![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
From 2000 to 2010, the federal criminal caseload increased roughly 50% with a large portion of this increase attributed to the rise in immigration prosecutions. These changes coupled with recent Supreme Court decisions rendering the guidelines advisory have renewed calls for research examining prosecutorial discretion, particularly with respect to the influence of legal and extralegal factors on charging and bargaining decisions. This study utilizes data (2002–2010) from the Federal Justice Statistics Program database housed within the National Archive of Criminal Justice Data (NACJD). More specifically, the current research examines prosecutorial decisions to decline to charge federal arrestees and to make any changes to the charge from the arresting offense. Results from the multilevel, multivariate models reveal that both extra-legal and legal factors were influential of these decisions. Disaggregated models also revealed considerable variation across different offense types. Finally, districts with higher caseloads had lower odds of a prosecutorial declination and charge change.
Notes
1 The collective findings of these studies reveal that male and minority defendants receive harsher outcomes (i.e., the in/out decision and sentence length), net of controls for other legal and extra-legal factors.
2 Johnson & Betsinger (Citation2009) reported that Asian American defendants were treated similarly to White defendants compared to either Black or Hispanic defendants.
3 A “year” represents activities occurring between October 1 and September 30.
4 Relatedly, the unmatched case rate was disproportionally higher in arrests between 2002 and 2007 and dropped roughly 50% in 2008–2010 arrests. This roughly coincides with the implementation of “Operation Streamline” by the United States Border Patrol (USBP) in conjunction with the U. S. Attorney’s Office in the Western District of Texas which resulted in increased referral and criminal prosecution of immigrants apprehended at the border for unlawful reentry (Light, Lopez, & Gonzalez-Barrera, Citation2014). Between 2005 and 2008, six Border Patrol sectors participated in Operation Streamline across three federal districts (Arizona and Western and Southern Texas); prior to this program, the Border Patrol rarely referred illegal reentry violators to the Department of Justice for criminal prosecution instead allowing them to voluntarily return (Office of Inspector General, Citation2015). This is likely an explanation for the large percentage of immigration offenses represented in the unmatched cases and the reduction in percentage of unmatched cases post 2007 due to more immigration arrests by USBP resulting in federal prosecutions rather than voluntary repatriation.
5 Transfers were not included in our measure of case declined (coded 1) because they only represented 0.3% of all cases, and less than 1% (0.9%) of “declinations.” Moreover, it was unclear whether a transfer in our data included movement to another federal district for prosecution (especially with interstate drug trafficking, and varying priorities at the different districts) or an outright dismissal. Given the interest in understanding prosecutors’ decision-making regarding whether to take further action in a case, transfers were not included in the dependent variable.
6 These data do not allow a mapping of case processing, but only whether the case was declined or prosecuted. The decision to decline a case could conceivably occur at any point through the process from when AUSAs initially receive a case all the way through trial (if it goes that far). In short, this is an analysis of outcome, not process.
7 Although recent research (Kutateladze, Citation2018) reveals that different stages in the prosecutorial process are also important for examining race effects in charge bargaining (i.e. case screening, arraignment, and final disposition), these data did not allow us to examine these processes specifically. In the multivariate analyses, our charge change variable only measures a change from one offense to another; there was no determination of whether the charge was more or less serious than the arresting charge. Consideration was given to applying a hierarchy to the charges to enable a determination of whether the prosecutor adjusted the charge to a more or less serious level compared to the arresting charge. This approach is somewhat problematic due to the challenges associated with determining an appropriate hierarchy with our offense type variable, and the fact that within the offense type variable there are multiple offenses, some of which are less serious than others. Another option was to consider the average sentence length (in months) given to these different crime types. Unfortunately, the 2016 USSC Interactive Sourcebook for Federal Sentencing Statistics (https://isb.ussc.gov/Login) indicates significant variability in the mean sentence length within crime type. For example, some violent offenses resulted in longer average sentences (murder – 241 months, kidnapping – 239 months), while others were considerably shorter (assault – 29 months). Across category similarities also made developing a hierarchy challenging. For example, drug trafficking charges resulted in a sentence length of 66 months on average, auto theft convictions produced an average sentence length of 94 months, and weapons cases resulted in an average sentence length of 75 months. In an attempt to further understand the charge change decision, however, we further examined cases for violent crime arrests in greater detail. The results of this supplemental analysis indicate that both male and Black arrestees were less likely to have their charge increased and more likely to have it stay the same as the arresting offense. Black arrestees, however, were also less likely to experience a change to a non-violent offense. All of these findings were consistent with Hypotheses 3 & 4. We include information on these results in the discussion section, and these supplemental models are available from the authors upon request.
8 Previous research has suggested that an arrestee’s prior criminal record might be influential of decisions to proceed with prosecution. In the federal criminal justice system, prior criminal record is determined by a federal probation officer via a presentence investigation report subsequent to a defendant being convicted to assist the judge in making an appropriate sentencing determination. The available data possess no information on the prior criminal history of these arrestees because it captures court actions prior to conviction. Further, the United States Sentencing Commission (USSC) data would only have prior criminal record information on arrestees who were prosecuted and convicted; arrestees whose cases were declined for prosecution would not appear in the USSC data.
9 Several studies in the past have excluded federal immigration offenses from their analyses because these cases are often handled differently than other federal crimes. Ulmer, Light, and Kramer (Citation2011), for example, note that because the majority of immigration offenders are non-citizens, deportation is a sentencing option that is not available to citizens who participate in immigration offenses, and some districts with relatively large numbers of immigration prosecutions utilize “fast-track” programs which expedite processing these cases whereas other districts do not regularly take advantage of these programs. These unique issues make analyses of federal court outcomes more difficult due to district variation in how these cases are processed (Tillyer & Hartley, 2013). Recent commentary, however, also reveals the critical importance of examining these cases in studies of federal court processes because of their dramatic growth as part of the federal caseload (Ulmer et al., Citation2011). Thus, initial analyses included immigration cases but subsequent model estimation separated cases by arrest type to identify differences in predictors of the dependent variables across different case types.
10 Ideally, the analysis plan would have included a tri-level model in which cases were nested within prosecutors and then districts. Unfortunately, these data do not identify the prosecutor or their characteristics.
11 One option to overcome the lack of data on arrestee ethnicity was to extrapolate using citizenship status. However, not all non-citizens are Hispanic, and many U.S. citizens are also Hispanic. As a result, such an imputation exercise would artificially produce bias results.
12 Immigration cases represented 23.7% of all arrests in 2002, but 61.3% of all arrests in 2010.
13 Results for year of arrest reveal, with a few exceptions, that compared to 2002, declinations and charge changes were more likely in subsequent years. We speculate that this may be related to the rising number of annual arrests over the years of data studied here which increases the caseload burden and the pressure to pursue cases that have a strong likelihood of conviction and dismiss those that are less likely to result in a conviction.
14 For example, an arrest for murder that resulted in a charge of assault was coded as a charge decrease, whereas an arrest for an assault that resulted in a charge of murder was coded as a charge increase.
15 Full tables are available upon request.