Corrections for racial disparities in law enforcement.

• Jurisdiction: United States
• Date: 01 April 2014
• Author: Griffin, Christopher L., Jr.

ABSTRACT

Much empirical analysis has documented racial disparities at the beginning and end stages of criminal cases. However, our understanding about the perpetuation of--and even corrections for--differential outcomes in the process remains less than complete. This Article provides a comprehensive examination of criminal dispositions using all DWI cases in North Carolina from 2001 to 2011, focusing on several major decision points in the process. Starting with pretrial hearings and culminating in sentencing results, we track differences in outcomes by race and gender. Before sentencing, significant gaps emerge in the severity of pretrial release conditions that disadvantage black and Hispanic defendants. Yet when prosecutors decide whether to pursue charges, we observe an initial correction mechanism: Hispanic men are almost two-thirds more likely to have those charges dropped relative to white men. Although few cases survive after the plea bargaining stage, a second correction mechanism arises: Hispanic men are substantially less likely to receive harsher sentences and are sent to jail for significantly less time relative to white men. The first mechanism is based, in part, on prosecutors' reviewing the strength of the evidence, but much more on declining to invest scarce resources in the pursuit of defendants who fail to appear for trial. The second mechanism seems to follow more directly from judicial discretion to reverse decisions made by law enforcement or prosecutors. We discuss possible explanations for these novel empirical results and review methods for more precisely identifying causal mechanisms in criminal justice.

TABLE OF CONTENTS INTRODUCTION I. THE THEORY AND PRACTICE OF DWI CASE DISPOSITION A. Theoretical Considerations B. DWI Case Processing in North Carolina II. DATA AND DESCRIPTIVE STATISTICS A. The ACIS Data B. Descriptive Statistics III. EMPIRICAL ANALYSIS A. The Empirical Model B. The First Correction Mechanism: Prosecutors C. The Second Correction Mechanism: Judges IV. DISCUSSION CONCLUSION INTRODUCTION

Although recognized for years, widespread concern over racial, ethnic, and income-based disparities persists in the criminal justice system. (1) These differentials arise at various stages of case processing, starting with the probability of being stopped and searched by law enforcement, and culminating in penalties assessed after conviction, irrespective of any prior plea agreement. (2) These bookend events have received ample attention in the literature. (3) But intermediate decision points--the terms of bail at arraignment, whether the prosecutor declines to charge an arrestee, the quality of legal representation, and plea agreements--are not well understood. (4) Moreover, the predominant analytical framework compares outcomes for black defendants relative only to their white counterparts with little to no discussion of the rapidly growing Hispanic population. (5)

This Article addresses both deficiencies. First, we examined the anatomy of a subset of criminal offenses from arrest to sentencing at several critical junctures, and, second, we accounted for puzzling differences between Hispanic and white defendants, as well as between Hispanic and black defendants. Our findings suggest that racial minorities (6) still fare worse than white defendants, especially at the time of arrest. (7) Even if such differences arise in law enforcement practices, we found strong evidence that the judicial process self-corrects by offsetting--or at least narrowing--these gaps as the prosecution's case develops. (8)

Our empirical analysis focuses exclusively on driving while intoxicated (DWI), also known as driving under the influence (DUI), a decidedly major public health issue. (9) As of 2009, DWI arrests were the second-most common offense across the United States. (10) State legislatures have responded to the threats posed by drunk driving through tough laws designed to deter DWI offenses and punish perpetrators. (11) The underlying assumption has been that, by enacting stricter laws, legislatures will be one step closer to accomplishing their task of reducing DWI incidence. (12) However, prior evaluations of these laws, which often used state-level panel data, have yielded mixed results with respect to the laws' effects on motor vehicle mortality, binge drinking, and DWI. (13) These inconclusive findings suggest that we cannot merely legislate our way out of the problem. After enactment, the manner in which local law enforcement implements new statutes greatly affects the realization of legislative objectives.

Our empirical understanding of the interaction between legal interventions and social behavior remains incomplete because there are many links in the causal chain between passing laws and improving safety. These factors include: (1) the extent to which police enforce the laws and investigate crimes; (2) how prosecutors and courts handle DWI arrests; (3) policies and practices of the corrections system; and (4) how people respond to sanctions and complementary policies involving substance abuse treatment. (14) Each stage can be highly complex, and most are understudied. In particular, very little is known about how prosecutors and judges handle the wide variety and volume of cases that appear before them. (15) To the extent that these issues have been studied at all, research has focused on higher-level offenses such as capital

crimes. (16) Although capital offenses are responsible for more harm per crime than DWIs, they are fewer in number. (17) Further, although a substantial amount of research has dealt with policing and arrests, comparatively little attention has been paid to downstream decisions following those same arrests. (18)

Based on the background fact that stop and arrest rates for DWI offenses are higher for black men--and even more so for Hispanic men--than their population shares would predict, (19) this Article primarily addresses two questions about racial disparities and criminal process: (1) Do decisions after arrest, both by prosecutors and the courts, significantly differ according to race and gender? and (2) Is there evidence of correction mechanisms that mitigate adverse outcomes experienced at earlier stages of criminal processing? We analyze the first question using regressions with and without controls for socioeconomic attributes that should be systematically related to defendants' demographic characteristics. Answers to the second question arise from comparing differential outcomes in successive stages of the prosecution and trial, namely, whenever all available evidence suggests errors in arrest or charging that a presiding judge later remedies. Correction of differential outcomes among DWI defendants sheds light on the roles of prosecutorial and judicial discretion in the U.S. criminal justice system. Although we detect beneficial, warranted discretion in these data, we recognize that discretion can also be used to socially undesirable and perhaps even unconstitutional ends. (20)

We investigate DWI arrests in North Carolina (NC) for important analytical reasons. First, although common, DWIs are atypical in that charges are often based on a biological test, which involves sampling a driver's breath or blood to gauge its blood alcohol concentration (BAC). (21) The implication is that evidence supporting charges and convictions is arguably more limited and straightforward than for many other crime types. Second, focusing on a single offense rather than on a wider array of offenses avoids many of the difficulties generated by unobserved heterogeneity across offense types. Third, studying one offense type across many or all states must account for jurisdiction-specific effects in, at best, crude ways--that is, using fixed effects regression models. Our exclusive focus on NC reduces jurisdictional variation to the prosecutorial and judicial districts within the state as opposed to across the country. (22) Although there are differences in severity, even for a single offense type such as DWI, such variation is small relative to offenses in the aggregate. Reducing both observed and unobserved heterogeneity in crime severity is a useful step toward addressing equality of outcomes among ostensibly similar arrestees. Fourth, the administrative data available for NC are exceptional, and although DWI laws, drinking culture, and the availability of public transportation certainly vary among the states, the underlying legal processes are very similar. Finally, one's prior convictions for drunk-driving offenses are the only criminal records that matter in sentencing for a misdemeanor DWI. Having constructed prior DWI records for every arrestee in the dataset, we can confidently attest that the data capture every objective measure that could influence sentencing outcomes. To our knowledge, no previous study has tracked outcomes for arrestees in any crime category from the initial arrest through sentencing, particularly for an administrative entity as large as a state. A host of other work has examined differential enforcement behavior by highway officers, with at least one study exclusively using data from NC. (23)

A complete overview of the criminal process permits a discussion that extends beyond identifying racial disparities. Pointing out ways in which law enforcement and prosecutors make decisions that seem to correlate more with demographic characteristics than underlying risk or true guilt assuredly matters to our understanding of how well the system functions. The most...