The Court Organizational Context of Diversity

Sentencing research suggests that social organizational characteristics of specific courts influence the outlooks, opportunities, and behaviors of court work groups, giving shape to cultures of case processing and thus sentencing and other outcomes (Reference EisensteinEisenstein et al. 1988; Reference StapletonStapleton et al. 1982; Reference DixonDixon 1995; Reference Ulmer and KramerUlmer & Kramer 1996, Reference Ulmer and Kramer1998; Reference UlmerUlmer 1997; Reference Ulmer and JohnsonUlmer & Johnson 2004).Footnote ² The “court community” concept has been developed to characterize this social organizational quality of legal decisionmaking in courts (Reference EisensteinEisenstein et al. 1988; Reference Ulmer and KramerUlmer & Kramer 1998; Reference UlmerUlmer 1995). The framework has been developed along the lines of classic organizational theory, examining the “processual order” (negotiated order) of court communities, to understand how court workers “organize themselves in the face of others attempting to organize them along different lines, all set against the backdrop of broader societal-structural [circumstances]” (Reference Fischer and DirsmithFischer & Dirsmith 1995:386; also see Reference WeberWeber 1954; Reference SudnowSudnow 1965; Reference CicourelCicourel 1968; Reference EmersonEmerson 1969; Reference Eisenstein and JacobEisenstein & Jacob 1977; Reference DixonDixon 1995; Reference Ulmer and KramerUlmer & Kramer 1998; Reference Kramer and UlmerKramer & Ulmer 2002).

Court cultures and outcomes may hinge on countless factors specific to court contexts and populations, tied to the identities and personalities of court workers, social demographics, economic and political pressures, and the resulting webbing of legal, public policy, and normative pressures bearing upon court participants. This view on courts as “social worlds, or communities of action and communication” (Reference UlmerUlmer 1995:589) suggests the importance of “localized, diverse processual orders in which case processing and sentencing practices develop through … interaction of courtroom workgroup members” (Reference Ulmer and KramerUlmer & Kramer 1998:251).

The framework may be particularly helpful to understanding differences between court contexts and changes in court processes, where levels and forms of decision maker participation are modified or constrained by social change (Reference Fischer and DirsmithFischer & Dirsmith 1995). Efforts to diversify decision makers and regulate decisionmaking through presumptive sentencing structures limiting judicial discretion represent two overlapping and perhaps contradictory changes affecting federal district courts in the period of interest.Footnote ³

Beyond Race and Threat: A Power Balance Perspective

Theories of “racial group threat” (Reference Bobo and HutchingsBobo & Hutchings 1996; Reference BlumerBlumer 1958) have been useful to interpreting how race and ethnicity relate to public and official orientations and behaviors related to criminal social control, especially in regard to policing and court processes (Reference Holmes and SmithHolmes, Smith, et al. 2008; Reference Ulmer and JohnsonUlmer & Johnson 2004; Reference BehrensBehrens et al. 2003; Reference CrawfordCrawford et al. 1998; Reference ChiricosChiricos et al. 1997). The threat perspective theorizes that nondominant racial groups are disadvantaged through racialized constructions of criminal tendency and responsibility and through dominant group mobilization of resources to punish and control them. Although issues of crime and fear of crime, according to these theories, are symbolic reflections of other political, economic, and social anxieties (i.e., integration, immigration, economic and political competition, etc.), these racial politics become mobilized in institutions of law and criminal social control where law enforcement resources and legal sanctions are used in accordance with dominant group interests.

Theoretical origins of the racial group threat perspective trace to the power threat thesis (Reference BlalockBlalock 1960, Reference Blalock1967; Reference Jackson and LiskaJackson 1992). The power threat thesis proposes that as minority groups grow in size and accumulate resources, they threaten majority group control, creating “a fear of political power [shifting to] the minority,” which encourages the majority group to intensify efforts to maintain social dominance (Reference BlalockBlalock 1967:147). When the power threat thesis is applied to studies of race, crime, and punishment, the size of the minority population is expected to have a positive, linear effect on mobilization of police resources and severity of criminal justice sanctions, especially where majority populations have experienced socioeconomic marginalization or decline. To test this relationship, studies typically consider the contextual significance of demographic factors, especially minority population concentration, and the relative socioeconomic standing of racial and ethnic groups, to criminal justice outcomes generally, and the sanctioning of minorities specifically. While some find little or no evidence to support the racial group threat thesis (Reference KauttKautt 2002; Reference BrittBritt 2000; Reference UlmerUlmer 1997), others report that minority concentration relates significantly to aspects of social control, including police resource allocation (Reference Holmes and SmithHolmes, Smith, et al. 2008; Reference KaneKane 2003), police use of force (Reference Liska, Yu and LiskaLiska & Yu 1992), and severity of court sanctions (Reference Kramer and UlmerKramer & Ulmer 1996; Reference Bridges and Kempf-LeonardBridges et al. 1995; Reference Kramer and SteffensmeierKramer & Steffensmeier 1993; Reference FeldFeld 1991; Reference Myers and TalaricoMyers & Talarico 1987), with minority population interacting with defendant racial status to predict disparity in sentencing severity (Reference Ulmer and JohnsonUlmer & Johnson 2004; Reference Bridges and CrutchfieldBridges & Crutchfield 1988; Reference CrawfordCrawford et al. 1998) and other sanctions (Reference BehrensBehrens et al. 2003).

Recent examinations of the power threat thesis have drawn attention to the variable and fluid nature of racial and ethnic group agency in processes of social control (Reference Holmes and SmithHolmes, Smith, et al. 2008; Reference Jackson and LiskaJackson 1992). As predicted, studies find curvilinear relationships between nonwhite group presence in the population and the racialization of social control, where increases in minority population seem to translate into redistributions of power and some ability for nonwhite groups to reshape racial politics of control (Reference KaneKane 2003; Reference Jackson and CarrollJackson & Carroll 1981).

In the twenty-first century United States, multiple racial and ethnic groups can “pressure authorities at differing levels of the crime control system” to advance shared and distinct social, economic, and political interests, such that “th[e] simple bifurcation between a dominant and subordinate group, one using the mechanisms of the state, the other its victim, is now outmoded” (Reference Jackson and LiskaJackson 1992:91). This shifting racial group power balance requires more theoretical and empirical attention. Thus while there is some truth to Reference Ulmer and JohnsonUlmer and Johnson's (2004:145) claim that “dominant [court] actors (judges, prosecutors) are white in the contemporary United States,” the reality is much browner. A racially and ethnically heterogeneous cast of authorities has emerged in federal and other court contexts over the past several decades, especially in certain jurisdictions (Reference Ward and PetersonWard 2006). In the present analysis of 89 federal districts, the percentage of white judges and prosecutors in a district varies from lows of 54 and 50 percent, respectively, to highs of 100 percent; on average, 86 percent of judges and prosecutors in a district are white (see Table 1).

Table 1. Descriptive Analysis (White and Black Offenders Only)

From a “power-balance” perspective, variable racial group agency in decisionmaking should differentiate the racialization of social control and ultimately, how race relates to court processual orders and outcomes, including sentencing disparity. In terms of criminal sentencing, for example, the degree to which and how nonwhite population concentration and crime are held to constitute threats, and the significance of race of offender and victim relationships to charging and sentencing decisions, is contingent upon the racial composition of those whose judgments shape sanctioning. For example, recent research finds that black jurors are uniquely receptive to mitigation evidence in capital cases, especially in cases involving black defendants and white victims (Reference BrewerBrewer 2004). The finding affirms the role of offender and victim race in sentencing, while also suggesting that this hinges on the race of those who influence sentence decisions.

A limitation of research on the power-threat thesis is that racial group “power” has typically been measured indirectly, as a function of population size, where a greater proportion of population is equated with greater influence over resources of social control (Reference Holmes and SmithHolmes, Smith, et al. 2008; Reference KaneKane 2003; Reference BodapatiBodapati 2008). Here power is typically defined as a threshold effect obtained as nondominant groups achieve a numerical majority in the population, at which point they are assumed to gain social, economic, and political influence. This influence is theorized in relation to both direct and indirect control of criminal justice agencies and resources; that is, as nonwhite groups gain presence and resources in the population, they are expected to gain the capacity to mobilize control resources on their own behalf (Reference HorowitzHorowitz 1985) and to negotiate accommodations of dominant group members in contexts of social control (Reference TurkTurk 1969).

In the federal criminal justice system there is little evidence that proportionally larger minority populations have greater direct influence over social control. In federal districts such as Southern Georgia, Southern Alabama, and Northern Mississippi, more than a third of residents are black, yet black judges and prosecutors on average constitute less than 10 percent of prosecutors and judges in these districts. In other districts, black representation among prosecutors and judges is more commensurate with the black residential population of the district. Few studies have measured these levels of relative racial group influence directly and assessed their relationship to outcomes. We perform a more direct test of the significance of racial group power in formal social control by measuring levels of equitable group representation in federal district courts and assessing their relation to racial parity in sentencing.

Does Racial Group Balance in Representation Increase Equity in Sentencing?

As we have noted, the contextual significance of court workforce representation has received limited attention in race and sentencing research, notwithstanding the problem of inequality and prevalent assumptions regarding the symbolic and substantive benefits of representation. Promoting ethnoracial group representation among authorities of social control, including police and court authorities, has long been considered a possible remedy to inequality in justice administration (Kerner Commission 1968; Reference IrwinIrwin 1980; Reference WelchWelch et al. 1988; Reference Ward and PetersonWard 2006). In his writing on “democratic social control,” for example, Braithwaite argues for a more equitable distribution of authority in processes of social control, holding this “control balance” to be an essential component of liberty (Reference Braithwaite and CoteBraithwaite 2002:166–7). Like the power-threat thesis, this argument looks to more balanced representation among authorities as a means of redistributing influence over norm definition and enforcement, thus enhancing the effectiveness, legitimacy, and ultimately the social justice of systems of social control.

Researchers, policy makers, and practitioners have often extolled the value of representation. Commemorative statements of the Just the Beginning Foundation, an association of minority federal court lawyers and judges founded in 1992, illustrate a common claim that equal justice may obtain from equal representation in the justice workforce. Citing the late Justice Thurgood Marshall, the foundation maintains that “African American lawyers have played a unique role in American history. Imbued with respect for the rule of law and the responsibility that such belief engenders, these lawyers have used their legal training not only to be masterful technicians but to force the legal system to live up to its creed: the promise of equal justice under law” (Just the Beginning Foundation 2009: n.p.).

Minority justice professionals are not alone in claiming that workforce representation is symbolically and substantively beneficial to equal justice administration. A recent U.S. Justice Department initiative to increase diversity among U.S. Attorneys asserts that “[O]ur pursuit of justice is stronger, and fulfillment of our national mission more effective, when we bring to bear the experience, judgment, and energy of colleagues from a wide spectrum of racial, ethnic, economic, and geographic backgrounds” (U.S. Department of Justice 2003:1). Research and policy focused on criminal justice and other public and private sectors also theorizes that representation in decisionmaking results in more substantively fair and symbolically legitimate processes, including more equitable outcomes (see Reference Thomas and ElyThomas & Ely 2001; Reference YoungYoung 1990; Reference KrislovKrislov 1974).

Expectations of substantive and symbolic benefits of justice workforce representation raise empirical questions that await more systematic scrutiny. In particular, there is a need for improved theorization and measurement of racial group representation, and greater attention to the specific range and mechanism of these race-effects. While several studies consider ethnoracial group diversity in justice administration, they typically focus on attitudes and behaviors of individual workers, such as prison authorities, police, and judges in particular (Reference Ward and PetersonWard 2006). Research on racial differences in professional orientation generally finds that status characteristics of justice workers (including race) hold limited significance to process outcomes, in comparison to occupational roles, political ideology, and professional goals (Reference Jacobs and CohenJacobs & Cohen 1978; Reference Crouch and AlpertCrouch & Alpert 1982; Reference Raganella and WhiteRaganella & White 2004). Yet others report that minorities bring distinct perspectives to bear on the idea and practice of justice (M. Reference SmithSmith 1983; Reference Bennett and JohnsonBennett & Johnson 2000; Reference CullenCullen et al. 1989; Reference JurikJurik 1985), which may affect court practices and outcomes. Research is less extensive and inconclusive on consequences of decision maker racial diversity for justice processes and outcomes. Again, several studies find no substantial race-of-decision maker affect on outcomes (Reference MastrofskiMastrofski et al. 2002; Reference Walker and BarrowWalker & Barrow 1985; Reference SpohnSpohn 1990a, Reference Uhlman1990b; Reference UhlmanUhlman 1978, Reference Spohn1979), while others report marginal or more substantial effects (Reference Luna-FirebaughLuna-Firebaugh 2003; Reference WeitzerWeitzer 2000; Reference WelchWelch et al. 1988; Reference BrewerBrewer 2004).

Studies of judicial diversity illustrate the limited and inconsistent picture of the relationship between decision maker racial diversity, professional orientations, and case outcomes. Some studies report that nonwhite judges do not treat defendants differently than white judges (Reference SpohnSpohn 1990b; Reference Walker and BarrowWalker & Barrow 1985), or that white and nonwhite judges both sentence nonwhite defendants more severely than white defendants, if for different reasons (Reference SpohnSpohn 1990b; Reference SchanzenbachSchanzenbach 2005). Others find that nonwhite judges sentence nonwhite defendants more leniently than white defendants (Reference WelchWelch et al. 1988), sentence nonwhite defendants more harshly than white defendants (Reference Steffensmeier and BrittSteffensmeier & Britt 2001), and sentence all defendants more consistently than white judges (Reference Holmes and HoschHolmes, Hosch, et al. 1993). Some find that nonwhite judges generally favor the defense at higher rates than white judges (Reference GottschallGottschall 1983) and are uniquely responsive to certain defendant claims of procedural impropriety, such as police misconduct (Reference SchererScherer 2004).

Workforce diversity appears likely to subtly and selectively differentiate court organizational norms and outcomes, rather than dramatically distinguishing courts, perhaps especially in the context of rigid guideline sanctioning. In one of few studies considering aggregate levels of court workforce representation, Reference SchanzenbachSchanzenbach (2005) finds that the significance of minority representation among federal judges is mainly limited to certain categories of cases and is associated with both advantages and disadvantages for minority defendants. In examining the racial composition of judges in federal districts in relation to district-level differences in sentencing, he finds proportions of judges in a district who are black, Hispanic, or female have little effect on sentencing overall but reduce racial and gender disparity in sentencing severity, especially in cases involving less serious offenses. Reference SchanzenbachSchanzenbach (2005) suggests that black decision makers may hold more empathy for black victims, leading to their stricter sentencing of black defendants, much like their white counterparts, albeit for different reasons. While disadvantageous to black defendants, such an orientation may be considered consistent with black community interests in crime control and equal protection under law (Reference KennedyKennedy 1997).

The inconsistency of prior research on justice workforce diversity likely reflects limitations in the conceptualization and measurement of representation and its relationship to legal decisionmaking (Reference SchererScherer 2004; Reference IfillIfill 2000). In the case of court organizational research, there is especially a need for more direct attention to group dynamics of race and representation. As we have noted, court organizational research stresses the contextual nature of court processes and outcomes but has given little attention to the contextual significance of race. Conflict theories of social control, and particularly the power-threat thesis, suggest that dominant group control over social control resources diminishes as minority groups gain presence and influence in the population and become capable of advancing their own interests in criminal social control.

(Re)distributing Racial Bias: The Diversity of Racial Group Preferences in Courts

Recent attitudinal research may also help explain how the racial composition of court authorities relates to race relations of sentencing in variably representative court contexts. First, studies find growing tendencies among whites to express indifference or lack of sympathy toward nonwhite groups (Reference Pettigrew and MeertensPettigrew & Meertens 1995), to claim unfamiliarity with issues of racial and ethnic inequality (Reference BerinskyBerinsky 1999), to express unwillingness to intervene in racially disparate treatment (Reference GoldbergGoldberg 2006), and to avoid contact with other ethnoracial groups (Reference Forman and LewisForman & Lewis 2006). Evidence ranging from the increased selection of neutral positions or “don't know” on social surveys related to race (Reference Forman, Krysan and LewisForman 2004) to candid expressions of a lack of interest in matters of racial and ethnic group relations in qualitative interviews (Reference Forman and LewisForman & Lewis 2006), suggests this increase in racial apathy in the contemporary United States and other racialized societies (Reference Pettigrew and MeertensPettigrew & Meertens 1995). Whereas racial prejudice is typically regarded as an overt expression of animus toward out-groups, these studies suggest that “modern racism is more likely expressed in a failure to help rather than a conscious desire to hurt (Reference SueSue 2005:108).

Insofar as racial apathy is more common to whites, nonwhite court authorities may be more responsive to racial inequality, and nonwhite defendants may be disadvantaged in contexts of disproportionately white representation among authorities, especially where issues of bias or discrimination are apparent or alleged (i.e., police misconduct). As such, racial apathy, and its potential disadvantage to black crime victims, witnesses, jurors, or defendants, for example, may be mitigated by black representation among decision makers, insofar as these court actors are more likely to engage racial issues. Furthermore, white racial apathy may generally be diminished in more diverse or representative courts, to the extent that nonwhite court personnel alter court cultures, through direct and indirect influence on court norms, including racial attitudes of white counterparts.

Studies of racial apathy call attention to its implicit discrimination through disengagement, yet there is also strong evidence of quite common if unconscious racial stereotyping, in the form of implicit racial bias, and its activation of racial preferences in people's choices and decisions (Reference Greenwald and KriegerGreenwald & Krieger 2006; Reference KangKang 2004). Like racial apathy research, these studies suggest that explicit bias has waned in recent times but that subtle, implicit biases still shape racial attitudes, group preferences, and behaviors. Implicit racial bias and related racial attitudes are likely important in accounting for the paradox of declining explicit bias and persistent racial stratification in the United States (Reference DunhamDunham et al. 2006:1269; Reference Greenwald and KriegerGreenwald & Krieger 2006), including continued inequality in policing and court contexts (Reference BanksBanks et al. 2006; Reference RachlinskiRachlinski et al. 2009; Reference Graham and LoweryGraham & Lowery 2004).

A range of experimental studies using Implicit Association Tests (IAT) illustrate that respondents commonly maintain attitudes that are implicitly racially biased, and that these biases affect some behaviors, even in the absence of consciously held racist beliefs.Footnote ⁴ Race IAT studies find that white Americans routinely express stronger biases of “white preference.” Black Americans hold less consistent biases, including a much stronger “black preference” than whites, and significant but lower levels of white preference, than among whites themselves (Reference Greenwald and KriegerGreenwald & Krieger 2006; Reference KangKang 2004). While the strength of the relationship between implicit bias and discriminatory behavior is not yet well-established (Reference BanksBanks et al. 2006:1188), there is evidence of an at least marginal influence of implicit bias on a range of behaviors (Reference GreenwaldGreenwald et al. 2009). Moreover, in the criminal justice context, even a slight effect on behavior is likely to compound over time to produce substantial inequality (Reference RachlinskiRachlinski et al. 2009:1201–2; also see Reference BrownBrown et al. 2003; Reference WesternWestern 2007).

Several studies have examined implicit biases of law enforcement and legal decision makers. Race IAT studies of trial court judges (Reference RachlinskiRachlinski et al. 2009) and parole and police officers (Reference Graham and LoweryGraham & Lowery 2004; Reference CorrellCorrell et al. 2007) find familiar patterns of implicit racial preference among these officials, suggesting that implicit and explicit biases likely impact legal attitudes and decisions. In a recent experimental design, Reference RachlinskiRachlinski and colleagues (2009) administered a Race IAT to 133 white and black trial court judges from three jurisdictions in the eastern and western United States. The authors found that judges harbor implicit biases much like others in the population, and these biases may have some bearing upon judgment. The authors found a strong white preference among white judges and that black judges demonstrated no clear bias in their IAT but relatively higher levels of black preference than their white counterparts. While black judges also held white preferences to a significant degree, this was far less so than in the case of their white judicial counterparts.

These racial biases were also shown to influence judgment. Judges who exhibited greater implicit white preferences gave harsher sentences to defendants when case vignettes were preceded by subliminal references to blacks than when they were preceded with a neutral priming. Similarly, judges indicating an implicit bias of black preference gave less harsh sentences to hypothetical defendants when primed with black references in case vignettes than when primed with neutral references. When the race of the defendant was explicitly indicated (in defendant descriptions), however, white judges were equally inclined to convict defendants identified as white or black, but black judges remained more inclined to convict when the defendant was identified as white rather than black. The authors conclude that black judges who harbored a white preference tended to judge more like white colleagues, in this experiment, while black judges harboring black preferences treated black defendants more leniently than others (Reference RachlinskiRachlinski et al. 2009). In general, findings suggest that black authorities counter white preferences with their own in-group favoritism, and that black defendants are advantaged at sentencing within districts with greater black representation.

Black representation among court authorities may also shift the overall magnitude of implicit racial bias, including stereotypical associations that activate anti-black bias, and contribute to white favoritism and black disadvantage in sentencing. Several scholars suggest that “automatic operation of stereotypes” may be attenuated through changes in the environment, including affirming positive associations, such as counter-stereotypical contact with racial and other out-group members (Reference Graham and LoweryGraham & Lowery 2004:500–1; Reference SinclairSinclair et al. 2005; Reference BlairBlair et al. 2001; Reference WittenbrinkWittenbrink et al. 2001; Reference PeresiePeresie 2005). Insofar as implicit bias can be controlled through these types of “social tuning,” court workforce diversity may both redistribute and limit automatic operation of racial bias by advancing specific in-group preferences and diminishing levels of out-group derogation, through counter-stereotypical contact with members of different racial groups. Thus black disadvantages in sentencing may be moderated by the balance of racial group representation in courts, both for an elevation of black preferences in districts with black representation and an attenuation of white preferences in more representative courts.

Black representation among prosecutors may be especially important for structural and social-psychological reasons. For one, presence among these substantively powerful court officials may allow black authorities to express their relative black preference, and lack of white preference, in case processing. Perhaps more significant, black representation among such high-prestige figures may symbolically disrupt a dominant racial paradigm of race, crime, and justice—pivoting on the black offender and white victim—and trigger counter-stereotypical considerations among other court workers, which may work to constrain automatic stereotypes and their impact on sentencing. As we discuss further in the conclusion, to the extent that court workforce diversity affects sentencing, this is likely for some combination of direct nonwhite group influence in case processing (i.e., the balance of white and black preference in proposed sanctions) and its indirect influence on the salience of race within particular case and court contexts.

In summary, this study aims to assess the significance of black representation among court authorities to racial parity in sentencing. Given persistent racial disparity in federal sentencing, notwithstanding sentencing guidelines and increased but variable decision maker diversity, we ask whether balanced racial group representation among federal judges and prosecutors predicts district-level racial parity in federal sentencing. The present research does not include individual-level measures of racial attitudes, or race-related behavior, but rather examines racial group dynamics of representation in relation to sentencing. Following conflict theories, including the power-threat perspective and racial attitudes research, we ask whether the variable balance of racial group authority across district courts predicts racial parity in sentencing outcomes.

Dependent Variables

Our dependent variable was severity in sentence decision. The sentencing literature has conceptualized sentence outcomes in a variety of ways, but the decision to incarcerate a convicted offender or grant probation and the length of the term of incarceration are the two most commonly examined outcomes. We defined sentence severity using the following outcome measures: (1) whether or not a defendant received a prison sentence (coded 1 if the offender was sentenced to a length of confinement of 1 month or more in a federal prison, and 0 if he or she was sentenced to any nonincarceration option such as house arrest, probation or a fine, or a term of confinement less than 1 month); and (2) the length of the sentence imposed (calculated as the natural log of the sentence length in months to compensate for the positively skewed nature of sentencing length data.Footnote ⁹

Individual Case-Level Variables

The Monitoring of Federal Sentences dataset provides detailed information on case and defendant characteristics expected to predict sentencing. The presumptive sentence, calculated here as the logged minimum months of incarceration recommended by the guidelines, measures the combined effect of a defendant's criminal history and the severity of the offense on sentence outcomes (Reference Engen and GaineyEngen & Gainey 2000). A separate measure of defendant criminal history scores was also included to account for the additional impact of an offender's criminal history beyond what was captured by the presumptive sentence measure (Reference Hofer and BlackwellHofer & Blackwell 2001). Criminal history scores ranged from 1, indicating no or limited criminal history, to 6, indicating a significant criminal history. The primary offense was measured through a series of dummy variables that captured whether the primary offense was a violent, property, drug, firearm, fraud, immigration, or other type of offense.

Case processing factors were also included to consider whether the defendant was released on bail pending sentencing (coded 0 for detention and 1 for bail). Defendants released on bail have opportunities to secure jobs, reconnect with family members, or make other meaningful changes that can impact sentence outcomes (Reference AlbonettiAlbonetti 1996). An additional measure of whether or not the case went to trial (coded 0 for a plea bargain and 1 for trial) was also included.

Departures from the guidelines were also measured. At the time when the data used in this analysis were collected, downward departures from the guidelines occurred in two main ways. Under Federal Sentencing Guidelines rule 5K2, a defendant could receive a sentence below the prescribed guideline range if the judge felt an extenuating circumstance not considered by the guidelines warranted a reduced sentence. Such departures reflected, for example, extraordinary family circumstances, evidence of rehabilitation, or an important role in the community, among other considerations. Under guidelines rule 5K1.1, defendants who provided assistance to the government could receive a sentence below the prescribed guidelines range. These departures could only be requested by prosecutors, though judges retained some discretion on the degree of departure warranted. Dummy variables measuring the presence of either a traditional downward departure (5K2) and/or a substantial assistance departure (5K1.1) were included. Recent research suggests that departures from the guidelines may be affected by defendant and district- level characteristics (Reference JohnsonJohnson 2003, Reference Johnson2005; Reference FarrellFarrell 2004; Reference JohnsonJohnson et al. 2008).

Finally, measures of defendant status characteristics were included. Defendant race was coded 0 for white and 1 for black. Defendant gender was coded 0 for male and 1 for female. Defendant age was coded in number of years at the time of sentencing. Finally, citizenship status was coded 0 for U.S. citizens and 1 for noncitizens. A full correlation matrix for individual-level variables is included in Appendix Table 1. All level one variables were grand-mean centered to test for effects of case and offender characteristics, net of effects of being sentenced in the same district.

Judicial District Variables

Of central interest here is whether greater racial group representation among key federal court authorities relates to equity in federal sentencing outcomes. Our judicial district variables included an aggregate measure of racial group representation among judges and prosecutors in 89 federal districts. By calculating the proportions of judges, prosecutors, and overall population who are black in each district, we developed an index measure of district-level representation. The index gauged dissimilarity between the racial group distribution of the district population and the racial group distribution of the district court workforce, measured as the proportion of black judges and the proportion of black prosecutors in a district divided by the black proportion of the overall district population. Values of 1 indicate equaled representation in a particular role, while values greater than 1 indicated overrepresentation and values less than 1 indicated underrepresentation. While some studies suggest the value of literal “balance” in racial group representation (Reference WeitzerWeitzer 2000), this is not clearly indicated in the power-threat literature and was unexamined here. Our models assessed whether greater black representation among district court authorities is related to equity in sentencing, where representation is defined in relational rather than absolute terms (i.e., in relation to population).Footnote ¹⁰

Our examination of the variable balance of racial group representation among both judges and prosecutors represented a departure from previous research, which has generally focused on single sets of actors, usually judges, and rarely contextualizes diversity among individual workers or occupations (Reference JohnsonJohnson 2006; Reference SchererScherer 2004; Reference SchanzenbachSchanzenbach 2005). As discussed above, the potential displacement of discretion from judges to prosecutors under structured sentencing necessitates attention to the potential independent effects of group representation among both judges and prosecutors. Measuring judge and prosecutor representation in relation to district population also provided a more direct test of the racial group power-threat thesis, which posits that group power over resources of social control varies in relation to population, and that greater equity should obtain in contexts of power balance. Our approach allowed us to empirically assess relationships between group population size and federal court representation and to compare effects of variable balance of black representation among judges and prosecutors on the sentencing of black defendants.

We incorporated several additional judicial district and court characteristics as control variables to better isolate the significance of workforce representation. The proportions of judges and proportion of prosecutors in a district who are female were included, though primarily as a control variable in the present analysis.Footnote ¹¹ Other measures of the district context included the violent crime rates of each judicial district, measured as the total number of violent crimes per 1,000 people in the district aggregated from the county to the district level,Footnote ¹² and the socioeconomic characteristics of the district, measured as a summed z-scale combining the percent of the population living in poverty and percent of the population unemployed (alpha=0.81). These measures have been found to relate to differences in sentencing across counties (Reference Myers and TalaricoMyers & Talarico 1987; Reference FearnFearn 2005).

A measure of whether the district sits in the South was included to control for unique characteristics of Southern court communities. While the South has been linked to more punitive attitudes toward sentencing and punishment in general and racial disparity in criminal justice, there are other reasons to suspect distinct politics of race, crime, and justice. Social survey data indicate that white support for punitive sentencing options such as the death penalty is significantly linked to racial antipathy, perhaps especially outside the South (Reference BorgBorg 1997). We measured Southern location to isolate and possibly compare regional differences in the race relations of federal sentencing.

Finally, we included measures of the political ideologies of governmental actors in states where the federal district courts reside. Previous research suggests that local, state, and federal court actors are influenced by local political ideologies (Reference HughesHughes 1995; Reference HuangHuang et al. 1996; Reference Helms and JacobsHelms & Jacobs 2002).Footnote ¹³ Recent research by Reference JohnsonJohnson and colleagues (2008) reports a significant relationship between political liberalness in federal court districts, measured by a composite score of senatorial voting patterns on civil rights issues, and decisions to grant downward departures from the guidelines, suggesting a potentially strong relationship between district-level political ideology, application of the guidelines, and eventual sentence outcomes. In the present research, we included a measure of state governmental ideology from 2001 commonly used by political scientists to help control for potential effects of the political context in which federal sentencing occurs on being sentenced to prison and prison length. State government ideology scores provide annual measures of “roll call voting scores of state congressional delegations, the outcomes of congressional elections, partisan divisions in state legislatures, the party of the governor and various assumptions regarding voters and state political elites” (Reference BerryBerry et al. 1998:327). They have been updated annually through 2007 by Richard Fording of the University of Kentucky. These measures have a number of advantages over the composite measure of civil rights voting utilized by Reference JohnsonJohnson and colleagues (2008). State government ideology scores measure a diverse array of congressional voting patterns and include measures of state legislator and gubernatorial activity. The scores provide a robust measure of the attitudes and political leanings of political actors in each state who directly and indirectly influence the nomination and confirmation of U.S. district court judges and prosecutors in each district (Reference GilesGiles et al. 2001) and the political ideologies that may affect the decisions of court actors operating in local contexts. Lower scores represent conservative political ideologies, and higher scores represent more liberal-leaning ideologies.

In addition to measures of the social contexts in which courts operate, we included control variables related to court organizations themselves. Numerous studies find that court structure and especially size affect the processing of cases, with larger courts often pursuing more efficiency in sentencing, given their heavier caseloads and other resource pressures (Reference WooldredgeWooldredge 1989; Reference DixonDixon 1995; Reference EisensteinEisenstein et al. 1988; Reference NardulliNardulli et al. 1988; Reference Ulmer and JohnsonUlmer & Johnson 2004). To control for these factors we included a measure of the size of the court, determined by the number of authorized judgeships, and caseload pressure, measured by the average annual criminal felony filings in a district divided by the number of judges.

It is also possible that qualitative caseload differences, such as the proportion of criminal filings for weapons-related offenses in a particular district, differentiate case processing routines in general and in relation to sentencing severity and disparity. During the early 2000s there was great variation in the volume of federal weapons filings across districts, representing a combination of different levels of offending across districts and variation in the emphasis on developing weapons cases among U.S. Attorney offices.Footnote ¹⁴ To control for such differences we included a measure of the proportion of filings in a district that are weapons-related compared to the proportion of filings for all other crime types.Footnote ¹⁵

Finally, given the centrality of federal guidelines to sentencing in the period considered, and to persistent racial disparity in federal sentencing, we included two district-level measures of court tendency to strictly adhere to guideline recommendations. Specially, we controlled for the proportion of cases that involve downward departures and/or substantial assistance departures in a district. A full correlation matrix for all level 2 variables is included in Appendix Table 2.

Analytic Strategy

Following prior research, we employed Hierarchical Generalized Linear Modeling (HGLM) and Hierarchical Linear Modeling (HLM) to assess separate and interactive effects of case-level and court contextual predictors on dichotomous prison sentence outcomes (no prison vs. prison) and linear sentence length outcomes (logged measure of months of incarceration), respectively (Reference BrittBritt 2000; Reference JohnsonJohnson 2006; Reference KauttKautt 2002; Reference Ulmer and JohnsonUlmer & Johnson 2004). Our analysis began by estimating effects of individual-level variables on being sentenced to prison and sentence length. Fixed- and random-effects models were then generated for all individual case-level variables.Footnote ¹⁶ By incorporating a unique random effect and allowing for variation by district, hierarchical modeling allowed us to examine the experiences of defendants nested in districts, which share characteristics that may similarly influence individual-level defendant outcomes. District-level variables were introduced into the models at level two,Footnote ¹⁷ and a series of cross-level analyses were conducted to assess effects of judge and prosecutor representation on sentencing by race. For all analyses, models examining incarceration were estimated using logistic regression, and models examining sentence length were estimated using linear regression.Footnote ¹⁸

Descriptive Analyses

Descriptive statistics for both case-level and district-level variables are provided in Table 1. In 2001–2002, approximately 9 percent of all federal judges and 8 percent of federal prosecutors were black, slightly below the proportion of blacks in the U.S. population (13 percent) and the average black population by federal judicial district (12 percent). Yet there was great variation among districts in levels of black representation, in terms of both population and key court authorities. At the low end, districts like Montana and Idaho had less than 1 percent black members of the overall population, while numerous districts in the South (e.g., northern and southern Mississippi, western Tennessee, and central Georgia) were more than 35 percent black.Footnote ¹⁹ It is therefore not surprising that black workforce representation varied among federal districts, with black prosecutors ranging from between 0 and 24 percent of the prosecutorial workforce of a district, and black judges constituting from 0 to 33 percent of the judicial workforce.

To assess relationships between representation in the court workforce and outcomes in sentencing, we created an index of dissimilarity between black representation in the district court workforce and black population in the district, where a value of 1 indicated proportional representation, values under 1 indicated underrepresentation, and values over 1 indicated overrepresentation. On average, blacks were underrepresented among judges (0.86) and prosecutors (0.75) across the federal districts.

It is not always the case, however, that districts with sizable black populations contain equally representative court organizations. While the two variables are strongly correlated,Footnote ²⁰ variation in representation among judges and prosecutors within districts, and levels of representation across districts, suggested the limits of using population proportion as an indirect indicator of relative group power in contexts of social control. In the state of Mississippi, for example, blacks made up large and nearly equal proportions of the northern (37 percent) and southern (35) federal districts yet were twice as underrepresented among judges and prosecutors in the northern (index=0.26) than in the southern (0.50) district. In the district of southern Georgia, blacks made up 33 percent of the population and 60 percent of defendants, yet they were severely underrepresented among judges and prosecutors (0.08), and far more so than in the central (0.60) and northern (0.68) districts of the state.

Few districts had high levels of black representation among both judges and prosecutors. After using a black representation index of 1.01 as a cutoff (meaning that blacks were 1 percent more likely to be in the court workforce as in the general population), we found that only 12 districts had overrepresentation of both black judges and prosecutors. Ten districts had overrepresentation of black prosecutors but not black judges, and 15 districts had overrepresentation of black judges but not black prosecutors. Fifty-two districts had underrepresentation among both black judges and prosecutors, including 16 districts with no black representation among either judges or prosecutors.

Similar to prior research on judicial diversity in state courts (Reference JohnsonJohnson 2006), we generally found more gender than racial diversity among both federal judges and prosecutors. In 2001, women held 20 percent of federal judgeships, and 48 percent of federal prosecutors were women. While proportions of federal judgeships held by women varied considerably across districts (0 to 54 percent), female prosecutors were more evenly distributed, ranging between 27 and 73 percent of the prosecutorial workforce. Our data do not presently allow us to disaggregate combined race and gender diversity.

Districts also varied on other dimensions, including size, caseload characteristics, liberalness, and the tendency to adhere to guidelines. While there were an average of seven judges per court, districts varied from those with only a few judges and fewer than 25 prosecutors to large districts of more than 25 judges and several hundred prosecutors. Similar variation existed in the quantity and quality of district caseloads. On average, judges received a total of 90 felony filings per year, but that figure ranged from 25 to 364. There was also significant variation among districts in the types of cases they handled, including the volume of cases involving weapons offenses. Two-thirds of districts (63 percent) clustered within one standard deviation of the mean for the proportion of cases involving weapons charges, (between 12 and 30 percent of criminal filings) but weapons offenses made up between 1 and 12 percent of criminal filings in 14 districts at the low end and between 30 and 45 percent of the filings in 12 districts at the high end. Another key difference between federal districts was the degree of adherence to sentencing guidelines. On average, districts granted a downward departure in 12 percent of cases and substantial assistance departures in 20 percent of cases, but the variation between districts was large (standard deviations of 11 and 9 percent, respectively). Some districts seemingly more inclined to grant downward departures also appeared less likely than average to grant substantial assistance departures, and vice-versa. Ideological differences also existed among districts, evident in variation in state government ideology scores.Footnote ²¹

Explaining Differences in Sentence Outcomes

We developed two-level unconditional models to measure the amount of variance in sentencing explained by individual and district factors. Individual case-level factors accounted for the majority of variance in imprisonment and sentence lengths. However, 8 percent of the variance in the decision to incarcerate and 7 percent of the variance in the sentence length was attributable to differences in the federal districts.

Despite the fact that the variance in sentences from district court level measures was small, we caution against a conclusion that racial stratification, including variable balance in representation, is unrelated to sentencing. First, as other research clearly shows (Reference ZatzZatz 1987), court outcomes can mask subtle, cumulative inequalities that can become embedded and amplified in “legal variables” (i.e., prior police contact, arrest, prior incarceration, etc.), some of which may pertain to racial group balance in representation among police and other court (i.e., juvenile and state) authorities (Reference BrownBrown et al. 2003; Reference Bridges and SteenBridges & Steen 1998). Further, we should expect little district-level variation in sentencing under presumptive sentencing systems. This is particularly true because the individuals examined in the present analyses were sentenced prior to the federal sentencing guidelines becoming advisory (United States v. Booker 543 U.S. 220 2005), where a core goal of the guidelines was to standardize sentences across judicial districts. The fact that any variation in sentence exists under such a structured sentencing system, after controlling for case factors, invites further scrutiny.

The effects of individual case- and district-level variables on being sentenced to a term of incarceration (coded 0 for no prison and 1 for prison) and length of incarceration are found in Table 2. The results presented in Table 2 include both individual- and district-level findings from the random-effects models because individual-level fixed-effects results were relatively unchanged with the addition of district-level variables. As expected, higher criminal history scores and presumptive sentences significantly increased the likelihood of incarceration and incarceration length. Conversely, departures from the guidelines and being released on bail before sentencing decreased the chance of incarceration and length of sentence.

Table 2. HGLM and HLM Regression Random Coefficient Models of Receiving a Prison Sentence and Prison Length

^* p<0.10, ^**p<0.05, ^***p<0.01

^a χ²=125.45, df=76; ^bχ²=437.01, df=67; ^cReference category is all other offenses.

Several defendant status characteristics also related to sentence outcomes. Noncitizens were more likely to be incarcerated than citizens. Female offenders had a significantly decreased likelihood of being incarcerated as compared to men and on average received shorter sentences. Older defendants had a decreased likelihood of incarceration, though the effect of age on length of sentence was not significant. Black defendants had a higher likelihood of being incarcerated than white defendants (14 percent higher odds) but did not differ from whites significantly in the length of their sentence.

In line with other research examining differences in sentence outcomes across courts (Reference FearnFearn 2005; Reference JohnsonJohnson 2006; Reference KauttKautt 2002), random-effects models indicated that the effect of nearly all the case-level factors varied across judicial districts, suggesting that these factors are weighted somewhat differently in sentencing decisions, depending upon the district. To measure the effect of district characteristics on sentence outcomes, we constructed two-level random-effects models. More liberal districts handed down shorter sentences than more conservative districts. Larger districts had shorter terms of incarceration than smaller districts, though caseload pressure increased sentences length across districts. Districts with higher departure rates not surprisingly sentenced less severely. Violent crime rate, location in the South, socioeconomic status of the district, and type of charges filed in the district were unrelated to sentencing outcomes in the multilevel model.

We turn now to findings regarding the relationship between race, workforce representation, and sentencing severity and parity. District-level variation in black representation among judges was not significantly related to either being sentenced to prison or sentence length. This finding appears to support literature suggesting a weak relationship between judge diversity and sentencing. However, while levels of racial group balance in judicial representation did not predict sentencing, offenders were less likely to be sentenced to prison, and likely to receive shorter sentences, in districts with greater black representation among prosecutors. This suggests that racial group balance in representation among prosecutors may have a stronger impact on sentence outcomes than representation among judges, perhaps especially in this context of presumptive guideline sentencing. While most research on racial group representation among legal decision makers has focused on the judicial role, there is clearly a need for more attention to the prosecutorial role.

To test our hypotheses that greater black representation among court authorities increases racial parity in sentencing, we examined cross-level interactions between black workforce representation and race of defendant separately for each occupational group. A cross-level effect identified the degree to which a district-level variable (here work group representation) conditioned the relationship between an individual-level independent variable (offender black) and the dependent variable (sentence). The coefficients for individual-level variables were treated as outcomes predicted by district-level variables. The cross-level effects models contained main effects and cross-level effect coefficients. Main effect terms were fixed-effects coefficients measuring the average effect of the individual-level variable on the sentence outcome. The analysis helped identify how district variation in the balance of black representation among judges and prosecutors conditions the relationship between defendant race (black) and imprisonment decisions.

Table 3 reports the cross-level interaction of black judge and prosecutor representation on the sentencing of black defendants.Footnote ²² As we noted in the previous analysis, black defendants were more likely to be sentenced to a term of incarceration than white defendants; however, when black defendants were sentenced in districts with greater black representation among judges and particularly prosecutors, this disparity between white and black offender sentences lessened. To help interpret the nature of the interaction effects, predicted probabilities of incarceration for black and white defendants at different levels of black representation of judges and prosecutors were calculated holding all other variables constant at their mean.Footnote ²³ Differences in the probabilities of black and white offender incarceration at various levels of black judge and prosecutor representation are plotted in Figure 1. The x axis in Figure 1 represents the level of black representation of the workforce group from 0 (indicating no black representation) to three (indicating that black workers were overrepresented by three times the general population). A measure of 1 on the x axis indicated that black decision makers in the district were represented at an identical rate to black members of the general population of the district.

Table 3. HGLM Cross-Level Interaction for Offender Race and Judge and Prosecutor Representation

Note: Cross-level models only run on prison y/n outcomes due to the face that race was not significant in Prison-length random-effects models.

^* p<0.10, ^**p<0.05, ^***p<0.01

Note: The x axis represents the level of black representation of the workforce group from 0 (indicating no black representation) to 3 (indicating that black workers are overrepresented by three times the general population). A measure of one on the x axis indicates that black decision makers in the district are represented at an identical rate to black members of the general population of the district The y axis represents the predicted probability of incarceration.

Figure 1. Predicted Probability of Incarceration for White and Black Defendants by Various Levels of Judge and Prosecutor Representation

Figure 1 illustrates that the main effect for black defendants on incarceration was conditioned by variable district-level workforce representation, such that the probability of a prison sentence for black defendants decreased, relative to black representation among judges and prosecutors. While an overwhelming majority of federal offenders were sentenced to a term of imprisonment regardless of the racial group balance in district court workforce representation, at lower levels of black judge and prosecutorial representation, black offenders were slightly more likely to be incarcerated than whites. Conversely, at higher levels of black representation, white offenders were slightly more likely to be incarcerated than blacks. For example, as black representation among prosecutors increased from zero (indicating no black representation among workers) to three (indicating a prosecutorial workforce that was three times more likely to be black as the general population of the district), the likelihood of incarceration for a black defendant decreased from 96.6 to 93.2 percent. The effect of black representation among prosecutors had a notably stronger effect on sentencing than did black representation among judges. These findings appear to confirm the importance of racial group balance among authorities of social control. Indeed, black and white offenders appeared to be disadvantaged by their respective underrepresentation among authorities of social control. These findings underscore the group dynamics of race, representation, and sentencing, and particularly how racially representative groups of prosecutors manage to alter the race relations of federal sentencing.