Examining the Impact of Ecological Contexts on Gender Disparity in Federal Sentencing

Abstract

Gender disparity in sentencing outcomes has been well established in literature. Recent research has increasingly paid attention to social contexts within which judicial decision-making occurs. This study combines these two lines of research by dissecting the nature of gender disparity through ecological lenses. Using 2008–2010 federal sentencing data, we examine the roles of religious and political conservatism in affecting gender-based sentencing disparity. We find that religious and political conservatism reduces gender disparity, with the female discount dissipating in court communities with higher levels of religious and political conservatism. We also find that the conditioning effects of both religious conservatism and political conservatism on gender disparity further interact with race, with black female defendants more likely to be influenced by religious and political conservatism than their white counterparts. Overall, this study contributes to sentencing literature by demonstrating that gender disparity is deeply entrenched in the ecological contexts of court communities.

Keywords:

• gender disparity
• ecological contexts
• federal sentencing

Acknowledgements

We want to thank Cassia Spohn and Kevin Wright for their feedbacks on the previous draft. We also want to thank the Editor and anonymous reviewers for their constructive feedback and insights.

Notes

¹ The sole exception is Helms and Jacobs (2002) that found political conservatism amplified gender disparity such that female defendants were accorded more preferential treatment over their male counterparts in courts with higher levels of political conservatism than female defendants in courts with lower levels of political conservatism.

² One exception is a study conducted by Ulmer and his colleagues (Ulmer, Bader, & Gault, 2008). Specifically, Ulmer and colleagues found that counties with higher levels of religious homogeneity were more likely to incarcerate defendants, and that both the religious heterogeneity and percent conservative Protestant intensified the effects of offense severity on the odds of incarceration. In sum, Ulmer and colleagues demonstrated that religious context may influence court actors’ perceptions of defendants and in turn shape court outcomes in a complex way.

³ On the other hand, it is also possible that female defendants adjudicated in communities with higher levels of religious conservatism may be treated more leniently than their counterparts in less conservative communities, which may lead to a greater gender disparity in communities with higher levels of religious conservatism. This may be due to the possibility that judges in more religiously conservative communities also encounter strong pressure from the communities to protect families and to value the role of the husband as a form of informal social control. In fact, this expectation is more in line with the theoretical argument made by Kruttschnitt (1982) and Kruttschnitt and Green (1984)—that is, female defendants may be accorded leniency in formal social control because of a higher level of informal social control in their lives. If this were the case, gender disparity favoring female defendants may be amplified by religious conservatism such that the extent of gender disparity in sentencing is greater in jurisdictions with higher levels of religious conservatism.

⁴ However, it is also plausible that gender differentials in sentencing severity may be greater in court communities characterized by higher levels of political conservatism. This could be due to the fact that political conservatives typically hold more traditional paternalistic views, and conservative political climates often reinforce the traditional gender roles by not endorsing policies in support of gender equality (Vieraitis et al., 2007). In addition, conservative political ideology is more likely to define women as the primary caretakers and puts more emphasis on the caregiving role than breadwinning role in maintaining family unity (Daly, 1987a, 1987b). These sentiments may manifest themselves in chivalrous attitudes among criminal justice actors in formal social control settings, and the paternalistic culture may prefer informal social control to be imposed by the husband and family over formal social control. If this were the case, gender disparity favoring female defendants may be amplified by political conservatism such that the extent of gender disparity in sentencing is greater in jurisdictions with higher levels of political conservatism.

⁵ Further, Viglione, Hannon, and DeFina (2011) found that black women deemed to have a lighter skin tone received more lenient prison sentences and served less time behind the bar than those with a darker skin tone.

⁶ We thank an anonymous reviewer for bringing this important issue to our attention.

⁷ However, it is not entirely impossible that white women may be more disadvantaged as compared to minority women because white women may be held to a higher standard. We thank an anonymous reviewer for suggesting this possibility.

⁸ Although we focus on the effects of religious and political conservativism on gender disparity in federal court sentencing, several differences between state and federal sentencing are worth mentioning. First, state courts have “general jurisdiction” and process all the cases not specifically selected for federal courts. As a result, state courts process the majority of all the civil and criminal cases that individual citizens are most likely to be involved in. By contrast, federal courts have “limited jurisdiction” and can only process cases in special circumstances, such as cases involving the U.S. Government, the U.S. Constitution or other federal laws and cases involving disputes between two parties not from the same state or country. Second, given that federal court cases are processed under federal sentencing guidelines that score a defendant’s offenses and criminal history with the mandatory minimum requirement, there may be a severity gap between federal and state sentencing, and sentences meted out by federal courts tend to be harsher than those meted out by state courts, including those for the overlapping crimes such as narcotics and firearms crimes (Wright, 2006). In 2009, for example, approximately 90% of convicted defendants in federal courts received a prison sentence and about seven percent were sentenced to probation (U.S. Sentencing Commission, 2009). However, 73% of convicted defendants in state courts were sent to jail (37%) or prison (36%), and 27% were sentenced to non-custodial sanctions in 2009 (Reaves, 2013). In terms of length of prison sentences, federal courts on average sentenced 56.4 months (median = 33 months) versus 52 months by state courts (median = 30 months). Third, whereas state court judges are selected by election and/or appointment for a given number of years or for life, federal court judges are nominated by the President and confirmed by the Senate. As a result, federal judges may be less influenced by public opinion than state court judges (Norphoth, Segal, Mishler, & Sheehan, 1994). That said, some studies found that the attitudes of some Supreme Court Justices changed, intentionally or not, in response to long-term shifts in the public mood or social forces (Epstein & Martin, 2010; Giles, Blackstone, & Vinig, 2008; Mishler & Sheehan, 1993). Overall, given these differences between state and federal sentencing, future research may want to examine the effects of religious and political conservatism on gender disparity in state courts and assess if these effects are more pronounced in state courts than federal courts.

⁹ As correctly noted by an anonymous reviewer, more recent federal sentencing data (i.e. 2013) are currently available. We used the 2008–2010 data, however, because work on this paper started several years ago and before the more recent data became available. Although no important changes have occurred in federal sentencing between 2010 and 2013, political and religious contexts may have changed over time. Thus, it may be a fruitful line of inquiry to replicate the current findings with more recent federal sentencing data.

¹⁰ A recurring theme in empirical sentencing research concerns the correct way to model judicial decision-making in federal courts (see a recent debate for example: Ulmer et al., 2011b versus Sentencing Commission, 2010). A more conventional approach is to estimate two separate equations to model judicial decision-making: one for incarceration and the other for sentence length (e.g. Steffensmeier et al., 1993; Wheeler, Weisburd, & Bode, 1982). However, some scholars have argued that judicial decision-making, especially under sentencing guidelines schemes, is better modeled utilizing a so-called one-stage model where the outcome is sentence length and the non-incarceration cases are included as a prison sentence of zero (Bushway & Piehl, 2001). Scholars in favor of this latter approach contend that similar criteria guide judicial decision-making that influence both the in/out and sentence length decisions, as judges are instructed to first calculate a sentencing range, followed by a specific prison term under guidelines schemes (Bushway & Piehl, 2001; Starr & Rehavi, 2013; see also Paternoster, 2011). We follow this latter approach in this study.

¹¹ It should be noted that there is more than one way of classifying Protestant denominations. In the current study, we include fundamentalist and evangelical Protestants, but exclude Mormons.

¹² With regard to the treatment of missing data, we followed prior studies in federal sentencing and used listwise deletion (see Johnson et al., 2008; Kautt, 2000; Kim, Cano, Kim, & Spohn, 2016). An anonymous reviewer raised an issue concerning the missing data problem on the number of dependents, as this variable has substantial missing values and the number of dependents may be theoretically relevant when examining gender-based sentencing disparity. To address this issue, we performed an ancillary analysis with an alternative approach. Specifically, following Johnson and Betsinger (2009), we coded the missing data on the number of dependents as 0 and reran our models. We found that the findings for our variables of interest—that is, religious and political conservatism—remained identical.

¹³ Some previous studies in federal sentencing employed both the offender’s final criminal history score and presumptive sentence (see Albonetti, 2002; Johnson et al., 2008). Since the final criminal history score is one of the two main determinants of the presumptive sentence, including both final criminal history score and the presumptive sentence may do more harm than good. Further, given the way in which the final criminal history score is structured—that is, the criminal history category 1 does not differentiate a case with a zero point from a case with one point, it is important to control for criminal history. Thus, we include a dummy variable indicating whether a defendant has a criminal history, instead of final criminal history scores. That said, following an anonymous reviewer’s suggestion, we included final criminal history score in an ancillary analysis and found that the results concerning the effects of religious and political conservatism were the same.

¹⁴ An anonymous reviewer suggests that we incorporate prosecutors’ characteristics as substantial assistance departures are largely the purview of the federal prosecutors. Per this suggestion, we performed an ancillary analysis by including the variables capturing prosecutors’ characteristics that Farrell, Ward, and Rousseau (2009) utilized. This ancillary analysis indicated that the results regarding religious and political conservatism remained almost identical. For the sake of parsimony, we reported the results without controls for prosecutors’ characteristics but the results with these controls are available upon request.

¹⁵ Departure is a term used in criminal courts to refer to departing from the sentencing guidelines from the applicable sentencing guidelines range. In federal sentencing, there exist different kinds of departures. First, judicial downward departure is one of the most frequently used departures and it is reflected by 18 U.S.C. 3353(b) that articulated that the court may depart from the guidelines if there exists a mitigating circumstance. Second, substantial assistance departure is another important avenue by which defendants receive sentence reduction. Section 5K1.1 states that, upon the motion of government, a defendant could receive substantial assistance departure when he or she provides significant assistance to government in securing justice. Third, a defendant could also receive upward departure if the court finds any aggravating circumstances not adequately taken into consideration into the guidelines. In federal sentencing research, it is by now an almost established practice that one controls for the offender’s departure status in both in/out and sentence length models (e.g. Albonetti, 1997; Doerner & Demuth, 2010; Feldmeyer & Ulmer, 2011; but see Steffenmeier & Demuth, 2000; Kim et al., 2015). However, it is important to note that a group of legal scholars have criticized this practice (see Fischman & Schanzenbach, 2011; Freeborn & Hartmann, 2010; Rehavi & Starr, 2012). Specifically, in criticizing US sentencing commission’s report on Booker and racial disparity (Sentencing Commission, 2010), Starr and Rehavi (2013, pp. 19, 20) poignantly argued that “(by controlling for departure decisions), the Commission … is filtering out a key part of the sentencing decision itself … In effect, the Commission is estimating race gaps in the size of departure (and in sentence choices within the narrow Guidelines range), but filtering out whether there is a departure and, if so, in what direction. That is, to say the least, a strange choice, and it could very easily produce misleading results.” Thus, we do not control for departure decisions in our main analyses. However, given the debate on this practice, we conduct supplementary analyses with the departure status variables included and present results with and without the departure status variables in Table 4.

¹⁶ We performed a series of analyses to check whether there exist any multicollinearity issues. First, all bivariate correlations fall between −.45 and .55, which is short of creating a potential collinearity issue. Second, we also performed multicollinearity diagnostics (e.g. VIF and condition index), which revealed no further issue.

¹⁷ In separate analyses (not reported here but available upon request), we ran models without the conservative measures and the results showed that some of the district-level variables (e.g. disadvantage) were significantly associated with sentence length. Thus, it is likely that some contextual variables, such as disadvantage, may affect sentence length through political conservatism. Future research may want to investigate the complex relationship between political conservatism and other court-level covariates, which, however, is beyond the scope of this study.

¹⁸ We do not present a figure representing the finding concerning hypothesis 6 due to space limitations. This figure, available upon request, is largely similar to Figure 2.

¹⁹ An anonymous reviewer notes that the type of cases, average sentence length, and average criminal history may vary across districts and suggests we assess how these measures at district level may affect our results. Per this suggestion, we ran a set of ancillary analyses by including district-level average sentence length, average criminal history, % property cases, and % drug cases (results are available upon request). We found that when these additional district-level controls were included in the analysis, the results concerning the effects of religious and political conservatism were almost identical. Further, of all the additional district-level controls, average sentence length was the only one that was significantly related to sentence length. This finding is not surprising, given that district-level average sentence length was derived from individual-level sentence length for all the defendants processed in each district. Because the results regarding the effects of religious and political conservatism are substantively the same and multilevel models allow us to control for any case compositional difference across districts, and for the sake of parsimony, we presented results without these additional controls.

²⁰ Our supplementary analyses show that departures account for meaningful portions of the effects of gender and religious/political conservatism on sentence length. Further, similar to prior studies (e.g. Albonetti, 1997; Doerner, 2012; Doerner & Demuth, 2010; Feldmeyer & Ulmer, 2011; Johnson et al., 2008; Tillyer & Hartley, 2016; Wu & Spohn, 2010), our supplementary analyses suggest that departures have a large effect on sentence length and there is a gender effect in departure decisions. Moreover, the use of departures also likely varies across districts (see Johnson et al., 2008), thus departures may contribute to sentence variation across districts. All of these suggest that departure decisions, unlike some legally relevant variables (e.g. criminal history and offense severity), are a part and parcel of judicial decision-making. Thus, controlling for departure decisions may filter out “a key part of the sentencing decision itself” and could “produce misleading results” (Starr & Rehavi, 2013, pp. 19, 20).

²¹ In addition, sentencing literature also suggests that a primary channel by which local political conservatism affects sentencing outcomes is the local political processes that influence the selection of judges and prosecutors (Eisenstein & Jacob, 1977; Eisenstein et al., 1988). In other words, court communities with higher levels of political conservatism are more likely to have judges and other courtroom workgroup members with conservative political affiliations. Because this study controlled for a proxy variable measuring the political affiliations of the federal judges, along with the gender and race of the judges, the possibility that the impact of local political conservatism works through the local partisan strategy may be at least partially ruled out.

Additional information

Notes on contributors

Byungbae Kim

Byungbae Kim is a policy analyst working for Ministry of Justice, South Korea. His research interests include sentencing, correctional policies, punishment theories, and quantitative data analysis.

Xia Wang

Xia Wang is an Associate Professor at Arizona State University’s School of Criminology and Criminal Justice. She is involved in studies of race and ethnicity and their effects on crime and criminal justice, and the use of various analyses to test and extend criminological theories. Her work has appeared in Criminology, the Journal of Research in Crime and Delinquency, and the Journal of Quantitative Criminology.

Hyunjung Cheon

Hyunjung Cheon is a graduate student at Arizona State University’s School of Criminology and Criminal Justice. Her research interests include cross race/ethnicity and social control, criminological theories, and quantitative data analysis.