Statistical aspects of cases concerning racial discrimination in drug sentencing: Stephens v. State and U.S. v. Armstrong.

• Author: Gastwirth, Joseph L.

1. Introduction

   Statistical evidence has been accepted in a wide variety of legal cases, including trademark confusion,(1) product liability,(2) and jury and employment discrimination.(3) Recently, statistics introduced on behalf of defendants who have challenged the fairness of sentencing practices have been poorly received by courts, as compared to their acceptability in equal pay and related fair employment cases. In this article we argue that the usefulness of aggregating the results of many prosecutorial decisions into a statistical analysis may not be adequately appreciated by the legal system. This lack of appreciation may be partially due to the fact that rarely is any data set perfect; one can almost always assert that information about some potentially relevant variable is lacking or that the sample is not sufficiently large. In the context of two recent cases involving claims of racial disparities in drug offense sentencing(4) we illustrate how a fuller use of statistical techniques applied to the data might have strengthened the claims of the defendants.(5)

2. Overview of the Use of Statistics in Discrimination Cases

   Ever since the United States Supreme Court adopted statistical testing in the Castenada v. Partida(6) jury discrimination case, statistical evidence has been used in many types of cases concerning discrimination regarding race, age, or sex.(7) When a plaintiff in a civil case claims disparate treatment, or a defendant in a criminal case introduces statistical evidence as part of their prima facie case, the purpose is to show that otherwise comparable individuals of the protected class are being treated less favorably than those from the majority group.

   In Texas Dep't. of Community Affairs v. Burdine,(8) the Supreme Court outlined the order of proof needed to show disparate treatment in an employment discrimination case.(9) In Batson v. Kentucky,(10) the Court adapted this standard to criminal cases in which the defendant claims that the prosecution abused its discretion in making peremptory challenges at trial by eliminating a disproportionate number of minority venirepersons.(11) Once a defendant raises a Batson claim, presumably based on the fact that most if not all minority venirepersons were eliminated, the prosecution must advance a neutral explanation for their removal. Then the defendant is given the opportunity to demonstrate that the state's reasons are insufficient or pretextual. The trial court then must determine whether the defendant has established purposeful discrimination. This procedure is sensible, as only the prosecution possesses the information that it uses to decide which venirepersons to challenge.

   In the typical jury discrimination case, the plaintiff uses data for a number of venires to compare the number of minority jurors to its expected number, derived from the minority fraction of persons eligible for jury service. A plaintiff in a hiring discrimination case would do much the same. This technique has been called "standard deviation" analysis, as the difference between the actual number of minorities and the numbers expected if selections were randomly chosen from the eligible pool is measured in units of the sampling variability, called standard deviation.(12)

   In cases concerning the discriminatory use of peremptory challenges, Batson allowed the defendant to establish a prima facie case based solely on evidence concerning the prosecution's exercise of its peremptory challenges in his own case.(13) The Court noted that the peremptory challenge process enables those who have a mind to discriminate to do So.(14) The decision in Batson overruled Swain v. Alabama,(15) which required the defendant to demonstrate a pattern of discriminatory peremptory strikes in a reasonable number of similar cases. In Batson the Court noted that this was an extremely difficult hurdle for defendants, and essentially impossible in jurisdictions where the requisite information -- the race of the jurors -- was not preserved.(16) The Court did not state that defendants must only use information from their own case; rather, statistical data on the practices of the same prosecutor or of prosecutors from the same office or in the same system is admissible evidence in conjunction with the pattern of strikes in the defendant's trial.

   Before describing the criteria courts have used to evaluate the completeness of a statistical analysis, we note that Justice Powell's opinion in McCleskey v. Kemp(17) distinguished death-sentencing cases from employment discrimination cases with respect to the number of potentially relevant variables, and asserted that there is no common standard by which to compare and evaluate all defendants who were eligible for the death penalty.(18) Since the death penalty may be given for a wide variety of very serious offenses it is unclear that this reservation concerning statistical analysis of sentencing data should apply to a more homogeneous set of crimes.

   In practice, virtually no data set is ideal; one can almost always suggest another potentially relevant variable or question the accuracy of the measurement of a factor: for example, is education adequately measured by years of school completed or does one need the grade point average too? In the context of a regression analysis, which predicts pay as a function of appropriate factors like seniority,(19) the Court in Bazemore v. Friday(20) stated that it is not necessary for a party offering a regression to incorporate all measurable variables relating to productivity as long as the model includes the major ones. In legal terms, omissions should affect the weight or importance of the evidence rather than its admissibility. Subsequently, in Allen v. Seidman,(21) Judge Posner noted that it is easy to take "pot shots" at a statistical analysis and that the defendant needs to do more than simply raise a potential flaw.(22) The critic should show that the flaw would seriously influence the ultimate inference. A number of statistical methods have been developed for assessing the potential effect of missing data,(23) omitted variables,(24) and errors of measurement.(25) These methods can be used to assist courts in evaluating statistical evidence.

3. Stephens v. State: Racial Disparities in Prosecutorial

   Requests for Life Sentences for Repeat

   Drug Offenders

   1. THE MAJORITY OPINION

      Mr. Stephens challenged the constitutionality of a Georgia law(26) which provided for mandatory life imprisonment upon a second conviction for the sale of a controlled substance or possession with intent to distribute a controlled substance. As the law gives prosecutors the discretion to seek a life sentence, he contended that the law was applied in a racially discriminatory manner in violation of both the United States and Georgia Constitutions.

      To support his claim, Stephens submitted both statewide and county-wide statistical data. A statewide study conducted by the Georgia Department of Corrections showed that of all persons eligible for a life sentence, only 1 of 168 whites sentenced for two or more convictions for drug sales is serving a life sentence compared to 202 out of 1219 blacks.(27) Apparently the statistical significance of this data was not calculated and submitted into evidence.(28) The court also found that 98.4% (369 of 375) of the state's prisoners serving life sentences for drug offenses were African-American, although they constituted only 27% of the state's population. In Hall County, where the defendant was convicted, all fourteen persons serving a life sentence for drug offenses were African-American.(29) The dissent noted that African-Americans form less than ten percent of the county's population, but accounted for 50% to 60% of those arrested in drug investigations.(30)

      In a divided opinion, the majority held that this statistical evidence was insufficient to support Stephens' claim of an equal protection violation under the Georgia Constitution. The opinion observed that Stephens failed to present critical evidence concerning persons eligible for life sentences under the statute in Hall County, but against whom the district attorney failed to seek the enhanced sentence.(31) The majority noted that in each judicial circuit the district attorney exercises discretion in seeking the increased sentence, so that a defendant must present some evidence addressing whether the prosecutor handling a particular case engaged in selective prosecution in order to prove a state equal protection violation.(32)

      Judge Fletcher went on to criticize the statistical study for ignoring other factors that may explain the disparity in sentencing. His opinion mentions the following possibly significant factors: the charge brought, concurrent offenses, prior offenses and sentences, the type of lawyer (retained or court-appointed), whether the defendant plead guilty, the circuit where convicted, and the defendant's legal status.(33) Furthermore, Judge Fletcher wrote, "[w]ithout more adequate information about what is happening ... we defer deciding whether statistical evidence alone can ever be sufficient to prove an allegation of discriminatory intent in sentencing under the Georgia constitution."(34) Thus, the issue of the type of proof defendants need to offer to establish differential treatment of eligible individuals and the role of statistical evidence in these cases remains unsettled.

   2. THE MINORITY OPINION

      The dissent, written by Judge Benham, described the statistics as "numbing and paralyzing,"(35) and objected to the majority opinion's failure to state that they substantiate a need for serious inquiry. After reviewing some of the major U.S. cases on jury discrimination, he noted that Batson held that an inference of discriminatory intent could be drawn from certain conduct or statistical data and recognized that the crucial information about an allegedly discriminatory decision could only come from the one who made...