Abstract
This article applies an econometric decomposition technique to analysis of racial group differences in incarceration and sentencing of cocaine possession offenders. The standard 2-stage modeling approach is used to analyze the incarceration decision first and then, for offenders who are incarcerated, the length of the sentence. About a third of the difference in incarceration rates between Blacks and others can be attributed to the endowments, or characteristics, of the offenders’ cases. This means that it is not possible to reject the hypothesis that there is racial bias in incarceration decisions. In contrast, for those who are incarcerated, the differences in sentence length across racial groups are attributed entirely to endowments.
Notes
1. Available at http://www.hrw.org/backgrounder/usa/race
2. The Kentucky and Jefferson County statistics are from http://factfinder.census.gov
3. The data on ethnicity in the sentencing records are relatively incomplete and identify very few offenders as Hispanic, making it impossible to study that group. The data allow only comparison of African Americans with a single other group that includes all other racial categories.
4. Black cocaine users may be more likely to be caught up in a “crime-and-cocaine street lifestyle” (CitationLockwood, Pottieger, & Inciardi, 1995, p. 231), while White users may be more likely to abuse drugs privately and less likely to simultaneously commit other crimes (the latter supposition is supported by our sample statistics on simultaneous convictions).
5. See CitationBushway, Johnson, and Slocum (2007) for a discussion of issues involved in implementing the Heckman procedure.
6. CitationDemuth (2003) finds a bias against Hispanics in pretrial release decisions.
7. Commonwealth v. Doughty, 869 S.W.2d 53 (Ky. App. 1994).
8. In our sample of 825 cocaine possession cases, 285 involved charges that were amended and 133 involved charges that were dismissed. Of the cases with amended charges, 203 involved cocaine trafficking charges amended down to cocaine possession. Of the cases with dismissed charges, 64 involved dismissed persistent felony offender (both first and second degree) charges.
9. See CitationGreene (1997, pp. 161–162) for a discussion of likelihood ratio tests.
10. See CitationGujarati (1995, pp. 262–265) for a simple explanation of this statistic.
11. A more technical treatment of the decomposition method can be found in, for example, CitationBourassa and Yin (2006).
12. Note that, in the case of the logit models, the marginal effects are not the estimated coefficients (see CitationGreene, 2002).
13. The calculations used to decompose the residual effect exclude the constant term. This is because the marginal effects of the constant terms are relatively large and overwhelm the effects of the other variables without adding any explanatory ability. (In the case of the endowment effect, the constant term has no impact because it is equal to 1 for each group.)