Mandatory sentencing and racial disparity: assessing the role of prosecutors and the effects of Booker.

• Author: Starr, Sonja B.
• Position: III. The Booker Question: Does Expanding Judicial Discretion Increase Racial Disparity? through Conclusion, with footnotes, p. 39-80

3. THE BOOKER QUESTION: DOES EXPANDING JUDICIAL DISCRETION INCREASE RACIAL DISPARITY?

The discussion above illustrates the serious limitations of an empirical approach that focuses on the sentencing decision in isolation. In this Part, we apply that insight to the question that so worried Justice Stevens in his Booker dissent: has freeing judges to sentence outside the Guidelines led to an increase in unwarranted disparities? The Sentencing Commission has given the most prominent answer to this question so far, and its answer is a resounding yes. Its race findings have garnered understandable attention, because they are shocking: Booker and its progeny appear to have led to a nearly fourfold increase in racial disparity in sentencing, from 5.5% to 19.5%. (126) This was an explosive finding, and it has led to calls (spearheaded by the Commission itself) to reinstate stronger constraints on judicial discretion. (127) However, we show here that the Commission's conclusions are unfounded. Properly analyzed, there is no evidence that unexplained racial disparity in sentences has increased since Booker--much less because of Booker.

There are two core problems with the Commission's analysis of Booker-problems that also pervade the rest of the empirical literature examining the disparity consequences of sentencing law reforms. The first is that the studies estimate disparity in a very limited way-the problem discussed in Part II. In Section III.A, we explain why the "presumptive sentence" approach is a particularly poor choice for analyzing Booker's effects, and we present a simple linear trend analysis showing that when disparity is estimated using our broader method, it has not increased in the years since Booker (and may have declined). In Section III.B, we discuss an additional serious problem with the existing studies: poor causal inference strategies. Even if it were true that disparity had increased after Booker, that is, these studies provide no reason to believe Booker was the cause. In Section III.C, we introduce a method that can be used to assess causation--a regression discontinuity-style approach. In Section III.D, we present the results of this analysis of Booker's effects on sentencing as well as charging and plea-bargaining. Finally, in Section III.E, we discuss the limitations on our analysis and explain why researchers may never be able to give an entirely definitive answer to the question of Booker's effects. (128)

1. The Changing Yardstick Problem

   A subset of the sentencing disparity literature focuses on measuring changes in disparity resulting from changes to sentencing law, such as Booker. Like other sentencing disparity analyses, these studies typically control for the presumptive Guidelines sentence as well as the statutory mandatory minimum. The problem with this approach is largely explained above, but it impacts sentencing-reform studies in a slightly different way. In principle, studies focusing on changes in disparities have an advantage over those that estimate the extent of "unwarranted" disparity: the ability to ignore the possibility of stable differences between groups that the observed variables do not capture. (129) Suppose the control variables amount to only a "broken yardstick" for measuring the defendant's underlying criminal behavior-for instance, suppose the presumptive sentence variable diverges from true case severity in racially disparate ways. In a policy-change study, so long as the same broken yardstick is used before and after the policy change, one can validly estimate the policy's relative effects on different groups. This advantage is a mixed blessing: estimates of changes in disparity are less policy-relevant if we do not know whether the disparity in either the pre- or the post-period is "real." Still, not every study needs to answer every question, and research that brackets the "is this real?" question can be useful.

   However, a serious problem arises if one cannot be confident that the yardstick itself has not been affected by the policy change. Consider again the 2012 Sentencing Commission report discussed above. It found that the black-white gap rose from 5.5% before Booker to 25.2% after, and finally to 29.5% after Booker's successor cases Kimbrough and Gall. (130) Other studies have likewise found at least some increase in disparity after Booker or after Kimbrough and Gall (although not as large). (131) Below, we discuss potential confounding factors that make it very problematic to infer that these changes were caused by either Booker or Kimbrough/Gall. But let's start with a more basic question: do these numbers actually tell us that racial disparity in sentences has grown?

   In each period, the Sentencing Commission estimates sentencing disparities conditional on the presumptive sentence (likely a "broken yardstick" for the reasons discussed above), and then compares the disparities across time periods. If one were certain that racial disparities in the processes determining the presumptive sentence remained constant pre- and post-Booker, then this would be a "same broken yardstick" comparison. Whatever biases were hidden in the presumptive sentence variable would affect the estimates for both time periods similarly, so the comparison would be apples-to-apples.

   But the problem is that Booker may have replaced one broken yardstick with a different one by affecting charging, plea-bargaining, or sentencing fact-finding in racially disparate ways. In other words, cases with the same presumptive sentences may represent different actual conduct pre-and post-Booker in ways that vary by race. Sample selection bias is also a potential problem: Booker may have changed which cases are winnowed out by the "funnel" of the criminal process, such that the samples of sentenced cases before and after Booker are not fairly comparable.

   There is good reason to worry about these potential biases. One clear lesson from the legal scholarship reviewed in Part I is that the stages in the criminal justice process are interrelated. Charging, plea-bargaining, and fact-finding all occur in anticipation of and in an attempt to influence the sentencing consequences. It is not even remotely safe to assume that changes in sentencing law do not affect decision-making at those earlier stages. After all, consider what happened after the Guidelines were adopted: a drastic increase in guilty pleas, which legal scholars have (very plausibly) attributed to prosecutors' sharp increase in leverage. (132)

   There are many theoretically plausible ways decision-making prior to sentencing could have changed after Booker. For example:

   * Prosecutors might have to offer more favorable plea deals to induce guilty pleas, potentially resulting in more favorable findings of fact,

   reduced charges and presumptive sentences, and perhaps more trials. (133)

   * Prosecutors could respond to the reduction in their power to manipulate the Guidelines to control the sentence by expanding use of their other tool for constraining judges: statutory mandatory minimums.

   * Judges might become less willing to make findings of fact that diverge from the plea stipulations, because doing so is no longer necessary to achieve what they perceive as a just sentencing result--they can depart instead.

   These changes would only bias estimates of post-Booker changes to racial disparity if they had a racially disparate impact on the presumptive sentence or on the composition of the sentenced sample. (134) It is possible that this is not so, of course, but one cannot simply assume it is not so-it must be tested. However, all of the existing studies of Booker (and prior studies of the initial shift to mandatory sentencing) do assume exactly that, usually implicitly. Other studies have criticized various other aspects of the Sentencing Commission's Booker study and have reached different conclusions. But these studies too have taken the sentencing-stage-only approach, controlling either for the presumptive sentence or for something closely related (the Guidelines "base offense level"), and thus are subject to the same concern. (135)

   These studies, in short, ignore the "hydraulic discretion" theory that has dominated theoretical scholarship about sentencing reform. (136) Conversely, key aspects of the hydraulic discretion theory remain almost completely untested empirically. (137) No empirical studies have yet used case data to assess changes in disparities in charging, plea-bargaining, or sentencing fact-finding in the wake of Booker. One study surveyed federal district court judges and defense attorneys about their perceptions of whether aspects of plea-bargaining had changed. (138) However, the researchers did not evaluate these perceptions' accuracy, and the perceptions of judges and defense counsel varied quite substantially. (139)

   Just a few studies have looked at changes in charging and plea-bargaining disparities in response to earlier changes to sentencing law and policy. Wooldredge et al. found that Ohio's shift to mandatory sentencing reduced racial disparities in charge-bargaining, yet increased racial disparities in sentencing (a surprising result). (140) But the authors did not evaluate changes in initial charging, without which the results are harder to interpret. In a 1987 study of Minnesota's adoption of mandatory sentencing guidelines, Miethe did evaluate initial charging and found a small but significant increase in gender disparity and no significant change in racial disparity; plea-bargaining disparities were unchanged. (141) No studies have evaluated changes in disparities in sentencing fact-finding.

   Beyond the failure to account for pre-sentencing stages of the process, recall that the Sentencing Commission's study of Booker has an additional problem: it also controls for departure status, thereby also filtering out some of the potential disparities in the sentencing decision as well. This is an especially...