This past week, the Government Accountability Office (GAO) published its’ latest report to the House Committee on Armed Services, entitled: DoD and the Coast Guard Need to Improve their Capabilities to Assess Racial and Gender Disparities.

From that title, one might presume that the GAO was merely advising lawmakers on the military services’ inadequate bookkeeping. And, indeed, the report does spend some time highlighting the armed forces’ deficient efforts “to collect and maintain consistent information about race and ethnicity in their investigations, military justice, and personnel databases.” Report at 22. However, despite lackluster database utilization by DoD and DHS agencies, enough data was apparently present for GAO reach this core conclusion:

Our analysis of available data identified racial and gender disparities in all of the military services for service members with recorded investigations, and for four of the military services for trials in special and general courts-martial, but these disparities generally were not present in the convictions or punishments of cases. These findings suggest disparities may be limited to particular stages of the military justice process for the period covered by our analysis.

The first key graphic illustrating this disparity is found in Figure 5, on page 41 of the report, which presents the Likelihood of Recorded Investigations for Alleged Uniform Code of Military Justice Violations by Race and Gender, After Controlling for Rank and Education, Fiscal Years 2013-2017. That table shows that, depending on the service, black military members were 1.58 to 2.36 times more likely to be investigated by military criminal investigative organizations (MCIOs) as compared to other service members. Hispanic service members were 1.07 to 1.54 times more likely to be investigated, and male service members were 1.43 to 3.03 times more likely to be investigated.

The rates for prosecution in trials by special or general court-martial showed similar disparity. Figure 6, on page 43 of the report, presents the Likelihood of Trial in General and Special Courts-Martial by Race and Gender, After Controlling for Rank and Education, Fiscal Years 2013–2017. There, it is shown that black service members were 1.51 to 2 times more likely to be tried by special or general court-martial as compared to other service members. Hispanic service members were 1.29 to 1.42 times more likely to be tried by such courts, and male service members were 2.56 to 5.75 times more likely.

Of interest, GAO found that the racial, ethnic, and gender disparities in the rate of prosecution by special and general court-martial were less pronounced if the case first involved an investigation by a MCIO, rather than being initiated by other means such as by a commander-directed investigation. Specifically, Figure 7 of the report shows that, for example, in cases involving a MCIO investigation, black service members in the Army were only 1.16 times (rather than 2 times) more likely to face trial by general or special court-martial than their peers of other races. Hispanic Soldiers were only 1.37 times (rather than 1.41 times) more likely to be tried by such courts, and male Soldiers were only 3.92 times (rather than 5.75) times more likely.

The GAO is careful to correctly warn its audience that, although the statistically-verifiable existence of disparity is cause for further study, it is not evidence of unlawful discrimination. Rather:

Our analyses of these data, taken alone, do not establish whether unlawful discrimination has occurred, as that is a legal determination that would involve other corroborating information along with supporting statistics. Further, we did not identify the causes of any racial or gender disparities, and the results of our work alone should not be used to make conclusions about the military justice process.

As a way ahead, GAO makes 11 recommendations. Those recommendations largely concern requiring the Departments of Defense and Homeland Security to gather more reliable data by which the scope of the military justice system’s racial, ethnic and gender disparities can be measured. GAO also recommends that the Secretaries publish canary-in-the-coal mine standards, identifying what sort of data, if seen, should trigger further review. Most broadly, in recommendation 11, GAO suggests:

The Secretary of Defense, in collaboration with the Secretaries of the military services and the Secretary of Homeland Security, should conduct an evaluation to identify the causes of any disparities in the military justice system, and take steps to address the causes of these disparities as appropriate.

That is an exceptionally broad recommendation, likely requiring the collection of more data than we currently possess. To that point: throughout its report, GAO advises that, to understand the genesis of the disparities it has documented, we need better “data on the service member’s race, ethnicity, gender, offense, and punishment.” It is certainly true that having ready access to such demographic data regarding the accused is important. But, those are not the only demographics of import. There is also utility in gathering and maintaining, by case, the same sort of demographic data for the other key actors in the military justice system: complaining witnesses, criminal investigators, investigative supervisors, commanders, convening authorities, staff judge advocates, court members, counsel, and judges. It seems unlikely that a reliable study on causation into the military justice system’s racial, ethnic and gender disparities could be conducted without, among other factors, considering the possible impact of racial, ethic, and gender under-representation in those key roles.

7 Responses to “Scholarship Saturday: The GAO alerts the House Committee on Armed Services to the existence of racial, ethic, and gender disparities in the administration of military justice”

  1. Shawn says:

    At a glance, the disparities seem far less pronounced than in the civilian justice system.  I suppose that, like Robert Mueller, if they had found discrimination, they would have said so.  What they actually say is that they cannot prove discrimination does not exist.
    The problem with this sort of multivariate linear regression analysis is that, having captured the data, you next torture it until it confesses.  Moreover, I see no mention of confidence intervals, suggesting that these were either disappointing or not computed.  That the disparities were limited to early stages of the process suggests Bayesian (rather than frequency based) analysis might be a more rigorous approach.
    Now for the politically incorrect thing that no one may say:  Would it not make sense to control for liklihood of criminal behavior?  That might explain, for example, the huge gender disparity.

  2. Stephen W says:

    I would be interested in seeing an analysis of NJP vs CM decisions broken down by ethnicity and gender, and also O vs E grades. 

  3. Isaac Kennen says:

    Shawn said: 

    Now for the politically incorrect thing that no one may say:  Would it not make sense to control for likelihood of criminal behavior?  That might explain, for example, the huge gender disparity.

    Pray tell: How should we “control for the likelihood of criminal behavior?”

    Should we use the rate at which members of different groups are investigated for alleged crimes? That is, after all, precisely the data most commonly used to assess “likelihood of criminal behavior.” But, it is also precisely the data collected by this GAO report. Your question presumes that measure is insufficient, so what alternative should be used?

    Perhaps we should instead use the rate at which members of a group, once accused, are actually convicted?  Of course, the GAO report has that data as well. And, that data shows there is no statistical difference in conviction rates by race, ethnicity or gender. So, if we use that conviction data to assess “likelihood of criminal behavior” we are left with one of two possibilities:

    (1) The military services tend to launch criminal investigations against men and racial and ethnic minorities at a rate that is unjustified by conviction statistics; or(2) The military services tend to not launch criminal investigations against women and non-minorities at a rate that is unjustified by conviction statistics.
     
     

  4. Shawn says:

    Isaac, “liklihood of criminal behavior” must inform not just the research method, as you point out, but also the very hypothesis being tested.  Do the disparities reflect bias in the justice system?  Or bias in the law itself?  Or lack of bias, suggesting a trait in the population?  An obvious example is sex crimes, where the law is biased by gender.  Another is drug crimes, where changing incarceration rates suggest that the law is biased by race.  (Crack being mostly a former black issue and meth mostly a current white issue.)  Another is the low incarceration rate among Asians, which many attribute to a population trait.  It’s complicated.  The report was careful not to conflate disparity with discrimination.  Nor should we.

  5. George says:

    stop assuming that crime is committed equally across races and genders. There is no, and never has been any basis to make this assumption.

  6. Advocaat says:

    GAO should look at these and other disparities in the context of the panel selection process.  Many CAs are responsive to requests for a diverse panel, especially when you appeal to their fairness and impartiality (if not, giving notice they will be a witness for a pretrial motion to dismiss is an avenue to reconsideration or necessary confrontation).  Defense counsel need to explore disparities of every kind in every context and in every type of proceeding, and raise them when it benefits their clients.

  7. stewie says:

    Oh George, you racist scamp you. Crime is, primarily a function of two things: socioeconomics and age. Younger people commit more crimes than older people. People lower on the socioeconomic scale commit more crimes than people higher on that scale.
    In the early 1900s, that included Irish, Italians, even Jewish people. People back then ascribed it to race too, like you do now, so not much has changed except the races involved have changed because socioeconomics have changed.
    A perfect example is we know that whites, blacks and Hispanics all use drugs at almost exactly the same rates. We also know for a fact that whites are not searched, arrested, prosecuted or sentenced in the same way as the other two groups. Criminality is not a “racial” trait. If it were, Irish and Italian-Americans would be committing crimes at the same rates they did in the early 20th century.
     
    It is certainly true that men commit more crimes than women. That’s also a fact. The whys are somewhat complex, although certainly for sex assaults, it’s not so much.
     
    Ultimately, this shows a disparity that needs to be examined. As stated, it does not mean that the system is racist or sexist. But it does mean there’s a disparity that needs to be looked at.