Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The Immediate Consequences of Federal Pretrial Detention

The Immediate Consequences of Federal Pretrial Detention Abstract Unlike the cash-bail regimes that are prevalent in state courts, federal courts rarely use money bail as a condition of pretrial release. Nonetheless, this article presents evidence that pretrial release influences case outcomes for federal defendants. Using case data spanning 71 federal district courts, the article suggests that pretrial release reduces a defendant’s sentence and increases the probability that they will receive a sentence below the recommended sentencing range. Pretrial release also appears to lessen the probability that a defendant will receive a mandatory minimum sentence when one is charged. The analysis exploits variation in magistrate judges’ propensities to release defendants pending trial, which allows magistrate judge leniency to serve as an instrumental variable for pretrial release. The article also provides suggestive evidence that pretrial release affects case outcomes through two channels: first, by giving defendants the opportunity to present mitigating evidence at sentencing and second, by making it easier for defendants to earn a sentencing reduction by providing assistance to the government. 1. Introduction Over the last several years, advocates, policymakers, and scholars have increasingly called for changes to bail regimes for those arrested in the United States. Much of the current effort to reform pretrial detention has focused on the money-bail systems that are ubiquitous in state courts. One concern that animates this debate—and finds support in recent literature—is that cash bail can induce indigent defendants to plead guilty simply to get out of jail and return home. It is plausible, however, that pretrial detention affects case outcomes in other ways and for defendants charged with serious offenses, as well. To test this possibility, this article considers a setting in which pretrial detention is not directly connected to financial means: the federal criminal system. Unlike bail practice in many state court systems, federal defendants are rarely detained based on their inability to pay. Nonetheless, this article finds that pretrial detention significantly affects how federal defendants fare in court, and provides evidence about the mechanisms animating these effects. Near the beginning of a typical criminal case, a judicial officer determines whether a criminal defendant should be held in pretrial detention or released for the duration of their case.1 It is possible—but not certain—that a defendant’s pretrial status could affect the outcome of their case, for several reasons. First, a released defendant is more equipped to participate in their defense, as Caleb Foote observed in early work on the administration of bail in Philadelphia (Foote, 1954). Subsequent qualitative work has identified many channels through which pretrial detention can hinder a defendant’s defense, for example by making it difficult for a defendant to meet with their lawyer, to secure witnesses who can help their case, and to obtain relevant case documents (Williams, 2003). These disadvantages may translate into a decreased likelihood of acquittal or an inferior position in plea bargaining and at sentencing. Second, pretrial release affords a defendant the opportunity to undertake activities that build a strong mitigation case at sentencing, which can be difficult for those who are held in custody while their case is pending (Foote, 1954; Williams, 2003; Sacks and Ackerman, 2014). For example, pretrial release provides a defendant with the chance to demonstrate to their prosecutor and sentencing judge that they can resist criminal activity.2 Pretrial release also offers a defendant the opportunity to undertake mitigating activities such as maintaining a job and supporting dependents (Sacks and Ackerman, 2014). Such mitigation is important to sentencing judges: in a recent survey, the defendant’s post-offense rehabilitative effort was one of the factors about which federal judges reported caring the most in sentencing (United States Sentencing Commission, 2010). Third, defendants might benefit or suffer from representativeness bias based on detention status (Kahneman and Tversky, 1972; Kahneman and Frederick, 2002). In particular, representativeness bias could cause released defendants near the margin of detention to be thought of as less likely to recidivate and less dangerous than defendants near the margin who are detained, simply because released defendants are in general less likely to recidivate and less dangerous. On the other hand, pretrial release could lead to worse outcomes for defendants who violate the terms of release—for example by failing a drug test, committing a crime while released, or failing to appear in court. The idea that a defendant’s pretrial detention status might affect the outcome of their case is troubling for several reasons. For one thing, judges make detention decisions quickly and early on in a case—before the defendant and their attorney have time to fully prepare. For another thing, to the extent that a defendant’s ability to earn release is related to their race (Ayres and Waldfogel, 1994; Arnold et al., 2018), racial disparities in pretrial release could easily translate into unwarranted disparities in sentencing. This article suggests strong effects of pretrial release at the sentencing phase, but not earlier on in a case. Pretrial release appears to decrease sentence length, increase the probability of receiving a below-Guidelines sentence, and decrease the probability of receiving a mandatory minimum sentence, if facing one. Pretrial release also increases the probability that a defendant will receive a sentencing reduction for providing substantial assistance to the government. The data suggest that the effects of detention chiefly work through two channels: by hampering a defendant’s ability to provide mitigating information at sentencing and by making it harder for a defendant to assist the government. The ability to present mitigation evidence at sentencing is particularly important in federal cases, in which defendants often face weighty sentences and judges routinely sentence defendants below the recommended sentencing range. The data are not consistent with the theory that pretrial detention impedes a defendant’s ability to bargain with their prosecutor over the charges. Several recent papers use a similar empirical approach to estimate the effects of pretrial detention on case outcomes in state courts (Gupta et al., 2016; Leslie and Pope 2017; Dobbie et al., 2018; Stevenson 2018). This recent work reaches mixed conclusions about the effect of pretrial detention on sentencing outcomes3—the primary class of outcomes studied here—and largely focuses on the effect of pretrial detention on the probability of conviction.4 Unlike work set in state courts, this article suggests that pretrial detention does not affect the probability of conviction in federal cases. Nearly all federal defendants are ultimately convicted.5 A body of work in criminology also attempts to estimate the effect of pretrial detention on case outcomes. Most of this literature addresses the risk of endogeneity in the estimating equation by controlling for observable characteristics of the defendant and his case. This research largely reports that pretrial detention is associated with longer sentences (Williams, 2003; Sacks and Ackerman, 2014; Oleson et al., 2016, Oleson et al., 2017). The work is less uniform as to whether pretrial detention influences the binary decision to incarcerate (Williams, 2003; Sacks and Ackerman, 2014; Oleson et al., 2016, Oleson et al., 2017). This research contributes to the literature in several ways. First, this article is able to evaluate the potential mechanisms through which pretrial release affects federal case outcomes by carefully coding the reasons articulated by the sentencing judge when they sentence a defendant outside their recommended Guidelines range. Second, this article examines a setting in which poverty is not directly connected to detention. Third, this article includes a much larger and more geographically and demographically representative sample of defendants than most previous work. The federal criminal system incarcerates more people than any state.6 Federal criminal defendants also receive much longer sentences than state criminal defendants7 and are more frequently detained pretrial. The article proceeds as follows. Section 2 describes the relevant law and practices governing pretrial release in federal court and discusses the role that federal magistrate judges play in federal criminal cases. Section 3 describes the data. Section 4 explains the empirical approach. Section 5 presents results. Section 6 examines the mechanisms responsible for the effects and the robustness of the results. Section 7 concludes. 2. Relevant Background This article uses data on felony defendants sentenced in 71 of the 94 federal district courts in the United States over fiscal years 2002 through 2014. This section summarizes the rules governing pretrial release in federal district courts and describes how federal criminal cases are initially assigned to judicial officers. 2.1. Pretrial Release in U.S. District Courts In federal district courts, the Bail Reform Act of 1984 governs pretrial release.8 For most defendants, the Bail Reform Act directs the presiding judicial officer—usually a federal magistrate judge—to release the defendant under the least restrictive conditions necessary to reasonably assure the appearance of the defendant in court and the safety of any other person and the community.9 If no such conditions exist, then pretrial detention is appropriate.10 Roughly two-thirds of released federal defendants are released at their initial appearance in court (Cohen, 2012). The remaining released defendants are initially detained and released after a detention hearing, which must occur within 5 days of the initial appearance.11 At a detention hearing, the judge considers many flexible factors in deciding whether to release the defendant.12 The majority of released defendants—78%—are released with conditions, which can include travel restrictions, substance abuse treatment requirements, weapons restrictions, or promises to remain employed or seek employment (Cohen, 2012). Unlike pretrial practice in many state courts, federal magistrate judges rarely impose financial conditions for release. If a defendant commits a new crime while released, they are subject to additional criminal penalties. 2.2. Federal Magistrate Judges and Their Responsibilities in Felony Cases Federal magistrate judges are members of federal district courts who serve for renewable 8-year terms. Although their duties vary by district, in felony cases, federal magistrate judges typically handle initial pretrial matters including: conducting the defendant’s first appearance in court and preliminary hearing, appointing counsel, taking the defendant’s initial plea, and deciding whether to release or detain the defendant pending trial (McCabe et al., 2014). After these initial steps, felony criminal matters fall under the authority of federal district judges, who rule on pretrial motions, conduct trials, and sentence defendants.13 Although all federal magistrate judges must follow the Bail Reform Act in deciding whether to detain a defendant, they vary considerably in their propensities to detain. Figure 1 depicts variation in release rates among magistrate judges both between and within federal district courts and courthouses. This variation is not particularly surprising given that the detention decision is discretionary and rarely reviewed. In the data, the mean rate of pretrial release among magistrate judges is 0.361. The most lenient magistrate judge released 70.8% of defendants, while the least lenient magistrate judge released 3.8% of defendants. There is also considerable variation between district courts: the most lenient district court released 59.9% of defendants, while the least lenient district court released 15.4% of defendants. Figure 1. Open in new tabDownload slide Variation in Pretrial Release. Notes: Histograms by magistrate judge (top-left and top-right) include 452 observations; histogram by courthouse (bottom-left) includes 126 observations; histogram by district (bottom-right) includes 71 observations. Figure 1. Open in new tabDownload slide Variation in Pretrial Release. Notes: Histograms by magistrate judge (top-left and top-right) include 452 observations; histogram by courthouse (bottom-left) includes 126 observations; histogram by district (bottom-right) includes 71 observations. It is common for federal district courts to assign felony criminal cases to magistrate judges based on a rotating duty schedule.14 One white paper describes the process as follows: Initial proceedings in criminal cases normally arise on short notice and must be handled promptly, ahead of other duties. For that reason, where two or more Magistrate Judges are located in the same courthouse, they normally rotate the handling of initial proceedings. Each Magistrate Judge will take a turn as the “duty judge” and handle all or virtually all initial proceedings for a set period, such as a week or month. The duty assignment allows the other Magistrate Judges to concentrate without interruption on other court business. (McCabe et al., 2014). As a tangible example, Figure 2 shows the criminal duty schedule used by the Northern District of California in 2017. In that year, the Northern District of California had twelve active magistrate judges, six of whom were assigned to the San Francisco courthouse, two to Oakland, three to San Jose, and one to McKinleyville.15 Figure 2. Open in new tabDownload slide Example Magistrate Judge Criminal Duty Schedule. Notes: U.S. District Court for the Northern District of California, “Magistrate Judge Criminal Duty Schedule,” <https://web.archive.org/web/20170301234946/https://cand.uscourts.gov/dutyschedules> (last accessed August 7, 2019). Judge Nandor J. Vadas was the only magistrate judge assigned to the McKinleyville courthouse, although he also presided over cases in San Francisco. See U.S. District Court for the Northern District of California, “About Magistrate Judge Nandor J. Vadas,” <https://web.archive.org/web/20170401055631/http://www.cand.uscourts.gov/njv> (last accessed August 7, 2019). Figure 2. Open in new tabDownload slide Example Magistrate Judge Criminal Duty Schedule. Notes: U.S. District Court for the Northern District of California, “Magistrate Judge Criminal Duty Schedule,” <https://web.archive.org/web/20170301234946/https://cand.uscourts.gov/dutyschedules> (last accessed August 7, 2019). Judge Nandor J. Vadas was the only magistrate judge assigned to the McKinleyville courthouse, although he also presided over cases in San Francisco. See U.S. District Court for the Northern District of California, “About Magistrate Judge Nandor J. Vadas,” <https://web.archive.org/web/20170401055631/http://www.cand.uscourts.gov/njv> (last accessed August 7, 2019). Following a duty schedule for criminal cases makes sense because magistrate judges handle the preliminary procedural requirements of felony cases before they are assigned to a presiding district judge. These preliminary duties take place in the first few days after a case begins. As described in more detail in Section 4, it is central to the estimation strategy that magistrate judge assignment is essentially random and that the pretrial release decision is the only way that magistrate judges meaningfully affect a defendant’s case. In Section 4, I present both qualitative and quantitative evidence that suggests these requirements are met. 3. Data The empirical analysis uses data from two sources: the United States Sentencing Commission’s annual records of federal defendants sentenced under the Sentencing Reform Act, and the Executive Office for U.S. Attorneys national caseload data. This section describes these data sources in detail and explains how I merged the two data sets. Additional details about the data construction are provided in Appendix A. 3.1. United States Sentencing Commission Data Every fiscal year, the United States Sentencing Commission (the “Commission”) publishes data files containing detailed information about nearly all defendants sentenced in federal courts.16 To compile these offender data files, the Commission culls information from documents submitted by federal district courts. The Commission data include detailed information about individual defendants and their cases, including: characteristics of the defendant; information about the case; the defendant’s Guidelines range; the sentence imposed; and the reasons for an out-of-range sentence, if applicable. For purposes of this article, however, the Commission data have a few important deficiencies. First, the Commission data do not identify—even anonymously—the defendant’s initial magistrate judge or sentencing judge. Second, the Commission data do not identify the courthouse in which the defendant was sentenced—it only identifies the federal district court. This omission presents a serious concern because roughly 80% of federal district courts have more than one courthouse, to which defendants, judges, and court personnel are non-randomly assigned. Ideally, an empirical researcher using Commission data would account for intra-courthouse correlation. Unfortunately, this is not possible with the Commission data. The LIONS data—discussed in the next subsection—address each of these shortcomings. 3.2. LIONS National Caseload Statistical Data The Executive Office for United States Attorneys regularly publishes case data from the 93 United States Attorney Offices (USAOs) located throughout the United States. These data originate from the Legal Information Office Network System (LIONS), which is a computer program that the EOUSA uses to track cases. The LIONS data are comprehensive—it covers all cases in which a USAO is involved. The LIONS data consist of many discrete text files that can be linked using district court, case, and participant identification numbers that are provided with the data. The LIONS data include abundant case-specific information but minimal defendant-specific information—only gender, citizenship, and, in some cases, whether the defendant was a juvenile. The LIONS data also have several additional shortcomings. First, unlike the Commission data, there does not appear to be any evidence that the data are cross-checked with court documents, which raises concerns about human error and incompleteness. Second, the LIONS data contain many more missing values than the Commission data. Third, many of the LIONS variables are coded differently by different USAOs. 3.3. Merged Sample Before attempting to merge the Commission and LIONS data sets, I restrict attention to a subset of federal defendants. First, I only consider defendants who are U.S. citizens. Non-citizen defendants are rarely released pretrial.17 In the event that a non-citizen defendant is released pretrial, they could be subject to an immigration detainer and thus could be held in U.S. Immigration and Customs Enforcement (ICE) custody while their case proceeds. Moreover, non-citizen defendants who are charged with immigration offenses—as most non-citizen defendants are—are often eligible for reduced sentences as part of the Department of Justice’s Fast Track program.18 Thus, pretrial release for citizen defendants is likely to affect case outcomes in a very different way than it does for non-citizen defendants. Second, defendants sentenced in district courts that border Mexico and district courts are excluded,19 although the results are robust to including these districts in the sample. Judges in border districts face uniquely large caseload pressures, primarily driven by immigration prosecutions.20 These caseload pressures might influence pretrial release decisions directly (for example, if pretrial detention facilities are pressed for space) or indirectly [for example, if judges’ behavior in other cases is affected by their immigration workload, a phenomenon documented in Huang (2011)].21 In total, 77.6% of such defendants from the Commission data match in the LIONS data. After matching, the matched sample is restricted in a few additional ways before beginning the empirical analysis. First, I entirely remove four federal district courts whose courthouse identifier variables appear to be unreliable (5.8% of defendants).22 Second, defendants whose pretrial detention status is missing are removed (2.0% of defendants). Third, I remove defendants who have no pretrial judge associated with their case in the data (26.9% of defendants) and those who were sentenced by their pretrial judge (29.2% of defendants).23 I also remove defendants whose initial judge was not a magistrate judge (7.5% of remaining defendants), defendants whose cases were initiated prior to 1999 or after 2013 (1.1% of remaining defendants), magistrate judges with fewer than twenty cases (2.6% of remaining defendants),24 and those who appear in fewer than 8 years (5.9% of remaining defendants).25 Finally, I remove observations containing missing values in any covariates26 (2.5% of remaining defendants), and courthouses with fewer than two magistrate judges (11.7% of remaining defendants). The resulting sample includes 107,246 defendants sentenced in 71 of the 94 federal district courts, comprising 126 unique courthouses and 382 unique magistrate judges. Although magistrate judges are usually assigned to a particular courthouse within a district court, 59 of the 382 unique magistrate judges appear in more than one courthouse. The analysis treats a magistrate judge who appears in two courthouses as two distinct judges. There are 452 unique magistrate judge-courthouse cells in the data, with a mean 241 cases per magistrate judge-courthouse. Table 1 presents summary statistics of the sample (column (1)), and assesses its representativeness relative to all observations matched between Commission and LIONS data (column (2)); Commission observations that did not match with LIONS (column (3)); and the full sample of Commission offenders (column (4)). Comparing columns (2) and (4) demonstrates that the matching process resulted in a matched sample that is very similar in terms of all demographic and case characteristics to the full universe. Comparing columns (1) and (4) shows that additional restrictions did not seem to compromise the representativeness of the sample.27 Defendants in the sample, however, appear to be very slightly more serious offenders than those in the full universe (column (4)) and the matched sample (column (2)). They are released at slightly lower rates pending trial and, on average, have slightly more serious final offense levels and sentencing outcomes. This discrepancy is due to the fact that the sample restrictions described above remove some of the smallest courthouses from the sample and offenders in small courthouses are, on average, less serious offenders. Overall, the data used in the article appear to be very representative of the full universe of U.S. citizen-defendants prosecuted in districts that do not border Mexico. Table 1. Summary Statistics and Represenatativeness of the Sample . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 Notes: Summary statistics of U.S. citizen defendants sentenced in fiscal years 2002 through 2014 in districts that do not border Mexico. The poverty variable is defined as follows: the variable equals zero if the defendant was represented by retained counsel or if their fines were not waived by the court. The variable equals one otherwise. The variable indicating whether the defendant received a government-sponsored sentencing reduction other than a substantial assistance reduction is not available for fiscal year 2002 and only contains 101,567 observations in the sample. The sentence length variable does not include sentences of alternative confinement, such as house arrest. Sentences are capped at 470 months—the Commission’s value for life sentences. The Criminal History, Mean Guidelines Range, Facing Mandatory Minimum, Final Offense Level, Substantial Assistance Reduction, Below Guidelines, Above Guidelines, and In-Range variables all contain fewer than 107,246 observations to missingness in the Commission data (less than 1% of observations for each of these variables). Open in new tab Table 1. Summary Statistics and Represenatativeness of the Sample . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 Notes: Summary statistics of U.S. citizen defendants sentenced in fiscal years 2002 through 2014 in districts that do not border Mexico. The poverty variable is defined as follows: the variable equals zero if the defendant was represented by retained counsel or if their fines were not waived by the court. The variable equals one otherwise. The variable indicating whether the defendant received a government-sponsored sentencing reduction other than a substantial assistance reduction is not available for fiscal year 2002 and only contains 101,567 observations in the sample. The sentence length variable does not include sentences of alternative confinement, such as house arrest. Sentences are capped at 470 months—the Commission’s value for life sentences. The Criminal History, Mean Guidelines Range, Facing Mandatory Minimum, Final Offense Level, Substantial Assistance Reduction, Below Guidelines, Above Guidelines, and In-Range variables all contain fewer than 107,246 observations to missingness in the Commission data (less than 1% of observations for each of these variables). Open in new tab 4. Empirical Strategy and Estimation 4.1. Empirical Specification For defendant |$i$|⁠, I propose a model that relates a case outcome, |$Y_{i}$|⁠, to pretrial release as: $$\begin{equation} Y_i =\beta _0 +\beta _1 PTR_i +\Theta X_i +\varepsilon_i, \end{equation}$$(1) where PTR|$_{i}$| is an indicator variable that equals one if the defendant was released pretrial, and |$X_{i}$| is a vector of observable characteristics of defendant |$i$| and their case. The essential difficulty in estimating the causal effect of pretrial detention on case outcomes is controlling for the non-random assignment of defendants to release. Many of the factors that affect the likelihood that a defendant will be released—such as the nature of the offense with which they are charged—also influence the defendant’s sentence, so it would be careless to assume that correlation between pretrial release and sentence length indicates an underlying causal relationship. Instead, it is very likely that the negative correlation between pretrial release and sentence length will generate an estimate of |$\beta_{1}$| that is biased downward. The empirical strategy uses the propensity of a defendant’s initial pretrial judge to grant pretrial release as an instrumental variable for release. Formally, the instrument for defendant |$i$| is defined as the release rate among all defendants seen by defendant |$i$|’s pretrial judge, |$j$|⁠, in defendant |$i$|’s courthouse, |$c$|⁠, excludingi (also called the judge’s “leave-one-out-mean” release rate) minus the release rate among all defendants in defendant |$i$|’s courthouse, excludingi:28 $$\begin{equation} z_{ijc} =\frac{Released_{jc} - Released_i}{Decided_{jc} -1} -\frac{Released_c - Released_i}{Decided_c -1}, \end{equation}$$(2) where Released|$_{jc}$| is the total number of defendants released by magistrate judge |$j$| in courthouse |$c$|⁠, Released|$_i$| is an indicator variable that equals one if defendant |$i$| was released pending trial and zero otherwise, Decided|$_{jc}$| is the total number of defendants seen by magistrate judge |$j$| in courthouse |$c$|⁠, Released|$_{jc}$| is the total number of defendants released in courthouse |$c$|⁠, and Decided|$_{jc}$| is the total number of defendants who appear in courthouse |$c$|⁠. The instrumental variable is computed at the magistrate judge-courthouse level. Both the first and second stages include fixed effects for the month and year of case initiation and sentencing fiscal year, courthouse fixed effects, and courthouse-specific linear trends to ensure that the instrumental variable captures a magistrate judge’s propensity to release relative to other magistrate judges in her courthouse and in the relevant month and year. To be a valid instrument for pretrial release, |$Z_{ijc}$|⁠: (a) must be correlated with pretrial release, conditional on covariates (the relevance requirement); (b) must not be correlated with the error term, |$\epsilon_i$|⁠, conditional on covariates (the exogeneity requirement, or, the exclusion restriction); and (c) must be monotonic across all defendants (the monotonicity assumption). Subsection 4.2 examines the exclusion restriction. Evidence of the relevance and monotonicity requirements is presented in subsection 5.1. 4.2. The Exclusion Restriction To be a valid instrument, the exclusion restriction requires that the instrumental variable only affect case outcomes by influencing the prospect that a defendant is released. This subsection presents both qualitative and quantitative evidence that the exclusion restriction is satisfied. 4.2.1. Qualitative Evidence of the Exclusion Restriction. As described in subsection 2.2, the institutional context—in which magistrate judges are initially assigned criminal cases on a rotating basis—ought to generate effectively random allotments of defendants to magistrate judges. However, if magistrate judges are able to influence case assignments, perhaps by delaying or transferring cases they do not want to decide, assignment will not be essentially random. This threat seems unlikely to materialize because, as described above, magistrate judges generally handle cases quickly as they come in. To address this concern, however, the instrumental variable, |$Z_{ijc}$|⁠, is computed based on the first magistrate judge before which the defendant appears.29 Even if case assignment is effectively random, the instrumental variable will violate the exclusion restriction if magistrate judge leniency affects case outcomes through channels other than pretrial release. Besides deciding whether to detain a defendant, magistrate judges in criminal cases typically have several other early case responsibilities, including holding the initial appearance, appointing counsel, and conducting a preliminary hearing. Among these initial duties, however, the detention decision seems to be the primary opportunity for a magistrate judge to substantially affect a defendant’s case. For one thing, the decision to appoint counsel is not particularly discretionary.30 If a defendant seeks appointed counsel, he must complete a financial affidavit at his initial appearance and, if his income qualifies, the magistrate judge will appoint an attorney.31 Similarly, the preliminary hearing rarely offers a magistrate judge the chance to affect the defendant’s case because defendants often waive their preliminary hearing, and among preliminary hearings that are not waived, it is extremely rare for a defendant to win.32 4.2.2. Quantitative Evidence of the Exclusion Restriction. One way to quantitatively test whether magistrate judge assignment appears to be random is to examine whether the instrument is correlated with defendant and case characteristics that are predictive of pretrial release. This test is suggestive rather than dispositive because random assignment can generate demographically different caseloads by chance (especially if caseloads are small). Still, it is important to verify that the instrumental variable for defendant |$i$|⁠, |$Z_{ijc}$|⁠, is not correlated with immutable characteristics of defendant |$i$| or their case. Table 2 presents results of regressing pretrial release (columns (1) through (4)) and the instrumental variable (column (5)) on demographic and case characteristics. Columns (1) through (4) demonstrate that, as one might expect, demographic and case characteristics are highly predictive of pretrial release.33 In particular, having no prior criminal history and being female are the personal traits that most strongly predict pretrial release. Controlling for geographic differences with courthouse-level fixed effects helps to clarify the relationship between race and ethnicity and pretrial release, as the coefficient estimates on these variables are heightened when those fixed effects are included in columns (3) and (4). On the other hand, controlling for seasonal and time variation—in the form of fixed effects for the month and year the defendant’s case was initiated and the fiscal year they were sentenced—is less meaningful given the similarity of the estimates in columns (1) and (2), and the estimates in columns (3) and (4) in Table 2. Table 2. Relating Pretrial Release (PTR) and the Instrumental Variable (IV) to Case Characteristics . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P<0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. OLS regressions of whether a defendant was released pretrial (columns (1)–(4)) and the IV (column (5)). Standard errors are reported in parentheses and are clustered at the courthouse level. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab Table 2. Relating Pretrial Release (PTR) and the Instrumental Variable (IV) to Case Characteristics . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P<0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. OLS regressions of whether a defendant was released pretrial (columns (1)–(4)) and the IV (column (5)). Standard errors are reported in parentheses and are clustered at the courthouse level. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab Column (5) demonstrates that, unlike pretrial release itself, the instrumental variable, |$Z_{ijc}$|⁠, is unrelated to demographic and case characteristics. In column (5), the point estimates are very close to zero for all demographic characteristics, and all are insignificant predictors of the instrumental variable. Thus, variation in the instrumental variable likely reflects true variation in judges’ personal propensities to release defendants rather than fixed characteristics of their caseloads. 5. Results 5.1. First Stage and Monotonicity To satisfy the relevance requirement, the instrument variable must be predictive of pretrial release. The first-stage equation estimates this relationship: $$\begin{equation} PTR_{ijc} =\alpha _0 +\alpha _1 z_{ijc} +\alpha _2 X_{ijc} +\gamma _{ic} +\delta _{im} +\mu _{iy} +\rho _{iy} +y\gamma _{ic} +\varepsilon_{ic} \end{equation}$$(3) where PTR|$_{ijc}$| is an indicator variable that equals one if defendant |$i $|in courthouse |$c$| was released pretrial by magistrate judge |$j$|⁠; |$z_{ijc}$| is the instrumental variable defined in equation (2); and |$X_{ijc}$| are control variables, including the defendant’s age, age-squared, and indicator variables for being: Black, Hispanic, female, a high school graduate, impoverished, having prior criminal history, and committing an offense that falls into one of six offense categories.34 The variables |$\gamma $||$_{ic}$|⁠, |$\delta $||$_{im}$|⁠, |$\mu $||$_{iy}$|⁠, |$\nu $||$_{im}$|⁠, and |$\rho $||$_{iy}$| are fixed effects that control for the defendant’s courthouse, month and year of case initiation, and sentencing fiscal year, respectively. The terms y|$\gamma $||$_{ic}$| are courthouse-specific time trends. The results in Table 3 show that the instrumental variable, |$z_{ijc}$|⁠, is strongly predictive of whether a defendant will be released. After controlling for relevant fixed effects and demographic and case characteristics in column (5), a ten percentage point increase in magistrate judge leniency is predictive of a 2.37 percentage point increase in the probability of release.35 The F-statistic for the instrumental variable is 21.94. Table 3. First Stage . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. OLS regression of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. Demographic and case controls include age and age-squared at sentencing, offense type, and indicators for being black, Hispanic, female, a high school but not college graduate, a college graduate, impoverished, and having prior criminal history. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab Table 3. First Stage . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. OLS regression of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. Demographic and case controls include age and age-squared at sentencing, offense type, and indicators for being black, Hispanic, female, a high school but not college graduate, a college graduate, impoverished, and having prior criminal history. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab To be a valid instrumental variable, |$z_{ijc}$| must also be monotonic across all defendants. That is, if Judge A is more lenient on average than Judge B, it must be that Judge A would release all of the defendants released by Judge B (and more). Violations of the monotonicity assumption can produce biased coefficient estimates (Mueller-Smith, 2015). Because the monotonicity assumption is a statement about the counterfactual—how Judge B would treat defendants seen by Judge A—it is not directly testable, but can be indirectly evaluated by examining first-stage regressions of pretrial release on the instrumental variable in subsamples of the data. Monotonicity will be violated if, for example, magistrate leniency varies based on defendant or case characteristics.36 A natural way to investigate monotonicity is to examine whether the first-stage regression of the instrumental variable on pretrial release [Equation (3)] produces coefficient estimates that are positive within all subgroups of defendants. Column (1) in Appendix Table A1 presents such first-stage regression results in subgroups of defendants based on: race, gender, age, education, offense type, criminal history, and poverty.37 In these regressions, the instrumental variable |$z_{ijc}$| is computed using the entire sample, but the first-stage regression is confined to subsamples of defendants. The first-stage estimates in all subgroups are greater than zero and statistically significant, most at the 1% level. Column (2) presents first-stage results in which the instrumental variable is computed based on the reverse sample as described in Bhuller et al. (Forthcoming).. Coefficient estimates in column (2) are all greater than zero and most are highly statistically significant. These results suggest that judges are not selectively lenient, and the monotonicity assumption is likely satisfied. 5.2. Two-Stage Least Squares Results Table 4 presents two-stage least squares (2SLS) estimates of the effect of pretrial release on case and intermediate outcomes. Case outcomes are presented in Panel A, and intermediate outcomes are presented in Panel B of the table. Table 4 also presents ordinary least squares (OLS) estimates for comparison with the 2SLS estimates. Significant effects of pretrial release on intermediate outcomes would support the hypothesis that pretrial release leads to lower sentences by helping defendants substantively contribute to their cases and bargain with prosecutors. On the other hand, evidence that pretrial release affects case outcomes alone would suggest that pretrial release affects judicial (and, perhaps, prosecutorial) behavior at sentencing. Subsection 6.1 presents additional suggestive evidence about the mechanisms at work. Appendix Table A2 presents reduced form results. Table 4. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; |$*P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table 4. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; |$*P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Beginning with case outcomes, pretrial release predicts a 1.124-month decrease in the inverse hyperbolic sine of sentence length,38 which can be interpreted as a 67.5% reduction in sentence length. Pretrial release also increases a defendant’s probability of receiving a below-Guidelines sentence by 56.2 percentage points and reduces the chance that a defendant facing a mandatory minimum will receive (at least) the mandatory minimum by 36.5 percentage points, although this result is only statistically significant at the 10% level. It is important to note that the 95% confidence intervals around the point estimates presented in Table 4 allow for considerably smaller or larger effects of pretrial release. For example, the 95% confidence interval of the effect of pretrial release on the sentence length variable is [|$-$|2.094, |$-$|0.154], which means I cannot reject sentence reductions in the realm of 14–88% as a result of pretrial release. I therefore interpret the results to broadly demonstrate sizeable effects of pretrial detention on sentencing outcomes, with the caveat that the effects are not precisely quantified. Turning to the intermediate outcomes, pretrial release does not statistically significantly affect whether a defendant faced a mandatory minimum. This result functions as a sort-of placebo check because the government usually charges the defendant before release is decided.39 The effects of pretrial release on the defendant’s final offense level are also insignificant in the 2SLS regressions, despite being large and strongly significant when estimated with OLS. The effect of pretrial release on the probability of receiving a government-sponsored below-Guidelines sentence is similarly statistically insignificant in the 2SLS regression, and only significant at the 5% level in the OLS regression, probably because such sentencing reductions are very rare. Overall, I interpret the results in Panel B to be inconsistent with the idea that pretrial release helps a defendant bargain with their prosecutor over their charges. Pretrial release, however, does seem to affect one intermediate outcome: increasing a defendant’s chance of receiving a substantial assistance reduction, although this result is only significant at the 10% level. Pretrial release could improve a defendant’s ability to earn a substantial assistance reduction because it is logistically easier for a released defendant to provide assistance to the government by participating in undercover investigations and providing tangible evidence, such as documents, as opposed to verbal information (Maxfield and Kramer, 1998). Taken together, the results suggest that prosecutorial behavior prior to sentencing bears little responsibility for the effects of pretrial release on case outcomes with one potential exception: the substantial assistance reduction. 6. Extensions and Robustness 6.1. Mechanisms The previous section presented evidence that pretrial release helps defendants at sentencing: significantly reducing sentence length and increasing the chance that a defendant will receive a below-Guidelines sentence and that they will avoid a mandatory minimum sentence if facing one. Pretrial release does not appear to affect a defendant’s ability to earn lesser charges or a prosecutor-supported below-Guidelines sentence other than a substantial assistance reduction. These findings suggest that pretrial release could affect case outcomes through two distinct channels: mitigation (which is reflected at sentencing),40 and cooperation. The results are largely inconsistent with the idea that a released defendant can achieve a lower sentence by more effectively bargaining with the prosecutor over the merits of their case. This subsection presents additional evidence about the mitigation and cooperation channels. The analysis contained in this subsection is premised on an important feature of legal doctrine, which is that there is a subset of federal criminal defendants for whom mitigation evidence is more valuable than others: defendants sentenced (a) under an advisory Guidelines regime who (b) do not face a mandatory minimum sentence. Defendants sentenced under mandatory Guidelines and those facing mandatory minimum sentences typically cannot rely on mitigation evidence to receive a sentence below their recommended Guidelines range. In contrast, all defendants are eligible to earn sentences below their Guidelines range or mandatory minimum if they provide cooperation to the government, regardless of whether the Guidelines are advisory or mandatory. The remainder of this subsection leverages this discrepancy to explore the ways in which pretrial detention affects case outcomes. The Commission codes the reason(s) articulated by the sentencing judge when sentencing a defendant outside of the Guidelines range. In total, the Commission data include nearly 300 unique reason codes, which I broadly categorize as relating to: (a) mitigation, (b) cooperation, (c) bargaining, or (d) avoiding unwarranted disparity. Appendix B provides a complete list of all of the Commission’s reason codes and how they are categorized. There are a few caveats to keep in mind with respect to the coding. First, reason coding is only available for below-Guidelines sentences. It is likely that mitigation, cooperation, bargaining, and avoiding unwarranted disparity also operate in cases that are sentenced within the Guidelines range. For example, if mitigation helps a defendant earn a sentence at the bottom of their Guidelines range, this will not be reported in the data. Similarly, the data suggest that bargaining to a below-Guidelines sentence is exceedingly rare. However, bargaining might be more relevant on the intensive margin (that is, the prosecutor might agree to ask for a sentence at the bottom end of the range as opposed to below the Guidelines) because AUSAs must get supervisory approval to seek out-of-range sentences.41 Second, around 20% of defendants sentenced below the Guidelines range have no reason code provided in the Commission data. Figure 3 shows that below-Guidelines sentencing responds to legal change in precisely the way one would expect. The share of defendants receiving below-Guidelines sentences based on cooperation (dashed black line) is stable over time while the share of defendants receiving below-Guidelines sentences for mitigation reasons (solid purple line) spikes after the Supreme Court’s decision in Booker, and again after the Supreme Court’s decisions in Gall and Kimbrough (which were decided on the same day).42Booker rendered the Guidelines advisory, which greatly expanded the extent to which judges can sentence defendants outside their Guidelines range. Gall and Kimbrough further bolstered out-of-range sentences by holding that appellate courts may not presume that such sentences are unreasonable, and by affirming that sentencing a defendant outside the Guidelines range based on a policy disagreement with the Guidelines is acceptable.43 Figure 3. Open in new tabDownload slide Share of Defendants Receiving a Below-Guidelines Sentence, Over Time. Notes: Monthly share of defendants receiving below-Guidelines sentences based on mitigation, cooperation, disparity, and bargaining. Figure 3. Open in new tabDownload slide Share of Defendants Receiving a Below-Guidelines Sentence, Over Time. Notes: Monthly share of defendants receiving below-Guidelines sentences based on mitigation, cooperation, disparity, and bargaining. To tease out the role that pretrial detention plays in mitigation- and cooperation-based sentencing reductions, I examine whether the effects of pretrial detention are larger for defendants sentenced after Booker (Gall/Kimbrough) and not facing mandatory minimums. Such a finding would be consistent with the idea that mitigation is a channel through which pretrial detention affects case outcomes. In Table 5, I define a defendant as treated by Booker (Gall/Kimbrough) if the defendant was sentenced after Booker (Gall/Kimbrough) and not charged with a mandatory minimum. Table 5. Effect of Pretrial Release on Below-Guidelines Sentences and Their Reasons . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 Notes: ***|$P<0.01$|⁠; **|$P< 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects. Column (1) includes courthouse-specific trends. Open in new tab Table 5. Effect of Pretrial Release on Below-Guidelines Sentences and Their Reasons . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 Notes: ***|$P<0.01$|⁠; **|$P< 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects. Column (1) includes courthouse-specific trends. Open in new tab Panel A demonstrates that the effect of pretrial release on earning a below-Guidelines sentence is larger in the treated groups, but still significant in the untreated groups, which suggests that mitigation is an important—but not singular—mechanism through which pretrial detention affects case outcomes. In Panels B and C, I look for direct evidence that pretrial detention helps defendants earn below-Guidelines sentences based on mitigation (Panel B) and cooperation (Panel C). For mitigation-based reductions, the effect of pretrial release on below-Guidelines sentencing is only significant in the treated groups, as one would expect. In the untreated groups, the estimates of the effect of pretrial release on earning a below-Guidelines sentence based on mitigation are not only statistically insignificant, but also economically small (0.031 and 0.051 for those untreated by Booker and Gall/Kimbrough, respectively). For cooperation-based reductions, on the other hand, Booker and Gall/Kimbrough do not seem to have affected the extent to which pretrial release helps a defendant earn a below-Guidelines sentence to the same extreme. Taken together, the results in Table 5 suggest that both cooperation and mitigation are important channels through which pretrial detention affects case outcomes. The cooperation channel is available to all defendants, while mitigation is significantly more valuable for defendants sentenced under advisory Guidelines who are not facing mandatory minimums, as one would expect. 6.2. Selection into the Sample As described in note 5, roughly 94% of federal felony defendants are ultimately convicted. Among the 6% of defendants who are not convicted, the vast majority have their charges dismissed. Only 0.4% of federal felony defendants are acquitted after trial. Because this work uses Commission data of defendants sentenced in federal district courts, it necessarily excludes defendants who were not ultimately convicted. This subsection examines whether this restriction biases the results of the article, and concludes that it does not. To investigate selection into the sample, I rely on the LIONS data. The LIONS data include defendants whose charges are declined, dismissed, or who are acquitted, making it possible to perform the analysis with LIONS-only data and examine the extent to which restricting the sample to sentenced defendants affects the results. It is important to note, however, that there are several shortcomings to using the LIONS data alone. First, it is much more difficult to determine a defendant’s pretrial detention status using LIONS data than Commission data, which includes pretrial detention status as a variable.44 Second, the LIONS data lack most of the control variables that are used in the main analysis, such as the defendant’s race and ethnicity, level of education, age, criminal history, and poverty status. The only control variables that can be constructed in the LIONS data are gender (with many missing values) and offense type. Third, the LIONS data do not include any of the outcome variables used in this article except the defendant’s sentence. Thus, the LIONS-only analysis serves to provide assurance that the results in the article are not biased by selection, but should be interpreted as less persuasive than the results in the article. To create the LIONS-only subsample, I apply the same restrictions to the data that I use when constructing the sample used in the article, as described in Section 3. The LIONS-only sample contains 44,669 defendants, of whom 41,921 are sentenced.45 Appendix Table A3 presents first-stage results in three samples of the data. Column (1) simply presents the first-stage results using the main data sample (reproduced from Table 3). Column (2) presents first-stage results using the data from the main sample, but restricted to variables that are available in the LIONS-only sample. As described above, this restriction removes many control variables. Column (3) presents first-stage results using the LIONS-only sample. The first-stage coefficients are quite similar in all three columns: 0.24 in the data and specification used in the article, 0.26 in the data used in the article restricted to LIONS-only variables, and 0.30 in the LIONS-only sample. All three first-stage regressions produce highly significant coefficient estimates. Appendix Table A4 presents coefficient estimates of the effect of pretrial release on the defendant’s case ultimately being dismissed, declined, or acquitted (hereafter “dismissal”). The OLS regression of dismissal on pretrial release generates a positive and statistically significant point estimate, suggesting that pretrial release is strongly related to dismissal, although this relationship should not be interpreted as causal. Column (2) shows no significant relationship between dismissal and the instrumental variable. The point estimate is small, and not statistically significant, which suggests that magistrate judge leniency is not strongly related—either economically or statistically—to dismissal. This in turn suggests that selection into the sample is unlikely to bias the results. A 2SLS regression that estimates the effect or pretrial release on dismissal generates a statistically insignificant result, although the point estimate is similar in magnitude to the coefficient estimate in the OLS regression, around 0.06. Finally, Appendix Table A5 displays coefficient estimates of the effect of pretrial release on sentence length. The first two columns present estimates using the full data from the paper but restricts to control variables that are available in the LIONS data. Columns (3) through (6) present results from four regressions in the LIONS-only sample of pretrial release on sentence length: first, an OLS regression; second, an OLS regression with a Heckman correction for selection into the sample; third, an instrumental variables regression; and fourth, an instrumental variables regression with a Heckman correction. The Heckman correction does not meaningfully alter the coefficient estimate in either the OLS or 2SLS regressions. 6.3. Robustness Checks 6.3.1. High Quality Sample. In the preceding analysis, the defendant’s pretrial judge was defined as the first magistrate judge appearing in the defendant’s case in the LIONS data. Constructing the instrumental variable this way has two advantages. First, it often identifies the person who decided whether the defendant was released pretrial—the first magistrate to see a defendant is typically the person to make this decision. Second, it is objective. In some cases, however, the event at which the defendant’s first magistrate judge appears is plainly an event in the case at which it is very unlikely that detention was decided. For example, 424 defendants in the sample have a magistrate judge that appears for the first time at a motion to suppress hearing. In this robustness check, I define the instrumental variable more carefully, restricting attention to defendants whose first magistrate judge is expressly associated with a pretrial event at which detention was likely determined, such as an initial appearance or detention hearing. The results—presented in Table 6—are very similar to the results in the full sample. Table 6. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes in the High-Quality Sample Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Notes: ***|$P <0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table 6. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes in the High-Quality Sample Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Notes: ***|$P <0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab This restriction removes three types of defendants from the sample. First, defendants whose first magistrate judge appeared at an event that likely preceded the detention decision (for example, ordering electronic surveillance) are excluded. Second, defendants whose first event with a magistrate judge likely occurred after a detention decision was made are excluded from the high-quality sample. For example, defendants whose first magistrate judge appeared at a suppression hearing are not part of the high-quality sample. Third, defendants whose first event with a magistrate judge is described in generic terms (for example “court appearance” or “hearing”) are excluded. Appendix A provides details of the construction of the high-quality sample—including a list of all event types that are included. The high-quality sample comprises 110 courthouses and 391 magistrate judges, while the full sample covers 126 courthouses and 452 magistrate judges. The high-quality sample includes 69,796 observations, while the full sample includes 107,246 observations. Regression results for the high-quality sample are quite similar to results using the full sample. In terms of sentencing outcomes, the coefficient estimates on the probability of receiving a below-Guidelines and below-mandatory minimum sentence are more precisely estimated compared to the estimates in the full sample, while the sentence length coefficient is less precisely estimated (⁠|$P$| = 0.068). The point estimates are similar to those in the full sample. Also similar to the results in the full sample, the coefficient estimate of the effect of pretrial release on the probability of receiving a substantial assistance reduction is the only intermediate outcome that is statistically significant, and it is slightly more precisely estimated in the high-quality sample than in the full sample (⁠|$P$| = 0.050), lending more support to the idea that pretrial release can reduce a defendant’s sentence by facilitating assistance to the government. Table 7. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes with Sentencing Judge Controls Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table 7. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes with Sentencing Judge Controls Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab 6.3.2. Sentencing Judge Fixed Effects. In light of evidence that some federal district courts pair magistrate judges and district judges after initial pretrial matters are completed,46Table 7 presents regression results that control for the defendant’s sentencing judge.47 The sample excludes sentencing judges who only appear with one magistrate judge. The sample also restricts attention to sentencing judges who appear in at least eight distinct years of the data and sentence at least 100 defendants. These restrictions reduce the sample size from 107,426 to 75,050. In total, there are 324 unique sentencing judges and 336 unique sentencing judge-courthouse cells. Regression results with sentencing judge fixed effects are presented in Table 7. The estimates are similar to those presented in the main analysis, although less precisely estimated due to the inclusion of many fixed effects. For example, while the coefficient estimates of the effects of pretrial release on sentence length and receiving a below-Guidelines sentence are significant at the 5% level, the coefficient estimate of the effect of pretrial release on receiving at least a mandatory minimum sentence is statistically insignificant (⁠|$P$| = 0.139). However, the coefficient estimate of the effect of pretrial release on the defendant’s probability of earning a substantial assistance reduction is statistically significant at the 5% level, again suggesting that cooperation is an important channel through which pretrial release affects case outcomes. 7. Discussion This article presents evidence that federal pretrial detention appears to significantly increase sentences, decrease the probability that a defendant will receive a below-Guidelines sentence, and decrease the probability that they will avoid a mandatory minimum sentence if facing one. Pretrial detention does not seem to affect a defendant’s ability to avoid a mandatory minimum charge, secure a lower offense level, or earn a prosecutor-supported below-Guidelines sentence other than a substantial assistance reduction. These findings suggest that pretrial detention affects case outcomes through two distinct channels: mitigation (which is reflected at sentencing) and substantial assistance. The results are not consistent with the idea that released defendants can achieve lower sentences by more effectively bargaining with prosecutors over the substance of their cases. It is also important to emphasize that judges might not be singularly responsible for the effects of pretrial detention on sentencing outcomes. Just as pretrial release could allow a defendant to better present a mitigation case to their judge, pretrial release might also influence the behavior of other actors, such as the defendant’s prosecutor. Although this article does not detect an effect of pretrial release on the probability of receiving a government-sponsored departure other than a substantial assistance reduction, it is possible that pretrial release helps a defendant earn a lower recommendation from the prosecutor within his Guidelines range. Unfortunately, the data used in this work do not include the prosecutor’s sentencing recommendation, so I cannot empirically verify whether pretrial release affects prosecutorial behavior on this margin. These findings carry several implications. First, as reported in the first stage, more than 20% of the pretrial release decision stems from judge leniency. This variance is likely to create unwarranted sentencing disparity between defendants based on differences in magistrate judges’ propensities to release defendants. The reduced form, reported in Appendix Table A2, highlights this concern: a ten percentage-point increase in magistrate judge leniency predicts a 2.3% decrease in sentence length. Second, large effects of pretrial release on sentencing outcomes suggest that race- and gender-based disparity in pretrial release could contribute to disparity at the sentencing phase, which researchers have long observed. The ability of released defendants to achieve lower sentences through mitigation also hints that expanding alternatives to incarceration could be a promising policy solution to address the large costs of incarceration in the United States. The finding that pretrial release leads to lower sentences is telling: it suggests that pretrial release produces mitigating information that sentencing judges find credible about defendants’ prospects outside of prison. Moreover, because pretrial release seems to lead to shorter sentences, reducing pretrial detention could be an avenue through which the federal government could reduce incarceration. The federal prison population has near tripled over the last three decades (although it has been slightly shrinking since 2012).48 Much of the popular debate around bail reform focuses on the immediate harms caused by cash bail systems, but this article suggests that the benefits of reform extend beyond the short-term benefits of leaving jail. Several important research questions remain unanswered. Further work might quantify the extent to which differences in detention status contribute to sentencing disparity. The long-term consequences of pretrial detention on outcomes such as recidivism and employment are also under-studied, and further work to disentangle the channels through which pretrial detention affects long-term outcomes would be valuable. A. Appendix A: Data Construction Matching Defendants. The data are matched between the Commission and LIONS data sets using an iterative matching process described here. The matching variables used for the process are: Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed |$^{\rm a}$|This variable is only available for defendants sentenced in fiscal years 2002 and 2003. Open in new tab Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed |$^{\rm a}$|This variable is only available for defendants sentenced in fiscal years 2002 and 2003. Open in new tab I designed the matching process to balance two competing challenges in selecting variables on which to match. First, with too few matching variables, the matching process is likely to generate excess many-to-many matches. In other words, as the number of matching variables decreases, it is increasingly difficult to uniquely identify matches. Second, however, with too many matching variables, the matching process is likely to miss matches because the LIONS data have many missing values and is potentially susceptible to human error. In other words, as the number of matching variables increases, it is increasingly difficult to identify slightly imperfect matches. The iterative process begins by trying to find perfect matches: defendants in the two data sources who uniquely match on every matching variable. After noting these matches and removing them from each data source, the process gradually relaxes the number of variables upon which the defendants must match, in order to find as many matches as possible. Identifying Pretrial Judges. Neither the Commission data nor the LIONS data explicitly identify a defendant’s pretrial judge(s). To determine each defendant’s pretrial judge, I use the event files of the LIONS data. These data files list events associated with a defendant’s case. Critically, events have a type (such as a detention hearing), a judge, and a date. For each defendant, I find the chronologically first event with which a magistrate judge is associated, and label this judge as the defendant’s pretrial judge. As described in the text, defendants who do not have any events in the data, or who do not have any magistrate judge associated with an event in the data, or are ultimately sentenced by the first magistrate to appear in the case, are not included in the analysis. Constructing the High-Quality Sample. As described above, a defendant’s pretrial judge is the first magistrate judge to appear in his case. For each defendant, I also define a “first pretrial event,” which is the type of event at which a magistrate judge first appears. In the full sample of 128,305 observations, there are 284 unique types of first pretrial events. The high-quality sample, with results presented in Table 8, restricts attention to defendants whose first pretrial event is one of the following: AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) Open in new tab AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) Open in new tab The high-quality sample also includes defendants whose first pretrial event is coded as ARGN (defined as initial appearance and/or arraignment) except those in the following districts, for which the code ARGN is defined as arraignment only: Alabama-Northern Colorado District of Columbia Iowa-Northern Indiana-Northern Louisiana-Eastern Louisiana-Middle Louisiana-Western Maryland Michigan-Western North Carolina-Middle North Dakota Ohio-Northern Ohio-Southern Pennsylvania-Middle Pennsylvania-Western South Carolina Texas-Northern Vermont Wyoming Determining Pretrial Detention Status in the LIONS-Only Sample. The main analysis uses the Sentencing Commission’s variable that reports whether the defendant was detained pending trial. The analysis in Section 6.2, however, uses the LIONS data alone. To determine pretrial detention with the LIONS data, I use variables that record the defendant’s status. A defendant is coded as released if their first status is: bond, fugitive, pretrial diversion, released on own recognizance, or unlawful flight to avoid prosecution. A defendant is coded as detained if their first status is custody prior to the detention decision or detained. If a defendant does not have any of these statuses recorded in the data, the pretrial detention variable is missing. B. Appendix B: Reasons for Out-Of-Range Sentences Codes are reproduced from pages A-11 through A-16 of the Commission codebook (United States Sentencing Commission, 2013). M = mitigation D = unwarranted disparity C = cooperation B = bargaining N = none of the above Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Open in new tab Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Open in new tab Table A1. Monotonicity Checks . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. In column (1), the independent variable of interest is the instrumental variable, |$z_{ijc}$|⁠, as defined among all defendants in equation (2). In column (2), the independent variable of interest is judge |$j$|’s release rate among all defendants in the reverse subgroup. Open in new tab Table A1. Monotonicity Checks . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. In column (1), the independent variable of interest is the instrumental variable, |$z_{ijc}$|⁠, as defined among all defendants in equation (2). In column (2), the independent variable of interest is judge |$j$|’s release rate among all defendants in the reverse subgroup. Open in new tab Table A2. Reduced Form: Effects of Magistrate Judge Leniency Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Notes: ***|$P<0.01$|⁠; **|$P<0.05$|⁠; *|$P<0.10$|⁠. OLS regressions of case and intermediate outcomes on the IV, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table A2. Reduced Form: Effects of Magistrate Judge Leniency Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Notes: ***|$P<0.01$|⁠; **|$P<0.05$|⁠; *|$P<0.10$|⁠. OLS regressions of case and intermediate outcomes on the IV, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table A3. First Stage Results in the LIONS-Only Sample . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 Notes: ***|$P< 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. OLS regressions of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. In the LIONS-only data, the only control variables are gender and offense type. In columns (1) and (2), the instrumental variable, |$Z_{ijc}$|⁠, is computed using only sentenced defendants; in column (3) it is computed using all defendants. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A3. First Stage Results in the LIONS-Only Sample . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 Notes: ***|$P< 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. OLS regressions of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. In the LIONS-only data, the only control variables are gender and offense type. In columns (1) and (2), the instrumental variable, |$Z_{ijc}$|⁠, is computed using only sentenced defendants; in column (3) it is computed using all defendants. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A4. Effects of Pretrial Release on Dismissal in the LIONS-Only Sample . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The dependent variable is a binary variable that equals one if all charges were dismissed, declined, or acquitted (or some combination of the three) and zero otherwise. The only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A4. Effects of Pretrial Release on Dismissal in the LIONS-Only Sample . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The dependent variable is a binary variable that equals one if all charges were dismissed, declined, or acquitted (or some combination of the three) and zero otherwise. The only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A5. Effects of Pretrial Release on Sentence Length in the LIONS-Only Sample . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The LIONS-only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. |$^{a}$|The mean of logged sentence in the LIONS-only data is computed among sentenced defendants so as to be comparable to the mean in the data used in this article. Open in new tab Table A5. Effects of Pretrial Release on Sentence Length in the LIONS-Only Sample . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The LIONS-only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. |$^{a}$|The mean of logged sentence in the LIONS-only data is computed among sentenced defendants so as to be comparable to the mean in the data used in this article. Open in new tab Acknowledgments Peter DiCola, Vikas Didwania, Craig Garthwaite, Benjamin Jones, Max Schanzenbach, Matt Schmitt, Alison Siegler, Crystal Yang, the editor, two anonymous referees, and participants in the Kellogg Strategy Department brownbag lunch seminar provided very helpful comments and suggestions. Footnotes 1. " Gerstein v. Pugh, 420 U.S. 103 (1975); County of Riverside v. McLaughlin, 500 U.S. 44 (1991). 2. " As evidence of this point, strong support networks outside custody appear to be more effective at reducing recidivism than in-person visitation while in custody (Lee, 2019). 3. " Among three recent papers to estimate the effect of pretrial detention on sentence length, one finds no significant effect of pretrial detention in the instrumental variables estimate (Dobbie et al., 2018); one finds the opposite (Leslie and Pope 2017); and one finds significant effects among defendants charged with misdemeanors but not felonies (Stevenson 2018). 4. " Heaton et al. (2017) similarly use day-of-the-week of the defendant’s bail hearing as an instrumental variable for release in light of evidence that defendants are more likely to be released near the end of the week. The authors find that for misdemeanor defendants in Harris County, Texas, pretrial detention causes an increase in the probability that a defendant pleads guilty, serves time in jail, and recidivates. The authors also find that pretrial detention causes defendants to receive longer sentences. 5. " In fiscal year 2014, 93.6% of federal felony defendants were ultimately convicted. Just 0.4% of such defendants were acquitted after a trial, while 6.0% had their cases dismissed (Motivans 2017, 17). 6. " At the end of 2015, the federal incarcerated population comprised 196,455 prisoners, while the incarcerated populations of the two largest states—California and Texas—were 129,593 prisoners and 163,909 prisoners, respectively. (Carson and Anderson 2016). 7. " In 2006, the most recent year in which data were available, the average prison sentence for a felony defendant was 64 months in federal court and 38 months in state court. 8. " 18 U.S.C. § 3142. 9. " 18 U.S.C. § 3142(b)–(c). 10. " 18 U.S.C. § 3142(e). 11. " 18 U.S.C. § 3142(f)(2)(B). A defendant may consent to pretrial detention. For example, non-citizen defendants often consent to pretrial detention because pretrial release puts them at risk of being transferred to immigration custody (Graber and Schnitzer, 2013). 12. " These factors include: the nature and circumstances of the offense; the weight of the evidence; the history and characteristics of the person; whether, at the time of the current offense or arrest, the person was on probation, parole, or other release for a criminal offense; and the nature and seriousness of the danger to any person or the community that would be posed by the person’s release. 18 U.S.C. § 3142(g). 13. " Magistrate judges are authorized to handle additional tasks in criminal cases, and the extent to which magistrate judges perform these tasks varies by district court (McCabe et al., 2014). See subsection 4.2. 14. " I consulted local rules to verify the ways in which magistrate judges are assigned to cases in each district. To the extent that information was available, all of the district courts that I found indicated that magistrate judges rotate on a weekly, bi-weekly, or monthly schedule. 15. " Months in which two magistrate judges are listed are those in which each magistrate judge spent time on duty that month. However, no courthouse had more than one magistrate judge on duty at any given time. 16. " The Commission reports data for defendants convicted of felonies and Class A misdemeanors (Reedt, 2013). 17. " In the Commission data, roughly 6% of non-citizen defendants were released in fiscal years 2002–2013, compared to roughly 40% of U.S. citizen-defendants. 18. " In a fast-track prosecution, the prosecutor requests a reduced sentence in exchange for the defendant quickly pleading guilty and relinquishing some of their rights (Cole, 2012). 19. " The districts that border Mexico are: the District of Arizona, the Southern District of California, the District of New Mexico, the Southern District of Texas, and the Western District of Texas. 20. " The five border districts (listed in note 19) account for 34% of defendants in the Commission data. Many of the defendants prosecuted in these districts are non-citizens. The five border districts account for 17% of U.S.-citizen defendants in the Commission data. 21. " Moreover, some news reporting suggests that due to their distinctively large workloads, border districts charge more cases as misdemeanors and rely more heavily on magistrate judges compared to other districts in the sample, which could threaten the exclusion restriction, as described in subsection 4.2. (Lydgate, 2010; Barnitz, 2019). 22. " Data from the Northern District of Alabama, the Northern District of Iowa, the District of Nebraska, and the District of South Carolina list many more courthouses than actually exist in those districts. 23. " Around half of these removals are because the defendant has no judge recorded in the data; and half are because the only judge associated with the defendant’s case was their sentencing judge. 24. " As described in more detail in subsection 4.1, the instrumental variable is computed as the leave-one-out-mean release rate for the defendant’s magistrate judge. For judges with small numbers of observations, this measure will exaggerate variation in leniency. 25. " Because the data include fifteen initial years—1999 through 2013—I restrict to magistrate judges who were in the data for at least half of the period to ensure a relatively stable group of magistrate judges. The results are robust, however, to including defendants whose initial judge was not a magistrate judge, defendants whose cases were initiated outside of 1999 to 2013, and whose magistrate judges appeared in fewer than 8 years and decided at least ten (rather than twenty) cases. 26. " These variables include race, Hispanic ethnicity, sex, age, education level, criminal history, and poverty. 27. " Defendants receiving sentences of no incarceration are responsible for the discrepancy in the share of defendants receiving sentences of any incarceration in the unmatched sample versus the full universe of Commission defendants. The mass of defendants with sentences of zero months creates a many-to-many matching problem that makes these defendants intrinsically difficult to match between the LIONS and Commission data. The match rate was 68% for defendants receiving sentences of no incarceration, compared to around 78% for all defendants. It is also difficult to match defendants who received life sentences which explains why the mean sentence for unmatched defendants is noticeably larger than the mean sentence in the full universe. The match rate was 22% for defendants receiving life sentences (sentences of at least 470 months), compared to around 78% for all defendants. Fortunately, less than 1% of defendants in the Commission data received a life sentence so this problem does little to compromise the representativeness of the sample used in this article, in which the mean sentence is 63.9 months. 28. " By constructing |$Z_{ij} $| to exclude the defendant, the instrument removes the mechanical correlation between the defendant’s outcome and the magistrate judge’s leniency. The resulting estimator is a Jackknife Instrumental Variables estimator (JIVE), as described in Angrist et al. (1999). JIVEs are recommended for models that contain many instruments (Stock et al., 2002). 29. " One might also worry that the exclusion restriction will be threatened if prosecutors strategically time their decisions to file charges based on which magistrate judge is on duty. This threat seems implausible because a defendant must be brought before a magistrate judge soon after arrest. Gerstein, 420 U.S. at 125 (cited in note 1). Moreover, there is no statistically significant relationship between the defendant’s base offense level (a proxy for charging severity) and magistrate judge leniency. 30. " For example, I find that a magistrate judge’s leave-one-out rate of appointing counsel is not strongly nor significantly related to the probability that a given defendant assigned to that magistrate judges actually receives appointed counsel. In contrast—as described in subsection 5.1 and central to the estimation strategy—a magistrate judge’s leniency in granting pretrial release is strongly and significantly related to the probability that a given defendant assigned to that magistrate is released. 31. " In all of the district courts for which I could find information, appointed counsel is assigned randomly. For example, the District of Massachusetts states that “The CJA Assignment Program is a weighted system of random selection and direct assignment in which the program logic continuously self-balances the weights across all panel attorneys.” (United States District Court for the District of Massachusetts, 2015). 32. " Winning a preliminary hearing is roughly equivalent to a grand jury finding of no probable cause, which happens less than 1 in 10,000 times in federal criminal cases (Mueller-Smith, 2015). 33. " Even though the outcome variable here—pretrial release—is binary, I use OLS regressions to easily compare the coefficients to those from regressions of the instrument on the same covariates in column (5). The results are robust to using logistic regression in columns (1)–(4). 34. " The offense categories are: (1) violent crimes; (2) drug offenses; (3) fraud and similar offenses; (4) administration of justice offenses; (5) regulatory offenses; and (6) sex offenses. 35. " The instrumental variable is quite far away from one, consistent with attenuation bias in the instrumental variable due to measurement error. 36. " For example, suppose Judge |$A$| releases 50% of female defendants and just 10% of male defendants, while Judge |$B$| releases 25% of all defendants, regardless of gender. Suppose the judges each see 100 defendants—15 women and 85 men. In this case, Judge |$A$|’s release rate will be 16%, while Judge |$B$|’s release rate will be 25%. Judge |$B$| is more lenient than Judge |$A$|⁠, but there will be female defendants that are released by Judge |$A$| that would not have been released by Judge |$B$|⁠. 37. " Where possible, I divided the data into subgroups that included roughly equal numbers of defendants. 38. " Following Burridge et al. (1988), I use the inverse hyperbolic sine (IHS) transformation rather than the log transformation for the sentence length variable because the log-likelihood function for IHS is well-defined when the dependent variable equals zero. The coefficient estimates are interpreted the same way that coefficients are interpreted when the dependent variable is logged. Alternatively, the results are robust to using ln(sentence + 1) as the dependent variable. 39. " Prosecutors may, however, alter the charges by using superseding indictments—including adding a mandatory minimum charge. For example, Yang (2014) suggests that prosecutors could charge mandatory minimum crimes with superseding indictments after learning the identity of the sentencing judge as a way to avoid below-Guidelines sentences. 40. " I use the term mitigation broadly to refer to personal characteristics of the defendant their offense conduct that might affect the sentencing behavior of the judge (or prosecutor). 41. " The Justice Manual, requires federal prosecutors to obtain supervisory approval before recommending a sentence outside of the sentencing Guidelines range (Justice Manual, 9-27.730(1)), so it is possible that prosecutors prefer to exercise leniency towards defendants on the intensive margin—asking for lower sentences within the Guidelines range, rather than asking for below-Guidelines sentences. 42. " The short-term bump in bargaining reductions (dash-dotted green line) just after Booker likely reflects defendants who were re-sentenced in light of Booker. 43. " In Kimbrough’s case, the sentencing judge disagreed with how the Guidelines treat offenses involving crack versus powder cocaine. 44. " Appendix A explains how I determine pretrial detention status using the LIONS data. 45. " In the LIONS-only sample, 6.6% of defendants have their charges dismissed, declined, or acquitted. This number is extremely close to the 6.4% of felony defendants that are not ultimately convicted, as reported in note 5. 46. " For example, the Eastern District of Michigan assigns magistrate judges to criminal cases randomly in its largest courthouses (United States District Court for the Eastern District of Michigan, 2007), while the Southern District of Florida pairs district judges and magistrate judges to work together (Southern District of Florida 2015). That magistrate and district judges might be paired threatens the exclusion restriction if lenient magistrate judges are more likely to be partnered with lenient district judges. 47. " Similar to magistrate judges, sentencing judges in the analysis are defined at the courthouse level. 48. " In 1987, there were 48,300 federal prisoners; in 2016, there were 189,192 federal prisoners. The federal prison population has been shrinking since 2012. References Angrist, Joshua , Imbens Guido, and Krueger Alan. 1999 . “ Jackknife Instrumental Variables Estimation ,” 14 Journal of Applied Econometrics 57 – 67 . Google Scholar Crossref Search ADS WorldCat Arnold, David , Dobbie Will, and Yang Crystal S.. 2018 . “ Racial Bias in Bail Decisions ,” 133 Quarterly Journal of Economics 1885 – 1932 . Google Scholar Crossref Search ADS WorldCat Ayres, Ian , and Waldfogel Joel. 1994 . “ A Market Test for Race Discrimination in Bail Setting ,” 46 Stanford Law Review 987 – 1047 . Google Scholar Crossref Search ADS WorldCat Barnitz, Katy . 2019 . “ Federal Judges in New Mexico are Stretched Thin with Massive Caseloads .” Albuquerque Journal , March 23 . Available at: https://www.abqjournal.com/1295725/fedear-judges-in-new-mexico-stretched-thin.html (accessed December 1, 2019 ). Bhuller, Manudeep , Dahl Gordon B., Løken Katrine V., and Mogstad Magne. Forthcoming . “ Incarceration, recidivism and employment ,” Journal of Political Economy ; DOI: 10.1086/705330. OpenURL Placeholder Text WorldCat Burbidge, John B. , Magee Lonnie, and Robb A. Leslie. 1988 . “ Alternative Transformations to Handle Extreme Values of the Dependent Variable ,” 83 Journal of the American Statistical Association 123 – 127 . Google Scholar Crossref Search ADS WorldCat Carson, E. Ann , and Anderson Elizabeth. 2016 . “ Prisoners in 2015. ” Burearu of Justice Statistics . Available at: https://www.bjs.gov/content/pub/pdf/p15.pdf. (accessed December 1, 2019 ). Cohen, Thomas H. 2012 . Pretrial Release and Misconduct in Federal District Courts, 2008–2010 . Special Report, U.S. Department of Justice Bureau of Justice Statistics . Cole, James M. 2012 . Department Policy on Early Disposition or “Fast-Track” Programs . Memorandum, U.S. Department of Justice . Dobbie, Will , Goldin Jacob, and Yang Crystal. 2018 . “ The Effects of Pretrial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges ,” 108 American Economic Review 201 – 240 . Google Scholar Crossref Search ADS WorldCat Foote, Caleb . 1954 . “ Compelling Appearance in Court: Administration of Bail in Philadelphia .” 12 University of Pennsylvania Law Review 1031 – 1079 . OpenURL Placeholder Text WorldCat Graber, Lena , and Schnitzer Amy. 2013 . “ The Bail Reform Act and Release from Criminal and Immigration Custody for Federal Criminal Defendants .” National Immigration Project of the National Lawyers Guild . Available at: http://www.nationalimmigrationproject.org/legalresources/practice_advisories/pa_Federal_Bail_Advisory.pdf (accessed October 14, 2015 ) Gupta, Arpit , Hansman Christopher, and Frenchman Ethan. 2016 . “ The Heavy Costs of High Bail: Evidence from Judge Randomization ,” 45 Journal of Legal Studies 471 – 505 . Google Scholar Crossref Search ADS WorldCat Heaton, Paul , Mayson Sandra, and Stevenson Megan. 2017 . “ The Downstream Consequences of Misdemeanor Pretrial Detention ,” 69 Stanford Law Review 711 . OpenURL Placeholder Text WorldCat Huang, Bert I. 2011 . “ Lightened Scrutiny ,” 124 Harvard Law Review 1109 – 1152 . OpenURL Placeholder Text WorldCat Kahneman, Daniel , and Tversky Amos. 1972 . “ Subjective Probability: A Judgment of Representativeness ” 3 Cognitive Psychology 430 – 454 . Google Scholar Crossref Search ADS WorldCat Kahneman, Daniel , and Frederick Shane. 2002 . “Representativeness Revisited: Attribute Substitution in Intuitive Judgment” in Griffin Dale W., Kahneman Daniel, and Gilovich Thomas eds., Heuristics and Biases: The Psychology of Intuitive Judgment . New York : Cambridge University Press , 49 – 81 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Lee, Logan M. 2019 . “ Far from Home and All Alone: The Impact of Prison Visitation on Recidivism ,” 21 American Law and Economics Review 431 – 481 . Google Scholar Crossref Search ADS WorldCat Leslie, Emily , and Pope Nolan G.. 2017 . “ The Unintended Impact of Pretrial Detention on Case Outcomes: Evidence from NYC Arraignments .” 60 Journal of Law and Economics 529 – 557 . Available at: http://home.uchicago.edu/~npope/pretrial_paper.pdf (accessed May 25, 2017 ). Lydgate, Joanna . 2010 . “ Operation Streamline: Border Enforcement That Doesn’t Work .” Los Angeles Times , May 14 . Available at: https://www.latimes.com/archives/la-xpm-2010-may-14-la-oe-lydgate-immigration-20100514-story.html. (accessed December 1, 2019 ). Maxfield, Linda Drazga , and Kramer John H.. 1998 . “ Substantial Assistance: An Empirical Yardstick Gauging Equity in Current Federal Policy and Practice .” January . Available at: http://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-publications/1998/199801_5K_Report.pdf (accessed Feburary 21, 2017 ). McCabe, Peter G. , Gelpi Gustavo A. Hon. Jr., and Newman Michael J. Hon.. 2014 . A Guide to the Federal Magistrate System . White Paper, Federal Bar Association . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Motivans, Mark . 2017 . “ Federal Justice Statistics, 2014 – Statistical Tables .” Bureau of Justice Statistics . March 23 . Available at: https://www.bjs.gov/index.cfm?ty=pbdetail&iid=5873 (accessed November 21, 2017 ). Mueller-Smith, Michael . 2015 . “ The Criminal and Labor Market Impacts of Incarceration .” Available at: https://sites.lsa.umich.edu/mgms/wp-content/uploads/sites/283/2015/09/incar.pdf (accessed Feburary 15, 2018 ). Oleson, J.C. , Lowenkamp Christopher T., Cadigan Timothy P., VanNostrand Marie, and Wooldredge John. 2016 . “ The Effect of Pretrial Detention on Sentencing in Two Federal Districts ,” 33 Justice Quarterly 1103 – 1122 . Google Scholar Crossref Search ADS WorldCat Oleson, James C. , Lowenkamp Christopher T., Wooldredge John, VanNostrand Marie, and Cadigan Timothy P.. 2017 . “ The Sentencing Consequences of Federal Pretrial Supervision ,” 63 Crime & Delinquency 313 – 333 . Google Scholar Crossref Search ADS WorldCat Reedt, Lou . 2013 . Effective Use of Federal Sentencing Data . Presentation to the American Society of Criminology, United States Sentencing Commission . Sacks, Meghan , and Ackerman Alissa R.. 2014 . “ Bail and Sentencing: Does Pretrial Detention Lead to Harsher Punishment? ” 25 Criminal Justice Policy Review 59 – 77 . Google Scholar Crossref Search ADS WorldCat Stevenson, Megan . 2018 . “ Distortion of Justice: How the Inability to Pay Bail Affects Case Outcomes .” 34 Journal of Law, Economics & Organization 511 – 542 . Available at: https://ssrn.com/abstract=2777615 (accessed May 25, 2017 ). Stock, James H. , Wright Jonathan H., and Yogo Motohiro. 2002 . “ A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments .” 20 Journal of Business and Economic Statistics 518 – 529 . Google Scholar Crossref Search ADS WorldCat United States District Court for the Distict of Massachusetts . 2015 . “ CJA Assignment Protocol .” February 26 . Available at: http://www.mad.uscourts.gov/attorneys/pdf/CJAAssignmentProtocol.pdf (accessed February 21, 2017 ). United States District Court for the Eastern District of Michigan . 2007 . “ Re: Plan for the Administration of the Magistrate Judge System .” July 26 . Available at: https://www.mied.uscourts.gov/PDFFIles/07-AO-015.pdf (accessed February 21, 2017 ). United States District Court for the Southern District of Florida . 2015 . “ District Judge-Paired Magistrate Judge Reference Table .” January 12 . Available at: https://www.flsd.uscourts.gov/wp-content/uploads/2015/01/JudgePairingReferenceSheetExternalEffective01122015.pdf (accessed February 20, 2017 ). United States District Court for the Southern District of Florida . 2010 . “ Results of Survey of United States District Judges January 2010 through March 2010 .” June . Available at: http://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-projects-and-surveys/surveys/20100608_Judge_Survey.pdf (accessed May 10, 2016 ). United States District Court for the Southern District of Florida . 2013 . “ Variable Codebook for Individual Offenders .” April 29 . Available at: https://www.ussc.gov/sites/default/files/pdf/research-and-publications/datafiles/Variable_Codebook_for_Individual_Offenders.pdf (accessed February 28, 2019 ). U.S. Department of Justice, Justice Manual . 2018 . https://www.justice.gov/jm/justice-manual (accessed December 9, 2019 ). Williams, Marian R. 2003 . “ The Effect of Pretrial Detention on Imprisonment Decisions ,” 28 Criminal Justice Review 299 – 316 . Google Scholar Crossref Search ADS WorldCat Yang, Crystal S. 2014 . “ Have Interjudge Disparities Increased in an Advisory Guidelines Regime? Evidence from Booker ” 89 New York University Law Review 1268 – 1342 . OpenURL Placeholder Text WorldCat © The Author 2020. Published by Oxford University Press on behalf of the American Law and Economics Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Law and Economics Review Oxford University Press

The Immediate Consequences of Federal Pretrial Detention

Loading next page...
 
/lp/oxford-university-press/the-immediate-consequences-of-federal-pretrial-detention-KzowBhiZmE

References (22)

Publisher
Oxford University Press
Copyright
© The Author 2020. Published by Oxford University Press on behalf of the American Law and Economics Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1465-7252
eISSN
1465-7260
DOI
10.1093/aler/ahz012
Publisher site
See Article on Publisher Site

Abstract

Abstract Unlike the cash-bail regimes that are prevalent in state courts, federal courts rarely use money bail as a condition of pretrial release. Nonetheless, this article presents evidence that pretrial release influences case outcomes for federal defendants. Using case data spanning 71 federal district courts, the article suggests that pretrial release reduces a defendant’s sentence and increases the probability that they will receive a sentence below the recommended sentencing range. Pretrial release also appears to lessen the probability that a defendant will receive a mandatory minimum sentence when one is charged. The analysis exploits variation in magistrate judges’ propensities to release defendants pending trial, which allows magistrate judge leniency to serve as an instrumental variable for pretrial release. The article also provides suggestive evidence that pretrial release affects case outcomes through two channels: first, by giving defendants the opportunity to present mitigating evidence at sentencing and second, by making it easier for defendants to earn a sentencing reduction by providing assistance to the government. 1. Introduction Over the last several years, advocates, policymakers, and scholars have increasingly called for changes to bail regimes for those arrested in the United States. Much of the current effort to reform pretrial detention has focused on the money-bail systems that are ubiquitous in state courts. One concern that animates this debate—and finds support in recent literature—is that cash bail can induce indigent defendants to plead guilty simply to get out of jail and return home. It is plausible, however, that pretrial detention affects case outcomes in other ways and for defendants charged with serious offenses, as well. To test this possibility, this article considers a setting in which pretrial detention is not directly connected to financial means: the federal criminal system. Unlike bail practice in many state court systems, federal defendants are rarely detained based on their inability to pay. Nonetheless, this article finds that pretrial detention significantly affects how federal defendants fare in court, and provides evidence about the mechanisms animating these effects. Near the beginning of a typical criminal case, a judicial officer determines whether a criminal defendant should be held in pretrial detention or released for the duration of their case.1 It is possible—but not certain—that a defendant’s pretrial status could affect the outcome of their case, for several reasons. First, a released defendant is more equipped to participate in their defense, as Caleb Foote observed in early work on the administration of bail in Philadelphia (Foote, 1954). Subsequent qualitative work has identified many channels through which pretrial detention can hinder a defendant’s defense, for example by making it difficult for a defendant to meet with their lawyer, to secure witnesses who can help their case, and to obtain relevant case documents (Williams, 2003). These disadvantages may translate into a decreased likelihood of acquittal or an inferior position in plea bargaining and at sentencing. Second, pretrial release affords a defendant the opportunity to undertake activities that build a strong mitigation case at sentencing, which can be difficult for those who are held in custody while their case is pending (Foote, 1954; Williams, 2003; Sacks and Ackerman, 2014). For example, pretrial release provides a defendant with the chance to demonstrate to their prosecutor and sentencing judge that they can resist criminal activity.2 Pretrial release also offers a defendant the opportunity to undertake mitigating activities such as maintaining a job and supporting dependents (Sacks and Ackerman, 2014). Such mitigation is important to sentencing judges: in a recent survey, the defendant’s post-offense rehabilitative effort was one of the factors about which federal judges reported caring the most in sentencing (United States Sentencing Commission, 2010). Third, defendants might benefit or suffer from representativeness bias based on detention status (Kahneman and Tversky, 1972; Kahneman and Frederick, 2002). In particular, representativeness bias could cause released defendants near the margin of detention to be thought of as less likely to recidivate and less dangerous than defendants near the margin who are detained, simply because released defendants are in general less likely to recidivate and less dangerous. On the other hand, pretrial release could lead to worse outcomes for defendants who violate the terms of release—for example by failing a drug test, committing a crime while released, or failing to appear in court. The idea that a defendant’s pretrial detention status might affect the outcome of their case is troubling for several reasons. For one thing, judges make detention decisions quickly and early on in a case—before the defendant and their attorney have time to fully prepare. For another thing, to the extent that a defendant’s ability to earn release is related to their race (Ayres and Waldfogel, 1994; Arnold et al., 2018), racial disparities in pretrial release could easily translate into unwarranted disparities in sentencing. This article suggests strong effects of pretrial release at the sentencing phase, but not earlier on in a case. Pretrial release appears to decrease sentence length, increase the probability of receiving a below-Guidelines sentence, and decrease the probability of receiving a mandatory minimum sentence, if facing one. Pretrial release also increases the probability that a defendant will receive a sentencing reduction for providing substantial assistance to the government. The data suggest that the effects of detention chiefly work through two channels: by hampering a defendant’s ability to provide mitigating information at sentencing and by making it harder for a defendant to assist the government. The ability to present mitigation evidence at sentencing is particularly important in federal cases, in which defendants often face weighty sentences and judges routinely sentence defendants below the recommended sentencing range. The data are not consistent with the theory that pretrial detention impedes a defendant’s ability to bargain with their prosecutor over the charges. Several recent papers use a similar empirical approach to estimate the effects of pretrial detention on case outcomes in state courts (Gupta et al., 2016; Leslie and Pope 2017; Dobbie et al., 2018; Stevenson 2018). This recent work reaches mixed conclusions about the effect of pretrial detention on sentencing outcomes3—the primary class of outcomes studied here—and largely focuses on the effect of pretrial detention on the probability of conviction.4 Unlike work set in state courts, this article suggests that pretrial detention does not affect the probability of conviction in federal cases. Nearly all federal defendants are ultimately convicted.5 A body of work in criminology also attempts to estimate the effect of pretrial detention on case outcomes. Most of this literature addresses the risk of endogeneity in the estimating equation by controlling for observable characteristics of the defendant and his case. This research largely reports that pretrial detention is associated with longer sentences (Williams, 2003; Sacks and Ackerman, 2014; Oleson et al., 2016, Oleson et al., 2017). The work is less uniform as to whether pretrial detention influences the binary decision to incarcerate (Williams, 2003; Sacks and Ackerman, 2014; Oleson et al., 2016, Oleson et al., 2017). This research contributes to the literature in several ways. First, this article is able to evaluate the potential mechanisms through which pretrial release affects federal case outcomes by carefully coding the reasons articulated by the sentencing judge when they sentence a defendant outside their recommended Guidelines range. Second, this article examines a setting in which poverty is not directly connected to detention. Third, this article includes a much larger and more geographically and demographically representative sample of defendants than most previous work. The federal criminal system incarcerates more people than any state.6 Federal criminal defendants also receive much longer sentences than state criminal defendants7 and are more frequently detained pretrial. The article proceeds as follows. Section 2 describes the relevant law and practices governing pretrial release in federal court and discusses the role that federal magistrate judges play in federal criminal cases. Section 3 describes the data. Section 4 explains the empirical approach. Section 5 presents results. Section 6 examines the mechanisms responsible for the effects and the robustness of the results. Section 7 concludes. 2. Relevant Background This article uses data on felony defendants sentenced in 71 of the 94 federal district courts in the United States over fiscal years 2002 through 2014. This section summarizes the rules governing pretrial release in federal district courts and describes how federal criminal cases are initially assigned to judicial officers. 2.1. Pretrial Release in U.S. District Courts In federal district courts, the Bail Reform Act of 1984 governs pretrial release.8 For most defendants, the Bail Reform Act directs the presiding judicial officer—usually a federal magistrate judge—to release the defendant under the least restrictive conditions necessary to reasonably assure the appearance of the defendant in court and the safety of any other person and the community.9 If no such conditions exist, then pretrial detention is appropriate.10 Roughly two-thirds of released federal defendants are released at their initial appearance in court (Cohen, 2012). The remaining released defendants are initially detained and released after a detention hearing, which must occur within 5 days of the initial appearance.11 At a detention hearing, the judge considers many flexible factors in deciding whether to release the defendant.12 The majority of released defendants—78%—are released with conditions, which can include travel restrictions, substance abuse treatment requirements, weapons restrictions, or promises to remain employed or seek employment (Cohen, 2012). Unlike pretrial practice in many state courts, federal magistrate judges rarely impose financial conditions for release. If a defendant commits a new crime while released, they are subject to additional criminal penalties. 2.2. Federal Magistrate Judges and Their Responsibilities in Felony Cases Federal magistrate judges are members of federal district courts who serve for renewable 8-year terms. Although their duties vary by district, in felony cases, federal magistrate judges typically handle initial pretrial matters including: conducting the defendant’s first appearance in court and preliminary hearing, appointing counsel, taking the defendant’s initial plea, and deciding whether to release or detain the defendant pending trial (McCabe et al., 2014). After these initial steps, felony criminal matters fall under the authority of federal district judges, who rule on pretrial motions, conduct trials, and sentence defendants.13 Although all federal magistrate judges must follow the Bail Reform Act in deciding whether to detain a defendant, they vary considerably in their propensities to detain. Figure 1 depicts variation in release rates among magistrate judges both between and within federal district courts and courthouses. This variation is not particularly surprising given that the detention decision is discretionary and rarely reviewed. In the data, the mean rate of pretrial release among magistrate judges is 0.361. The most lenient magistrate judge released 70.8% of defendants, while the least lenient magistrate judge released 3.8% of defendants. There is also considerable variation between district courts: the most lenient district court released 59.9% of defendants, while the least lenient district court released 15.4% of defendants. Figure 1. Open in new tabDownload slide Variation in Pretrial Release. Notes: Histograms by magistrate judge (top-left and top-right) include 452 observations; histogram by courthouse (bottom-left) includes 126 observations; histogram by district (bottom-right) includes 71 observations. Figure 1. Open in new tabDownload slide Variation in Pretrial Release. Notes: Histograms by magistrate judge (top-left and top-right) include 452 observations; histogram by courthouse (bottom-left) includes 126 observations; histogram by district (bottom-right) includes 71 observations. It is common for federal district courts to assign felony criminal cases to magistrate judges based on a rotating duty schedule.14 One white paper describes the process as follows: Initial proceedings in criminal cases normally arise on short notice and must be handled promptly, ahead of other duties. For that reason, where two or more Magistrate Judges are located in the same courthouse, they normally rotate the handling of initial proceedings. Each Magistrate Judge will take a turn as the “duty judge” and handle all or virtually all initial proceedings for a set period, such as a week or month. The duty assignment allows the other Magistrate Judges to concentrate without interruption on other court business. (McCabe et al., 2014). As a tangible example, Figure 2 shows the criminal duty schedule used by the Northern District of California in 2017. In that year, the Northern District of California had twelve active magistrate judges, six of whom were assigned to the San Francisco courthouse, two to Oakland, three to San Jose, and one to McKinleyville.15 Figure 2. Open in new tabDownload slide Example Magistrate Judge Criminal Duty Schedule. Notes: U.S. District Court for the Northern District of California, “Magistrate Judge Criminal Duty Schedule,” <https://web.archive.org/web/20170301234946/https://cand.uscourts.gov/dutyschedules> (last accessed August 7, 2019). Judge Nandor J. Vadas was the only magistrate judge assigned to the McKinleyville courthouse, although he also presided over cases in San Francisco. See U.S. District Court for the Northern District of California, “About Magistrate Judge Nandor J. Vadas,” <https://web.archive.org/web/20170401055631/http://www.cand.uscourts.gov/njv> (last accessed August 7, 2019). Figure 2. Open in new tabDownload slide Example Magistrate Judge Criminal Duty Schedule. Notes: U.S. District Court for the Northern District of California, “Magistrate Judge Criminal Duty Schedule,” <https://web.archive.org/web/20170301234946/https://cand.uscourts.gov/dutyschedules> (last accessed August 7, 2019). Judge Nandor J. Vadas was the only magistrate judge assigned to the McKinleyville courthouse, although he also presided over cases in San Francisco. See U.S. District Court for the Northern District of California, “About Magistrate Judge Nandor J. Vadas,” <https://web.archive.org/web/20170401055631/http://www.cand.uscourts.gov/njv> (last accessed August 7, 2019). Following a duty schedule for criminal cases makes sense because magistrate judges handle the preliminary procedural requirements of felony cases before they are assigned to a presiding district judge. These preliminary duties take place in the first few days after a case begins. As described in more detail in Section 4, it is central to the estimation strategy that magistrate judge assignment is essentially random and that the pretrial release decision is the only way that magistrate judges meaningfully affect a defendant’s case. In Section 4, I present both qualitative and quantitative evidence that suggests these requirements are met. 3. Data The empirical analysis uses data from two sources: the United States Sentencing Commission’s annual records of federal defendants sentenced under the Sentencing Reform Act, and the Executive Office for U.S. Attorneys national caseload data. This section describes these data sources in detail and explains how I merged the two data sets. Additional details about the data construction are provided in Appendix A. 3.1. United States Sentencing Commission Data Every fiscal year, the United States Sentencing Commission (the “Commission”) publishes data files containing detailed information about nearly all defendants sentenced in federal courts.16 To compile these offender data files, the Commission culls information from documents submitted by federal district courts. The Commission data include detailed information about individual defendants and their cases, including: characteristics of the defendant; information about the case; the defendant’s Guidelines range; the sentence imposed; and the reasons for an out-of-range sentence, if applicable. For purposes of this article, however, the Commission data have a few important deficiencies. First, the Commission data do not identify—even anonymously—the defendant’s initial magistrate judge or sentencing judge. Second, the Commission data do not identify the courthouse in which the defendant was sentenced—it only identifies the federal district court. This omission presents a serious concern because roughly 80% of federal district courts have more than one courthouse, to which defendants, judges, and court personnel are non-randomly assigned. Ideally, an empirical researcher using Commission data would account for intra-courthouse correlation. Unfortunately, this is not possible with the Commission data. The LIONS data—discussed in the next subsection—address each of these shortcomings. 3.2. LIONS National Caseload Statistical Data The Executive Office for United States Attorneys regularly publishes case data from the 93 United States Attorney Offices (USAOs) located throughout the United States. These data originate from the Legal Information Office Network System (LIONS), which is a computer program that the EOUSA uses to track cases. The LIONS data are comprehensive—it covers all cases in which a USAO is involved. The LIONS data consist of many discrete text files that can be linked using district court, case, and participant identification numbers that are provided with the data. The LIONS data include abundant case-specific information but minimal defendant-specific information—only gender, citizenship, and, in some cases, whether the defendant was a juvenile. The LIONS data also have several additional shortcomings. First, unlike the Commission data, there does not appear to be any evidence that the data are cross-checked with court documents, which raises concerns about human error and incompleteness. Second, the LIONS data contain many more missing values than the Commission data. Third, many of the LIONS variables are coded differently by different USAOs. 3.3. Merged Sample Before attempting to merge the Commission and LIONS data sets, I restrict attention to a subset of federal defendants. First, I only consider defendants who are U.S. citizens. Non-citizen defendants are rarely released pretrial.17 In the event that a non-citizen defendant is released pretrial, they could be subject to an immigration detainer and thus could be held in U.S. Immigration and Customs Enforcement (ICE) custody while their case proceeds. Moreover, non-citizen defendants who are charged with immigration offenses—as most non-citizen defendants are—are often eligible for reduced sentences as part of the Department of Justice’s Fast Track program.18 Thus, pretrial release for citizen defendants is likely to affect case outcomes in a very different way than it does for non-citizen defendants. Second, defendants sentenced in district courts that border Mexico and district courts are excluded,19 although the results are robust to including these districts in the sample. Judges in border districts face uniquely large caseload pressures, primarily driven by immigration prosecutions.20 These caseload pressures might influence pretrial release decisions directly (for example, if pretrial detention facilities are pressed for space) or indirectly [for example, if judges’ behavior in other cases is affected by their immigration workload, a phenomenon documented in Huang (2011)].21 In total, 77.6% of such defendants from the Commission data match in the LIONS data. After matching, the matched sample is restricted in a few additional ways before beginning the empirical analysis. First, I entirely remove four federal district courts whose courthouse identifier variables appear to be unreliable (5.8% of defendants).22 Second, defendants whose pretrial detention status is missing are removed (2.0% of defendants). Third, I remove defendants who have no pretrial judge associated with their case in the data (26.9% of defendants) and those who were sentenced by their pretrial judge (29.2% of defendants).23 I also remove defendants whose initial judge was not a magistrate judge (7.5% of remaining defendants), defendants whose cases were initiated prior to 1999 or after 2013 (1.1% of remaining defendants), magistrate judges with fewer than twenty cases (2.6% of remaining defendants),24 and those who appear in fewer than 8 years (5.9% of remaining defendants).25 Finally, I remove observations containing missing values in any covariates26 (2.5% of remaining defendants), and courthouses with fewer than two magistrate judges (11.7% of remaining defendants). The resulting sample includes 107,246 defendants sentenced in 71 of the 94 federal district courts, comprising 126 unique courthouses and 382 unique magistrate judges. Although magistrate judges are usually assigned to a particular courthouse within a district court, 59 of the 382 unique magistrate judges appear in more than one courthouse. The analysis treats a magistrate judge who appears in two courthouses as two distinct judges. There are 452 unique magistrate judge-courthouse cells in the data, with a mean 241 cases per magistrate judge-courthouse. Table 1 presents summary statistics of the sample (column (1)), and assesses its representativeness relative to all observations matched between Commission and LIONS data (column (2)); Commission observations that did not match with LIONS (column (3)); and the full sample of Commission offenders (column (4)). Comparing columns (2) and (4) demonstrates that the matching process resulted in a matched sample that is very similar in terms of all demographic and case characteristics to the full universe. Comparing columns (1) and (4) shows that additional restrictions did not seem to compromise the representativeness of the sample.27 Defendants in the sample, however, appear to be very slightly more serious offenders than those in the full universe (column (4)) and the matched sample (column (2)). They are released at slightly lower rates pending trial and, on average, have slightly more serious final offense levels and sentencing outcomes. This discrepancy is due to the fact that the sample restrictions described above remove some of the smallest courthouses from the sample and offenders in small courthouses are, on average, less serious offenders. Overall, the data used in the article appear to be very representative of the full universe of U.S. citizen-defendants prosecuted in districts that do not border Mexico. Table 1. Summary Statistics and Represenatativeness of the Sample . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 Notes: Summary statistics of U.S. citizen defendants sentenced in fiscal years 2002 through 2014 in districts that do not border Mexico. The poverty variable is defined as follows: the variable equals zero if the defendant was represented by retained counsel or if their fines were not waived by the court. The variable equals one otherwise. The variable indicating whether the defendant received a government-sponsored sentencing reduction other than a substantial assistance reduction is not available for fiscal year 2002 and only contains 101,567 observations in the sample. The sentence length variable does not include sentences of alternative confinement, such as house arrest. Sentences are capped at 470 months—the Commission’s value for life sentences. The Criminal History, Mean Guidelines Range, Facing Mandatory Minimum, Final Offense Level, Substantial Assistance Reduction, Below Guidelines, Above Guidelines, and In-Range variables all contain fewer than 107,246 observations to missingness in the Commission data (less than 1% of observations for each of these variables). Open in new tab Table 1. Summary Statistics and Represenatativeness of the Sample . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 . Sample . All Matched . All Unmatched . All Commission . . (1) . (2) . (3) . (4) . A. Defendant characteristics Black 0.41 0.39 0.41 0.40 Hispanic 0.11 0.10 0.12 0.11 Other race 0.06 0.05 0.05 0.05 White 0.42 0.45 0.41 0.44 Female 0.15 0.16 0.16 0.16 No high school 0.31 0.30 0.31 0.30 High school grad only 0.62 0.61 0.59 0.60 College graduate 0.07 0.08 0.08 0.08 Age (years) 36.3 36.7 36.2 36.6 Criminal history 0.83 0.82 0.80 0.81 Poverty 0.85 0.84 – – B. Intermediate outcomes Released pretrial 0.37 0.41 0.41 0.41 Mean guidelines range (months) 87.1 83.7 94.3 86.0 Facing mandatory minimum 0.35 0.34 0.37 0.34 Final offense level (1–43) 21.1 20.6 20.5 20.6 Substantial assistance reduction 0.19 0.21 0.20 0.20 Other Govt sponsored reduction 0.04 0.04 0.04 0.04 C. Case outcomes Any incarceration 0.87 0.84 0.77 0.83 Sentence (months) 63.9 61.1 70.7 63.2 Sentence (% of guidelines mean) 0.69 0.66 0.62 0.65 Below guidelines 0.48 0.47 0.46 0.47 Above guidelines 0.02 0.02 0.02 0.02 In-range 0.50 0.51 0.51 0.51 Observations 107,246 354,830 102,319 457,149 Notes: Summary statistics of U.S. citizen defendants sentenced in fiscal years 2002 through 2014 in districts that do not border Mexico. The poverty variable is defined as follows: the variable equals zero if the defendant was represented by retained counsel or if their fines were not waived by the court. The variable equals one otherwise. The variable indicating whether the defendant received a government-sponsored sentencing reduction other than a substantial assistance reduction is not available for fiscal year 2002 and only contains 101,567 observations in the sample. The sentence length variable does not include sentences of alternative confinement, such as house arrest. Sentences are capped at 470 months—the Commission’s value for life sentences. The Criminal History, Mean Guidelines Range, Facing Mandatory Minimum, Final Offense Level, Substantial Assistance Reduction, Below Guidelines, Above Guidelines, and In-Range variables all contain fewer than 107,246 observations to missingness in the Commission data (less than 1% of observations for each of these variables). Open in new tab 4. Empirical Strategy and Estimation 4.1. Empirical Specification For defendant |$i$|⁠, I propose a model that relates a case outcome, |$Y_{i}$|⁠, to pretrial release as: $$\begin{equation} Y_i =\beta _0 +\beta _1 PTR_i +\Theta X_i +\varepsilon_i, \end{equation}$$(1) where PTR|$_{i}$| is an indicator variable that equals one if the defendant was released pretrial, and |$X_{i}$| is a vector of observable characteristics of defendant |$i$| and their case. The essential difficulty in estimating the causal effect of pretrial detention on case outcomes is controlling for the non-random assignment of defendants to release. Many of the factors that affect the likelihood that a defendant will be released—such as the nature of the offense with which they are charged—also influence the defendant’s sentence, so it would be careless to assume that correlation between pretrial release and sentence length indicates an underlying causal relationship. Instead, it is very likely that the negative correlation between pretrial release and sentence length will generate an estimate of |$\beta_{1}$| that is biased downward. The empirical strategy uses the propensity of a defendant’s initial pretrial judge to grant pretrial release as an instrumental variable for release. Formally, the instrument for defendant |$i$| is defined as the release rate among all defendants seen by defendant |$i$|’s pretrial judge, |$j$|⁠, in defendant |$i$|’s courthouse, |$c$|⁠, excludingi (also called the judge’s “leave-one-out-mean” release rate) minus the release rate among all defendants in defendant |$i$|’s courthouse, excludingi:28 $$\begin{equation} z_{ijc} =\frac{Released_{jc} - Released_i}{Decided_{jc} -1} -\frac{Released_c - Released_i}{Decided_c -1}, \end{equation}$$(2) where Released|$_{jc}$| is the total number of defendants released by magistrate judge |$j$| in courthouse |$c$|⁠, Released|$_i$| is an indicator variable that equals one if defendant |$i$| was released pending trial and zero otherwise, Decided|$_{jc}$| is the total number of defendants seen by magistrate judge |$j$| in courthouse |$c$|⁠, Released|$_{jc}$| is the total number of defendants released in courthouse |$c$|⁠, and Decided|$_{jc}$| is the total number of defendants who appear in courthouse |$c$|⁠. The instrumental variable is computed at the magistrate judge-courthouse level. Both the first and second stages include fixed effects for the month and year of case initiation and sentencing fiscal year, courthouse fixed effects, and courthouse-specific linear trends to ensure that the instrumental variable captures a magistrate judge’s propensity to release relative to other magistrate judges in her courthouse and in the relevant month and year. To be a valid instrument for pretrial release, |$Z_{ijc}$|⁠: (a) must be correlated with pretrial release, conditional on covariates (the relevance requirement); (b) must not be correlated with the error term, |$\epsilon_i$|⁠, conditional on covariates (the exogeneity requirement, or, the exclusion restriction); and (c) must be monotonic across all defendants (the monotonicity assumption). Subsection 4.2 examines the exclusion restriction. Evidence of the relevance and monotonicity requirements is presented in subsection 5.1. 4.2. The Exclusion Restriction To be a valid instrument, the exclusion restriction requires that the instrumental variable only affect case outcomes by influencing the prospect that a defendant is released. This subsection presents both qualitative and quantitative evidence that the exclusion restriction is satisfied. 4.2.1. Qualitative Evidence of the Exclusion Restriction. As described in subsection 2.2, the institutional context—in which magistrate judges are initially assigned criminal cases on a rotating basis—ought to generate effectively random allotments of defendants to magistrate judges. However, if magistrate judges are able to influence case assignments, perhaps by delaying or transferring cases they do not want to decide, assignment will not be essentially random. This threat seems unlikely to materialize because, as described above, magistrate judges generally handle cases quickly as they come in. To address this concern, however, the instrumental variable, |$Z_{ijc}$|⁠, is computed based on the first magistrate judge before which the defendant appears.29 Even if case assignment is effectively random, the instrumental variable will violate the exclusion restriction if magistrate judge leniency affects case outcomes through channels other than pretrial release. Besides deciding whether to detain a defendant, magistrate judges in criminal cases typically have several other early case responsibilities, including holding the initial appearance, appointing counsel, and conducting a preliminary hearing. Among these initial duties, however, the detention decision seems to be the primary opportunity for a magistrate judge to substantially affect a defendant’s case. For one thing, the decision to appoint counsel is not particularly discretionary.30 If a defendant seeks appointed counsel, he must complete a financial affidavit at his initial appearance and, if his income qualifies, the magistrate judge will appoint an attorney.31 Similarly, the preliminary hearing rarely offers a magistrate judge the chance to affect the defendant’s case because defendants often waive their preliminary hearing, and among preliminary hearings that are not waived, it is extremely rare for a defendant to win.32 4.2.2. Quantitative Evidence of the Exclusion Restriction. One way to quantitatively test whether magistrate judge assignment appears to be random is to examine whether the instrument is correlated with defendant and case characteristics that are predictive of pretrial release. This test is suggestive rather than dispositive because random assignment can generate demographically different caseloads by chance (especially if caseloads are small). Still, it is important to verify that the instrumental variable for defendant |$i$|⁠, |$Z_{ijc}$|⁠, is not correlated with immutable characteristics of defendant |$i$| or their case. Table 2 presents results of regressing pretrial release (columns (1) through (4)) and the instrumental variable (column (5)) on demographic and case characteristics. Columns (1) through (4) demonstrate that, as one might expect, demographic and case characteristics are highly predictive of pretrial release.33 In particular, having no prior criminal history and being female are the personal traits that most strongly predict pretrial release. Controlling for geographic differences with courthouse-level fixed effects helps to clarify the relationship between race and ethnicity and pretrial release, as the coefficient estimates on these variables are heightened when those fixed effects are included in columns (3) and (4). On the other hand, controlling for seasonal and time variation—in the form of fixed effects for the month and year the defendant’s case was initiated and the fiscal year they were sentenced—is less meaningful given the similarity of the estimates in columns (1) and (2), and the estimates in columns (3) and (4) in Table 2. Table 2. Relating Pretrial Release (PTR) and the Instrumental Variable (IV) to Case Characteristics . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P<0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. OLS regressions of whether a defendant was released pretrial (columns (1)–(4)) and the IV (column (5)). Standard errors are reported in parentheses and are clustered at the courthouse level. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab Table 2. Relating Pretrial Release (PTR) and the Instrumental Variable (IV) to Case Characteristics . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 . PTR . PTR . PTR . PTR . IV . Dependent Variable . (1) . (2) . (3) . (4) . (5) . Black -0.044*** -0.046*** -0.073*** -0.073*** -0.0004 (0.009) (0.009) (0.005) (0.005) (0.0005) Hispanic -0.033*** -0.034*** -0.049*** -0.050*** -0.00009 (0.011) (0.011) (0.009) (0.010) (0.0005) Other Race 0.006 0.009 0.027*** 0.026** -0.0003 (0.029) (0.027) (0.010) (0.010) (0.002) Female 0.159*** 0.162*** 0.164*** 0.164*** 0.0005 (0.007) (0.007) (0.007) (0.007) (0.0004) Age (years) -0.003*** -0.003*** -0.003*** -0.004*** 0.00009 (0.0008) (0.0008) (0.0007) (0.0007) (0.00006) Age 0.00006*** 0.00007*** 0.00007*** 0.00007*** -0.000001 (0.00001) (0.00001) (0.000009) (0.000009) (0.0000007) HS only 0.057*** 0.057*** 0.055*** 0.055*** -0.00006 (0.004) (0.004) (0.005) (0.005) (0.0003) College grad 0.135*** 0.128*** 0.116*** 0.116*** 0.00006 (0.008) (0.008) (0.007) (0.007) (0.0005) Criminal history (0/1) -0.221*** -0.215*** -0.210*** -0.210*** -0.0007 (0.008) (0.007) (0.007) (0.007) (0.0004) Poverty (0/1) -0.120*** -0.112*** -0.108*** -0.109*** -0.0006 (0.009) (0.009) (0.007) (0.007) (0.0004) F-stat: offense type 193.66*** 166.13*** 170.55*** 172.95*** 1.67 Month and year fixed effects X X X X Courthouse fixed effects X X X Courthouse trends X X Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P<0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. OLS regressions of whether a defendant was released pretrial (columns (1)–(4)) and the IV (column (5)). Standard errors are reported in parentheses and are clustered at the courthouse level. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab Column (5) demonstrates that, unlike pretrial release itself, the instrumental variable, |$Z_{ijc}$|⁠, is unrelated to demographic and case characteristics. In column (5), the point estimates are very close to zero for all demographic characteristics, and all are insignificant predictors of the instrumental variable. Thus, variation in the instrumental variable likely reflects true variation in judges’ personal propensities to release defendants rather than fixed characteristics of their caseloads. 5. Results 5.1. First Stage and Monotonicity To satisfy the relevance requirement, the instrument variable must be predictive of pretrial release. The first-stage equation estimates this relationship: $$\begin{equation} PTR_{ijc} =\alpha _0 +\alpha _1 z_{ijc} +\alpha _2 X_{ijc} +\gamma _{ic} +\delta _{im} +\mu _{iy} +\rho _{iy} +y\gamma _{ic} +\varepsilon_{ic} \end{equation}$$(3) where PTR|$_{ijc}$| is an indicator variable that equals one if defendant |$i $|in courthouse |$c$| was released pretrial by magistrate judge |$j$|⁠; |$z_{ijc}$| is the instrumental variable defined in equation (2); and |$X_{ijc}$| are control variables, including the defendant’s age, age-squared, and indicator variables for being: Black, Hispanic, female, a high school graduate, impoverished, having prior criminal history, and committing an offense that falls into one of six offense categories.34 The variables |$\gamma $||$_{ic}$|⁠, |$\delta $||$_{im}$|⁠, |$\mu $||$_{iy}$|⁠, |$\nu $||$_{im}$|⁠, and |$\rho $||$_{iy}$| are fixed effects that control for the defendant’s courthouse, month and year of case initiation, and sentencing fiscal year, respectively. The terms y|$\gamma $||$_{ic}$| are courthouse-specific time trends. The results in Table 3 show that the instrumental variable, |$z_{ijc}$|⁠, is strongly predictive of whether a defendant will be released. After controlling for relevant fixed effects and demographic and case characteristics in column (5), a ten percentage point increase in magistrate judge leniency is predictive of a 2.37 percentage point increase in the probability of release.35 The F-statistic for the instrumental variable is 21.94. Table 3. First Stage . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. OLS regression of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. Demographic and case controls include age and age-squared at sentencing, offense type, and indicators for being black, Hispanic, female, a high school but not college graduate, a college graduate, impoverished, and having prior criminal history. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab Table 3. First Stage . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 . (1) . (2) . (3) . (4) . (5) . |$Z_{ijc}$|⁠: Pretrial judge’s |$0.580^{***}$| |$0.479^{***}$| |$0.498^{***}$| |$0.346^{***}$| |$0.237^{***}$| |$\quad$| propensity to release (0.047) (0.050) (0.047) (0.052) (0.051) Month and year FEs X X X X Courthouse FEs X X X Courthouse trends X X Case/demographic FEs X F-statistic 153.28 90.39 111.33 43.79 21.94 Observations 107,246 107,246 107,246 107,246 107,246 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. OLS regression of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. Demographic and case controls include age and age-squared at sentencing, offense type, and indicators for being black, Hispanic, female, a high school but not college graduate, a college graduate, impoverished, and having prior criminal history. Month and year fixed effects include month and year of case initiation and sentencing fiscal year. Courthouse trends are courthouse-specific linear time trends. Open in new tab To be a valid instrumental variable, |$z_{ijc}$| must also be monotonic across all defendants. That is, if Judge A is more lenient on average than Judge B, it must be that Judge A would release all of the defendants released by Judge B (and more). Violations of the monotonicity assumption can produce biased coefficient estimates (Mueller-Smith, 2015). Because the monotonicity assumption is a statement about the counterfactual—how Judge B would treat defendants seen by Judge A—it is not directly testable, but can be indirectly evaluated by examining first-stage regressions of pretrial release on the instrumental variable in subsamples of the data. Monotonicity will be violated if, for example, magistrate leniency varies based on defendant or case characteristics.36 A natural way to investigate monotonicity is to examine whether the first-stage regression of the instrumental variable on pretrial release [Equation (3)] produces coefficient estimates that are positive within all subgroups of defendants. Column (1) in Appendix Table A1 presents such first-stage regression results in subgroups of defendants based on: race, gender, age, education, offense type, criminal history, and poverty.37 In these regressions, the instrumental variable |$z_{ijc}$| is computed using the entire sample, but the first-stage regression is confined to subsamples of defendants. The first-stage estimates in all subgroups are greater than zero and statistically significant, most at the 1% level. Column (2) presents first-stage results in which the instrumental variable is computed based on the reverse sample as described in Bhuller et al. (Forthcoming).. Coefficient estimates in column (2) are all greater than zero and most are highly statistically significant. These results suggest that judges are not selectively lenient, and the monotonicity assumption is likely satisfied. 5.2. Two-Stage Least Squares Results Table 4 presents two-stage least squares (2SLS) estimates of the effect of pretrial release on case and intermediate outcomes. Case outcomes are presented in Panel A, and intermediate outcomes are presented in Panel B of the table. Table 4 also presents ordinary least squares (OLS) estimates for comparison with the 2SLS estimates. Significant effects of pretrial release on intermediate outcomes would support the hypothesis that pretrial release leads to lower sentences by helping defendants substantively contribute to their cases and bargain with prosecutors. On the other hand, evidence that pretrial release affects case outcomes alone would suggest that pretrial release affects judicial (and, perhaps, prosecutorial) behavior at sentencing. Subsection 6.1 presents additional suggestive evidence about the mechanisms at work. Appendix Table A2 presents reduced form results. Table 4. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; |$*P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table 4. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial -1.382*** -1.124** 0.171*** 0.562** -0.233*** -0.365* |$\quad$| release (0.041) (0.495) (0.005) (0.219) (0.008) (0.220) Mean of DV 3.927 0.479 0.661 Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0/1) Govt Sponsored Down Departure (not 5K1.1) (0/1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial -0.128*** 0.047 -3.590*** 3.859 0.050*** 0.327* 0.004** 0.060 |$\quad$| release (0.006) (0.180) (0.121) (3.311) (0.004) (0.190) (0.002) (0.078) Mean of DV 0.350 21.075 0.189 0.043 Observations 107,134 106,353 106,290 101,567 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; |$*P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Beginning with case outcomes, pretrial release predicts a 1.124-month decrease in the inverse hyperbolic sine of sentence length,38 which can be interpreted as a 67.5% reduction in sentence length. Pretrial release also increases a defendant’s probability of receiving a below-Guidelines sentence by 56.2 percentage points and reduces the chance that a defendant facing a mandatory minimum will receive (at least) the mandatory minimum by 36.5 percentage points, although this result is only statistically significant at the 10% level. It is important to note that the 95% confidence intervals around the point estimates presented in Table 4 allow for considerably smaller or larger effects of pretrial release. For example, the 95% confidence interval of the effect of pretrial release on the sentence length variable is [|$-$|2.094, |$-$|0.154], which means I cannot reject sentence reductions in the realm of 14–88% as a result of pretrial release. I therefore interpret the results to broadly demonstrate sizeable effects of pretrial detention on sentencing outcomes, with the caveat that the effects are not precisely quantified. Turning to the intermediate outcomes, pretrial release does not statistically significantly affect whether a defendant faced a mandatory minimum. This result functions as a sort-of placebo check because the government usually charges the defendant before release is decided.39 The effects of pretrial release on the defendant’s final offense level are also insignificant in the 2SLS regressions, despite being large and strongly significant when estimated with OLS. The effect of pretrial release on the probability of receiving a government-sponsored below-Guidelines sentence is similarly statistically insignificant in the 2SLS regression, and only significant at the 5% level in the OLS regression, probably because such sentencing reductions are very rare. Overall, I interpret the results in Panel B to be inconsistent with the idea that pretrial release helps a defendant bargain with their prosecutor over their charges. Pretrial release, however, does seem to affect one intermediate outcome: increasing a defendant’s chance of receiving a substantial assistance reduction, although this result is only significant at the 10% level. Pretrial release could improve a defendant’s ability to earn a substantial assistance reduction because it is logistically easier for a released defendant to provide assistance to the government by participating in undercover investigations and providing tangible evidence, such as documents, as opposed to verbal information (Maxfield and Kramer, 1998). Taken together, the results suggest that prosecutorial behavior prior to sentencing bears little responsibility for the effects of pretrial release on case outcomes with one potential exception: the substantial assistance reduction. 6. Extensions and Robustness 6.1. Mechanisms The previous section presented evidence that pretrial release helps defendants at sentencing: significantly reducing sentence length and increasing the chance that a defendant will receive a below-Guidelines sentence and that they will avoid a mandatory minimum sentence if facing one. Pretrial release does not appear to affect a defendant’s ability to earn lesser charges or a prosecutor-supported below-Guidelines sentence other than a substantial assistance reduction. These findings suggest that pretrial release could affect case outcomes through two distinct channels: mitigation (which is reflected at sentencing),40 and cooperation. The results are largely inconsistent with the idea that a released defendant can achieve a lower sentence by more effectively bargaining with the prosecutor over the merits of their case. This subsection presents additional evidence about the mitigation and cooperation channels. The analysis contained in this subsection is premised on an important feature of legal doctrine, which is that there is a subset of federal criminal defendants for whom mitigation evidence is more valuable than others: defendants sentenced (a) under an advisory Guidelines regime who (b) do not face a mandatory minimum sentence. Defendants sentenced under mandatory Guidelines and those facing mandatory minimum sentences typically cannot rely on mitigation evidence to receive a sentence below their recommended Guidelines range. In contrast, all defendants are eligible to earn sentences below their Guidelines range or mandatory minimum if they provide cooperation to the government, regardless of whether the Guidelines are advisory or mandatory. The remainder of this subsection leverages this discrepancy to explore the ways in which pretrial detention affects case outcomes. The Commission codes the reason(s) articulated by the sentencing judge when sentencing a defendant outside of the Guidelines range. In total, the Commission data include nearly 300 unique reason codes, which I broadly categorize as relating to: (a) mitigation, (b) cooperation, (c) bargaining, or (d) avoiding unwarranted disparity. Appendix B provides a complete list of all of the Commission’s reason codes and how they are categorized. There are a few caveats to keep in mind with respect to the coding. First, reason coding is only available for below-Guidelines sentences. It is likely that mitigation, cooperation, bargaining, and avoiding unwarranted disparity also operate in cases that are sentenced within the Guidelines range. For example, if mitigation helps a defendant earn a sentence at the bottom of their Guidelines range, this will not be reported in the data. Similarly, the data suggest that bargaining to a below-Guidelines sentence is exceedingly rare. However, bargaining might be more relevant on the intensive margin (that is, the prosecutor might agree to ask for a sentence at the bottom end of the range as opposed to below the Guidelines) because AUSAs must get supervisory approval to seek out-of-range sentences.41 Second, around 20% of defendants sentenced below the Guidelines range have no reason code provided in the Commission data. Figure 3 shows that below-Guidelines sentencing responds to legal change in precisely the way one would expect. The share of defendants receiving below-Guidelines sentences based on cooperation (dashed black line) is stable over time while the share of defendants receiving below-Guidelines sentences for mitigation reasons (solid purple line) spikes after the Supreme Court’s decision in Booker, and again after the Supreme Court’s decisions in Gall and Kimbrough (which were decided on the same day).42Booker rendered the Guidelines advisory, which greatly expanded the extent to which judges can sentence defendants outside their Guidelines range. Gall and Kimbrough further bolstered out-of-range sentences by holding that appellate courts may not presume that such sentences are unreasonable, and by affirming that sentencing a defendant outside the Guidelines range based on a policy disagreement with the Guidelines is acceptable.43 Figure 3. Open in new tabDownload slide Share of Defendants Receiving a Below-Guidelines Sentence, Over Time. Notes: Monthly share of defendants receiving below-Guidelines sentences based on mitigation, cooperation, disparity, and bargaining. Figure 3. Open in new tabDownload slide Share of Defendants Receiving a Below-Guidelines Sentence, Over Time. Notes: Monthly share of defendants receiving below-Guidelines sentences based on mitigation, cooperation, disparity, and bargaining. To tease out the role that pretrial detention plays in mitigation- and cooperation-based sentencing reductions, I examine whether the effects of pretrial detention are larger for defendants sentenced after Booker (Gall/Kimbrough) and not facing mandatory minimums. Such a finding would be consistent with the idea that mitigation is a channel through which pretrial detention affects case outcomes. In Table 5, I define a defendant as treated by Booker (Gall/Kimbrough) if the defendant was sentenced after Booker (Gall/Kimbrough) and not charged with a mandatory minimum. Table 5. Effect of Pretrial Release on Below-Guidelines Sentences and Their Reasons . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 Notes: ***|$P<0.01$|⁠; **|$P< 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects. Column (1) includes courthouse-specific trends. Open in new tab Table 5. Effect of Pretrial Release on Below-Guidelines Sentences and Their Reasons . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 . All Defs, All Years . Untreated B . Untreated G/K . Treated B . Treated G/K . . (1) . (2) . (3) . (4) . (5) . Panel A. Outcome: Any Below-Guidelines Sentence (0/1) Pretrial release 0.562** 0.317** 0.309** 0.613** 0.830* (0.219) (0.136) (0.127) (0.256) (0.459) Panel B. Outcome: Below-Guidelines Sentence Based on Mitigation without Cooperation (0/1) Pretrial release 0.235 0.031 0.051 0.403** 0.782** (0.166) (0.122) (0.106) (0.201) (0.371) Panel C. Outcome: Below-Guidelines Sentence Based on Cooperation without Mitigation (0/1) Pretrial Release 0.402** 0.199* 0.165 0.376* 0.700 (0.175) (0.115) (0.102) (0.203) (0.493) Observations 106,373 51,956 68,482 54,410 37,859 Notes: ***|$P<0.01$|⁠; **|$P< 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects. Column (1) includes courthouse-specific trends. Open in new tab Panel A demonstrates that the effect of pretrial release on earning a below-Guidelines sentence is larger in the treated groups, but still significant in the untreated groups, which suggests that mitigation is an important—but not singular—mechanism through which pretrial detention affects case outcomes. In Panels B and C, I look for direct evidence that pretrial detention helps defendants earn below-Guidelines sentences based on mitigation (Panel B) and cooperation (Panel C). For mitigation-based reductions, the effect of pretrial release on below-Guidelines sentencing is only significant in the treated groups, as one would expect. In the untreated groups, the estimates of the effect of pretrial release on earning a below-Guidelines sentence based on mitigation are not only statistically insignificant, but also economically small (0.031 and 0.051 for those untreated by Booker and Gall/Kimbrough, respectively). For cooperation-based reductions, on the other hand, Booker and Gall/Kimbrough do not seem to have affected the extent to which pretrial release helps a defendant earn a below-Guidelines sentence to the same extreme. Taken together, the results in Table 5 suggest that both cooperation and mitigation are important channels through which pretrial detention affects case outcomes. The cooperation channel is available to all defendants, while mitigation is significantly more valuable for defendants sentenced under advisory Guidelines who are not facing mandatory minimums, as one would expect. 6.2. Selection into the Sample As described in note 5, roughly 94% of federal felony defendants are ultimately convicted. Among the 6% of defendants who are not convicted, the vast majority have their charges dismissed. Only 0.4% of federal felony defendants are acquitted after trial. Because this work uses Commission data of defendants sentenced in federal district courts, it necessarily excludes defendants who were not ultimately convicted. This subsection examines whether this restriction biases the results of the article, and concludes that it does not. To investigate selection into the sample, I rely on the LIONS data. The LIONS data include defendants whose charges are declined, dismissed, or who are acquitted, making it possible to perform the analysis with LIONS-only data and examine the extent to which restricting the sample to sentenced defendants affects the results. It is important to note, however, that there are several shortcomings to using the LIONS data alone. First, it is much more difficult to determine a defendant’s pretrial detention status using LIONS data than Commission data, which includes pretrial detention status as a variable.44 Second, the LIONS data lack most of the control variables that are used in the main analysis, such as the defendant’s race and ethnicity, level of education, age, criminal history, and poverty status. The only control variables that can be constructed in the LIONS data are gender (with many missing values) and offense type. Third, the LIONS data do not include any of the outcome variables used in this article except the defendant’s sentence. Thus, the LIONS-only analysis serves to provide assurance that the results in the article are not biased by selection, but should be interpreted as less persuasive than the results in the article. To create the LIONS-only subsample, I apply the same restrictions to the data that I use when constructing the sample used in the article, as described in Section 3. The LIONS-only sample contains 44,669 defendants, of whom 41,921 are sentenced.45 Appendix Table A3 presents first-stage results in three samples of the data. Column (1) simply presents the first-stage results using the main data sample (reproduced from Table 3). Column (2) presents first-stage results using the data from the main sample, but restricted to variables that are available in the LIONS-only sample. As described above, this restriction removes many control variables. Column (3) presents first-stage results using the LIONS-only sample. The first-stage coefficients are quite similar in all three columns: 0.24 in the data and specification used in the article, 0.26 in the data used in the article restricted to LIONS-only variables, and 0.30 in the LIONS-only sample. All three first-stage regressions produce highly significant coefficient estimates. Appendix Table A4 presents coefficient estimates of the effect of pretrial release on the defendant’s case ultimately being dismissed, declined, or acquitted (hereafter “dismissal”). The OLS regression of dismissal on pretrial release generates a positive and statistically significant point estimate, suggesting that pretrial release is strongly related to dismissal, although this relationship should not be interpreted as causal. Column (2) shows no significant relationship between dismissal and the instrumental variable. The point estimate is small, and not statistically significant, which suggests that magistrate judge leniency is not strongly related—either economically or statistically—to dismissal. This in turn suggests that selection into the sample is unlikely to bias the results. A 2SLS regression that estimates the effect or pretrial release on dismissal generates a statistically insignificant result, although the point estimate is similar in magnitude to the coefficient estimate in the OLS regression, around 0.06. Finally, Appendix Table A5 displays coefficient estimates of the effect of pretrial release on sentence length. The first two columns present estimates using the full data from the paper but restricts to control variables that are available in the LIONS data. Columns (3) through (6) present results from four regressions in the LIONS-only sample of pretrial release on sentence length: first, an OLS regression; second, an OLS regression with a Heckman correction for selection into the sample; third, an instrumental variables regression; and fourth, an instrumental variables regression with a Heckman correction. The Heckman correction does not meaningfully alter the coefficient estimate in either the OLS or 2SLS regressions. 6.3. Robustness Checks 6.3.1. High Quality Sample. In the preceding analysis, the defendant’s pretrial judge was defined as the first magistrate judge appearing in the defendant’s case in the LIONS data. Constructing the instrumental variable this way has two advantages. First, it often identifies the person who decided whether the defendant was released pretrial—the first magistrate to see a defendant is typically the person to make this decision. Second, it is objective. In some cases, however, the event at which the defendant’s first magistrate judge appears is plainly an event in the case at which it is very unlikely that detention was decided. For example, 424 defendants in the sample have a magistrate judge that appears for the first time at a motion to suppress hearing. In this robustness check, I define the instrumental variable more carefully, restricting attention to defendants whose first magistrate judge is expressly associated with a pretrial event at which detention was likely determined, such as an initial appearance or detention hearing. The results—presented in Table 6—are very similar to the results in the full sample. Table 6. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes in the High-Quality Sample Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Notes: ***|$P <0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table 6. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes in the High-Quality Sample Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release |$-1.383^{***}$| |$-1.077^{*}$| |$0.175^{***}$| |$0.731^{***}$| |$-0.235^{***}$| |$-0.657^{***}$| (0.050) (0.590) (0.006) (0.260) (0.008) (0.234) Mean of DV 3.934 0.467 0.665 Observations 69,796 69,206 24,199 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release |$-0.125^{***}$| |$-0.066$| |$-3.541^{***}$| |$4.321$| |$0.049^{***}$| |$0.454^{*}$| 0.004|$^*$| 0.053 (0.007) (0.273) (0.140) (4.466) (0.005) (0.232) (0.002) (0.094) Mean of DV 0.348 20.953 0.177 0.040 Observations 69,723 69,195 69,042 65,742 Notes: ***|$P <0.01$|⁠; **|$P < 0.05$|⁠; *|$P<0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab This restriction removes three types of defendants from the sample. First, defendants whose first magistrate judge appeared at an event that likely preceded the detention decision (for example, ordering electronic surveillance) are excluded. Second, defendants whose first event with a magistrate judge likely occurred after a detention decision was made are excluded from the high-quality sample. For example, defendants whose first magistrate judge appeared at a suppression hearing are not part of the high-quality sample. Third, defendants whose first event with a magistrate judge is described in generic terms (for example “court appearance” or “hearing”) are excluded. Appendix A provides details of the construction of the high-quality sample—including a list of all event types that are included. The high-quality sample comprises 110 courthouses and 391 magistrate judges, while the full sample covers 126 courthouses and 452 magistrate judges. The high-quality sample includes 69,796 observations, while the full sample includes 107,246 observations. Regression results for the high-quality sample are quite similar to results using the full sample. In terms of sentencing outcomes, the coefficient estimates on the probability of receiving a below-Guidelines and below-mandatory minimum sentence are more precisely estimated compared to the estimates in the full sample, while the sentence length coefficient is less precisely estimated (⁠|$P$| = 0.068). The point estimates are similar to those in the full sample. Also similar to the results in the full sample, the coefficient estimate of the effect of pretrial release on the probability of receiving a substantial assistance reduction is the only intermediate outcome that is statistically significant, and it is slightly more precisely estimated in the high-quality sample than in the full sample (⁠|$P$| = 0.050), lending more support to the idea that pretrial release can reduce a defendant’s sentence by facilitating assistance to the government. Table 7. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes with Sentencing Judge Controls Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table 7. Effects of Pretrial Release on Case Outcomes and Intermediate Outcomes with Sentencing Judge Controls Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) . . . . OLS . 2SLS . OLS . 2SLS . OLS . 2SLS . . . Pretrial release -1.346*** -0.865** 0.174*** 0.420** -0.232*** -0.327 (0.050) (0.441) (0.006) (0.208) (0.009) (0.215) Mean of DV 3.961 0.480 0.652 Observations 75,050 74,427 26,337 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 Govt Sponsored Down Departure (not 5K1.1) OLS 2SLS OLS 2SLS OLS 2SLS OLS 2SLS Pretrial release -0.127*** 0.049 -3.510*** 2.498 0.053*** 0.325** 0.005* 0.105 (0.007) (0.151) (0.145) (2.499) (0.005) (0.160) (0.002) (0.074) Mean of DV 0.356 21.203 0.199 0.042 Observations 74,978 74,412 74,648 71,311 Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab 6.3.2. Sentencing Judge Fixed Effects. In light of evidence that some federal district courts pair magistrate judges and district judges after initial pretrial matters are completed,46Table 7 presents regression results that control for the defendant’s sentencing judge.47 The sample excludes sentencing judges who only appear with one magistrate judge. The sample also restricts attention to sentencing judges who appear in at least eight distinct years of the data and sentence at least 100 defendants. These restrictions reduce the sample size from 107,426 to 75,050. In total, there are 324 unique sentencing judges and 336 unique sentencing judge-courthouse cells. Regression results with sentencing judge fixed effects are presented in Table 7. The estimates are similar to those presented in the main analysis, although less precisely estimated due to the inclusion of many fixed effects. For example, while the coefficient estimates of the effects of pretrial release on sentence length and receiving a below-Guidelines sentence are significant at the 5% level, the coefficient estimate of the effect of pretrial release on receiving at least a mandatory minimum sentence is statistically insignificant (⁠|$P$| = 0.139). However, the coefficient estimate of the effect of pretrial release on the defendant’s probability of earning a substantial assistance reduction is statistically significant at the 5% level, again suggesting that cooperation is an important channel through which pretrial release affects case outcomes. 7. Discussion This article presents evidence that federal pretrial detention appears to significantly increase sentences, decrease the probability that a defendant will receive a below-Guidelines sentence, and decrease the probability that they will avoid a mandatory minimum sentence if facing one. Pretrial detention does not seem to affect a defendant’s ability to avoid a mandatory minimum charge, secure a lower offense level, or earn a prosecutor-supported below-Guidelines sentence other than a substantial assistance reduction. These findings suggest that pretrial detention affects case outcomes through two distinct channels: mitigation (which is reflected at sentencing) and substantial assistance. The results are not consistent with the idea that released defendants can achieve lower sentences by more effectively bargaining with prosecutors over the substance of their cases. It is also important to emphasize that judges might not be singularly responsible for the effects of pretrial detention on sentencing outcomes. Just as pretrial release could allow a defendant to better present a mitigation case to their judge, pretrial release might also influence the behavior of other actors, such as the defendant’s prosecutor. Although this article does not detect an effect of pretrial release on the probability of receiving a government-sponsored departure other than a substantial assistance reduction, it is possible that pretrial release helps a defendant earn a lower recommendation from the prosecutor within his Guidelines range. Unfortunately, the data used in this work do not include the prosecutor’s sentencing recommendation, so I cannot empirically verify whether pretrial release affects prosecutorial behavior on this margin. These findings carry several implications. First, as reported in the first stage, more than 20% of the pretrial release decision stems from judge leniency. This variance is likely to create unwarranted sentencing disparity between defendants based on differences in magistrate judges’ propensities to release defendants. The reduced form, reported in Appendix Table A2, highlights this concern: a ten percentage-point increase in magistrate judge leniency predicts a 2.3% decrease in sentence length. Second, large effects of pretrial release on sentencing outcomes suggest that race- and gender-based disparity in pretrial release could contribute to disparity at the sentencing phase, which researchers have long observed. The ability of released defendants to achieve lower sentences through mitigation also hints that expanding alternatives to incarceration could be a promising policy solution to address the large costs of incarceration in the United States. The finding that pretrial release leads to lower sentences is telling: it suggests that pretrial release produces mitigating information that sentencing judges find credible about defendants’ prospects outside of prison. Moreover, because pretrial release seems to lead to shorter sentences, reducing pretrial detention could be an avenue through which the federal government could reduce incarceration. The federal prison population has near tripled over the last three decades (although it has been slightly shrinking since 2012).48 Much of the popular debate around bail reform focuses on the immediate harms caused by cash bail systems, but this article suggests that the benefits of reform extend beyond the short-term benefits of leaving jail. Several important research questions remain unanswered. Further work might quantify the extent to which differences in detention status contribute to sentencing disparity. The long-term consequences of pretrial detention on outcomes such as recidivism and employment are also under-studied, and further work to disentangle the channels through which pretrial detention affects long-term outcomes would be valuable. A. Appendix A: Data Construction Matching Defendants. The data are matched between the Commission and LIONS data sets using an iterative matching process described here. The matching variables used for the process are: Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed |$^{\rm a}$|This variable is only available for defendants sentenced in fiscal years 2002 and 2003. Open in new tab Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed Federal judicial district Length of probation imposed Sentencing year Gender Sentencing month Offense type Sentencing day, where available|$^{\rm a}$| Substance type, where available Months of incarceration imposed Lead charge title in the U.S. Code Days of incarceration imposed Lead charge section in the U.S. Code Length of supervised release imposed |$^{\rm a}$|This variable is only available for defendants sentenced in fiscal years 2002 and 2003. Open in new tab I designed the matching process to balance two competing challenges in selecting variables on which to match. First, with too few matching variables, the matching process is likely to generate excess many-to-many matches. In other words, as the number of matching variables decreases, it is increasingly difficult to uniquely identify matches. Second, however, with too many matching variables, the matching process is likely to miss matches because the LIONS data have many missing values and is potentially susceptible to human error. In other words, as the number of matching variables increases, it is increasingly difficult to identify slightly imperfect matches. The iterative process begins by trying to find perfect matches: defendants in the two data sources who uniquely match on every matching variable. After noting these matches and removing them from each data source, the process gradually relaxes the number of variables upon which the defendants must match, in order to find as many matches as possible. Identifying Pretrial Judges. Neither the Commission data nor the LIONS data explicitly identify a defendant’s pretrial judge(s). To determine each defendant’s pretrial judge, I use the event files of the LIONS data. These data files list events associated with a defendant’s case. Critically, events have a type (such as a detention hearing), a judge, and a date. For each defendant, I find the chronologically first event with which a magistrate judge is associated, and label this judge as the defendant’s pretrial judge. As described in the text, defendants who do not have any events in the data, or who do not have any magistrate judge associated with an event in the data, or are ultimately sentenced by the first magistrate to appear in the case, are not included in the analysis. Constructing the High-Quality Sample. As described above, a defendant’s pretrial judge is the first magistrate judge to appear in his case. For each defendant, I also define a “first pretrial event,” which is the type of event at which a magistrate judge first appears. In the full sample of 128,305 observations, there are 284 unique types of first pretrial events. The high-quality sample, with results presented in Table 8, restricts attention to defendants whose first pretrial event is one of the following: AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) Open in new tab AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) AEAP (initial appearance) INDS (indictment-sealed) APPT (order for appointment of counsel) INDT (indictment) ARRU (arrest) INDW (waiver of indictment) ARST (defendant arrested) INFI (information filed) AWAR (arrest warrant issued) INFO (information filed) AWIS (arrest warrant issued) INIT (initial appearance) BEWR (bench warrant issued) INPP (initial appearance) BNCH (bench warrant issued) NINE (first appearance after indictment) BOHR (bond hearing) PCHR (probable cause hearing) BOIF (bill of information filed) PRDT (preliminary/detention hearing) BOND (bond hearing) PREX (preliminary examination) BSET (bail set) PRLE (preliminary examination) CMPL (complaint) PRLM (preliminary hearing) CMSD (complaint served on defendant) RUAR (Rule 5A hearing and arraignment) COFI (complaint filed) RULE (Rule 5, i.e. initial appearance) CSID (counsel identification) SECR (secret indictment/information) DEHE (detention hearing) SIDF (sealed information filed) DTHR (detention hearing) SIIF (sealed information filed) DTMT (detention motion filed) UNSL (unsealed indictment/information) FAPP (first appointment with counsel) VFIR (first appearance) IADT (initial appearance-detention hearing) VPRE (preliminary hearing) IAPP (initial appearance) WARR (warrant issued) INDC (indictment) WARS (secret warrant) Open in new tab The high-quality sample also includes defendants whose first pretrial event is coded as ARGN (defined as initial appearance and/or arraignment) except those in the following districts, for which the code ARGN is defined as arraignment only: Alabama-Northern Colorado District of Columbia Iowa-Northern Indiana-Northern Louisiana-Eastern Louisiana-Middle Louisiana-Western Maryland Michigan-Western North Carolina-Middle North Dakota Ohio-Northern Ohio-Southern Pennsylvania-Middle Pennsylvania-Western South Carolina Texas-Northern Vermont Wyoming Determining Pretrial Detention Status in the LIONS-Only Sample. The main analysis uses the Sentencing Commission’s variable that reports whether the defendant was detained pending trial. The analysis in Section 6.2, however, uses the LIONS data alone. To determine pretrial detention with the LIONS data, I use variables that record the defendant’s status. A defendant is coded as released if their first status is: bond, fugitive, pretrial diversion, released on own recognizance, or unlawful flight to avoid prosecution. A defendant is coded as detained if their first status is custody prior to the detention decision or detained. If a defendant does not have any of these statuses recorded in the data, the pretrial detention variable is missing. B. Appendix B: Reasons for Out-Of-Range Sentences Codes are reproduced from pages A-11 through A-16 of the Commission codebook (United States Sentencing Commission, 2013). M = mitigation D = unwarranted disparity C = cooperation B = bargaining N = none of the above Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Open in new tab Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Code . Description . Category . 0 No reason given N 1 Nature and circs of offense and history and char. of def (18 3553(a)(1)) M 2 Reflect seriousness of offense, promote respect for law, provide just punishment for the offense (18 3553(a)(2)(A)) M 3 Afford adequate deterrence to criminal conduct (18 3553(a)92)(B)) M 4 Protect the public from further crimes of the defendant (18 3553(a)(2)(C)) M 5 Provide the def with needed educational or vocational training, medical care or other correctional treatment in most effective manner (18 3553(a)(2)(D)) M 6 Avoid unwarranted sentencing disparities among defendants D 7 Provide Restitution to any victims of the offense N 10 (5G1.3) Convictions on related counts N 11 (5H1.1) Age (Defendant’s youth or old age) M 12 (5H1.2) Educational and vocational skills M 13 (5H1.3) Mental and emotional conditions M 14 (5H1.4) Physical condition M 15 (5H1.4) Drug dependence and alcohol abuse M 16 (5H1.5) Previous employment record M 17 (5H1.6) Family ties and responsibilities M 18 (5H1.6) Community ties M 19 (5K1.1) Substantial assistance WITH govt. motion C 20 (not 5K1.1) Cooperation without motion C 21 (5K1.1) Cooperation (motion unknown) C 22 Several persons injured M 23 (5K2.1) Death M 24 (5K2.2) Physical injury M 25 (5K2.3) Extreme psychological injury M 26 (5K2.4) Abduction or unlawful restraint M 27 (5K2.5) Property damage or loss M 28 (5K2.6) Weapons and dangerous instrumentalities M 29 (5K2.7) Disruption of governmental function M 30 (5K2.8) Extreme conduct M 31 (5K2.9) Criminal purpose M 32 (5K2.10) Victim’s conduct M 33 (5K2.11) Lesser harm M 34 (5K2.12) Coercion and duress M 35 (5K2.13) Diminished capacity M 36 (5K2.14) Public welfare M 37 Fast Track M 38 Totality of Circumstances/Combination of Factors (Cook/Koon) N 39 Circumstances not Considered by the Guidelines N 40 EDP Disparity D 41 (4A1.3) Criminal History Adequacy Explain M 42 (5K2.17) High-Capacity Semiautomatic Firearm M 43 (5K2.18) Violent Street Gang M 44 (5K2.20) Aberrant Behavior M 45 (5K2.21) Dismissed and Uncharged Conduct M 46 (5K2.22) Age or Health of Sex Offenders M 47 (5K2.23) Discharge Terms of Imprisonment M 48 (5K3.1) Early Disposition. Fast Track program M 49 (5K2.24) Commit offense while wearing/displaying unauth/counterfeit uniform M 51 Charge Bargaining B 52 Unspecified government motion (Not 5K1.1 or 5K3.1) B 53 Defendant Unaware of Type/Quantity/Purity of Drugs M 54 Mitigating Factors Regarding Illegal Reentry M 55 Victim of Abuse (Not Childhood Abuse) M 56 (2D2.1) Application note 1 (Intended Consumption by Person Other than def) M 57 No Benefit of Incarceration to Def M 58 Charitable Service/Good Works M 59 Mens Rea M 60 The nature and circumstance of the offense pursuant to 18 U.S.C. 3553(a)(1) M 61 The history and characteristics of the defendant pursuant to 18 U.S.C. 3553(a)(1) M 62 To provide the defendant with needed educational or vocational training (18 U.S.C. 3553(a)(2)(D)) M 63 To provide the defendant with medical care (18 U.S.C. 3553(a)(2)(D)) M 64 To provide the defendant with other correctional treatment in the most effective manner (18 U.S.C. 3553(a)(2)(D)) M 65 Policy Disagreement with the Guidelines (Kimbrough v. U.S. 552 U.S. 85 (2007) D 66 Victim Impact M 67 Pre-Sentence Rehabilitation M 100 (2A1.1) Death not caused intentionally M 101 (2A2.1, 2A2.2) Degree of injury falls between two categories M 103 (2A3.2) Criminal sexual act for commercial purpose M 105 (2A6.1) Factors not incorporated in guideline M 106 (2B1.3) Monetary value does not reflect extent of harm M 108 (2B3.1) intended to murder the victim M 110 (2C1.1) Systematic or pervasive corruption of governmental function M 111 (2D1.1) Unusually high drug purity M 112 (2D1.1) Unusually high drug amount M 113 Drug Quantity M 114 (2D1.5) Sanctioned used of violence in enterprise M 115 Crack/Powder disparity D 118 (2F1.1) Loss substantially exceeds maximum from loss table M 119 (2F1.1) Dollar loss overstates seriousness of offense M 120 (2F1.1) Fraudulent statements could be covered by more specific statute M 121 (2G1.1) Offense did not involve profit nor physical force or coercion M 124 2H2.1 Corruption a public official/bodily injury/property damage M 125 (2L1.1) Large number of aliens M 126 (2L1.1) Dangerous or inhumane treatment M 127 (2L2.2, 2L2.4) Deported on one or more prior occasions M 128 (2M3.1, 2M3.7, 2M3.9) Revelation causes little or no harm M 131 (2M5.1, 2M5.2) Extreme threat to national security M 132 (2M5.1, 2M5.2) Extreme volume of commerce involved M 134 (2M5.1, 2M5.2) Extreme number of multiple occurrences M 137 (2N2.1, 2Q1.2) Negligence involved M 139 (2Q1.2, 2Q1.3) Harm resulting from emission, release or discharge M 140 (2Q1.2, 2Q1.3) Harm resulting from quantity & nature of substance or pollutant M 141 (2Q1.2, 2Q1.3) Harm resulting from duration of offense M 142 (2Q1.2, 2Q1.3) Harm resulting from risk M 143 (2Q1.2, 2Q1.3) Nature of risk M 144 (2Q1.2, 2Q1.3) Number of people at risk M 145 (2Q1.2, 2Q1.3) Nature of contamination M 146 (2Q1.2, 2Q1.3) Similar conduct by civil adjudication M 147 (2Q1.2, 2Q1.3) Failure to comply with administrative order M 148 (2T1.6) Tax evasion and embezzlement M 151 (2T3.1, 2) Import. of drugs, obscene matter, firearms, pelts of endangered species M 152 (2T3.1) Duties may not adequately reflect harm M 200 (3A1.2) Exceptionally high level victim M 201 (3C1.1) Further obstruction of justice M 202 5C1.2 (Safety Valve) N 250 (3D1.4) Offenses significantly more than 5 units M 300 (4A1.3) Pattern of conduct M 301 (4A1.3) Pending cases M 302 (4A1.3) Related cases M 304 Gen. adequacy of crim. history; does not reflect seriousness of crim. hist M 305 (4A1.3) Outside applicable time period, but similar misconduct M 306 (4A1.3) outside applicable time period, but substantial portion of income M 307 (4A1.3) Significance or similarity of past conduct M 308 Other adequacy of criminal history M 309 Criminal history category over-represents the defendant’s involvement M 310 Prior criminal record; risk of future criminal conduct M 311 Age of Priors M 312 First Federal Conviction M 313 Little/No Incarceration for Priors M 314 Age when prior was committed M 400 Unconvicted behavior added under relevant conduct M 401 (4B1.3) Criminal livelihood M 402 (4B1.1) Career offender M 500 Pursuant to a NONbinding (or unknown type) plea agreement B 501 Due to stipulations B 502 Other plea agreement reasons B 503 Pursuant to a binding plea 11(c)(1)(C) B 504 Statmin/Max N 648 Alien Status M 649 Cultural Assimilation M 650 Mule/Role in the offense M 651 Interpolation M 652 Deportation M 653 Local conditions M 654 Adequate punishment to meet the purposes of sentencing M 655 Deterrence M 656 Prey to other inmates/Susceptibility to abuse in prison M 657 Reduce disparity D 658 Charge/Plea does not reflect the seriousness of the offense M 659 Guidelines do not reflect the seriousness of the offense M 660 Justifiable Reasons M 661 No prior record/first offender M 662 Put sentence in line with codefendant’s (i.e. Reduce disparity among codefendants) D 663 High speed chase M 664 Lack of culpability/accountability of defendant M 665 Lost job/punishment enough M 666 Lack of available facilities/overcrowding N 667 Non-Guideline Sentence N 668 Nonviolent Offense/Offender M 669 Indigent background M 670 Impact on Employment of Self/Others M 671 Time or cost involved in the investigation M 672 Defendant’s propensity for violence M 673 Acceptance of responsibility M 674 Limited/Minor prior record M 675 Training/Treatment Opportunities M 676 Rehabilitation M 677 Restitution M 678 Incapacitation M 679 Sufficient punishment M 680 Nature/seriousness of the offense M 681 First felony conviction M 683 On-going nature of activity M 684 Terminated Offense Prior to Arrest/Discovery M 685 Low Likelihood of Recidivism M 686 Conduct on Release/Bond/Supervision M 687 Limited/No Loss or Monetary Benefit M 688 Dollar amount involved in crime (general) M 689 Minors involved M 690 Untruthful testimony M 691 Terrifying the victim M 692 Influenced/Used by Others M 693 Cost of Incarceration N 695 Currently receiving punishment under state or federal jurisdiction N 696 Mitigating Factors regarding Firearms M 697 Defendant’s positive background/good character M 698 Defendant’s negative background/bad character M 700 Not representative of the ‘heartland’ M 701 Guidelines too high/offense level over-represented M 702 Guidelines too law/offense level under-represented M 703 (5K2.0) General aggravating or mitigating circumstance M 704 (4A1.3) Seriousness of criminal history substantially underrepresented by CHC M 705 (4A1.3) Likelihood of recidivism underrepresented by CHC M 706 (2C1.1) Payment not reflect seriousness of offense M 707 (2L1.1) Knowingly smuggled alien who intended to engage in serious crim. activity M 708 (2B1.1) Offense level understate offense seriousness M 709 (2B1.1) Offense level overstates offense seriousness M 710 (2B1.1) Offense level disrupts critical infrastructure M 711 (5G1.3) N 712 Advisory nature of guidelines M 713 Judge specifics presence of variance/deviation M 719 Cooperation/Attempted Cooperation (not 5K, not with prosecution, per se) C 720 5K1.1 Subst. Assist. WITHOUT govt. motion C 721 5H1.11 Military Record/Charitable Service/Good Works M 722 Charitable Service/Good Works M 730 Early Plea B 731 Global Plea B 740 Savings to government B 741 Waiver of appeal B 742 Waiver of indictment B 743 Global disposition N 744 As stated on record N 745 SOR references an exhibit (eg., “See Exhibit C”) N 746 See Transcript/Rider/PSR N 747 Party Motion/Agreement/Consent (Reason Unspecified) B 750 Reasonableness N 751 Interest of Justice N 760 US v. Booker / US v. Fanfan N 761 18 USC 3553(a) M 762 Language from 18 USC 3553(a) statute text M 763 2007 New Crack Table Amendment D 777–799 Other (SPECIFY) N 801 8C4.2, risk of death or bodily injury M 822 Defendants is a law enforcement officer or ex-law enforcement officer M 823 Offense behavior was an isolated incident M 824 A lower sent. gives defend. a chance to become a productive member of society M 825 (5K2.16) Voluntary Disclosure N 826 Lack of Youthful Guidance M 827 Delay in prosecution; evidentiary concerns N 828 Time Served N 830 Child abuse/childhood abuse/battered child syndrome M 831 Remorse M 832 Abuse of trust/skill/position M 833 2L1.2 Criminal History Issues M 834 Limited Duration of Involvement M 835 Conditions of Confinement M 836 Kimbrough/Gall Supreme Court Decision D 837 Good Time N 838 Number of Images M 839 No Inappropriate Conduct with Children/Not a Pedophile M 840 Defendant did not Distribute Child Pornography M 841 Rule 35b 842 Due to Spears Supreme Court Decision 843 Has had inappropriate contact with or exploited minors M 844 Age of Victim (Only Porn or Sex Offenses) M 845 Def participated in psychosexual assessment M 846 No/Minor/Limited History of Drug/Alcohol Abuse/Defendant is not an Addict M 847 Crime motivated by financial distress M 848 Victim request/letter from victim to judge/PO M 849 Age of Instant Offense M 850 First immigration offense M 851 Number of documents M 853 No History of Mental Health Issues M 854 Lack of Family Support M 855 Forfeiture M 856 No victims/no harm M 857 Letters of support - unspecified from whom M 858 Proposed changes to the 2D1.1 drug quantity table D 859 Loss of Border crossing Card/Privilege or LPR Status M 860 Increase Over Similar Prior’s Sentence M 861 Collateral Consequences of Offense M 862 Conduct in jail/prison/custody M 863 No Deportation/Immigration History M 864 Limited Deportation/Immigration History M 865 Non-Crack Drug Table Disparity M 866 Gambling Addiction M 867 Traumatic/Recent Loss of Family Members M 868 Applied proposed guideline amendment N 869 No Criminal Conduct (while in the US) other than Illegal Reentry M 870 Length of Time Since Last Deportation/Removal M 871 No Firearm/Weapons Involved in the Offense M 872 Lack of Treatment in Prior Sentences/Prior to Instant Offense M 873 Issue with Child Porn Use of a Computer SOC M 1200 Criminal History Issues (aggregated reason) M 1201 Unusually High Drug Amount/Purity (aggregated reason) M 1202 General Guideline Adequacy Issues M 1203 Loss Issues (aggregated reason) M 1205 Insufficient Documentation Provided on SOR to Determine Reason (aggregated reason) M 1206 Reduce Disparity (aggregated reason) D 1207 § 5H1.11 - Military Record/Charitable Works/Good Deeds (aggregated reason) M 1208 § 5G1.3 (aggregated reason) N 1209 Cooperation without motion (not § 5K1.1) (aggregated reason) C 1210 (§ 4A1.1e) Recency Amendment (aggregated reason) N 1211 Mental and Emotional Conditions (aggregated reason) M 1212 First immigration offense (aggregated reason) M 1213 Charitable Service/Good Works (aggregated reason) M Open in new tab Table A1. Monotonicity Checks . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. In column (1), the independent variable of interest is the instrumental variable, |$z_{ijc}$|⁠, as defined among all defendants in equation (2). In column (2), the independent variable of interest is judge |$j$|’s release rate among all defendants in the reverse subgroup. Open in new tab Table A1. Monotonicity Checks . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] . Subgroup First . Reverse First . Mean Release . . Stage (1) . Stage (2) . Rate (3) . Dependent variable: pretrial release (0/1) Full sample 0.237*** |$-$| 0.365 (0.051) [N = 107,246] Non-White 0.242*** 0.042 0.314 (0.059) (0.040) [N = 61,998] White 0.235*** 0.096 0.435 (0.071) (0.064) [N = 45,205] Female 0.214* 0.260** 0.616 (0.109) (0.108) [N = 16,172] Male 0.236*** 0.062** 0.320 (0.050) (0.027) [N = 90,883] Under 35 years old 0.288*** 0.187*** 0.302 (0.061) (0.054) [N = 54,981] At Least 35 years old 0.191*** 0.160*** 0.431 (0.062) (0.050) [N = 52,265] High school graduate 0.203*** 0.114*** 0.414 (0.049) (0.041) [N = 73,956] Not high school graduate 0.312*** 0.208** 0.257 (0.095) (0.090) [N = 33,290] Drug offense 0.379*** 0.224*** 0.294 (0.067) (0.066) [N = 41,329] Not drug offense 0.136** 0.051 0.410 (0.054) (0.032) [N = 65,889] Two or fewer CH points 0.175** 0.155*** 0.559 (0.076) (0.058) [N = 54,460] More than two CH points 0.233*** 0.112** 0.166 (0.052) (0.045) [N = 52,036] Impoverished 0.218*** 0.022 0.336 (0.059) (0.017) [N = 91,454] Not impoverished 0.319*** 0.293*** 0.533 (0.084) (0.075) [N = 15,485] Notes: ***|$P < 0.01$|⁠; **|$P < 0.05$|⁠; *|$P < 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions include case and demographic controls listed in the notes to Table 3. All regressions also include month and year of case initiation, sentencing year, and sentencing judge fixed effects, as well as courthouse-specific linear trends. In column (1), the independent variable of interest is the instrumental variable, |$z_{ijc}$|⁠, as defined among all defendants in equation (2). In column (2), the independent variable of interest is judge |$j$|’s release rate among all defendants in the reverse subgroup. Open in new tab Table A2. Reduced Form: Effects of Magistrate Judge Leniency Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Notes: ***|$P<0.01$|⁠; **|$P<0.05$|⁠; *|$P<0.10$|⁠. OLS regressions of case and intermediate outcomes on the IV, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table A2. Reduced Form: Effects of Magistrate Judge Leniency Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Panel A. Case outcomes . . Sentence in Months (Inverse Hyperbolic Sine) . Below-Guidelines Sentence (0/1) . Received Mandatory Minimum (if facing) (0/1) . . . Magistrate judge leniency (⁠|$Z_{ijc})$| -0.266** 0.132** -0.100 (0.134) (0.055) (0.062) Observations 107,246 106,373 37,477 Panel B. Intermediate outcomes Facing Mandatory Minimum (0/1) Final Offense Level Substantial Assistance 5K1.1 (0 /1) Govt Sponsored Departure (not 5K1.1) (0/1) Magistrate judge leniency (⁠|$Z_{ijc})$| 0.011 0.905 0.076* 0.013 (0.043) (0.706) (0.042) (0.018) Observations 107,134 106,353 106,290 101,567 Notes: ***|$P<0.01$|⁠; **|$P<0.05$|⁠; *|$P<0.10$|⁠. OLS regressions of case and intermediate outcomes on the IV, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. All regressions also include month and year of case initiation, sentencing fiscal year, and courthouse fixed effects, as well as courthouse-specific linear time trends. Government sponsored departure variable is not available for fiscal year 2002. Open in new tab Table A3. First Stage Results in the LIONS-Only Sample . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 Notes: ***|$P< 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. OLS regressions of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. In the LIONS-only data, the only control variables are gender and offense type. In columns (1) and (2), the instrumental variable, |$Z_{ijc}$|⁠, is computed using only sentenced defendants; in column (3) it is computed using all defendants. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A3. First Stage Results in the LIONS-Only Sample . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 . Data Used in the Paper—All Variables . Data Used in the Paper—LIONS Only Variables . LIONS-Only Data . . (1) . (2) . (3) . |$Z$|⁠: Pretrial judge’s propensity to release 0.237*** 0.256*** 0.298*** (0.051) (0.051) (0.082) F-statistic 21.94 24.81 13.26 Observations 107,246 107,246 44,698 Notes: ***|$P< 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. OLS regressions of whether the defendant was released pretrial on the instrumental variable, |$Z_{ijc}$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. In the LIONS-only data, the only control variables are gender and offense type. In columns (1) and (2), the instrumental variable, |$Z_{ijc}$|⁠, is computed using only sentenced defendants; in column (3) it is computed using all defendants. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A4. Effects of Pretrial Release on Dismissal in the LIONS-Only Sample . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The dependent variable is a binary variable that equals one if all charges were dismissed, declined, or acquitted (or some combination of the three) and zero otherwise. The only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A4. Effects of Pretrial Release on Dismissal in the LIONS-Only Sample . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 . OLS: All . Reduced . 2SLS: All . . Defendants . Form . Defendants . . (1) . (2) . (2) . Z: Pretrial judge’s propensity to release |$0.056^{***}$| 0.019 0.063 (0.009) (0.030) (0.109) Observations 44,669 44,669 44,669 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The dependent variable is a binary variable that equals one if all charges were dismissed, declined, or acquitted (or some combination of the three) and zero otherwise. The only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. Open in new tab Table A5. Effects of Pretrial Release on Sentence Length in the LIONS-Only Sample . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The LIONS-only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. |$^{a}$|The mean of logged sentence in the LIONS-only data is computed among sentenced defendants so as to be comparable to the mean in the data used in this article. Open in new tab Table A5. Effects of Pretrial Release on Sentence Length in the LIONS-Only Sample . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 . Sentence (Inverse Hyperbolic Sine) Original Data with LIONS-Only Variables . Sentence (Inverse Hyperbolic Sine) LIONS-Only Data . . OLS: . 2SLS: . OLS: . OLS . 2SLS: . 2SLS . . Sentenced . Sentenced . Sentenced . Heckman: . Sentenced . Heckman: . . Defendants . Defendants . Defendants . All Defendants . Defendants . All Defendants . . (1) . (2) . (3) . (4) . (5) . (6) . Pretrial Release |$-1.474^{***}$| |$-1.429^{***}$| |$-1.181^{***}$| |$-1.182^{***}$| |$-1.476^{*}$| |$-1.478^{*}$| (0.040) (0.477) (0.056) (0.055) (0.822) (0.836) Mean of DV 3.927 3.493|$^{\rm a}$| Observations 107,246 41,921 44,698 41,921 41,642 Notes: ***|$P < 0.01$|⁠; **|$P< 0.05$|⁠; *|$P< 0.10$|⁠. Standard errors are reported in parentheses and are clustered at the courthouse level. The LIONS-only control variables are gender and offense type. All regressions also include fixed effects for the month and year of case initiation and courthouse. |$^{a}$|The mean of logged sentence in the LIONS-only data is computed among sentenced defendants so as to be comparable to the mean in the data used in this article. Open in new tab Acknowledgments Peter DiCola, Vikas Didwania, Craig Garthwaite, Benjamin Jones, Max Schanzenbach, Matt Schmitt, Alison Siegler, Crystal Yang, the editor, two anonymous referees, and participants in the Kellogg Strategy Department brownbag lunch seminar provided very helpful comments and suggestions. Footnotes 1. " Gerstein v. Pugh, 420 U.S. 103 (1975); County of Riverside v. McLaughlin, 500 U.S. 44 (1991). 2. " As evidence of this point, strong support networks outside custody appear to be more effective at reducing recidivism than in-person visitation while in custody (Lee, 2019). 3. " Among three recent papers to estimate the effect of pretrial detention on sentence length, one finds no significant effect of pretrial detention in the instrumental variables estimate (Dobbie et al., 2018); one finds the opposite (Leslie and Pope 2017); and one finds significant effects among defendants charged with misdemeanors but not felonies (Stevenson 2018). 4. " Heaton et al. (2017) similarly use day-of-the-week of the defendant’s bail hearing as an instrumental variable for release in light of evidence that defendants are more likely to be released near the end of the week. The authors find that for misdemeanor defendants in Harris County, Texas, pretrial detention causes an increase in the probability that a defendant pleads guilty, serves time in jail, and recidivates. The authors also find that pretrial detention causes defendants to receive longer sentences. 5. " In fiscal year 2014, 93.6% of federal felony defendants were ultimately convicted. Just 0.4% of such defendants were acquitted after a trial, while 6.0% had their cases dismissed (Motivans 2017, 17). 6. " At the end of 2015, the federal incarcerated population comprised 196,455 prisoners, while the incarcerated populations of the two largest states—California and Texas—were 129,593 prisoners and 163,909 prisoners, respectively. (Carson and Anderson 2016). 7. " In 2006, the most recent year in which data were available, the average prison sentence for a felony defendant was 64 months in federal court and 38 months in state court. 8. " 18 U.S.C. § 3142. 9. " 18 U.S.C. § 3142(b)–(c). 10. " 18 U.S.C. § 3142(e). 11. " 18 U.S.C. § 3142(f)(2)(B). A defendant may consent to pretrial detention. For example, non-citizen defendants often consent to pretrial detention because pretrial release puts them at risk of being transferred to immigration custody (Graber and Schnitzer, 2013). 12. " These factors include: the nature and circumstances of the offense; the weight of the evidence; the history and characteristics of the person; whether, at the time of the current offense or arrest, the person was on probation, parole, or other release for a criminal offense; and the nature and seriousness of the danger to any person or the community that would be posed by the person’s release. 18 U.S.C. § 3142(g). 13. " Magistrate judges are authorized to handle additional tasks in criminal cases, and the extent to which magistrate judges perform these tasks varies by district court (McCabe et al., 2014). See subsection 4.2. 14. " I consulted local rules to verify the ways in which magistrate judges are assigned to cases in each district. To the extent that information was available, all of the district courts that I found indicated that magistrate judges rotate on a weekly, bi-weekly, or monthly schedule. 15. " Months in which two magistrate judges are listed are those in which each magistrate judge spent time on duty that month. However, no courthouse had more than one magistrate judge on duty at any given time. 16. " The Commission reports data for defendants convicted of felonies and Class A misdemeanors (Reedt, 2013). 17. " In the Commission data, roughly 6% of non-citizen defendants were released in fiscal years 2002–2013, compared to roughly 40% of U.S. citizen-defendants. 18. " In a fast-track prosecution, the prosecutor requests a reduced sentence in exchange for the defendant quickly pleading guilty and relinquishing some of their rights (Cole, 2012). 19. " The districts that border Mexico are: the District of Arizona, the Southern District of California, the District of New Mexico, the Southern District of Texas, and the Western District of Texas. 20. " The five border districts (listed in note 19) account for 34% of defendants in the Commission data. Many of the defendants prosecuted in these districts are non-citizens. The five border districts account for 17% of U.S.-citizen defendants in the Commission data. 21. " Moreover, some news reporting suggests that due to their distinctively large workloads, border districts charge more cases as misdemeanors and rely more heavily on magistrate judges compared to other districts in the sample, which could threaten the exclusion restriction, as described in subsection 4.2. (Lydgate, 2010; Barnitz, 2019). 22. " Data from the Northern District of Alabama, the Northern District of Iowa, the District of Nebraska, and the District of South Carolina list many more courthouses than actually exist in those districts. 23. " Around half of these removals are because the defendant has no judge recorded in the data; and half are because the only judge associated with the defendant’s case was their sentencing judge. 24. " As described in more detail in subsection 4.1, the instrumental variable is computed as the leave-one-out-mean release rate for the defendant’s magistrate judge. For judges with small numbers of observations, this measure will exaggerate variation in leniency. 25. " Because the data include fifteen initial years—1999 through 2013—I restrict to magistrate judges who were in the data for at least half of the period to ensure a relatively stable group of magistrate judges. The results are robust, however, to including defendants whose initial judge was not a magistrate judge, defendants whose cases were initiated outside of 1999 to 2013, and whose magistrate judges appeared in fewer than 8 years and decided at least ten (rather than twenty) cases. 26. " These variables include race, Hispanic ethnicity, sex, age, education level, criminal history, and poverty. 27. " Defendants receiving sentences of no incarceration are responsible for the discrepancy in the share of defendants receiving sentences of any incarceration in the unmatched sample versus the full universe of Commission defendants. The mass of defendants with sentences of zero months creates a many-to-many matching problem that makes these defendants intrinsically difficult to match between the LIONS and Commission data. The match rate was 68% for defendants receiving sentences of no incarceration, compared to around 78% for all defendants. It is also difficult to match defendants who received life sentences which explains why the mean sentence for unmatched defendants is noticeably larger than the mean sentence in the full universe. The match rate was 22% for defendants receiving life sentences (sentences of at least 470 months), compared to around 78% for all defendants. Fortunately, less than 1% of defendants in the Commission data received a life sentence so this problem does little to compromise the representativeness of the sample used in this article, in which the mean sentence is 63.9 months. 28. " By constructing |$Z_{ij} $| to exclude the defendant, the instrument removes the mechanical correlation between the defendant’s outcome and the magistrate judge’s leniency. The resulting estimator is a Jackknife Instrumental Variables estimator (JIVE), as described in Angrist et al. (1999). JIVEs are recommended for models that contain many instruments (Stock et al., 2002). 29. " One might also worry that the exclusion restriction will be threatened if prosecutors strategically time their decisions to file charges based on which magistrate judge is on duty. This threat seems implausible because a defendant must be brought before a magistrate judge soon after arrest. Gerstein, 420 U.S. at 125 (cited in note 1). Moreover, there is no statistically significant relationship between the defendant’s base offense level (a proxy for charging severity) and magistrate judge leniency. 30. " For example, I find that a magistrate judge’s leave-one-out rate of appointing counsel is not strongly nor significantly related to the probability that a given defendant assigned to that magistrate judges actually receives appointed counsel. In contrast—as described in subsection 5.1 and central to the estimation strategy—a magistrate judge’s leniency in granting pretrial release is strongly and significantly related to the probability that a given defendant assigned to that magistrate is released. 31. " In all of the district courts for which I could find information, appointed counsel is assigned randomly. For example, the District of Massachusetts states that “The CJA Assignment Program is a weighted system of random selection and direct assignment in which the program logic continuously self-balances the weights across all panel attorneys.” (United States District Court for the District of Massachusetts, 2015). 32. " Winning a preliminary hearing is roughly equivalent to a grand jury finding of no probable cause, which happens less than 1 in 10,000 times in federal criminal cases (Mueller-Smith, 2015). 33. " Even though the outcome variable here—pretrial release—is binary, I use OLS regressions to easily compare the coefficients to those from regressions of the instrument on the same covariates in column (5). The results are robust to using logistic regression in columns (1)–(4). 34. " The offense categories are: (1) violent crimes; (2) drug offenses; (3) fraud and similar offenses; (4) administration of justice offenses; (5) regulatory offenses; and (6) sex offenses. 35. " The instrumental variable is quite far away from one, consistent with attenuation bias in the instrumental variable due to measurement error. 36. " For example, suppose Judge |$A$| releases 50% of female defendants and just 10% of male defendants, while Judge |$B$| releases 25% of all defendants, regardless of gender. Suppose the judges each see 100 defendants—15 women and 85 men. In this case, Judge |$A$|’s release rate will be 16%, while Judge |$B$|’s release rate will be 25%. Judge |$B$| is more lenient than Judge |$A$|⁠, but there will be female defendants that are released by Judge |$A$| that would not have been released by Judge |$B$|⁠. 37. " Where possible, I divided the data into subgroups that included roughly equal numbers of defendants. 38. " Following Burridge et al. (1988), I use the inverse hyperbolic sine (IHS) transformation rather than the log transformation for the sentence length variable because the log-likelihood function for IHS is well-defined when the dependent variable equals zero. The coefficient estimates are interpreted the same way that coefficients are interpreted when the dependent variable is logged. Alternatively, the results are robust to using ln(sentence + 1) as the dependent variable. 39. " Prosecutors may, however, alter the charges by using superseding indictments—including adding a mandatory minimum charge. For example, Yang (2014) suggests that prosecutors could charge mandatory minimum crimes with superseding indictments after learning the identity of the sentencing judge as a way to avoid below-Guidelines sentences. 40. " I use the term mitigation broadly to refer to personal characteristics of the defendant their offense conduct that might affect the sentencing behavior of the judge (or prosecutor). 41. " The Justice Manual, requires federal prosecutors to obtain supervisory approval before recommending a sentence outside of the sentencing Guidelines range (Justice Manual, 9-27.730(1)), so it is possible that prosecutors prefer to exercise leniency towards defendants on the intensive margin—asking for lower sentences within the Guidelines range, rather than asking for below-Guidelines sentences. 42. " The short-term bump in bargaining reductions (dash-dotted green line) just after Booker likely reflects defendants who were re-sentenced in light of Booker. 43. " In Kimbrough’s case, the sentencing judge disagreed with how the Guidelines treat offenses involving crack versus powder cocaine. 44. " Appendix A explains how I determine pretrial detention status using the LIONS data. 45. " In the LIONS-only sample, 6.6% of defendants have their charges dismissed, declined, or acquitted. This number is extremely close to the 6.4% of felony defendants that are not ultimately convicted, as reported in note 5. 46. " For example, the Eastern District of Michigan assigns magistrate judges to criminal cases randomly in its largest courthouses (United States District Court for the Eastern District of Michigan, 2007), while the Southern District of Florida pairs district judges and magistrate judges to work together (Southern District of Florida 2015). That magistrate and district judges might be paired threatens the exclusion restriction if lenient magistrate judges are more likely to be partnered with lenient district judges. 47. " Similar to magistrate judges, sentencing judges in the analysis are defined at the courthouse level. 48. " In 1987, there were 48,300 federal prisoners; in 2016, there were 189,192 federal prisoners. The federal prison population has been shrinking since 2012. References Angrist, Joshua , Imbens Guido, and Krueger Alan. 1999 . “ Jackknife Instrumental Variables Estimation ,” 14 Journal of Applied Econometrics 57 – 67 . Google Scholar Crossref Search ADS WorldCat Arnold, David , Dobbie Will, and Yang Crystal S.. 2018 . “ Racial Bias in Bail Decisions ,” 133 Quarterly Journal of Economics 1885 – 1932 . Google Scholar Crossref Search ADS WorldCat Ayres, Ian , and Waldfogel Joel. 1994 . “ A Market Test for Race Discrimination in Bail Setting ,” 46 Stanford Law Review 987 – 1047 . Google Scholar Crossref Search ADS WorldCat Barnitz, Katy . 2019 . “ Federal Judges in New Mexico are Stretched Thin with Massive Caseloads .” Albuquerque Journal , March 23 . Available at: https://www.abqjournal.com/1295725/fedear-judges-in-new-mexico-stretched-thin.html (accessed December 1, 2019 ). Bhuller, Manudeep , Dahl Gordon B., Løken Katrine V., and Mogstad Magne. Forthcoming . “ Incarceration, recidivism and employment ,” Journal of Political Economy ; DOI: 10.1086/705330. OpenURL Placeholder Text WorldCat Burbidge, John B. , Magee Lonnie, and Robb A. Leslie. 1988 . “ Alternative Transformations to Handle Extreme Values of the Dependent Variable ,” 83 Journal of the American Statistical Association 123 – 127 . Google Scholar Crossref Search ADS WorldCat Carson, E. Ann , and Anderson Elizabeth. 2016 . “ Prisoners in 2015. ” Burearu of Justice Statistics . Available at: https://www.bjs.gov/content/pub/pdf/p15.pdf. (accessed December 1, 2019 ). Cohen, Thomas H. 2012 . Pretrial Release and Misconduct in Federal District Courts, 2008–2010 . Special Report, U.S. Department of Justice Bureau of Justice Statistics . Cole, James M. 2012 . Department Policy on Early Disposition or “Fast-Track” Programs . Memorandum, U.S. Department of Justice . Dobbie, Will , Goldin Jacob, and Yang Crystal. 2018 . “ The Effects of Pretrial Detention on Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges ,” 108 American Economic Review 201 – 240 . Google Scholar Crossref Search ADS WorldCat Foote, Caleb . 1954 . “ Compelling Appearance in Court: Administration of Bail in Philadelphia .” 12 University of Pennsylvania Law Review 1031 – 1079 . OpenURL Placeholder Text WorldCat Graber, Lena , and Schnitzer Amy. 2013 . “ The Bail Reform Act and Release from Criminal and Immigration Custody for Federal Criminal Defendants .” National Immigration Project of the National Lawyers Guild . Available at: http://www.nationalimmigrationproject.org/legalresources/practice_advisories/pa_Federal_Bail_Advisory.pdf (accessed October 14, 2015 ) Gupta, Arpit , Hansman Christopher, and Frenchman Ethan. 2016 . “ The Heavy Costs of High Bail: Evidence from Judge Randomization ,” 45 Journal of Legal Studies 471 – 505 . Google Scholar Crossref Search ADS WorldCat Heaton, Paul , Mayson Sandra, and Stevenson Megan. 2017 . “ The Downstream Consequences of Misdemeanor Pretrial Detention ,” 69 Stanford Law Review 711 . OpenURL Placeholder Text WorldCat Huang, Bert I. 2011 . “ Lightened Scrutiny ,” 124 Harvard Law Review 1109 – 1152 . OpenURL Placeholder Text WorldCat Kahneman, Daniel , and Tversky Amos. 1972 . “ Subjective Probability: A Judgment of Representativeness ” 3 Cognitive Psychology 430 – 454 . Google Scholar Crossref Search ADS WorldCat Kahneman, Daniel , and Frederick Shane. 2002 . “Representativeness Revisited: Attribute Substitution in Intuitive Judgment” in Griffin Dale W., Kahneman Daniel, and Gilovich Thomas eds., Heuristics and Biases: The Psychology of Intuitive Judgment . New York : Cambridge University Press , 49 – 81 . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Lee, Logan M. 2019 . “ Far from Home and All Alone: The Impact of Prison Visitation on Recidivism ,” 21 American Law and Economics Review 431 – 481 . Google Scholar Crossref Search ADS WorldCat Leslie, Emily , and Pope Nolan G.. 2017 . “ The Unintended Impact of Pretrial Detention on Case Outcomes: Evidence from NYC Arraignments .” 60 Journal of Law and Economics 529 – 557 . Available at: http://home.uchicago.edu/~npope/pretrial_paper.pdf (accessed May 25, 2017 ). Lydgate, Joanna . 2010 . “ Operation Streamline: Border Enforcement That Doesn’t Work .” Los Angeles Times , May 14 . Available at: https://www.latimes.com/archives/la-xpm-2010-may-14-la-oe-lydgate-immigration-20100514-story.html. (accessed December 1, 2019 ). Maxfield, Linda Drazga , and Kramer John H.. 1998 . “ Substantial Assistance: An Empirical Yardstick Gauging Equity in Current Federal Policy and Practice .” January . Available at: http://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-publications/1998/199801_5K_Report.pdf (accessed Feburary 21, 2017 ). McCabe, Peter G. , Gelpi Gustavo A. Hon. Jr., and Newman Michael J. Hon.. 2014 . A Guide to the Federal Magistrate System . White Paper, Federal Bar Association . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Motivans, Mark . 2017 . “ Federal Justice Statistics, 2014 – Statistical Tables .” Bureau of Justice Statistics . March 23 . Available at: https://www.bjs.gov/index.cfm?ty=pbdetail&iid=5873 (accessed November 21, 2017 ). Mueller-Smith, Michael . 2015 . “ The Criminal and Labor Market Impacts of Incarceration .” Available at: https://sites.lsa.umich.edu/mgms/wp-content/uploads/sites/283/2015/09/incar.pdf (accessed Feburary 15, 2018 ). Oleson, J.C. , Lowenkamp Christopher T., Cadigan Timothy P., VanNostrand Marie, and Wooldredge John. 2016 . “ The Effect of Pretrial Detention on Sentencing in Two Federal Districts ,” 33 Justice Quarterly 1103 – 1122 . Google Scholar Crossref Search ADS WorldCat Oleson, James C. , Lowenkamp Christopher T., Wooldredge John, VanNostrand Marie, and Cadigan Timothy P.. 2017 . “ The Sentencing Consequences of Federal Pretrial Supervision ,” 63 Crime & Delinquency 313 – 333 . Google Scholar Crossref Search ADS WorldCat Reedt, Lou . 2013 . Effective Use of Federal Sentencing Data . Presentation to the American Society of Criminology, United States Sentencing Commission . Sacks, Meghan , and Ackerman Alissa R.. 2014 . “ Bail and Sentencing: Does Pretrial Detention Lead to Harsher Punishment? ” 25 Criminal Justice Policy Review 59 – 77 . Google Scholar Crossref Search ADS WorldCat Stevenson, Megan . 2018 . “ Distortion of Justice: How the Inability to Pay Bail Affects Case Outcomes .” 34 Journal of Law, Economics & Organization 511 – 542 . Available at: https://ssrn.com/abstract=2777615 (accessed May 25, 2017 ). Stock, James H. , Wright Jonathan H., and Yogo Motohiro. 2002 . “ A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments .” 20 Journal of Business and Economic Statistics 518 – 529 . Google Scholar Crossref Search ADS WorldCat United States District Court for the Distict of Massachusetts . 2015 . “ CJA Assignment Protocol .” February 26 . Available at: http://www.mad.uscourts.gov/attorneys/pdf/CJAAssignmentProtocol.pdf (accessed February 21, 2017 ). United States District Court for the Eastern District of Michigan . 2007 . “ Re: Plan for the Administration of the Magistrate Judge System .” July 26 . Available at: https://www.mied.uscourts.gov/PDFFIles/07-AO-015.pdf (accessed February 21, 2017 ). United States District Court for the Southern District of Florida . 2015 . “ District Judge-Paired Magistrate Judge Reference Table .” January 12 . Available at: https://www.flsd.uscourts.gov/wp-content/uploads/2015/01/JudgePairingReferenceSheetExternalEffective01122015.pdf (accessed February 20, 2017 ). United States District Court for the Southern District of Florida . 2010 . “ Results of Survey of United States District Judges January 2010 through March 2010 .” June . Available at: http://www.ussc.gov/sites/default/files/pdf/research-and-publications/research-projects-and-surveys/surveys/20100608_Judge_Survey.pdf (accessed May 10, 2016 ). United States District Court for the Southern District of Florida . 2013 . “ Variable Codebook for Individual Offenders .” April 29 . Available at: https://www.ussc.gov/sites/default/files/pdf/research-and-publications/datafiles/Variable_Codebook_for_Individual_Offenders.pdf (accessed February 28, 2019 ). U.S. Department of Justice, Justice Manual . 2018 . https://www.justice.gov/jm/justice-manual (accessed December 9, 2019 ). Williams, Marian R. 2003 . “ The Effect of Pretrial Detention on Imprisonment Decisions ,” 28 Criminal Justice Review 299 – 316 . Google Scholar Crossref Search ADS WorldCat Yang, Crystal S. 2014 . “ Have Interjudge Disparities Increased in an Advisory Guidelines Regime? Evidence from Booker ” 89 New York University Law Review 1268 – 1342 . OpenURL Placeholder Text WorldCat © The Author 2020. Published by Oxford University Press on behalf of the American Law and Economics Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

American Law and Economics ReviewOxford University Press

Published: Apr 1, 2020

There are no references for this article.