Russell Sage Foundation

Despite growing interest in bail and pretrial detention among both academic researchers and policymakers, systematic research on pretrial release remains limited. In this article, we examine bail and pretrial release practices across seventy-five large U.S. counties from 1990 to 2009 and look at the contextual correlates of bail regime severity. We find tremendous intra-county variation in bail practices, as well as a nationwide decline in the use of nonfinancial release and doubling of bail amounts during this period. This variation is not accounted for by differences in case composition across jurisdictions or over time. Patterns of bail practices are associated with political, socioeconomic, and demographic factors, however. Implications of these findings for future research on bail and pretrial detention are discussed.

Keywords

bail, pretrial detention, inequality

Pretrial detention and release are an important but often overlooked source of inequality in the criminal justice system. Two of every three jail inmates in the United States—20 percent of the total incarcerated population—are being held in pretrial detention (Minton and Zeng 2015). These approximately half a million people have been arrested but not yet convicted of a crime. The burden of detention does not fall evenly on the unconvicted population, however. Black men are significantly more likely than white men to be arrested without engaging in illegal activity (Weaver, Papachristos, and Zanger-Tishler 2019). And once in custody, the overwhelming majority of pretrial detainees remain behind bars because they are unable to pay for the bail needed to secure their release (Phillips 2008). Yet research also shows that detention has considerable collateral consequences. Studies find that pretrial detention fuels further inequality in criminal justice outcomes (Stevenson 2016; Sacks and Ackerman 2014; [End Page 126] Lowenkamp et al. 2013a; Phillips 2008; Williams 2003), increases recidivism (Gupta, Hansman, and Frenchman 2016; Heaton, Mayson, and Stevenson 2016; Lowenkamp et al. 2013b), and undermines the socioeconomic stability of detainees and their families (Dobbie, Goldin, and Yang 2016; Comfort 2016). Although reformers from across the political spectrum have drawn attention to the inequity of this approach (Harris and Paul 2017), research on the subject remains limited. Despite the size and impact of this system, we know surprisingly little about the bail practices that govern pre-trial detention and release.

Existing work on bail and pretrial detention focuses almost exclusively on features of individual cases, looking either in-depth at single jurisdictions or with a broad national sweep to determine how case characteristics shape relevant outcomes. In this article, we shift the focus to consider instead how the varied social, political, and organizational contexts in which cases are processed are associated with patterns of bail and pretrial release. To do this, we look at systematic variation in pretrial practices across time and place. Drawing on the National Archive of Criminal Justice Data State Court Processing Statistics (SCPS) series, we examine variation in bail regimes across the seventy-five largest counties in the United States between 1990 and 2009. We then look at the contextual correlates of bail regime severity, considering how case composition, organizational features, politics, economic conditions, and demographic factors relate to bail-setting practices.

Our analysis reveals considerable variation in pretrial release practices both across the country and over the nearly twenty-year period captured in our data. During this period, counties came to rely increasingly on money bail to determine pretrial release for defendants, while the average (inflation adjusted) cost of bail nearly doubled. This shift toward increasing use of money bail aligns closely with the broader shift toward increasing imprisonment over the same period (Western 2006). Further, given that the burden of arrest falls unevenly across racial categories, that the ability to meet money bail is sharply graded by class, and that pretrial detention is likely to have negative consequences for individual and family well-being, this shift in practices is also likely to have increased inequality in multiple domains.

Specific counties, meanwhile, vary considerably, both in the rates at which they use money bail to determine release and in the amounts at which that bail is set. With few exceptions, however, individual case characteristics do little to explain variation in either time or place. Indeed, our analysis suggests that we would expect more between-county variation in pretrial practices than we see given between-county variation in case characteristics. We do find, however, that patterns of bail outcomes are associated with county- and state-level factors. Higher proportions of African American residents, non-Democrats in the district attorney’s office, higher state-level income inequality, and Republican governors were all associated with lower levels of nonfinancial release. Higher unemployment rates, non-Democrats in the district attorney’s office, Republican governors, and elected judges were all associated with higher bail amounts.

Together, these findings reveal important sources of inequality in criminal justice contact. They document real differences in bail and release practices over time and across jurisdictions, and they show how local regimes of pre-trial practices vary along with the social, political, and economic contexts in which they operate. By highlighting the regimes that produce this inequality, these findings also point to explanations that extend beyond differences in individual detainees to consider instead what drives the choices local jurisdictions make about how to manage the pretrial process.

background

As with many aspects of its carceral system, the United States is a global outlier in how it manages pretrial detention. Officially, the U.S. Constitution outlines a system of bail in which persons awaiting trial may deposit money or property as collateral to ensure they appear in court. As long as defendants do not fail to appear, their bail should be returned. In practice, however, most people turn to commercial bail bonds to secure pretrial release, paying a registered bail agent a nonrefundable fee (usually 10 percent of the bail amount) to purchase a surety bond. The United States and the Philippines [End Page 127] are the only two countries in the world that allow this practice (Devine 1991).

Actual bail amounts are determined by a number of factors. Judges and magistrates may consider a defendant’s criminal history, ties to the community, or the circumstances of the alleged crime (Spohn 2009). Many jurisdictions, however, rely on bail schedules that base amounts largely on criminal charges. These schedules vary a great deal across state and county lines, and judges and bail magistrates sometimes maintain considerable discretion even when such schedules are used. Yet court officials need not set financial bail on a case at all. Bail may be denied if a person is considered excessively dangerous or likely to flee, and a person considered less risky may be released under supervision conditions or with a promise to appear for subsequent court dates.

Both the decision to set money bail and the amount at which bail is set are important determinants of pretrial detention and release. Research suggests that bail amount is among the most important predictors of the length of pretrial detention (Phillips 2007, 2008; Cohen and Reaves 2007). Even small amounts of bail may keep many people behind bars. One study finds that only one in eight defendants nationwide (12.5 percent) can secure pretrial release when bail is assessed at $50,000 or more, and that number only grows to nearly five in eight (60 percent) when bail is assessed at $5,000 or less (Beck, Bonczar, and Gilliard 1993). While higher bail amounts are thus an important determinant of pretrial detention, many indigent defendants are only able to access pretrial freedom if they are granted release on nonfinancial terms.

The implications of bail decisions can be serious for defendants. In addition to the immediate consequence of incarceration, a small but growing body of research finds that pretrial detention has a number of collateral consequences for detainees. Studies find that pretrial detainees are more likely to be convicted (Stevenson 2016), receive harsher sentences (Sacks and Ackerman 2014; Lowenkamp et al. 2013a; Phillips 2008; Williams 2003), and have higher rates of recidivism (Gupta, Hansman, and Frenchman 2016; Heaton, Mayson, and Stevenson 2016; Lowenkamp et al. 2013b) than similar defendants who are granted pretrial release. Evidence also suggests that pretrial detention is associated with decreased employment and loss of government benefits (Dobbie, Goldin, and Yang 2016). Indeed, even short periods of detention can create considerable instability, because people can quickly lose jobs, housing, and custody of their children while detained. Short-term incarceration can also put great strain on families, extending the consequences of this disruption well beyond individual detainees (Comfort 2016). Indeed, the financial and logistical burdens of securing bail often fall disproportionately on partners and other female caregivers (Page, Piehowski, and Soss 2019).

Yet though bail and pretrial detention have important consequences for inequality within the criminal justice system and for subsequent socioeconomic stability, we know little about patterns of inequality in bail and pretrial release practices. One important exception is a report from the Bureau of Justice Statistics that examines change over time in bail and pretrial detention in large U.S. counties. The report documents a nationwide shift from nonfinancial release to the use of surety bonds in the late 1990s (Cohen and Reaves 2007). This empirical analysis highlights the growing use of financial practices that disadvantage low-income defendants and put additional strain on their families, but it provides no explanation for the trend. And because it pools data from across the country, it provides no information about geographic differences in bail and pre-trial patterns. Indeed, because research on bail and pretrial detention tends to look either in-depth at single jurisdictions or broadly at national trends, we know very little about the breadth or variety of bail and pretrial release systems at work across the United States at any given time.

Research on other criminal justice outcomes suggests that criminal case processing can vary widely by time and place, however. The substantial over-time variation in incarceration is very well documented (Garland 2001; Western 2006; Wakefield and Uggen 2010). But, there is also evidence of geographic variation. Bruce Western, for example, shows that although incarceration rates are historically unprecedented [End Page 128] in the United States as a whole, states actually vary considerably in the rates at which they imprison people (2006). Whereas in 2003 Louisiana had an incarceration rate of 801 per hundred thousand residents, Maine’s rate was only 149 per hundred thousand. A number of studies similarly find systematic differences in sentencing across county lines (Johnson 2006, 2005; Fearn 2005). Studies of geographic differences in pretrial detention, however, have been limited to juvenile offenders. Barry Feld shows that juveniles in Minnesota are detained at higher rates in urban counties than suburban and rural ones, a difference he suggests results from local preferences for the use of formal versus informal social control (1991). Although pretrial detention of juveniles is quite rare and does not rely on the system of money bail so central in adult corrections, this work does suggest that systematic variation in pretrial practices may exist across counties.

This study explores patterns of bail and pre-trial practices across time and place and considers the individual and contextual factors associated with variation in each. The overwhelming majority of research looking specifically at bail and pretrial detention focuses on the impact of individual case characteristics on pretrial outcomes, but research on other aspects of criminal case processing suggests a variety of contextual factors that may affect these practices as well. The following section reviews this research and outlines possible determinants of bail and pretrial practices.

determinants of bail and pretrial practices

We next review existing research on the determinants of bail and pretrial practices. We begin with a look at the literature on individual determinants of pretrial outcomes, then turn to research on other aspects of criminal case processing for insights into contextual factors that may shape the pretrial process.

Individual Case Characteristics

Most systematic research on determinants of bail and pretrial detention focuses on the salience of individual case characteristics for pretrial outcomes. This work is interested primarily in the extent to which outcomes are influenced either by official legal criteria or by extralegal characteristics that might reflect discriminatory decision-making. Studies consistently show that legally relevant factors are important determinants of pretrial outcomes. Severity of charges and criminal history are particularly strong predictors of bail amount, pretrial release, and pretrial detention (Goldkamp and Gottfredson 1985; Cohen and Reaves 2007). Other characteristics that judges are legally allowed to consider, such as community ties, help predict pretrial outcomes as well (Petee 1994; Spohn 2009).

Yet considerable work also finds that extra-legal factors play an important role in bail and pretrial release. Most of this research focuses on race and sex and finds that pretrial detention is more common for defendants who are black or Hispanic (Spohn 2009; Cohen and Reaves 2007; Leiber and Fox 2005; Katz and Spohn 1995) and for those who are men (Spohn 2009; Katz and Spohn 1995). Other studies consider the specific components of pretrial detention in more detail. Stephen Demuth and Traci Schlesinger each find that black and Hispanic defendants are less likely to be granted nonfinancial release and more likely to receive higher bail amounts, controlling for other relevant predictors (Demuth 2003; Schlesinger 2005). Studies similarly find that bail amounts are higher for male defendants (Demuth and Steffensmeier 2004; Katz and Spohn 1995).

Given the abundance of research finding that individual-level case characteristics play an important role in pretrial outcomes, it is possible that variation in the composition of cases could drive variation in bail and pretrial release decisions. If the legal characteristics of individual cases and demographic characteristics of individual defendants are important determinants of pretrial decisions, then patterns of bail and pretrial release may simply reflect the cases processed at a given time and place. Variation over time or across jurisdiction would then result from changing case composition from year to year or across county lines.

Organizational Context

Organizational features may also play an important role in bail and pretrial release practices. A sizable literature argues that daily concerns [End Page 129] about organizational efficiency and practical constraints drive many of the decisions made by courtroom officials (Eisenstein, Flemming, and Nardulli 1988; Dixon 1995). Studies find evidence that such constraints shape criminal case outcomes. Jeffrey Ulmer and Brian Johnson find that caseload pressures help predict differences in sentencing across Pennsylvania counties, for example, and Johnson finds separately that caseload pressures similarly affect rates of downward departures (Ulmer and Johnson 2004; Johnson 2005). These findings suggest that the relative pressure on courts to efficiently process heavy case-loads shapes the decisions prosecutors and judges make about criminal case processing.

A pair of older studies suggest that similar practical considerations might shape pretrial decisions as well. Comparing pretrial practices in Detroit and Baltimore, Roy Flemming finds that the availability of resources, and particularly the availability of space in local jails, is a critical factor in bail policy (1982). Crowded jails increased the likelihood of pretrial release, while excess jail space resulted in more punitive pretrial detention practices. Jeffrey Roth and Paul Wice similarly find that the occupancy rate of local jails influenced the conditions set for pretrial release in Washington, D.C. (1980). This work suggests that judges consider jail crowding when making decisions about bail, pretrial detention, and release. When jail space is limited, that space may be reserved for the most serious offenders. Jail capacity and occupancy rates may create practical constraints on bail and release decisions, and differences in the availability of jail beds may thus affect differences in patterns of bail setting and pretrial release practices.

Political Context

Research suggests that the political context in which cases are processed may affect outcomes as well. Concerns about law and order have frequently been at the center of American electoral politics in recent decades, and politicians have repeatedly appealed to voters by promising to be tough on crime (Helms and Jacobs 2002). While Democrats and Republicans have both made these appeals, and both parties have been complicit in the expansion of the U.S. carceral system (Gottschalk 2014), there do appear to be meaningful partisan differences in criminal justice policies and outcomes. Republican political leaders have spent more than Democrats on police, courts, and corrections (Caldeira 1983; Davey 1998; Jacobs and Helms 1999). Republican strength at the state and national levels is also associated with higher rates of incarceration (Jacobs and Helms 1996, 1997; Western 2006). Local partisan preferences may also shape sentencing outcomes. Ronald Helms and David Jacobs find that courts embedded in conservative communities produced more sentencing disparities, with longer sentences issued for African Americans and for men (2002).

Political context is likely to affect bail and pretrial detention practices in one of two ways. First, it may shape the discretionary decisions of local officials. District attorneys and judges are often elected and accountable to local voters. A conservative electorate may therefore select candidates with more punitive orientations, and officials may consider local preferences when making discretionary decisions about setting bail or allowing pretrial release.

But partisan politics may also shape policies that affect pretrial outcomes. Elected officials may pass laws that set bail schedules, regulate the commercial sale of bail bonds, or affect the funding of pretrial services agencies that facilitate the use of nonfinancial release. Such policies may be shaped by pro-business partisan politics as much as partisan preferences for tough on crime approaches. Indeed, the conservative American Legislative Exchange Council has worked with Republican state law-makers to pass laws strengthening commercial bail bonds by undermining pretrial service agencies, organizations that facilitate pretrial release largely at the expense of bail industry profits (NAPSA 2009). All of these policies could affect the amounts at which bail is set and the extent to which detainees are released or detained pretrial.

Socioeconomic Context

Socioeconomic context may also play a role in bail and pretrial detention. Social control scholars argue that criminal justice policies are not used merely to control crime, but also more broadly to manage unruly populations that [End Page 130] might pose a threat to broader social order. Because of this, criminal punishment is directly related to economic conditions.

One line of research in this area considers the relationship between unemployment and criminal punishment. In 1939, Georg Rusche and Otto Kirchheimer laid out an early and influential argument, proposing that punishment historically grows harsher amid labor surpluses and milder when that surplus shrinks (2003). Later scholars have suggested that in a modern capitalist system the state uses incarceration to neutralize the threat posed by unemployed workers, who have a particular propensity to become unruly and disrupt the status quo (Spitzer 1975; Box and Hale 1992). This argument has some empirical support. Studies have found that unemployment rates are related to rates of incarceration independent of the effects of crime (Chiricos and Delone 1992; Yeager 1979; Lessan 1991). Stewart D’Alessio and Lisa Stolzenberg also find evidence that this argument may apply to pretrial outcomes (2002). They find that unemployed defendants have a substantially higher probability of pretrial detention in cities with high unemployment rates and argue that the unemployed population only poses a threat that warrants detention in the context of broader economic decline. This finding suggests that unemployment rates may prompt bail-setting patterns that make it harder to secure pretrial release.

But punishment may also be related to economic inequality itself. Scholars have argued that the wider social distance between marginal populations and criminal justice decision-makers may lead to more punitive outcomes for members of marginal groups (Bridges and Crutchfield 1988). Income inequality may reflect social disadvantage better than unemployment alone. Indeed, Western and his colleagues show that though recent incarceration trends in the United States bear little association with actual crime, they do track closely with income inequality. Income inequality may drive more punitive pretrial decisions as well (2004).

Demographic Context

Last, considerable research finds that criminal justice outcomes may be affected by a jurisdiction’s racial composition. Most studies explain this relationship in terms of racial threat. Minority or racial threat occurs when a majority white population perceives a threat from the size of a racial or ethnic minority group. Some scholars suggest that this threat stems from concerns that minority groups will challenge the dominant group’s economic and political dominance (Blalock 1967). More recent work, however, suggests that it arises from the white majority’s association of large minority populations with crime (Bontrager, Bales, and Chiricos 2005). In either case, empirical evidence supports the racial threat theory. Studies have found that black population size in particular is linked to several criminal justice outcomes, one of which is higher rates of imprisonment (Myers and Talarico 1987; Britt 2000; Weidner, Frase, and Schultz 2005) and with longer sentences (Wang and Mears 2015). Less work considers other minority groups, although Xia Wang and Daniel Mears find that larger Hispanic populations increase the decision to incarcerate (2015).

One study does look at whether racial threat accounts for decisions about bail and pretrial release. Marvin Free compares racial disparities in pretrial outcomes across jurisdictions with different size minority populations, and finds a curvilinear relationship between population size and disparate outcomes (2004). The data underlying this finding comes largely from the 1970s—a very different era in criminal case processing—but the analysis nevertheless suggests that racial threat may continue to shape bail and pretrial release decisions. In jurisdictions where white majority populations feel threatened by the size of nonwhite groups, officials may be more likely to embrace bail and pretrial release policies that increase pretrial detention.

contributions

Despite growing interest in bail and pretrial detention among both academic researchers and policymakers, systematic research on pretrial release practices remains limited. This study contributes to knowledge of this important but relatively understudied topic in two ways.

First, we address the surprising lack of systematic information about the bail practices [End Page 131] that govern pretrial detention and release by providing an empirical description of bail setting patterns both over time and across major U.S. counties. Because most research on bail and pretrial detention focuses on single locations or collapses geographic differences in national data, we know little about whether and how pretrial practices vary in the United States. Mapping the contours of this variation is thus an important empirical contribution in itself.

Second, we provide insight into the contextual correlates of bail regime severity. Much of what we know about variation in pretrial practices comes from studies that focus exclusively on the influence of individual case characteristics. We assess the relative importance of these characteristics for explaining variation in bail and pretrial release over time and across places. But we also draw on a wide range of literature to consider how the larger organizational, political, economic, and social context in which cases are processed may shape bail and pretrial release decisions. In this way, we shift the focus from individual-level determinants of pretrial outcomes to look instead at how policy choices, and the factors that influence them, shape patterns of pretrial detention and release across the country.

data and methods

In the following section, we describe the primary individual-level data that we use in our analysis, the State Court Processing Statistics (SCPS), and then detail our construction of state- and county-level variables that we merge to the SCPS. We then describe our analysis plan.

State Court Processing Statistics

We primarily draw on individual-level data from the SCPS. The SCPS were compiled biannually by the Bureau of Justice Statistics from administrative criminal justice records in a sample of the seventy-five largest counties by population. These counties include more than one-third of the United States population and half of all reported crimes (ICPSR 2016). In a given year, forty of the seventy-five counties are sampled based on a four-strata and two-stage design. First, the counties are allocated to four strata based on the number of filings. The ten counties in the first strata are selected with certainty; those in the second, third, and fourth strata are selected at random with decreasing probability of selection by strata. Second, defendants within counties are sampled by selecting from all felony cases filed in May of the survey year and, depending on the number of filings, either taking a week’s worth of cases (strata 1), two weeks (strata 2 and 3), or the full month (strata 4). The resulting individual-level data include detailed information on arrest charges, demographic characteristics, criminal history, pretrial release and detention, adjudication, and sentencing. We pool the available waves of the micro-data, covering the years 1990, 1992, 1994, 1996, 1998, 2000, 2002, 2004, 2006, and 2009. Unfortunately, no data in this series is available after 2009.

We merge this data with county-year-level information on several measures designed to capture the contextual processes that may shape pretrial release. These measures and their sources are described in the following section.

Key Variables

We first describe our three key dependent variables that capture the severity of pretrial detention practices and then detail the individual-level measures of demographics, case charges, and prior criminal justice history that we draw on from the SCPS. Finally, we detail the county-level measures that we have assembled from a variety of data sources.

Pretrial Detention

We construct three dependent variables that capture the severity of pretrial detention practices. First, we code whether a defendant is granted nonfinancial release. Defendants who receive nonfinancial release are coded as 1, and respondents who receive financial release or who are held on bail are coded as 0. Respondents who are granted emergency release, who are denied bail, whose release conditions are unknown, who are detained for unknown reasons, or whose cases are closed (7 percent of cases) are coded as missing.

Second, we measure the amount of bail for respondents who have bail set (and are either held on bail or released having made bail). We [End Page 132] inflation adjust the amount of bail to 2012 dollars using the CPI-U series and set bail values above the 99th percentile (approximately $800,000) to missing. Respondents in other release categories are set to missing values.

Third, we recode our measure of the amount of bail, by imputing $0 values for respondents who are granted nonfinancial release rather than setting these respondents to missing. Here too, we set bail values above the 99th percentile (approximately $600,000) to missing. Respondents in the residual release categories are set to missing.

Case-Level Predictors

We code three sets of case-level characteristics from the SCPS data: demographics, charges, and prior criminal justice contact.

Demographics

We code a dichotomous variable for gender equal to 1 if defendants are male. We include a continuous measure of age in years, setting ages greater than ninety to missing. We categorize defendants in terms of race-ethnicity as being white, non-Hispanic; black, non-Hispanic; Hispanic of any race; and other race, non-Hispanic.

Charges

We construct a fifteen-category variable that captures the most series-specific category of arrest charge, distinguishing murder, rape, robbery, assault, other violent crime, burglary, larceny-theft, motor vehicle theft, forgery, fraud, other property offense, drug sales, other drug offense, weapons offense driving related, or other public order. Although the dataset includes only those with felony charges, we also adjust for an indicator of whether the adjudication charge was ultimately a felony or a misdemeanor.

Prior Criminal Justice Contact

We construct a number of measures that capture defendants’ prior involvement with the criminal justice system. First, we code if the defendant had an active criminal justice status at arrest, a measure of whether the defendant had any prior arrests, and a measure of whether the defendant had any prior failures to appear. All are coded as dichotomous variables. We also construct a measure of the severity of any prior convictions coded 0 for those with none, 1 for those with misdemeanors, and 2 for those with a felony. Finally, we include a measure of the number of prior convictions that ranges from 0 to 10 (where 10 is inclusive of those with more than 10).

Contextual Predictors

Organizational Context

We construct a measure of the county jail occupancy rate using data from the Bureau of Justice Statistics’ Annual Survey of Jails. We divide the average daily population of jail facilities located in the county each year by the capacity of those jail facilities to derive the jail occupancy rate for each of the county-year observations represented in the SCPS data.

Political Context

We drew on information from government websites to construct a county-year time series of judicial selection processes, coding the initial selection as involving nonpartisan elections, partisan elections, or appointment, and then retention as involving nonpartisan elections, partisan elections, retention election, or reappointment. We coded counties dichotomously by whether judges faced election to maintain their positions. Judges facing popular elections (either contested or retention) were coded 1 and judges who maintained their position by appointment or political confirmation were coded 0. We also constructed a county-year time series of the partisan affiliation of county district attorneys (DAs) based on data from multiple sources in the public record, including county websites, election records, and newspaper coverage. All but three counties used in our analysis have a publicly elected district attorney. We construct a measure of DA partisanship that is equal to 0 if the DA is a Democrat and 1 otherwise. Finally, we code whether the governor is a Republican 1 or not 0 based on the University of Kentucky’s Center for Poverty Research’s National Welfare Data base (1980–2015).

Economic Context

We construct two measures of the economic context. First, we measure the county-level unemployment rate based on annual county-level data published by the Bureau of Labor Statistics through the Local Area Unemployment [End Page 133] Statistics program. Second, we use the series of state-level Gini coefficients assembled by Mark Frank from the Internal Revenue Service Statistics of Income (2014).

Demographic Context

We capture the demographic context of the county with a measure of the percentage of the county population that is black, dividing the number of black county residents by the total county population in each given county-year using data from the Census Bureau’s intercensal estimates.

Analysis

Our analysis proceeds in three parts. First, we take advantage of the merged 1990–2009 SCPS micro-data to document the variation in pre-trial detention practices across U.S. counties and over this nearly twenty-year period. We generate the county comparisons by selecting defendants in the forty counties covered by the data in either 2006 or 2009 and estimating ordinary least squares (OLS) regressions of our outcomes as a function of a set of county indicators and year fixed effects. We generate the time trends by selecting the full available analysis sample across all years and counties and estimating OLS regressions of our outcomes as a function of a set of year indicators and county fixed effects.1 We then present plots of the share of defendants in a given county or in a given year who are released on nonfinancial terms; the mean bail amounts adjusted to 2012 dollars in a given county or a given year, for those who are granted financial release or held on bail; and the mean bail amount set adjusted to 2012 dollars in a given county or in a given year, and those released on nonfinancial terms set to a bail of $0.

Second, we examine whether the between-county and over-time variation in pretrial detention practices can be explained by the composition of cases. We examine the extent to which these two sources of variation are accounted for by the demographic characteristics of those charged, the charges brought, and the prior criminal justice history of those charged.

For the between-county analysis, we again select defendants in the forty counties covered by the data in either 2006 or 2009. As before, we have three key dependent variables: defendants released on nonfinancial terms, bail conditional on financial release or being held on bail, and the amount of bail set unconditional on release type. In the first step, we regress each of the dependent variables on a vector of county indicators and a dummy for year. In the second step, we add a set of measures of defendant demographic characteristics—gender, age, and race-ethnicity. In the third step, we add measures of charges—a set of fifteen dummies for the charge and a dichotomous indicator of the level of the most serious adjudication charge. Finally, in the fourth step, we add measures of defendants’ prior criminal justice contact—status at arrest, number of prior arrests, prior failures to appear, number of prior convictions, and most serious prior conviction. After each step, we estimate predicted values (percentage receiving nonfinancial release or mean bail amount) for each county. We assess whether the coefficient of variation for the estimated county-specific fixed effects is reduced by controlling for each additive set of case characteristics.

For the over-time analysis, we pool all of the available data from 1990 through 2009 and follow a set of analytical steps similar to those described for counties. We estimate a first-step model by regressing each of the dependent variables on a vector of county indicators and a vector of year indicators. Then, as before, we sequentially and cumulatively add sets of indicators for defendant demographics, charges, and prior criminal justice history. We estimate predicted values for each dependent variable from each of the models. We assess whether the coefficient of variation for the estimated year-specific fixed effects is reduced by controlling for case characteristics.

Third, we turn from this individual-level analysis to a county-year level analysis in which we examine how county contexts are associated with the severity of pretrial detention policies. To do so, we collapse the SCPS micro-data to the county-year level and estimate OLS [End Page 134] models that take either the share of defendants in a given county-year released on nonfinancial terms, the mean bail amount, conditional on financial release or being held on bail, or the mean bail amount for all defendants (with those granted nonfinancial release assigned a bail of $0) in the county-year as the dependent variables. We estimate a first model that includes our county-level predictors as well as year and region fixed effects and then a second model that also adds attributes of the SCPS defendants, but aggregated to the county-year (with weights).2 We conduct this exercise in two steps because the same county-level factors that bear on pretrial detention may also shape who is arrested, what they are charged with, and what their prior criminal justice contact has been. By controlling for these factors, we may underestimate any total effects of county-level characteristics on pretrial detention.

Although no data are missing at the county-year level (because the file is built by aggregating up from all available individual values), we are required to censor county-year cell values based on small numbers of individual observations in the SCPS. In total, of 395 possible county-year observations, we are missing information on one or more variables suppressed in this way for forty-five observations. We impute forty-three of these forty-five using the county-specific mean for available years. However, in two instances, all of the county-year observations on a variable are missing, so we have no basis for the mean imputation. We delete these two rows. We are also missing data on at least one of the county-level predictors constructed from outside sources for eleven county-year observations.3 We also remove these by list-wise deletion. Finally, we also remove three county-year observations for which we suppressed data on the outcome variables due to small samples in the SCPS. Our final analysis sample, then, is 379 county-year observations, which is 95 percent of those available.

results

We organize our presentation of results into three parts, first describing the observed variation in nonfinancial release and bail amount across counties and over time, then examining whether this variation can be accounted for by case characteristics, and then examining the county-level correlates of pretrial detention severity.

Variation in Nonfinancial Release and Bail Amount Across Counties and Over Time

We first examine the degree of variation between counties and over time in our key measures of pre trial detention.

County Variation in Pretrial Detention

We begin by plotting the share of defendants, by county, who are granted nonfinancial release prior to adjudication against the mean amount of bail set (conditional on nonfinancial release not being granted). We graph this data for the forty counties represented in either the 2006 or 2009 waves of the SCPS. The result is shown in figure 1.

Examining the values for each county on the y-axis (share granted nonfinancial release), we see that there is significant variation across counties in the share of defendants who are granted a nonfinancial release (as opposed to either being released after making bail or being held after failing to make bail). Being released without making some financial bail is vanishingly uncommon in Harris County, Texas (home to Houston); Tarrant County, Texas (home to Fort Worth); and Orange County, Florida (home to Orlando), where fewer than 5 percent of those charged were granted nonfinancial release in those counties. Other large counties in Texas, Florida, and California released [End Page 135] a somewhat larger share—between 10 percent and 20 percent—on nonfinancial release, but still a relatively small one. At the other end of the distribution, the three New York City counties represented in the data (New York, Bronx, and Kings) each granted nonfinancial release to more than half of defendants; two-thirds of those in the Bronx granted nonfinancial release. The only other counties that came close to granting nonfinancial release to this share of defendants were King County, Washington; Salt Lake, Utah; Pima, Arizona; and Maricopa, Arizona.

Figure 1. Between-County Variation in Terms of Pre-Trial Detention, 2006/2009 (SCPS) Source: Authors’ calculations from the 1990–2009 SCPS (). Note: Plotted values county-specific estimates, adjusting for year of survey and weighted using the survey weights.
Click for larger view
View full resolution
Figure 1.

Between-County Variation in Terms of Pre-Trial Detention, 2006/2009 (SCPS)

Source: Authors’ calculations from the 1990–2009 SCPS (ICPSR 2016).

Note: Plotted values county-specific estimates, adjusting for year of survey and weighted using the survey weights.

The amount of cash bail, conditional on financial release, for each of these counties is plotted along the x-axis. Although the New York City counties are still among the least severe in terms of bail amount, they are now in the company of Orange, Florida; Harris, Texas; and Tarrant, Texas, counties that were very unlikely to grant nonfinancial release. At the other end of the scale, the large California counties have the highest mean bail amounts (conditional on granting financial release or holding on bail), joined by King County, Washington, and Cook County, Illinois, as well as several non–New York City eastern urban counties.

Taken together, we see some evidence of clustering. The counties with high rates of nonfinancial release and low mean bail conditional on financial terms, are in the top left of the plot—primarily the New York City counties. In contrast, counties with low rates of release and high levels of bail are at the far right—primarily California counties. In the bottom left corner, we see a set of Florida and Texas counties in which nonfinancial release is very uncommon, but bail amounts are relatively low.

Over-Time Variation in Pretrial Detention

These analyses focus only on the period between 2006 and 2009, pooling across those years to assess between-county variation in pre-trial detention practices, holding temporal change more or less constant. However, just as post-adjudication imprisonment has increased dramatically over the past several decades, so too may have the severity of pretrial practices. [End Page 136] Figure 2 plots the time trend in two of our key measures of pretrial detention. In each panel, the hollow circles represent predicted values of the dependent variable estimated from a model that also adjusts for a county fixed effect and is weighted using the survey weights. The lines show this series smoothed with a loess regression.

Figure 2. Time Trends in Terms of Pre-Trial Detention, 1990–2009 (SCPS) Source: Authors’ calculations from the 1990–2009 SCPS (). Note: Hollow circles represent estimates of year-specific predicted values from model with county fixed effects with survey weights. Solid lines apply loess smoother (bandwidth = 0.8).
Click for larger view
View full resolution
Figure 2.

Time Trends in Terms of Pre-Trial Detention, 1990–2009 (SCPS)

Source: Authors’ calculations from the 1990–2009 SCPS (ICPSR 2016).

Note: Hollow circles represent estimates of year-specific predicted values from model with county fixed effects with survey weights. Solid lines apply loess smoother (bandwidth = 0.8).

The left panel of figure 2 shows the dramatic decline between 1990 and 2009 in the share of defendants granted nonfinancial release. This share declines from about 40 percent in 1990 to just 27 percent in 2009. The most dramatic decline seems to be between the early 1990s and the late 1990s to early 2000s. The right panel of figure 2 plots the series for the mean level of bail (in 2012 dollars) over the same period. This measure is not conditional on having bail set; instead, it includes defendants released on nonfinancial bail and assigns them a $0 value. Here, mirroring the decline in nonfinancial release, we see a steady rise in the level of bail that is set, from about $12,000 in 1990 to almost $30,000 in 2009. Some of this increase is due to the decline in nonfinancial release, but that is by no means the whole story. If we separately examine the trend in bail, conditional on bail being set, we see that the mean level increased from $20,000 in 1990 to $40,000 by 2009 (not shown in figure 2).

Accounting for Variation as a Function of Case Characteristics

Variation is substantial both across counties and over time in the severity of pretrial detention as measured by the share of defendants granted nonfinancial release and by the level of bail set for those who are not. One possibility is that the variation is the product of compositional differences in the cases brought in different counties and in different periods. Research has carefully documented the role that case characteristics play in shaping pretrial detention (Goldkamp and Gottfredson 1985; Katz and Spohn 1995; Demuth 2003; Schlesinger 2005; Leiber and Fox 2005; Cohen and Reaves 2007; Spohn 2009). If some counties have different kinds of cases than others or if the composition [End Page 137] of cases changed over time, then that could explain county or temporal variation in pretrial detention practices.

Table 1. Coefficient of Variation on Adjusted County FE Estimates
Click for larger view
View full resolution
Table 1.

Coefficient of Variation on Adjusted County FE Estimates

Accounting for Between-County Variation

We begin with county-level variation, first augmenting the model that underlies figure 1 with adjustments for defendant gender, age, and race-ethnicity. Second, we add measures of the current charges brought in the case. Third, we add measures of the defendant’s history of criminal justice contact. Adjusting for these factors does essentially nothing to account for the between-county variation in the share of defendants granted nonfinancial release. Column 1 of table 1 presents the coefficient of variation calculated from the forty county-specific estimates of the share of defendants granted nonfinancial release, for the baseline model (with just year fixed effects) and then for each of the three models that add in other case characteristics. The coefficient of variation (scaled from 0 to 100) increases as we account for case characteristics. Rather than explain why some counties have less nonfinancial release than others, this analysis suggests that we would expect more between-county variation than we see given the between-county variation in case characteristics.

The same holds true when we examine the extent to which these case characteristics might account for between-county variation in the level of bail conditional on bail being set (column 2), and when we examine the inclusive measure of the level of bail unconditional on whether bail was set and assign those granted nonfinancial release a bail of $0 (column 3). Much of the between-county variation is simply not a function of differences in measurable case characteristics. In short, case composition is not the reason for the substantial variation between counties in pretrial detention practices.

Accounting for Over-Time Variation

We conduct a similar exercise to assess whether these case characteristics might account for the sharp decline between 1990 and 2009 in the share of defendants granted nonfinancial release and the sharp increase in the level of bail set. As before, we estimate a series of models that progressively add groups of case characteristics. For each model and for each outcome, we estimate the predicted values for each of the years in the data. We then calculate the coefficient of variation across the year estimates for each outcome and for each model, assessing the extent to which accounting for these characteristics reduces over-time variation (which we saw previously is essentially monotonic) in pretrial detention practices.

The only hint that these case characteristics matter is found in column 1 of table 2. We see that though accounting for defendant demographics and charges does not reduce the overtime variation, accounting for prior criminal justice contact does play a small role, reducing the coefficient of variation from 12.3 in the baseline model to 9.4. An implication of this result is that accrued criminal justice contact could serve to justify less nonfinancial release, potentially perpetuating entanglement in the criminal justice system. However, this logic does not hold for our other two measures of pretrial detention practices, as adjusting for case characteristics somewhat increases the coefficient of variation on the measures of bail amount. [End Page 138]

Table 2. Coefficient of Variation on Adjusted Year FE Estimates
Click for larger view
View full resolution
Table 2.

Coefficient of Variation on Adjusted Year FE Estimates

Table A1 reports the coefficients on the case characteristics from the full model (including defendant demographics, charges, and history of criminal justice contact). We see that even after adjusting for charges and history of criminal justice contact, men are less likely to be granted nonfinancial release and to have higher bail amounts than women, and that older defendants are more likely to be granted nonfinancial release. Hispanic defendants are less likely to be granted nonfinancial release and have higher bail amounts than white, non-Hispanic defendants. Interestingly, we do not see significant black-white gaps in nonfinancial release or in bail amount after adjusting for charges and prior criminal justice contact. However, we note a significant gap between blacks and whites in nonfinancial release without these adjustments and with the adjustments for charges but not criminal justice contact. The gap in bail amount is also significant when only adjusting for demographics. In short, charge severity and prior interactions with the criminal justice system contribute to black-white inequality in pretrial detention practices at the case-level. Turning to charges, we see the expected relationships between severity of charge and nonfinancial release and bail amount. Those charged with felonies are much less likely to be granted nonfinancial release and have much higher bail amounts. Finally, history of criminal justice contact also shapes pretrial detention. Those with prior failure to appears, prior arrests, and prior felony and violent felony convictions are significantly less likely to be granted nonfinancial release, and those with prior felony and violent felony convictions also receive higher bail amounts.

The Role of County Contexts in Shaping Pretrial Detention

These case-level characteristics play an important role in shaping pretrial detention outcomes for individual cases. But, we would also expect that contextual features of counties would affect pretrial detention practices. In table 3, we report estimates from a set of regression models that examine the association between time-varying county-level characteristics and county-level pretrial detention practices, aggregated up to the county-year level from the individual-level SCPS.

Models 1a and 1b examine the association between these county characteristics and the share of defendants in the county-year who are granted nonfinancial release. Models 2a and 2b examine bail for defendants who have financial terms set, and then models 3a and 3b examine the level of bail for all defendants, where those granted nonfinancial release are imputed a $0 bail.

Models 1a, 2a, and 3a include year and region fixed effects and models 1b, 2b, and 3b include both year and region fixed effects as well as time-varying county-year level measures of the defendant characteristics calculated from the individual-level SCPS data. We are somewhat agnostic about the two models. The risk with models 1a, 2a, and 3a is that we undercontrol for case characteristics that could confound the relationship between county characteristics and pretrial detention practices. The risk with models 1b, 2b, and 3b is that we overcontrol, and block one or more of the pathways by which county-level characteristics shape pretrial practices (for instance, if the proportion of blacks in a county shapes the characteristics of those arrested or if the partisan [End Page 139] affiliation of the district attorney shapes prior criminal justice contact).

Table 3. Relationship Between County-Level Characteristics and Pre-Trial Detention Practices
Click for larger view
View full resolution
Table 3.

Relationship Between County-Level Characteristics and Pre-Trial Detention Practices

We begin by examining how organizational context, operationalizing with the jail occupancy rate, shapes pretrial practices. In models 1a and 1b, we see that the coefficient on jail occupancy is positive, indicating that when occupancy rates are higher, defendants are more likely to be granted nonfinancial release. However, the coefficient is not statistically significant and the estimates of the relationship between jail occupancy rates and bail amounts in models 2a through 3b are inconsistent in direction and not significant.

We next examine how the political context is associated with pretrial detention practices. We measure three political dimensions of local criminal justice systems—the method by which judges are selected, the partisan affiliation of county district attorneys, and the partisan affiliation of the state governor. Here, we see no relationship between judges being elected rather than appointed and the share of defendants granted nonfinancial release. However, the relationship between judges being selected by election and the amount of bail set is positive. Those subject to election tend to set higher bail amounts as seen in the positive coefficients for models 2a and 3a and the positive and significant coefficients in models 3a and 3b.

The partisan affiliation of district attorneys is also strongly related to pretrial detention practices. In counties with non-Democratic district attorneys, defendants are granted nonfinancial release at significantly lower rates and set bail at significantly higher amounts. Figure 3 contrasts the mean bail amounts predicted from model 3b for DAs by partisan affiliation. Mean bail under Democratic DAs is $15,000 versus $22,000 under non-Democrats—a nearly 50 percent difference.

Defendants who go through the pretrial detention process when Republican governors are in office are significantly less likely to be granted nonfinancial release (models 1a and 1b), though the effect is only between one-third [End Page 140] and one-half as large as the effect on having a non-Democratic district attorney in the county. However, there is a significant negative association between Republican governors and bail amount in models 2a and 2b and no association in models 3a and 3b. One implication is that Republican governors preside over bail regimes that are less likely to grant nonfinancial release and then set lower bail on average given the pool of cases subject to a bail determination.

Figure 3. Mean Financial Bail by Political Party of District Attorney Source: Authors’ calculations from the 1990–2009 SCPS () and originally collected data. Note: Predicted values from model 3b.
Click for larger view
View full resolution
Figure 3.

Mean Financial Bail by Political Party of District Attorney

Source: Authors’ calculations from the 1990–2009 SCPS (ICPSR 2016) and originally collected data.

Note: Predicted values from model 3b.

Figure 4. Amount of Money Bail by County Unemployment Rate Source: Authors’ calculations from the 1990–2009 SCPS () and the Bureau of Labor Statistics Local Area Unemployment Statistics program. Note: Predicted values from model 2b.
Click for larger view
View full resolution
Figure 4.

Amount of Money Bail by County Unemployment Rate

Source: Authors’ calculations from the 1990–2009 SCPS (ICPSR 2016) and the Bureau of Labor Statistics Local Area Unemployment Statistics program.

Note: Predicted values from model 2b.

County-level economic context may also shape pretrial detention practices. There is a negative, but not statistically significant, association between county-level unemployment rate and the share of defendants granted nonfinancial release. However, a higher unemployment rate is positively and significantly related to the amount of bail in models 2a and 2b and in models 3a and 3b. Figure 4 plots the predicted amount of bail for those with financial terms set from model 2b against the county-level unemployment rate (over the range of values from the 5th through 95th percentile). We see that the predicted bail amount is about $25,500 when unemployment is very low, but [End Page 141] rises to almost $37,000 in the highest unemployment contexts.

Figure 5. Share of Defendants Granted Nonfinancial Release by Percentage of County Population African American Source: Authors’ calculations from the 1990–2009 SCPS () and Census Bureau intercensal estimates. Note: Predicted values from model 1b.
Click for larger view
View full resolution
Figure 5.

Share of Defendants Granted Nonfinancial Release by Percentage of County Population African American

Source: Authors’ calculations from the 1990–2009 SCPS (ICPSR 2016) and Census Bureau intercensal estimates.

Note: Predicted values from model 1b.

Our second measure of economic context is also associated with pretrial detention practices. The state-level Gini coefficient is strongly and significantly negatively related to the rate of nonfinancial release in models 1a and 1b. Defendants in more unequal places are less likely to be granted nonfinancial release. But, there is no such relationship for the amount of bail.

Finally, the demographic context, in particular the racial composition of the county, may also shape pretrial detention practices. In model 1a, we see that a larger percentage of blacks in the county is negatively associated with lower rate of nonfinancial release. Further, in model 1b, after controlling for case characteristics (including the share of defendants who are black), we find a large and statistically significant negative association between county percent black and the probability of nonfinancial release. Figure 5 plots the predicted share of defendants granted nonfinancial release by the percentage of the county that is black over the effective range of 0 percent (~p5) to 45 percent (~p95). We see that, controlling for the racial composition of defendants, 40 percent of defendants in counties with essentially no black population are granted nonfinancial release against just 25 percent of those in the counties with the largest share of black residents. Although county racial composition matters for nonfinancial release, we find no consistent evidence of a relationship with the amount of bail, but a negative and significant association in model 2b.

discussion

Bail and pretrial detention have long been an important but overlooked part of the criminal justice system. Yet in spite of growing interest from researchers and policymakers, systematic research on the topic remains limited. Most of what we do know comes from studies that focus narrowly on the relationship between individual case characteristics and pretrial outcomes, ignoring or eliding the considerable [End Page 142] geographic and temporal variation that exists in bail and pretrial practices. In this article, we focus instead on that variation. We document patterns of bail and pretrial release across both place and time, and look at whether and how those patterns relate to the larger organizational, political, socioeconomic, and demographic context in which pretrial decisions are made.

Our analysis reveals considerable variation in pretrial practices both across U.S. counties and over the nearly twenty-year period covered in our data. Whereas some counties released more than 50 percent of defendants awaiting trial with no financial conditions, others released fewer than 5 percent on nonfinancial terms. Average bail amounts ranged from less than $10,000 to more than $100,000. Nationwide, nonfinancial release has declined steadily over time, while bail amounts have doubled from $20,000 to $40,000. Yet, for the most part, this variation does not seem to result from differences in the kinds of cases being processed from one place to another or from one year to the next. Defendant demographics, charges, and prior record did nothing to account for pretrial differences between counties. Criminal history does account for some of the decrease in nonfinancial release over time, but declining use of nonfinancial release is unrelated to changes in demographics or charges, while the doubling of bail amounts during this period is not explained by any of these individual case characteristics. These patterns of bail and pretrial release are related to several contextual factors, however. The politicization of judicial offices, partisan affiliations of district attorneys and governors, income inequality, unemployment rates, and the size of the black population all seem to be related to bail-setting practices. Jail occupancy rates, however, are not.

Although variation in both time and place are associated with political, economic, and demographic factors, some of these relationships are easier to interpret than others. Local politics shape pretrial decisions in predictable ways. Elected judges are associated with higher bail amounts than their appointed counterparts, and Democrats in the district attorney’s office are linked to both higher rates of nonfinancial release and cheaper bail. This suggests, on the one hand, that judges concerned about reelection make more conservative pretrial decisions, and on the other that Democrats reflect a partisan preference for less-punitive pretrial regimes. Socioeconomic conditions are also important in predictable ways. Higher income inequality is linked to lower financial release, and higher unemployment rates are linked to more expensive bail amounts. In both cases, poorer socioeconomic conditions are related to more punitive pretrial practices.

Other contextual factors, however, show more complicated relationships to bail and pre-trial release. Republican governors are associated with lower rates of nonfinancial release, but they are also associated with cheaper bail. Larger black populations are similarly associated with both lower nonfinancial release but cheaper bail. These patterns may suggest a more complicated relationship between state-level politics and pretrial practices, as well as between those practices and local black populations. Alternatively, these patterns might suggest a more complicated relationship between nonfinancial release and bail amounts. Although both nonfinancial release and bail amount are important mechanisms for regulating pretrial detention, they may be governed by different dynamics not well captured in our analysis.

Our findings are descriptive in that they primarily are intended to characterize variation in pretrial detention practices over time and across counties and to show that case-level explanations appear largely insufficient for explaining this variation. However, we also pre sent evidence of associations between contextual-level factors and pretrial detention practices. The Bureau of Justice Statistics has itself stressed that the SCPS are insufficient on their own for making causal inferences (Cohen and Kyckelhahn 2010), and we stress here that we have not attempted to identify the causal effect of any of these measures or underlying constructs on pretrial detention practices. Future work that focuses on a particular explanatory factor could very usefully advance knowledge by attempting such identification. Further, we again emphasize that our descriptive work here is limited in representing only those particular [End Page 143] large urban counties that are sampled as part of the complex SCPS design. We are not able to generalize to pretrial detention practices more generally—whether for other urban counties or nationally (National Research Council 2009).

Still, these results suggest tremendous variation in the choices local jurisdictions make about how to manage the pretrial process, choices that do not simply reflect differences in the cases they handle. With distinct bail regimes operating in different counties, the costs of criminal justice contact can differ radically across county lines. Heavy reliance on money bail and high bail amounts increase the price of pretrial freedom for detainees and their families, putting greater financial burden on defendants who do bail out and leaving defendants unable to pay this price facing the long-term costs of detention. Studies show that detainees are more likely than released defendants to be convicted (Stevenson 2016) and receive harsher sentences when they are (Sacks and Ackerman 2014; Lowenkamp et al. 2013a; Phillips 2008; Williams 2003), while even short stays behind bars put people at risk of losing jobs, housing, property, access to government benefits, or custody of their children (Dobbie, Goldin, and Yang 2016; Comfort 2016). Different bail regimes thus have important consequences for inequality both in the criminal justice system and in society more broadly. Yet amid local variation, the trend nevertheless has been toward more punitive pretrial practices over time. Both the steady decline of nonfinancial release and the massive increase in bail amounts during these twenty years make pre-trial freedom substantially more expensive and increase the likelihood of pretrial detention across the country.

Going forward, future research on bail and pretrial detention should consider the sizable inequality in pretrial practices and the importance of the contextual factors we have identified here. Indeed, these findings suggest a number of possible avenues for further study. One valuable avenue for future research would be to decompose the within-county and between-county (and within-year and between-year) variation in pretrial detention practices and assess the contribution of case-level factors and county-level characteristics to this variance. Additionally, more research is needed to tease apart the role of political, socioeconomic, and demographic factors in pretrial practices, particularly the seemingly inconsistent relationships of both governor partisanship and race to bail decisions. In light of growing calls for bail reform, future studies should also investigate new types of pretrial management systems, and particularly the use of risk assessment tools. Studies suggest these tools hold considerable promise for reducing the total number of people held in pretrial detention, as well as reducing racial and ethnic disparities in detention practices (Kleinberg et al. 2017). We know little, however, about whether and how their efficacy might be shaped by the patterns we document here. Last, more historical and ethnographic accounts of bail and pretrial release practices could greatly enhance our understanding of the mechanisms driving the patterns we document, and perhaps provide a more complete understanding of how these different contextual factors interact and shape different pretrial regimes. [End Page 144]

Katherine Hood

Katherine Hood is a graduate student in sociology at the University of California, Berkeley.

Daniel Schneider

Daniel Schneider is assistant professor of sociology at the University of California, Berkeley.

Direct correspondence to: Katherine Hood at khood@berkeley.edu, 410 Barrows Hall, Department of Sociology, University of California Berkeley, Berkeley, CA 94703; and Daniel Schneider at djschneider@berkeley.edu, 480 Barrows Hall, Department of Sociology, University of California Berkeley, Berkeley, CA 94703.
Table A1. Associations Between Case-Level Characteristics and Pre-Trial Detention and Bail Amount
Click for larger view
View full resolution
Table A1.

Associations Between Case-Level Characteristics and Pre-Trial Detention and Bail Amount

REFERENCES

Angrist, Joshua D., and Jörn-Steffen Pischke. 2008. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton, N.J.: Princeton University Press.
Beck, Allen J., Thomas P. Bonczar, and Darrell K. Gilliard. 1993. Jail Inmates, 1992. Washington: U.S. Department of Justice, Bureau of Justice Statistics.
Blalock, Hubert M. 1967. Toward a Theory of Minority Group Relations. New York: John Wiley & Sons.
Bontrager, Stephanie, William Bales, and Ted Chiricos. 2005. “Race, Ethnicity, Threat, and the Labeling of Convicted Felons.” Criminology 43(3): 589–622.
Box, Steven, and Chris Hale. 1992. “Economic Crisis and the Rising Prisoner Population.” Crime and Social Justice 17(1): 20–35.
Bridges, George S., and Robert D. Crutchfield. 1988. “Law, Social Standing, and Racial Disparities in Imprisonment.” Social Forces 66(3): 699–724.
Britt, Chester L. 2000. “Social Context and Racial Disparities in Punishment Decisions.” Justice Quarterly 17(4): 707–32.
Caldeira, Greg A. 1983. “Elections and the Politics of Crime: Budgetary Choices and Priorities in America.” In The Political Science of Criminal Justice, edited by Stuart Nagel, Erika Fairchild, and Anthony Champaign. Springfield, Ill.: Charles C. Thomas.
Caldeira, Greg A., and Andrew T. Cowart. 1980. “Budgets, Institutions, and Change: Criminal Justice Policy in America.” American Journal of Political Science 24(3): 413–38.
Chiricos, Theodore G., and Miram A. Delone. 1992. “Labor Surplus and Punishment: A Review and Assessment of Theory and Evidence.” Social Problems 39(4): 421–46.
Cohen, Thomas H., and Tracey Kyckelhahn. 2010. Data Advisory: State Court Processing Statistics Data Limitation. Washington: U.S. Department of Justice, Bureau of Justice Statistics.
Cohen, Thomas H., and Brian A. Reaves. 2007. Pre-trial Release of Felony Defendants in State Court: State Court Processing Statistics, 1990–2004. Washington: U.S. Department of Justice, Bureau of Justice Statistics.
Comfort, Megan. 2016. “A Twenty-Hour-A-Day Job: The Impact of Frequent Low-Level Criminal Justice Involvement on Family Life.” Annals of the American Academy of Political and Social Science 665(1): 63–79.
D’Alessio, Stewart J., and Lisa A. Stolzenberg. 2002. “A Multilevel Analysis of the Relationship Between Labor Surplus and Pretrial Incarceration.” Social Problems 49(2): 178–93.
Davey, Joseph D. 1998. The Politics of Prison Expansion: Winning Elections by Waging War on Crime. Westport, Conn.: Praeger.
Demuth, Stephen. 2003. “Racial and Ethnic Differences in Pretrial Release Decisions and Outcomes: A Comparison of Hispanic, Black, and White Felony Arrestees.” Criminology 41(3): 873–908.
Demuth, Stephen, and Darrell Steffensmeier. 2004. “The Impact of Gender and Race-Ethnicity in the Pretrial Release Process.” Social Problems 51(2): 222–42.
Devine, F. E. 1991. Commercial Bail Bonding: A Comparison of Common Law Alternatives. New York: Praeger.
Dixon, Jo. 1995. “The Organizational Context of Criminal Sentencing.” American Journal of Sociology 100(5): 1157–98.
Dobbie, Will, Jacob Goldin, and Crystal Yang. 2016. “The Effects of Pre-Trial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges.” NBER working paper no. 22511. Cambridge, Mass.: National Bureau of Economic Research.
Eisenstein, James, Roy Flemming, and Peter Nardulli. 1988. The Contours of Justice: Communities and Their Courts. Boston, Mass.: Little, Brown.
Fearn, Noelle E. 2005. “A Multilevel Analysis of Community Effects on Criminal Sentencing.” Justice Quarterly 22(4): 452–87.
Feld, Barry C. 1991. “Justice by Geography: Urban, Suburban, and Rural Variations in Juvenile Justice Administration.” Journal of Criminal Law and Criminology 82(1): 156–210.
Flemming, Roy. 1982. Punishment Before Trial: An Organizational Perspective on Felony Bail Process. New York: Longman.
Frank, Mark. 2009. “Inequality and Growth in the United States: Evidence from a New State-Level Panel of Income Inequality Measure.” Economic Inquiry 47(1): 55–68.
Free, Marvin. 2004. “Bail and Pretrial Release Decisions.” Journal of Ethnicity in Criminal Justice 2(4): 23–44.
Garland, David. 2001. The Culture of Control: Crime and Social Order in a Contemporary Society. Chicago: University of Chicago Press.
Goldkamp, John, and Michael Gottfredson. 1985. Judicial Guidelines for Bail: The Philadelphia Experiment. Philadelphia, Pa.: Temple University Press.
Gottschalk, Marie. 2014. Caught: The Prison State and the Lockdown of American Politics. Princeton, N.J.: Princeton University Press.
Gupta, Arpit, Christopher Hansman, and Ethan Frenchman. 2016. “The Heavy Costs of High Bail: Evidence from Judge Randomization.” Columbia Law and Economics working paper no. 531. New York: Columbia Law School.
Harris, Kamala D., and Rand Paul. 2017. “To Shrink Jails, Let’s Reform Bail.” New York Times, July 20, A27.
Heaton, Paul, Sandra Mayson, and Megan Stevenson. 2017. “The Downstream Criminal Justice Consequences of Pretrial Detention.” Stanford Law Review 69(3): 711–94.
Helms, Ronald, and David Jacobs. 2002. “The Political Context of Sentencing: An Analysis of Immunity and Individual Determinants.” Social Forces 81(2): 577–604.
ICPSR. 2016. “State Court Processing Statistics, 1990–2009: Felony Defendants in Large Urban Counties (ICPSR 2038).” Machine Readable Database and Documentation.
Jacobs, David, and Ronald E. Helms. 1996. “Toward a Political Model of Incarceration: A Time-Series Examination of Multiple Explanations for Prison Admission Rates.” American Journal of Sociology 102(2): 323–57.
———. 1997. “Testing Coercive Explanations for Order: The Determinants of Law Enforcement Strength over Time.” Social Forces 75(4): 1361–92.
———. 1999. “Collective Outbursts, Politics, and Punitive Resources: Toward a Political Sociology of Spending on Social Control.” Social Forces 77(4): 1497–524.
Johnson, Brian. 2005. “Contextual Disparities in Guidelines Departures: Courtroom Social Contexts, Guidelines Compliance, and Extralegal Disparities in Criminal Sentencing.” Criminology 43(3): 761–96.
———. 2006. “The Multilevel Context of Criminal Sentencing: Integrating Judge- and County-Level Influences.” Criminology 44(2): 259–98.
Katz, Charles M., and Cassia C. Spohn. 1995. “The Effect of Race and Gender on Bail Outcomes: A Test of an Interactive Model.” American Journal of Criminal Justice 19:161–84.
Kleinberg, Jon, Himabindu Lakkaraju, Jure Leskovec, Jens Ludwig, and Sendhil Mullainathan. 2017. “Human Decisions and Machine Predictions.” Technical report. Cambridge, Mass.: National Bureau of Economic Research.
Leiber, Michael J., and Kristan C. Fox. 2005. “Race and the Impact of Detention on Juvenile Justice Decision-Making.” Crime & Delinquency 51(4): 470–97.
Lessan, Gloria T. 1991. “Macroeconomic Determinants of Penal Policy: Estimating the Unemployment and Inflation Influences on Imprisonment Rate Changes in the United States, 1948–1985.” Crime, Law and Social Change 16(2): 177–98.
Lowenkamp, Christopher T., Marie VanNostrand, and Alexander Holsinger. 2013a. “Investigating the Impact of Pretrial Detention on Sentencing Outcomes.” Houston, Tex.: Laura and John Arnold Foundation.
———. 2013b. “The Hidden Costs of Pretrial Detention” Houston, Tex.: Laura and John Arnold Foundation.
Minton, Todd D., and Zhen Zeng. 2015. “Jail Inmates at Midyear 2014.” Technical Report. Washington: U.S. Department of Justice, Bureau of Justice Statistics.
Myers, Martha A., and Susette M. Talarico. 1987. The Social Contexts of Criminal Sentencing. New York: Springer-Verlag.
National Association of Pretrial Services Agencies (NAPSA). 2009. “The Truth About Commercial Bail Bonding in America.” Advocacy Brief 1(1).
National Research Council. 2009. Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics, edited by Robert M. Groves and Daniel L. Cook. Washington, D.C.: National Academies Press.
Page, Joshua, Victoria Piehowski, and Joe Soss. 2019. “A Debt of Care: Commercial Bail and the Gendered Logic of Criminal Justice Predation.” RSF: The Russell Sage Foundation Journal of the Social Sciences 5(1): 150–72. DOI:10.7758/RSF.2019.5.1.07.
Petee, Thomas A. 1994. “Recommended for Release on Recognizance: Factors Affecting Pretrial Release Recommendations.” Journal of Social Psychology 134(3): 375–82.
Phillips, Mary T. 2007. “Bail, Detention, and Nonfelony Case Outcomes.” New York: City Criminal Justice Agency.
———. 2008. “Bail, Detention, and Felony Case Outcomes.” Research Brief Series no. 18. New York: Criminal Justice Agency.
Roth, Jeffrey, and Paul Wice. 1980. Pretrial Release and Misconduct in the District of Columbia. Washington, D.C.: Institute for Law and Social Research.
Rusche, Georg, and Otto Kirchheimer. 2003 [1939]. Punishment and Social Structure. New Brunswick, N.J.: Transaction Books.
Sacks, Meghan, and Alissa R. Ackerman. 2014. “Bail and Sentencing: Does Pretrial Detention Lead to Harsher Punishment?” Criminal Justice Policy Review 25(1): 59–77.
Schlesinger, Traci. 2005. “Racial and Ethnic Disparity in Pretrial Criminal Processing.” Justice Quarterly 22(2): 170–92.
Spitzer, Steven. 1975. “Toward a Marxian Theory of Deviance.” Social Problems 22(5): 638–51.
Spohn, Cassia C. 2009. “Race, Sex, and Pretrial Detention in Federal Court: Indirect Effects and Cumulative Disadvantage.” Kansas Law Review 57:879–901.
Stevenson, Megan. 2016. “Distortion of Justice: How the Inability to Pay Bail Affects Case Outcomes.” Unpublished working paper, University of Pennsylvania Law School.
Ulmer, Jeffery T., and Brian Johnson. 2004. “Sentencing in Context: A Multilevel Analysis.” Criminology 42(1): 137–77.
Wakefield, Sara, and Christopher Uggen. 2010. “Incarceration and Stratification.” Annual Review of Sociology 36: 387–406.
Wang, Xia, and Daniel Mears. 2015. “Sentencing and State-Level Racial and Ethnic Contexts.” Law and Society Review 49(4): 883–915.
Weaver, Vesla M., Andrew Papachristos, and Michael Zanger-Tishler. 2019. “The Great Decoupling: The Disconnection Between Criminal Offending and Experience of Arrest Across Two Cohorts.” RSF: The Russell Sage Foundation Journal of the Social Sciences 5(1): 89–123. DOI:10.7758/RSF.2019.5.1.05.
Weidner, Robert R., Richard S. Frase, and Jennifer S. Schultz. 2005. “The Impact of Contextual Factors on the Decision to Imprison in Large Urban Jurisdictions: A Multilevel Analysis.” Crime and Delinquency 51(3): 400–24.
Western, Bruce. 2006. Punishment and Inequality in America. New York: Russell Sage Foundation.
Western, Bruce, Meredith Kleykamp, and Jake Rosenfeld. 2004. “Crime, Punishment, and American Inequality.” In Social Inequality, edited by Katherine Neckerman. New York: Russell Sage Foundation.
Williams, Marion R. 2003. “The Effects of Pretrial Detention on Imprisonment Decisions.” Criminal Justice Review 28(2): 299–316.
Yeager, Milton G. 1979. “Unemployment and Imprisonment.” Journal of Criminal Law and Criminology 70(4): 586–88.

The authors gratefully acknowledge financial support from the UC Berkeley Department of Sociology and the Institute for Research on Labor and Employment. We received helpful comments and advice from David Harding, Chris Muller, Kristin Turney, Sara Wakefield, and participants in the Russell Sage Foundation conference “Criminal Justice Contact and Inequality.”

Footnotes

1. We use linear probability models for the dichotomous financial release—held on bail versus nonfinancial release outcome (Angrist and Pischke 2008).

2. Here, we use a slightly different set of case characteristics because of high multicolinearity in the models. We adjust for the share of defendants who are male, mean age, share white, non-Hispanic, share charged with violent offense, share with a property offense, share with a drug offense, share with a felony charge at adjudication, criminal justice status at arrest, mean number of prior arrests, share with any prior failure to appears, share with a prior misdemeanor conviction, share with a prior felony conviction, and mean number of prior convictions.

3. These eleven county-year cases are dropped due to missing data on jail capacity for Honolulu, New Haven, and Hartford.

Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.

inline graphic

This work is licensed under a Creative Commons Attribution 4.0 International License.

Share