Pretrial detention functions differently depending on why an individual is referred to the legal system. For those charged with misdemeanor offenses, the pretrial process is often the primary punishment, irrespective of guilt or innocence. For those charged with felonies, the primary punishment often comes from the resulting adjudicative sentence. Thus, the consequences of such detention on guilty pleas and carceral sentences could vary starkly between misdemeanors and felonies. This study draws on a unique dataset combining complete individual-level arrest and pretrial incarceration data for all adults arrested in New York City in 2016 and 2017. Using logistic regression and Cox survival models, the study identifies differential impacts of time detained pretrial on downstream consequences. Compared to people detained on felonies, people detained on misdemeanors are more likely to plead guilty, plead faster, and receive a carceral sentence. The article concludes with a discussion of policy implications of these differential pretrial justice regimes
JCJ final version: https://doi.org/10.1016/j.jcrimjus.2022.102008
More than three-fifths of the approximately 11 million admissions into local jails across the U.S. each year involve people who are legally innocent and are being detained pretrial (Zeng, 2018). Historically, the purpose of pretrial detention has been to ensure that people charged with crimes show up to court, but its justification has evolved over time to include incapacitation, such that people deemed threats to public safety can be detained pretrial (Rabinowitz, 2021; Scott-Hayward & Fradella, 2019). The impacts of pretrial detention can be severe and long-lasting. Local jails tend to be overcrowded, under-resourced, and chaotic, with legally innocent individuals exposed to high rates of disease, violence from staff and other inmates, and other serious dangers (Csete, 2010; Futrell, 2020; Jacobson et al., 2017; Lippman et al., 2017). Additionally, even brief periods in jail are highly disruptive to people’s lives, causing them to lose jobs, housing, and sometimes parental rights (Cochran & Toman, 2018; Menefee et al., 2021; Thomas, 2022). Additionally, studies have shown that upstream pretrial detention can lead to downstream negative consequences like unfavorable case dispositions (Campbell et al., 2020; Lum, Ma, & Baiocchi, 2017) or more severe sanctions (Oleson et al., 2016; Sacks & Ackerman, 2012). There is also reason to believe that longer stays in pretrial detention contribute to more punitive outcomes (Martinez, Omori, & Petersen, 2020; Petersen, 2019; Smith & Hu, 2021).
Further, poor conditions can add pressure for people to plead guilty (Lerman, Green, & Dominguez, 2022; McCoy, 2007). Indeed, the average detained person can expect to spend at least a month in jail, whether or not they are convicted (Natapoff, 2018). It is not surprising that pretrial detention could be experienced as “coercive,” especially by those detained on low-level or non-violent charges (Kirk & Wakefield, 2018).
Although a few states have begun to experiment with pretrial detention reform (Rengifo, Flores & Jackson, 2021), in most states, the ability to pay cash bail still determines who spends time in jail (Turney and Wakefield, 2019). As a result, most people detained pretrial are not flight risks nor threats to safety, but are low-income and often also marginalized due to race, ethnicity, social status, mental health, housing status, substance use, and other social characteristics (Foote, 1954; Irwin, 1985; Jeffreys, 2018; Martinez, Petersen, & Omori, 2020; Walker, 2016). Thus, pretrial detention has become a major part of the “shadow” carceral state (Beckett & Murakawa, 2012), in which millions of people who are legally innocent are punished by being held in pretrial detention.
However, not all detention is the same, and scholars contend that pretrial detention functions differently for those who come into the system with more serious versus less serious charges. For those who were charged with lower-level offenses (i.e., misdemeanors), many scholars have echoed Feeley (1979) that the pretrial process is the punishment, irrespective of guilt or innocence (Natapoff, 2021; Ristroph, 2020). Similarly, Kohler-Hausmann (2018) argues that misdemeanor courts are driven by “managerial justice,” in which pretrial detention is an end in itself – a process that both punishes and manages marginalized populations – rather than determining actual guilt.
In contrast, those charged with felonies (more severe charges) experience adjudicative justice in which the punishment (i.e., the sentence received) is separate from court processing and is administered after adjudication (e.g., a prison sentence). For those charged with felonies, pretrial detention (as opposed to being released pretrial) contributes to the impression that these individuals will inevitably receive carceral sentences (e.g., prison or more jail; Jeffreys, 2018; Rabinowitz, 2021).
The current study examines whether the duration of pretrial detention is related to the probability of pleading guilty, the timing of pleading guilty, and whether it increases the chance of a carceral (i.e., prison or jail) sentence. We are able to investigate these questions by drawing on a unique dataset combining arrest and pretrial detention data for adults arrested in New York City in 2016 and 2017. We also investigate the possibility that the effects of detention vary substantially between misdemeanor and felony cases. We anticipate that people who are held in pretrial detention will be more likely to plead guilty, will plead faster, and will be more likely to receive carceral sentences than those who are released pretrial. Further, we anticipate that people charged with misdemeanors will be more sensitive to the potentially coercive pressures of detention. We believe that those held on misdemeanors will plead as quickly as possible, whereas people charged with felonies will spend more time detained pretrial. We anticipate similar dynamics when examining carceral sentences, since plea bargaining determines sentencing for the vast majority of criminal cases (McCoy, 2005). Notably, we move the literature forward by examining not just the occurrence of pretrial detention but its duration, and by examining pleas and sentences as separate outcomes. Before presenting our analyses, we review the existing research on the consequences of pretrial detention, highlighting how our work adds to this growing body of literature.
OUTCOMES OF PRETRIAL DETENTION
A number of studies have investigated the impact of pretrial detention on downstream criminal legal system outcomes. Two of the most important consequences are guilty pleas and carceral sentences. While some theories have emerged regarding differences based on charge severity (discussed above and below), very little empirical literature has focused on disaggregating the effects of pretrial detention based on charge severity. Some recent studies have focused on sub-felony cases only, but very few have directly compared misdemeanor and felony cases in the same analysis, as ours does, making the current study an important contribution to empirical literature. Additionally, our analyses examine the duration of pretrial detention as an ordinal variable, rather than solely as dichotomous (i.e., length of pretrial detention versus pretrial detention or not), which also contributes to the field’s understanding of the impacts of pretrial detention.
Impact of pretrial detention on guilty pleas
Pretrial detention has consistently been associated with an increased likelihood of pleading guilty to a crime (Ares et al., 1963; Bergin et al., 2021). Theoretically and empirically, literature indicates this association is related to one of two factors: the desire to escape harsh conditions of confinement, or the consequences of deprivation of liberty in itself. First, even brief periods in jails weigh heavily on detained people, leading them to feel pressure to plead (Appleman, 2012; Bibas, 2004). Second, deprivation of liberty has a coercive effect on many detained people; individuals become consumed with seeking their immediate freedom (Campbell, 2021; Lerman et al., 2022). Deprivation of liberty via pretrial detention also limits access to lawyers and other resources, exposes people to various parties trying to convince them to plead guilty (including their own lawyers), and isolates them from those who would encourage them to assert their right to trial (Blume & Helm, 2014; Rabinowitz, 2021). Further, given that most of the people detained pretrial are low-income or otherwise marginalized, the mere threat of losing employment, housing, custody or other factors due to prolonged detention can induce pressure to plead guilty (Blume & Helm, 2014; Manns, 2005; Oleski, 1977; Wakefield & Andersen, 2020).
Empirical evidence supports the relationship between detention and increased likelihood of pleas. One study using New York County District Attorney data regarding 185,275 felony and misdemeanor cases found that pretrial detention increases the likelihood of guilty pleas due to successive and interrelated decision points in the adjudication process (Kutateladze et al., 2014). Qualitative and mixed-methods studies have also shown that pretrial detention adds pressure to accept pleas, in New York City (Kohler-Hausmann, 2018) and elsewhere (Cheng, 2012; Kellough & Wortley, 2002). Notably, these studies operationalize pretrial detention as a dichotomous variable, looking only at whether someone was held pretrial or not.
A growing body of research has also examined the effect of pretrial detention on the pace of pleas (Euvrard & Leclerc, 2017; Kellough & Wortley, 2002). For example, Petersen (2020) found that individuals in Florida who were held pretrial pled guilty 2.86 times faster than those who were not detained. Similar findings from previous research in other jurisdictions suggest that individuals detained pretrial pled faster than those not detained (Leslie and Pope, 2017; Ostrom and Hanson, 1999; Sacks & Ackerman, 2012).
Further, the impact of pretrial detention on guilty pleas may differ between misdemeanor and felony cases, but this potential differential effect has not yet been decisively theorized. Some scholars have speculated that people detained on misdemeanor charges have more to gain by taking a guilty plea than do their felony counterparts, since people detained on misdemeanors can usually leave detention if they accept a guilty plea (Bowers, 2007; Rabinowitz, 2021), which is perceived as a major benefit. Indeed, many people detained on misdemeanors are faced with the promise of either no further incarceration, or a sentence of time served, allowing them to leave jail immediately (Bibas, 2004). Conversely, people held on felony charges may be less likely or slower to plead because they are less likely to attain their freedom by admitting guilt.
Euvrard and Leclerc (2017) qualitatively explore differential reasons for taking guilty pleas in interviews with 24 people detained in Canada.1 They developed a three-tiered framework regarding pleas and the coercive effects of pretrial detention: hard coercion, soft coercion, and neutral custody. Within this typology, those detained on misdemeanors would be most likely to experience hard coercion and would be motivated to take a plea to get out of jail quickly, while those charged with felonies would experience soft coercion or neutral custody, resulting in longer jail stays. Although this coercion typology is intriguing, it is inchoate and suggestive rather than an established explanatory mechanism.
Further, the limited quantitative studies on the differential impact of likelihood of detention by severity have yielded conflicting results. For example, Kutateladze et al. (2014), compared misdemeanor and felony cases in New York City, finding the largest effect of detention on guilty pleas among individuals charged with person- and drug-related misdemeanors. Stevenson (2018) used judge leniency as an independent variable in a study of Philadelphia cases, and found similar results. However, in another New York City study, Leslie and Pope (2017) compared the effects in both misdemeanor and felony cases, and found the opposite pattern. Detention increased the probability of guilty pleas by approximately 10% in felony cases but only by 7% in misdemeanor cases. However, their study’s causal identification strategy relies on randomization of judges, but misdemeanor judges in their sample were not assigned as randomly as felony judges. Thus, Leslie and Pope (2017: 548) state they are “more cautious” and “less confident in the magnitude of the effect for misdemeanors” than for felonies.
The limited research that investigates how charge severity moderates the impact of the duration of pretrial detention on the pacing of pleas suggests that there is an association here. Petersen (2019) used a survival analysis to demonstrate that individuals charged with felonies waited more than 2.3 times as long to plead guilty compared to those charged with misdemeanors. Similarly, Leslie and Pope (2017) found that individuals detained and accused of felonies waited more than twice as long as those charged with misdemeanors before pleading guilty. In other words, those charged with misdemeanors plead faster, presumably to remove themselves from the coercive effects of jail.
Altogether, there is a growing body of evidence that pretrial detention has a significant effect on both the likelihood and the pacing of guilty pleas, only three studies have directly examined whether these effects differ based on charge severity. To date, no prior research has assessed these differences using both data and analytical techniques capable of determining the substantive and statistical significance. Further, since Leslie and Pope’s results contradict Stevenson’s and Kutateladze’s, there is an open question about whether the effects of pretrial detention on accepting guilty pleas are different for individuals facing felony or misdemeanor charges, and if so, in which direction.
Impact of pretrial detention on sentencing
Pretrial detention is also associated with an increased likelihood of a jail or prison sentence, as well as the length of that sentence (Heaton, Mayson & Stevenson, 2018; Oleson et al., 2014). In investigating the impact of the duration of pretrial detention, Clarke and Kurtz (1983) found that the longer an individual was detained pretrial, the less likely their case is dismissed and more likely to have a longer sentence. Importantly, Clarke and Kurtz’s early study is one of the few that investigates duration of pretrial detention and its impact on sentencing. Phillips (2012) found that 20 percent of those released during the pretrial period received a jail or prison sentence while 87 percent of the people who were detained for the entire pretrial period were incarcerated.
Several explanations have been proposed, many of which follow the logic of pleas. First, not being detained affords the individual the opportunity to demonstrate that they can remain crime free and build a case for reduced sentencing (Foote, 1954, Sacks and Ackerman, 2014). Second, most closely aligned with pleas, pretrial detention may make it more difficult for individuals to meet with their court representative, to secure character witnesses, or to obtain documents that may impact the outcome of their case (Williams, 2003). Third, detention impairs the ability to demonstrate remorse or attempts at self-improvement (e.g., entering a treatment program) which may improve a judge’s outlook on the individual being sentenced (Goldkamp, 1980). Fourth, since an individual who is already in custody is perceived as belonging in custody, court actors deciding sentences are more likely to assume the culpability of individuals who have been detained compared to those who were released pretrial (Campbell, 2021; Katz et al., 1972). These early appraisals of guilt and dangerousness may send signals to later system actors, setting into motion a pattern of compounding stigma and disadvantages (Kurlychek and Johnson, 2019). Such cumulative disadvantages may translate into a decreased likelihood of a non-carceral sentence (Spohn, 2009; Schlesinger, 2007).
Empirical research also supports the idea that pretrial detention has a substantive and causal effect on sentencing. In a study of misdemeanor cases in Harris County, Texas, Heaton and colleagues (2017) found that detained individuals were more likely to receive jail sentences than similarly situated individuals who were released. Spohn and DeLone (2000) found that pretrial detention was associated with greater odds of an incarceration sentence for felony convictions in three counties from Florida, Illinois, and Missouri. Others have demonstrated that among both misdemeanor and felony cases pretrial detention was associated with a higher likelihood of a carceral sentence and increased sentence length (Donnelly & Macdonald, 2018; Phillips, 2008, 2009; Sacks & Ackerman, 2014; Williams, 2003; c.f. Dobbie, Goldin, & Yang, 2018).
As with pleas, the impact on carceral sentences may differ between misdemeanor and felony cases. Similarly, this issue has not been decisively theorized, and existing research on how offense severity may moderate the impact of pretrial detention on sentencing is very limited, with mixed results. In a study of 153,407 people booked into a jail in Kentucky, Lowenkamp et al. (2013) found the effect of pretrial detention was more pronounced for those charged with misdemeanors than those charged with felonies. Even after accounting for other factors such as charge type, demographics, and criminal history, individuals in Kentucky detained on misdemeanor charges until disposition were 4.44 times more likely to receive a jail sentence compared to those released at some point prior to disposition, while those detained on felony charges until disposition were 3.32 times more likely to receive a prison sentence, compared to those released prior to disposition. Similarly, Stevenson’s (2018) and Phillips’ (2007, 2008) work suggest the consequences of pretrial detention are often more severe for people charged with less serious crimes, while others have found the effect of being detained is greatest for those charged with violent felony offenses (Bergin et al. 2021; Petersen, 2019). Similarly, Leslie and Pope (2017) found that while pretrial detention in New York City had a significant effect on the sentence received among both misdemeanor and felony cases, the effect was much larger among felony cases, with pretrial detention associated with a minimum sentence over 150 days longer than those who were not detained.
In summary, much like the literature on plea deals, it is likely that pretrial detention has a substantial impact on sentencing outcomes. What is missing from the literature, however, is an explicit examination of whether the effects of pretrial detention on these carceral sentences are moderated by charge severity. Like Tartaro and Sedelmaier (2009), who found that an individual’s race and geographical context may moderate the impact of pretrial detention on the likelihood of incarceration, we suggest a complex interplay between charge severity and pretrial detention may exist.
THE CURRENT STUDY
The current study adds to the existing knowledge on the impact of pretrial detention on criminal legal system outcomes in a number of ways. Using a unique and comprehensive dataset on arrests and pretrial detention in New York City, we explore how the effect of pretrial detention may vary between misdemeanor and felony cases. We examine these charge-specific effects across different points in the court process: the likelihood of entering a plea, the speed of entering a plea, and the likelihood of receiving a carceral sentence.
Based on our synthesis of prior findings and theory, we expect time in pretrial detention to increase the likelihood and speed of both pleading guilty and receiving a carceral sentence. However, there is disagreement among the few studies examining whether these effects might vary between misdemeanor and felony cases. Based on the limited prior empirical evidence, we expect the effect on the likelihood and pace of guilty pleas to be stronger in misdemeanor cases than felony cases. We also expect that while individuals charged with felonies are more likely to be given a carceral sentence, on average, the effect of being detained pretrial on the likelihood of a carceral sentence will be greater among individuals charged with misdemeanors. In testing these assertions, we account for a number of relevant covariates, including both individual- and case-level characteristics, as well as controls for both geographical and temporal context. Further, we build upon previous studies by using a novel second-difference approach to assess interaction effects on our nonlinear outcomes. Through a rigorous empirical analysis of this comprehensive dataset, the current study fills an important empirical gap and, by extension, sets the stage for gaining a sharper understanding of the mechanisms at play.
DATA AND METHODS
Sources and sample
New York City is a particularly apt setting in which to study pretrial detention since its courts process over one million criminal cases annually and detain almost 9,000 people daily at the time of the study period (Criminal Court of the City of New York, 2018). New York’s jail population is also made up of a larger share of individuals being held pretrial. Nationally, approximately 65% of people held in jail are being held pretrial (Zeng, 2018), while in New York City in 2016 that proportion is around 75% (Rempel & Rodriguez, 2019). Further, in New York City in 2016, 72% of people detained pretrial were held because of failure to post bail as opposed to having been remanded, or held for other reasons (e.g., warrants; Lowenstein, 2017).
This study leverages a unique dataset that links all fingerprintable2 misdemeanor or felony arrests, provided by the New York State Division of Criminal Justice Services (DCJS), to pretrial detention data provided by the New York City Department of Correction (DOC). The unique linking of these two agencies’ data allows us to gather information about people’s arrests, time spent in pretrial detention (if any), case outcome, and sentence. After limiting the cases to arrests that occurred within the five counties of New York City, the arrest records were linked to DOC detention data using a unique case-level identifier that is available to both agencies (a court-assigned case number) and a deidentified state-issued person-level identifier (pseudo-NYSID). Linked cases were checked for accuracy using arrest and admission dates, as well as demographic characteristics.
Using the DCJS data as a baseline for the population under study, the current analysis focuses on the population of adults arrested for fingerprintable offenses as a top charge in New York City between January 1, 2016 and December 31, 2017. In order to ensure each arrest represented an independent observation, the original arrest-level dataset was reduced to the person-level by retaining only an individual’s first arrest during the study period (i.e., all subsequent arrests were excluded). A small number of cases were excluded if information about the individual’s race was missing from the DCJS data (< 1%). We also excluded a small number of cases that were not disposed of within two years (730 days; n=3,057) as well as any cases that were still pending at the time of data extraction in early 2019. The resulting dataset consists of 255,966 unique individuals over the age of 18 who were arrested in New York City in 2016 or 2017.
Our outcomes were created using the case disposition provided by DCJS. The first two outcomes of interest are the existence of, and time until accepting a guilty plea. A plea is captured by dichotomizing the case disposition variable as it appears in the DCJS dataset. The original case disposition variable is a categorical variable that includes case outcomes such as conviction by verdict, conviction by plea, and dismissal. Cases where this disposition variable is listed as conviction by plea are coded as 1; all others are coded as 0. Time to pleas, for those cases disposed of by plea, was calculated using the number of days between arraignment and case disposition. Individuals who did not plead guilty were right-censored in the survival analyses discussed below.
The last outcome captures whether the individual received a carceral sentence (jail or prison) following case disposition (Yes=1). Carceral sentences are captured by dichotomizing the sentence variable as it appears in the DCJS dataset. The original sentence variable includes outcomes like probation, credit for time served (CTS), or fine. Cases where the sentence includes jail or prison are coded as 1; all others are coded as 0 (including CTS).
Although some scholars argue that jail and prison sentences should be assessed separately (Freiburger & Hilinski, 2013), the focus of the current study is to assess whether individuals who are detained pretrial receive harsher sentences at the time of disposition. Thus, it is reasonable to assume that any carceral sentence is more severe than community-based alternatives (e.g., probation), CTS, or outright dismissal. This dichotomous operationalization is also consistent with past research that examines the effect of various factors on case-level outcomes (e.g., Ulmer & Johnson, 2004).
Our primary independent measure captures pretrial detention, coded as an ordinal measure of time spent in detention (as it was measured in Dobbie et al. (2018), Heaton et al. (2016), and Lowenkamp et al. (2013)). This operationalization allows us to compare the outcomes across groups of individuals who were not detained during the pretrial process (= 0) with those who were detained but released on the same day (=1) 3, as well as those detained for longer periods of time (2-3 days=2; 4-14 days=3; 15-30 days=4; and 31+ days=5). These categories most closely follow Lowenkamp et al. (2013) 4, who observed that measuring detention duration in such categories allows for the examination of each particular length of detention as a predictor. Prior researchers have found turning points such as the first 48-72 hours detained to be substantively meaningful (Chevrier, 2021; Dobbie et al., 2018; Smith & Hu, 2021), as have reform advocates (Pretrial Justice Institute, 2017). This includes differences between felony and misdemeanor detention – for example, Heaton and colleague (2016: 757) note that for people charged with misdemeanors, much of “the damage is done in the first few days” and that “for a crime with an expected punishment of a few days’ imprisonment, after a few days a quick guilty plea may become relatively more attractive than posting bail.” Such ordinal treatment of pretrial detention expands on existing research that uses dichotomous measures of detention or does not capture variation in duration of detention (e.g., Bergin et al., 2021; Leslie & Pope 2017).5
Importantly, we are unable to assess whether pretrial detention was a result of monetary bail being set and not paid or whether the individual was remanded to custody. Although there is considerable overlap between judicial decision-making (e.g., bail set) and the incidence of pretrial detention, our measure cannot tease apart these differences.
We also capture the nature and severity of the most serious offense using a series of dichotomous variables. The first, offense severity, was created to indicate whether the most serious charge was a misdemeanor (=0) or a felony (=1). The interaction between charge severity and pretrial detention will allow us to assess whether the effects of pretrial detention differ between felony and misdemeanor cases.
All models also include a measure of offense type by coding the most serious charge into seven categories: (1) possession of dangerous weapons, (2) sex offenses, (3) drug offenses, which include sale and possession of illicit substances, (4) violent offenses, which include homicide, manslaughter, robbery, kidnapping, arson, and assault; (5) property crimes, which includes burglary, larceny, and other types of theft such as stolen property, bribery, extortion, fraud, embezzlement, and forgery, (6) driving while intoxicated (DWI), including under the influence of drugs or alcohol, and (7) other offenses, which includes gambling, crimes against public order, and prostitution. Unsurprisingly the majority of individuals in the current sample were charged with misdemeanors (69.1%) while the most common offense type was property offenses (31.3%) followed by violent offenses (29.2%) and drug crimes (21.6%).
Because prior criminal history is consistently one of the strongest predictors of both pretrial detention and case outcomes, we also include two measures of criminal conviction history, prior felony convictions and prior misdemeanor convictions (see Ulmer, 2012). Although each of these measures were significantly skewed, no transformations improved the distribution in any meaningful way, thus the original metrics were retained.
Demographic control measures include sex, race, and age as reported by DJCS. The individual’s biological sex (reported at arrest) is captured using a dichotomous measure (male=1; 78.8%). Race and ethnicity are categorized as White (14.2%), Black (44.4%), Latinx (34.4%), or other (7.0%). We also include age at the time of arrest as a continuous measure, with an average age of 34.6 years old (SD=12.2).6
Finally, as this analysis includes cases from all five boroughs of New York City (Brooklyn, the Bronx, Manhattan, Queens, and Staten Island), each distinct in terms of population size, enforcement rates, and case processing procedures (Chauhan et al., 2015), we account for borough-level variation by including a binary variable for each borough, using Manhattan as the reference category. We also account for seasonal and period effects by including a series of binary “dummy” variables for each month and each year in the analysis sample. The inclusion of these measures account for any temporal shocks that are common across all cases in a given period and allow us to more accurately assess the focal relationship present among this sample of cases. The descriptive statistics for all measures included in the current study are shown in Table 1.
[TABLE 1 ABOUT HERE]
The current analyses use bivariate and multivariate methods to examine the relationship between the duration of detention and two case outcomes, guilty pleas and receiving a carceral sentence. Prior to our multivariate assessment, we use a series of cross-tabulations to explore the bivariate relationship between the focal variables used in the analysis for the full sample. Then, we estimate a series of logistic regression models to assess the relationship between duration of detention and each case outcome, while controlling for other relevant individual- and case-level characteristics. To answer our question of whether outcomes vary by severity, we estimate a logistic regression model for each outcome that includes a series of interaction terms between our measure of pretrial detention (captured using 6 dichotomous indicators of detention duration, with no pretrial detention representing the reference category) and our measure of offense severity (0= misdemeanor, 1= felony). The results of these models allow us to easily compare whether the association between pretrial detention and our outcomes of interest differs between misdemeanor and felony cases.
Importantly, methodological advances suggest caution is necessary when interpreting the significance of interactions in non-linear regression models (Long & Mustillo, 2018). Following recent recommendations to the field (Mize, 2019), we evaluate the substantive differences between misdemeanor and felony cases using adjusted predictions rather than significance of the coefficient of the interaction term alone. More specifically, when examining the interaction between case severity and pretrial detention, we assess the significance of this interaction by examining the second differences (i.e., the difference in differences) of the marginal effects of the focal independent variable (pretrial detention) between misdemeanor and felony cases (Long & Mustillo, 2018; Mize, 2019). Here the first differences represent the effect of pretrial detention on the probability of each of our focal measures for both misdemeanor and felony cases separately. The second step of this analysis is to calculate the difference in differences (i.e., the second derivative), to determine whether the effect of being detained for a certain amount of time is significantly different between misdemeanor and felony cases. This approach has been used in recent criminological research examining the effect of race across urban and rural areas (c.f. Pupo & Zane, 2021; Zane, Mears & Welch, 2020) and allows the associations between our focal variables to be assessed with confidence.
Finally, to assess the association between pretrial detention, charge severity, and time-to-plea we use a series of Cox proportional hazard models. Proportional hazard models are ideal for analyzing time-dependent outcomes, such as days to plea, within a specified follow-up period. They are also capable of accommodating censored data, in which we do not observe the outcome of interest (a guilty plea). The Cox method is also a special form of the generalized proportional hazard model that is less restrictive and allows the baseline hazard ratio (HR) to take any form, rather than being parametrically specified a priori (Cox, 1972). The model is written as follows:
where hi(t) is the risk of pleading guilty on day t for individual I, α(t) is the baseline hazard or intercept, DETAINED is the measure of pretrial detention (captured as a series of dichotomous variables), DEMO is a matrix of demographic covariates (age, gender, race and criminal history), and CASE is a matrix of the case characteristics described above. The outcome h(t) is defined as the instantaneous probability of a guilty plea on day t knowing the actual failure day T, dropping those who have already failed, and adjusting for the passage of time during the 2-year analysis window Δt.
Bivariate association between pretrial detention and case outcomes
Table 2 displays the results of bivariate associations between duration of pretrial detention and the prevalence of pleas or carceral sentences, broken out by charge severity. The results suggest a roughly positive relationship between the duration of pretrial detention and each outcome.
First, individuals who were detained, if even for a short amount of time, were far more likely to plead to the charges they faced (43.8% for those who were not detained, and over 60% for those who were detained, regardless of the duration). A closer look at felony and misdemeanor cases also suggests that differences in detention are larger among individuals facing misdemeanor charges than for those charged with felonies. Among felony cases, there is a clear positive association between the duration of detention and the percentage of cases that ended in a guilty plea. However, for misdemeanor cases, the largest difference observed is actually a dichotomous split between those who were detained any number of days compared to those who were not detained at all. The ordinal distribution is notably bimodal: compared to no detention, a higher proportion who were released same-day pled. From same-day to the next duration, the proportion decreases, and then steadily increases, with longer periods of pretrial detention associated with a higher likelihood of a plea. This bimodal pattern will recur throughout our results for misdemeanor cases.
[TABLE 2 ABOUT HERE]
On the right side of Table 2 we see a similar pattern with respect to the percentage of each group who received a jail or prison sentence. In general, the longer a person was detained, the more likely they were to receive a carceral sentence. Among felony cases, however, longer periods of pretrial detention were associated with a higher percentage of individuals in each category receiving a jail or prison sentence. Echoing the bimodal pattern seen in misdemeanor guilty pleas, the one interesting exception to this relatively linear pattern is a spike in the proportion of individuals who spent less than 24 hours detained.
[TABLE 3 ABOUT HERE]
Pretrial detention and guilty pleas
Table 3 displays the results of our multivariate assessment of the relationship between the pretrial detention and cases resulting in a plea7. The first model in Table 3 displays the results of a logistic regression in which plea (=1) was regressed on our measure of pretrial detention and the matrix of other predictors considered. Although not shown to conserve space, these models also include a total of 24 period effects, one for each month in the study period as well as one for each year (minus the two left out as reference categories).
Results shown in the first model suggest that pretrial detention (of any duration) is associated with a greater likelihood of a plea. For example, individuals who were detained and then released on the same day were 60% more likely to enter a plea than those who were not detained (OR = 1.59; CI = 1.47 – 1.72). The magnitude of the observed effect was similar for those who were detained between 2 and 3 days (OR = 1.38; CI = 1.31 – 1.46), as well as 4-14 days (OR = 1.45; CI = 1.39 – 1.52), but was substantially larger for those detained for longer periods of time. In fact, among the full sample of cases analyzed, individuals who were detained 15 to 30 days had odds of pleading nearly three times as large as their counterparts, and those detained longer than a month had odds of entering a plea more than three and a half times as large (OR = 2.99 and OR = 3.70, respectively).
In addition to our focal measure of pretrial detention, a number of covariates were also significantly related to entering a plea. Felony cases had odds of entering a guilty plea more than twice as large as their counterparts (OR = 2.11; CI = 2.07 – 2.15). Violent and drug or sex-related offenses were less likely to result in a plea, while weapons offenses as well as DWI and other offense types were more likely to result in a guilty plea. In addition, individuals with a greater number of felony and/or misdemeanor convictions were also more likely to plead guilty. Of the demographic characteristics considered, sex was significantly related to entering a plea, with males having odds of entering a guilty plea 54% larger than females (OR = 1.54; CI = 1.51 – 1.57). Consistent with prior research on pleas (e.g., Kutateladze et al., 2014), there were few racial differences, with the only substantial difference being that individuals of other racial/ethnic groups were less likely to enter a guilty plea. Finally, pleas were more likely in Queens and Staten Island (as compared to Manhattan) and less likely in the Bronx or in Brooklyn.
The results presented of the proportional hazard model, shown in Model 3 of Table 3, paint the same picture in terms of the direction and magnitude of the effects of the case and individual characteristics. Here a hazard ratio of greater than 1.0 can be interpreted as an individual with those characteristics pleads faster than those in the reference category. These results are consistent with prior research in this area (Ostrom & Hanson, 1999; Petersen, 2020; Sacks & Ackerman, 2012), and suggest that individuals who were detained pretrial plead between 5-58% faster than individuals who were not detained. Similar to the results of the logistic regression model, a number of individual and case-level characteristics were also related to the timing of a guilty plea.
Differences by charge severity
After establishing the overall association between pretrial detention and the likelihood and timing of pleading guilty, our focus shifts to determining whether these associations are conditional on the charge severity. To answer this question, we introduce a series of interaction terms (one for each level of our focal independent measure, time in detention) into each model. This allows us to directly estimate the difference in the effects of pretrial detention between misdemeanor and felony cases (see Models 2 and 4 of Table 3). For ease of interpretation, regression coefficients are presented instead of odds ratios or hazard ratios. Significant coefficients greater than zero are associated with an increase in the probability of a guilty plea over misdemeanor cases, while negative coefficients are associated with a decreased odds of pleading.
Model 2 represents the effects observed among misdemeanor cases (the reference category). Interpretation of these coefficients allows us to calculate the odds of a felony plea by summing the main effects odds ratio and the interaction term coefficient. For example, individuals facing misdemeanor charges who were detained for 15 to 30 days had odds of entering a guilty plea over 3.5 times higher than their counterparts (OR = exp(1.33) = 3.78). For felonies, the odds of a plea were only 2.5 times higher for those who were detained (exp(1.33-.39) = 2.56). We see a similar pattern across each of the categories of pretrial detention, where among misdemeanor cases there is a significant positive effect of pretrial detention on pleas, but this effect is reduced among felony cases. There is one exception, among those detained longer than 31 days, the impact of detention is stronger among felony cases.
Examining the results as predicted probabilities of a plea allows us to further assess whether the association between detention and pleas is moderated by case severity (see Figure 1). Findings indicate that effect of pretrial detention on the probability of a plea does, in fact, vary based on case severity. Among misdemeanor cases, the predicted probability of a plea is .41 for those who were not detained and .62 for those who were detained for a single day, a 20.8% difference (p < .001). However, among felony cases, this difference was much smaller (.62 - .58 = .035; p < .01). The difference between these differences (i.e., second difference)—17.3%—is statistically significant (p < .001), indicating that case severity moderated the (positive) association between pretrial detention and entering a plea. This pattern of significant second differences held across levels of our measure of pretrial detention, with the exception of 31+ days detained, where case severity did not significantly moderate the association between detention and a plea.8
[FIGURE 1 ABOUT HERE]
Figure 1 suggests some other patterns relevant to our hypotheses about the differences between misdemeanor and felony cases. Echoing the unmodeled bivariate relationships in Table 2, in felony cases, the trend is monotonic (i.e., never decreases) and approximately quadratic. Each increase in our ordinal days of detention measure within in the first two weeks is associated with an almost constant predicted probability of pleading guilty (just above 60%). Then at two weeks, there is a sharply accelerating increase for people held for longer periods of time (from about 64% for those held 15-30 days to 85% for those held longer than a month). This suggests that in felony cases, there is a flat or, at most, a very modestly increasing association between detention and pleas in the first two weeks of pretrial detention that then accelerates.
However, for misdemeanor cases, the pattern is quite different. As in the unmodeled bivariate data, the pattern is non-monotonic and bimodal, with one sharp increase among same-day release, followed by a decrease in probability of a plea. There is then a steady increase to the second mode at 15-30 days detained, followed by another drop in the 31+ days group. Although the spike among same-day release misdemeanor cases is consistent with our hypotheses, the drop among those detained more than 31 days is unexpected (though as noted above, the second-difference test is not significant at 31+ days). Unlike with felony cases, the pattern of likelihood of pleading is less straightforward for misdemeanor cases.
Of note, the predicted probabilities shown in Figure 1 are significantly different from one another, but the substantive differences across many shorter durations of pretrial detention are relatively small. For example, Figure 1 shows there is only a range of about 5% across the predicted probabilities of taking a plea for people charged with misdemeanors who are detained for less than two weeks. It could be argued that the statistically significant differences identified might be driven by the sample size, since over a quarter million cases are included. However, the overall range of predicted probabilities across the entire span of time detained for people charged with misdemeanors is more than 30 percentage points, which is a substantive difference. Intriguingly, the overall range is almost 10 percentage points smaller for those charged with felonies. That difference in range magnitudes is driven by the smaller difference between those not detained and those detained for less than two weeks on felony charges, as compared to those detention durations for people charged with misdemeanors.
Last, the results of the Cox models shown in Table 3 confirm that the main effects of the logistic regression model hold when the timing of a plea is considered. Pretrial detention of any duration specified is associated with a shorter time-to-plea, as denoted by positive coefficients obtained from the proportional hazard model. This model also indicates that time to plea is moderated by severity because the magnitude of these effects is significantly reduced among felony cases. These conditional effects are shown graphically in the survival curves in Figure 2, which were derived from the Cox survival model in Table 3. Here, we show the cumulative risk of pleading guilty as time goes on for each of the six categories of time detained. The distance between each of the estimated curves (e.g., the effect of pretrial detention) is much smaller among felony cases (shown on the right) as compared to misdemeanor cases (shown on the left), again indicating a differential impact by severity.
[FIGURE 2 ABOUT HERE]
Overall, the results in Table 3 and Figures 1 and 2 suggest that a number of individual- and case-level factors are associated with the likelihood of a guilty plea. After accounting for those factors, our focal variable remained significant, indicating that individuals detained pretrial, and especially those detained for longer periods of time, were more likely to plead and plead more quickly than those who were not detained. Additionally, the observed effects of pretrial detention varied significantly between misdemeanor and felony cases, with estimated effects being larger among misdemeanor cases.
[TABLE 4 ABOUT HERE]
Pretrial detention and carceral sentences
Table 4 presents the results that examine the association between pretrial detention and receiving a carceral sentence (prison or jail). Results of the first logistic regression model suggest that individuals detained at any point during the pretrial period were between 2.8 and 40.8 times more likely to be sentenced to jail or prison compared to those who were not detained. Not surprisingly, felony cases were more likely to result in a jail or prison sentence as were cases involving sex-related offenses, DWI, and “other” crimes. Similarly, individuals with a greater number of prior convictions were more likely to receive a carceral sentence. While males had higher odds of being incarcerated than females (OR = 2.12; CI = 1.99 – 2.25), there were few racial differences observed. Individuals of other racial/ethnic backgrounds had lower odds than Whites of receiving a prison sentence (OR = 0.50; CI =.44 – .56). Finally, there were significant differences observed across boroughs with cases in the Bronx, Brooklyn, and Queens compared to Manhattan being less likely to result in a jail or prison sentence, while individuals in Staten Island were more likely to receive carceral sentences net of other factors considered.
The second model of Table 4 displays the results of the model designed to test whether the association between pretrial detention and later incarceration varies based on case severity. Again, regression coefficients are shown in place of odds ratios for ease of interpretation and model-adjusted predicted probabilities are used to examine differences in the association between pretrial detention and the probability of receiving a carceral sentence (second differences). Much like the results obtained for pleas, regardless of the length of pretrial detention, individuals who were detained pretrial were more likely to receive a carceral sentence than those who were not detained, and the majority of these odds were significantly larger among misdemeanor cases. For example, among misdemeanor cases, the predicted probability of a carceral sentence was .016 for those who were not detained and .245 who were detained for a single day, a 22.9% difference (p < .001). However, among felony cases, this difference was smaller (.13 - .06 = .07; p < .001) and the difference between these differences (i.e., second difference)—16.1%—was statistically significant (p < .001), indicating that case severity moderated the (positive) association between pretrial detention and a carceral sentence. Importantly, the examination of second differences suggests that the differences in sentences between misdemeanor and felony cases are quite a bit smaller than obtained above for pleas. Additionally, the association between 2-3 days in pretrial detention and the likelihood of a carceral sentence did not significantly differ between misdemeanor and felony cases (p > .05). Additionally, the second differences suggest the effect at 31+ days is larger for felonies.
The estimated probabilities of incarceration for both misdemeanor and felony cases are shown in Figure 3. The patterns are similar as those for pleas, except that the misdemeanor and felony cases move even more in tandem when examining this outcome. As in the previous figures, for felonies the effect on incarceration is monotonic, whereas for misdemeanors the effect is starkly bimodal, with the first mode at same-day release and the second at 31+ days. People charged with felonies are expected to be more likely to be sentenced to jail or prison than are people charged with misdemeanors, as is shown in almost all categories in this figure. What is unexpected based on our hypotheses but consistent with Model 2 (Table 4) is that for those who spent less than 24 hours detained pretrial, those charged with misdemeanors are more likely to receive a carceral sentence than are people charged with felonies.
[FIGURE 3 ABOUT HERE]
This study contributes to resolving the uncertainties around differential effects of misdemeanor and felony pretrial detention on the likelihood and speed of pleas and the likelihood of carceral sentences. Like Kutateladze et al. (2014), Petersen (2019), and Stevenson (2018), but unlike Leslie and Pope (2017), we do find evidence that the effect of pretrial detention on the chances of pleading guilty is stronger for people charged with misdemeanors, compared to those charged with felonies. Our second-difference analysis found the differences between charge severities was significant across almost all lengths of pretrial detention. Both the studies by Kutateladze and colleagues, and by Leslie and Pope examined New York City data, but Leslie and Pope’s study had weaknesses in the causal identification strategy for misdemeanor judge data (as discussed above in the literature review), so it is not entirely surprising that our results align with the other New York City study by Kutateladze and colleagues. Overall, we believe it is likely that people detained on misdemeanors have more rational incentives to plead guilty to get out of jail and regain their liberty through credit for time served or other outcomes, compared to those detained on felonies.
With respect to timing, our findings are consistent with Petersen (2019) and Leslie and Pope (2017) that people plead faster the longer they are detained pretrial, with misdemeanor cases yielding stronger effects than felony cases. Our data do not allow us to determine why misdemeanor pretrial detention would lead to such faster plea negotiations, but our results are consistent with the nascent theory that those detained on misdemeanors feel especially pressured to swiftly accept the ‘going rate’ of punishment for their charges.
With respect to carceral sentences, our findings accord with Stevenson (2018) that the effect of detention on sentences is stronger for people experiencing pretrial detention for misdemeanor cases. This is in contrast to the findings of Petersen (2019), Bergin et al. (2021), and Leslie and Pope (2017), who found more severe effects among felony cases. The differences between our findings could be due to differences in modeling strategies, since two of those studies use random assignment to judges as an exogenous instrument. Our contribution to the empirical debate about the relationship between detention and carceral sentencing does not decisively resolve the contradictory findings in prior literature, but rather prompts more research in this area. By comparing detention effects across these two groups, we set the stage for future studies to develop theories, replicate our investigations, and explain these policy-relevant differences.
Further, our results showed an intriguing bimodal pattern among misdemeanor cases not previously noticed in the literature. There was a spike among misdemeanors in likelihood of plea, speed of plea, and likelihood of carceral sentences that was not observed in felony same-day releases. This subset of people may be especially motivated to avoid additional time in pretrial detention. It is worth noting that these dynamics among same-day-releases are only detectable because of our operationalization of the duration of pretrial detention categorically rather than as a continuous measure. Overall, the striking differences between individuals charged with misdemeanor and felony crimes are consistent with the coercion typology identified by Euvrard and Leclerc. Future studies should continue to test this early theory and investigate the instances where people are spending less than 24 hours in pretrial detention.
Altogether, our findings support our hypotheses about the association between pretrial detention and subsequent negative consequences on case outcomes. Pretrial detention was found to increase both the odds and the speed of pleading guilty, as well as to increase the chances of receiving a carceral sentence. Crucially, these relationships were found to be conditional on the seriousness of charges. The relationships were significantly stronger for people charged with misdemeanors than those charged with felonies.
Another contribution of this study is examining the temporal effects of the duration of detention rather than examining detention dichotomously, and we found that duration amplifies the effects of detention. These findings of differential dynamics accord with Kohler-Hausmann’s (2018) argument that there are two fundamentally different models of justice in U.S. courts: adjudicative justice for felonies and managerial justice for misdemeanors. Future research should continue to explore the varying effects of the duration of pretrial detention on individual decision making and resulting criminal justice outcomes.
We opted to study pretrial detention in New York City precisely because New York has one of the largest jail systems in the country, and a sizeable proportion of individuals detained pretrial in the U.S. are detained there. Further, the uniqueness and completeness of our dataset provides an unprecedented opportunity to examine these important criminal justice outcomes in new and robust ways. However, because of our geographically limited sample, the external validity of our results may be likewise limited. Additionally, because of the time period examined in the current study, we are unable to assess the impact of New York 2020 bail reform legislation, which prohibited bail setting for most misdemeanors and non-violent felonies (Lu, Bond & Chauhan, 2021). However, following a historical pattern of cyclical experiments with and challenges to pretrial detention reform (Rabinowitz, 2021), the reforms were criticized and amended (Bates, 2022; Chasin, 2021; Lu, Bond, Chauhan & Rempel, 2022). Therefore, although geographically limited samples are often seen as lacking generalizability, and although this paper may not apply to current New York City policies, its findings still apply to criminal legal system decision makers. The fact that pretrial detention was associated with higher odds of pleading guilty and receiving a carceral sentence contribute to a growing empirical knowledge base that complements anecdotal evidence regarding the harms of pretrial detention. Although pretrial detention in New York City may be particularly coercive given conditions at the infamous Rikers Island, the largest jail facility in New York City, pretrial conditions in other jurisdictions around the U.S. are reportedly also dangerous and harmful (e.g., Schmadeke, 2014). Therefore, if, as we theorize, coercion operates to increase pleas and sentences, these findings could hold true in other settings. Additionally, even if conditions are less harmful, research supports the notion that the deleterious effects of pretrial detention can have far-reaching consequences (as discussed in this paper’s introduction). Therefore, while the current study may be limited in its generalizability, its findings are still useful in policy discussions regarding reform of pretrial detention in many settings.
Further, our dataset does not allow us to identify whether individuals were remanded to custody or simply did not make bail. St. Louis (2022) finds that people detained preventatively are less likely than those detained on (unaffordable) bail to plead guilty, but more likely to be sentenced to prison and get longer sentences, demonstrating the salience of these distinctions in pretrial detention type. As noted in the methods section, this means that although there is considerable overlap between judicial decision-making and the incidence of pretrial detention, our measure cannot tease apart the differences among the individuals who were detained for different reasons. As discussed in the Results section above, it is important to acknowledge that although our results are statistically significant, in some of the analyses the substantive differences are not as large. For example, the predicted probabilities shown in Figure 1 are significantly different from one another, but it is possible that the significant differences identified are influenced by the large number of cases included. Nonetheless, our results demonstrate many substantively meaningful patterns and differences, such as the predicted probabilities of guilty pleas for people charged with misdemeanors varying by 30 percentage points across the levels of time detained. Another limitation is that our findings could be confounded by omitted variable bias and other sources of endogeneity. Our dataset does not include judge identifiers, so we could not use judges as quasi-instruments to minimize endogeneity as has been done in prior research. We work within this limitation, common to observational studies in social science, of balancing the trade-offs between over-fitting and controlling for relevant potential confounders by using a parsimonious set of theoretically driven covariates as well controls for both geographical (i.e., borough) and temporal (i.e., month and year) context.
Relatedly, the current study does not include information that would allow the assessment of the potential importance of representation type (i.e., public, appointed, flat-fee, or hourly-fee attorneys) with regard to these specific research questions. There could be differences in pretrial communication, attorney incentives, and other relevant factors between attorney types. For example, Bibas (2004: 2477) notes that private defense attorneys retained on a flat fee (often the only private option for people with modest means) might have “financial incentives to plead cases out quickly in order to handle larger volumes.” An important area for future research could be how attorney types might affect plea negotiations, sentences, or even earlier processes during pretrial detention.
Similarly, prosecutorial discretion in charging decisions shape the downstream consequences and could explain some important differences between misdemeanor and felony processing, but such considerations are outside the scope of this study. As discussed in the data section above, these are arrest-level data, so data on prosecutors’ charges are not included. However, future studies empirically exploring the role of charging decisions could make an important contribution to the literature.
Another potentially relevant consideration not examined in the current study is the type of guilty pleas defendants may accept. For example, the key findings may be affected by the presumably small number of individuals (particularly individuals charged with a felony) not being offered guilty pleas, but data on who was or was not offered pleas was not available in this study. Similarly, patterns in who is offered CTS pleas might shed light on the bi-modal effects for individuals charged with a misdemeanor. Due to data limitations, these considerations are outside the scope of the current paper, but these could be fruitful areas for future research.
Another potential limitation is that combining jail and prison sentences into one carceral sentence outcome for both misdemeanor and felony cases might elide differences in the significance of sentences in the two groups. A sentence of serving additional time in jail is a negative outcome for people charged with misdemeanors—for whom it is as punitive a sentence as can legally happen. For those charged with felonies, a jail sentence may be less punitive along a spectrum of outcomes compared with the potential of a long prison sentence. In this case, combining jail and prison sentences for felonies may bias our estimates of the impacts of pretrial detention. To test this, as a sensitivity analysis, we also ran our models with an alternate operationalization of carceral sentences, bifurcated into jail sentences for misdemeanor and prison sentences for felony cases (results available upon request). The results are substantively the same, suggesting this was not a source of distortion in our study.
An additional consideration not addressed in this study involves whether the individual is detained at the time of disposition. If someone makes bail after a stint of pretrial detention, there may no longer be as much incentive to plead guilty and/or agree to a more punitive sentence. Alternately, the pressure from the detention already experienced might still be motivating. To determine the perseverance of pretrial detention pressure, future studies could fruitfully compare the effects of detention on those no longer detained to the effects on those still detained at disposition.
The findings of this study have powerful implications about the differential impact of the downstream consequences of pretrial detention in New York City. Most significantly, compared to people detained on felonies, people detained on misdemeanors are particularly likely to plead guilty, plead guilty faster, and receive a carceral sentence. Considering the demographics and social marginalization of many people detained on misdemeanors in cities like New York, the collateral consequences of this carceral pipeline sweep highly vulnerable people much more deeply into the criminal legal system than their lower-level charges might warrant. Although some misdemeanors are harmful to other people or society, many people charged with misdemeanors have not committed serious offenses. However, as discussed in the introduction to this article, the negative consequences of pretrial detention are real. Thus, it is important to also understand the harms that people detained on misdemeanors experience through pretrial detention. This is especially necessary because those personal harms spring from carceral sorting prior to any adjudication of guilt or innocence, and our results support early theorizing about the downstream impacts of pretrial detention and the coercive effect that detention can have. Thus, some individuals may feel coerced into pleading guilty to low-level crimes they have not actually committed, in order to escape the ordeal of pretrial detention.
Since deprivation of liberty in pretrial detention is one factor driving these substantively consequential outcomes for people detained on misdemeanors, one implication could be to shift away from detention for this group in particular, perhaps by simply releasing individuals on their own recognizance. In states like New York where the only legal justification for pretrial detention is ensuring court appearance (rather than public safety), other alternatives have been found to improve appearance rates without using potentially coercive pretrial detention. For example, the New York City Criminal Justice Agency evaluated pretrial behavioral nudges like text message reminders about court appearances and simplified court information forms, finding them to significantly reduce failure to appear (Ferri, 2022; Fishbane, Ouss & Shah, 2020). Such programs narrowly tailored to inducing court appearance have minimal negative externalities and promise to achieve similar results as pretrial detention, without the differentially experienced costs documented in our study.
New York and many other cities continue to experiment with bail and jail reform, and those experiments continue to be as deeply contested as previous historical waves of reform (Rabinowitz, 2021). Until innovative reforms are found to deliver the legal functions of pretrial detention without the persistent unintended consequences of previous waves of pretrial justice reforms, however, the findings in this study suggest that pretrial detention will continue to pressure people into taking guilty pleas that prompt a cascade of compounding negative consequences in their lives. Further, our results support a differential effect, consistent with Kohler-Hausmann’s (2018) contention that pretrial detention in misdemeanor and felony cases has two fundamentally different functions: in misdemeanor courts, pretrial detention can be seen as ultimately less about adjudicating guilt or innocence than about carceral sorting and hassling (focused on individuals who often already experience social disadvantages that prevent them from posting bail). For people charged with misdemeanors, in particular, our study demonstrates that the stakes of pretrial detention reform are high.
Appleman, L. I. (2012). Justice in the shadowlands: Pretrial detention, punishment, & the Sixth Amendment. Washington & Lee Law Review, 69, 1297.
Ares, C. E., Rankin, A., & Sturz, H. (1963). The Manhattan Bail Project. NYU Law Review, 38, 67.
Barabas, C., Doyle, C., Rubinovitz, J. B., & Dinakar, K. (2020). Studying up: Reorienting the study of algorithmic fairness around issues of power. Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, 167-176.
Bates, L. (2022). Supplemental Pretrial Release Data: Summary Analysis: 2019 - 2021. New York State Division of Criminal Justice Services Office of Justice Research and Performance. https://datacollaborativeforjustice.org/wp-content/uploads/2022/09/FINAL-DCJS-Public-Briefing-on-Supplemental-Pretrial-Release-Data-9-21-22.pdf
Beckett, K., & Murakawa, N. (2012). Mapping the shadow carceral state: Toward an institutionally capacious approach to punishment. Theoretical Criminology, 16(2), 221-244.
Bergin, T., Koppel, S., Ropac, R., Randolph, I. & Joseph, H. (2021). The Pretrial Detention Penalty: How Detention Impacts Case Outcomes. Available at: https://ssrn.com/abstract=3710879.
Bibas, S. (2004). Plea bargaining outside the shadow of trial. Harvard Law Review, 2463–2547.
Blume, J. H., & Helm, R. K. (2014). The unexonerated: Factually innocent defendants who plead guilty. Cornell Law Review, 100, 157.
Bowers, J. (2007). Punishing the innocent. U. Pennsylvania Law Review, 156, 1117.
Campbell, C., Henderson, K., & Renauer, B. (2021). Final Report on Examining Pretrial Detention in Oregon. Oregon Criminal Justice Commission.
Campbell, C. M., Labrecque, R. M., Weinerman, M., & Sanchagrin, K. (2020). Gauging detention dosage: Assessing the impact of pretrial detention on sentencing outcomes using propensity score modeling. Journal of Criminal Justice, 70, 101719.
Chasin, D. (2021). Two Steps Forward, One Step Back: How New York's Bail Reform Saga Tiptoes around Addressing Economic Inequality. Cardozo Law Review, 43 (1), 273-312.
Chauhan, P., Welsh, M., Fera, A. G., Balazon, E.. (2015). The summons report: Trends in issuance and dispositions of summonses in New York City, 2003-2013. The Misdemeanor Justice Project.
Cheng, K. K. (2012). Pressures to plead guilty: Factors affecting plea decisions in Hong Kong’s magistrates’ courts. British Journal of Criminology, 53(2), 257–275.
Chevrier, C. (2021). Why individuals who are held pretrial have worse case outcomes: how our reliance on cash bail degrades our criminal legal system. In Handbook on Pretrial Justice (pp. 67-87). Routledge.
Clarke, S. H., & Kurtz, S. T. (1983). The importance of interim decisions to felony trial court dispositions. J. Criminal Law & Criminology, 74, 476.
Cochran, J. C., & Toman, E. L. (2018). Mass jail incarceration and its consequences. In Handbook on the Consequences of Sentencing and Punishment Decisions (pp. 245-270). Routledge.
Cox, D. R. (1972). Regression models and life‐tables. Journal of the Royal Statistical Society: Series B (Methodological), 34(2), 187-202.
Criminal Court of the City of New York. (2018). 2017 Annual Report. Retrieved from https://www.nycourts.gov/LegacyPDFs/COURTS/nyc/criminal/2017-Annual-Report.pdf.
Csete, J. (2010). Consequences of injustice: pre-trial detention and health. International Journal of Prisoner Health, 6(2), 47–58.
Dobbie, W., Goldin, J., & Yang, C. S. (2018). The effects of pretrial detention on conviction, future crime, and employment: Evidence from randomly assigned judges. American Economic Review, 108(2), 201-240.
Donnelly, E. A., & MacDonald, J. M. (2018). The downstream effects of bail and pretrial detention on racial disparities in incarceration. J. Criminal Law & Criminology, 108, 775.
Euvrard, E., & Leclerc, C. (2017). Pre-trial detention and guilty pleas: Inducement or coercion?. Punishment & Society, 19(5), 525-542.
Feeley, M. M. (1979). The process is the punishment. Russell Sage.
Ferri, R. (2022). The benefits of live court date reminder phone calls during pretrial case processing. Journal of Experimental Criminology, 18, 149-169.
Fishbane, A., Ouss, A., & Shah, A. K. (2020). Behavioral nudges reduce failure to appear for court. Science, 370(6517).
Foote, Caleb (1954). Compelling Appearance in Court: Administration of Bail in Philadelphia. University of Pennsylvania Law Review 12, 1031-1079.
Freiburger, T. L., & Hilinski, C. (2013). The effects of race, gender and age on sentencing using a trichotomous dependent variable. Crime & Delinquency, 59(1), 69-86.
Futrell, N. S. (2020). Decarcerating New York City: Lessons From a Pandemic. Fordham Urban Law Journal, 48(1), 57.
Heaton, P., Mayson, S., & Stevenson, M. (2017). The downstream consequences of misdemeanor pretrial detention. Stanford Law Review, 69, 711.
Irwin, J. (1985). The Jail. University of California Press.
Jacobson, M., DeWolf, E., Egan, M., & Hafetz, D. (2017). Beyond the Island: changing the culture of New York City jails. Fordham Urb. LJ, 45, 373.
Jeffreys, D. (2018). America's Jails: The Search for Human Dignity in an Age of Mass Incarceration. NYU Press.
Jordan, K. L., & Freiburger, T. L. (2015). The effect of race/ethnicity on sentencing. Journal of Ethnicity in Criminal Justice, 13(3), 179-196.
Katz, L., Litwin, L., & Bamberger, R. (1972). Justice is the crime: Pretrial delay in felony cases. Case Western University Press.
Kellough, G., & Wortley, S. (2002). Remand for plea. Bail decisions and plea bargaining as commensurate decisions. British Journal of Criminology, 42(1), 186–210.
Kirk, D. S., & Wakefield, S. (2018). Collateral consequences of punishment: A critical review and path forward. Annual Review of Criminology, 1, 171-194.
Kohler-Hausmann, I. (2018). Misdemeanorland: Criminal Courts and Social Control in an Age of Broken Windows Policing. Princeton University Press.
Kurlychek, M. C., & Johnson, B. D. (2019). Cumulative disadvantage in the American criminal justice system. Annual Review of Criminology, 2, 291-319.
Kutateladze, B. L., Andiloro, N. R., Johnson, B. D., & Spohn, C. C. (2014). Cumulative disadvantage: Examining racial and ethnic disparity in prosecution and sentencing. Criminology, 52(3), 514-551.
Lerman, A. E., Green, A. L., & Dominguez, P. (2022). Pleading for Justice: Bullpen Therapy, Pre-Trial Detention, and Plea Bargains in American Courts. Crime & Delinquency, 68 (2), 159-182.
Leslie, E., & Pope, N. G. (2017). The unintended impact of pretrial detention on case outcomes: Evidence from New York City arraignments. The Journal of Law and Economics, 60(3), 529–57.
Lippman, J., et al. (2017). A More Just New York City. INDEP.COMM’N ON N.Y.C. CRIMINAL JUSTICE & INCARCERATION REFORM.
Long, J. S., & Mustillo, S. A. (2018). Using predictions and marginal effects to compare groups in regression models for binary outcomes. Sociological Methods & Research, 1-36.
Lowenkamp, Christopher T., Marie VanNostrand, & Alexander Holsinger. (2013). Investigating the impact of pretrial detention on sentencing outcomes. New York: Laura and John Arnold Foundation.
Lowenstein, Ronnie (2017). The City of New York Independent Budget Office report on pretrial detention rates to the New York City Council (May 16, 2017). https://ibo.nyc.ny.us/iboreports/pretrial-detention-rates-may-2017.pdf
Lu, O., Bond, E., Chauhan, P., & Rempel, M. (2022). Bail Reform in Action: Pretrial Release Outcomes in New York State, 2019-2020. Data Collaborative for Justice.
Lu, O., Bond, E., & Chauhan, P. (2021). Assessing the Potential Impact of 2020 Bail Reforms on 2019 New York City Criminal Court Cases. Data Collaborative for Justice.
Lum, K., Ma, E., & Baiocchi, M. (2017). The causal impact of bail on case outcomes for indigent defendants in New York City. Observational Studies, 3, 39-64.
Manns, J. (2005). Liberty takings: A framework for compensating pretrial detainees. Cardozo Law Review, 26, 1947.
Martinez, B. P., Petersen, N., & Omori, M. (2020). Time, money, and punishment: Institutional racial-ethnic inequalities in pretrial detention and case outcomes. Crime & Delinquency, 66(6), 837-863.
McCoy, C. (2005). Plea bargaining as coercion: The trial penalty and plea bargaining reform. Crim. LQ, 50, 67.
McCoy, C. (2007). Caleb was right: Pretrial decisions determine mostly everything. Berkeley J. Crim. L., 12, 135.
Menefee, M. R., Harding, D. J., Nguyen, A. P., Morenoff, J. D., & Bushway, S. D. (2022). The effect of split sentences on employment and future criminal justice involvement: evidence from a natural experiment. Social Forces, 101(2), 829-863.
Mize, T. D. (2019). Best practices for estimating, interpreting, and presenting nonlinear interaction effects. Sociological Science, 6, 81-117.
Natapoff, A. (2018). Punishment without crime: How our massive misdemeanor system traps the innocent and makes America more unequal. Basic Books.
Natapoff, A. (2021). Atwater and the misdemeanor carceral state. Harv. L. Rev. F., 133, 147.
Oleski, M. S. (1977). The effect of indefinite pretrial incarceration on the anxiety level of an urban jail population. Journal of Clinical Psychology, 33(4), 1006–1008.
Oleson, J., Lowenkamp, C., Cadigan, T., VanNostrand, M., & Wooldredge, J. (2016). The effect of pretrial detention on sentencing in two federal districts. Justice Quarterly, 33(6), 1103-1122.
Ostrom, B. J., & Hanson, R. A. (1999). Efficiency, timeliness, and quality: A new perspective from nine state criminal trial courts. Citeseer.
Petersen, N. (2020). Do detainees plead guilty faster? A survival analysis of pretrial detention and the timing of guilty pleas. Criminal Justice Policy Review, 31(7), 1015-1035.
Petersen, N. (2019). Low-level, but high speed?: Assessing pretrial detention effects on the timing and content of misdemeanor versus felony guilty pleas. Justice Quarterly, 36, 1314–1335.
Phillips, M. T. (2012). A Decade of Bail Research in New York City. NYC Criminal Justice Agency.
Phillips, M. T. (2008). Pretrial Detention and Case Outcomes, Part 2: Felony Cases. NYC Criminal Justice Agency.
Phillips, M. T. (2007). Pretrial Detention and Case Outcomes, Part 1: Nonfelony Cases. NYC Criminal Justice Agency.
Pretrial Justice Institute (2017). 3 Days Count: Commonsense Pretrial. http://www.pretrial.org/download/advocacy/3DaysCount_overview.pdf
Pupo, J. A., & Zane, S. N. (2021). Assessing Variations in Juvenile Court Processing in Urban Versus Rural Courts. Youth Violence and Juvenile Justice, 19(3), 330-354.
Rabinowitz, M. (2021). Incarceration without Conviction: Pretrial Detention and the Erosion of Innocence in American Criminal Justice. Routledge.
Rabner, S. (2019). Hobbs Lecture: Criminal Justice Reform in New Jersey. Seton Hall L. Rev., 50, 619.
Rengifo, A. F., Flores, S. G., & Jackson, A. N. (2021). From Bright Plots to Blind Spots: Mapping Departures in Case Review Post-Bail Reform in Two New Jersey Courts. Criminal Justice and Behavior, 48(1), 96-115.
Rempel, M., & Rodriguez, K. (2019). Bail Reform in New York: Legislative Provisions and Implications for New York City. Center for Court Innovation.
Ristroph, A. (2020). Criminal law as public ordering. U. Toronto Law Journal, 70(1), 64-83.
Sacks, M., & Ackerman, A.R. (2014). Bail and Sentencing: Does Pretrial Detention Lead to Harsher Punishment?. Criminal Justice Policy Review 25 (1): 59-77.
Sacks, M., & Ackerman, A. R. (2012). Pretrial detention and guilty pleas: If they cannot afford bail they must be guilty. Criminal Justice Studies, 25(3), 265–278.
Schmakede, S. (2014). Expert: Cook County Jail one of the most dangerous in country. The Chicago Tribune. https://www.chicagotribune.com/news/ct-cook-jail-violence-met-20141008-story.html
Schlesinger, T. (2007). The cumulative effects of racial disparities in criminal processing. Journal of the Institute of Justice & International Studies, 7, 261–278.
Scott-Hayward, C. S., & Fradella, H. F. (2019). Punishing poverty: How bail and pretrial detention fuel inequalities in the criminal justice system. University of California Press.
Smith, S. S., & Hu, C. (2021). Exploring the Causal Mechanisms Linking Pretrial Detention and Future Penal System Involvement 1. In Handbook on Pretrial Justice (pp. 88-109). Routledge.
Spohn, C. (2009). Race, sex, and pretrial detention in federal court: Indirect effects and cumulative disadvantage. University of Kansas Law Review, 57, 879–901.
Spohn, C., & DeLone, M. (2000). When does race matter? An analysis of the conditions under which race affects sentence severity. Sociology of Crime, Law, and Deviance, 2(1), 3-37.
Stemen, D., & Olson, D. (2020). Dollars and Sense in Cook County. Loyola University Chicago.
Stevenson, M. T. (2018). Distortion of justice: How the inability to pay bail affects case outcomes. The Journal of Law, Economics, and Organization, 34(4), 511-542.
St. Louis, S. (2022). Bail denied or bail too high? Disentangling cumulative disadvantage by pretrial detention type. Journal of Criminal Justice, 82, 101971.
Tartaro, C., & Sedelmaier, C. M. (2009). A tale of two counties: The impact of pretrial release, race, and ethnicity upon sentencing decisions. Criminal Justice Studies, 22(2), 203-221.
Thomas, C. (2022). The Racialized Consequences of Jail Incarceration on Local Labor Markets. Race and Justice, https://doi.org/10.1177/21533687221101209.
Turney, K., & Wakefield, S. (2019). Criminal justice contact and inequality. RSF: The Russell Sage Foundation Journal of the Social Sciences, 5(1), 1-23.
Ulmer, J. T., & Johnson, B. (2004). Sentencing in context. Criminology, 42(1), 137-178.
Ulmer, J. T. (2012). Recent developments and new directions in sentencing research. Justice Quarterly, 29(1), 1-40.
Wakefield, S., & Andersen, L. H. (2020). Pretrial Detention and the Costs of System Overreach for Employment and Family Life. Sociological Science, 7, 342-366.
Walker, M. L. (2016). Race making in a penal institution. American J. of Sociology, 121(4), 1051-78.
Williams, M. (2003). The effect of pretrial release in imprisonment decisions. Criminal Justice Review, 28(2), 299–316.
Zane, S. N., Mears, D. P., & Welsh, B. C. (2020). How Universal is disproportionate minority contact? Justice Quarterly, 37(5), 817-841.
Zeng, Z. (2018). Jail inmates in 2016. NCJ, 251210.
TABLES AND FIGURES
Figure 1: Effect of pretrial detention on predicted probability of guilty plea, by charge severity
Figure 2: Cumulative probability of guilty plea over days detained pretrial, by charge severity
Figure 3: Effect of pretrial detention on predicted probability of carceral sentence, by charge severity