![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
Research has examined the role of race and ethnicity in the punishment of offenders. Narrative and meta-analytic reviews have indicated that race/ethnicity influences key sentencing outcomes, at least under certain conditions. This research relies almost exclusively on regression-based analyses for determining race and ethnicity effects. While this technique is useful, recent statistical advances may provide more accurate race/ethnicity estimates. The current study employs propensity score analysis to compare punishment outcomes across White, Black, and Hispanic offenders sentenced in US federal courts during the years 2006 through 2008. Results suggest that (a) during the in/out decision the effect of minority status is frequently smaller than that estimated by regression modeling and (b) during the sentence length decision the effect of minority status is frequently larger than that estimated by regression modeling. Consequently, the modeling strategy may produce different conclusions regarding the presence of race- and ethnic-based disparity in sentencing outcomes.
Notes
1. It should be noted that some studies of federal sentencing do not reveal direct race effects or reveal very limited effects (e.g. Hartley & Tillyer, 2012; LaFrentz & Spohn, 2006; Wu & Spohn, Citation2010). These studies tend to focus on a limited subset of federal offenders (e.g. drug cases) or a small subsample of federal districts. Since the present study differs in that it is a large scale assessment of federal sentencing, these studies are not reviewed in detail here.
2. In the purest sense, causality cannot be established in the absence of manipulability, and race/ethnicity is not able to be manipulated in the present analysis. Based on this perspective, causal inferences cannot be drawn with regard to the race/sentence relationship. Similar to Paternoster and Brame (Citation2008), the current study adopts a less restrictive view about causality as it relates to racial disparity and argues that through the creation of balanced racial groups, it is reasonable to engage in careful causal thinking (even if causal inferences cannot be established in the purest sense). Moreover, even in the absence of establishing causality, it remains useful to compare the outcomes of similarly situated offenders across racial/ethnic groups for practical purposes such as examining sentencing outcomes under sentencing guidelines.
3. Franklin (Citation2011) examined conditional race effects in a smaller subset of 28 US Districts where Native American populations were more prevalent, and as a result this study has limited generalizability. The current study offers greater generalizability by examining these interactions with data from all 90 US District Courts located in the USA (with only the exclusion of four district courts located in US territories).
4. Some studies that examine the in/out and sentence length decision include a hazard rate in the sentence length model to correct for selection bias. Bushway, Johnson, and Slocum (Citation2007) demonstrated that researchers rarely employ proper exclusion restrictions in the selection model (i.e. predictors of the in/out decision that are not used in the prediction of sentence length), which can be problematic. Since exclusion restrictions are not available in the current analysis, the advice of Solzenberg and Relles (Citation1990) is followed, and uncorrected estimates are presented.
5. This measure is naturally logged when included in the multivariate regression analyses that examine the naturally logged sentence length outcome.
6. Upward departures were captured through the use of two measures (i.e. upward departure and upward departure with Booker). Since the occurrence of these departures in the sample is relatively rare, these two measures were collapsed to form a single measure of any upward departure for use in the analyses.
7. This can be accomplished by estimating a basic logistic regression model. The current study, however, follows in the footsteps of Paternoster and Brame (Citation2008) by estimating a generalized boosted regression model. Boosted regression models have the potential to estimate more accurate propensity scores since they are impervious to the functional form of the relationship between the independent variables and the dependent variable. Boosted regression models automatically account for nonlinearity and interactions, even if the researcher does not directly specify them (for a discussion see McCaffrey et al., Citation2004).
8. It is not entirely clear, a priori, which method will produce better group balance in a given study sample. Thus, several researchers have suggested examining covariate balance produced by alternative methods and then selecting the one that does, in fact, demonstrate the best balance (e.g., Harder et al., 2010; Ho et al., 2007).
9. A 1-to1 matching approach was also used and yielded very similar findings to the 3-to-1 matching approach.
10. As an example, in the current analysis, all but seven cases in the sample of Blacks (n = 45,323) were able to be matched with comparison cases from the White sample. Similar matching results were produced for each of the additional comparisons examined in the present study. By contrast, when matching without replacement, a substantial portion of the cases in the target groups were unable to be matched to similar cases from the comparison groups (e.g. just under half of the cases from the Black group had suitable matches from the White group).
11. The standardized difference statistic is calculated by taking the difference between the means for the target (e.g. Black) and comparison group (e.g. White) and dividing it by the standard deviation for the target group.
Additional information
Notes on contributors
Travis W. Franklin
Travis W. Franklin earned his PhD in criminal justice from Washington State University and is currently an assistant professor in the College of Criminal Justice at Sam Houston State University. His recent work has appeared in Justice Quarterly, Crime & Delinquency, Criminal Justice and Behavior, and Journal of Criminal Justice. His current research focuses on understanding racial and ethnic disparity in the prosecution and sentencing of offenders in state and federal courts.