EXTRALEGAL DISPARITIES IN THE SENTENCING OF FEDERAL FRAUD OFFENDERS By Mark Speers A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Criminal Justice-Master of Science 2021 ABSTRACT EXTRALEGAL DISAPRITIES IN THE SENTENCING OF FEDERAL FRAUD OFFENDERS By Mark Speers This study aims to examine the main and combined effects of gender and race on the odds of incarceration and downward departure for fraud offenders processed in federal courts during the fiscal year 2015-2016. Data was obtained from the United States Sentencing Commission’s Monitoring of Federal Criminal Sentences dataset. A sequence of multilevel logistic regressions is performed to elucidate the main and combined effects of race and gender on the decision to incarcerate and depart downward from the federal sentencing guidelines. Findings attest to the importance of studying the interaction of extralegal characteristics in studies of sentencing disparities. Black males were the only race-gender subgroup to experience a statistically significant disparity in the odds of incarceration compared to White males. Additionally, findings reveal significant inter-district disparity in the odds of incarceration and downward departure receipt. Future qualitative and quantitative research avenues for sentencing research are discussed in light of prior research and the findings of this study. TABLE OF CONTENTS LIST OF TABLES..........................................................................................................................iv Introduction......................................................................................................................................1 Background......................................................................................................................................5 Relevant Empirical and Theoretical Literature................................................................................8 Focal Concerns Perspective and Relevant Literature..........................................................8 Federal Sentencing Departure Disparities and the Liberation Hypothesis………………12 Offense-Specific Stereotypes…………………………………………………………….16 Methodology..................................................................................................................................25 Data....................................................................................................................................25 Sample................................................................................................................................25 Measures............................................................................................................................26 Dependent Variables..............................................................................................26 Independent Variables...........................................................................................26 Control Variables...................................................................................................26 District-level controls.................................................................................28 Analytic Procedure.................................................................................................28 Findings.........................................................................................................................................30 Descriptive Statistics.........................................................................................................30 Bivariate Statistics.............................................................................................................32 Explained Variation at Levels 1 and 2...............................................................................36 Multivariate Mixed Effects Logistic Regression Models..................................................37 Multicollinearity Diagnostics.................................................................................37 Odds of Incarceration ............................................................................................38 Odds of Downward Departure...............................................................................40 Analyses of Combined Race-Gender Effects........................................................41 Discussion......................................................................................................................................45 Conclusion.....................................................................................................................................51 BIBLIOGRAPHY..........................................................................................................................52 iii LIST OF TABLES Table 1. Descriptive Statistics of Convicted Federal Fraud Offenders..........................................31 Table 2. Bivariate Relationships of Categorical Variables.............................................................32 Table 3. Bivariate Relationships of Continuous Variables to Incarceration....................................33 Table 4. Bivariate Relationships of Continuous Variables to Downward Departure.....................34 Table 5. Multilevel Mixed Effects Logistic Regression of Race, Gender, and Other Legal and Extralegal Variables on the Odds of Incarceration and Downward Departure...............................38 Table 6. Multilevel Mixed Effects Logistic Regression of Race-by-Gender intersections and Other Legal and Extralegal Variables on the Odds of Incarceration and Downward Departure...............42 iv Introduction Early studies of white-collar crime sentencing questioned whether some white-collar offenders were given more lenient sentences compared to street crime offenders or other white- collar offenders based on their social statuses, such as education level and professional occupational status (Hagan et al., 1980; Nagel & Hagen, 1982; Tillman & Pontell, 1992; Wheeler et al.,1982). This line of inquiry was drawn from offender-based definitions of white- collar crime (e.g., Sutherland, 1983) that emphasize the high social status of the offenders committing the crime, as well as the use of their privileged occupational roles to commit them. In general, these dimensions of social status were unrelated to sentencing outcome (Hagan et al., 1980; Nagel & Hagan, 1982; Tillman & Pontell, 1992: for an exception, see Wheeler et al., 1980). More contemporary analyses of white-collar crime sentencing tend to focus on the influence of other status-related variables, typically race, gender, and age (Cassidy & Gibbs, 2019; Holtfreter, 2013; Maddan, Hartley, Walker, Miller 2010; Testa, 2019; Van Slyke & Bales, 2012, 2013). These studies often use definitions of white-collar crime that highlight the nature of the offense, such as non-violent or non-physical crimes committed through deception (Van Slyke & Bales, 2013). Though findings are mixed, they generally show enhanced sentence severity for males compared to females (Maddan et al., 2010; Testa, 2019; Van Slyke & Bales, 2013) and Blacks compared to Whites (Cassidy & Gibbs, 2019). This work is representative of and consistent with a much larger body of sentencing literature on street crime. The primary goal of this literature is to assess the extent to which “extralegal” factors, or characteristics of an offense or offender that are not legally relevant to the sentencing decision, influence sentencing outcomes. Though sentencing decisions are supposed 1 to be based on legally relevant factors (e.g., offense seriousness, criminal history), discretion can produce disparities in sentencing outcomes linked to these extralegal characteristics (e.g., race, gender, age, occupation). The focal concerns perspective is now the most common framework for studying and understanding both street and white-collar sentence disparities (Frase & Mitchell, 2020). This perspective postulates that judges are guided by three primary factors when deciding upon the appropriate sentence for an individual defendant (Steffensmeier et al., 1998): blameworthiness, protection of the community, and practical constraints and consequences of the sentence. Blameworthiness is an assessment of culpability; protection of the community (i.e., threat reduction) is a consideration of the offender’s dangerousness and the protection afforded to the community through incapacitation and general deterrence; practical constraints and consequences refers to constraints on the criminal justice system (e.g., prison capacity) and the consequences of the sentence for the offender and their family. Additional theoretical concepts have been developed to further explain how extralegal disparities operate in specific sentencing outcomes for specific offenses. First, the liberation hypothesis asserts that sentencing decisions or criminal cases which involve more discretion will be more heavily influenced by extralegal variables (Frase & Mitchell, 2020). The name of the liberation hypothesis comes from the elevated discretion or liberty court actors have during certain decisions and cases. In other words, heightened discretion liberates, or frees, judges to rely on their own biases related to extralegal factors to a greater degree. This hypothesis is employed in this study to expand the traditional sentencing research approach. Instead of using the liberation hypothesis to assess differences across cases, this study considers two sentencing decisions with disparate levels of discretion. 2 Second, street crime sentencing scholars have theorized that extralegal disparities may be driven by stereotypes related to specific offenses. For example, criminologists theorized that harsher sentencing outcomes for Hispanics for drug trafficking and Blacks for drug distribution may be the result of racial stereotypes related to each offense (Steffensmeier & Demuth, 2000). However, mock jury studies, one of the primary sources of evidence pertaining to offense- specific stereotypes, indicate that stereotypes of white-collar and street offenders differ (Gordon, 1990; Jones & Kaplan, 2003; Skorinko & Spellman, 2013). For example, one study found that the black burglar’s offense conduct was more likely to be attributed to the personality of the defendant compared to a black embezzler (Gordon, 1990). This study will draw upon the stereotypes hypothesis by examining extralegal disparities for a specific (and understudied) form of white-collar crime—fraud. Shifting racial demographics among white-collar offenders underpin the theoretical need to examine specific offense types. Benson and colleagues (2020) found that minorities comprise a larger percentage of white-collar offenders in recent years compared to the 1970s. This growth in minority involvement was especially large in fraud offenses and bribery. As fraud offender demographics have shifted, so too may the stereotypes of fraud offenders held by court actors. Thus, these findings further demonstrate the need to examine extralegal sentencing disparities for specific forms of white-collar offending. In sum, this study draws on the focal concerns perspective, the liberation hypothesis, and the stereotypes hypothesis to examine multiple sentencing outcomes among fraud offenders. Data are drawn from the United States Sentencing Commission’s Monitoring of Federal Criminal Sentences series. Fiscal year 2015-16 is used in order to account for recent reformations to the guidelines (described below). 3 This work contributes to the sentencing literature in several ways. This study represents a current examination of white-collar crime sentencing patterns, adding to a very small body of extant literature. From a theoretical perspective, this study will provide an additional assessment of the generalizability of the theoretical frameworks designed to explain street crime sentencing decisions. Additionally, this study applies the liberation hypothesis in a novel way by assessing extralegal sentencing disparities across two decision points with varying levels of discretion. Finally, by testing the white-collar crime stereotypes found in mock jury studies using actual sentencing data, this study can inform whether court actors sentencing decisions are similar to the perceptions of mock jurors. In the following sections, I begin with a description of background information on the federal sentencing guidelines and relevant reforms, accompanied by a brief explanation of how the sentencing decisions relevant to this study are made. Next, I provide a detailed discussion of the three theoretical perspectives informing this study, as well as the literature upon which these perspectives are based. The focal concerns perspective is presented first, followed by the liberation hypothesis and the evidence for offense-specific stereotypes. Hypotheses informed by these perspectives are presented after each theory and the relevant literature is described. Following the empirical and theoretical literature section, the methodology is described and the results are presented. I conclude with a description of limitations and implications of this work, as well as directions for future research. 4 Background The federal sentencing guidelines were created in 1987 by the United States Sentencing Commission (USSC, 2015) as a result of the Sentencing Reform Act of 1984, which aimed to reduce disparities in the indeterminate sentencing regime. In a typical case, the guidelines provide a sentencing range based on two factors—the offense seriousness and the criminal history of the specific defendant (USSC, 2015). Though initially mandatory, in the 2000s, these guidelines were transformed by case-driven reforms. In United States v. Booker, the U.S. Supreme Court concluded that the guidelines must be advisory instead of mandatory to maintain their constitutionality. Subsequently, Gall v. United States (2007) decided that federal district court judges “may not automatically presume the Guidelines range to be reasonable and must” make their own assessment of the facts presented in the case (Ulmer, Light, Kramer, 2011. p. 1079). As a result of these reforms the federal sentencing guidelines act as recommendations or standards upon which to sentence. Should a judge desire to sentence outside the range provided, all they must do is provide a reason for doing so, which in some cases is as simple as writing U.S. v. Booker in the appropriate statement of reasons form (Kaiser & Spohn, 2018). In addition to determining whether to sentence within or outside of the guidelines, the sentencing process involves many other decisions, each of which can be made with some amount of discretion on the part of the sentencing entity (i.e., judge, prosecutor). The level of discretion involved varies from one decision to the next and is partially dependent upon the structure of the guidelines and legal requirements and considerations of each decision. The two decisions of primary concern to this study are the decision to incarcerate and the decision to grant a downward departure. 5 The decision to incarcerate is the culmination of decisions regarding the calculation of the final offense level and criminal history score, which are also based on previous decisions. For example, final offense level is determined by the “offense conduct charged in the count of the indictment or information of which the defendant was convicted” (USSC, 2015. p. 20). Thus, the final offense level is partially dependent upon the charges levelled at the defendant, affording the prosecutor some weight in determining the sentence. In addition, one’s criminal history score is dependent upon the severity of sentences imposed in prior convictions. Thus, the decision to incarcerate for a certain offense is partially based on the decisions made in a separate sentencing incident. Once the final offense level and criminal history score are decided upon, the advised sentence range is calculated. Following this, the court considers the relevance of specific offender characteristics (e.g., age, education and vocational skills, mental and emotional conditions) that when present to an atypical degree may warrant a departure. Additionally, the court considers the applicability of specific departure provisions delineated in the sentencing guidelines such as: departures for substantial assistance to authorities, victim’s conduct, coercion and duress, or early disposition programs (USSC, 2015. p. 461-478). The court then concludes the sentencing process by considering “the applicable factors in 18 U.S.C § 3553(a) as a whole.” (USSC, 2015. p. 16). U.S.C. § 3553(a) is the portion of the United States Code pertaining to the determination of a sentence. This section of the code describes the relevant conduct to be considered when imposing a sentence upon a defendant. This last step is well conceptualized as a judgement made while viewing all relevant conduct and purposes of punishment, as a whole, and a decision that best accounts for this conduct and purpose. 6 In sum, a judge is responsible for the singular decision of whether to impose a sentence within the range of the guidelines, and in doing so potentially sentence a defendant to incarceration; however, there are many structural elements of the guidelines that grant influence to prosecutors and defense attorneys (e.g., offense level based on conduct the defendant is convicted of) and judges who have previously sentenced this defendant (e.g., criminal history score based on severity of previous sentences). That being said, the judge does make the decision in that moment of sentencing, though the rationale for the decision is predicated upon the actions of lawyers and other judges. Additional input from other court actors may further shape the decision for some forms of departure from the guidelines, as there are various types, such as downward departures below the advised range. Prosecutor initiated (aka. government-sponsored) downward departures are motioned for by the prosecution. The judge can either grant or deny these motions. Forms of prosecutor initiated downward departures include substantial assistance departures and early disposition departures. Prosecutors motion for this kind of departure when the defendant has provided “substantial assistance in the investigation or prosecution of another person who has committed an offense” (USSC, 2016. p.461). Early disposition departures are intended to be applied to defendants who enter into an early disposition program which may save court and government resources. In contrast to prosecutor-initiated departures that involve input from multiple court actors, judge-initiated departures are granted at the judge’s discretion based on statutory factors in United States Code 18 U.S.C §3553(a). Some examples of factors include characteristics of the defendant, defendant criminal history, and family circumstances, but could also include the judge’s personal and subjective philosophies of punishment as has been evidenced in studies of judge-initiated departures (Kaiser & Spohn, 2018). 7 Relevant Empirical and Theoretical Literature The focal concerns perspective is the overarching theoretical framework that spells out the relevant variables to assess when studying sentencing. Focal concerns perspective also provides an explanation, albeit one that lacks direct examination, for why the identified variables relate to and influence sentencing decisions. This study utilizes focal concerns as the guiding perspective informing the choice of control variables. The other two theoretical elements, namely the liberation hypothesis and offense-specific stereotypes, can be conceptualized as theoretical “a la carte” additions to the entrée that is the focal concerns perspective. These theoretical frameworks present explanations for why sentencing disparities may manifest in specific scenarios based on characteristics of the case or defendant. In this study, the liberation hypothesis provides a rationale for expecting sentencing disparities to manifest differently across sentencing decisions based on unequal levels of discretion. Offense-specific stereotypes are used to explain racial and gendered sentencing disparities based on stereotypes specific to white-collar offenses. Focal concerns is used to frame the general nature of sentencing decisions. A large number of studies have examined disparity in various sentencing decisions over the past 50 years, particularly after the establishment of the federal sentencing guidelines (USSC, 2015). Many of these studies have examined racial and other extralegal sentencing disparities (Doerner & Demuth, 2010; Hartley, Maddan, Spohn, 2007; Maddan & Hartley, 2018; Ulmer and Johnson, 2004; Ulmer, Light, Kramer, 2011). The overall findings are fairly consistent and have primarily been interpreted using the focal concerns perspective. Focal Concerns Perspective and Relevant Literature The most consistent finding of this body of research is that legally relevant variables, such as offense seriousness and criminal history, are the most robust and powerful predictors of 8 sentencing severity (Baumer, 2013; Cassidy & Gibbs, 2019; Steffensmeier, Painter-Davis, Ulmer, 2017). For extralegal variables, studies have overwhelmingly found that minorities and males experience harsher sentencing than whites and women, respectively (Doerner & Demuth, 2010; Kutateladze, Andiloro, Johnson, Spohn, 2014; Mustard, 2001; Steffensmeier et al., 2017). Some recent studies the interaction of extralegal variables, such as how racial sentencing disparities may be conditioned by other extralegal variables such as gender and age (Doerner & Demuth, 2010; Steffensmeier et al., 2017). Although the results of studies focused on intersectionality are somewhat mixed, overall they support the notion that young adult males of minority races most often experience the harshest sentencing outcomes (Steffensmeier et al., 2017). Focal concerns perspective (FCP) is the dominant theory employed to interpret research into unwarranted sentencing disparities (Frase & Mitchell, 2020). Initially conceptualized as a framework for judges’ decisions, this theoretical framework has been extended to include the decision-making process of prosecutors (Steffensmeier, Ulmer, Kramer, 1998; Holmes & D’Amato, 2019). FCP asserts that in making decisions, court actors are guided by three focal concerns: blameworthiness, protection of the community, and practical implications of sentencing decisions (Maddan & Hartley, 2018; Steffensmeier, Painter-Davis, Ulmer, 2017). Thus, the sentencing decision (i.e., the decision to incarcerate and the length of the period of incarceration) made by the judge is based on their assessment of these three factors. Assessments of blameworthiness are grounded in retributive philosophies of punishment. This assessment is concerned with having the punishment be adequately proportional to the crime committed. The primary factors considered in this assessment are the seriousness of the offense, criminal history, and offender’s role in the offense (Steffensmeier et al., 2017). The second focal concern is 9 protection of the community. The goal of protecting the community is obtained by incapacitating or deterring the offender. Put differently, the judge evaluates the dangerousness of the defendant. This evaluation is conditioned by criminal history and seriousness of the offense. The third focal concern(s) are the implications or consequences of the sentencing decision. This assessment is made in light of implications of the sanction pertaining to the defendant, the court, and more broadly, the criminal justice system. Assessments of the implications of criminal punishment for the defendant take into consideration the individual’s familial connections and the health of the offender (Frase & Mitchell, 2020). Evaluations of criminal sentencing implications on the court and the criminal justice system consider the capacities of correctional institutions, efficiency of the court, and inherent biases within criminal justice policies (e.g., powder-to-rock cocaine sentencing discrepancy). The theoretical mechanisms through which these three focal concerns result in unwarranted sentencing disparities are grounded in Albonetti’s uncertainty avoidance and causal attribution hypotheses (1991). The assertions of uncertainty avoidance and causal attribution center around the inability of humans to make accurate predictions about the future, especially when the decision-making environment is contextualized by limited time and information. In the case of sentencing specifically, judges are perceived as unable to make accurate predictions of a defendant’s likelihood of recidivism. Because of this inability, judges may incorporate extralegal information, such as criminal stereotypes based on the defendant’s gender, race, or age into their sentencing assessment. For instance, females may be perceived as less blameworthy because of societal expectations of passivity and dependence (Spohn & Brennan, 2011).1 These prejudicial 1 Women may also benefit from elevated association with familial roles as judges may be more inclined to consider the negative effects of incarceration on children and other dependents (Van Slyke & Bales, 2013). Alternatively, minorities, and particularly Blacks, may be perceived as more likely to recidivate, more violence-prone, or more dangerous (Bridges & Steen, 1998; Steen, Engen, Gainey, 2005). 10 stereotypes result in a sentencing decision whose rationality is bounded by the uncertain circumstances characterizing assessments of the defendant’s likelihood of future criminal behavior. The resulting assessment, based upon legally relevant as well as legally irrelevant offense and offender characteristics, is termed a perceptual shorthand. Research into sentencing disparities has typically focused on the decision to incarcerate and the sentence length. These decisions are the last steps in a long process that includes charging decisions, case dismissals, record sealing, and ends with guidelines departures. All of these decisions involve discretionary action from court actors and therefore could potentially disadvantage defendants of certain social statuses more or less than others. The required legal considerations and levels of discretion of sentencing decisions influence the effect of legal and extralegal variables on said decisions. Empirical analyses bear this assertion out. Meta-analyses and assessments of sentencing disparity research indicate that the influence of race and other extralegal variables on sentencing is dependent upon the outcome examined (Mitchell, 2005; Baumer, 2013), making investigation of multiple decision points crucial. Accordingly, assessing multiple decision points within one study facilitates a more comprehensive understanding of how the influence of extralegal variables shifts (or accumulate) between sentencing decision points. Departures are an important decision point to consider, as the previously described reforms increased the amount of judicial discretion in sentencing. Recall that United States v. Booker changed the guidelines from mandatory to advisory and Gall v. United States required federal district court judges to make their own assessments rather than presuming that the guidelines were reasonable. US v. Booker brought additional changes to the standard of review that appellate courts apply when appraising district court sentencing decisions. Subsequent to Booker, the standard of review for sentences shifted from precise application of the guidelines to 11 a “reasonableness” standard (Kaiser & Spohn, 2018; USSC, 2016). This change effectively made it much more difficult for appellate courts to overturn outside-of-range sentence decisions and as a result decreased the oversight over these decisions (Holmes, Feldmeyer, Kulig, 2020). The result was an increase in outside-of-range sentences (USSC, 2010). Between 2005 and 2016, the percentage of within range sentences has fallen from 70% to approximately 45% (USSC, 2010, 2017). This corresponds to an increase in both government sponsored and judge-initiated departures by 6% and 16%, respectively, during the same time period (USSC, 2010, 2016). Due to their increased use and the near complete lack of formal review of outside-of-range sentencing decisions, departures epitomize an area of critical investigation as it pertains to sentencing disparities. Federal Sentencing Departure Disparities and the Liberation Hypothesis Yet, the bulk of knowledge regarding disparities in departure decisions is based on pre- Booker data. An early study of federal sentencing departures examined the odds of receiving substantial assistance departures (also known as a 5K1.1 departure) for crack and powder cocaine offenders (Hartley, Maddan, and Spohn, 2007). Offense level, along with offender race, were the most consistent and powerful predictors of substantial assistance departure odds (Hartley et al., 2007). Black drug offenders were consistently less likely to receive a 5K1.1 departure than White drug offenders. In an intersectional analysis of race and gender, Hartley et al., (2007) found that Black males were less likely to receive a substantial assistance departure than White males in both powder and crack scenarios, regardless of whether the sentence included a mandatory minimum penalty. They found significant inter-circuit disparity in the odds of 5K1.1 departure. Johnson, Ulmer and Kramer (2008) expanded upon these findings, examining judicial departures and substantial assistance departures. Like Hartley et al. (2007), 12 Blacks, as well as Hispanics, were less likely to be granted both 5K1.1 and judicial departures. Some of the district-level variables driving the inter-district disparity in downward departure odds include caseload pressure and political context (Johnson et al., 2008). In addition, Spohn and Fornango (2009) found that females had better odds of receiving a 5K1.1 departure compared to males, and again found that Blacks had less chance of receiving a 5K1.1 departure then White offenders (Spohn & Fornango, 2009). In the single study of departures for white- collar offenders, Schanzenbach and Yaeger (2006) also found that Blacks and Hispanics were more likely to be imprisoned and less likely to receive judicial departures than Whites.2 Findings remain consistent in the few studies of downward departures that employ post- Booker data. Ulmer and Johnson (2017) found that females and Whites experienced heightened chances of judicial (i.e. judge-initiated) downward departure receival compared to males and Blacks respectively (2017). Holmes & D’Amato (2020) and Holmes et al., (2020) also found that Black and Hispanic defendants had lower odds of receiving downward departures compared to White defendants. Additionally, females were found to have greater chances of departure than males. The effect of gender on the odds of downward departure was more consistent and powerful than those of race (Holmes & D’Amato, 2020; Holmes, Feldmeyer, Kulig, 2020).3 In addition to these departure studies, a few studies have examined both the decision to depart and the decision to incarcerate, either interpreting these findings separately (Ulmer et al., 2011) or explicitly comparing extralegal disparities across the two decision points (Mustard, 2001). Mustard (2001) compared the influence of extralegal factors on sentence length for all cases versus cases sentenced within the advised range. Approximately 56% and 67% of racial 2 The racial disparities in these two outcomes were heavily influenced by the ability to pay a fine (Schanzenbach & Yaeger, 2006) 3 It is important to note that studies of the decision to incarcerate have similarly found the effect of gender to be more consistent and powerful than the effect of race (Steffensmeier et al., 2017). 13 and gendered disparities in sentence length were due to out-of-range sentences, respectively. Ulmer, Light, and Kramer (2011) provide a more recent and basic comparison of extralegal disparities associated with the decision to incarcerate and the decision to depart. In their analysis of the odds of downward departures, Black and Hispanic males had lower odds of departure receival (b=-.661, b=-.432) compared to White males (Ulmer et al., 2011). For the incarceration decision, in the years post-Gall, Black and Hispanic males had higher odds of incarceration (b=.175, b=.437) than White males. Similar to Mustard (2001), Ulmer et al. (2011) demonstrate the increased predictive power of gender and race when the outcome is downward departure odds relative to the incarceration decision. The liberation hypothesis provides one potential explanation. This hypothesis emerged from empirical findings on juror decisions. Namely, Kalven and Zeisel (1966) found that jurors were more likely to incorporate extralegal factors into sexual assault verdicts when the case was characterized as having weak or contradictory evidence. In cases where the evidence is weak, juries and judges are more apt to draw on personal stereotypes to complete their decision-making process. In a sense, court actors have more freedom, or are more at liberty, to employ their subjective perceptual shorthand because the evidence does not provide a clear conclusion upon which to substantiate a verdict. In applying these ideas to sentencing, Spohn and Cederblom (1991) argued that the influence of race on sentence outcomes would be conditioned by the seriousness of the offense, asserting that increased discretion for less serious offenses would result in larger extralegal disparities. This was confirmed for the decision to incarcerate (but not the length of the incarceration), as race was a significant factor in the decision to incarcerate only for the least serious offense type in their analysis (assault), and for cases where there were 14 mitigating conditions (i.e., cases where the offender did not injure the victim) (Spohn & Cederblom, 1991). The liberation hypothesis has also been employed in examinations of the effects of sentencing guidelines reforms on racial sentencing disparity. In one such study, Ulmer and colleagues (2011) explored the question of whether the elevated levels of discretion afforded to judges following the guideline reforms (i.e., U.S. v. Booker and Gall v. U.S.) resulted in more extralegal disparity in sentence outcomes. Their results were mixed, but generally they did not identify significantly elevated levels of extralegal disparity concerning the decision to incarcerate and sentence length subsequent to these reforms. However, they did find significant increases in extralegal disparity in substantial assistance departures post-Gall (Ulmer et al., 2011). The basic idea of the liberation hypothesis, regardless of whether it is being used to test the influence of elevated discretion due to legislative reform (Ulmer et al., 2011) or offense seriousness (Spohn & Cederblom, 1991), is that the increased discretion in the sentencing decision results in extralegal variables having a greater impact on the outcome. In sum, the liberation hypothesis asserts that extralegal sentencing disparities are greater when the amount of discretion involved in the specific sentencing decision is higher, as judges are free to rely on their biases (Frase & Mitchell, 2020). As previously described, studies of sentencing departures commonly emphasize the highly discretionary nature of departures relative to the decision to incarcerate (Engen, Gainey, Crutchfield, Weis, 2003; Holmes & D’Amato, 2020). Furthermore, studies of guideline departures have found extralegal sentencing disparities to be concentrated in outside-of-range sentences (Hartley, Maddan, Spohn, 2007; Johnson et al., 2008; Mustard, 2001; Ulmer et al., 2011). Based on the liberation hypothesis and these empirical 15 findings, it can be argued that court decisions with elevated discretion may involve greater extralegal disparities. Hypothesis 1: In this sample of fraud offenders, extralegal factors will exhibit more significant explanatory power over the decision to depart as opposed to the decision to incarcerate. Offense-Specific Stereotypes In addition to theorizing that discretion is associated with the magnitude of extralegal disparities, scholars have inferred that stereotypes may help explain the direction of racial disparities in sentencing decisions.4 In the sentencing literature, the idea was first introduced as general stereotypes by Steffensmeier, Ulmer and Kramer (1998). Sociological research preceding their study found that males who were Black were often presented in media depictions as dangerous and crime-prone (Gibbs, 1988). Steffensmeier and colleagues (1998) hypothesized that these racial stereotypes would influence court actors’ perceptions of defendants’ dangerousness and amenability to conditions of incarceration, which would then result in elevated sentencing severity for young Black males. They found young Black males experienced the highest odds of incarceration compared to both White and Black females as well as all ages of White males and older Black males (Steffensmeier et al., 1998). Since then, scholars have argued that offense specific stereotypes may play a similar role, but produce some variability by race. For example, Steffensmeier and Demuth (2000) hypothesized that ethnicity (Hispanic, non- Hispanic) would influence sentencing outcomes more heavily in drug offense cases, in part based on the notion that minorities (i.e., Hispanics, Blacks) are more readily stereotyped as drug 4 At the time of writing the author is unaware of any direct examinations of whether court actors hold general or offense-specific stereotypes based on race or gender. 16 traffickers and drug distributors. They did find enhanced racial disparities for Black and Hispanic males in the incarceration decision and the sentence length for drug offenses compared to nondrug offenses, ostensibly due to such stereotypes (Steffensmeier & Demuth, 2000). Embry and Lyons (2012) also interpreted sentencing disparities through the lens of stereotypical offense types, though the focus was on how gender would interact with offense type instead of race. Their sample was comprised of only sex offenders in order to see if females who violated gender role expectations by committing an atypical female offense (i.e. a sex offense) would be sentenced similarly to, or more severely than, men. A finding such as this would conflict with the chivalry stereotype/hypothesis which asserts that males feel obligated to protect females from negative experiences (Embry & Lyons, 2012). If the leniency generally experienced by female offenders was significantly reduced or eliminated in the case of sex offenders, then the selective chivalry hypothesis may be at work. This hypothesis asserts that women who violate gender role expectations by committing a sex crime will lose the typical gender-based leniency (Van Slyke & Bales, 2013). Embry and Lyons found no such effect; females sex offenders still experienced comparably shorter sentences than male sex offenders (2012). Thus, their study did not provide support for gendered offense-specific stereotypes. Instead, their findings bolstered the more general gender-based stereotype presented by the chivalry hypothesis, which asserts that females receive more lenient sentences than men, across offense types, because of a perceived obligation to protect women from harm due to patriarchal conceptualizations of women’s gender roles. The idea of stereotypic offenses has its roots in mock jury studies in which stereotypes have been directly assessed.5 Early studies compared stereotypes for “street” versus “white- 5 Mock jury studies have also found support for general stereotypes, namely that minorities and males are perceived as more violence-prone and dangerous than Whites and females (Singh & Sprott, 2017). 17 collar” offenses. For example, in one study, undergraduate students presumed that white-collar crimes were more likely to be committed by White defendants and violent crimes were more likely to be committed by Black defendants (Gordon, Michels, Nelson, 1996). In another, undergraduate students were presented with four vignette scenarios, varied by race of defendant (White and Black) and offense (embezzlement and burglary). Respondents were then asked whether they thought the offender’s behavior in each was due to disposition or situational conditions. Gordon (1990) found that dispositional attributions were made for Black burglars significantly more than for Black embezzlers. Consistent with Albonetti’s (1991) idea of causal attribution, where personal characteristics of the defendant are factored into attributions of causality, when the defendant was charged with a crime perceived as typical of their race, the juror tended to attribute the cause of the offense to the offenders’ disposition. In contrast, when the offense is not considered stereotypical of the defendant’s race, the casual attribution is situationally based (Gordon, 1990). More recent studies have largely supported the association between race-offense stereotypes and sentence outcomes. In another mock jury study of undergraduate students as jurors, Jones and Kaplan (2003) found Black embezzlers less likely to be found guilty than White embezzlers and Black auto thieves more likely to be found guilty than White auto thieves. In a similar study, Skorinko and Spellman (2013) conducted a series of experiments using undergraduate students as mock jurors to assess if students associated certain crimes with certain races and how these associations (i.e., crime stereotypes) influenced memory and sentencing outcomes. The first experiment found that students were more likely to associate Whites with white-collar crimes such as embezzlement, credit fraud, and insurance fraud, as well as hate crimes, and Blacks with burglary, auto theft, and drug-related crimes. Skorinko and Spellman 18 then investigated how these crime-specific stereotypes affected guilty verdicts and sentence length for violent and nonviolent crimes. They found that violent offense stereotypes influenced the students’ guilty verdict, but not sentence length. Specifically, students who were presented with offenders who committed a violent offense stereotypic of the offender’s race were more likely to find the offender guilty. The nonviolent offense scenarios produced a similar effect, but the effect was on the sentence length rather than the guilty verdict. In the nonviolent offense scenario, students who were presented with a nonviolent offender who committed an offense stereotypic of their race typically gave longer sentences. Because students were more likely to find violent offenders guilty and give nonviolent offenders longer sentences if the offense was stereotypic of the defendant’s race, Skorinko and Spellman’s (2013) findings support the idea that offense-specific stereotypes influence sentence outcomes. However, other mock jury studies argue that racial stereotypes of minorities are the true source of extralegal sentencing disparities. For example, Singh and Sprott (2017) find support for general racial stereotypes, rather than offense-specific ones, as the most common predictor of the sentence outcome was the perceived dangerousness of the defendant and Blacks were perceived as more dangerous regardless of offense type. Though the findings are not without challenge, findings from mock jury studies seem to indicate that white-collar offenders who are White will face elevated punitiveness because white- collar offenses are perceived as stereotypically White (Gordon, 1990; Jones & Kaplan, 2003). However, these studies are based on respondents imagining the characteristics of the defendant. Mock jurors may not have an accurate conceptualization of the typical white-collar offender processed through the sentencing system. As a result, their attributions may not reflect those made by court actors. 19 In the limited sentencing literature on such decisions, findings of extralegal disparity are very mixed. Some findings show that Black white-collar offenders are more likely to be incarcerated (Cassidy & Gibbs, 2019), or face longer sentences (Albonetti, 1998; Schanzenbach & Yaeger, 2006). In contrast, some studies find that white-collar offenders who are White receive longer sentences (Cassidy & Gibbs, 2019; Hagan & Nagel, 1982). Still others find that race has no significant influence on white-collar offenders’ odds of incarceration (Van Slyke & Bales, 2013). As for gender effects, the results are less mixed, but not without contradiction. A few studies find that male white-collar offenders are more likely to be incarcerated than female white-collar offenders (Cassidy & Gibbs, 2019; Van Slyke & Bales, 2013), and others find males face longer sentences when imprisoned (Cassidy & Gibbs, 2019; Maddan et al., 2012; Schanzenbach & Yaeger, 2006). Yet, Albonetti (1998) found that female white-collar offenders were given longer sentences than males, though this was partially due to the reduced chance of female white-collar offenders to plea bargain. Furthermore, other studies found no significant gender disparities in in the odds of imprisonment and sentence length (Holtfreter, 2013; Tillman & Pontell, 1992). These contradictory findings may be due to variation in the type of white-collar crime examined across studies. For instances, studies using federal (Maddan et al., 2012; Schanzenbach & Yaeger, 2006) as opposed to state (Cassidy & Gibbs, 2019; Van Slyke & Bales, 2013) data likely include more serious white-collar offenders, as more serious offenses are typically regulated by federal law. In addition, some studies focus on a single type of white-collar crime (Maddan et al., 2012), while others group many (potentially disparate) forms of white-collar crime together (Schanzenbach & Yaeger, 2006; Van Slyke & Bales, 2013). Meta-analyses of race and sentencing research purport that studies of sentencing disparity performed at higher 20 levels of aggregation may misrepresent findings because offense type may condition extralegal variables’ influence on sentencing outcomes (Mitchell, 2005). Some studies of white-collar crime sentencing have focused their analyses on a particular offense type, and in doing so illustrate how an offense-specific stereotype may influence the sentence outcomes of white- collar criminals (Holtfreter, 2013; Maddan et al., 2010). However, the literature is not expansive. For example, only one such study focuses primarily on fraud offenders, as this study does. Holtfreter (2013) used data collected from certified fraud examiners to analyze the influence of gender on sentence outcomes for fraud offenders.6 Gender did not significantly influence sentence outcomes (Holtfreter, 2013). In contrast, Maddan et al., (2010) found that gender did influence the sentence length of embezzlement offenders such that females were sentenced to approximately three months less than males on average. Though the data drawn from these studies is from two decades, the lack of gendered sentencing disparities for fraudsters (Holtfreter, 2013) and the presence of this disparity for embezzlement offenders illustrates how extralegal sentencing disparities may be conditioned by white-collar offense type. Though Holtfreter (2013) is informative, the overall body of sentencing literature consistently illustrates consistent gender- based sentencing disparities (Doerner & Demuth, 2010; Hartley et al., 2007; Spohn & Brennan, 2011; Steffensmeier et al., 2017). Due to the overwhelming evidence in opposition to Holtfreter (2013), it is hypothesized that gendered sentencing disparities observed in this study will reflect those found in the general sentencing literature. Hypothesis 2: Female fraud offenders will have lower odds of incarcerated than male fraud offenders. 6 5.7% of Holtfreter’s cases were related to corruption instead of fraud. 21 Hypothesis 3: Female fraud offenders will have higher odds of downward departures than male fraud offenders. In light of the dearth of sentencing disparity research focused on specific white-collar offense types, more generalized studies of the demographics of white-collar offenders in court systems may offer a clearer picture of potential racial or gender-based stereotypes among court actors. Early white-collar crime research on offenders sentenced in seven federal district courts in the 1970s indicated that approximately 71.4% and 78.5% of credit and mail fraudsters were White (Weisburd et al, 1991). Therefore, as with mock jury studies (Gordon et al., 1996; Skorinko & Spellman, 2013), court actors may presume that most white-collar offenders are White and therefore should be subject to harsher sentencing due to notions of how stereotypic offenders invoke the focal concerns. This would also be consistent with the few recent white- collar crime studies in which White offenders received harsher sentencing outcomes (Cassidy & Gibbs, 2019). Based on these findings, sentencing disparities for fraud may result in harsher sentences for White offenders. Therefore, I offer the following hypotheses: Hypothesis 4: Black fraud offenders will have lower odds of incarceration than White fraud offenders Hypothesis 5: White fraud offenders will have lower odds of receiving a downward departure compared to Black fraud offenders However, recent research into the racial composition of white-collar offenders described a gradual increase in minority representation amongst white-collar defendants over the past 50 years, particularly for white-collar offenses that are of low or mid-level complexity (Benson, Feldmeyer, Gabbidon, Chio, 2019). Comparison of the demographic characteristic of white- 22 collar criminals in 2015 to the Weisburd et al (2001) study indicates that the percentage of non- White fraud offenders has more than doubled for tax, false claims, and credit fraud, and nearly doubled for mail fraud (Benson et al., 2019). Additionally, some white-collar sentencing studies have found that Black offenders receive harsher sentence outcomes compared to similarly situated Whites (Albonetti, 1998; Schanzenbach & Yaeger, 2006). Thus, it is questionable whether court actors still stereotype fraud offenders as White. If courtroom actors have taken note of the growth of minority involvement in fraud offenses, it could be reasonably expected that they will not see fraud offenses as stereotypically White. It follows that racial sentencing disparities amongst federal fraud offenders may reflect those of the general sentencing literature focusing on street crime. As such, it is also possible that the following competing hypotheses will be supported: Hypothesis 6: Black and Hispanic fraud offenders will have higher odds of incarceration than White fraud offenders Hypothesis 7: White fraud offenders will have higher odds of receiving a downward departure compared to Black and Hispanic fraud offenders. Additionally, extant sentencing literature finds that extralegal variables have more influence over sentence outcomes when they are combined. (i.e. Race x Gender). These characteristics each have their proposed influence on sentencing outcomes, and these traits do not appear in a vacuum. Previous research has identified Black males as the most disadvantaged race-gender subgroup when it comes to odds of incarceration (Cassidy & Gibbs, 2019) and odds of downward departure (Holmes et al., 2020). 23 Hypothesis 8: Black male fraud offenders will face higher odds of incarceration than White male fraud offenders Hypothesis 9: Black male fraud offenders will face the lower odds of downward departure receipt than White male fraud offenders 24 Methodology Data The data for this study come from the United States Sentencing Commission’s (USSC) Monitoring of Federal Criminal Sentences annual series. These annual datasets contain all federal cases that are received by the United States Sentencing Commission. This study uses the file containing cases with sentencing dates falling between the 1st of October, 2015 and the 30th of September, 2016. The unit of analysis for the dataset is an individual case. The dataset contains demographic information, statutes of conviction, criminal history points, as well as information on the relative position of the sentence compared to the guideline range (Kitchens, 2019). The total number of cases included in the dataset is 67,742. Sample Because this study employs the receipt of a downward departure as a dependent variable, certain groups of cases were eliminated because they are subject to different case procedures. Cases with defendants who were not U.S. citizens are excluded because they can commit offenses citizens cannot and because deportation is a viable sanction and thus influences the odds of departure. Juvenile defendants were also excluded. Those cases with a minimum guideline sentence of zero months or a guideline maximum of 470 (i.e., life sentences) were excluded because a judge or prosecutor cannot depart downward from zero and life sentences are subject to limited availability for departures (Holmes & D’Amato, 2020). Following previous research, I also excluded cases sentenced in Guam, Puerto Rico, and Mariana Islands (Holmes, Feldmeyer, & Kulig, 2020; Johnson, Ulmer, & Kramer, 2008). Lastly, those cases where a defendant received the formal classification of ‘armed criminal career’ or ‘repeat and dangerous sex 25 offender against minors’ were excluded, as they are subject to limited eligibility for departure (United States sentencing Commission, 2015). As this study is only concerned with fraud offenses, cases against defendants convicted of other offense types were excluded. Of the initial 6,517 fraud cases, 1830 (i.e., 28%) were omitted due to case processing exclusions discussed above, resulting in a sample size of 4,687. Important to note is that only a subsample of these fraudsters received some form of downward departure (N=3,011). Measures Dependent Variables There are two dependent variables in this study. First, a dichotomous variable indicating whether the defendant was incarcerated (N=3,379) or not (N=784). The second dependent variable is an indicator variable measuring whether the defendant received a downward departure (N=2,711) or was sentenced within the guidelines range (N=1,452). Those defendants who were the recipients of upward departures were excluded from the analysis. Independent Variables The main independent variables used are the defendant’s race and sex. Race is a trichotomous measure with Blacks, Hispanics, and Whites as the values. Because the “Other” category of the dataset’s variable contained so few cases (n=265), these cases were excluded from the analysis. Whites serve as the reference group. Sex of the defendant is coded dichotomously with males functioning as the reference. Control Variables The extant literature on sentencing disparities has identified a number of variables found to influence sentence severity. This study includes two sets of control variables. One set 26 characterizes various aspects of the offense and a second set that depicts elements of the offender. The control variables that are characteristics of the offense are derived from many studies of sentencing disparity. Many of these variables are hypothesized to influence sentencing severity based on their influence on one of the three focal concerns. As one could assume, the final offense level has been found to be positively correlated with the odds incarceration and length of sentence (Steffensmeier & Demuth, 2000; Van Slyke & Bales, 2013). Thus, final offense level is controlled for in this study. Criminal history, scored on a six-point scale, is also controlled for as this is also positively related to sentence severity (Doerner & Demuth, 2010; Holmes & D’Amato, 2020; Holmes et al., 2020; Steffensmeier & Demuth, 2000; Steffensmeier et al., 2017; Van Slyke & Bales, 2013). The mode of disposition, that is, whether the case was dealt with through a guilty plea or a trial, is also controlled for. A few studies have controlled for this variable based on the assertion that the judge considers the organizational resources (i.e., the court’s time) expended on a case when deciding the sentence (Doerner & Demuth, 2010; Holmes & D’Amato, 2020; Holmes et al., 2020; Steffensmeier & Demuth, 2000; Steffensmeier et al., 2017; Van Slyke & Bales, 2013). To discourage trials, or to punish those who take up the time of the court through a trial, sentence severity is heightened for those whose cases are disposed of through trials. The last characteristic of the offense that is controlled for is the pretrial detention status. This variable has been employed as a control in previous examinations of guidelines departures (Holmes & D’Amato, 2020; Holmes et al., 2020). The second set of control variables, those pertaining to the offender, includes three variables. First, the defendant’s age, measured in years, is controlled for. Some studies have found age to be negatively associated with incarceration or downward departure odds (Doerner 27 & Demuth, 2010; Holmes et al., 2020), while others, using more parsimonious analyses of age effects, have identified a curvilinear relationship (Steffensmeier et al., 2017). Second, the education level of the offender is included as a control because previous studies have found that increases in education are associated with elevated odds of downward departure (Holmes & D’Amato, 2020; Holmes et al., 2020). Third, the number of dependents the defendant is responsible for is controlled using a dummy variable (0 = no dependents, 1 = one or more dependents). The number of dependents has been identified in previous research as potentially influencing a judge’s consideration of the consequences of the sentence on the offender and their family (Holmes et al., 2020). District-level controls. Several extant studies have identified significant inter-district disparity in criminal sentencing (Johnson, Ulmer, Kramer, 2008; Ulmer & Johnson, 2004; Ulmer & Johnson, 2017). Significant variation in sentencing outcomes has been found at the county (Ulmer & Johnson, 2004), district (Johnson et al., 2008), and circuit levels (Hartley et al., 2007). Additionally, the effects of case-level predictors of sentence severity have been found to vary across districts (Ulmer & Johnson, 2004). As a result of these findings, it is imperative that analyses of sentencing disparity in the federal court system account for this inter-district disparity. Though a comprehensive examination of district-level effects is beyond the scope of the current inquiry, this study will test whether there are significant variations in extralegal sentencing disparities by estimating a multi-level model with cases nested at the district level. Analytic Procedure Two multivariate regression models will be estimated to examine the direct effects of race and gender on (1) the odds of being sentenced to incarceration and (2) the odds of being granted a downward departure. All cases with missing data on any of the variables included 28 across models were removed. This procedure resulted in a loss of 524 cases or 12.6% of the sample (n = 4687). Prior to estimating the regression models for extralegal disparity, a multi- level model, nested at the district level, was constructed to test for inter-district disparity for the two dichotomous outcome variables. Since significant variation at the district level was observed, multi-level models were employed in the subsequent logistic regressions examining racial and gendered sentencing disparities to account for this variation. 29 Findings Descriptive Statistics Table 1 presents the descriptive statistics of the sample of fraud offenders. In order to make comparisons between the two multivariate models, only those cases which had no missing values on any variables included were selected for the sample. 81.2% of the cases in the sample were incarcerated. A smaller majority (65.1%) of the cases received downward departures. Turning now to the racial and gendered composition of the sample. 49.6% of the cases involved a defendant who was White. Blacks and Hispanics comprised 37.5% and 12.9% of the sample respectively. Males comprised 66.9% of the sample. Table 1 also presents characteristics of the offense (legal) and offender (extralegal) used as control variables in this study. The final offense level had an average of 17.17, with scored of 11.1 to 24.4 falling within one standard deviation. The average criminal history score was 1.84 out of 6. Approximately 93.9% of cases were dealt with through plea agreement. The last offense-related control variable, pretrial detention status, showed that 75.6% of the sample was released on bail, bond, or on their own recognizance prior to sentencing. Turning to the offender- related control variables, the average age of the sample was 43.66, reflecting the tendency of white-collar offenders to be older than perpetrators of violent crimes. The educational attainment of the sample was relatively evenly distributed. The smallest portion of the sample (12.8%) had not graduated high school. In contrast 28% had finished high school, 35% had attended some college and 24.3% had graduated college. Lastly, 56,2% of defendants had 1 or more dependents. 30 Table 1. Descriptive Statistics of Convicted Federal Fraud Offenders Standard Min- N % Mean Deviation Max Dependent Variables Decision to Incarcerate Not incarcerated 784 18.8 Incarcerated 3379 81.2 Downward departure Sentenced within range 1452 34.9 Received downward departure 2711 65.1 Independent Variables Race White 2066 49.6 Black 1561 37.5 Hispanic 536 12.9 Gender Male 2783 66.9 Female 1380 33.1 Control Variables Offense Characteristics Final Offense Level 17.75 6.606 4-41 Criminal History Score 1.84 1.43 1-6 Mode of Disposition Plea 3908 93.9 Trial 255 6.1 Pretrial Detention Status In Custody 1014 24.4 Bail/Bond or ROR 3149 75.6 Offender Characteristics Age 43.66 13.116 19-88 Education Level Less than H.S graduate 532 12.8 High School graduate 1164 28 Some college 1455 35 College graduate 1012 24.3 Dependents No Dependents 1825 43.8 1 or More Dependents 2338 56.2 31 Bivariate Statistics Tables 2, 3, and 4, display the bivariate relationships of all analysis variables to both outcome variables. Table 2 describes the relationships of categorical variables, tables 3 and 4 depict the associations of continuous variables to the incarceration outcome (table 3) and the downward departure outcome (table 4). Table 2 uses the Chi-square (x2) test statistic to describe the relationships between the categorical or ordinal independent and control variables and the two outcome variables. Tables 3 and 4 present two-tailed t-tests of means for those analysis variables that were measured at the interval level. Two tailed t-tests were selected because the outcome variables are dichotomous, and t-tests are used to assess relationships between an interval level variable and a dichotomous variable. Thus, tables 3 and 4 present the bivariate relationships of Age, measured in years, the 43-point offense level scale, and the 6-point criminal history scale, with the two outcome variables. Table 3 is dedicated to the incarceration outcome and table 4 presents the downward departure outcome for these three interval level variables. The description of bivariate associations proceeds in the order the variables appear in table 2, drawing on information from tables 3 and 4 when necessary. Table 2. Bivariate Relationships for Categorical Variables Incarceration Downward Departure Yes No (received (within Yes No departure range) % % % % Dependent Variables Decision to Incarcerate Not Incarcerated - - 13.1 86.9 Incarcerated - - 39.9 60.1 X2 201.00*** Downward Departure Within Range 92.9 7.1 - - Received Downward Departure 74.9 25.1 - - X2 201.00*** 32 Independent Variables Race Table 2 (cont’d) White 79.1 20.9 68.6 31.4 Black 83.0 17.0 61.0 39.0 Hispanic 83.8 16.2 63.6 36.4 X2 11.72** 23.51*** Gender Male 84.2 15.8 64.5 35.5 Female 75.1 24.9 66.4 33.6 X2 48.98*** 1.60 Control Variables Offense Variables Final Offense Level - - - - - - Criminal History Score - - - - - - Mode of Disposition Plea 80.4 19.6 65.4 34.6 Trial 93.3 6.7 61.6 38.4 X2 26.30*** 1.50 Pretrial Detention Status In Custody 99.1 0.9 44.0 56.0 Bail/Bond or ROR 75.4 24.6 71.9 28.1 X2 282.40*** 263.68*** Offender Variables Age - - - - - - Education Level Less than H.S. 82.9 17.1 56.6 43.4 High School Graduate 81.9 18.1 60.9 39.1 Some College 81.0 19.0 65.2 34.8 College Graduate 79.6 20.4 74.4 25.6 X2 2.97 64.59*** Dependents No Dependents 80.2 19.8 65.0 35.0 1 or More Dependents 81.9 18.1 65.2 34.8 X2 1.91 0.02 * p ≤ .05 ** p ≤ .01 *** p ≤ .001 Table 3. Bivariate Relationships of Continuous Variables to Incarceration Incarceration Yes No Min- Min- t Mean SD Max Mean SD Max Final Offense Level 5.817*** 18.68 6.64 4-41 13.71 4.68 4-35 Criminal History Score -14.562*** 1.95 1.51 1-6 1.36 0.891 1-6 Age 6.013*** 43.08 12.93 19-86 46.19 13.63 19-88 * p ≤ .05 ** p ≤ .01 *** p ≤ .001 33 Table 4. Bivariate Relationships of Continuous Variables to Downward Departures Downward Departure Yes No Min- t Mean SD Min-Max Mean SD Max Final Offense Level -6.485*** 18.24 6.37 4-41 16.82 6.94 4-41 Criminal History Score 9.369*** 1.68 1.32 1-6 2.14 1.57 1-6 Age -7.975*** 44.84 13.27 19-88 41.47 12.54 19-80 * p ≤ .05 ** p ≤ .01 *** p ≤ .001 As can be seen from tables 2, 3, and 4, the majority of the analysis variables have significant independent relationships with one or both outcome variables. The only variable that was insignificantly associated with both outcomes was the variable indicating whether the defendant had dependents. In table 2, a notable result is found in the relationship between the two outcome variables. The majority of those who were not incarcerated also did not receive downward departures (86.9%). This seems to indicate that departures are reserved for those cases where the advised sentence includes a period of incarceration. The great majority of those who were sentenced within range were also sentenced to incarceration (92.9%). Additionally, of those who received downward departures, the majority still were sentenced to incarceration (74.9%), perhaps illustrating that downward departures are used primarily to reduce and not eliminate periods of incarceration. Turning now to race and gender, both were significantly related to incarceration (x2 = 11.72, x2 = 48.98). Similar proportions of defendants were incarcerated across the three races, though the proportion of Whites that were incarcerated is approximately 4% less than the other two races. The distribution of gender across the incarceration outcome was less similar. The 34 proportion of males incarcerated was approximately 11% higher than the proportion of females. Race, but not gender, was significantly associated with downward departure outcomes (x2 = 23.51). Though the racial distribution is only slightly skewed, the proportion of Whites (68.6%) that received downward departures is about 5-8% higher than Blacks (61.0%) or Hispanics (63.6%). Described next are the bivariate associations of offense-related control variables to the outcome variables. The defendant’s final offense level is significantly associated with the incarceration outcome. As one can imagine, those who were incarcerated had a higher average offense level (x̅=18.68) than those who did not (x̅=13.71). Criminal history score was also significantly associated with incarceration. Those who were incarcerated had slightly higher scores (x̅=1.95) than those who were not incarcerated (x̅=1.36). Both final offense level and criminal history score are significantly related to departures (p ≤ .001, p ≤ .001). The final offense levels of defendants who received downward departures were higher (x̅=18.24) than those who did not (x̅=16.82). In contrast, the criminal history scores of those defendants who received downward departures were lower (x̅=1.68) than those who did not (x̅=2.14). Mode of disposition was significantly related to incarceration but not departures. A larger proportion of those who went to trial were incarcerated (93.3%) compared to the proportion of those who pled out and were incarcerated (80.4%). The last offense-related control variable, pretrial detention status, was significantly related to both outcomes. Nearly all defendants (99.1%) who were in custody prior to their trial were incarcerated. A smaller majority (75.4%) of those who were out of custody prior to trial were incarcerated. Turning to the offender-related control variables, the relation of age to the outcome variables is presented in tables 3 and 4. Age was significantly associated with both incarceration 35 and departures. The average age of those incarcerated (x̅=43.08) was about 3 years lower than those who were not incarcerated (x̅=46.19). The average age of those who received downward departures (x̅ =44.84) was about 3 years higher than those who did not (x̅=41.47). Thus, it seems that being older decreases one’s odds of incarceration while increasing the odds of downward departure. Returning to table 2 for the last two offender-related control variables, the defendant’s education level was significantly related to departures but not incarceration. With each increase in education level, the proportion of defendant’s who received a downward departure grew by approximately 4-5%. This incremental disparity results in the proportion of defendants who received a downward departure being approximately 19% higher for college graduates (74.4%) than for those with less than a high school diploma (56.6%). Finally, as was stated earlier, the variable measuring dependents was insignificantly associated with both outcomes. Explained Variation at Levels 1 and 2 To determine how much of the outcome variation was explainable by the level 2 variables (i.e., district case was sentenced in), the intraclass correlations for each outcome grouped by district were calculated. The intraclass correlation for the prison outcome was 0.24. For the downward departure outcome, the intraclass correlation was 0.33. Thus, the district of sentencing explained about 24% of the observed variation in the incarceration outcome and 33% of the observed variation in departure outcome. In other words, depending upon the district one was sentenced in, one could be 23.7% more or less likely to receive a sentence of incarceration, all other things equal. A similar conclusion can be drawn from the departure outcome’s intra- class correlation. The considerable ICC for both outcomes attests to the need for multilevel modeling in this study. 36 If the intra-class correlations measure the proportion of explained variation that is explained by the level 2 variables, then we can subtract the ICCs from 100 to obtain the proportion of explained variation that is explained by the level 1 variables. In the case of the incarceration outcome, the level 1 predictors were responsible for approximately 76% of the explained variation. For downward departures, 66% of the explained variation was due to level 1 variables. Multivariate Mixed Effects Logistic Regression Models For the multivariate, mixed effects models, I first estimate the independent effects of race and gender on the odds of incarceration and downward departure. After a description of the main effects models, models assessing the combined influence of race and gender are presented. Prior to describing the models, I first discuss a series of checks for multicollinearity. Multicollinearity Diagnostics In order to consider issues of multicollinearity in the multivariate models, three checks were made. First, the two bivariate correlation matrices were constructed--one for the race and gender variables and a second for the combined race-gender variables7. None of the correlations exceeded the .60 threshold employed in previous sentencing disparity research (Holmes et al., 2020). Notably, nearly all correlations were below the .5 level. The only correlation to exceed this was the relationship between the dummy variables for Whites and Blacks, only one of which is included in the model (as White is the reference category). Second, the tolerance values were examined. Tolerance values did not fall below .521. In other words, the percentage of the variance of any one predictor that cannot be explained by the other predictors never was below 7 Within the correlation matrix, some analysis variables are not measured as accurately by the Pearson’s R statistic and would be better assessed by Spearman’s Rho. Both statistics could not be incorporated into one correlation matrix, so Pearson’s R was chosen as it is the more accurate measure for the majority of analysis variables. 37 52. Thirdly, the variance inflation values (VIF) were calculated. None of the VIF values surpassed 1.918. Recommendations from previous literature suggest that VIF values greater than 2.5 may be indicative of multicollinearity for logistic models (Senaviratna & Cooray, 2019). Since the VIF values did not exceed the thresholds of multicollinearity concern for tolerance values nor variance inflation scores, it appears that multicollinearity is not a serious concern for this study. Odds of Incarceration Model 1 in table 5 illustrates the multivariate model for incarceration. Hypotheses 2, 4, and 6 are addressed by model 1. Hypothesis 2 predicts females will have lower odds of being incarcerated. Hypotheses 4 and 6 predict that minorities will either have lower or higher odds of being incarcerated compared to Whites. Table 5. Multilevel Mixed Effects Logistic Regression of Race, Gender, and Other Legal and Extralegal Variables on the Odds of Incarceration and Downward Departure Incarceration Downward Departure Model 1 Model 2 b p SE Odds b p SE Odds Ratio Ratio Independent Variables Race White (Ref) - - - - - - - - Black 0.21 0.07 0.12 1.23 -0.07 0.42 0.09 0.93 Hispanic 0.02 0.89 0.17 1.02 0.03 0.82 0.12 1.03 Gender Male (Ref) - - - - - - - - Female -0.32** 0.00 0.10 0.73 0.15 0.06 0.08 1.16 Control Variables Offense Variables Final Offense Level 0.22*** 0.00 0.01 1.25 0.03*** 0.00 0.01 1.03 Criminal History Score 0.43*** 0.00 0.06 1.53 -0.04 0.21 0.03 0.96 Mode of Disposition 38 Plea (Ref) - - - - - - - - Trial 0.25 0.39 0.29 1.28 -0.49** 0.00 0.15 0.61 Pretrial Detention Status Table 5 (cont’d) In Custody (Ref) - - - - - - - - Bail/Bond or ROR -3.50*** 0.00 0.36 0.03 1.07*** 0.00 0.09 2.93 Offender Variables Age -0.02*** 0.00 0.00 0.98 0.00 0.18 0.00 1.00 Education Level Less than H.S. 0.29 0.11 0.18 1.33 -0.31* 0.02 0.13 0.74 High School Graduate 0.06 0.65 0.14 1.07 -0.21 0.06 0.11 0.81 Some College 0.10 0.42 0.13 1.11 -0.17 0.10 0.10 0.85 College Graduate (ref) - - - - - - - - Dependents No Dependents (Ref) - - - - - - - - 1 or More Dependents -0.04 0.66 0.10 0.96 0.00 0.97 0.07 1.00 * p ≤ .05 ** p ≤ .01 *** p ≤ .001 As can be observed, whether a fraud offender was Black, Hispanic, or White, did not have a significant influence on the odds of incarceration. Hypotheses 4 and 6 are not supported by this finding as there was no statistically significant difference in odds of incarceration based on the race of the defendant. Model 1 indicates that women have significantly lower odds of being incarcerated than male offenders. Specifically, the odds of incarceration are 28% lower for females than for males. This finding is in line with the prediction of hypothesis 2. Findings for the legal control variables are similar to many previous studies of sentencing. Specifically, final offense level (OR = 1.25) and criminal history score (OR = 1.53), are very significantly (p<.001, p<.001) related to the odds of incarceration. Those who were in 39 custody prior to sentencing had 97% (OR = 0.03) lower odds of incarceration than those who were released on bail, bond, or on their own recognizance (p <.001). For control variables related to the offender, only age was significant (p < .001) while education level and number of dependents were not. Each increase in age made a defendant have 2% lower odds of being incarcerated. Odds of Downward Departure Model 2 in table 5 provides the findings the receipt of a downward departure. Hypotheses 3, 5, and 7 are relevant for model 2. Hypothesis 3 predicted females would have higher odds of receiving downward departures than males. Hypotheses 5 and 7 predicted either elevated or reduced odds of downward departures for minorities compared to Whites. In terms of hypothesis 3, gender was not significantly related to the receipt of downward departures at the traditional .05 level (p = 0.06). However, a strict adherence to the .05 p-value is not necessarily a substantive rationale for dismissing a result as insignificant (Senaviratna & Cooray, 2019). Model 1 indicated that females had 27% lower odds of being incarcerated (OR=0.73), model 2, significance notwithstanding, indicates that females had 16% lower odds of downward departure (OR=1.16). These results demonstrate consistent leniency for females, which provides evidence to support the assertion that the marginally significant results of model 2 can be considered noteworthy. Notably, the odds coefficients cannot be compared across models, here I am just drawing attention to the consistent direction of the coefficients across models. However, these are just the results of one study, and the results of model 2 have a 6% likelihood of being due to chance. In this case – that is, the case where one result is marginally significant – it is informative to look at prior research. In past research, females have been shown to have significantly lower odds of incarceration (Doerner & Demuth, 2010; Steffensmeier et al., 40 2017) and significantly higher odds of downward departures (Holmes & D’Amato, 2020; Spohn & Brennan, 2011; Spohn & Fornango, 2009). This, prior research presents the same results as displayed by this study. It is highly unlikely that the results of all these studies are due to chance, as each demonstrated similar gender disparities that were significant at the .05 level. These findings inform my interpretation of this study’s results. While the gender disparity is technically not significant (p=0.06) in model 2 and thus, hypothesis 3 is technically rejected; this study still demonstrated leniency towards female defendants that is consistent with the prior literature. As for race effects, being Black, Hispanic, or White was not significantly associated with receiving a downward departure and thus, hypotheses 5 and 7 are not supported. Four control variables reached significance in model 2. First, an increase in final offense level was accompanied by a 3% increase in downward departure odds. Second, defendants who went to trial (rather than plea) had approximately 39% higher odds of receiving a downward departure (OR=0.61). Third, pretrial detention status, was significantly associated with downward departures (p<.001). Specifically, those defendants who were in custody prior to sentencing had 193% elevated odds (OR=2.93) of receiving a downward departure than defendants who were out on bail/bond or released on recognizance. The fourth and final control variable to reach significance in model 2 is the defendant’s level of education. Defendants with less than a high school degree had significantly (p=0.02) lower odds (OR=0.74) of downward departure receipt compared to defendants who graduated college. Analyses of Combined Race-Gender Effects As has been described, in models 1 and 2, the independent effects of race on incarceration and downward departure odds did not reach statistical significance. Based on previous findings regarding the combined impact of extralegal characteristics, an analysis of the 41 effect of the intersection between gender and race on the same outcomes was performed. The results of this analysis are presented in table 6 and are relevant for hypothesis 8 and 9. Table 6. Multilevel Mixed Effects Logistic Regression of Race-by-Gender intersections and Other Legal and Extralegal Variables on the Odds of Incarceration and Downward Departure Incarceration Downward Departure Model 3 Model 4 Odds Odds b p SE b p SE Ratio Ratio Race-Gender Intersections White Male (ref) - - - - - - White Female -0.19 0.17 0.14 0.82 -0.08 0.51 0.11 0.93 Black Male 0.30* 0.05 0.15 1.35 -0.20 0.06 0.10 0.82 Black Female -0.11 0.47 0.15 0.90 0.11 0.38 0.12 1.11 Hispanic Male 0.18 0.42 0.22 1.19 -0.14 0.36 0.15 0.87 Hispanic Female -0.38 0.09 0.23 0.68 0.27 0.13 0.18 1.30 Control Variables Offense Variables Final Offense Level 0.23*** 0.00 0.01 1.25 0.03*** 0.00 0.01 1.03 Criminal History Score 0.42*** 0.00 0.06 1.53 -0.03 0.26 0.03 0.97 Mode of Disposition Plea (Ref) - - - - - - - - Trial 0.23 0.43 0.29 1.26 -0.48** 0.00 0.15 0.62 Pretrial Detention Status In Custody (Ref) - - - - - - - - Bail/Bond or ROR -3.50*** 0.00 0.36 0.03 1.06*** 0.00 0.09 2.90 Offender Variables Age -0.02*** 0.00 0.00 0.98 0.00 0.20 0.00 1.00 Education Level Less than H.S. 0.29 0.10 0.18 1.34 -0.31* 0.02 0.13 0.74 High School Graduate 0.06 0.67 0.14 1.06 -0.20 0.06 0.11 0.82 Some College 0.10 0.45 0.13 1.10 -0.16 0.12 0.10 0.85 College Graduate - - - - - - - - Dependents No Dependents (Ref) - - - - - - - - 1 or More Dependents -0.04 0.66 0.10 0.96 -0.01 0.89 0.07 0.99 42 * p ≤ .05 ** p ≤ .01 *** p ≤ .001 Model 3 examines the effects of the intersection between race and gender on the odds of incarceration. The statistical significance and odds ratios of control variables in model 3 mirror those of model 1, the main effects model for incarceration odds. Most of the race-gender combinations are insignificantly related to incarceration; however, the relationship of Black males does reach statistical significance (p=4.6E-02). As was asserted by hypothesis 8, the odds of incarceration for Black males are 35% higher than White males. All other race-gender groupings are insignificant, and the only group to near significance was Hispanic females (p=0.09). The influence of the combined effects of gender and race on the odds of downward departure are presented in model 4. Models 4 and 2 are similar to models 3 and 1. That is, combining race and gender had nearly no effect on the associations of control variables with the outcome variable, in this case, the odds of downward departure receipt. Black male was the race- gender grouping with the most significant association, however it did not reach the 5% threshold typically adhered to (p=.06). As it was with the gender disparity difference between models 1 and 2, attention must be drawn to the substantive value of the significance level, because blind adherence to the .05 threshold is not necessarily the most substantive rationale for dismissing results, especially when they are marginally significant (Senaviratna & Cooray, 2019). The p- value of Black Males in the downward departure model was .06. The p-value of Black Males in the incarceration model was .046. This means that the downward departure disparity between Black Males and White Males (OR=0.82) is only 1.4% more likely to be due to chance than the 43 disparity in incarceration odds for these same groups (OR=1.35). Thus, Black Males do seem to be about 18% less likely to receive a downward departure compared to White Males. Therefore, hypothesis 9 is technically rejected based on significance criteria of .05, but the results are nonetheless informative. 44 Discussion This inquiry examined the effects of race and gender on federal fraud offenders’ odds of incarceration and downward departure from sentencing guidelines. In interaction analysis, the combined effects of race and gender on these outcome odds was examined. Results reveal a few key findings. First, in contrast to previous literature, race, by itself, was not significantly associated with the odds of incarceration or downward departures. In other words, Blacks and Hispanics did not have significantly different odds of incarceration or downward departure than Whites. Second, females’ odds of incarceration were about 23% less than that of males and though it was of marginal significance (p=0.06), females had approximately 16% elevated odds to receive a downward departure compared to males. This gender disparity is commonly identified in street crime sentencing research (Doerner & Demuth, 2000; Steffensmeier et al., 2017; Ulmer et al., 2011), and in white-collar crime sentencing research (Cassidy & Gibbs, 2019; Testa, 2019; Van Slyke & Bales, 2013). Third, in the sample of fraud offenders, the odds of incarceration for Black males were 35% higher than White males. Some previous studies have observed elevated odds of incarceration for the same race-gender grouping (Cassidy & Gibbs, 2019; Doerner & Demuth, 2010; Steffensmeier et al., 2017; Ulmer et al., 2011). The first hypothesis of this study asserted that race and gender would exert more influence over the decision to depart downwards from the sentencing guidelines than the decision to incarcerate. Considering each decision independently, models 1 and 2 reveal that the race of the defendant was not significantly associated with odds of incarceration or downward departure. In contrast, gender was significantly associated with the odds of incarceration and the effects of gender are marginally significant in model 2 (p=0.06). In other words, women had 45 27% lower odds of being imprisoned, and about 16% higher odds of receiving a downward departure. Though a statistical test for more rigorous comparisons between these models is not available, the interaction models are also relevant for hypothesis one. Black males had significantly higher odds of being incarcerated compared to White males (OR=1.35), and their odds of receiving a downward departure were 18% lower (OR=0.82). Additional research is needed, but these preliminary results do not seem to support hypothesis. A handful of implications can be drawn from the key findings of this study. First, because this study found race to have no independent direct effects on the odds of incarceration, it could be argued that the offense-specific stereotypes demonstrated in mock jury studies and inferred from studies of white-collar offender demographics may not be active in the minds of sentencing judges. Or, if active, court actors may be successful in separating their stereotypes from their judicial decisions. Another interpretation is possible when considering the effects of race and gender combined. The models examining the interaction of race and gender found the Black male variable to be either significant at the .05 level or near enough that dismissal of the results would be arbitrary. In previous studies, extralegal variables have demonstrated higher significance and explanatory power when combined than when examined independently (Steffensmeier et al., 2017). This elevated influence and the theoretical concept of intersectionality (Crenshaw, 2017) suggest that the effect of extralegal variables should be assessed in a format more aligned with the real-life experiences of court actors. Focal concerns perspective asserts that stereotypes of race and gender may influence sentencing. It has also been asserted, and findings concerning the intersection of race and gender support the notion, that the stereotypes associated with a certain extralegal status (e.g., gender) may be influenced by the stereotypes associated with another extralegal characteristic (e.g., race) (Steffensmeier et al., 46 2017). Thus, it may be that studies of extralegal sentencing disparity, whether approached through the use of existing data or through the administration of mock jury studies, should combine extralegal variables in their analyses to account for intersectional identities. Significant inter-district disparity was observed in the odds of both incarceration and downward departure. This is a consistent finding in the literature (Hartley, 2020; Johnson, 2004; Johnson, et al., 2008; Ulmer & Johnson, 2004; Ulmer & Johnson, 2017). One previous study found caseload pressure to be positively correlated with elevated odds of downward departure (Johnson, 2008). It was speculated that as caseload pressure goes up, the need for the expeditious processing of cases increases the use of departures (Johnson, 2008). This same study found no association between district size and downward departure odds. However, other studies have found cases in larger courts to have lower odds of incarceration (Ulmer & Johnson, 2004), and higher odds of downward judicial departures (Ulmer & Johnson, 2017). The diversity of cases sentenced and elevated autonomy of larger courts has been hypothesized as a mechanism linking such districts with more lenient sentences (Eisenstein et al., 1988). Findings of the present inquiry and these previous studies suggest avenues for future research. Inter-district variation in case-processing and sentencing outcomes outside of incarceration and downward departures has yet to be comprehensively investigated, and studies of these outcomes could help to elucidate the mechanisms behind the inter-district variation observed in this study. Should these investigations of alternative sentencing outcomes (i.e., charging decisions, bail/bond, restitution amount, plea offers) reveal significant interdistrict disparity, researchers could then set about the task of identifying specifically where amidst the sentencing process this inter-district disparity is present. If an early case processing outcome (i.e. charging decisions) is the source of inter-district disparity then findings of such disparity in 47 outcomes subsequent to this (i.e., incarceration odds), may simply be reflecting the disparity of the earlier decision. In addition to these contributions to the sentencing literature, this study, as with all studies, also has some limitations. First and foremost, the study of extralegal disparities, and the theoretical grounding of these disparities in racial and gendered stereotypes, assumes that these stereotypes are held by and actively influencing the sentencing practices of court officials. There are precious few studies which include the use of qualitative data to investigate these propositions (Steffensmeier et al., 2017; Kaiser & Spohn, 2018; Ulmer & Johnson, 2017). Future qualitative studies that examine the perceptions of court actors through a qualitative phenominological approach combining in-depth interviews with observations of court cases to examine the shared sentencing experiences of judges, prosecutors, and defense attorneys may be warranted. Such studies could help to identify whether extralegal sentencing disparities are due to appropriately individualized sentencing practices or to stereotypes. The accumulation of such research could in turn support sentencing reform. The lack of qualitative research upon which to base the theoretical assumptions of the focal concerns perspective is closely tied to a second limiting factor, namely, the lack of direct measures for focal concerns. Where can one obtain direct measures of the blameworthiness and dangerousness of the defendant, or the practical constraints and consequences of the sentence? Additionally, how can we know that these factors are actually important to court actors through quantitative examiantions of sentencing patters? Quantitative research illustrates sentencing trends and interprets them using the focal concerns theory. To evaluate the accuracy of these interpretations, qualitative research is needed to understand if these interpretations are aligned with the actual perceptions of court actors. In other words, researchers need to ask court actors 48 what factors they consider when deciding the culpability and dangerousness of the defendant, as well as what they perceive to be important factors in the consideration of the practical constraints and consequences of the sentence. A third limitation precludes the possibility of conclusively answering hypothesis one, and also restricts sentencing work in general. This limitation is the inability to compare coefficients across logistic regression models. This is partly due to the unobserved heterogeneity in the latent variable being measured across logistic regression models (Brzoska et al., 2017). Certain types of logistic regression models may be compared; those that are nested (Brzoska et al., 2017) or those with the same outcomes but with a comparison of the coefficients of different groups (i.e., males and females) (Kuha & Mills, 2020). Additionally, cross-model comparisons of coefficients belonging to ordinary least squares (OLS) regression models may be compared, so extralegal disparities in the lengths of incarceration and downward departures may be compared. However, the lack of a statistical test to compare across logistic models with different outcomes persists as a limiting factor to the results of this study. Fourth, the data used in this study come from only one year of sentencing, and therefore cannot speak to trends in sentencing. This study provides merely a piece to the puzzle that is sentencing patterns across time. Investigation of long-term trends in sentencing patterns and disparities could investigate how inter-district disparity has changed over time. A fifth limitation to this study is found in the aggregation of downward departure types. Just as reducing the types of offenses, or separating them for analysis, may produce a more accurate estimate of extralegal disparities in sentencing (Mitchell, 2005), so too may disaggregating types of departures. Perhaps judges are more lenient than federal prosecutors in their application of departures because of structural features of their occupation (Holmes et al., 49 2020). This interpretation may be evidenced by the findings of previous studies which indicate variation in the odds of departure across departure types (i.e., judicial and prosecutorial) (Holmes & D’Amato, 2020). Perhaps the lack of significant extralegal disparities in the odds of downward departure observed in this study are due to the combining of all downward departures into one category. This study is also limited by the data utilized. Those offenders who were charged under the fraud guidelines and coded as fraud offenders in the data, could have committed a wide variety of offenses to justify such a label. Thus, although this study has narrowed down the sample to a single offense type, the relationships observed in this study could be concentrated in a specific type of fraud offense. Unfortunately, because the data did not include the type of fraud offense, this was a limitation for which there was no choice but to acquiesce. 50 Conclusion This study presents an examination of the extralegal sentencing disparities for federal fraud offenders. Further quantitative and qualitative examinations of sentencing are necessary to improve upon and strengthen the theoretical frameworks currently utilized in sentencing research. This study attested to the enhanced effect of combined extralegal variables relative to their independent effects. Additionally, significant inter-district disparity was identified but not explored in this study, and this topic provides numerous avenues for future explorations of sentencing disparities, whether legal or extralegal. Qualitative studies of the stereotypes held by judges, prosecutors, and other court actors could help advance development of the theoretical frameworks utilized in this study, and in doing so provide support for these frameworks’ internal validity. 51 BIBLIOGRAPHY 52 BIBLIOGRAPHY Albonetti, C. A. (1991). An integration of theories to explain judicial discretion. Social problems, 38(2), 247-266. Albonetti, C. A. (1998). Direct and Indirect Effects of Case Complexity, Guilty Pleas, and Offender Characteristics on Sentencing for Offenders Convicted of a White-Collar Offense Prior to Sentencing Guidelines. Journal of Quantitative Criminology, 14(4), 353–378. Amrhein V, Korner-Nievergelt F, Roth T. 2017. The earth is flat (p > 0.05): significance thresholds and the crisis of unreplicable research. PeerJ 5:e3544 https://doi.org/10.7717/peerj.3544 Baumer, E. P. (2013). Reassessing and Redirecting Research on Race and Sentencing. Justice Quarterly, 30(2), 231–261. https://doi.org/10.1080/07418825.2012.682602 Benson, M. L., Chio, H. L., Feldmeyer, B., & Gabbidon, S. L. (2021). Race, ethnicity, and social change: The democratization of middle‐class crime*. Criminology, 59(1), 10–41. Bridges, G. S., & Steen, S. (1998). RACIAL DISPARITIES IN OFFICIAL ASSESSMENTS OF JUVENILE OFFENDERS: ATTRIBUTIONAL STEREOTYPES AS MEDIATING MECHANISMS. American Sociological Review, 63(4), 554–570. Cassidy, M., & Gibbs, C. (2019). Examining Sentencing Patterns and Outcomes for White- Collar and Property Crime Offenders. Victims & Offenders, 14(1), 75–95. https://doi.org/10.1080/15564886.2018.1547996 Crenshaw, K. W. (2017). On intersectionality: Essential writings. The New Press. Doerner, J. K., & Demuth, S. (2010). The Independent and Joint Effects of Race/Ethnicity, Gender, and Age on Sentencing Outcomes in U.S. Justice Quarterly, 27(1), 1–27. https://doi.org/10.1080/07418820902926197 Eisenstein, J., Flemming, R. B., & Nardulli, P. F. (1988). The contours of justice: Communities and their courts. Boston, MA: Little, Brown. Embry, R., & Lyons, P. M. (2012). Sex-Based Sentencing. Feminist Criminology, 7(2), 146–162. https://doi.org/10.1177/1557085111430214 Engen, R. L., Gainey, R. R., Crutchfield, R. D., & Weis, J. G. (2003). DISCRETION AND DISPARITY UNDER SENTENCING GUIDELINES: THE ROLE OF DEPARTURES AND STRUCTURED SENTENCING ALTERNATIVES*. Criminology, 41(1), 99–130. https://doi.org/10.1111/j.1745-9125.2003.tb00983.x 53 Frase, R. S., & Mitchell, K. L. (2019). SENTENCING GUIDELINES IN THE UNITED STATES. In C. Spohn & P. Brennan (Eds.), Handbook on Sentencing Policies and Practices in the 21st Century. (pp. 43-67) Routledge. Gordon, R. A. (1990). Attributions for Blue-Collar and White-Collar Crime: The Effects of Subject and Defendant Race on Simulated Juror Decisions1. Journal of Applied Social Psychology, 20(12), 971–983. https://doi.org/10.1111/j.1559-1816.1990.tb00385.x Gordon, R. A., Michels, J. L., & Nelson, C. L. (1996). Majority Group Perceptions of Criminal Behavior: The Accuracy of Race-Related Crime Stereotypes1. Journal of Applied Social Psychology, 26(2), 148–159. https://doi.org/10.1111/j.1559-1816.1996.tb01843.x Hartley, R. D., Maddan, S., & Spohn, C. C. (2007). Prosecutorial Discretion: An Examination of Substantial Assistance Departures in Federal Crack‐Cocaine and Powder‐Cocaine Cases. Justice Quarterly, 24(3), 382–407. https://doi.org/10.1080/07418820701485379 Holmes, B., & D’Amato, C. (2020). Judicial and prosecutorial decision-making: Assessing the effects of race, gender, and age on federal downward sentencing departures, 2013 – 2016. Journal of Crime and Justice, 43(4), 449–466. https://doi.org/10.1080/0735648X.2019.1704838 Holmes, B., Feldmeyer, B., & Kulig, T. C. (2020). Sentencing departures and focal concerns: The joint effect of race and gender on departures in United States district courts, 2014 – 2016. Journal of Crime and Justice, 43(5), 598–622. https://doi.org/10.1080/0735648X.2020.1730933 Holtfreter, K. (2013). Gender and “Other People’s Money”: An Analysis of White-Collar Offender Sentencing. Women & Criminal Justice, 23(4), 326–344. https://doi.org/10.1080/08974454.2013.821015 Johnson, B. D., Ulmer, J. T., & Kramer, J. H. (2008). THE SOCIAL CONTEXT OF GUIDELINES CIRCUMVENTION: THE CASE OF FEDERAL DISTRICT COURTS*. Criminology, 46(3), 737–783. https://doi.org/10.1111/j.1745-9125.2008.00125.x Jones, C. S., & Kaplan, M. F. (2003). The Effects of Racially Stereotypical Crimes on Juror Decision-Making and Information-Processing Strategies. Basic and Applied Social Psychology, 25(1), 1–13. https://doi.org/10.1207/S15324834BASP2501_1 Kaiser, K., & Spohn, C. (2018). A Review of Reasons for Judicial Departures in Federal Sentencing. Criminology, Criminal Justice, Law & Society, 19(2), 44–62. Kutateladze, B. L., Andiloro, N. R., Johnson, B. D., & Spohn, C. C. (2014). CUMULATIVE DISADVANTAGE: EXAMINING RACIAL AND ETHNIC DISPARITY IN PROSECUTION AND SENTENCING: CUMULATIVE DISADVANTAGE. Criminology, 52(3), 514–551. https://doi.org/10.1111/1745-9125.12047 54 Maddan, S., Hartley, R. D. (2018). Towards the Development of a Standardized Focal Concerns Theory of Sentencing. In J. T. Ulmer & M. S. Bradley (Eds.), Handbook on Punishment Decisions: Locations of Disparity (311-335). Routledge. Maddan, S., Hartley, R. D., Walker, J. T., & Miller, J. M. (2012). Sympathy for the Devil: An Exploration of Federal Judicial Discretion in the Processing of White-Collar Offenders. American Journal of Criminal Justice, 37(1), 4–18. https://doi.org/10.1007/s12103-010- 9094-y Mitchell, O. (2005). A Meta-Analysis of Race and Sentencing Research: Explaining the Inconsistencies. Journal of Quantitative Criminology, 21(4), 439–466. https://doi.org/10.1007/s10940-005-7362-7 Mustard, D. B. (2001). RACIAL, ETHNIC, AND GENDER DISPARITIES IN SENTENCING: EVIDENCE FROM THE U.S. FEDERAL COURTS. The Journal of Law and Economics, 44(1), 285–314. Nagel, I. H., & Hagan, J. L. (1982). The Sentencing of White-Collar Criminals in Federal Courts: A Socio-Legal Exploration of Disparity. Michigan Law Review, 80(7), 1427– 1465. https://doi.org/10.2307/1288555 Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36(4), 859-866. Schanzenbach, M., & Yaeger, M. L. (2006). Prison Time, Fines, and Federal White-Collar Criminals: The Anatomy of a Racial Disparity. Journal of Criminal Law and Criminology, 96(2), 757–794. Senaviratna, N. A. M. R., & Cooray, T. M. J. A. (2019). Diagnosing multicollinearity of logistic regression model. Asian Journal of Probability and Statistics, 1-9. Singh, A.-M., & Sprott, J. B. (2017). Race Matters: Public Views on Sentencing . Canadian Journal of Criminology and Criminal Justice, 59(3), 285–312. https://doi.org/10.3138/cjccj.2016.E26 Skorinko, J. L., & Spellman, B. A. (2013). Stereotypic Crimes: How Group-Crime Associations Affect Memory and (Sometimes) Verdicts and Sentencing. Victims & Offenders, 8(3), 278–307. https://doi.org/10.1080/15564886.2012.755140 Spohn, C., & Brennan, P. K. (2011). The Joint Effects of Offender Race/Ethnicity and Gender on Substantial Assistance Departures in Federal Courts. Race and Justice, 1(1), 49–78. https://doi.org/10.1177/2153368710396228 55 Spohn, C., & Cederblom, J. (1991). Race and disparities in sentencing: A test of the liberation hypothesis. Justice Quarterly, 8(3), 305–327. https://doi.org/10.1080/07418829100091071 Spohn, C., & Fornango, R. (2009). U.S. ATTORNEYS AND SUBSTANTIAL ASSISTANCE DEPARTURES: TESTING FOR INTERPROSECUTOR DISPARITY. Criminology, 47(3), 813–846. https://doi.org/10.1111/j.1745-9125.2009.00163.x Steen, S., Engen, R. L., & Gainey, R. R. (2005). IMAGES OF DANGER AND CULPABILITY: RACIAL STEREOTYPING, CASE PROCESSING, AND CRIMINAL SENTENCING*. Criminology, 43(2), 435–468. https://doi.org/10.1111/j.0011-1348.2005.00013.x Steffensmeier, D., & Demuth, S. (2000). Ethnicity and Sentencing Outcomes in U.S. Federal Courts: Who is Punished More Harshly? American Sociological Review, 65(5), 705. https://doi.org/10.2307/2657543 Steffensmeier, D., Painter-Davis, N., & Ulmer, J. (2017). Intersectionality of Race, Ethnicity, Gender, and Age on Criminal Punishment. Sociological Perspectives, 60(4), 810–833. https://doi.org/10.1177/0731121416679371 Steffensmeier, D., Ulmer, J., & Kramer, J. (1998). THE INTERACTION OF RACE, GENDER, AND AGE IN CRIMINAL SENTENCING: THE PUNISHMENT COST OF BEING YOUNG, BLACK, AND MALE. Criminology, 36(4), 763–798. https://doi.org/10.1111/j.1745-9125.1998.tb01265.x Sutherland, E. H. (1983). White collar crime: The uncut version. Yale University Press. Testa, A. (2019). Examining Federal Criminal Sentencing of White-Collar and Common Property Offenders: The Case of Embezzlement and Larceny. Criminal Justice Policy Review, 30(5), 681–707. https://doi.org/10.1177/0887403417709313 Tillman, R., & Pontell, H. N. (1992). IS JUSTICE “COLLAR-BLIND”?: PUNISHING MEDICAID PROVIDER FRAUD*. Criminology, 30(4), 547–574. https://doi.org/10.1111/j.1745-9125.1992.tb01116.x Ulmer, J. T., & Johnson, B. (2004). SENTENCING IN CONTEXT: A MULTILEVEL ANALYSIS. Criminology, 42(1), 137–178. https://doi.org/10.1111/j.1745- 9125.2004.tb00516.x Ulmer, J. T., & Johnson, B. D. (2017). Organizational Conformity and Punishment: Federal Court Communities and Judge-Initiated Guidelines Departures. Journal of Criminal Law and Criminology, 107(2), 253–292. Ulmer, J. T., Light, M. T., & Kramer, J. H. (2011). An alternative analysis to the USSC’s 2010 report. Criminology & Public Policy, 10(4), 1077–1118. https://doi.org/10. 1 1 1 1/j. 1745-9133.2011.00761.x 56 Van Slyke, S., & Bales, W. D. (2012). A contemporary study of the decision to incarcerate white‐collar and street property offenders. Punishment & Society, 14(2), 217–246. https://doi.org/10.1177/1462474511434437 Van Slyke, S. R., & Bales, W. D. (2013). Gender dynamics in the sentencing of white-collar offenders. Criminal Justice Studies, 26(2), 168–196. https://doi.org/10.1080/1478601X.2012.729707 Weisburd, D., Waring, E., & Chayet, E. F. (2001). White-collar crime and criminal careers. Cambridge University Press. 57