The Pernicious Misunderstanding of Effects of Policies on Racial Differences in Criminal Justice Outcomes
On September 12, 2017, the Sentencing Project released a “Fact Sheet: Black Disparities in Youth Incarceration.” The announcement of the release on the organization’s web site reads:
Despite long-term declines in youth incarceration, the disparity at which black and white youth are held in juvenile facilities has grown. As of 2015 African American youth are five times as likely as white youth to be detained or committed to youth facilities.
The fact sheet itself indicates that during a period from 2001 to 2015 in which overall juvenile placements in such facilities dropped by more than half, racial disparities increase in 37 states while declining in 13.
The organization had released a similarly themed study on April 1, 2016, then reporting that, as of 2013, the black youth commitment rate was four times the white rate.
In a March 29, 2017 post here titled “Racial Impact Statement Laws in New Jersey and Elsewhere” I discussed the role of another 2016 Sentencing Project study in causing New Jersey to believe it needed a racial impact statement law to ameliorate the effects of criminal justice policies on racial disparities in incarceration in the state’s prisons. The proposed legislation, which then had just been reported favorably by the New Jersey Assembly Public Safety Committee, became law in slightly modified form in July.
The Sentencing Project study, titled “The Color of Justice: Racial and Ethnic Disparities in State Prisons,” found that New Jersey, with a black incarceration rate 12.2 times the white rate, had the largest racial disparity in incarceration rates in the nation. As reflected in a Star-Ledger editorial, the finding was apparently a source of consternation in New Jersey because the state considered itself to be a leader in reducing prison populations and had incarceration rates below the national average. According to 2009 article by the director of the Sentencing Project, an earlier study by the organization, which had shown Iowa to have a black incarceration rate that was 13 times the white rate, had prompted that state, in 2008, to be the first to an to enact a law requiring racial impact statements for proposed legislation affecting sentencing. Activities of the Sentencing Project probably have a significant role in much of the support for racial impact laws around the country.
In the March 29 post I explained that discussion of the New Jersey legislation, and the legislations itself, accorded with the near universal belief – shared and promoted by the United States Department of Justice and the last President – that generally reducing adverse criminal justice outcomes will tend to reduce (a) relative (percentage) racial differences in rates of experiencing the outcomes and (b) the proportion racial minorities make up of persons experiencing the outcomes. I also explained that, as I had discussed here previously – including in “Things the President Doesn’t Know About Racial Disparities” (Aug. 5, 2016) and “Compliance Nightmare Looms for Baltimore Police Department” (Feb. 8, 2017) – reducing any outcome in fact tends to increase, not reduce, (a) and (b) as to the outcome.
The August 5, 2016 post had also explained that there may be situations in criminal justice context where generally reducing outcomes will not tend to increase (a) and (b). Some departures from the pattern I describe relate to the way racial differences vary by type of crime. For example, I assume that actions directly related to crack cocaine sentencing may well not exhibit the pattern, though I have not examined the issue carefully. The February 8, 2017 post makes similar points about the likely effects of the Baltimore Police consent decree on measures of racial disparity in adverse interactions between the police and the citizens of Baltimore.
One example of the departure from the pattern may be found in New Jersey itself. In the March 29 post I observed that although I have not seen discussion of the situation in New Jersey before the general reductions in incarceration, I assumed that prior to those reductions the black-to-white ratio was lower than after the reductions. That assumption was not correct. In fact, a table in 2016 Sentencing Project study that I had overlooked showed that the ratio was slightly higher before the general reductions in incarceration.
But such departures do not detract from the need to understand that in the criminal justice context, as in essentially every context, the belief that generally reducing an outcome will tend to reduce (a) and (b) as to the outcome is the exact opposite of reality.
The Sentencing Project’s presentation of findings in its September 2016 release as well the April 2016 release reveal both (a) that with respect to a key aspect of its mission the organization proceeds on the belief that reducing adverse criminal justice outcomes should tend to reduce relative racial differences in rates of experiencing those outcomes and (b) that, as an informed individual should expect, reducing such outcomes in fact usually increases such differences. Meanwhile, the organization continues to promote general reductions in discipline rates as a means of reducing those differences, while relying on data that suggests there is an increasing urgency to the organization’s message. Thus, we have the pernicious cycle where actions prompted by misunderstandings of data lead to beliefs that the matters warranting those actions are growing ever worse and, hence, that there is increasing need for such actions.
And in a society that is increasingly divided by perceptions about bias in the criminal justice system, those who believe that racial differences in criminal justice outcomes are substantially driven by racial bias are led falsely to believe that such bias is growing. Indeed, findings like those reported by the Sentencing Project will lead ill-informed observers to believe that something must be especially amiss given that disparities are increasing in the face or actions that should be decreasing them.
We have been observing the same thing, of course, with regard to school discipline, where the Departments of Education, Justice, and Health and Human Services have been leading the public and school administrators to mistakenly believe that relaxing standards and otherwise generally reducing adverse school discipline outcomes will tend to reduce, rather than increase, (a) and (b) as to those outcomes. See my posts here of August 24, 2017 (“Innumeracy at the Department of Education and the Congressional Committees Overseeing It”) and July 24, 2017 (“The Government’s Uncertain Path to Numeracy”).
I have discussed here my various efforts to explain this to the government, including by letter of April 13, 2017, and by letter of July 17, 2017 to the Departments of Education, Health and Human Services, and Justice. The July 17 letter impresses on the agencies the obligation at least to explain to school administrators that what the agencies had previously led those administrators to believe about the effects of generally relaxing public school discipline standards on measures of racial disparity was incorrect. Once the agencies understand the matter, the obligation is clear enough. But whether the agencies ever will understand the matter remains to be seen.
I have occasionally noted here that government policies based on beliefs that are the opposite of reality have managed to persist for decades because the misunderstanding of this subject is as pervasive in the scientific community as in the government. In that regard, by letter of October, 8, 2015, I requested that American Statistical Association (ASA) take certain actions to reform analyses of demographic differences and to explain to arms of the federal government that reducing the frequency of an adverse outcome will tend to increase, rather than reduce, relative racial and other demographic differences in rates of experiencing the outcome. While ASA was considering the matter, by letter of March 28, 2016, I requested the same actions of the Population Association of American (PAA) and Association of Population Centers (APC). PAA/APC leadership responded in a month to say that they were taking none of the requested actions because they did not prescribe statistical methods to the government or to their members.
By letter to ASA of July 25, 2016, I followed up with ASA urging it to explain to President Barrack Obama that understandings about effects of modifications of practices on measures of racial/ethnic differences in criminal justice outcomes reflected in a July 7, 2016 speech on criminal justice disparities (the subject of my August 5, 2016 post here) were incorrect. In September 2016, the ASA informed me that it believed that I was effectively highlighting the issues raised in my communications to the organization and did not see an additional role ASA could play in disseminating information on such issues.
I tend to doubt that PAA/APC leadership ever understood the issues. ASA leadership should have. But the belief that ASA could add not anything to what I was doing is hardly realistic. The organization could overnight cause to be almost universally understood things that my three decades of publications on the subject have caused only a handful of ASA members themselves to understand. In any case, the decision of the organizations to allow the government to continue to wrongly believe that reducing adverse criminal justice and other outcomes tends to reduce relative racial differences in rates of experiencing the outcomes was also a decision to allow virtually all of their members to continue to believe it. Thus, members of those organizations will continue to contribute to the misunderstandings that cloud debate of one of the most sensitive issues in American society.
As to the Sentencing Project, its failures of understanding are no more serious than those found in the United States government or the world’s leading research institutions. But the organization and many like it play an especially unfortunate role in promoting mistaken beliefs about racial disparities.