Abstract
In the last decade, immigration prosecutions in federal court have increased 165%, with immigration offenses comprising over 28% of the federal criminal caseload in 2008. Despite this increase, research has yet to fully examine the sentencing outcomes for these offenders. Exploration of sentencing outcomes for immigration offenders is particularly salient due to the racial/ethnic composition of these defendants and the documented history of disparate treatment of minorities at this decision point. To explicate these issues, we examine the sentence length of immigration cases to assess the impact of legal, extra-legal, and case-processing factors. We further disaggregate by offense type to explore if correlates of sentencing are consistent across specific categories of immigration violations. Finally, we examine southwestern border districts, which process over 70% of all cases, to assess their specific sentencing practices. Model results, theoretical implications, and avenues for future research are also discussed.
Notes
1. Several US Attorneys General have issued memorandums outlining the types of cases in which fast-track departures can be used including several criteria that must be met, some of which include: the presence of a large number of a certain type of offense where there is strain on prosecutorial and judicial resources, where the case cannot be turned over for state prosecution, and where the offense was not a crime of violence (Gorman, Citation2009). These programs vary in scope and by district.
2. Researchers often model imprisonment decisions prior to an analysis of sentence length decisions, in which case a correction for sample selection bias is suggested for inclusion in the sentence length model (Heckman, 1979). We did not follow this approach for two reasons. First, since over 97% of immigration defendants in these data received a prison sentence, we contend that sample selection bias would be minimal in this sample. Second, other researchers argue that including Heckman’s correction introduces bias (Stolzenberg & Relles, Citation1997); thus, the statistical cure is worse than the statistical disease. Therefore, we model the sentencing decisions here without this control. For a more in-depth discussion of selection bias issues related to modeling sentencing decisions, see Bushway, Johnson, and Slocum (Citation2007).
3. After logarithmic transformation, dependent variable residuals were normally distributed and the skewness statistic was within an acceptable range.
4. For a more in-depth discussion of the debate surrounding correctly specifying the presumptive sentencing model using legally relevant variables see, Ulmer and Light (Citation2009).
5. Collinearity diagnostics also reveal that all variance inflation factor scores were in acceptable ranges (1.1–1.9).
6. Defendant education was not included due to the high degree of missing data on this variable.
7. There is some debate about whether using age as a dichotomous variable is appropriate due to a loss of information. We ran additional models using age as a continuous variable; however, the significance of the variable did not change for any of the models. We chose to leave age as a dichotomous variable for two reasons: (1) the average age of defendants in the data set was 33; therefore, using a cut-off of 30 years is justified and (2) the interpretation of age becomes less meaningful when used as a continuous variable.
8. Since illegal-entry or visa overstay offenses can only be committed by those who are non-citizens (permanent residents and non-US citizens), the variable measuring citizenship status was eliminated from this model.
9. The fast-track variable was removed from this model due to a lack of variation.
10. Additional analyses (not reported but available upon request) examined the interaction between defendant gender and ethnicity. Results indicated no significant differences between the models reported here and those additional models. In general, male defendants, regardless of whether they were Hispanic or not, received longer sentences compared to female defendants (both Hispanic and non-Hispanic females).