A Criminal Justice Reform Premise That Is Statistically Flawed

By James Scanlan | April 4, 2021, 8:02 PM EDT

James Scanlan
James Scanlan
There exists an essentially universal belief that generally reducing adverse criminal justice outcomes will tend to reduce (a) relative racial differences in rates of experiencing the outcomes (as commonly presented in terms of the ratio of the rate for Black individuals to that of white individuals) and (b) the proportion Black individuals make up of persons experiencing the outcomes (compared with the proportion they make up of the population).

The belief underlies the calls for defunding the police that were heard in many places over the past year. The belief also plays an important role in support for criminal justice reforms aimed at generally reducing prison populations, expanding options for pretrial release and diversion programs for defendants with low risk of recidivism, de-incentivizing traffic stops, and modifying police practices in ways that can reduce all adverse interactions between the police and the public.[1]

The belief, however, is the exact opposite of reality.

That is, when two groups differ in their susceptibility to an outcome, generally reducing the outcome tends to increase, not reduce, relative differences in rates of experiencing the outcome while reducing relative differences in rates of avoiding the outcome, i.e., experiencing the opposite outcome.

Correspondingly, reducing an outcome tends to increase the proportion the more susceptible group makes up of persons experiencing the outcome and persons avoiding the outcome. The pattern can be easily illustrated with test score data where two groups differ in the average performance on the test, as in Table 1 below.

The table is a slightly modified version of a table I used in my testimony at a December 2017 U.S. Commission on Civil Rights hearing on discipline disparities in public schools, where I attempted to explain the effects of reducing adverse discipline outcomes.[2]



The first row of the table presents a situation where test pass rates are 80% for an advantaged group and 63% for a disadvantaged group. The advantaged group's pass rate is 1.27 times, or 27% greater than, the disadvantaged group's pass rate, as shown in column 6.

But if the test cutoff is lowered to the point where the advantaged group's pass rate is 95%, assuming normal test score distributions, the disadvantaged group's pass rates would be about 87%. With the lower cutoff, the advantaged group's pass rate is only 1.09 times, or 9% greater than, the disadvantaged group's pass rate.

That lowering a cutoff tends to reduce relative differences in pass rates is well known in civil rights circles. That is why lowering cutoffs is universally considered a means of reducing the disparate impact of tests on which some groups outperform others. It is also a reason why observers generally consider stringent criteria to be especially hard on minority groups.

But the same data also show that lowering the cutoff increased the relative difference between rates of test failure.

As shown in column 7, with the higher cutoff, the disadvantaged group's failure rate was 1.85 times, or 85% greater than, the advantaged group's failure rate (37% versus 20%). But with the lower cutoff, the disadvantaged group's failure rate would be 2.6 times, or 160% greater than, the advantaged group's failure rate (13% versus 5%).

Lowering the cutoff would also increase the proportion the disadvantaged group makes up of persons who pass the test and persons who fail the test. This would hold regardless of the proportion the disadvantaged group makes up of persons who take the test.

But, for example, if the disadvantaged group makes up half of the test-takers, lowering the cutoff in the manner described above would increase the proportion the disadvantaged group makes up of persons who pass the test, from 44% to 48% (column 8), and increase the proportion the disadvantaged group makes up of persons who fail the test, from 65% to 72% (column 9).

This pattern is not peculiar to test score data or the numbers I used to illustrate it. Rather, the pattern is a function of most risk distributions and is evident in myriad types of data.

For example, data on income, credit ratings, blood pressure, folate level or body mass index all show how generally reducing the adverse outcome associated with the indicator tends to increase relative racial differences in rates of experiencing the outcome at the same time that it reduces relative racial differences in rates of avoiding the outcome.[3]

A useful illustration of the pattern with particular relevance to criminal justice outcomes may be found in data in a 2016 ProPublica analysis of a common algorithm for evaluating recidivism risk.[4]

The data show how relaxing a standard for pretrial release, while reducing relative racial differences in rates of being granted release, will tend to increase relative differences in rates of being denied release.

For example, if only persons in the decile with the least recidivism risk are granted release, the ratio of the release rate of white individuals to the release rate of Black individuals will be 2.53, while the ratio of the rate of denial of release for Black individuals to that for white individuals would be 1.25.

If standards are modified such as to grant release to all persons other than in the decile of greatest recidivism risk, the former figure would decline to 1.05 while the latter figure would increase to 2.88. The same data also show the aforementioned pattern of changes in the proportion Black defendants make up of persons granted release and denied release as release rates are generally increased and denial rates are generally decreased.[5]

One would observe a similar pattern when data on recidivism risk are used to inform decisions about eligibility for diversion programs. But in the case of diversion programs, that Black defendants are more likely to have prior convictions than white defendants ought to make it evident that programs that enable first offenders to avoid incarceration will tend to increase relative racial differences in incarceration rates.[6]

While that fact alone does not make it evident that the same programs will tend to reduce relative racial differences in rates of avoiding incarceration, such reduction in fact will typically occur.

In any case, numerous entities that tout their ability to provide guidance on analyzing criminal justice statistics in order to reduce disproportionate minority group contact with the criminal justice system invariably promote the mistaken belief that diversion programs will tend to reduce relative racial differences in incarceration rates.

Further, the reality is that general reductions in adverse criminal justice outcomes, like general reductions in public school suspensions,[7] have typically been accompanied by increases in relative racial differences in rates of experiencing the outcomes and the proportions Black individuals make up of persons experiencing the outcomes.

Such a pattern was recently observed, for example, in Missouri following governmental decisions to generally reduce traffic stops that were mistakenly premised on the belief that the reductions would reduce the proportion Black individuals make up of persons stopped.[8] As with countless other matters, however, reportage of such situations commonly suggests that the increases occurred "despite," rather than "because of," the general reductions.

That general reductions in adverse criminal justice outcomes tend to lead to increases in relative racial differences in rates of experiencing the outcomes and the proportion Black individuals make up of persons experiencing the outcomes does not mean that the reductions have increased racial disparities in any meaningful sense.

Rather, the general reductions in the outcomes are simply causing increases in certain unsound measures of the differences in the circumstances of Black and white individuals regarding the outcome and its opposite at the same time that the general reductions are reducing other unsound measures of those differences.

Nevertheless, there are many perverse consequences of leading the public to believe that reducing adverse criminal justice or other outcomes will tend to reduce racial disparities.

For one thing, when measures of disparity increase in the face of policies that the public has been led to believe should reduce those measures, observers will believe that the forces causing Black and white outcome rates to differ, including racial bias, must be increasing.

Further, law enforcement authorities and individual officers that assiduously endeavor to reduce adverse criminal justice outcomes — including the use of force and other adverse interactions with the public — increase the chances that they will be accused of discrimination.

Consider, for example, the police force or officer that is particularly skilled at de-escalation techniques, and thus will tend to have comparatively large relative racial differences in the use of force while making arrests (which may receive great attention) but comparatively small relative racial differences in avoiding the use of force while making arrests (which will receive no attention whatever).

Equally important, it will be impossible to determine whether policies, including implicit bias and other training, are increasing or decreasing racial bias,[9] and whether particular law enforcement authorities or officers are more likely to be biased than others, without understanding the ways the measures employed to quantify disparities tend to be affected by the prevalence of an outcome.

Quantifying the forces causing the outcome rates of advantaged and disadvantaged groups to differ and understanding the role of bias in those forces is task of some complexity. It will take much effort and much time for governmental authorities and scholars to master this subject sufficiently to provide useful insight into disparity issues or useful interpretations of what is actually happening in society.[10]

But it is crucial that all observers immediately come at least to understand that generally reducing adverse criminal justice outcomes does not tend to reduce (a) and (b) for the outcomes, but in fact tends to increase (a) and (b) for the outcomes.



James P. Scanlan is an attorney in Washington, D.C. 

"Perspectives" is a regular feature written by guest authors on access to justice issues. To pitch article ideas, email expertanalysis@law360.com.

The opinions expressed are those of the author(s) and do not necessarily reflect the views of the firm, its clients, or Portfolio Media Inc., or any of its or their respective affiliates. This article is for general information purposes and is not intended to be and should not be taken as legal advice.


[1] See Subramanian, Ram, et al., A Federal Agenda for Criminal Justice, Brennan Center for Justice (Dec. 11, 2020), for a range of recent reform recommendations.

[2] Scanlan, James, Measuring Discipline Disparities, Testimony U.S. Commission on Civil Rights Briefing "The School to Prison Pipeline: The Intersection of Students of Color with Disabilities" (see Table 3 at 3) (Dec. 8, 2017).

[3] Collected Illustrations subpage of Scanlan's Rule page of jpscanlan.com.

[4] Larson, Jeff, et al., How We Analyzed the COMPASS Recidivism Algorithm, ProPublica (May 23, 2016).

[5] See Recidivism Illustration subpage of the Scanlan's Rule page of jpscanlan.com.

[6] See Scanlan, James, United States Exports Its Most Profound Ignorance About Racial Disparities to the United Kingdom, Federalist Society Blog (Nov. 2, 2017).

[7] See Scanlan, James, Maryland Discipline Study Shows Usual – But Misunderstood – Effects of Policies on Measures of Racial Disparity, Gunpowder Gazette (Dec. 16, 2019). See also the subpages of Discipline Disparities page of jpscanlan.com bearing names of states, counties, or cities, as well as the Truancy Illustration subpage (which uses student truancy data to show that increasing the level of truancy necessary to warrant disciplinary action will tend to increase relative racial differences in rates of being disciplined for truancy while reducing relative racial differences in rates of avoiding discipline for truancy).

[8] See Cheney-Rice, Zak, 5 Years After Ferguson, Racial Disparities in Traffic Stops Have Gotten Worse; Scanlan, James, Usual, But Wholly Misunderstood, Effects of Policies on Measures of Racial Disparity Now Being Seen in Ferguson and the UK and Soon to Be Seen in Baltimore, Federalist Society Blog (Dec. 4, 2019).

[9] In situations where racial bias contributes to a racial disparity, reducing bias will tend to reduce all measures of racial disparity. But it will be impossible to determine whether bias has been reduced without understanding the way any general change in the prevalence of an outcome affects measures of disparity. For example, even though an implicit bias training program significantly reduces bias, (a) and (b) for pertinent adverse outcomes may well increase if there have also been general reductions in the outcomes. 

[10] See Scanlan, James, The Mismeasure of Discrimination, Faculty Workshop, University of Kansas School of Law (Sept. 20, 2013), Scanlan, James, Amicus Curiae Brief in Texas Department of Housing and Community Development, et al. v. The Inclusive Communities Project, Inc., Supreme Court No. 13-1731 (Nov. 17, 2014), and Scanlan, James, Comments for the Commission on Evidence-Based Policymaking (Nov. 14, 2016).

Hello! I'm Law360's automated support bot.

How can I help you today?

For example, you can type:
  • I forgot my password
  • I took a free trial but didn't get a verification email
  • How do I sign up for a newsletter?
Ask a question!