Racial/Ethnic Differentials in Sentencing to Incarceration

Abstract

Few criminological topics are as controversial as the relationships between race, ethnicity, crime, and criminal justice outcomes—especially incarceration. This paper assesses whether Blacks and Hispanics are disadvantaged at the sentencing phase of the justice system and whether the findings depend on the use of traditional regression-based methods to control for legally relevant variables vs. the use of precision matching methods, which attend to potential sample selection bias that occurs when there are not exact matches for those sentenced to incarceration and non-incarceration. Analysis of the population of Florida offenders from 1994 to 2006 using both methodologies indicates that Black offenders continue to be disproportionately incarcerated compared to White or Hispanic offenders, and that Hispanic offenders were slightly more likely than White offenders to be incarcerated.

Keywords:

• incarceration
• over-representation
• disparity
• minorities
• precision matching

Notes

1. FBI arrest statistics do not currently provide data for Hispanics, thus similar comparisons cannot be made.

2. Recent research has calculated offending ratios (number of arrests/number of self-reported offenses) to examine this issue more directly (Blumstein, Cohen, Piquero, & Visher, 2010). Piquero and Brame (2008) compared self-report and official records using a large sample of serious juvenile offenders and found little evidence of racial and ethnic differences in either self-reported offending (frequency or variety) or officially-based arrests leading to a court referral in the preceding year.

3. The sample selection bias we are concerned with is not the same as that which may be a function of individuals’ penetration through criminal justice decision points and the extent to which individuals are a representative sample (e.g. Bushway, Johnson, & Slocum, 2007; Zatz & Hagan, 1985).

4. An anonymous reviewer observed that Blumstein also noted that 1/4 of the race differential in incarceration could not be explained by race differentials in arrests (and thus, assumedly, in offending).

5. Steffensmeier, Feldmeyer, Harris, and Ulmer (2011) re-visited Tonry and Melewski’s “Ongoing racial discrimination” conclusion by re-assessing the extent to which Blacks were “over-incarcerated” given their arrest percentages. They (p. 231) examined trends in Black percentages for arrest and incarceration after partialing out the Hispanic effect on in-stock prisoner statistics for 1980-2005. Finding mixed patterns of both under- and over-incarceration of Blacks, Steffensmeier et al. (2011, p. 233) report “little evidence … of aggregate increased punitiveness toward Blacks throughout the 1980-2005 period, at least as indicated by incarceration-arrest imbalances.”

6. There were 1,156,531 cases in the original sentencing guidelines file. Less than 1% of the cases (n = 7,174) were excluded from the final file of 1,149,357 cases due to missing data. The following variables had missing data; age at sentencing (n = 494), race (n = 1,993), current offense (n = 3), and sentence outcome (n = 4,684).

7. An anonymous reviewer observed that the disparate treatment of Hispanics in Florida may be less evident in Florida relative to other states because Hispanics (and specifically Cubans) have exercised considerable political clout in Florida—far more than in many other states, including those with large numbers of Hispanics.

8. The distribution of cases across these groups is available upon request.

9. Florida has mandatory sentencing laws for various violations of drug trafficking crimes. The state attorney may request that the sentencing court reduce the sentence for offenders who provide assistance to law enforcement in other cases. Data on whether the cases included in our analysis were subjected to a mandatory sentence were not available. Thus, it was not possible to determine the impact of the use of these laws on the sentencing outcomes.

10. It is important to note that precision matching and propensity score methods (PSM) are different ways of dealing with the selection bias issue. What precision matching offers is just that, precise matching on key variables. This may be especially relevant for variables such as race, sex, and especially age, which are related to offending, recidivism, and various criminal justice decisions (Nagin et al., 2009). Further, while precision matching and PSM are both designed to derive equivalent control and experimental groups to control for counterfactuals, they differ in the following respect. PSM seeks to derive equivalent control and experimental groups by matching cases based on their probability of being placed in the experimental group through the creation of propensity scores, derived typically from logistic regression models, based on the covariates available. Cases in the two groups are then matched based on these probabilities to derive presumably equivalent groups except for the treatment. The important feature of this approach is that cases are matched on the propensity score rather than explicitly on the covariates included in its estimation. This means that two matched individuals can have the same propensity score, but different values on the covariates that are indexed by it. In contrast, precision matching results in cases in the control and experimental groups that are exactly the same on all of the covariates included in the matching process.

11. We also categorized the current offense into 53 different groups, but are too cumbersome to present in Table 1. The maximum number of variables matched on was 21: sex (male/female), race/ethnicity (White, Black, Hispanic), age, current offense types-53 categories, current offense(s) seriousness points, number of prior felony convictions, number of prior prison commitments, number of prior community supervision violations, prior record guidelines points, number of prior convictions—murder/manslaughter, number of prior convictions—sexual crimes, number of prior convictions—robbery, number of prior convictions—other violence, number of prior convictions—burglary, number of prior convictions—property, number of prior convictions—drug, number of prior convictions—weapons, sentencing guidelines enhancement points, recommended prison sentence, year of sentencing, and judicial circuit.

12. When conducting precision matching using increasing numbers of variables to create equivalency across the control and experimental groups, an occurrence that appears to be counter-intuitive is that the number of cases that precision matching generates will increase or decrease when more variables are added to the precision matching variable. Here, we provide an example to explain why this occurs. Let us assume we match the prison and community control datasets based on a matched variable which includes five control variables and the number of cases in the prison group that have a unique value on the matching variable is 10 and the number of cases with the same unique value on the matching variable in the community control group is 15. The final dataset of matched cases across the two sanctioning groups will be one after merging the two groups on unique values on the match variable and retaining one of these records. When a sixth variable is added to the match variable, by definition, the same number of cases will have the same value across the two groups based on the five matching variables and one or more of the prison group cases can have the same matching variable value as one or more of the cases in the community control group. After retaining only one case from each unique value on the matching variable, the number of cases in the final dataset of matched prison and community control records will now be two. When a seventh or subsequent variables are added to the matching variable, the potential of adding additional cases to the final dataset continues, resulting in the final dataset with more cases than when fewer variables are used in the matching process. In contrast, the occurrence of a decrease in the number of matched cases as the number of matching variables increases occurs in the following circumstances. If there are no cases in the prison or community control samples that have the same value on an additional matching variable when increasing the number of matching variables by one, the cases that matched on the next lowest number of matching variables will no longer be included in the final matched dataset.

13. Because these differences are statistically significant with such a large sample, focus should be placed on substantive differences across groups, such as incarceration, which shows that Blacks have a much greater likelihood of being imprisoned than both Whites and Hispanics.

14. Although we lack the data needed to explain this finding, an anonymous reviewer speculated that it may be due to the differing political clout Hispanics hold in various counties, tied also to whether the Hispanic population in that jurisdiction is primarily Cuban (with its strong ethnic enclaves) or Dominican, Central American, or some other group. The finding may be due to the flow and type of cases in specific judicial circuits that either increase or decrease the likelihood of incarceration (e.g. more serious crimes, which tend to be followed by incarceration, may be over-represented in certain judicial districts), or to local courtroom workgroup interactions and decision-making styles. All of these interpretations need further research.

15. To address the question of whether the effect of being Black or Hispanic has unique consequences on the incarceration sentencing decision, 10 different comparisons in the sentencing outcomes across: White, Black, and Hispanic defendants are presented based on datasets created using the precision matching process described earlier through the inclusion of additional covariates on a cumulative basis to produce the resulting matched pairs of cases incarcerated and not incarcerated. While the specific cases that are retained in each of these datasets will change as a result of the inclusion of additional matching variables, the resulting comparison in the sanctioning outcomes across the race and ethnic groups will be based on cases that are identical on all of the values of the matching covariates. An occurrence that is somewhat counterintuitive in the outcome of the matching process as more control variables is included is when the number of cases increases. This occurs because one matched pair is retained when building each unique precision matching dataset based on a set number of matching criteria. As more variables are added to the matching criteria, the possibility for more cases to be retained occurs when there is at least one case in the incarceration and non-incarceration group which has at least one record with identical values on the additional matching covariates. We also conducted t-tests of the covariates across two groups: those that matched (precision) in Model #10 in Table 5 (n = 38,218) and those that did not match (n = 111,139). Due to the large sample size, there were significant differences on many of the covariates. However, results from logistic regression and precision matching yield substantively similar conclusions about the relative disadvantage that Blacks and Hispanics experience.

16. Table 5 does provide some evidence of variation in the effects of being Hispanic (and in some instances breaks in the general pattern for Blacks). Although the data do not permit a detailed investigation, we speculate that this may be a function of the relatively low percentage of the total sample that is Hispanic (8.3% of those incarcerated and 10.3% among the community supervision cases, figures not shown in tables) which could produce less stability across the 10 models compared with the relative stability in the effects among the Whites and Blacks.