Corrections for racial disparities in law enforcement.

Author:Griffin, Christopher L., Jr.
Position:III. Empirical Analysis B. The First Correction Mechanism: Prosecutors through Conclusion, with footnotes and tables, p. 1394-1427 - North Carolina
  1. The First Correction Mechanism: Prosecutors

    The first set of outcomes we explored relates to how magistrates set pretrial release bail or bond. Recalling the four broad classes of release conditions--(1) promise to appear/custodial release/pretrial release; (2) unsecured bond; (3) cash; and (4) secured bond--we estimate Equation (1) with Bond_Class as the dependent variable and various subsets of the remaining regressors included. The ordered structure of the dependent variable means that larger odds ratios correspond to increased odds of the higher bond values. Thus, although we do not estimate the likelihood of receiving each pretrial release condition, we can say something about the relative likelihood of receiving more onerous conditions. Column 1 of Table 4 displays odds ratios and 95 percent confidence intervals from an ordered logit regression in which the dependent variable is sequenced as above with respect to perceived severity. (130) The point estimates suggest that Hispanic men are 86 percent more likely to receive a more taxing release mandate than white men and that this relationship is less pronounced for black men. White women, on the other hand, are 23 percent less likely to receive more onerous pretrial release conditions.

    Column 2 indicates that the same pattern holds for whether the release condition carries any monetary payment, and Column 3 reports differences with respect to the amount of those payments using ordinary least square regressions. (131) In total, magistrates seem to set cash payments or bond amounts $330 higher for black men than white men and add stricter qualifications, for example, collateral property, for Hispanics and black men. One reason for Hispanic arrestees receiving less lenient treatment could be the higher perceived flight risk among undocumented immigrants. Whether black males suffer from outright animus or statistical discrimination over the probability of flight or culpability is a question we cannot reliably address with these data alone. (132)

    Felony DWI charges and prior DWI convictions predict bail/bond amounts. As we expect, defendants charged with a felony DWI must overcome significant barriers to release, with bond amounts set about $8000 more than for non-felony arrestees. (133) In addition to the index offense, a prior DWI conviction also significantly increases the set bail amount by about $240. When we account for the fact that some of these conditions are cash payments versus bonds in Column 3, more restrictive pretrial release conditions lead to monotonic increases in the monetary value of the release condition.

    Despite the fact that the use of public defenders versus court-appointed counsel is exogenous to the individual defendant, (134) the decision to appear pro se or pay for representation merits some exploration. The first notable feature of Table 5 is the robustness of the odds ratios across all specifications. (135) Reading across the rows for each race/gender group, black and Hispanic men are much more likely than white defendants of either gender to "choose" public or self-representation. (136) Black men are almost twice as likely as white men not to retain private counsel, whereas white women are almost indistinguishable from their male counterparts. (137) The estimates in Column 4 include all relevant explanatory variables yet essentially tell the same story as the parsimonious model in Column 1. (138) The most interesting result is that felony DWI charges dramatically increase the likelihood of public counsel or pro se representation. Even after controlling for socioeconomic characteristics--which unsurprisingly show that higher income reduces the odds of public or pro se representation--defendants charged with the most serious offense do not seek assistance from the private bar.

    A prosecutor's decision to withdraw or never pursue charges opens the door to even more unobserved heterogeneity, primarily regarding the quality of underlying evidence or office priorities among different sorts of crimes. Nonetheless, the identity of opposing counsel, which should act as a proxy for the effort required to secure a conviction; the chance of recidivism, perhaps elevated by an existing prior DWI conviction; and the existence of concurrent charges should explain a great deal of the ADA's decision calculus. In Table 6 we observe pronounced differences among black males, Hispanic males, and white females relative to the omitted group of white males, but particularly among Hispanic male defendants. (139) Specifically, when controlling for attorney type and characteristics of the crime in the fully specified model in Column 5, Hispanic men are 58 percent more likely to see charges dropped. Increases in the fraction of adults with less than a high school diploma in the block group increases the probability of declination. With block group characteristics included, arrests of black men are no more likely to be dismissed than arrests for white men. Even in specifications excluding all other covariates, as in Column 1, the difference in declination probabilities between black and white men is small.

    Table 6 offers the first robust evidence that case processing by an officer of the court positively corrects for potential error in arrest and pretrial release patterns. Recall that Hispanic men were much more likely to be arrested at rates exceeding their underlying DWI risk than any other race/gender combination. (140) ACIS distinguishes between declinations with and without leave, that is, those for which the prosecutor may refile charges or not. Hispanic men are defendants in about 30 percent of all cases declined by prosecutors. However, 78 percent of those declinations permit the prosecutor to refile charges. The most common reason the state offers for abandoning the case, at least temporarily, is that the defendant cannot be located. (141) Therefore, some of the more onerous pretrial release conditions for Hispanic males could be justified by their greater flight risk.

    These data support a reading that, perhaps inadvertently, the system is "self-correcting" at the declination stage. Prosecutors' decisions to pursue a DWI case plausibly reflect their judgments about guilt and risk after reviewing evidence collected by law enforcement and the magistrate's initial assessment. (142) And one could infer that the choice to decline serves as a correction of both police officers' and magistrates' inaccuracies. Alternatively, higher declinations could and seem to follow from a cost-benefit analysis whereby office budget constraints or crime priorities counsel against tracking down defendants that have eluded the criminal justice system. Finally, a combination of both accuracy correction and resource constraints could operate in tandem. Given the skew toward declinations with leave among Hispanic male defendants, we infer that prosecutors are opting to preserve attention and resources on other defendants or crime categories. Presumably, some arrested Hispanic males who should not be detained initially exit the criminal justice system for legitimate reasons. Others may be guilty of DWI and still escape justice. Regardless of the underlying reasons, Tables 1, 3, and 6 provide substantial evidence that prosecutor review offsets the disproportionately high arrest rates for Hispanic males.

    Our findings on the determinants of not guilty pleas are at once fascinating and troubling. Hispanic men--and to a much lesser extent black men--are significantly (both statistically and practically) less likely to enter a not guilty plea than their white counterparts, ranging from 51 to 59 percent, depending on the specification. (143) Including other important explanatory variables does little to upset these estimates. In accordance with reasonable expectations, the propensity to plead not guilty decreases monotonically with the severity of pretrial release conditions. In other words, assuming that bail conditions somewhat relate to underlying guilt or innocence, defendants who have to pay more to secure their release are also more likely to plead guilty. After accounting for presumed attorney effort and the seriousness of the index offense, mainly through the felony charge indicator, it is difficult to pinpoint why minority men would exhibit such different pleading behavior relative to white men. The answer could lie in differential priors about fairness in the criminal justice system, which would prompt Hispanic men to take the certain outcome rather than a gamble over a higher expected punishment level. Another possibility lies in the accounts of prosecutors pressuring racial and ethnic minorities into accepting plea deals even when prosecutors sense that the probability of conviction is much less than certain. (144) All we can say for sure is that, after controlling for observable characteristics of the defendant and the crime, the differential by race--but not by gender--is robust and larger than one might naturally expect.

    Our results for not guilty pleas therefore set up a reversal of the prosecutorial correction story based on declination rates independent of the reason for declination. Once prosecutors decide to pursue a DWI conviction, disparities reemerge in terms of the probability of a guilty plea. Keeping in mind that only 11 percent of all arrests in the data lead to a not guilty plea, (145) the rest are disposed of through either guilty pleas on the stated charge or plea bargains with the ADA. Controlling for all other relevant factors, especially observable characteristics of the crime, should generate indistinguishable estimates on the race/gender variables if prosecutors are offering plea bargains or otherwise equally communicating with defendants across the race/gender groups. Yet Table 7 clearly shows that Hispanic men who remain in the system after declinations are much more likely to be convicted through a guilty plea. (146) We cannot...

To continue reading