U.S. Department of Justice Office of Justice Programs Bureau of Justice Statistics Special Report State Court Processing Statistics, 1990-2004 Pretrial Release of Felony Defendants in State Courts November 2007, NCJ 214994 ---------------------------------------------------------- This file is text only without graphics and many of the tables. A Zip archive of the tables in this report in spreadsheet format (.csv) and the full report including tables and graphics in .pdf format are available from: http://www.ojp.usdoj.gov/bjs/abstract/prfdsc.htm ----------------------------------------------------------- By Thomas H. Cohen, Ph.D. and Brian A. Reaves, Ph.D. BJS Statisticians ------------------------------------------------------------ Between 1990 and 2004, 62% of felony defendants in State courts in the 75 largest counties were released prior to the disposition of their case. Beginning in 1998, financial pretrial releases, requiring the posting of bail, were more prevalent than non-financial releases. This increase in the use of financial releases was mostly the result of a decrease in the use of release on recognizance (ROR), coupled with an increase in the use of commercial surety bonds. These findings are from a multi-year analysis of felony cases from the biennial State Court Processing Statistics (SCPS) program, sponsored by the Bureau of Justice Statistics. Among defendants detained until case disposition, 1 in 6 had been denied bail and 5 in 6 had bail set with financial conditions required for release that were not met. The higher the bail amount set, the lower the probability of release. About 7 in 10 defendants secured release when bail was set at less than $5,000, but this proportion dropped to 1 in 10 when bail was set at $100,000 or more. Murder defendants were the least likely to be released pretrial. Defendants charged with rape, robbery, burglary, and motor vehicle theft also had release rates lower than the overall average. The highest release rate was for defendants charged with fraud. Defendants were less likely to be released if they had a prior arrest or conviction or an active criminal justice status at the time of arrest (such as those on probation or parole). A history of missed court appearances also reduced the likelihood that a defendant would be released. About a third of released defendants were charged with one or more types of pretrial misconduct. Nearly a fourth had a bench warrant issued for failing to appear in court, and about a sixth were arrested for a new offense. More than half of these new arrests were for felonies. Logistic regression analyses that controlled for factors such as offense and criminal history found that Hispanics were less likely than non-Hispanic defendants to be released, and males were less likely than females to be released. Logistic regression was also used to calculate the probability of pretrial misconduct for defendants with a given characteristic, independent of other factors. Characteristics associated with a greater probability of being rearrested while on pretrial release included being under age 21, having a prior arrest record, having a prior felony conviction, being released on an unsecured bond, or being part of an emergency release to relieve jail crowding. Compared to release on recognizance, defendants on financial release were more likely to make all scheduled court appearances. Defendants released on an unsecured bond or as part of an emergency release were most likely to have a bench warrant issued because they failed to appear in court. The probability of failing to appear in court was higher among defendants who were black or Hispanic, had an active criminal justice status at the time of arrest, or had a prior failure to appear. About 3 in 5 felony defendants in the 75 largest counties were released prior to case disposition From 1990 to 2004, an estimated 62% of State court felony defendants in the 75 largest counties were released prior to the disposition of their case (table 1). Defendants were about as likely to be released on financial conditions requiring the posting of bail (30%) as to be granted a non-financial release (32%). Among the 38% of defendants detained until case disposition, about 5 in 6 had a bail amount set but did not post the financial bond required for release. From 1990 to 2004, surety bond (33%) and release on recognizance (32%) each accounted for about a third of all releases. Other release types that accounted for at least 5% of releases during this period were conditional release (12%), deposit bond (9%), unsecured bond (7%), and full cash bond (5%). (See box on page 3 for definitions of release types.) Since 1998 a majority of pretrial releases have included financial conditions Except for a decline to 57% in 2004, the percentage of defendants released each year varied only slightly, from 62% to 64%. A more pronounced trend was observed in the type of release used (figure 1). From 1990 to 1998, the percentage of released defendants under financial conditions rose from 24% to 36%, while non-financial releases dropped from 40% to 28%. Surety bond surpassed release on recognizance in 1998 as the most common type of pretrial release The trend away from non-financial releases to financial releases was accompanied by an increase in the use of surety bonds and a decrease in the use of release on recognizance (ROR) (figure 2). From 1990 through 1994, ROR accounted for 41% of releases, compared to 24% for surety bond. In 2002 and 2004, surety bonds were used for 42% of releases, compared to 23% for ROR. Two-thirds of defendants had financial conditions required for release in 2004, compared to half in 1990 Including both released and detained defendants, the percentage required to post bond to secure release rose from 53% in 1990 to 68% in 2004 (not shown in table). Overall, about half (48%) of defendants required to post bail for release did so. From 1998 through 2004, 51% posted bail, compared to 45% in prior years. The higher the bail amount the lower the probability of pretrial release The median bail amount for detained defendants ($15,000) was 3 times that of released defendants ($5,000); the mean amount was about 5 times higher ($58,400 versus $11,600) (not shown in table). For all defendants, the median bail amount was $9,000 and the mean was $35,800. There was a direct relationship between the bail amount and the probability of release. When the bail was under $10,000, most defendants secured release, including 7 in 10 defendants with bail under $5,000 (figure 3). The proportion released declined as the bail amount increased, dropping to 1 in 10 when bail was $100,000 or higher. Defendants arrested for violent offenses or who had a criminal record were most likely to have a high bail amount or be denied bail Courts typically use an offense-based schedule when setting bail. After assessing the likelihood that a defendant, if released, will not appear in court and assessing any danger the defendant may present to the community, the court may adjust the bail higher or lower. In the most serious cases, the court may deny bail altogether. The use of a high bail amount or the denial of bail was most evident in cases involving serious violent offenses. Eighty percent of defendants charged with murder had one of these conditions; with rape, 34%; and with robbery, 30% (table 2). Defendants who had an active criminal justice status (13%) were about 4 times as likely as other defendants (3%) to have bail denied. Defendants with 1 or more prior felony convictions (10%) were more than twice as likely as those without such a conviction (4%) to have bail denied. ------------------------------------------------------------ Commercial bail and pretrial release An estimated 14,000 commercial bail agents nationwide secure the release of more than 2 million defendants annually, according to the Professional Bail Agents of the United States. (See Methodology for other sources on bail and pretrial release.) Bond forfeiture regulations and procedures vary by jurisdiction, but most States regulate commercial bail and license bail agents through their departments of insurance. Four States do not allow commercial bail: Illinois, Kentucky, Oregon, and Wisconsin. Also, the District of Columbia, Maine, and Nebraska have little commercial bail activity. Bail agents generally operate as independent contractors using credentials of a surety company when posting appearance bond for their client. For a fee, the surety company allows the bail agent to use its financial standing and credit as security on bonds. In turn, the bail agent charges the defendant a fee (usually 10% of the bail amount) for services. In addition, the bail agent often requires collateral from the defendant. A bail agent usually has an opportunity to recover a defendant if they fail to appear. If the defendant is not returned, the agent is liable to the court for the full bail amount. Most jurisdictions permit revocation of the bond, which allows the agent to return the defendant to custody before the court date, freeing the agent from liability. The agent may be required to refund the defendant's fee in such cases. Courts can also set aside forfeiture judgments if good cause is shown as to why a defendant did not appear. Commercial bail has been a target of critics since the 1960s. Some organizations, such as the American Bar Association and the National District Attorney's Association, have recommended its abolishment. Some critics have succeeded in obtaining reforms in the release process, beginning with the Manhattan Bail Project in 1961. This project showed that defendants could be successfully released pretrial without the financial guarantee of a surety bail agent if verified information concerning their stability and community ties were presented to the court. The success of the Manhattan Bail Project resulted in a wide range of pretrial reforms in the Federal system, culminating in the Bail Reform Act of 1966. This Act created a presumption in favor of release for most non-capital defendants and led to the creation of non-surety release options, such as refundable deposit bail and conditional release. Many States followed the Federal system and created such release options. The Bail Reform Act of 1984 set forth new procedures which allowed the pretrial detention of defendants believed to be a danger to the community in addition to a flight risk. ----------------------------------------------------------- Financial releases took longer on average than non-financial releases About half of all pretrial releases occurred within 1 day of arrest, and about three-fourths within 1 week. Non-financial releases (59%) were more likely to occur within a day of arrest than financial releases (45%). For all release types, more than 90% occurred within 1 month of arrest. Among defendants released under financial conditions, the amount of time from arrest to release increased with bail amounts, ranging from a mean of 8 days for those with a bail amount of less than $5,000 to 22 days for bail amounts of $50,000 or more (not shown in table). About a quarter of released defendants had failed to appear in court during a prior case A majority (61%) of the defendants released into the community to await disposition of their case had been arrested previously (table 3). This included 27% who had failed to appear in court during a prior case. About half had 1 or more prior convictions (48%), and nearly a third (30%) had at least one prior felony conviction. About 1 in 4 released defendants had an active criminal justice status from a prior case at the time of their arrest. ----------------------------------------------------------- The role of pretrial services programs in the release process According to a BJA nationwide study, about 300 pretrial services programs were operating in the U.S. during 2001.***Footnote: John Clark and D. Alan Henry, Pretrial Services Programming at the Start of the 21st Century: A Survey of Pretrial Services Programs, Washington D.C.: Bureau of Justice Assistance, July 2003 (NCJ 199773)***. More than two-thirds of the programs had begun since 1980 and nearly half since 1990. The programs operated in a variety of administrative settings, including probation offices, courts, sheriffs' offices, independent agencies, and private non-profit organizations. Pretrial programs play an important role in the release process. Standards published by the American Bar Association and the National Association of Pretrial Services Agencies have specified core functions a model pretrial program should provide. Information gathering and assessment An important function of a pretrial program is to conduct a pretrial investigation to assist judicial officers in making release decisions. Prior to the initial court appearance, the pretrial program gathers information about the defendant, primarily through voluntary interviews and records checks. Some defendants may not be eligible for pretrial release because of the severity of the charged offense or an existing criminal justice status such as parole, probation, or an outstanding warrant. Information collected from the pretrial investigation typically includes: *residency *employment status *community ties *criminal record *court appearance record *criminal justice status *mental health status *indications of substance abuse Often a risk assessment tool is used to incorporate the information from the pretrial investigation into a score that guides the release decision. Periodic validation of the instrument ensures that it provides an accurate, unbiased measure of a defendant's potential for misconduct if released. Supervision and follow-up Pretrial services programs provide supervision and monitoring of a defendant's compliance with release conditions, such as testing for drug or alcohol use and electronic monitoring of defendants confined to a restricted area. These programs also assist with locating and returning defendants who fail to appear in court. Such assistance may include providing information to law enforcement officials or working directly with defendants to persuade them to return. Pretrial programs may regularly review the status of detained defendants for changes in their eligibility for release and facilitate their release where appropriate. -------------------------------------------------------- Prior criminal activity was more prevalent among pretrial detainees. About half had a criminal justice status at the time of arrest. A large majority had prior arrests (83%) and convictions (75%). More than half (57%) had a prior felony conviction, including 15% with a conviction for a violent felony. Nearly half (44%) had a prior failure to appear. Many factors influence the pretrial release decision SCPS collects information on some of the factors courts consider when making pretrial release decisions, such as arrest offenses, criminal justice status, prior arrests, prior court appearance record, and prior convictions. It does not collect data on residency, employment status, community ties, mental health status, or substance abuse history. The unique contribution of the factors collected in SCPS to the release decision can be assessed using logistic regression techniques. Logistic regression produces nonlinear estimations for each independent variable which can be transformed into predicted probabilities (table 4). In the case of pretrial release, the logistic regression analyses yielded patterns similar to that of the bivariate results. (See Methodology for more information on the logistic regression techniques). Murder defendants (19%) had the lowest probability of being released, followed by those charged with robbery (44%), burglary (49%), motor vehicle theft (49%), or rape (53%). Defendants charged with fraud (82%) were the most likely to be released. Male and Hispanic defendants less likely to be released than females and whites Female defendants (74%) were more likely than males (60%) to be released pretrial. By race and Hispanic origin, non-Hispanic whites (68%) had a higher probability of release than Hispanics (55%). Pretrial detention rates for Hispanics may have been influenced by the use of immigration holds to detain those illegally in the U.S. Defendants with a prior criminal record less likely to be released than those without a prior arrest Defendants on parole (26%) or probation (43%) at the time of their arrest for the current offense were less likely to be released than those without an active criminal justice status (70%). Defendants who had a prior arrest, whether they had previously failed to appear in court (50%) or not (59%), had a lower probability of release than those without a prior arrest (79%). Defendants with a prior conviction (51%, not shown in table) had a lower probability of being released than those without a conviction (77%). This was true even if the prior convictions were for misdemeanors only (63%). The effect of a conviction record on release was more pronounced if the defendant had at least one prior felony conviction (46%). About 1 in 5 detained defendants eventually had their case dismissed or were acquitted Sixty percent of released defendants were eventually convicted -- 46% of a felony and 14% of a misdemeanor (table 5). Conviction rates were higher for detained defendants, with 78% convicted, including 69% of a felony. On average, released defendants waited nearly 3 times longer than detainees for case adjudication Released defendants waited a median of 127 days from time of arrest until adjudication, nearly 3 times as long as those who were detained (45 days). For those released, the average time from release to adjudication was nearly 1 month longer for those on financial release (125 days) than for those released under non-financial conditions (101 days) (table 6). By specific release type, defendants released on recognizance had the shortest wait (98 days), while those released on property bond had the longest (140 days). Incidents of pretrial misconduct increased with length of time in release status The number of defendants charged with pretrial misconduct increased with the length of time spent in a release status. About a third (32%) of failure-to-appear bench warrants were issued within a month of release and about two-thirds (68%) within 3 months. The pattern was similar for rearrests, with 29% occurring within 1 month of release and 62% within 3 months. A third of released defendants were charged with pretrial misconduct within 1 year after release From 1990 through 2004, 33% of defendants were charged with committing one or more types of misconduct after being released but prior to the disposition of their case (figure 4). A bench warrant for failure to appear in court was issued for 23% of released defendants. An estimated 17% were arrested for a new offense, including 11% for a felony. Pretrial misconduct rates stable from 1990-2004 Overall misconduct rates varied only slightly from 1990 through 2004, ranging from a high of 35% to a low of 31% (figure 5). For failure to appear, the range was from 21% to 24%, and the fugitive rate ranged from 5% to 8%. Overall rearrest rates ranged from 13% to 21%, and felony rearrest rates from 10% to 13%. Pretrial misconduct rates highest for emergency releases About half (52%) of the 1% of defendants released under an emergency order to relieve jail crowding were charged with some type of misconduct (table 7). Pretrial misconduct rates for other types of releases ranged from 27% to 36%. After emergency release (45%), the highest failure-to-appear rate was for defendants released on unsecured bond (30%). Property bond (14%), which also accounted for just 1% of releases, had the lowest failure-to-appear rate followed by surety bond (18%). About 1 in 4 defendants who failed to appear in court were fugitives at end of a 1-year study period By type of release, the percent of the defendants who were fugitives after 1 year ranged from 10% for unsecured bond releases to 3% of those released on surety bond. Overall, 28% of the defendants who failed to appear in court and had a bench warrant issued for their arrest were still fugitives at the end of a 1-year study period. This was 6% of all defendants released pretrial (not shown in table). Compared to the overall average, the percentage of absconded defendants who remained a fugitive was lower for surety bond releases (19%). Likelihood of pretrial misconduct lower for defendants released after being charged with murder or rape Defendants released after being charged with murder (19%) or rape (18%) had misconduct rates that were about half that for defendants charged with motor vehicle theft (39%), drug trafficking (39%), or burglary (37%). Younger, male, black, and Hispanic defendants more likely to be charged with pretrial misconduct Released defendants age 20 or younger (33%) had higher misconduct rates than those age 40 or older (28%). This pattern also existed for rearrest and failure-to-appear rates. Male defendants (34%) had a higher misconduct rate than females (28%). Black (36%) and Hispanic (34%) defendants had a higher misconduct rate than whites (28%). Prior criminal activity associated with greater probability of pretrial misconduct Defendants who had an active criminal justice status at the time of arrest -- such as pretrial release (48%), parole (47%), or probation (44%) -- had a higher misconduct rate than those who were not on a criminal justice status (27%). This difference was observed for both failure to appear and rearrest. Defendants with a prior failure to appear (49%) had a higher misconduct rate than defendants who had previously made all court appearances (30%) or had never been arrested (23%). Defendants with a prior failure to appear (35%) were about twice as likely to have a bench warrant issued for failing to appear during the current case than other defendants (18%). Defendants with at least one prior felony conviction (43%) had a higher rate of pretrial misconduct than defendants with misdemeanor convictions only (34%) or no prior convictions (27%). -------------------------------------------------------- Logistic regression analysis of pretrial misconduct Logistic regression was used to assess the impact of given characteristics independent of other factors on the probability of a released defendant being charged with pretrial misconduct. The predicted probabilities generated from these analyses are presented in the adjacent table. (See Methodology for more information on logistic regression). Type of release Predicted overall misconduct rates were higher for unsecured bond (42%) and emergency (56%) releases. This was also the case for rearrest and failure to appear rates. Property (17%), surety (20%), deposit (20%), and full cash (20%) bonds all had lower predicted failure-to-appear rates than recognizance (24%). The percent of released defendants predicted to be fugitives after 1 year was lowest for property (3%) and surety bonds (4%). Emergency release and property bonds each accounted for 1% of all releases, compared to about 30% each for surety bonds and recognizance. (See table 7 for the number of defendants accounted for by each type of pretrial release). Arrest offense Drug trafficking defendants (38%) had higher predicted rates of overall misconduct, rearrest and failure-to-appear than defendants charged with murder (19%), rape (21%), assault (26%), fraud (29%), or a weapons offense (31%). Demographic characteristics Defendants age 20 or younger (39%) had a higher predicted misconduct rate than those ages 21 to 39 (35%) or age 40 or older (30%). This pattern held for rearrest, but for court appearance record only defendants age 40 or older were predicted to perform better than those under age 21. Male defendants (35%) were predicted to have a higher misconduct rate than females (32%). Hispanic (37%) and black (36%) defendants were predicted to be charged with misconduct more often than whites (32%). This difference also existed for failure to appear, but not rearrest. Criminal history Defendants with an active criminal justice status at the time of arrest, such as parole (42%), probation (39%), or pretrial release (42%), had higher predicted misconduct rates than those without such a status (33%). This difference was observed for both failure to appear and rearrest. Compared to those without prior arrests (29%), defendants with an arrest record were predicted to be charged with misconduct more often, especially if they had previously failed to appear in court (47%). This pattern was observed for both failure to appear and rearrest. Defendants with prior felony convictions (39%) had a higher predicted misconduct rate than other defendants (33%). This pattern also existed for rearrest, but not failure to appear. ------------------------------------------------------- Methodology Data utilized This report analyzed data from the State Court Processing Statistics (SCPS) series, covering felony cases filed in May of even-numbered years from 1990 through 2004. SCPS is a biennial data collection series that examines felony cases processed in a sample of 40 of the Nation's 75 most populous counties. The counties included in the sample have varied over time to account for changing national population patterns. For a year-by-year summary of the counties participating in SCPS, see Appendix table 1. For more information on the SCPS methodology see the BJS report Felony Defendants in Large Urban Counties, 2002 at http://www.ojp.usdoj.gov/bjs/abstract/fdluc02.htm>. Each SCPS data collection tracks approximately 15,000 felony cases for up to one year, with the exception of murder defendants who are followed for up to two years. In addition to defendant demographic characteristics and criminal history, SCPS also obtains data on a variety of felony case processing factors, including the types of arrest charges filed, conditions of pretrial release such as bail amount and type of release, and instances of pretrial misconduct including failure to appear in court, rearrest while on pretrial release, and other violations that resulted in the revocation of release. Adjudication and sentencing outcomes are also recorded. Using multivariate statistical techniques This report analyzes pretrial release and misconduct through both bivariate and multivariate statistical techniques. While the bivariate statistics provide a descriptive overview of pretrial release and misconduct among felony defendants in the 75 most populous counties, multivariate analysis can help disentangle the impacts that independent variables such as demographic characteristics, prior criminal history, severity of arrest charges, and release type have on dependent variables such as the probability of pretrial release and misconduct. Logistic regression models were used to estimate the probability of pretrial release and misconduct. This is one widely accepted method for analyzing the effects of multiple independent factors on dichotomous or binomial outcomes. The regression analyses excluded data from 1990 because of the large number of cases missing data on race or Hispanic origin. The regression models also excluded cases that had missing data on either the independent or dependent variables. This resulted in reductions in the number of cases analyzed. From 1992 through 2004, 99,899 felony defendants were either released or detained, but when missing data were excluded from the regression models, the number of cases analyzed declined to 71,027. To determine the impact of missing data, logistic regression models excluded certain independent factors to increase the number of analyzed cases. Since the results from these analyses did not differ appreciably from the full model, missing data did not affect the results. SCPS data are drawn from a sample and weighted to represent cases processed in the 75 most populous counties during the month of May. When the regressions used these weighted data, the large number of weighted cases resulted in statistical significance for nearly all the variables in the model. Effect weighting was employed to address this issue. Through effect weighting, the SCPS data were weighted to the number of cases actually sampled rather than the number of cases in the universe represented by the sample. Generalized estimating equation techniques One primary assumption of binary logistic regression is that all observations in the dataset are independent. This assumption is not necessarily appropriate for the SCPS series because the data are collected on a county basis. The county-based nature of SCPS creates a presumption of clustered data. In clustered datasets, "the data can be grouped into natural or imposed clusters with observations in the same clusters tending to be more alike than observations in different clusters."*** Footnote: Paul D. Allison, 2001. Logistic Regression Using the SAS System: Theory and Application, Cary, N.C.: SAS Institute Inc., page 179***. The clustered nature of the SCPS data was handled by utilizing generalized estimating equation (GEE) techniques. Logistic regression modeling with generalized estimating equation (GEE) techniques provides for more efficient computation of regression coefficients and more robust standard error estimates. Interpreting logistic regression probabilities Logistic regression produces nonlinear estimations for each independent variable that can be difficult to interpret. In this report, the logistic regression coefficients are made interpretable by transforming them into predicted probabilities (see table 4 and box on page 10). The predicted probabilities were calculated by setting all independent variables to their mean levels, setting the independent variable of interest to a value of one, multiplying the means of each independent variable by their respective logistic regression parameter estimates, taking the exponential function of the summed product of means and parameter estimates, and then calculating the probability of that exponential function. Limitations of models The logistic regression analyses were limited and intended to reflect the effects of only selected factors that were available in the SCPS data. Other factors could potentially be related to pretrial release and misconduct. Examples of these include: defendants' residence, employment status, community ties, mental health status, and substance abuse. If data on these variables were available, the logistic regression results could be altered. Additional sources on bail and pretrial release include: Demuth, Stephen. 2003. "Racial and Ethnic Differences in Pretrial Release Outcomes: A Comparison of Hispanic, Black, and White Felony Arrestees." Criminology: 41(3): 873. Feeley, Malcolm M., 1983. "Court Reform on Trial: Why Simple Solutions Fail." New York, NY: Basic Books, Inc., pp. 40-79. Goldkamp, John. 1985. "Danger and Detention: A Second Generation of Bail Reform." Journal of Criminal Law and Criminology: 76:1. Helland, Eric and Alexander Tabarrok, 2004. "Public versus Private Law Enforcement: Evidence from Bail Jumping." Journal of Law and Economics: (47): pp. 93-122. Kennedy, Spurgeon and D. Alan Henry. 1996. Commercial Surety Bail: Assessing Its Role in the Pretrial Release and Detention Decision . Watson, Jerry W. and L. Jay Labe, 2001. "Bail Bonds". The Law of Miscellaneous and Commercial Surety Bonds, Chicago, IL: American Bar Association, pp. 127-142. ----------------------------------------------------------- ----------------------------------------------------------- The Bureau of Justice Statistics is the statistical agency of the U.S. Department of Justice. Jeffrey L. Sedgwick is the director. This Special Report was written by Thomas H. Cohen, Ph.D., and Brian A. Reaves, Ph.D. William J. Sabol, Ph.D. provided technical assistance. Tracey Kyckelhahn provided verification. Tina Dorsey produced and edited the report, under the supervision of Doris J. James. Jayne Robinson prepared the report for final printing. November 2007, NCJ 214994 ----------------------------------------------------------- ----------------------------------------------------------- This report in portable document format and in ASCII and its related statistical data and tables are available at the BJS World Wide Web Internet site: . ----------------------------------------------------------- End of file 11/19/2007 JR