top charge at disposition, prior arrests, and prior warrants predicted post-sentencing re-arrest for most of the Supreme Court offenders. When predicting post-sentencing re-arrests for Criminal Court offenders, the variable reflecting prior misdemeanor convictions also proved to be useful. Despite our inability to identify truly low-risk offenders, we suggest that these factors be taken into account when considering eligibility for options other than incarceration.
Furthermore, the large proportion of offenders in our sample with prior warrants suggests that these offenders may also present a risk for non-appearance, in addition to an increased risk of re-arrest. This finding may aid in determining which option would be an appropriate alternative to incarceration.
Research Brief #24
7
Research Brief from
CJA
No. 24 (September 2010) Predicting Post-Sentencing Re-Arrest Forthcoming:
No. 25 (January 2011) Adolescent Male Domestic Violence Offenders by Richard R. Peterson
Most recently published in this series:
No. 23 (May 2010) Making Bail In New York City (Phillips) No. 22 (January 2010) Young Male DV Offenders (Peterson) No. 21 (September 2009) Juvenile Offenders And Weapons (Gewirtz) No. 20 (May 2009) The CASES Day Custody Program (Solomon) No. 19 (January 2009) Pretrial Failure Among New York City Defendants (Siddiqi) No. 18 (September 2008) Bail, Detention, and Felony Case Outcomes (Phillips) No. 17 (May 2008) Pretrial Misconduct Among Domestic Violence Defendants (Peterson) No. 16 (January 2008) Predicting The Likelihood of Pretrial Re-Arrest For Violent Felony Offenses (Siddiqi) No. 15 (September 2007) The Risk of Re-Arrest For Serious Juvenile Offenders (Gewirtz) No. 14 (May 2007) Bail, Detention, & Nonfelony Case Outcomes (Phillips)
www.nycja.org/research/research.htm
Research Brief from
CJA
No. 24 (September 2010)
Predicting Post-Sentencing Re-Arrest The New York City Criminal Justice Agency, Inc. 52 Duane Street New York, NY 10007
TO: