Cost of Discretion: Judicial Decision-Making, Pretrial Detention, and Public Safety in New York City

By: Oded Oren, Chad M. Topaz, and Courtney Machi Oliva.
PDF version: Cost of Discretion.
Published: May 2023.

Executive Summary

Key Findings:

  1. An analysis of public pretrial data from 2020 - 2022 reveals that some New York City judges are disproportionately carceral, i.e., these judges are substantially more likely to order pretrial detention than their peers, even when accounting for factors such as the severity of the case and the defendant’s prior criminal history.

  2. The fourteen judges who exhibited the most carceral discretion compared to their peers are Felicia Mennin, Gerald Lebovits, Quynda Santacroce, Josh Hanshaft, Kerry Ward, Bruna DiBiase, Gerianne Abriano, Beth Beller, Phyllis Chu, Alan Schiff, Tara Collins, Derefim Neckles, Joseph McCormack, and Lumarie Maldonado-Cruz.

  3. These fourteen judges’ disproportionately carceral decisions over 2.5 years resulted in an estimated 580 additional people detained, 154 additional years of pretrial detention, and over $77 million of additional costs borne by New York City taxpayers

Key Recommendations:

  1. Closer scrutiny of judges’ bail decisions is crucial because of the link between pretrial detention and increased recidivism rates, exacerbated racial disparities, and influence over case outcomes.

  2. New York (and other jurisdictions) must evaluate whether judicial discretion should be constrained given that legislative efforts to reform bail have not prevented some judges from exercising discretion in disproportionately carceral ways.

  3. New York lawmakers should consider the following approaches to constraining disproportionately carceral judges:

  • Making additional judge-level data publicly available to all New Yorkers.
  • Removing disproportionately carceral judges from overseeing criminal cases.
  • Limiting judges’ discretion to detain, including by mandating release from detention upon the preparation of a release plan by holistic teams of experts.

Introduction

Lower court judges are powerful actors in the criminal legal system. Although recent policy and political debates about criminal legal system reforms focus on prosecutors and law enforcement, judges should not be overlooked. In criminal cases, their power and discretion are often unfettered and rarely subject to review. Some judges exercise their power and discretion more punitively than their peers, opting to detain defendants where their peers would opt for release, supervision, or rehabilitation. These disproportionately punitive judges—whom we refer to as disproportionately carceral judges—have largely been left out of the discourse about reforming the criminal legal system.

This report aims to expand the debate about the criminal legal system to include the role that judges can and should play in reforming the system. To achieve this, we analyze public data that details New York City judges’ decisions to detain or release defendants at arraignment. We employ statistical modeling techniques to analyze judges’ decisions while accounting for the differences in the cases that each judge arraigned.1 After identifying the most carceral judges, we estimate their impact in terms of additional people detained, additional time spent in detention, and monetary cost to New York City taxpayers.

The fourteen New York City judges who were the most carceral compared to their 231 peers are Felicia Mennin, Gerald Lebovits, Quynda Santacroce, Josh Hanshaft, Kerry Ward, Bruna DiBiase, Gerianne Abriano, Beth Beller, Phyllis Chu, Alan Schiff, Tara Collins, Derefim Neckles, Joseph McCormack, and Lumarie Maldonado-Cruz. Based on our model, the estimated impact of these judges’ disproportionately carceral decisions over 2.5 years amounts to 580 additional people detained, 154 additional years of pretrial detention, and over $77 million of additional costs borne by New York City taxpayers.

These judges’ carceral decisions negatively impact the criminal legal system and undermine public safety. Their decisions contravene the substantial research that has linked pretrial detention to an increased likelihood of recidivism and exacerbation of racial disparities. Relatedly, research has also shown that the decision to detain has an outsized effect on criminal case outcomes. All told, this research suggests that disproportionately carceral judges are making decisions that harm public safety, increase racial inequality, and contribute to a legal system that metes guilt and punishment based on factors extraneous to the individual’s conduct.

Our findings suggest three main pathways to curbing the power and discretion of these disproportionately carceral judges:

  1. Making additional judge-level data publicly available to all New Yorkers.

  2. Removing disproportionately carceral judges from overseeing criminal cases.

  3. Limiting judges’ discretion to detain, including by mandating release from detention upon the preparation of a release plan by holistic teams of experts.

In the words of the New York City Comptroller’s 2022 report on bail in New York City: “[Our] findings make clear…that additional judicial and prosecutorial accountability and oversight of the bail system are needed.”2

Arraignment and Bail Reform in New York

In 2020, New York passed a law requiring the publication of data related to individual criminal cases arraigned in state courts.3 Pursuant to this new law, New York’s Office of Court Administration (“OCA”) and Division of Criminal Justice Services (“DCJS”) began publishing this data in 2020. The data includes information about each case arraigned in New York City since January 1, 2020, including, but not limited to, information about the defendant’s arrest and conviction record, the charges against the defendant, the judge at arraignment, and the judge’s bail decision. This data has enabled researchers to analyze New York’s pretrial practices and to study the effects of bail reform and represents an important step towards increased transparency in the state’s criminal legal system.4

Our report uses data taken from cases arraigned in New York City criminal courts. Accordingly, before we detail our findings, it is helpful to understand how criminal cases are arraigned. A person arrested in New York City will appear in court and have their criminal case heard before a judge. This court appearance is the “arraignment,” and different judges preside over arraignments on any given day across New York City.5 At arraignment, the judge will decide whether to release or detain a defendant pending the outcome of their case.6 A judge may decide to detain a defendant on monetary bail or without monetary bail (remand). Alternatively, the judge may release the defendant under specific conditions (such as supervision) or without any conditions.

Before bail reform was enacted in January 2020, judges arraigning cases in New York could choose to set bail on any criminal case that came before them.7 However, when bail reform took effect, it removed some of this discretion: judges were no longer allowed to set bail in most misdemeanor and non-violent felony offenses. In other words, defendants charged with these offenses were arraigned and released without any monetary conditions.8 Critics of the state’s bail reform law pushed back immediately,9 and by April 2020 state lawmakers caved: they made more offenses (once again) bail eligible,10 thereby returning power and discretion to criminal court judges. In 2022, state lawmakers again expanded the list of offenses that were eligible for bail,11 and in 2023 they rolled back the bail reform law yet again.12

In practice, the judge at arraignments decides whether a defendant should be detained or released. As part of this determination, the judge decides if the case is bail eligible. If it is, the judge decides whether to detain the defendant. If the judge decides to detain the defendant, they can choose to order monetary bail: the defendant will be detained until the monetary bail amount is paid. Alternatively, the judge may order the defendant remanded: the defendant will be detained for the duration of the case and no monetary payment for release is allowed.

In theory, these judicial decisions are guided by New York state law, which imposes limitations on what factors a judge can consider when deciding whether to detain a defendant.13 One key limitation is that a judge is only supposed to consider a defendant’s risk of flight to avoid prosecution in deciding whether to detain; they are not permitted to consider whether a defendant is a danger to the community. State law also directs judges to consider information from all available sources when assessing whether a defendant is a flight risk.14 If the judge finds that the defendant poses a risk of flight, state law mandates that they impose the “least restrictive” condition(s) for assuring the defendant’s return to court.15 This mandate was removed as part of the 2023 rollbacks to the bail reform law, but it was the applicable law during the time period we analyze. In making the bail determination, the judge relies on several, yet limited, sources of information.16 Importantly, New York state law also includes a presumption of release: a judge should detain a defendant only if other means of securing the defendant’s presence in court, such as non-monetary conditions, are found insufficient.17 If a judge decides not to remand the defendant and instead opts to set monetary bail, the law requires them to consider the defendant’s financial circumstances and the ability to pay bail when setting a bail amount.18

While appellate courts may, on rare occasion, modify a detention decision,19 lower court judges have virtually unchecked discretion in bail eligible cases to release a person with or without conditions, set bail, or detain without bail (remand).

Results

Our statistical model uses the OCA/DCJS data to study judges’ bail decisions: whether each judge chooses to order detention (remand or set monetary bail) or to release the defendant. The statistical model controls for differences in the types of cases, including the severity of the charges and the defendant’s history of contacts with the criminal legal system. The model thus identifies the most carceral judges: those who are more likely than their peers to order detention. A detailed discussion of our methodology and data appears in the Methodology section of this report.

Of the 245 judges whose decisions we analyze, our statistical model identifies fourteen (~6%) judges as the most carceral as compared to their peers. Our methodology identifies additional judges with disproportionately carceral tendencies beyond these fourteen. However, in this report, we choose to apply stringent criteria to report only the most carceral of those judges.

Tables 1A–1D report our primary results.

Table 1A: The Most Carceral Judges20

We next calculate the estimated impact of these judges’ exercise of disproportionate carcerality in terms of detention and cost to the New York City taxpayer. The additional people detained by a judge in the list are calculated by subtracting the number of people that a theoretical mean judge would have detained from the number of people that the judge in question actually detained (per Table 1A). To calculate the additional days/years of detention, we use a mean length of detention of 97 days, based on the most recent annual data published by the Data Collaborative for Justice.21 To calculate the additional cost to the taxpayer, we multiply the number of additional days of detention by the cost per day of detention per individual, $1375, as reported by the NYC Comptroller.22

Table 1B: Most Carceral Judges—Estimated Impacts (January 1, 2020, to June 30, 2022)23

The estimated number of people impacted (Column 2) is dependent on the number of cases that the judge had arraigned: the more cases they arraigned, the more people they had detained compared to the theoretical mean judge.

Overall, we estimate that these fourteen judges detained an additional 580 people, resulting in 56,260 additional days, or 154 additional years, of detention. In monetary terms, we estimate that the additional detention ordered by these judges resulted in a cost of over $77 million borne by New York City taxpayers

The previous estimate only accounts for the immediate costs of pretrial detention and the costs generated by New York City’s Department of Corrections. The true impact of these judges on incarceration and taxpayer dollars is even greater because pretrial detention has been linked to increased rates of conviction and lengthier sentences. For example, one study finds that pretrial detention (defined as detention for a minimum of three days) is linked to a 42% increase in sentence length.24 Moreover, in 2020, New York State had an average minimum felony sentence of 49 months,25 and a cost of $315 per day per incarcerated individual.26 These figures indicate that the judges’ additional impact on sentence length and prison costs would be substantial. If OCA/DCJS were to make additional data available publicly, a more precise calculation of costs would be possible.

Table 1C reports general information about the fourteen judges from publicly available sources, including the New York court system’s Judicial Directory, the New York City Mayor’s Office’s website, Trellis and Ballotpedia.27

Table 1C: Most Carceral Judges—General Information28

Mennin and Lebovits, as relatively senior judges, are less likely to work in arraignments, a fact reflected by the lower number of bail eligible cases that they each arraigned. Consequently, their disproportionate carcerality will affect fewer people, leading to lower impact figures per Table 1B. Judges elected to civil court—as in the case of the judges named above—are often assigned to criminal court after their election for several years, before finally being moved to civil court.29

Table 1D reports the results and rankings assigned to each of the reported judges per the selection models described in our Methodology section.

Table 1D: Most Carceral Judges—Ranking and Related Values30

Tables 2-5 in the Methodology section provide summary statistics for our dataset. Table 6 in the Methodology section provides the results of our regression model. These results demonstrate that we capture the substantive analysis that the law asks of judges at arraignments. More serious cases and cases where a defendant had more prior contacts with the criminal system are significantly associated with higher probabilities of detention. Specifically, the coefficients for more serious cases (A/B felonies) and for more serious involvement with the criminal legal system (pending violent felonies; violent felony convictions) are higher than both the referenced (dropped) categories and less serious cases (C/D felonies, pending nonviolent felonies, and nonviolent felony convictions). These results are consistent with the plain reading of C.P.L. § 510.10(1) as to the types of information a judge could take into consideration when making bail determinations.31

Discussion

Our analysis shows that some judges are disproportionately carceral compared to their colleagues when they exercise discretion at arraignment. These findings have important policy implications because a judge’s detainment decision has not only an immediate impact on the defendant, but broader impacts on public safety and on the community. Thus, identifying these disproportionately carceral judges and attempting to measure their carceral impact is relevant both for policymakers concerned with reforming the criminal legal system and for the public, to assess whether these judges are helping or harming the community.

The Negative Effects of Pretrial Detention on Public Safety and Criminal Case Outcomes: Evidence from Multiple Jurisdictions

Research has linked pretrial detention with an increased risk of recidivism.32 A recent study of almost 1.5 million cases in Kentucky from 2009–2018 found that even a short period of pretrial detention is associated with increased recidivism.33 Likewise, data from Philadelphia and Pittsburgh showed a 6–9 percent increase in recidivism among those detained pretrial.34 Moreover, the increase in recidivism associated with pretrial detention does not ameliorate over time. Data from Harris County, Texas, showed that being detained pretrial was associated with a 30% increase in picking up new felony charges and a 20% increase in being charged with new misdemeanors for defendants as long as 18 months after the bail determination in their earlier case.35

In addition, research undermines the oft-repeated claim that bail reform harms public safety. Using New York City data from 2017 to 2022, researchers found that New York bail reform’s elimination of bail for most misdemeanor and nonviolent felony charges reduced recidivism.36 Additionally, there was no impact on recidivism, i.e., it did not increase or decrease, when judges imposed non-monetary conditions of release instead of money bail.37 Similar conclusions have been found in other jurisdictions. For instance, in Cook County, IL, researchers found no increase in criminal activity or crime after the county’s Chief Judge issued new bail procedures that led to an increase in pretrial release.38 In Philadelphia, the District Attorney implemented bail reform and released data that showed no increase in recidivism rates post reform.39 The same conclusion is reached in research from multiple states (New Mexico, Kentucky, New Jersey), counties and cities, and the federal criminal legal system.40 Literature reviews by the Center for Court Innovation and New York City Comptroller found little change in rearrest rates before and after bail reform.41 Even the New York Post “debunk[ed] claims of bail reform leading to [a] spike in gun violence” in a preliminary analysis of NYPD 2020 data.42

Research has also linked pretrial detention to increased rates of conviction, likelier sentences of imprisonment, and lengthier periods of incarceration when prison sentences are imposed. In New York City, a study by the Criminal Justice Agency found that defendants who were detained pretrial were “more likely to be convicted; if convicted they [were] more likely to be sentenced to incarceration; and if incarcerated, their sentences [were] likely to be longer.”43 Studies from other jurisdictions are consistent with these results. Research has shown that pretrial detention was associated with a 12% increase in the likelihood of conviction (Philadelphia and Pittsburgh) as well as a 42% increase in the length of sentences (Philadelphia).44 In Harris County, Texas, a study found that detained defendants were 25% more likely to plead guilty than released defendants, 43% more likely to be sentenced to jail, and more likely to have received jail sentences that were more than twice as long on average.45 In Kentucky, data showed that people released pretrial were “about one-half to three-quarters as likely to receive a sentence to prison or jail” compared to detained counterparts.46 Similar results were found in New Jersey47 and the federal system.48

These studies raise concerns that pretrial detention exerts outsized influence over a defendant’s decision to resolve a criminal case and the outcome of such resolutions. Specifically, these studies depict a criminal legal system in which guilt and punishment are being decided in part by factors extraneous to the alleged conduct itself.

Pretrial detention also reflects and perpetuates racial disparities in the criminal legal system. A study using 2010-2011 Manhattan criminal court data found that Black and Latinx defendants were more likely to be detained pretrial than similarly situated white defendants.49 Another study using New York City data found that approximately two-thirds of the average pretrial release rate disparity between white and Black defendants was due to racial discrimination.50 Similar studies using Miami and Philadelphia data found evidence of “substantial bias against Black defendants” in pretrial bail determinations,51 with a later study assessing that Black defendants were four percentage points less likely to be released compared to white counterparts.52 A multitude of other studies confirm that racial disparities exist in bail determinations across multiple jurisdictions.53

Policy Implications

Our findings demonstrate that New York City defendants run a substantially greater risk of pretrial detention when their cases are arraigned by the fourteen judges identified in our report. This finding is consistent with other research that has found that the identity of the arraigning judge is a strong predictor of the outcome of the bail determination.54 Given that decisions made at arraignments have important long-term implications for public safety, case outcomes, and racial equality, policymakers should ensure that arraignments are fair and that judges do not exercise their discretion in a disproportionately carceral manner that is antithetical to the laws that govern arraignments and bail. We now present possible policy responses to address the disproportionately carceral exercise of judicial discretion during arraignments that we have identified.

As an initial matter, policymakers must ensure increased transparency into the decisions made by individual judges. Release of more judge-level data by OCA/DCJS—which the two organizations maintain but have chosen not to make publicly available to New Yorkers55—would permit further scrutiny of judges. Without such data, policymakers, advocates, journalists, and the public will be unable to hold accountable this branch of the government.

One solution to addressing the use of disproportionately carceral judicial discretion can be found in the democratic process. With more data about judicial decision-making made publicly available, stakeholders handling judicial appointments and the public would be better situated to understand and evaluate individual judges. In addition, advocates would be able to use such increased transparency to educate the voting public about the importance of judicial elections and to appeal to stakeholders overseeing judicial reappointments. In this way, judges who are facing reappointment and reelection will be scrutinized and held accountable for the decisions they make.

A second solution lies in creating policies that will remove or limit the scope of judicial discretion at arraignment. As a starting point, our findings suggest that certain judges were disproportionately carceral: they exercised discretion in a manner that produced statistically measurable effects when compared to their counterparts. Moreover, these judges had done so despite legal rules that were supposed to constrain their discretion, such as the presumption of release and the “least restrictive” condition,56 as well as legal mechanisms to review and appeal their detention decisions.57 In the face of these, the existence, and hence the possibility, of disproportionate judicial carcerality suggests that these legal rules and review mechanisms did not effectively constrain some judges. Although these rules may have led certain judges to act less carcerally than they would otherwise, it seems reasonable to question whether that was the case with the most carceral judges. If these judges did not feel constrained by legal rules and review mechanisms when they made the detention decisions, countering the detrimental effects of their decisions on public safety requires new limitations on their discretion.

Further limitations on judicial discretion at arraignments may also be necessary because of the ongoing resistance within the New York judiciary to legislative reforms. An article by New York City public defender Angelo Petrigh documents explicit judicial opposition to state lawmakers’ reforms through “opinions, administrative adjustments, and routine court actions.”58 The developments that the article describes are concerning because they indicate that the judiciary is willing to exercise its discretion in a manner designed to undermine legislative intent and explicit statutory text. This suggests that one interpretation of our findings is that disproportionate carcerality is an expression of opposition to criminal legal system reform. If this is the case, then trying to guide judicial discretion away from disproportionate carcerality through additional due process procedures—e.g., hearings, presumptions, and bail review procedures—will not suffice to prevent certain judges from thwarting legislative intent. Instead, strict limitations on judges’ discretion are needed.

Finally, limiting judges’ discretion to detain would help neutralize an extrajudicial factor that may be influencing judges including, possibly, the judges identified here: negative media coverage. A 2010 study of New York City judges revealed that the “judicial nightmare” among almost every judge interviewed was the possibility of receiving negative coverage because of their decision to release defendants at arraignments.59 Even after bail reform, judges may still be wary of such coverage, as some outlets continue to criticize their release decisions.60 Indeed, negative media coverage of bail decisions has reportedly resulted in the transfer of a Bronx judge from criminal to civil court and spurred Governor Hochul to push for further rollbacks of bail reform.61 If tabloid coverage can influence administrative and executive decisions in this manner, then judges may choose to exercise their discretion more carcerally to avoid such coverage. Removing the decision to release from judges’ hands may blunt some harsh criticisms and reduce the influence of tabloids over judges’ bail determinations.

Thus, instead of implementing non-binding provisions or oversight procedures, policymakers should implement strict limitations on judicial discretion. One example of such a mechanism was introduced in New York’s bail reform law, which rendered some charges ineligible for bail. As another example, policymakers could mandate that when a defendant is detained but unable to pay bail, the defendant would be released per an individually-tailored release plan. Under such an approach, a group of professionals from different disciplines would work together to provide comprehensive, individualized, and coordinated services to the defendant. Such teams would be made up of social workers, supervision specialists, and mental health and substance abuse experts, when needed—but not judges.62 The defendant would then be released and their compliance with the release program would be monitored closely by their team. If the holistic team determined that the defendant failed to comply with the conditions, the defendant would be detained once again.

The 2023 bail law rollbacks, which removed the “least restrictive” mandate among other changes,63 will test our conclusion, namely, that such legal rules are not as effective at constraining carceral discretion as removing judicial discretion altogether. The impetus for the 2023 rollbacks was to expand judicial discretion to detain, a fact that Governor Hochul, who pushed for the rollbacks, did not hide.64 In 2023 and beyond, pretrial data will reveal whether the “least restrictive” mandate effectively curbed carceral judicial discretion, and whether its removal led to an expansion of such discretion.

Conclusion

New York City judges have long been exempted from having their individual decisions rigorously scrutinized outside of the appellate process. This is slowly changing, as policymakers recognize the importance of data transparency and as calls grow for greater judicial accountability. Given the growing body of research linking pretrial detention to negative impacts on public safety, greater scrutiny of New York City judges’ bail determinations is essential.

Such scrutiny may not come easily. Judges have historically defended their exercise of discretion in individual cases and have tended to resist legislative efforts to constrain their decision-making. Prior to New York’s bail reform, advocates argued that judicial discretion posed obstacles to a fairer implementation of bail laws,65 with one study citing New York judges who admitted to exercising discretion in ways that diverged from the legal standard.66 After the bail reform law went into effect, some judges spoke out against it—specifically, against the limitations it imposed on their discretion. Judge George Grasso characterized bail reform as a “significant threat” to public safety, lamenting that the “scope” of the limits that it imposed on judicial discretion were “breathtaking.”67 More recently, former Chief Administrative Judge of New York State Lawrence Marks expressed a similar opinion: “Many judges—we’ve got most of our judges who sit on criminal cases—would like more discretion in making determinations about bail and release of people accused of crimes.”68 Some judges went so far as to intentionally violate the bail reform law and set bail on ineligible cases.69

These remarks and actions raise concerns that judges may attempt to curb the effect and scope of legislation that limits their discretion and power.70 There is thus a need to further scrutinize judges, because their decisions impact—and could potentially undermine—public safety. In the words of the New York City Comptroller’s 2022 report on bail in New York City: “[Our] findings make clear…that additional judicial and prosecutorial accountability and oversight of the bail system are needed.”71

Methodology

The OCA/DCJS data contains information about each case arraigned in New York City courts from January 1, 2020, through June 30, 2022. We analyze this data and ask whether certain judges were disproportionately likely to make decisions resulting in detention. We develop an analysis that evaluates a judge’s choice to either release the defendant or order their detention.

The initial dataset we downloaded from OCA/DCJS contained data on 582,981 arraigned cases (which we refer to as observations) from January 1, 2020 through June 30, 2022. A description of this data is available on OCA’s website.72 We filter out observations that met one or more of the following conditions: the judge name was not the name of an individual (e.g., “Judge/JHO/Hearing Examiner, Visiting”); the case was arraigned outside of New York City; there did not appear to be a criminal court arraignment per the explanation in OCA’s Data Dictionary;73 or the case was terminated at arraignment. These filters reduced the data set to 222,879 observations.

We next filter out cases that were not bail eligible at the time that they were arraigned. Since the data does not contain information on whether the case is bail eligible, we determine bail eligibility by cross referencing the relevant statutory provisions with the information contained in the data.74 Our method of identifying bail eligibility cannot account for all bail eligible offenses: we eliminate 14,263 cases from our dataset where bail or remand was mandated, but where the information in the data was insufficient to establish the bail eligibility of the case. This insufficiency owes to the fact that some cases are bail eligible due to the nuanced circumstances of the case or the defendant that are not reflected in OCA/DCJS data or that require individualized judicial determination, such as the defendant’s failure to register as a sex offender or a defendant’s eligibility for persistent felony offender sentencing.75 After filtering for bail eligible cases, we are left with 49,634 observations.

We keep all cases with cash bail equal to zero, cash bail equal or greater than $100, or cases where the defendant was remanded. Effectively, this step removes all cases where the cash bail amount was $1-$99. One dollar bail is often set in cases when the defendant is detained on a warrant or on another case, and therefore detention does not accurately reflect a judicial choice to order pretrial detention. We assume that cash bail of $2–$9 indicates $1 bail set in multiple cases (and a data entry error, aggregating bail on multiple cases together) and that cash bail of $10–$99 is a data entry error. All told, this step removes 640 observations, leaving us with 48,848 observations.

We study one outcome variable, detained. This binary variable indicates whether or not the judge decided to detain the defendant. Specifically, it takes a value of one if the judge remanded or set bail, or a value of zero if the defendant was released with or without conditions. Summary statistics for this variable appear in Table 2A while summary statistics for the number of cases per judge appear in Table 2B.

Table 2A: Summary Statistics—Detained

Table 2B: Summary Statistics—Cases Per Judge

Our 16 explanatory variables (not including judge) can be categorized into three groups. The first group includes variables relating to the defendant’s contact with the criminal legal system and general background: gender, age group, supervision status (on probation/parole), the number of convictions at the time of arraignment (broken down to misdemeanor, non-violent felony, and violent felony), and whether the defendant had any pending cases at the time of arraignment (broken down to misdemeanor, non-violent felony, and violent felony).76 We expect judges to consider these variables (aside from gender) under C.P.L. § 510.10(1)(a) and (1)(c), and, more generally, as “available” information pursuant to C.P.L. § 510.10(1). Summary statistics for these variables appear in Table 3.

Table 3: Summary Statistics—Defendant Background Variables

The second group of explanatory variables includes case-specific legal factors: the severity of the top charge (violent/non-violent felony, misdemeanor); the level of the top charge offense (Felony A-E, Misdemeanor A-B); the category of the charge (robbery, sex offense, etc.); and the type of arrest (police custody or a ticketed arrest).77 We expect judges to consider these variables under C.P.L. § 510.10(1)(a), (b), (g), and (i), and, more generally, as “available” information pursuant to C.P.L. § 510.10(1). Summary statistics for these variables appear in Table 4.78

Table 4: Summary Statistics—Case Severity Variables

Finally, the third group of explanatory variables includes factors relating to the judge and to the case more generally: borough of arraignment, the arraignment’s month and year, and the number of months that had passed between the offense and the arraignment. While these variables should not play a role in bail determinations per a plain-text reading of C.P.L. § 510.10(1), we include them as controls. The date of the offense, especially, is relevant, since the time period encompassed by the OCA/DCJS data includes several time-dependent phenomena: the beginning of the COVID pandemic and the ensuing lockdowns in New York City; the seasonality of crime;79 and the extensive reporting on increase in violent crime in New York City and across the country.80 We expect that these would have affected judges’ decisions on bail. Summary statistics for these variables are in Table 5.81

Table 5: Summary Statistics—Other Variables

We use two analyses to identify the most carceral judges. The judges we report as disproportionately carceral are those who are ranked in the top 15 in each of these two analyses. The use of the second analysis serves as a robustness check to our identification of the judges.

Analysis 1: We fit an ordinary least squares (OLS) regression model for detained using all 16 explanatory variables described above. The regression results are reproduced below in Table 6. We group residual by judge and perform a t-test on each group to discern whether the mean residual for each judge differs significantly from zero. After applying the Bonferroni correction for multiple comparisons, we retain only judges with significant adjusted p-values (p < 0.05). We rank the judges in descending order based on their mean residual: a higher mean residual indicates higher carcerality. The use of OLS residuals to measure the effect of individual actors—and specifically of an individual judge—on a binary outcome has been adopted in leading research papers, including studies of pretrial detention.82

Analysis 2: We perform 245 logistic regressions, each of which uses all 16 explanatory variables (as before) as well as a dummy variable for each judge per regression. This approach compares each of the 245 judge’s detainment decisions to those of all other judges (grouped together). After adjusting p-values for the judge coefficients using the Bonferroni method, we retain only judges with significant p-values (p < 0.05). We rank the judges in descending order based on their coefficient: a higher positive coefficient indicates higher carcerality.

We identify the most carceral judges as those ranking in the top 15 in both analyses, yielding a total of fourteen judges.

We next estimate these judges’ impact compared to their peers. For each judge, we calculate their observed mean probability of detention from Analysis 1 based on all cases they arraigned in our dataset. We then calculate the mean probability of detention that the regression model indicates would have applied to these same cases based on the decisions by all 245 judges. To restate, we calculate the predicted probability of detention for each case arraigned by the judge based on our regression and then take the mean of these predicted probabilities. This mean predicted probability of detention we designate as the “theoretical mean judge” probability. Based on these two mean probabilities—the observed probability and the probability of the theoretical mean judge—we calculate how many more people would have been released had the judge’s cases been arraigned by the theoretical mean judge. We call our measure the Estimated Additional People Detained Due to Disproportionate Carcerality, which appears in the Results section.

We use this measure to calculate the estimated additional time and cost associated with each judge’s disproportionate carcerality. To calculate the estimated number of additional days/years of detention that each judge is responsible for, we multiply the number of additional people detained due to disproportionate carcerality by the average length of detention in days, 97, a number we obtain from research conducted by the Data Collaborative for Justice.83 To calculate the additional cost to the taxpayer, we multiply the number of additional days of detention by the cost per day of detention per individual, $1375, as reported by the NYC Comptroller.84

Our model is constrained by the data made public by OCA/DCJS. That data does not contain defendant information from the New York City Criminal Justice Agency, such as the algorithmic recommendation of whether to release the defendant;85 failure to appear record;86 and juvenile delinquent and/or youthful offender adjudication record.87 The OCA/DCJS data also does not indicate how many pending misdemeanor, non-violent felony, or violent felony cases a defendant has—only whether there was a pending case in each of these categories at arraignment.88 In addition, the data does not include information on whether the prosecutor requested bail and at what amount. This omission is important because a prosecutor’s request can “anchor” a judge’s bail determination.89 Overall, we expect that some of the missing information would have reduced the residuals in our results, although it would not necessarily change the relative ranking of judges’ mean residuals. In summary, had this data been available, it would have permitted us to reach a more precise understanding of individual judge effects and estimated impacts. Interestingly, it appears that both OCA and DCJS have access to at least some of the missing data but have chosen not to make it available publicly.90

Table 6: Analysis 1 Regression Results91

Authors

Oded Oren, Founder and Executive Director, Scrutinize
Chad M. Topaz, Co-Founder and Executive Director of Research, QSIDE Institute
Courtney Machi Oliva, Executive Director of the Zimroth Center, NYU School of Law

About

Scrutinize uses data-driven advocacy to promote judicial accountability. We believe state court judges are the frontier of the struggle for decarceration and for a more just and equitable criminal legal system. Scrutinize uses data to identify the individual judges who drive the criminal legal system’s most extreme negative impacts and advocates to hold them accountable.

Institute for the Quantitative Study of Inclusion, Diversity, and Equity (QSIDE) uses the cutting edge of data science and mathematical modeling to shine a light on societal wrongs and to promote justice. Existing as a network of quantitative scholars, social justice subject matter experts, activists, and other stakeholders, QSIDE has put its discoveries into practice by creating the first public dataset that allows scrutiny of federal criminal sentencing equity at the individual judge level; exposing the structural role universities play in perpetuating racial disparities in STEM; partnering with museum curators who want to diversify their collections; establishing action-to-justice laboratories on human trafficking, small town policing, and criminal sentencing disparities; and much more. QSIDE’s work has been covered in Black Enterprise, Forbes, The Guardian, Inside Higher Ed, MIT Technology Review, Mother Jones, Nonprofit Quarterly, People, Science, Smithsonian, and other venues.

Zimroth Center on the Administration of Criminal Law at NYU School of Law’s mission is to promote good government practices in criminal matters, with a special focus on the exercise of prosecutorial power and discretion. It pursues this mission through a mix of academic and public policy research, and litigation advocacy. The academic and public policy components analyze good criminal justice practices at all levels of government, produce scholarship and policy reports on criminal justice issues, and host symposia and conferences to address significant topics in criminal law and procedure and enhance the public dialogue on criminal justice matters. The litigation component uses the center’s research and experience with criminal justice practices to inform courts in important criminal justice matters, particularly in cases in which exercises of prosecutorial discretion create significant legal issues.

Acknowledgments The authors would like to express their gratitude to Lauren May for her contribution to the research conducted for this report.

Cost of Discretion by Oded Oren, Chad M. Topaz, and Courney Machi Oliva is licensed under CC BY-NC-SA 4.0

Footnotes

  1. By using a statistical model rather than summary statistics, we can more accurately measure the effect of individual judges. Cf. Anna Maria Barry-Jester, You’ve Been Arrested. Will You Get Bail? Can You Pay It? It May All Depend On Your Judge, FiveThirtyEight (June 19, 2018); George Joseph and Akash Mehta, Death at Rikers: How NYC Judges Fueled the Swelling Jail Population, NY Focus (September 27, 2021) ↩

  2. Tammy Gamerman and Alyson Silkowski, NYC Bail Trends Since 2019, Office of the New York City Comptroller, page 3 (March 2022) (emphasis added). ↩

  3. Judiciary Law § 216(5); Executive Law § 837-U; Pretrial Release Data, NY Courts (2023). ↩

  4. See, e.g., Olive Lu, Erica Bond, Preeti Chauhan, and Michael Rempel, Bail Reform in Action: Pretrial Release Outcomes in New York State, 2019-2020, Data Collaborative for Justice (May 2022); Olive Lu and Michael Rempel, Two Years In: 2020 Bail Reforms in Action in New York State, Data Collaborative for Justice (December 2022); Michael Rempel, Krystal Rodriguez, Tyler Nims, Joanna Weill, Madison Volpe, and Zachary Katznelson, Closing Rikers Island: A Roadmap for Reducing Jail in New York City, Center for Court Innovation (July 2021). ↩

  5. Like us, other researchers have treated defendants’ assignments to NYC judges at arraignments as random. See, e.g., David Arnold, Will Dobbie, and Peter Hull, Measuring Racial Discrimination in Bail Decisions, American Economic Review, 112(9): 2992-3038 (2022); Jon Kleinberg, Himabindu Lakkaraju, Jure Leskovec, Jens Ludwig, and Sendhil Mullainathan, Human Decisions and Machine Predictions, The Quarterly Journal of Economics, 133(1): 237–293 (2018). ↩

  6. Our methodology defines “detention” as a judge’s decision to set bail or remand a defendant into custody without monetary bail and stands in opposition to a judge’s decision to release a defendant, whether with conditions or without them. We have defined detention in this way because when bail is set, a person is held in custody until they are able to pay bail and are thus effectively detained. ↩

  7. C.P.L. § 510.10 (pre-January 1, 2020, version). ↩

  8. C.P.L. § 510.10(4) (January 1, 2020, version). ↩

  9. See, e.g., Christopher Robbins, Tabloids Sow More Bail Reform Confusion, Claiming Laws Set Man “Free To Rape”, Gothamist (February 3, 2020); Jesse McKinley, The Bail Reform Backlash That Has Democrats at War, New York Times (February 16, 2020). ↩

  10. C.P.L. § 510.10(4) (July 2, 2020, version). ↩

  11. Id. (March 9, 2022, version). ↩

  12. Shantel Destra and Rebecca Lewis, What made it into the 2024 New York budget?, City & State (May 1, 2023); NY Focus, Your One-Stop Guide to the 2023 New York State Budget (May 3, 2023). ↩

  13. C.P.L. § 510.10(1) (post January 1, 2020, version). See Insha Rahman, New York, New York: Highlights of the 2019 Bail Reform Law, Vera Institute of Justice, pages 6-8 (July 2019). ↩

  14. C.P.L. § 510.10(1) delineates some of the information, or factors, that a judge must consider. ↩

  15. C.P.L. § 510.10(1) (post January 1, 2020, version). See also Michael Rempel and Joanna Weill, One Year Later Bail Reform and Judicial Decision-Making in New York City, Center for Court Innovation, page 2 (April 2021) (discussing the least restrictive provision). The options available to the judge, from least restrictive to most restrictive, are release on recognizance (“ROR”); release under non-monetary conditions (supervision, programming, etc.); monetary bail; and remand. ↩

  16. These sources include the prosecutor and defense attorney’s bail applications, as well as the criminal complaint and the defendant’s criminal history. The only third-party source of information is produced by the New York City Criminal Justice Agency (“CJA”). The CJA report includes background details about the defendant, such as employment status, living arrangement, and phone number. The CJA report also includes an algorithmically-generated release recommendation for the judge. See Release Assessment, New York City Criminal Justice Agency (2023). This recommendation is generated by a statistical algorithm that takes, as input, various details about the defendant, such as warrant history and reachability by phone. The CJA algorithmic recommendation is not binding on the judge but rather, is available to them to consult. ↩

  17. See C.P.L. § 510.10(1). See also Rempel and Weill, One Year Later Bail Reform and Judicial Decision-Making in New York City, page 14 (discussing the presumption of release). ↩

  18. See C.P.L. § 510.10(1)(f). See also Rempel and Weill, One Year Later Bail Reform and Judicial Decision-Making in New York City, page 2 (discussing the ability to pay provision). ↩

  19. See, e.g., Jan Ransom, A Look Inside Rikers: ‘Fight Night’ and Gang Rule, Captured on Video, New York Times (January 12, 2022); Matter of State of N.Y. ex rel. Meyer v Brann, 186 A.D.3d 1163 (1st Dep’t 2020). ↩

  20. Probabilities are rounded to two decimal places. The calculation for column five uses unrounded probabilities. We then round up the result to the nearest integer to avoid overestimating the carcerality of each judge relative to the theoretical mean judge. ↩

  21. Michael Rempel, Decarceration in the Bail Reform Era: New York City’s Changing Jail Population Since 2019, Data Collaborative for Justice, page 20, figure 6.2 (December 2022). ↩

  22. NYC Department of Correction FYs 2011-21 Operating Expenditures, Jail Population, Cost Per Incarcerated Person, Staffing Ratios, Performance Measure Outcomes, and Overtime, Office of The New York City Comptroller, pages 2-3 (December 6, 2021). We arrived at the average length of detention by taking a weighted mean from the 2020-2021 data in this report. We are not aware of data for 2022. ↩

  23. Column two is the difference of columns six and five in Table 1A. Year estimates in column three are rounded to the nearest year. Column four is calculated from the days given in column three. ↩

  24. Megan Stevenson, Distortion of Justice: How the Inability to Pay Bail Affects Case Outcomes, The Journal of Law, Economics, and Organization 34(4):511–542 (November 2018). ↩

  25. Statistical Overview: Year 2020 Court Commitments, New York State Corrections and Community Supervision, page 12 (2023). ↩

  26. Jullian Harris-Calvin, Sebastian Solomon, Benjamin Heller, and Brian King, The Cost of Incarceration in New York State, Vera Institute of Justice (October 31, 2022). Note that Vera cites a different cost per day of detention than the one we use in our calculation because Vera’s number represents the cost of prison incarceration, not pretrial detention. ↩

  27. Judicial Directory, New York Courts; New York City Mayor’s Office; Trellis; Ballotpedia. ↩

  28. Santacroce and Schiff currently preside in Queens. They were presiding in Brooklyn during the period covered by the OCA/DCJS data. ↩

  29. For general information about how judicial appointments and elections work in New York, see Council on Judicial Administration, Judicial Selection Methods in the State of New York: A Guide to Understanding and Getting Involved in the Selection Process, New York City Bar (March 2014). ↩

  30. Mean residuals and estimates are rounded to two decimal places. ↩

  31. Of course, per the C.P.L., judges are to consider this information only insofar as they deem it pertinent to an evaluation of risk of flight and the necessary conditions to assure return to court, and not as it pertains to perceived dangerousness. ↩

  32. The destabilizing effects of pretrial detention may explain the increased likelihood to recidivate. Research has documented how pretrial detention disrupts people’s community and family ties, reduces their ability to support themselves and their families, and interferes with their ability to access housing, education, and any mental health or other medical treatment options they might need. Such destabilization could lead to recidivism. See, e.g., Nick Pinto, The Bail Trap, New York Times Magazine (August 13, 2015); Tiffany Bergin, René Ropac, Imani Randolph, and Hannah Joseph, The Initial Collateral Consequences of Pretrial Detention, Criminal Justice Agency, page 1 (September 27, 2022) (“Pretrial detention predicts job issues, loss of employment, and becoming homeless. Nearly a quarter of participants reported that they missed at least one important family event due to their arrest or pretrial detention.”); Isabelle Leipziger, The Collateral Effects of Criminal Orders of Protection on Parent Defendants in Cases of Intimate Partner Violence, Fordham Law Review 19:274-308 (2022). Some studies found that the recidivism may be offset by detention’s incapacitation effect, so that detention neither increases nor reduces the recidivism rate. See Will Dobbie, Jacob Goldin & Crystal S. Yang, The Effects of Pretrial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges, American Economic Review, 108(2): 201-240, 203 (2018); Emily Leslie and Nolan Pope, The Unintended Impact of Pretrial Detention on Case Outcomes: Evidence from New York City Arraignments, The Journal of Law and Economics 60(3): 529-557, page 529 (2017) (“pretrial detention lowers the probability of rearrest while cases are being adjudicated, [but] this reduction in criminal activity is mostly offset by an increase in recidivism within 2 years after disposition”). ↩

  33. Christopher Lowenkamp, The Hidden Costs of Pretrial Detention Revisited, Arnold Foundation (March 21, 2022). See also Christopher Lowenkamp, Marie VanNostrand, and Alexander Holsinger, The Hidden Costs of Pretrial Detention, Arnold Foundation (November 2013) (finding similar results with an earlier, smaller dataset). ↩

  34. Arpit Gupta, Christopher Hansman, and Ethan Frenchman, The Heavy Costs of high bail: Evidence from judge randomization, The Journal of Legal Studies 45(2):471–505 (2016). ↩

  35. Paul Heaton, Sandra Mayson, and Megan Stevenson, The Downstream Consequences of Misdemeanor Pretrial Detention, 69 Stanford Law Review, page 718 (2017). See also Amanda Agan, Jennifer Doleac, and Anna Harvey, Misdemeanor Prosecution and Recidivism, Cato Institute (December 8, 2021) (non-prosecution of misdemeanors reduces recidivism). ↩

  36. René Ropac and Michael Rempel, Does New York’s Bail Reform Law Impact Recidivism? A Quasi-Experimental Test in New York City, Data Collaborative for Justice, page 42 (March 2023). ↩

  37. Id. at 43. ↩

  38. Don Stemen and David Olson, Dollars and Sense in Cook County: Examining the Impact of General Order 18.8A on Felony Bond Court Decisions, Pretrial Release, and Crime, Safety and Justice Challenge (2020). ↩

  39. Oren M. Gur, Michael Hollander, and Pauline Alvarado, Prosecutor-Led Bail reform: Year One, Philadelphia District Attorney’s Office (February 2019). ↩

  40. Tiana Herring, Releasing people pretrial doesn’t harm public safety, Prison Policy Initiative (November 17, 2020); Molly Gill, Thousands were released from prison during covid. The results are shocking, The Washington Post (September 29, 2022). ↩

  41. Gamerman and Silkowski, NYC Bail Trends Since 2019, page 16; Krystal Rodriguez, Michael Rempel, and Matt Watkins, The Facts on Bail Reform and Crime in New York City, Center for Court Innovation (February 2021). See also Brennan Center for Justice, The Facts on Bail Reform and Crime Rates in New York State (2022); Ames Grawert and Noah Kim, Myths and Realities: Understanding Recent Trends in Violent Crime, Brennan Center for Justice (July 12, 2022). ↩

  42. Craig McCarthy, Carl Campanile, and Aaron Feis, NYPD’s own stats debunk claims of bail reform leading to spike in gun violence, New York Post (July 8, 2020); See also CBS New York, During Questioning In Albany, NYPD Commissioner Shea Backtracks On Bail Reform Law As Big Reason For Gun Violence, CBS News (October 14, 2021). ↩

  43. Mary Phillips, A Decade of Bail Research in New York City, New York City Criminal Justice Agency, page 127 (August 1, 2012). ↩

  44. Stevenson, Distortion of Justice: How the Inability to Pay Bail Affects Case Outcomes; Gupta, Hansman, and Frenchman, The Heavy Costs of high bail: Evidence from judge randomization. ↩

  45. Heaton, Mayson, and Stevenson, The Downstream Consequences of Misdemeanor Pretrial Detention, page 718. ↩

  46. Lowenkamp, The Hidden Costs of Pretrial Detention Revisited. ↩

  47. Meghan Sacks and Alissa Ackerman, Bail and Sentencing: Does Pretrial Detention Lead to Harsher Punishment?, Criminal Justice Policy Review 25(1):59–77 (2012). ↩

  48. Stephanie Holmes Didwania, The Immediate Consequences of Federal Pretrial Detention, American Law and Economics Review 22(1): 24-74 (2020); Oleson, Lowenkamp, Wooldredge, VanNostrand, and Cadigan, The Sentencing Consequences of Federal Pretrial Supervision, Crime & Delinquency, 63(3):313–333 (2017). ↩

  49. Besiki Luka Kutateladze and Nancy Andiloro, Prosecution and racial justice in New York County—Technical report, Vera Institute of Justice (January 2014). ↩

  50. Arnold, Dobbie, and Hull, Measuring Racial Discrimination in Bail Decisions. ↩

  51. David Arnold, Will Dobbie, Crystal Yang, Racial Bias in Bail Decisions, The Quarterly Journal of Economics, 133(4):1885–1932 (2018). ↩

  52. Dobbie, Goldin, and Yang, The Effects of Pretrial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges. ↩

  53. See, e.g., Summary of research studies related to racial disparities in pretrial detention, Prison Policy Initiative (October 2019); Meghan Sacks, Vincenzo A. Sainato, and Alissa R. Ackerman, Sentenced to Pretrial Detention: A Study of Bail Decisions and Outcomes, American Journal of Criminal Justice 40(3):661–81 (2015). ↩

  54. See, e.g., Leslie and Pope, The Unintended Impact of Pretrial Detention on Case Outcomes: Evidence from New York City Arraignments, page 540 (analyzing NYC data and finding that “arraignment judge is a stronger predictor of the defendant’s arraignment outcome than of the defendant’s observable characteristics going into arraignment”) (emphasis added); Gupta, Hansman, and Frenchman, The Heavy Costs of high bail: Evidence from judge randomization, page 497 (“We find substantial variation among individual magistrates in setting money bail, which suggests that the imposition of money bail, and therefore pretrial detention, is a function of the judge one receives”). ↩

  55. OCA/DCJS’s actions suggest that they would resist steps to voluntarily release judge-level data. Judiciary Law § 216(5) and Executive Law § 837-U mandated the release of pretrial data from January 1, 2020 and going forward, and OCA/DCJS complied with these state laws. Yet even after these laws were passed, OCA/DCJS were reluctant to release additional judge-level data. For instance, in 2022, OCA/DCJS made available pretrial data for 2019. See Pretrial Release Data, NY Courts. Notably, the 2019 data was anonymized to remove the names of individual judges, thus effectively blocking any attempt to analyze individual judges’ bail determinations prior to 2020. It appears that because the new laws did not require the publication of pretrial data prior to 2020, OCA and DCJS took the initiative to anonymize the data it released from a prior year. See also Shelby Davis, Jennie Brooks, and Fiona Maazel, A Look Inside the Black Box of New York State’s Criminal Justice Data, Measures for Justice, page 1 (March 2021) (finding that “the mechanisms for criminal justice data collection and release in New York State are broken”). ↩

  56. C.P.L. § 510.10(1) (post January 1, 2020, version). See also Rempel and Weill, One Year Later Bail Reform and Judicial Decision-Making in New York City, page 2 (discussing the least restrictive provision). The options available to the judge, from least restrictive to most restrictive, are release on recognizance (“ROR”); release under non-monetary conditions (supervision, programming, etc.); monetary bail; remand. ↩

  57. See C.P.L. § 510.20(1) (application to review securing order); C.P.L. § 530.30 (permitting de novo review by supreme court judge of criminal court judge bail decision); N.Y. C.P.L.R.§ 7002 (codifying a writ of habeas corpus procedure through New York State Court). ↩

  58. Angelo Petrigh, Judicial Resistance to New York’s 2020 Criminal Legal Reforms, Journal of Criminal Law and Criminology 113:1, page 1 (March 10, 2023). See also Sam Mellins, New York Judges Lock the Accused Out of Their Homes, Skirting Review Required by Landmark Ruling, Critics Charge, NY Focus (July 23, 2021) (discussing an OCA memo that guided judges to limit the scope of a hearing to scrutinize the issuance of orders of protection). As a matter of disclosure, a co-author of this report, Oded Oren, is a former colleague of Mr. Petrigh in a New York City public defender office. ↩

  59. Jamie Fellner, The Price of Freedom: Bail and Pretrial Detention of Low Income Nonfelony Defendants in New York City, Human Rights Watch, page 46-47 (December 2010). ↩

  60. See, e.g., Amanda Woods and Jorge Fitz-Gibbon, NYC judge went easy on gangbanger now charged in shooting near school, New York Post (March 9, 2023). The Post’s “judge” section often features articles criticizing judges for their bail decisions. See Judges, New York Post (2023). ↩

  61. George Joseph, Bronx Judge About To Be Removed From Criminal Cases After Bail Blowback, Sources Say, The City (April 20, 2023); Peter Sterne, Hochul cites media coverage in decision to roll back bail reforms, City & State (April 28, 2023). ↩

  62. Similar bail review procedures, albeit led by judges, have been recommended before. See, e.g., Krystal Rodriguez and Michael Rempel, Advancing Just and Equity in the New York State Courts: The Crucial Policymaker Role of the Next Chief Judge, Data Collaborative for Justice, page 2 (November 2022); New York Courts, Chief Judge Jonathan Lippman Announces Series of Reforms to Address Injustices of NY’s Current Bail System (2015). ↩

  63. Destra and Lewis, What made it into the 2024 New York budget?, City & State (May 1, 2023). ↩

  64. Joshua Solomon, Hochul proposes greater bail discretion for judges in ‘serious’ criminal cases, Times Union (January 11, 2023); Peter Sterne, Hochul cites media coverage in decision to roll back bail reforms. ↩

  65. Letter from Over 100 Community and Advocacy Groups Across New York State to Governor Andrew Cuomo (November 2017). See also Fellner, The Price of Freedom: Bail and Pretrial Detention of Low Income Nonfelony Defendants in New York City. ↩

  66. Specifically, judges admitted to evaluating a defendant’s dangerousness, rather than solely risk of flight, in making their choice on bail. Fellner, The Price of Freedom: Bail and Pretrial Detention of Low Income Nonfelony Defendants in New York City, page 46 (“judges acknowledged public safety does play a role in some cases.”). Both pre-reform and the reform bail laws permitted detention solely based on risk of flight, not dangerousness or public safety. In basing their decisions, at least in part, on dangerousness or public safety, New York judges violated New York state law. Nor are New York judges the only ones to violate the laws they are tasked with upholding. A comprehensive study of pretrial detention in the federal system, the University of Chicago’s Federal Criminal Justice Clinic found that, “federal judges routinely violate the very bail laws that they are tasked with upholding, which drives up detention rates, jails people for poverty, and exacerbates racial disparities.” Alison Siegler, Freedom Denied: How the Culture of Detention Created a Federal Jailing Crisis, Federal Criminal Justice Clinic, University of Chicago Law School (October 2022). ↩

  67. Andrew Denney and Bruce Golding, NYC judge slams bail reform as ‘significant threat to public safety’, New York Post (February 6, 2020). ↩

  68. Bernadette Hogan and Bruce Golding, NY judges agree with Mayor Adams on fixing bail-reform law, court official says, New York Post (January 25, 2022). See also Douglass Dowty, Judge: NY killers, burglars, robbers, bail jumpers must be freed under ‘dangerous’ bail law, Syracuse.com (January 3, 2020) (Judge calls bail reform law “dangerous” and claims legislature “usurped” the discretion of the judiciary with the reform law). ↩

  69. Judge Louis Nock set $20,000 cash bail and $30,000 bond on a man accused of vandalizing four synagogues although the charges were not bail eligible—a point expressed to the judge by the prosecutor at arraignments. Brittany Kriegstein, Clayton Guse, and Wes Parnell, Man responsible for vandalizing Bronx synagogues jailed after contentious arraignment; judge cites ‘gravity of the allegations’, New York Daily News (May 3, 2021). In Nassau, Judge McAndrews set $10,000 cash bail and $20,000 bond in a case that he acknowledged was not bail eligible, telling the defendant, “I don’t want you walking around my neighborhood.” Lorena Mongelli, Long Island judge ignores bail law, refuses release of ‘menace to society’, New York Post (January 28, 2020). In Cohoes City, Judge Marcelle set $100 cash bail on a man accused of driving with a suspended license despite recognizing that the charges were not bail eligible. Bernadette Hogan and Bruce Holding, Upstate judge challenges bail reform law with traffic case ruling, New York Post (February 5, 2020); People v. Johnston, 67 Misc. 3d 267 (N.Y. City Ct. 2020). ↩

  70. See generally Petrigh, Judicial Resistance to New York’s 2020 Criminal Legal Reforms. After the passage of a bail reform law in Illinois, a lower court judge ruled that the legislation violated the separation of powers because it “wrested” discretion from judges over pretrial detention. See Bryce Covert, With Illinois Cash Bail Case, Courts May Wall Themselves Off from Reform, Bolts (February 14, 2023). ↩

  71. Gamerman and Silkowski, NYC Bail Trends Since 2019, page 3. ↩

  72. Pretrial Release Data, NY Courts. ↩

  73. Pretrial Release Data Dictionary, NY Courts (2023). ↩

  74. See C.P.L. § 510.10(4). See also Krystal Rodriguez, New York’s Amended Bail Statute Pretrial Options, Center for Court Innovation (2020); Krystal Rodriguez, New York’s Bail Statute Pretrial Options Updated May 2022, Data Collaborative for Justice and Center for Court Innovation (May 2022). ↩

  75. See, e.g., C.P.L. § 510.10(4)(p) (failure to register as a sex offender when the defendant is designated a level three offender); C.P.L. § 510.10(4)(t) (any felony or A misdemeanor involving harm to an identifiable person or property while the defendant was released on a separate felony or class A misdemeanor involving harm to an identifiable person or property); C.P.L. § 510.10(4)(s) (a felony where the defendant qualifies for sentencing as a persistent felony offender). ↩

  76. An overwhelming body of scholarship has shed light on the disparities experienced by racially minoritized groups and individuals at every stage of the criminal justice process. Quantifying the role of race in arraignments in New York City is an avenue for future research. ↩

  77. For an explanation on ticketed arrests, or DATs, see Olive Lu, Erica Bond, and Preeti Chauhan, Desk Appearance Tickets in New York State in 2019, Data Collaborative for Justice (February 2021). ↩

  78. The OCA/DCJS data included a variable containing information on whether an order of protection was issued at arraignments, and whether the order was issued for a “Family” or “Non-Family” offense. This information would be available to the judge and could be considered “available” information within the meaning of C.P.L. § 510.10(1). However, the data quality for this variable is unreliable: there is missing information on whether an order was issued in many observations where an order of protection would have been issued. For example, there were 481 observations of robberies involving physical injuries and 971 robberies involving the use of a weapon that did not have order of protection data, although those types of cases involve orders of protection in almost all circumstances. ↩

  79. See, e.g., David McDowall, Colin Loftin, and Matthew Pate, Seasonal Cycles in Crime, and Their Variability, Journal of Quantitative Criminology 28:389–410 (2012); Martin A. Andresen and Nicolas Malleson, Crime seasonality and its variations across space, Applied Geography 43:25–35 (2013). ↩

  80. See, generally, Grawert and Kim, Myths and Realities: Understanding Recent Trends in Violent Crime; Ali Watkins, Violent Year in New York and Across U.S. as Pandemic Fuels Crime Spike, New York Times (December 29, 2020). ↩

  81. For readability, Table 5 presents the breakdown of date of arraignment by year instead of month and year. The OCA/DCJS only includes month and year of arraignment, not the exact date. ↩

  82. See, e.g., Dobbie, Goldin, and Yang, The Effects of Pretrial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges; Gordon Dahl, Andreas Ravndal Kostøl, and Magne Mogstad, Family Welfare Cultures, The Quarterly Journal of Economics, 129(4):1711–1752 (November 2014). We also perform a robustness check using a logistic model and deviance residuals, a more traditional modeling setting for a binary outcome, instead of OLS. With this model, the per-judge predicted mean probabilities of detention are within 0.01 of the values we obtain using the OLS model. The logit model also finds 11 of the 14 judges we report as ranked in the top 15. Three judges—Collins, McCormack, and DiBiase—do not make the top 15 in the logit model because they each have an adjusted p-value smaller than 0.05. We nevertheless report their names since they are ranked in the top 15 by the standard econometric approach (Analysis 1) and our robustness check (Analysis 2). ↩

  83. Rempel, Decarceration in the Bail Reform Era: New York City’s Changing Jail Population Since 2019, page 20, figure 6.2. ↩

  84. NYC Department of Correction FYs 2011-21 Operating Expenditures, Jail Population, Cost Per Incarcerated Person, Staffing Ratios, Performance Measure Outcomes, and Overtime, Office of The New York City Comptroller, pages 2-3. We arrived at the average length of detention by taking a weighted mean from the 2020-2021 data in this report. We are not aware of data for 2022. ↩

  85. See C.P.L. § 510.10(1)(a) (“activities and history”). The New York City Criminal Justice Agency is a nonprofit organization that conducts interviews of all people brought to arraignments from police custody. ↩

  86. See C.P.L. § 510.10(1)(e) (“record with respect to flight to avoid criminal prosecution”). ↩

  87. See C.P.L. § 510.10(1)(d) (“record of adjudication as juvenile delinquent or youthful offender”). ↩

  88. The data similarly does not account for the type of prior convictions a defendant has, aside from their severity, such as whether the prior conviction was for a drug offense or a property offense. It also does not indicate if the defendant has any out of state convictions. ↩

  89. A study by the New York City Criminal Justice Agency found that the prosecutor’s bail request was the most influential factor in whether individuals were released and almost solely predicted what bail amount the judge would set. See Phillips, A Decade of Bail Research in New York City, pages 58-68. See also Unlocking the Black Box of Prosecution, Vera Institute of Justice, footnote 1 (2023) (discussing other studies of the anchoring effect of prosecutors’ bail requests on judicial bail determinations). ↩

  90. OCA/DCJS have data corresponding to an individual’s failure to appear record, the charges and quantity of their prior convictions and pending cases, and their adjudication record. The adjudication/youthful offender record is also available to OCA/DCJS. That information is confidential, but given the ability to anonymize data, it would have been possible to make it public. It is unclear whether OCA/DCJS have access to CJA data, and specifically the algorithm-generated release recommendations. We do not know whether OCA tracks the prosecution’s requests regarding bail, although we are aware that judges often write down requested bail amounts on the physical court file. ↩

  91. Estimate, std.error, and statistic are rounded to two decimal places. ↩