GAO Report

United States Government Accountability Office

Report to the Committee on Armed
Services, House of Representatives

May 2019

MILITARY JUSTICE
DOD and the Coast
Guard Need to
Improve Their
Capabilities to Assess
Racial and Gender
Disparities

GAO-19-344

May 2019

MILITARY JUSTICE
DOD and the Coast Guard Need to Improve Their
Capabilities to Assess Racial and Gender Disparities
Highlights of GAO-19-344, a report to the
Committee on Armed Services, House of
Representatives

Why GAO Did This Study

What GAO Found

The Uniform Code of Military Justice
(UCMJ) was established to provide a
statutory framework that promotes fair
administration of military justice. Every
active-duty servicemember is subject
to the UCMJ, with more than 258,000
individuals disciplined from fiscal years
2013-2017, out of more than 2.3 million
unique active-duty servicemembers. A
key principle of the UCMJ is that a fair
and just system of military law can
foster a highly disciplined force.

The military services collect gender information, but they do not collect and
maintain consistent information about race and ethnicity in their investigations,
military justice, and personnel databases. This limits their ability to collectively or
comparatively assess these data to identify any disparities (i.e., instances in
which a racial, ethnic, or gender group was overrepresented) in the military
justice system within and across the services. For example, the number of
potential responses for race and ethnicity across the military services’ databases
ranges from five to 32 options for race and two to 25 options for ethnicity, which
can complicate cross-service assessments. The services also are not required to
and, thus, do not report demographic information in their annual military justice
reports—information that would provide greater visibility into potential disparities.

House Report 115-200, accompanying
a bill for the National Defense
Authorization Act for Fiscal Year 2018,
included a provision for GAO to assess
the extent that disparities may exist in
the military justice system. This report
assesses the extent to which (1) the
military services collect and maintain
consistent race, ethnicity, and gender
information for servicemembers
investigated and disciplined for UCMJ
violations that can be used to assess
disparities, and (2) there are racial and
gender disparities in the military justice
system, and whether disparities have
been studied by DOD. GAO analyzed
data from the investigations, military
justice, and personnel databases from
the military services, including the
Coast Guard, from fiscal years 20132017 and interviewed agency officials.

GAO’s analysis of available data found that Black, Hispanic, and male servicemembers were more likely than White or female members to be the subjects of
investigations recorded in databases used by the military criminal investigative
organizations, and to be tried in general and special courts-martial in all of the
military services when controlling for attributes such as rank and education. GAO
also found that race and gender were not statistically significant factors in the
likelihood of conviction in general and special courts-martial for most services,
and minority servicemembers were either less likely to receive a more severe
punishment than White servicemembers or there was no difference among racial
groups; thus, disparities may be limited to particular stages of the process. The
Department of Defense (DOD) has taken some steps to study disparities, but
has not comprehensively evaluated the causes of racial or gender disparities in
the military justice system. Doing so would better position DOD to identify actions
to address disparities and help ensure the military justice system is fair and just.
Likelihood that Servicemembers Were Subjects of Recorded Investigations and
Tried in General and Special Courts-Martial, Fiscal Years 2013-2017

What GAO Recommends
GAO is making 11 recommendations,
including that the services develop the
capability to present consistent race
and ethnicity data, and DOD include
demographic information in military
justice annual reports and evaluates
the causes of disparities in the military
justice system. DOD and the Coast
Guard generally concurred with GAO’s
recommendations.
View GAO-19-344. For more information,
contact Brenda S. Farrell at (202) 512-3604 or
[email protected].

Note: These analyses, taken alone, should not be used to make conclusions about the presence or
absence of unlawful discrimination. These multivariate regression analysis results estimate whether
a racial or gender group is more likely or less likely to be the subject of an investigation or a trial in
general or special courts-martial after controlling for race, gender, rank, and education, and in the Air
Force, years of service. GAO made all racial comparisons to White servicemembers and all gender
comparisons to females. GAO grouped individuals of Hispanic ethnicity together, regardless of race.
United States Government Accountability Office

Contents

Letter

1
Background
The Military Services Collect and Maintain Gender Information,
but Do Not Collect and Maintain Consistent Information about
Race and Ethnicity, Limiting Their Ability to Collectively or
Comparatively Assess Data to Identify Any Disparities
Racial and Gender Disparities Exist in Military Justice
Investigations, Disciplinary Actions, and Case Outcomes, but
Have Not Been Comprehensively Studied to Identify Causes
Conclusions
Recommendations for Executive Action
Agency Comments and Our Evaluation

38
66
68
71

Appendix I

Objectives, Scope, and Methodology

73

Appendix II

Summary Statistics and Bivariate Results for Regression Analyses

91

Appendix III

Analysis of Drug Offenses, Sexual Assault Offenses, and All Other
Offenses

111

Appendix IV

Army Data and Analyses

118

Appendix V

Navy Data and Analyses

126

Appendix VI

Marine Corps Data and Analyses

135

Appendix VII

Air Force Data and Analyses

145

Page i

7

22

GAO-19-344 Military Justice

Appendix VIII

Coast Guard Data and Analyses

157

Appendix IX

Key Indicators for Military Justice Actions

162

Appendix X

Comments from the Department of Defense

165

Appendix XI

Comments from the Department of Homeland Security

169

Appendix XII

GAO Contact and Staff Acknowledgments

171

Tables
Table 1: Overview of Nonjudicial Punishment and Different Types
of Courts-Martial
Table 2: Collection of Data on Servicemembers’ Gender in Military
Services’ Investigations, Military Justice, and Personnel
Databases
Table 3: Collection of Data on Servicemembers’ Race in Military
Services’ Investigations, Military Justice, and Personnel
Databases
Table 4: Collection of Data on Servicemembers’ Ethnicity in
Military Services’ Investigations, Military Justice, and
Personnel Databases
Table 5: Overview of Unmatched Military Justice Cases by Military
Service and Database
Table 6: Military Service Personnel, Investigations, and Military
Justice Databases
Table 7: Independent and Dependent Variables Included in GAO’s
Regression Analyses
Table 8: Modeling Groups Used in Regression Analyses for Each
Military Service
Table 9: Groups Used to Measure Punishment Severity
Table 10: Total Population of the Army by Race, Fiscal Years
2013–2017

Page ii

8
27
29
30
34
74
83
84
85
119

GAO-19-344 Military Justice

Table 11: Summary Statistics by Race for Army Military Justice
Actions, Fiscal Years 2013–2017
Table 12: Summary Statistics by Gender for Army Military Justice
Actions, Fiscal Years 2013–2017
Table 13: Summary Statistics by Rank for Army Military Justice
Actions, Fiscal Years 2013–2017
Table 14: Summary Statistics by Age for Army Military Justice
Actions, Fiscal Years 2013–2017
Table 15: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Army Military Justice Actions,
Fiscal Years 2013–2017
Table 16: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Army Military Justice
Actions, Fiscal Years 2013–2017
Table 17: Odds Ratios for Army Multivariate Regression Analyses
Table 18: Total Population of the Navy by Race, Fiscal Years
2013–2017
Table 19: Summary Statistics by Race for Navy Military Justice
Actions, Fiscal Years 2013–2017
Table 20: Summary Statistics by Gender for Navy Military Justice
Actions, Fiscal Years 2013–2017
Table 21: Summary Statistics by Rank for Navy Military Justice
Actions, Fiscal Years 2013–2017
Table 22: Summary Statistics by Education for Navy Military
Justice Actions, Fiscal Years 2013–2017
Table 23: Summary Statistics by Age for Navy Military Justice
Actions, Fiscal Years 2013–2017
Table 24: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Navy Military Justice Actions,
Fiscal Years 2013–2017
Table 25: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Navy Military Justice
Actions, Fiscal Years 2013–2017
Table 26: Odds Ratios for Navy Multivariate Regression Analyses
Table 27: Total Population of the Marine Corps by Race, Fiscal
Years 2013–2017
Table 28: Summary Statistics by Race for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017
Table 29: Summary Statistics by Gender for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017
Table 30: Summary Statistics by Rank for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017

Page iii

119
120
121
122
123
124
125
127
128
129
129
130
130
132
133
134
136
137
138
139

GAO-19-344 Military Justice

Table 31: Summary Statistics by Education for Marine Corps
Military Justice Actions, Fiscal Years 2013–2017
Table 32: Summary Statistics by Age for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017
Table 33: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Marine Corps Military Justice
Actions, Fiscal Years 2013–2017
Table 34: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Marine Corps Military Justice
Actions, Fiscal Years 2013–2017
Table 35: Odds Ratios for Marine Corps Multivariate Regression
Analyses
Table 36: Total Population of the Air Force by Race, Fiscal Years
2013–2017
Table 37: Summary Statistics by Race for Air Force Military
Justice Actions, Fiscal Years 2013–2017
Table 38: Summary Statistics by Gender for Air Force Military
Justice Actions, Fiscal Years 2013–2017
Table 39: Summary Statistics by Rank for Air Force Military
Justice Actions, Fiscal Years 2013–2017
Table 40: Summary Statistics by Education for Air Force Military
Justice Actions, Fiscal Years 2013–2017
Table 41: Summary Statistics by Age for Air Force Military Justice
Actions, Fiscal Years 2013–2017
Table 42: Summary Statistics by Rank and Years of Service
Hybrid Variable for Air Force Military Justice Actions,
Fiscal Years 2013–2017
Table 43: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Air Force Military Justice
Actions, Fiscal Years 2013–2017
Table 44: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Air Force Military Justice
Actions, Fiscal Years 2013–2017
Table 45: Odds Ratios for Air Force Multivariate Regression
Analyses
Table 46: Total Population of the Coast Guard by Race, Fiscal
Years 2013–2017
Table 47: Summary Statistics by Race for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017
Table 48: Summary Statistics by Gender for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017

Page iv

140
141
142
143
144
146
147
148
149
150
151
152
153
154
156
158
158
159

GAO-19-344 Military Justice

Table 49: Summary Statistics by Rank for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017
Table 50: Summary Statistics by Education for Coast Guard
Military Justice Actions, Fiscal Years 2013–2017
Table 51: Summary Statistics by Age for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017
Table 52: Odds Ratios for Coast Guard Multivariate Regression
Analyses
Table 53: Servicemember Rank Groups Most Likely to Be Subject
to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Rank Groups
Table 54: Servicemember Education Groups Most Likely to Be
Subject to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Education
Groups
Table 55: Servicemember Length of Service Groups Most Likely
to Be Subject to Investigations, Courts-Martial, and
Nonjudicial Punishments when Compared with All Other
Length of Service Groups
Table 56: Servicemember Age Groups Most Likely to Be Subject
to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Age Groups

159
160
160
161
163

163

164
164

Figures
Figure 1: Servicemembers Disciplined for Uniform Code of Military
Justice Violations during Fiscal Years 2013–2017
Figure 2: Overview of the Typical Military Justice Process
Figure 3: Number and Percent of General and Special CourtMartial Cases, with and without Recorded Investigations
by Service, Fiscal Years 2013–2017
Figure 4: Military Services’ Investigations, Courts-Martial, and
Nonjudicial Punishment Databases
Figure 5: 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
Figure 6: Likelihood of Trial in General and Special Courts-Martial
by Race and Gender, After Controlling for Rank and
Education, Fiscal Years 2013–2017
Figure 7: Likelihood of Trial in General and Special Courts-Martial
Following a Recorded Investigation by Race and Gender,

Page v

12
13
16
23

41
43

GAO-19-344 Military Justice

After Controlling for Rank and Education, Fiscal Years
2013–2017
Figure 8: Likelihood of Trial in General and Special Courts-Martial
without a Recorded Investigation by Race and Gender,
After Controlling for Rank and Education, Fiscal Years
2013–2017
Figure 9: Likelihood of Trial in Summary Courts-Martial in the Air
Force and the Marine Corps by Race and Gender, After
Controlling for Rank and Education, Fiscal Years 2013–
2017
Figure 10: Likelihood of Nonjudicial Punishments in the Air Force
and the Marine Corps by Race and Gender, After
Controlling for Rank and Education, Fiscal Years 2013–
2017
Figure 11: Army, Navy, and Coast Guard Reported Nonjudicial
Punishments Compared to Nonjudicial Punishments in
Military Justice Databases, Fiscal Years 2013–2017
Figure 12: Likelihood of Conviction in General and Special CourtsMartial by Race and Gender, Fiscal Years 2013–2017
Figure 13: Likelihood of Dismissal or Discharge in General and
Special Courts-Martial by Race and Gender in the Navy
and the Marine Corps, Fiscal Years 2013–2017
Figure 14: Likelihood of More Severe Punishment in General and
Special Courts-Martial by Race and Gender in the Army
and the Air Force, Fiscal Years 2013–2017
Figure 15: Rate and Likelihood of Recorded Investigations for
Alleged Violations of the Uniform Code of Military Justice
by Race and Gender, without Controlling for Any Other
Attributes, Fiscal Years 2013–2017
Figure 16: Rate and Likelihood of Trial in General and Special
Courts-Martial by Race and Gender, without Controlling
for Any Other Attributes, Fiscal Years 2013–2017
Figure 17: Rate and Likelihood of Trial in General and Special
Courts-Martial Following a Recorded Investigation by
Race and Gender, without Controlling for Any Other
Attributes, Fiscal Years 2013–2017
Figure 18: Rate and Likelihood of Trial in General and Special
Courts-Martial without a Recorded Investigation by Race
and Gender, without Controlling for Any Other Attributes,
Fiscal Years 2013–2017
Figure 19: Rate and Likelihood of Trial in Summary Courts-Martial
in the Air Force and the Marine Corps by Race and

Page vi

45

47

49

53
54
58
60
62

93
95

98

100

GAO-19-344 Military Justice

Gender, without Controlling for Any Other Attributes,
Fiscal Years 2013–2017
Figure 20: Rate and Likelihood of Nonjudicial Punishments in the
Air Force and the Marine Corps by Race and Gender,
without Controlling for Any Other Attributes, Fiscal Years
2013–2017
Figure 21: Rate and Likelihood of Conviction in General and
Special Courts-Martial by Race and Gender, without
Controlling for Any Other Attributes, Fiscal Years 2013–
2017
Figure 22: Rate and Likelihood of Dismissal or Discharge in
General and Special Courts-Martial in Navy and Marine
Corps by Race and Gender, without Controlling for Any
Other Attributes, Fiscal Years 2013–2017
Figure 23: Rate and Likelihood of More Severe Punishment in
General and Special Courts-Martial in Army and Air
Force by Race and Gender, without Controlling for Any
Other Attributes, Fiscal Years 2013–2017
Figure 24: Recorded Investigation Rates for Drug Offenses,
Sexual Assault Offenses, and All Other Offenses by Race
and Gender, Fiscal Years 2013–2017
Figure 25: General and Special Courts-Martial Trial Rates for Drug
Offenses, Sexual Assault Offenses, and All Other
Offenses by Race and Gender, Fiscal Years 2013–2017
Figure 26: Likelihood that Charges of Drug Offenses and Sexual
Assault Offenses Resulted in Convictions in General and
Special Courts-Martial, After Controlling for Race,
Gender, Rank, and Education, Fiscal Years 2013–2017

Page vii

102

104

106

108

109
113
115

116

GAO-19-344 Military Justice

Abbreviations
DOD
MCIO
ODEI
OMB
UCMJ

Department of Defense
military criminal investigative organization
Office of Diversity, Equity and Inclusion
Office of Management and Budget
Uniform Code of Military Justice

This is a work of the U.S. government and is not subject to copyright protection in the
United States. The published product may be reproduced and distributed in its entirety
without further permission from GAO. However, because this work may contain
copyrighted images or other material, permission from the copyright holder may be
necessary if you wish to reproduce this material separately.

Page viii

GAO-19-344 Military Justice

Letter

441 G St. N.W.
Washington, DC 20548

May 30, 2019
The Honorable Adam Smith
Chairman
The Honorable Mac Thornberry
Ranking Member
Committee on Armed Services
House of Representatives
The Uniform Code of Military Justice (UCMJ) provides the statutory
framework of the military justice system and establishes the complete
code of military criminal law. 1 It also outlines the jurisdiction and basic
procedure of the military justice system, and provides the legal framework
for conducting investigations and prosecutions of allegations of
misconduct by servicemembers. Every active-duty member of the Army,
the Navy, the Marine Corps, the Air Force, and the Coast Guard is
subject to the UCMJ. According to the Manual for Courts-Martial, the
purpose of military law is to promote justice, assist in maintaining good
order and discipline in the armed forces, promote efficiency and
effectiveness in the military establishment, and thereby strengthen the
national security of the United States. 2 The Military Justice Review Group
elaborated on this purpose, stating that the current structure and practice
of the UCMJ embodies a single overarching principle: a system of military
law can foster a highly disciplined force if it is fair and just, and is
recognized as such by both members of the armed forces and by the
American public. 3
The military justice system has rules, proceedings, and consequences
that are different from the rights and obligations in the civilian criminal
1

10 U.S.C. §§801-946a.

2

The President has implemented the UCMJ through the Manual for Courts-Martial, which
became effective on May 31, 1951, and was initially prescribed by Executive Order 10214
(Feb. 8, 1951). The Manual for Courts-Martial contains the Rules for Courts-Martial, the
Military Rules of Evidence, and the UCMJ. Each military service may supplement the
Manual for Courts-Martial with its own guidance to meet the service’s needs when
authorized to do so by the President.
3

Military Justice Review Group, Report of the Military Justice Review Group Part I: UCMJ
Recommendations, at 16 (Dec. 22, 2015). The Military Justice Review Group was
established at the direction of the Secretary of Defense to conduct a comprehensive
review of the UCMJ and the military justice system.

Page 1

GAO-19-344 Military Justice

court system. 4 In addition to articles that punish traditional crimes such as
unlawful drug use and assault, the UCMJ includes unique military
offenses including desertion, failure to obey orders or regulations, and
misbehavior before the enemy, among others. These unique military
offenses are specifically proscribed in the military context because of their
deleterious effect on morale and mission accomplishment.
In 1995, we reported that studies conducted in the 1970s and 1980s
showed no disparities—instances in which a racial, ethnic, or gender
group was overrepresented—in discipline rates between Black and White
servicemembers 5 and found no evidence that minority groups received
courts-martial or nonjudicial punishments out of proportion to certain
types of violations. 6 In that same report, however, we found that studies
published in the 1990s by the Navy and the Defense Equal Opportunity
Management Institute showed that Black servicemembers were
overrepresented in the number of servicemembers receiving judicial and
nonjudicial punishments. In 2017, a non-profit organization reported that
Black servicemembers were substantially more likely than White
servicemembers to face military justice action. 7
House Report 115-200, accompanying a bill for the National Defense
Authorization Act for Fiscal Year 2018, included a provision for us to
review differences in the way that the military services collect and
maintain information about the race and gender of servicemembers
convicted of violations of the UCMJ and to assess the extent that

4

Article III of the Constitution of the United States governs the federal judiciary, but does
not give it any explicit role in the military. Military courts, referred to as courts-martial, are
not considered to be Article III courts and thus are not subject to all of the rules that apply
in federal courts. For example, the U.S. Constitution specifically exempts military
members accused of a crime from the Fifth Amendment right to a grand jury indictment,
from which the Supreme Court has inferred that there is no right to a civil jury in courtsmartial. See Ex parte Milligan, 71 U.S. (4 Wall.) 2 (1866). Military courts are established
pursuant to Article I of the U.S. Constitution and as a result are of limited jurisdiction.
5

For purposes of this report, we use the term disparities to describe instances in which a
racial or gender group was overrepresented among the servicemembers who were
investigated or disciplined for violations of the UCMJ.
6
GAO, Equal Opportunity: DOD Studies on Discrimination in the Military,
GAO/NSIAD-95-103 (Washington, D.C.: Apr. 7, 1995).
7

Protect Our Defenders, Racial Disparities in Military Justice (May 2017).

Page 2

GAO-19-344 Military Justice

disparities may exist in the military justice system. 8 This report assesses
the extent to which (1) the military services collect and maintain
information about the race, ethnicity, and gender of servicemembers
investigated and disciplined for violations of the UCMJ that can be used
to assess disparities; and (2) there are racial or gender disparities in
investigations, disciplinary actions, and case outcomes in the military
justice system, and whether the Department of Defense (DOD) and the
military services have taken steps to study any identified disparities.
For our first objective, we reviewed service guidance, user manuals, and
other documentation to determine the types of data officials are required
to collect and maintain as well as internal procedures the services follow
to input information about race, ethnicity, and gender into their
investigations, military justice, and personnel databases. For example, we
determined whether the collection of this information was mandatory, and
how this information was entered into and recorded in each database. We
also interviewed agency officials who manage and use the databases to
determine which fields in each database track the race, ethnicity, and
gender of the accused; how these data are input in the databases; and
their insights regarding the reliability of these data. We also analyzed the
data we received from the investigations, military justice, and personnel
databases to determine the completeness of the race, ethnicity, and
gender information that was recorded in each of the databases. We
assessed service systems and procedures for collecting data against
DOD and service guidance and relevant federal internal control
standards. 9
For our second objective, we analyzed military justice actions initiated
and recorded in service investigations and military justice databases
between fiscal years 2013 through 2017—the most recent data available
at the time of our review—as well as record-level data from each of the
military services’ personnel, investigations, and military justice
databases. 10 To prepare the data for our analyses and ensure that we
8

The scope of our review included all five military services: the Army, the Navy, the Marine
Corps, the Air Force, and the Coast Guard. Although the Coast Guard is part of the
Department of Homeland Security, the Coast Guard is a military service and a branch of
the armed forces at all times.
9

GAO, Standards for Internal Control in the Federal Government, GAO-14-704G
(Washington, D.C.: September 2014).
10

We chose this time period because it provided the most recent history of available
military justice data.

Page 3

GAO-19-344 Military Justice

had consistent profiles for the race, ethnicity, and gender of the
servicemembers, we merged records using unique identifiers, such as
social security number or DOD employee identification number, that were
common among a particular service’s databases. Based on discussions
with service officials, we treated the personnel databases as the
authoritative sources for servicemembers’ demographic and
administrative data. In addition, as part of our data preparation, we
consolidated the various race and ethnicity values in the service
personnel databases to the five groups for race and the two groups for
ethnicity established by the Office of Management and Budget (OMB)
standards for maintaining, collecting, and presenting data on race and
ethnicity for federal reporting purposes. 11 When military service personnel
databases included different or additional possible options for race and
ethnicity than the groups established by the OMB standards, we
consolidated the options in accordance with the definitions for each race
and ethnicity option listed in the OMB standards. We grouped individuals
of Hispanic ethnicity together, regardless of their racial identification, so
that we could compare those of Hispanic ethnicity to other racial groups.
Throughout this objective in our report, we refer to the combined race and
ethnicity values as race.
We analyzed data from the military services’ investigations, military
justice, and personnel databases to determine the extent to which racial
and gender groups were the subjects of recorded investigations, tried in
courts-martial, and subject to nonjudicial punishments at higher or lower
rates than each racial and gender group’s proportion of the overall service

11

Office of Management and Budget, Revisions to the Standards for the Classification of
Federal Data on Race and Ethnicity, 62 Fed. Reg. 58,782 (Oct. 30, 1997). In 2016, the
Office of Management and Budget issued a proposed revision to the standards. See,
Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity,
81 Fed. Reg. 67,398 (Sept. 30, 2016). As of May 2019, the Office of Management and
Budget had not issued the revised standards.

Page 4

GAO-19-344 Military Justice

population. 12 We analyzed data for trials in general and special courtsmartial separately from trials in summary courts-martial because general
and special courts-martial result in a criminal conviction if the
servicemember is found guilty, while summary courts-martial are not a
criminal forum and do not result in a criminal conviction. 13 Our analyses
only counted cases that were ultimately tried at general, special, or
summary courts-martial, and excluded those cases where charges were
dismissed, withdrawn, or subject to some alternate resolution.
We also conducted bivariate analyses to estimate the association
between select attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. 14 We
reviewed relevant literature and interviewed agency officials to determine
which demographic attributes would be most appropriate to include in our
analyses. Our bivariate analyses examined attributes such as race,
gender, age, rank, years of service, education, and offense. We then
conducted multivariate regression analyses to test the association
between servicemember characteristics, such as race and gender, and
the odds of a military justice action, while holding other servicemember

12
We reviewed investigation data from the military criminal investigative organizations’
databases where the subject of the investigation was an active-duty servicemember. We
analyzed investigation information from the databases used by the Army’s Criminal
Investigation Command, which included cases investigated by military police and Criminal
Investigation Command; by the Navy and Marine Corps’ Naval Criminal Investigative
Service, which included cases investigated by the Naval Criminal Investigative Service
and military police; by the Air Force’s Office of Special Investigations, which included only
Office of Special Investigations cases; and by the Coast Guard Investigative Service,
which included only Coast Guard Investigative Service cases. This analysis does not
include investigations that were recorded in databases that were not used by the military
criminal investigative organizations, or investigations performed by other military law
enforcement entities or command investigations.
13

For this review, we used the preferral date, or the date when an accused
servicemember was first charged with a violation, to count the number of courts-martial
that occurred in a given fiscal year. However, each military service uses the date in which
the court-martial judgment was given when reporting the number of each type of courtmartial in their annual reports to the Court of Appeals for the Armed Forces. As a result,
the number of court-martial cases in a given year analyzed for our review differs from what
was reported in the annual reports to the Court of Appeals for the Armed Forces.
However, this variation in determining the year in which a case occurred does not impact
the findings of racial and gender disparities.

14

For additional explanation of how we conducted our bivariate analyses, see Appendix I,
and see Appendix II for all of our bivariate analyses results.

Page 5

GAO-19-344 Military Justice

attributes constant, such as gender, rank, and education. 15 Our
multivariate regression analyses controlled for attributes such as race,
gender, rank, years of service, and education. 16 We conducted data
reliability assessments on the datasets we received from the databases in
our review. We examined the documentation related to the databases,
conducted electronic tests on the data we received, and discussed data
reliability with database managers. Based on these actions, for the
purposes of our analysis, we found the variables we ultimately reported
on to be sufficiently reliable. 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. We
also reviewed publications about disparities in the military justice system
and the civilian justice system and summarized them in order to enhance
our understanding of the complexities of the issues, including how others
have attempted to measure disparities. A more detailed description of our
scope and methodology appears in appendix I.
We conducted this performance audit from November 2017 to May 2019
in accordance with generally accepted government auditing standards.
Those standards require that we plan and perform the audit to obtain
sufficient, appropriate evidence to provide a reasonable basis for our
findings and conclusions based on our audit objectives. We believe that
the evidence obtained provides a reasonable basis for our findings and
conclusions based on our audit objectives.

15

A multivariate regression analysis examines several variables simultaneously to
estimate whether each of these variables are more likely or less likely to be associated
with a certain outcome. For the purposes of consistency, in our multivariate analyses, we
made all racial comparisons with White servicemembers and all gender comparisons with
female servicemembers as the reference categories. See Appendix I for a full explanation
of the attributes we used in each service multivariate regression model, and Appendixes
IV through VIII for the demographic breakdowns of each of those attributes in each of the
military services.

16

We could not include education in our multivariate models for the Army due to variability
and overlapping values in the data, which resulted in a data reliability problem, but we
were able to control for age.

Page 6

GAO-19-344 Military Justice

Background
Overview of the Military
Justice System

According to the 2015 report ordered by the Secretary of Defense and
issued by the Military Justice Review Group, the military justice system is
designed to ensure discipline and order in the armed forces, since crimes
committed by servicemembers have the potential to destroy the bonds of
trust, seriously damage unit cohesion, and compromise military
operations. 17 The jurisdiction of the UCMJ extends to all places and
applies to all active-duty servicemembers. UCMJ jurisdiction applies to
other individuals as well, such as members of the National Guard or
reserves who are performing active-duty service; retired members who
are entitled to pay or are receiving hospitalization in a military hospital;
prisoners of war in custody of the armed forces; persons serving with or
accompanying the armed forces in the field in time of declared war or
contingency operations, such as contractors; and members of
organizations such as the National Oceanic and Atmospheric
Administration and the Public Health Service when assigned to and
serving with the armed forces. 18
In creating the military justice system, Congress established three types
of military courts, called courts-martial: summary, special, and general.
Each of these types respectively is intended to deal with progressively
more serious offenses, and each court-martial type may adjudicate more
severe maximum punishments as prescribed under the UCMJ. 19 In
addition, an accused servicemember can receive nonjudicial punishment
under Article 15 of the UCMJ, by which a commander can punish a
servicemember without going through the court-martial process. Table 1
provides an overview of nonjudicial punishments and the three different
types of courts-martial.

17

Military Justice Review Group, Report of the Military Justice Review Group Part I: UCMJ
Recommendations, at 17 (Dec. 22, 2015).

18

The phrase “in the field” regarding the jurisdiction of the UCMJ over civilians and
contractors has been interpreted to mean an area of actual fighting. See, e.g., United
States v. Ali, 71 M.J. 256, 264 (C.A.A.F. 2012), cert. denied, 569 U.S. 972 (2013).
19

In addition to the maximum punishments that may be adjudicated by each type of courtmartial, various relevant executive orders prescribe a maximum punishment for each
offense.

Page 7

GAO-19-344 Military Justice

Table 1: Overview of Nonjudicial Punishment and Different Types of Courts-Martial
Nonjudicial punishment

Summary court-martial

Special court-martial

General court-martial

Purpose

Discipline minor offenses
committed by enlisted
servicemembers or
officers

Adjudicate noncapital
offenses committed by
enlisted servicemembers
Not a criminal forum, so a
guilty finding is not a
criminal conviction

Adjudicate any noncapital
and some capital
a
offenses committed by
enlisted servicemembers
or officers

Adjudicate any offenses
committed by enlisted
servicemembers or
officers, including capital
offenses

Right to counsel

None, but the accused is
generally entitled to be
accompanied by a
spokesperson
The accused may demand
a court-martial in lieu of
nonjudicial punishment
(unless serving on a
vessel)

None, but the accused
must consent to the
proceedings, and will
generally be allowed to
have civilian counsel
represent the accused if
funded by the accused
and counsel’s appearance
will not delay proceedings

The accused is entitled to
an appointed military
attorney, a military
counsel of his or her own
selection (if reasonably
available), or may hire
civilian counsel

The accused is entitled to
an appointed military
attorney, a military
counsel of his or her own
selection (if reasonably
available), or may hire
civilian counsel

Decided by

Commanding officer (or,
for the Coast Guard,
officers-in-charge)

One commissioned officer One military judge and
four members on a panel;
or, one military judge
sitting alone if (1) the
accused requests a
military judge sitting alone,
or (2) the case is referred
to a military judge sitting
alone (which decreases
maximum possible
punishment). (If case is
referred to military judge
sitting alone, and parties
consent, military judge
may designate a military
magistrate to preside.) If a
panel, at least 3/4 of
members must agree on a
b
guilty verdict

Page 8

One military judge and
eight members on a
panel; or, by request of
the accused, one military
judge sitting alone
Panels require twelve
members for all capital
cases and eight members
for all noncapital cases
If a panel, at least 3/4 of
members must agree on a
guilty verdict; capital
verdicts must be
c
unanimous

GAO-19-344 Military Justice

Maximum possible
punishments

Nonjudicial punishment

Summary court-martial

Special court-martial

General court-martial

Depending on the grade
of the commander
imposing the punishment
and of the member being
punished, maximum
punishments range
widely; for example:
Officer: Reprimand,
restrictions with or without
suspension from duties for
up to 30 days; arrest in
quarters for up to 30 days,
forfeiture of one-half
month’s pay for two
months, etc.
Enlisted: Reprimand,
correctional custody or
forfeiture of pay for up to
30 days; reduction in
grade; extra duties for up
d
to 14 days; etc.

Confinement for up to 30
days; hard labor without
confinement for up to 45
days; forfeiture of 2/3 pay
for 1 month; reduction to a
lower pay grade

If referred to a courtmartial consisting of
military judge and panel:
Confinement for up to 1
year; hard labor without
confinement for up to 3
months; forfeiture of 2/3
pay for up to 1 year;
reduction to a lower pay
grade; bad conduct
e
discharge
If referred to a courtmartial consisting of
military judge alone:
Confinement for up to 6
months; hard labor without
confinement for up to 3
months; forfeiture of 2/3
pay for up to 6 months;
reduction to a lower pay
f
grade.

Any punishment within the
limits prescribed by the
Manual for Courts-Martial
for the offenses of which
the accused is found
guilty, including the death
penalty for certain
offenses

Source: GAO analysis of Uniform Code of Military Justice and Rules for Courts-Martial. | GAO-19-344

Note: The Uniform Code of Military Justice overview provided in this table reflects the changes to the
rules as of January 1, 2019. Since the scope of this report covers periods prior to these enacted
changes, the table notes explain the rules that were in effect during the period of our review.
a

A capital offense means an offense for which death is an authorized punishment under the UCMJ
and the Manual for Courts-Martial. A capital offense may be referred to special-court martial if and
only if a mandatory punishment is not prescribed that is beyond the punitive power of a special courtmartial.
b
Prior to changes effective on January 1, 2019, under the Military Justice Act of 2016, special courtsmartial were decided by one military judge and not less than three members on a panel; a panel of
not less than three members; or (with the consent of the accused) one military judge sitting alone.
Furthermore, in special courts-martial decided by panel members, 2/3 of the panel members were
required to agree on a guilty verdict.
c

Prior to changes effective on January 1, 2019, under the Military Justice Act of 2016, general courtsmartial were decided by one military judge and not less than five members on a panel; or (with the
consent of the accused) one military judge sitting alone. General courts-martial decided by a panel in
a capital case were to consist of “not less than” 12 members and required 2/3 of the panel members
to agree on a guilty verdict for offenses that did not carry a mandatory capital sentence; offenses with
mandatory capital sentences required unanimous verdicts.
d

Prior to changes effective on January 1, 2019, under the Military Justice Act of 2016, one possible
nonjudicial punishment was diminished rations of bread and water for 3 days or less, if serving
aboard a vessel.

e

Prior to changes effective on January 1, 2019, under the Military Justice Act of 2016, the maximum
punishment available to a special court-martial—regardless of its composition—was confinement for
up to 1 year; hard labor without confinement for up to 3 months; forfeiture of 2/3 pay for up to 1 year;
reduction to a lower pay grade; bad conduct discharge.
f

Prior to changes effective on January 1, 2019, under the Military Justice Act of 2016, cases were not
referred to a special court-martial consisting of a military judge sitting alone and, consequently, there
was no related requirement for reduced maximum punishments under such circumstances.

Page 9

GAO-19-344 Military Justice

The Military Justice Act of 2016 enacted significant reforms to the UCMJ,
most of the provisions of which became effective on January 1, 2019. 20
These reforms included changes such as limitations on the types of
punishments permitted with nonjudicial punishments, 21 changes to
required size of the panel, or jury, 22 and changes to what judicial
outcomes are subject to automatic appeal. 23 There are some areas where
individual services supplement but remain consistent with the UCMJ. For
example, the Air Force provides a right to counsel in certain forums where
the services are not required to do so.
In addition to the reforms affecting the UCMJ, the Military Justice Act of
2016 also directed changes to military justice data collection and
accessibility. 24 Specifically, section 5504 of the Military Justice Act of
2016 directed the Secretary of Defense to prescribe uniform standards
and criteria pertaining to case management, data collection, and
accessibility of information in the military justice system. 25 As a result, the
DOD Office of General Counsel authorized the establishment of the
Article 140A Implementation Subcommittee of the Joint Service
Committee on Military Justice to, among other things, assess each
service’s case management system, recommend what data fields the
services should collect, propose uniform definitions for the data fields the
services should collect, and recommend standardized methods and data
field definitions to improve the collection of data concerning race and
20

Military Justice Act of 2016, passed as part of the National Defense Authorization Act for
Fiscal Year 2017, Pub. L. No. 114-328 §§5001-5542 (Dec. 23, 2016).

21

For example, Section 5141 of the Military Justice Act of 2016 removes the authority to
restrict a servicemember’s diet to bread and water or to diminish rations during
confinement as a potential nonjudicial punishment.
22

The number of panel members, or jurors, required for special courts-martial, general
courts-martial with noncapital offenses, and general courts-martial with capital offenses
were set at 4, 8, and 12 members respectively. A capital offense means an offense for
which death is an authorized punishment under the UCMJ and the Manual for CourtsMartial.

23

Where the sentence does not also include death, dismissal, dishonorable discharge, or
a bad conduct discharge, automatic appellate review is now limited to those cases that
result in a sentence of confinement of two years or more, instead of the previous one-year
minimum confinement requirement.

24

Military Justice Act of 2016, passed as part of the National Defense Authorization Act for
Fiscal Year 2017, Pub. L. No. 114-328 §§5001-5542 (Dec. 23, 2016).

25

Pub. L. No. 114-328 §5504 (codified at 10 U.S.C. §940a). This section is also known as
Article 140a of the UCMJ.

Page 10

GAO-19-344 Military Justice

ethnicity of individuals involved in the military justice system. 26 The
subcommittee conducted a study and submitted its recommendations to
the Joint Service Committee Voting Group on July 2, 2018, and the
Voting Group submitted a report and its agreed upon recommendations to
the DOD Office of General Counsel on August 24, 2018. 27 The Military
Justice Act of 2016 provides that the Secretary of Defense was to carry
out this mandate by December 23, 2018, and that the Secretary’s
decisions shall take effect no later than December 23, 2020. On
December 17, 2018, the General Counsel of the Department of Defense
issued uniform standards and criteria, which directed that each military
justice case processing and management system be capable of collecting
uniform data concerning race and ethnicity.

Military Justice Process

From fiscal years 2013 through 2017, more than 258,000 active-duty
servicemembers were disciplined for a violation of the UCMJ, out of more
than 2.3 million unique active-duty servicemembers who served across all
of the military services during this period. Figure 1 shows the number of
cases of each type of court-martial and of nonjudicial punishments in
each of the military services.

26

The Joint Service Committee conducts an annual review of the Manual for Courts
Martial; prepares proposed amendments to the Manual for Courts-Martial and, as
appropriate, the UCMJ; and carries out other tasks related to the military justice system as
assigned by the Office of the General Counsel of the Department of Defense. The Article
140A Implementation Subcommittee of the Joint Service Committee was comprised of
legal and information technology experts from each service branch.

27

The Joint Service Committee Voting Group assesses recommendations and proposed
amendments to the Manual for Courts-Martial during periods of review and presents
recommendations to the DOD Office of General Counsel. Among other roles, the Office of
General Counsel advises the Secretary of Defense regarding all legal matters and
services performed within or involving DOD, including all military justice matters requiring
the attention of the Secretary of Defense; oversees the annual review of the Manual for
Courts-Martial; provides legal advice to DOD organizations and other DOD components;
and oversees, as appropriate, legal services performed within DOD. The Deputy General
Counsel of Personnel and Health Policy received the recommendations and forwarded
them to the General Counsel of the Department of Defense for action.

Page 11

GAO-19-344 Military Justice

Figure 1: Servicemembers Disciplined for Uniform Code of Military Justice Violations during Fiscal Years 2013–2017

Note: The number of summary courts-martial tried in the Army and the Navy, and the number of
nonjudicial punishments in the Army, the Navy, and the Coast Guard were computed using
information from the annual reports of the Court of Appeals for the Armed Forces. Nonjudicial
punishments are reported as a combined number for the Navy and the Marine Corps in the Court of
Appeals for the Armed Forces annual reports. To calculate this reported figure for the Navy, we
subtracted the number of Marine Corps nonjudicial punishment cases we identified in the Marine
Corps personnel database from the reported totals. All other disciplinary action totals were from
GAO’s analysis of the services’ military justice and personnel databases.
a
General courts-martial are used to adjudicate any offenses committed by enlisted servicemembers
or officers, including capital offenses. Special courts-martial are used to adjudicate any noncapital
and some capital offenses committed by enlisted servicemembers or officers. A capital offense
means an offense for which death is an authorized punishment under the UCMJ and the Manual for
Courts-Martial.
b

Summary courts-martial are a non-criminal forum used to adjudicate noncapital offenses committed
by enlisted servicemembers.

c

Nonjudicial punishments are used to discipline minor offenses committed by enlisted
servicemembers or officers.

d

The same servicemember could be involved in multiple cases.

e

The total number of servicemembers presented in this figure represents the number of unique activeduty servicemembers who served during fiscal years 2013 through 2017.

Page 12

GAO-19-344 Military Justice

There are several steps in the discipline of a servicemember who
allegedly commits a crime under the UCMJ, which are summarized in
figure 2 below.
Figure 2: Overview of the Typical Military Justice Process

a

Nonjudicial punishments are used to discipline minor offenses committed by enlisted
servicemembers or officers.

b

Summary courts-martial are a non-criminal forum used to adjudicate noncapital offenses committed
by enlisted servicemembers. Special courts-martial are used to adjudicate any noncapital and some
capital offenses committed by enlisted servicemembers or officers. General courts-martial are used to
adjudicate any offenses committed by enlisted servicemembers or officers, including capital offenses.
A capital offense means an offense for which death is an authorized punishment under the UCMJ and
the Manual for Courts-Martial.

Page 13

GAO-19-344 Military Justice

c
A preliminary hearing is required before referral of charges to a general court-martial, unless waived
by the accused, and is intended to determine issues such as whether there is probable cause to
believe that the accused committed the offense charged.

The military justice process begins once an offense is alleged and an
initial report is made, typically to law enforcement, an investigative entity,
or the suspect’s chain of command. Policies for initiating criminal
investigations by military criminal investigative organizations (MCIO) and
procedures for investigating criminal allegations are set forth in DOD and
service guidance. 28 At this time, the commanding officer or law
enforcement will conduct an inquiry or investigation into the accusations
and gather all reasonably available evidence. MCIOs have the authority
and independent discretion to assume investigative jurisdiction, and do
not require approval from any authority outside of the MCIO to conduct
such an investigation—commanders outside of the organization are not to
impede or interfere with such decisions or investigations by the MCIO. If
an MCIO is involved in the inquiry, the investigative entity is to gather all
reasonably available evidence and provide the commanding officer with
unbiased findings that reflect impartiality as required by DOD
instruction. 29 According to service officials, during the conduct of the
criminal investigation, the subject of the investigation has the right to
obtain legal counsel at any time. 30
28

MCIOs conduct criminal investigations in cases with a DOD nexus, such as if a crime
occurred on a DOD installation, or the subject of the investigation is currently affiliated with
DOD or was subject to the UCMJ at the time of the offense. The DOD service specific
MCIOs are the Army Criminal Investigation Command, the Naval Criminal Investigative
Service, and the Air Force Office of Special Investigations. In addition, the Coast Guard
criminal investigative organization is the Coast Guard Investigative Service. For purposes
of this report, we refer to all four of these entities as MCIOs. Instruction and guidance
related to initiating an investigation include DOD Instruction 5505.03, Initiation of
Investigations by Defense Criminal Investigative Organizations (Mar. 24, 2011)
(incorporating Change 2, Feb. 13, 2017); Army Regulation 195-2, Criminal Investigation
Activities (June 9, 2014); Secretary of the Navy Instruction 5430.107, Mission and
Functions of the Naval Criminal Investigative Service (Dec. 28, 2005); Air Force
Instruction 71-101, Vol. 1, Criminal Investigations Program (Oct. 8, 2015) (certified
current, Dec. 7, 2017); Commandant Instruction 5520.5F, Coast Guard Investigative
Service Roles and Responsibilities (Nov. 30, 2011).

29

DOD Instruction 5505.03, Initiation of Investigations by Defense Criminal Investigative
Organizations (Mar. 24, 2011) (incorporating change 2, Feb. 13, 2017).

30

According to the Rules for Courts-Martial, in cases of pre-trial confinement the accused
has the right to retain civilian counsel at no cost to the government, and may request to be
assigned military counsel for representation during pretrial confinement proceedings,
which must be honored.

Page 14

GAO-19-344 Military Justice

After an investigation, the first step toward initiation of a court-martial is
when the accused is presented with a list of charges signed by the
accuser under oath, which is called preferral of charges; the accuser who
prefers the charges may be anyone subject to the UCMJ. After charges
are preferred, the charges are forwarded to an officer with sufficient legal
authority to convene a court-martial, also known as the “convening
authority.” The convening authority in receipt of preferred charges may,
among other actions and depending on the nature of the charges and the
level of the convening authority, refer the case to its own court or forward
the case to a superior commander for disposition, for example, to a
general court-martial convening authority. The general court-martial
convening authority would have similar options: to dismiss the charges,
refer them to a general or special court-martial, or take some lesser
action. Before any case is referred to a general court-martial, the case
must proceed through a preliminary hearing under Article 32 of the
UCMJ, unless waived by the accused. The Article 32 hearing is presided
over by an impartial judge advocate, or another individual with statutory
authority, who is appointed by the convening authority and makes a
recommendation to the convening authority.
We analyzed general and special courts-martial that were preceded by
investigations recorded in databases maintained by MCIOs, which we
refer to as recorded investigations, and general and special courts-martial
that did not have a record within an MCIO database. 31 As shown in figure
3 below, the majority of general and special courts-martial, ranging from
53 percent to 74 percent across the services, had a recorded
investigation, while the remaining cases would have been investigated by
other sources, such as local civilian law enforcement, command

31

Investigations are recorded in the MCIO databases when a servicemember is the
subject of a criminal allegation; for purposes of this report, we say the servicemember had
a “recorded investigation” to describe these cases. To conduct our analyses, we used
data from the databases used by the Army’s Criminal Investigation Command, which
included cases investigated by military police and Criminal Investigation Command; by the
Navy and Marine Corps Naval Criminal Investigative Service, which included cases
investigated by the Naval Criminal Investigative Service and military police; by the Air
Force’s Office of Special Investigations, which included only Office of Special
Investigations cases; and by the Coast Guard Investigative Service, which included only
Coast Guard Investigative Service cases.

Page 15

GAO-19-344 Military Justice

investigations, or in the case of the Air Force, their military law
enforcement forces. 32
Figure 3: Number and Percent of General and Special Court-Martial Cases, with and
without Recorded Investigations by Service, Fiscal Years 2013–2017

a
General courts-martial are used to adjudicate any offenses committed by enlisted servicemembers
or officers, including capital offenses. Special courts-martial are used to adjudicate any noncapital
and some capital offenses committed by enlisted servicemembers or officers. A capital offense
means an offense for which death is an authorized punishment under the UCMJ and the Manual for
Courts-Martial.
b

General and special courts-martial without a recorded investigation are cases where the
servicemember was not the subject of an investigation recorded in the military criminal investigative
organization (MCIO) databases; these cases would have been investigated by other entities, such as
civilian law enforcement or command investigations.

c
General and special courts-martial with a recorded investigation are cases where the servicemember
was the subject of an investigation recorded in an MCIO database.

32

Our analysis of recorded investigations data did not include investigations conducted by
a servicemember’s command, because those investigations are not recorded in the MCIO
databases. Command investigations are included in our analysis of general and special
courts-martial cases without a recorded investigation.

Page 16

GAO-19-344 Military Justice

Once referred to a general or special court-martial, an accused
servicemember may be tried by a military judge alone or by a military
judge with a military jury, referred to as members of the court-martial. 33 If
the accused servicemember is tried by a military jury, the members of the
court-martial determine whether the accused is proven guilty and, if the
accused requests sentencing by the members, adjudicate a sentence.
Otherwise, the military judge adjudicates the sentence. If the accused is
tried by a military judge alone, the judge determines guilt and any
sentence. In a summary court-martial, a single commissioned officer who
is not a military judge adjudicates minor offenses and a sentence.
Convictions at the general and special court-martial level are subject to a
post-trial process and may be appealed to higher courts in cases where
the sentence reaches a certain threshold. For example, depending on the
forum and the adjudged sentence, the accused may be entitled to
appellate review by the service Court of Criminal Appeals, and may be
able to request or waive assignment of appellate defense counsel, or
waive appellate review entirely. Depending, again, on forum and
sentence, some cases that do not qualify for appellate review will receive
review by a judge advocate to, among other things, determine that the
court had jurisdiction and that the sentence was lawful. Some cases may
then be further reviewed by the Court of Appeals for the Armed Forces,
as well as by the U.S. Supreme Court at their discretion, if the case was
reviewed by the Court of Appeals for the Armed Forces.
The military justice system, like the civilian criminal justice system,
provides avenues for accused servicemembers to raise allegations of
discrimination, improprieties in investigations, improprieties in disposition,
and improprieties in the selection of panel members at the court-martial
proceeding, before a military judge and on appellate review. The Military
Justice Act of 2016 requires that legal training be provided to all officers,

33

Members may be commissioned officers in all cases, warrant officers in all cases except
when the accused is a commissioned officer, or enlisted members when the accused is an
enlisted member. If an enlisted accused so requests, at least one-third of the members
must be enlisted members.

Page 17

GAO-19-344 Military Justice

with additional training for commanders with authority to take disciplinary
actions under the UCMJ. 34

Definitions of Race,
Ethnicity, and Gender

The Office of Management and Budget (OMB) has established standards
for collecting, maintaining, and presenting data on race and ethnicity for
all federal reporting purposes. 35 These standards were developed in
cooperation with federal agencies to provide consistent data on race and
ethnicity throughout the federal government. 36 OMB standards establish
the following five categories of race:
•

American Indian or Alaska Native: A person having origins in any of
the original peoples of North and South America (including Central
America), and who maintains tribal affiliation or community
attachment.

•

Asian: A person having origins in any of the original peoples of the
Far East, Southeast Asia, or the Indian subcontinent including, for
example, Cambodia, China, India, Japan, Korea, Malaysia, Pakistan,
the Philippine Islands, Thailand, and Vietnam.

•

Black or African American: A person having origins in any of the
black racial groups in Africa.

34

Specifically, the Military Justice Act of 2016 included a provision that requires that the
UCMJ be carefully explained to each officer at the time of or within six months after initial
entrance of the officer on active duty, or the initial commissioning of the officer in a reserve
component. The act further requires officers with the authority to convene courts-martial or
impose nonjudicial punishment to receive periodic training regarding the purpose and
administration of the UCMJ. Military Justice Act of 2016, passed as part of the National
Defense Authorization Act for Fiscal Year 2017, Pub. L. No. 114-328, §5503 (Dec. 23,
2016) (codified at 10 U.S.C. §937).

35
Office of Management and Budget, Revisions to the Standards for the Classification of
Federal Data on Race and Ethnicity, 62 Fed. Reg. 58,782 (Oct. 30, 1997). In 2016, the
Office of Management and Budget issued a proposed revision to the standards. See
Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity,
81 Fed. Reg. 67,398 (Sept. 30, 2016). As of May 2019, the Office of Management and
Budget had not issued the revised standards.
36

According to officials from all of the military services, the racial and ethnic information in
their databases is self-reported, and so servicemembers may not follow these definitions
when they identify their race and ethnicity. We used these definitions in our analyses so
that we could group the various racial and ethnic categories in the services’ personnel
databases into consistent categories across all of the military services.

Page 18

GAO-19-344 Military Justice

•

Native Hawaiian or Other Pacific Islander: A person having origins
in any of the original peoples of Hawaii, Guam, Samoa, or other
Pacific Islands.

•

White: A person having origins in any of the original peoples of
Europe, the Middle East, or North Africa.

The OMB standards also establish two categories of ethnicity.
•

Hispanic or Latino: A person of Cuban, Mexican, Puerto Rican,
South or Central American, or other Spanish culture or origin,
regardless of race.

•

Not Hispanic or Latino: A person not having the above attributes.

In addition to defining race and ethnicity for federal administrative
reporting and record keeping requirements, OMB standards provide two
methods for federal agencies to follow regarding the collection of data on
race and ethnicity.
1. Separate questions shall be used for collecting information about race
and ethnicity wherever feasible. In this case, there are 5 categories of
race noted above which individuals can select, and individuals can
identify with more than one category of race. In addition to race,
individuals can select one of the two ethnicity categories above.
2. If necessary, a single question or combined format can be used to
collect information about race and ethnicity, where the following
categories are provided for individuals: American Indian or Alaska
Native, Asian, Black or African American, Hispanic or Latino, Native
Hawaiian or other Pacific Islander, and White. In this instance,
individuals can also select more than one category.
Information collected on servicemembers’ gender is governed by DOD
guidance. DOD Instruction 1336.05 provides that information collected on
a servicemember’s gender is based on reproductive function. 37 It provides
that there are three options that can be selected when inputting a
servicemember’s gender: male, female, or unknown.

37

DOD Instruction 1336.05, Automated Extract of Active Duty Military Personnel Records
(July 28, 2009) (incorporating Change 2, effective Mar. 31, 2015).

Page 19

GAO-19-344 Military Justice

Racial and Gender
Disparities in the Civilian
Justice System

Racial and gender disparities in the civilian criminal justice system have
been the subject of several studies in the past decade. While the civilian
and military justice systems differ from each other, we reviewed
information about racial and gender disparities in the civilian criminal
justice system to enhance our understanding of the complexities of the
issues, including how others had attempted to measure disparities. Some
studies have assessed the rates at which minority groups are policed. For
example, a Department of Justice study of data from the Bureau of
Justice Statistics’ 2011 Police-Public Contact survey found that Black
drivers were more likely than White or Hispanic drivers to be pulled over
in a traffic stop; specifically, the study found that 10 percent of White
drivers and 10 percent of Hispanic drivers were pulled over in a traffic
stop, compared to 13 percent of Black drivers. 38 This study also found
that Black and Hispanic drivers were more likely to be searched once
they were pulled over by the police; specifically, the study found that 2
percent of White drivers stopped by police were searched, compared to 6
percent of Black drivers and 7 percent of Hispanic drivers.
In addition, U.S. government data shows that racial disparities exist
among individuals who are arrested. For example, data from the Federal
Bureau of Investigation’s Uniform Crime Reporting Program, which
compiles data from law enforcement agencies across the country,
indicates that in 2016, Black individuals represented 26.9 percent of total
arrests nationwide, but comprised 13.4 percent of the U.S. population
according to U.S. census data estimates as of July 1, 2017. 39 This data
also shows that 69.6 percent of all arrested individuals were White, while
White individuals comprised 76.6 percent of the U.S. population.
Studies have also identified racial and gender disparities in civilian justice
sentencing. In 2010 and 2017, the U.S. Sentencing Commission reported
that Black male offenders received longer sentences than similarly
situated White male offenders. 40 Specifically, in 2017, the Commission
analyzed federal sentencing data and reported that Black male offenders
38

U.S. Department of Justice, Bureau of Justice Statistics, Police Behavior during Traffic
and Street Stops (September 2013).
39

U.S. Department of Justice, Uniform Crime Report, Crime in the United States 2016,
Table 21; U.S. Census Bureau, Quick Facts, population estimates as of July 1, 2017.

40

U.S. Sentencing Commission, Demographic Differences in Sentencing: An Update to
the 2012 Booker Report (November 2017); U.S. Sentencing Commission, Demographic
Differences in Federal Sentencing Practices: An Update of the Booker Report’s
Multivariate Regression Analysis (March 2010).

Page 20

GAO-19-344 Military Justice

received sentences that on average were 19.1 percent longer than
similarly situated White males for fiscal years 2012 to 2016. This analysis
controlled for factors such as type of offense, race, gender, citizenship,
age, education level, and criminal history. This study also found that
female offenders of all races received shorter sentences than White male
offenders. Similarly, the Commission’s 2010 report found that Black
offenders received sentences that were 10 percent longer than those
imposed on White offenders from December 2007 through September
2009, and male offenders received sentences that were 17.7 percent
longer than female offenders, after controlling for the same factors as
noted for the 2017 study, among others.
Finally, racial and gender disparities have been identified among
incarcerated populations. According to data from the Bureau of Justice
Statistics, for prisoners with sentences of 1 year or more under the
jurisdiction of state or federal correctional officials in 2016, Black males
were six times more likely to be imprisoned than White males, and
Hispanic males were 2.7 times more likely to be imprisoned than White
males. 41 The racial disparities were more pronounced for younger males,
where Black males aged 18 to 19 were approximately 11.8 times more
likely than White males of the same age to be imprisoned. The Bureau
also reported that Black females were imprisoned at approximately twice
the rate of White females. We did not assess the methodologies used in
any of these studies or the reliability of the data cited in the studies; these
studies are discussed here to provide broader context for the discussion
about racial and gender disparities in the military justice system. 42

41
U.S. Department of Justice, Bureau of Justice Statistics, Prisoners in 2016 (January
2018).
42

The findings in these studies, 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.

Page 21

GAO-19-344 Military Justice

The Military Services
Collect and Maintain
Gender Information,
but Do Not Collect
and Maintain
Consistent
Information about
Race and Ethnicity,
Limiting Their Ability
to Collectively or
Comparatively
Assess Data to
Identify Any
Disparities

The military services collect and maintain gender information, but they do
not collect and maintain consistent information about race and ethnicity in
their investigations, military justice, and personnel databases. This limits
the military services’ ability to collectively or comparatively assess these
demographic data to identify any racial or ethnic disparities in the military
justice system within and across the services. The military services use
different databases to collect and maintain information for investigations,
courts-martial, and nonjudicial punishments. All of the databases collect
and maintain gender information, but the Coast Guard’s military justice
database does not have the capability to query or report on gender data.
While the military services’ databases collect and maintain complete data
for race and ethnicity, the information collected and maintained about
race and ethnicity is not consistent among the different databases within
and across the services. Moreover, the Coast Guard, the Navy, and the
Marine Corps do not collect and maintain complete and consistent
servicemember identification data, such as social security number or
employee identification number, in their respective military justice
databases, although DOD leadership recently directed improvements in
this area. Finally, the military services do not report data that provides
visibility into disparities in the military justice system, and DOD and the
services lack guidance about when potential racial, ethnic, or gender
disparities should be further reviewed, and what steps should be taken to
conduct such a review if needed.

The Military Services Use
Different Databases to
Collect and Maintain
Information for
Investigations, CourtsMartial, and Nonjudicial
Punishments

Each military service uses a different database to collect and maintain
information on investigations and courts-martials, and, in some services,
nonjudicial punishments, as shown in figure 4. For three of the military
services—the Army, the Navy, and the Coast Guard—the databases
listed in figure 4 include information about some, but not all, of their
nonjudicial punishment cases.

Page 22

GAO-19-344 Military Justice

Figure 4: Military Services’ Investigations, Courts-Martial, and Nonjudicial
Punishment Databases

Note: Depending on the military service, information about nonjudicial punishments can be found in
the military justice or the personnel database. The databases listed for nonjudicial punishments
collect and maintain information about some but not all nonjudicial punishments, with the exception of
AMJAMS and MCTFS, which collect information about all nonjudicial punishments according to Air
Force and Marine Corps officials.

Additionally, the nature of the information collected by each of the
services’ databases varies, as noted below.
Army
•

Investigations. The Army collects and maintains information on
investigations conducted by the Army Criminal Investigation
Command in the Army Law Enforcement Reporting and Tracking
System database. 43 According to Army officials, the Office of the
Provost Marshal General and the Army Criminal Investigation
Command developed this database to replace a 2003 system, the
Army Criminal Investigation and Intelligence System, and a significant

43

According to Army officials, the Army started using an automated system, the Army
Criminal Investigative Reporting System (ACIRS), for felony investigations in 1989.

Page 23

GAO-19-344 Military Justice

part of the military police’s 2002 system, the Centralized Operations
Police Suite. The officials said that the Army Law Enforcement
Reporting and Tracking System has been operational since 2015, and
has become the primary case management system for all Army law
enforcement professionals. However, Army officials said that cases
involving commander-led investigations are unlikely to be recorded in
this database.
•

Courts-martial and nonjudicial punishments. The Army uses
Military Justice Online and the Army Courts-Martial Information
System to collect data on court-martial cases. According to Army
officials, Military Justice Online, created in 2008, is a documentgenerating system that primarily is used by the Army’s judge advocate
general corps and promotes uniformity in case processing among the
Army’s staff judge advocate offices. Military Justice Online includes
information about courts-martial, some nonjudicial punishments,
administrative separations, and administrative reprimands of
servicemembers. Army officials said that the Army Courts-Martial
Information System, which has been used since 1989, serves as the
Army trial judiciary’s case tracking system and is used by the Army’s
trial judiciary to track court-martial cases.

Air Force
•

Investigations. The Air Force military criminal investigative
organization, the Office of Special Investigations, uses a system
called the Investigative Information Management System to collect
and maintain information related to investigations. According to Air
Force officials, the Investigative Information Management System has
been in use since 2001.

•

Courts-martial and nonjudicial punishments. The Air Force uses
the Automated Military Justice Analysis and Management System,
which is designed to be a case management system to collect
comprehensive information for both court-martial cases and
nonjudicial punishments. According to Air Force officials, the
Automated Military Justice Analysis and Management System has
been in use since 1974.

Navy and Marine Corps
•

Investigations. According to Navy officials, the Navy and Marine
Corps’ joint system for maintaining and collecting information related
to investigations is the Consolidated Law Enforcement Operations

Page 24

GAO-19-344 Military Justice

Center, which has been in use since 2004. Navy officials said that this
database initially contained information regarding Navy and Marine
Corps law enforcement incidents and criminal investigations, but
began to include investigations conducted by the Naval Criminal
Investigative Service in 2012.
•

Courts-martial. The Navy and the Marine Corps both use the Case
Management System to collect and maintain information about
military justice matters with involvement by a Navy or Marine Corps
legal office, including special and general court-martial cases. This
system was initially developed by the Marine Corps to track
information about legal services provided by their legal offices.
According to Navy and Marine Corps officials, the system has been in
use by the Marine Corps since 2010 and by the Navy since 2013.
Officials from the Marine Corps said that although the Case
Management System has been in use since 2010, the system was not
widely used until 2012.

•

Nonjudicial punishments. The Marine Corps Total Force System,
the Marine Corps personnel database, collects and maintains
information on summary courts-martial and nonjudicial punishments
for cases where there was a conviction or punishment. According to
Marine Corps officials, this system has been in use since 1995. Navy
officials said that their personnel database records information about
nonjudicial punishments if the punishment involved a change in pay or
grade. The services’ military justice Case Management System
includes information on some nonjudicial punishment cases in the
Navy and the Marine Corps, which Navy and Marine Corps officials
said was for those cases that had involvement by their legal offices.

Coast Guard
•

Investigations. The Coast Guard Investigative Service uses the Field
Activity Case Tracking System to collect and maintain information on
servicemembers investigated for violations of the UCMJ. According to
Coast Guard officials, this system has been in use since July 2014.

•

Courts-martial. According to Coast Guard officials, the Coast Guard
uses Law Manager to collect and maintain administrative information
on court-martial cases. Law Manager has been in use since 2000, but
was not used for court-martial data until 2003.

•

Nonjudicial punishments. Coast Guard officials said that their
military justice database contains records of nonjudicial punishments
if a case involved their legal offices. In addition, according to Coast

Page 25

GAO-19-344 Military Justice

Guard officials, Direct Access, the Coast Guard’s personnel database,
also collects and maintains information about some court-martial
cases and nonjudicial punishments if the punishment resulted in a
change in rank or pay or an administrative action against the accused
servicemember.

The Military Services
Collect and Maintain
Gender Data, but the
Coast Guard Can Not
Query or Report on
Gender Data from its
Military Justice Database

All of the military services collect and maintain gender information in their
investigations, military justice, and personnel databases, but are
inconsistent in whether they allow an unknown or unspecified gender, and
the Coast Guard’s military justice database does not allow Coast Guard
officials to query or report on gender data. Table 2 below summarizes
how data regarding the servicemember’s gender is entered into the
services’ databases and the number of potential gender options. 44 Each
database identifies at least two potential options—male and female—for
data related to the servicemember’s gender, while about half of the
databases (8 of 15) provide a third option to indicate that the gender is
either unknown or not specified. 45 Each of the military services’
investigations, military justice, and personnel databases maintained
gender data for almost 100 percent of servicemembers, except we were
unable to determine this completion rate for the Coast Guard’s military
justice database. We could not determine the completeness of the Coast
Guard’s gender data in its military justice database because, as

44

The collection and maintenance of information pertaining to the gender identity of
transgender servicemembers within the armed forces has changed during the time period
of the data collected for our review. The Secretary of Defense released a memorandum in
2015 lifting service restrictions placed on servicemembers based on their gender identity.
Secretary of Defense Memorandum, Transgender Service Members (July 28, 2015). In
October 2016, DOD issued an instruction to establish a process for servicemembers to
transition during service and change their gender status within the Defense Enrollment
Eligibility Reporting System. DOD Instruction 1300.28, In-Service Transition for
Transgender Service Members (Oct. 1, 2016). In August 2017, the President issued a
memorandum directing DOD to reverse the policy changes regarding transgender
servicemembers. Memorandum for the Secretary of Defense and the Secretary of
Homeland Security, Military Service by Transgender Individuals, 82 Fed. Reg. 41,319
(Aug. 30, 2017). The provisions outlined in the memorandum became effective on January
1, 2018.

45

One database (the Navy and Marine Corps shared Consolidated Law Enforcement
Operations Center database) provides a fourth option—indeterminate. According to Army
officials, the criminal investigations databases collect information about gender to support
reporting requirements of 3 options associated with the Defense Incident-Based Reporting
System, which is a central database used by DOD to provide military crime statistics to the
Department of Justice.

Page 26

GAO-19-344 Military Justice

previously noted, its military justice database does not have the capability
to query on gender data. 46
Table 2: Collection of Data on Servicemembers’ Gender in Military Services’ Investigations, Military Justice, and Personnel
Databases
Gender information
Service

Database

Entry method

Number of potential
gender options

Army

Army Law Enforcement Reporting and
Tracking System (I)

Manual input (drop down)

3 options

Military Justice Online (MJ)

Auto-populated from Total Army Personnel
Database

2 options

Army Courts-Martial Information System
(MJ)

Manual input (drop down)

3 options

Total Army Personnel Database (P)

Manual input (drop down)

3 options

Navy and Marine
Corps Shared

Consolidated Law Enforcement
Operations Center (I)

Manual input (drop down)

4 options

Navy

Case Management System (MJ)

Manual input (drop down)

3 options

Navy Personnel Database (P)

Manual input (drop down)

2 options

Case Management System (MJ)

Manual input (drop down)

3 options

Marine Corps Total Force System (P)

Manual input (drop down)

2 options

Investigative Information Management
System (I)

Auto-populated from Defense Enrollment
Eligibility Reporting System

3 options

Automated Military Justice Analysis and
Management System (MJ)

Auto-populated from Military Personnel
Data System

2 options

Military Personnel Data System (P)

Manual input (drop down)

2 options

Field Activity Case Tracking System (I)

Manual input (drop down)

3 options

Law Manager (MJ)

Manual input (drop down)

2 options

Direct Access (P)

Manual input (drop down)

2 options

Marine Corps
Air Force

Coast Guard

a

Legend: (I)=Investigations database; (MJ)=military justice database; (P)=personnel database.
Source: GAO analysis of each military service’s investigations, military justice, and personnel database information. | GAO-19-344

Note: Each database identifies at least two potential options—male and female—for data related to
the servicemember’s gender, while about half of the databases provide a third option to indicate that
the gender is either unknown or not specified.
a
Gender options in the Consolidated Law Enforcement Operations Center database are male, female,
unknown, and indeterminate.

46

The military services differ regarding whether their databases require the collection of
information about gender.

Page 27

GAO-19-344 Military Justice

Standards for Internal Control in the Federal Government states that
management should use quality information and obtain data on a timely
basis so they can be used for effective monitoring. 47 However, the Coast
Guard does not have visibility over the gender of servicemembers
prosecuted for UCMJ violations without merging data from multiple
databases, which can be a labor-intensive and time-consuming process.
According to Coast Guard officials, information regarding the gender of
servicemembers prosecuted for UCMJ violations can be recorded in its
military justice database, but gender is not a field that can be searched on
or included in the reports they run using information from their military
justice database, because of the way the military justice module in the
database was designed. Coast Guard officials told us that the military
justice database—Law Manager—was designed to determine the status
of court-martial cases, and captures attributes that are generated by
relevant UCMJ documents. Those official documents do not require the
annotation of demographics such as gender, so this information is not
used in Law Manager. A Coast Guard official indicated that it would be
feasible to modify Law Manager to make it easier to run reports and
queries that include gender information. The ability to query and report on
the gender of servicemembers in its military justice database would
provide the Coast Guard with more readily available data to identify or
assess any gender disparities that may exist in the investigation and trial
of military justice cases.

The Military Services Do
Not Collect and Maintain
Consistent Data for Race
and Ethnicity

Each of the military services’ databases collect and maintain complete
data for race and ethnicity, but the military services do not collect and
maintain consistent information regarding race and ethnicity in their
investigations, military justice, and personnel databases. Additionally, the
military services have not developed a mechanism to aggregate the data
into consistent categories of race and ethnicity to allow for efficient
analysis and reporting of consistent demographic data. The number of
potential responses for race and ethnicity within the 15 databases across
the military services ranges from 5 to 32 options for race and 2 to 25
options for ethnicity, which can complicate cross-service assessments.
For example, the Army’s personnel database maintains 6 options for race

47

GAO-14-704G.

Page 28

GAO-19-344 Military Justice

and 23 options for ethnicity, 48 whereas the Coast Guard’s personnel
database maintains 7 options for race and 3 for ethnicity. 49 Table 3
summarizes how the databases used by the military services vary in how
the servicemember’s race is entered and the number of potential race
options. 50
Table 3: Collection of Data on Servicemembers’ Race in Military Services’ Investigations, Military Justice, and Personnel
Databases
Race information
Service

Database

Entry method

Number of potential
race options

Army

Army Law Enforcement Reporting and
Tracking System (I)

Manual input (drop down)

8 options

Military Justice Online (MJ)

Auto-populated from Total Army Personnel
Database

6 options

Army Courts-Martial Information System
(MJ)

Manual input (drop down)

8 options

Total Army Personnel Database (P)

Manual input (drop down)

6 options

Navy and Marine
Corps Shared

Consolidated Law Enforcement Operations
Center (I)

Manual input (drop down)

6 options

Navy

Case Management System (MJ)

Manual input (drop down)

7 options

Navy Personnel Database (P)

Manual input (drop down)

32 options

Case Management System (MJ)

Manual input (drop down)

7 options

Marine Corps Total Force System (P)

Manual input (drop down)

6 options

Marine Corps

48

The six options for race available within the Army’s personnel database include
American Indian or Alaska Native, Asian/Pacific Islander, Black, Other, Unknown, and
White. The options for ethnicity include Aleut, Chinese, Cuban, Eskimo, Filipino,
Guamanian, Indian, Japanese, Korean, Latin American with Hispanic Descent,
Melanesian, Mexican, Micronesian, None, Other, Other Asian Descent, Other Hispanic
Descent, Other Pacific Island Descent, Polynesian, Puerto Rican, United States/Canadian
Indian Tribes, Unknown, and Vietnamese.

49

The options for race in the Coast Guard’s personnel database include American
Indian/Alaska Native, Asian, Black or African American, Declined to Respond, Native
Hawaiian/Pacific Islander, White, and every potential mixed racial group from the provided
races. Additionally, the database has three options for ethnicity: declined to respond,
Hispanic or Latino, and Not Hispanic or Latino.

50

According to Army officials, the criminal investigations databases collect information
about race to support reporting requirements of 6 options associated with the Defense
Incident-Based Reporting System, which is a central database used by DOD to provide
military crime statistics to the Department of Justice. For example, Army officials said that
the 8 race options in their investigations database are converted to support the 6 options
in the DOD system.

Page 29

GAO-19-344 Military Justice

Race information
Service

Database

Entry method

Number of potential
race options

Air Force

Investigative Information Management
System (I)

Auto-populated from Defense Enrollment
Eligibility Reporting System

7 options

Automated Military Justice Analysis and
Management System (MJ)

Auto-populated from Military Personnel
Data System

5 options

Military Personnel Data System (P)

Manual input (drop down)

7 options

Field Activity Case Tracking System (I)

Manual input (drop down)

6 options

Law Manager (MJ)

N/A; does not track race

N/A; does not track
race

Direct Access (P)

Manual input (drop down)

7 options

Coast Guard

Legend: (I)=investigations database; (MJ)=military justice database; (P)=personnel database; N/A= not available.
Source: GAO analysis of each military service’s investigations, military justice, and personnel database information. | GAO-19-344

Table 4 shows that the military services’ databases also vary in how
information about servicemembers’ ethnicity is entered into the databases
and the number of potential ethnicity options that are collected. 51
Table 4: Collection of Data on Servicemembers’ Ethnicity in Military Services’ Investigations, Military Justice, and Personnel
Databases
Ethnicity information
Service

Database

Entry method

Number of potential
ethnicity options

Army

Army Law Enforcement Reporting and
Tracking System (I)

Manual input (drop down)

3 options

Military Justice Online (MJ)

N/A; collected as part of race field

N/A

Army Courts-Martial Information System (MJ) N/A; collected as part of race field

N/A

Total Army Personnel Database (P)

Manual input (drop down)

23 options

Navy and Marine
Corps Shared

Consolidated Law Enforcement Operations
Center (I)

Manual input (drop down)

3 options

Navy

Case Management System (MJ)

N/A; collected as part of race field

N/A

Navy Personnel Database (P)

Manual input (drop down)

23 options

Case Management System (MJ)

N/A; collected as part of race field

N/A

Marine Corps Total Force System (P)

Manual input (drop down)

25 options

Marine Corps

51

According to Army officials, the criminal investigations databases collect information
about ethnicity to support reporting requirements of 3 options associated with the Defense
Incident-Based Reporting System, which is a central database used by DOD to provide
military crime statistics to the Department of Justice.

Page 30

GAO-19-344 Military Justice

Ethnicity information
Service

Database

Entry method

Number of potential
ethnicity options

Air Force

Investigative Information Management
System (I)

Manual input (drop down)

3 options

Automated Military Justice Analysis and
Management System (MJ)

Auto-populated from Military
Personnel Data System

3 options

Military Personnel Data System (P)

Manual input (drop down)

23 options

Field Activity Case Tracking System (I)

Manual input (drop down)

2 options

Law Manager (MJ)

N/A; does not track ethnicity

N/A

Direct Access (P)

Manual input (drop down)

3 options

Coast Guard

Legend: (I)=Investigations database; (MJ)=military justice database; (P)=personnel database; N/A= not available.
Source: GAO analysis of each military service’s investigations, military justice, and personnel database information. | GAO-19-344

Although the data collected and maintained was not consistent within and
across the military services, each of the military services’ databases
maintained race and ethnicity data for at least 99 percent of the
servicemembers, with the exception of the Coast Guard. 52 The Coast
Guard does not track information about race or ethnicity in its military
justice database. 53 Coast Guard officials stated that this is because Law
Manager was designed to determine the status of court-martial cases,
and captures attributes that are needed to generate relevant UCMJ
documents, such as court pleadings. Demographic information such as
race and ethnicity is not included in these official documents, so this
information is not input into Law Manager. Further, four of the databases
we reviewed—including both of the Army’s military justice databases, and
the Navy and the Marine Corps’ military justice databases—collect
information on race and ethnicity in a combined data field as shown in
table 4, whereas the other databases collect and maintain race and
ethnicity information in two separate fields.
Standards for Internal Control in the Federal Government states that
management should use quality information to achieve the entity’s
objectives. 54 Among other things, attributes of this internal control
52

According to officials from all of the military services, the information about race and
ethnicity in their databases is self-reported by individual servicemembers, and there is no
way to verify whether the reported information is accurate.

53

The military services differ regarding whether their databases require the collection of
information about race and ethnicity.

54

GAO-14-704G.

Page 31

GAO-19-344 Military Justice

principle call for management to identify information requirements; obtain
relevant data from reliable sources that are reasonably free from error;
ensure that the data it receives is timely and reliable; and process the
data obtained into quality information— information that is appropriate,
current, complete, and accurate. In addition, federal internal control
standards call for management to design the entity’s information system
and related control activities to achieve objectives and respond to risks,
thereby enabling information to become available to the entity on a
timelier basis. Further, the Military Justice Act of 2016 required the
Secretary of Defense to prescribe uniform standards and criteria for
various items, including data collection and analysis for case
management at all stages of the military justice system, including pretrial,
trial, post-trial, and appellate processes, by December 2018. 55
On December 17, 2018, the General Counsel of the Department of
Defense issued the uniform standards and criteria required by article
140a of the Military Justice Act of 2016. 56 As part of these uniform
standards, the services were directed to collect data related to race and
ethnicity in their military justice databases, and to collect racial and ethnic
data in separate data fields. The standards provide that the services may
have their military justice databases capture expanded ethnic or racial
categories; however, for reporting purposes, expanded categories will
aggregate to those categories listed in the standards. For race, the
services will choose from six designations: (1) American Indian/Alaska
Native, (2) Asian, (3) Black or African American, (4) Native Hawaiian or
Other Pacific Islander, (5) White, or (6) Other. For ethnicity, the services
will choose from two options: (1) Hispanic or Latino, or (2) Not Hispanic or
Latino. These categories are consistent with the OMB standards for
collecting and presenting such data. The military services are to
implement the Secretary’s direction no later than December 23, 2020.
However, DOD has applied these newly issued standards only to the
military justice databases and not to the investigations and personnel
databases. DOD officials stated that the investigations and personnel
databases do not fall under the charter of the DOD General Counsel,
55

Military Justice Act of 2016, passed as part of the National Defense Authorization Act for
Fiscal Year 2017, Pub. L. No. 114-328, §§5001-5542 (Dec. 23, 2016).

56

The Coast Guard is a voting member of the Joint Service Committee on Military Justice,
and participated in the Joint Service Committee’s subcommittee that developed the
recommendations leading to the issuance of these standards. Coast Guard officials told
us that they consider these standards to be binding on the Coast Guard.

Page 32

GAO-19-344 Military Justice

which issued the standards for the military justice databases. Hence,
these uniform standards do not apply to the military services’
investigations and personnel databases. We were able to analyze data
across the investigations, military justice, and personnel databases by
merging data from these databases, but this took multiple, detailed steps
and would not be an efficient approach for routine analyses. Taking steps
to develop the capability to present the race and ethnicity data in the
military services’ personnel and investigations databases using the same
categories included in the December 2018 standards for the military
justice databases would allow for more efficient analysis of consistent
demographic data. This could be done through either collecting and
maintaining race and ethnicity data in the investigations and personnel
databases using the December 2018 uniform standards or developing a
capability to aggregate the data into the race and ethnicity categories
included in the standards.

The Navy, the Marine
Corps, and the Coast
Guard Did Not Collect and
Maintain Complete
Servicemember
Identification Data, but
Improved Collection Has
Been Directed

The Navy, the Marine Corps, and the Coast Guard did not collect and
maintain complete servicemember identification data, such as social
security number or employee identification number, in their military justice
or investigations databases; however, DOD recently directed them to do
so. 57 In the course of conducting our analysis, in some instances, we
could not match personnel records with military justice records because
the social security number or employee identification number in the
military justice database did not match the information in the personnel
database. In other instances, we could not match personnel records with
military justice records because the military justice records did not contain
a social security number or employee identification number to match with
information found in their personnel record. As shown in table 5, we
initially were unable to match 5 percent of Navy military justice cases, 12
percent of Marine Corps military justice cases, 18 percent of Coast Guard
investigation cases, and 6 percent of Coast Guard military justice cases.
57

While the data we received from the Army and the Air Force did not have
servicemember identification numbers for 100 percent of their cases, their data were
sufficiently complete for our purposes. The DOD employee identification number is known
as the electronic data interchange personal identifier (EDIPI). To conduct our disparities
analyses, we merged an accused servicemember’s personnel database records with their
investigations or military justice database records, to ensure that we had consistent
profiles for the race, ethnicity, and gender of servicemembers investigated, prosecuted,
and charged with a violation of the UCMJ. We merged the records by matching a
servicemember’s unique identifiers, such as social security number or employee
identification number, in the personnel databases with their social security number or
employee identification number in the investigations or military justice database records.

Page 33

GAO-19-344 Military Justice

Table 5: Overview of Unmatched Military Justice Cases by Military Service and Database
Military service/database

Total number
of cases

Number of
unmatched cases

Navy military justice
database

4,809

246

5%

216

4%

Marine Corps military
justice database

6,875

848

12%

188

3%

Coast Guard investigative
database

1,428

250

18%

23

2%

289

16

6%

2

1%

Coast Guard military
justice database

Missing Number of unmatched cases
rate
after manual lookup

Missing rate after
manual lookup

Source: GAO analysis of services’ personnel, investigations, and military justice data. | GAO-19-344

Note: We provided the services with lists of our unmatched cases, and service officials manually
looked up these data so that we could increase our match rates and complete our analyses.

On December 17, 2018, the General Counsel of the Department of
Defense issued the uniform standards and criteria required by article
140a of the Military Justice Act of 2016. 58 As part of these uniform
standards, the services were directed to collect either the social security
number or DOD identification number in their military justice databases.
The military services are to implement the Secretary’s direction no later
than December 23, 2020.

The Military Services Do
Not Consistently Report
Data that Provides
Visibility into Any
Disparities, and DOD Has
Not Identified When
Disparities Should Be
Examined Further

Although some military services report demographic information about the
subjects of military justice actions internally, the military services do not
externally report data that provides visibility into, or would enable an
analysis of, the extent of racial, ethnic, or gender disparities in the military
justice system. Service officials from all of the military services told us that
they compile internal quarterly or monthly staff judge advocate reports,
which include the total number of each type of court-martial handled by
their legal offices and of nonjudicial punishments. According to service
officials, in the Air Force and the Army these reports include demographic
information about servicemembers involved in these cases, such as the
total number of each type of case broken out by the subject’s race,
ethnicity, or gender, but the Navy, Marine Corps, and Coast Guard
58

The Coast Guard is a voting member of the Joint Service Committee on Military Justice,
and participated in the Joint Service Committee’s subcommittee that developed the
recommendations leading to the issuance of these standards. Coast Guard officials told
us that they consider these standards to be binding for the Coast Guard.

Page 34

GAO-19-344 Military Justice

reports do not include this demographic information, and there is no
requirement to do so.
Regarding external reporting, the UCMJ directs the Court of Appeals for
the Armed Forces, the Judge Advocates General, and the Staff Judge
Advocate to the Commandant of the Marine Corps to submit annual
reports on the military justice system to the Congressional Armed
Services Committees, the Secretary of Defense, the secretaries of the
military departments, and the Secretary of Homeland Security. 59 These
reports are to include information on the number and status of pending
cases handled in the preceding fiscal year, among other information. The
annual reports include the total number of cases each service handled for
each type of court-martial and for nonjudicial punishments. However,
these annual reports do not include demographic information about
servicemembers who experienced a military justice action, such as
breakdowns by race or gender, because the reporting requirement does
not direct the services to include such information. A DOD official
expressed concern about expanding the reporting requirement to have
public dissemination of race, ethnicity, and gender information due to the
potential for misinterpretation, but stated that such reporting requirements
for internal use would be beneficial. However, Congress and members of
the public have expressed an interest in this information.
Standards for Internal Control in the Federal Government state that
management should externally communicate the necessary quality
information to achieve the entity’s objectives. 60 Furthermore, these
standards state that management should use quality information to make
informed decisions and evaluate the entity’s performance. According to
DOD guidance, the Joint Service Committee on Military Justice, a
committee comprised of representatives from each service’s legal office,
is responsible for reviewing the Manual for Courts-Martial and the UCMJ
on an annual basis. The Joint Service Committee can consider suggested
changes to the UCMJ or the Manual for Courts-Martial or its
59

The reporting requirement for information about the number and status of pending cases
is in UCMJ Article 146a, and requires different reports from each of the services. The
Military Justice Act of 2016 amended this reporting requirement as of June 8, 2018. The
previous requirement, which had been in UCMJ Article 146 required one combined annual
report. The Judge Advocates General and the Staff Judge Advocate to the Commandant
of the Marine Corps are the senior officials within each service responsible for the overall
supervision and administration of military justice within their respective services.

60

GAO-14-704G.

Page 35

GAO-19-344 Military Justice

supplementary materials from the services or from the general public. The
Joint Service Committee then determines whether to propose any desired
amendments to the UCMJ, or the Manual for Courts-Martial or its
supplementary materials. 61 If the Joint Service Committee finds that an
amendment to either the Manual for Courts-Martial or the UCMJ is
required, the committee will provide the General Counsel of DOD with a
draft executive order containing the recommended amendments or will
forward a legislative proposal to amend the UCMJ. While it is unclear
whether the committee has ever considered or proposed an amendment
to the UCMJ or Manual for Courts-Martial that would require the external
reporting on an annual basis of demographic information about the race,
ethnicity, and gender of servicemembers charged with violations of the
UCMJ, no such change has been made. Reporting this information would
provide servicemembers and the public with greater visibility into potential
disparities and help build confidence that DOD is committed to a military
justice system that is fair and just.
Furthermore, DOD has not issued guidance that establishes criteria to
specify when any data indicating possible racial, ethnic, or gender
disparities in the investigations, trials, or outcomes of cases in the military
justice system should be further reviewed, and to describe what steps
should be taken to conduct such a review if it were needed. GAO’s
Standards for Internal Control in the Federal Government provides that an
agency needs to establish a baseline in order to perform monitoring
activities. 62 The baseline helps the agency understand and address
deficiencies in its operations.
While equal employment opportunity enforcement is a very different
context than the military justice system, other federal agencies have
developed such criteria in the equal employment opportunity context that
can indicate when disparities should be examined further. For example,
the Department of Justice, the Department of Labor, the Equal
Employment Opportunity Commission, and the Office of Personnel
Management use a “four-fifths” test to determine when differences
between subgroups in the selection rates for hiring, promotion, or other

61

DOD Instruction 5500.17, Role and Responsibilities of the Joint Service Committee on
Military Justice (JSC) (Feb. 21, 2018).

62

GAO-14-704G.

Page 36

GAO-19-344 Military Justice

employment decisions are significant. 63 These criteria, though inexact,
provide an example of the type of criteria that DOD could consider using
as a basis for determining when disparities among racial or gender
groups in the military justice process could require further review or
analysis. By issuing guidance that establishes criteria for determining
when data indicating possible racial and gender disparities in the
investigations, trials, or outcomes of cases in the military justice system
should be further examined, and describes the steps that should be taken
to conduct such further examination, DOD and the services would be
better positioned to monitor the military justice system to help ensure that
it is fair and just, a key principle of the UCMJ.

63

According to the Equal Employment Opportunity Commission, under the four-fifths test,
a selection rate for any race, sex, or ethnic group that is less than four-fifths or 80 percent
of the rate for the group with the highest selection rate will be regarded as substantially
different. This is considered a rule of thumb and not a legal definition, but is considered a
practical means of keeping the attention of enforcement agencies on discrepancies. It
establishes a numerical basis for drawing an initial inference and requiring additional
information. Equal Employment Opportunity Commission, Uniform Guidelines on
Employee Selection Procedures, 44 Fed. Reg. 11,996 (Mar. 2, 1979).

Page 37

GAO-19-344 Military Justice

Racial and Gender
Disparities Exist in
Military Justice
Investigations,
Disciplinary Actions,
and Case Outcomes,
but Have Not Been
Comprehensively
Studied to Identify
Causes

Racial and gender disparities exist in investigations, disciplinary actions,
and punishment of servicemembers in the military justice system, and
gender disparities exist in convictions in the Marine Corps. Our analysis
of available data from fiscal years 2013 through 2017, which controlled for
attributes such as race, gender, rank, education, and years of service,
found racial and gender disparities were more likely in actions that first
brought servicemembers into the military justice system. 64 Specifically, we
found that:
•

Black, Hispanic, and male servicemembers were more likely than
White and female servicemembers to be the subjects of recorded
investigations in all of the military services, and were more likely to be
tried in general and special courts-martial in the Army, the Navy, the
Marine Corps, and the Air Force. 65

•

There were fewer statistically significant racial and gender disparities
in most military services in general and special courts-martial that
were preceded by a recorded investigation than in general and special
courts-martial overall. We also found that statistically significant racial

64

Our findings of racial and gender disparities, 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. We conducted multivariate
regression analyses, which analyzed the degree to which one racial or gender group was
more likely or less likely than another racial or gender group to be the subject of recorded
investigations while controlling for race, gender, rank, and education. In the Air Force, we
also controlled for years of service among the lower enlisted ranks (E1-E4) at the request
of Air Force officials. In the Army, we could not control for education, but we were able to
control for age. A multivariate regression analysis examines several variables
simultaneously to estimate whether each of these variables are more likely or less likely to
be associated with a certain outcome. A multivariate regression analysis allows us to test
the association between a servicemember’s race and the odds of a particular military
justice action, while holding other servicemember attributes, such as rank, education, and
gender, constant. For the purposes of consistency, in our multivariate regression
analyses, we made all racial comparisons with White servicemembers and all gender
comparisons with female servicemembers as the reference categories. For purposes of
this report, we use the term “likelihood” when discussing the odds ratios from the results of
our regression analyses. Odds ratios that are statistically significant and greater than 1.00
or lower than 1.00 indicate that individuals with that characteristic are more likely or less
likely, respectively, to be subject to a particular military justice action. See Appendix I for a
more detailed explanation of how we conducted our multivariate regression analysis, and
a full explanation of the attributes we used in each service model. In addition, see
Appendix II for the summary statistics and bivariate regression analyses for the racial and
gender groups in each of the services, and Appendixes IV through VIII for the
demographic breakdowns of the modeled attributes in each of the military services.

65

We did not analyze general and special courts-martial in the Coast Guard due to the
small number of cases adjudicated from fiscal years 2013 through 2017.

Page 38

GAO-19-344 Military Justice

and gender disparities in general and special courts-martial that did
not follow a recorded investigation were similar to those we identified
for general and special courts-martial overall.
•

Black and male servicemembers were more likely than White and
female servicemembers to be tried in summary courts-martial and to
be subjects of nonjudicial punishment in the Air Force and the Marine
Corps. The Army and the Navy did not maintain complete data, and
the Coast Guard had too few summary courts-martial for us to
analyze, and did not maintain complete nonjudicial punishment data.

We identified fewer statistically significant racial or gender disparities in
case outcomes—convictions and punishment severity. Specifically:
•

Race was not a statistically significant factor in the likelihood of
conviction in general and special courts-martial in the Army, the Navy,
the Marine Corps, and the Air Force, but gender was a statistically
significant factor in the Marine Corps.

•

Black servicemembers were less likely to receive a more severe
punishment in general and special courts-martial compared to White
servicemembers in the Navy but there was no statistically significant
difference for Black servicemembers in the Marine Corps, the Army,
and the Air Force. Additionally, there were no statistically significant
differences for Hispanic servicemembers in the Navy, the Marine
Corps, the Army, or the Air Force; and males were more likely than
females to receive a more severe punishment in the Marine Corps,
the Army, and the Air Force.

Finally, DOD and the military services have taken some steps to study
racial and gender disparities in the military justice system over the last
several decades, but they have not comprehensively studied the extent or
causes of any disparities.

Page 39

GAO-19-344 Military Justice

Black, Hispanic, and Male
Servicemembers Were
More Likely to Be Subjects
of Recorded Investigations
and Tried in General and
Special Courts-Martial
Black, Hispanic, and Male
Servicemembers Were More
Likely to Be Subjects of
Recorded Investigations in All
of the Military Services

Black, Hispanic, and male servicemembers were more likely than White
or female servicemembers to be the subjects of recorded investigations in
all of the military services, after controlling for other attributes, as shown
in figure 5. 66 Servicemembers in the Other race category were more likely
than White servicemembers to be the subjects of recorded investigations
in the Navy, but were less likely in the Army. 67 Our analyses did not
identify any statistically significant differences for servicemembers in the
Other race category from the Air Force, the Marine Corps, or the Coast
Guard. 68

66

These analyses, taken alone, should not be used to make conclusions about the
presence or absence of unlawful discrimination. Our analysis focused on violations of the
UCMJ that were recorded in databases used by service-specific investigative entities
known as military criminal investigative organizations (MCIO). MCIOs conduct criminal
investigations in cases with a DOD nexus, such as if a crime occurred on a DOD
installation, or the subject of the investigation is currently affiliated with DOD or was
subject to the UCMJ at the time of the offense. Investigations are recorded in the MCIO
databases when a servicemember is the subject of a criminal allegation made by another;
for purposes of this report, we say the servicemember had a “recorded investigation” to
describe these cases. We used data from the databases used by the Army’s Criminal
Investigation Command, which included cases investigated by military police and Criminal
Investigation Command; by the Navy and Marine Corps Naval Criminal Investigative
Service, which included cases investigated by the Naval Criminal Investigative Service
and military police; by the Air Force’s Office of Special Investigations, which included only
Office of Special Investigations cases; and by the Coast Guard Investigative Service,
which included only Coast Guard Investigative Service cases. This analysis does not
include investigations that were recorded in databases that were not used by the MCIOs,
or other investigations conducted within the military, such as command investigations.
When we merged the military services’ investigation data with their personnel information,
we found the data to be sufficiently reliable for assessing gender and racial disparities.
See Appendix III for information regarding recorded investigations of drug and sexual
assault offenses.

67

The Other race category includes servicemembers that identify as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

68

Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case, a
recorded investigation.

Page 40

GAO-19-344 Military Justice

Figure 5: 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

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be the subject of an investigation recorded in the services’ military
criminal investigative organizations databases for alleged violations of the Uniform Code of Military
Justice after controlling for race, gender, rank, and education. We also controlled for years of service
among the lower enlisted ranks (E1-E4) in the Air Force. In the Army, we could not control for
education, but we were able to control for age. We made all racial comparisons with White
servicemembers and all gender comparisons with female servicemembers as the reference
categories. Odds ratios that are statistically significant (p-value < 0.05) and greater than 1.00 or lower
than 1.00 indicate that individuals with that characteristic are more likely or less likely, respectively, to
be the subject of a recorded investigation. Not statistically significant means that we could not
conclude there was an association between race and the likelihood of a recorded investigation. The
Other race category includes individuals who identified as American Indian/Alaska Native, Asian,
Native Hawaiian/Other Pacific Islander, and multiple races. The racial breakdowns in this figure may

Page 41

GAO-19-344 Military Justice

not sum to 100 percent because we excluded from this figure data for White servicemembers and
those whose race was unknown.

Black, Hispanic, and Male
Servicemembers Were More
Likely to Be Tried in General
and Special Courts-Martial

For the Army, the Navy, the Marine Corps, and the Air Force, Black,
Hispanic, and male servicemembers were more likely than White and
female servicemembers to be tried in general and special courts-martial
after controlling for other attributes, as shown in figure 6 below. 69
Servicemembers in the Other race category were more likely than White
servicemembers to be tried in general and special courts-martial in the
Navy, but we found no statistically significant differences in the likelihood
of servicemembers in the Other race category in the Army, the Marine
Corps, and the Air Force to be tried in general and special courts-martial
compared to White servicemembers. 70 We could not analyze Coast
Guard cases due to the small number of general and special courtsmartial adjudicated in the Coast Guard from fiscal years 2013 through
2017.

69

These analyses, taken alone, should not be used to make conclusions about the
presence or absence of unlawful discrimination. When we merged the military services’
military justice data with their personnel information, we found the data to be sufficiently
reliable for assessing gender and racial disparities. General and special courts-martial are
used to adjudicate more serious violations of the UCMJ, and therefore have the potential
for more severe judicial punishment. We conducted multivariate regression analyses,
which analyzed the degree to which one racial, ethnic, or gender group was more likely or
less likely than another racial, ethnic, or gender group to be tried in general and special
courts-martial while controlling for race, gender, rank, and education. In the Air Force, we
also controlled for years of service among the lower enlisted ranks (E1-E4). In the Army,
we could not control for education, but we were able to control for age. A multivariate
regression analysis examines several variables to estimate whether each of these
variables are more likely or less likely to be associated with a certain outcome. See
Appendix I for a more detailed explanation of how we conducted our multivariate
regression analysis, and a full explanation of the attributes we used in each service model.
In addition, see Appendix II for the summary statistics and bivariate regression analyses
for the racial and gender groups in each of the services, and see Appendixes IV through
VIII for the demographic breakdowns of the modeled attributes in each of the military
services. See Appendix III for information regarding general and special courts-martial of
drug and sexual assault offenses.

70

Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case, trial
in general and special courts-martial.

Page 42

GAO-19-344 Military Justice

Figure 6: Likelihood of Trial in General and Special Courts-Martial by Race and Gender, After Controlling for Rank and
Education, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be tried in general and special courts-martial after controlling for race,
gender, rank, and education. We also controlled for years of service among the lower enlisted ranks
(E1-E4) in the Air Force. In the Army, we could not control for education, but we were able to control
for age. We made all racial comparisons with White servicemembers and all gender comparisons with
female servicemembers as the reference categories. Odds ratios that are statistically significant (pvalue < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic
are more likely or less likely, respectively, to be tried in general or special courts-martial. Not
statistically significant means that we could not conclude there was an association between race and
the likelihood of trial in general and special courts-martial. The Other race category includes
individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific
Islander, and multiple races. The racial breakdowns in this figure may not sum to 100 percent
because we excluded from this figure data for White servicemembers and those whose race was
unknown.

Page 43

GAO-19-344 Military Justice

More Statistically
Significant Racial and
Gender Disparities Found
in General and Special
Courts-Martial Cases
without a Recorded
Investigation than with a
Recorded Investigation

When separating general and special court-martial cases into those that
either were or were not preceded by an investigation recorded in an
MCIO database, we found fewer statistically significant racial and gender
disparities in most of the military services in general and special courtsmartial that were preceded by a recorded investigation. 71 However,
statistically significant racial and gender disparities were also present in
general and special courts-martial that did not follow a recorded
investigation in all services included in this analysis, which would include
cases where the investigation was performed by the servicemember’s
command.
Specifically, as shown in figure 7 below, we found that Black, Hispanic,
Other, and male servicemembers in the Army, Hispanic servicemembers
in the Marine Corps, and males in the Air Force were more likely than
White or female servicemembers to be tried in general and special courtsmartial following a recorded investigation, after controlling for other
attributes. 72 We found no statistically significant differences in the
likelihood of any other racial or gender groups to be tried in general and
special courts-martial following a recorded investigation in any other
services. 73 Our analyses of general and special courts-martial with a
71

Investigations are recorded in the MCIO databases when a servicemember is the
subject of a criminal allegation made by another; for purposes of this report, we say the
servicemember had a “recorded investigation” to describe these cases. For additional
explanation of the databases we used to analyze investigations, please see appendix I. As
discussed in the background section above, and in figure 3, the majority of general and
special courts-martial, ranging from 53 percent to 74 percent, had a recorded
investigation, while the remaining general and special courts-martial cases, ranging from
26 percent to 47 percent, would have been investigated by other sources, such as local
civilian law enforcement, command investigations, or in the case of the Air Force, their
military law enforcement forces.

72

We conducted multivariate regression analyses, which analyzed the degree to which
one racial, ethnic, or gender group was more likely or less likely than another racial,
ethnic, or gender group to be tried in general and special courts-martial that followed a
recorded investigation while controlling for race, gender, rank, and education. In the Air
Force, we also controlled for years of service among the lower enlisted ranks (E1-E4). In
the Army, we could not control for education, but we were able to control for age. A
multivariate regression analysis examines several variables to estimate whether each of
these variables are more likely or less likely to be associated with a certain outcome. See
Appendix I for a more detailed explanation of how we conducted our multivariate
regression analysis, and a full explanation of the attributes we used in each service model.
In addition, see Appendixes II through VI for the demographic breakdowns of the modeled
attributes in each of the military services.

73

Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case, trial
in general and special courts-martial following a recorded investigation.

Page 44

GAO-19-344 Military Justice

recorded investigation generally found fewer statistically significant
differences compared to the results of our analyses for all special and
general courts martial.
Figure 7: Likelihood of Trial in General and Special Courts-Martial Following a Recorded Investigation by Race and Gender,
After Controlling for Rank and Education, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be tried in general and special courts-martial following an investigation
recorded in the services’ military criminal investigative organizations databases after controlling for
race, gender, rank, and education. We also controlled for years of service among the lower enlisted
ranks (E1-E4) in the Air Force. In the Army, we could not control for education, but we were able to
control for age and investigative entity. In the Navy and the Marine Corps, we also controlled for type
of offense, investigative entity, and composition of the deciding panel. Odds ratios that are statistically
significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that
characteristic are more likely or less likely, respectively, to be tried in general and special courtsmartial following a recorded investigation. Not statistically significant means that we could not
conclude there was an association between race and the likelihood of trial in general and special
courts-martial following a recorded investigation. We made all racial comparisons with White
servicemembers and all gender comparisons with female servicemembers as the reference
categories. The Other race category includes individuals who identified as American Indian/Alaska
Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races. The racial breakdowns in
this figure may not sum to 100 percent because we excluded from this figure data for White
servicemembers and those whose race was unknown.

Page 45

GAO-19-344 Military Justice

We also found that Black and male servicemembers in all of the military
services were more likely than White and female servicemembers to be
tried in general and special courts-martial without a recorded investigation
after controlling for other attributes, as shown in figure 8 below. Further,
Hispanic servicemembers in the Army were more likely than White
servicemembers to be tried in general and special courts-martial without a
recorded investigation, but we found no statistically significant differences
in the likelihood of Hispanic servicemembers to be tried in general and
special courts-martial without a recorded investigation in the Marine
Corps, the Navy, or the Air Force. We found no statistically significant
differences in the likelihood of servicemembers in the Other race category
to be tried in general and special courts-martial compared to White
servicemembers in all of the military services. Our findings of racial and
gender disparities in general and special courts-martial without a
recorded investigation found statistically significant differences for Black
and male servicemembers consistent with the differences we identified for
general and special courts-martial overall, as shown in figure 6 above.

Page 46

GAO-19-344 Military Justice

Figure 8: Likelihood of Trial in General and Special Courts-Martial without a Recorded Investigation by Race and Gender,
After Controlling for Rank and Education, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be tried in general and special courts-martial without an investigation
recorded in the services’ military criminal investigative organizations databases after controlling for
race, gender, rank, and education. We also controlled for years of service among the lower enlisted
ranks (E1-E4) in the Air Force. In the Army, we could not control for education, but we were able to
control for age. Odds ratios that are statistically significant (p-value < 0.05) and greater than 1.00 or
lower than 1.00 indicate that individuals with that characteristic are more likely or less likely,
respectively, to be tried in general and special courts-martial without a recorded investigation. Not
statistically significant means that we could not conclude there was an association between race and
the likelihood of trial in general and special courts-martial without a recorded investigation. We made
all racial comparisons with White servicemembers and all gender comparisons with female
servicemembers as the reference categories. The Other race category includes individuals who
identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. The racial breakdowns in this figure may not sum to 100 percent because we
excluded from this figure data for White servicemembers and those whose race was unknown.

Page 47

GAO-19-344 Military Justice

Black and Male
Servicemembers Were
More Likely to Be Subject
to Summary CourtsMartial and Nonjudicial
Punishment in the Air
Force and Marine Corps,
and the Other Services
Lack Data
Black and Male
Servicemembers Were More
Likely to Be Tried in Summary
Courts-Martial in the Air Force
and Marine Corps, and the
Army and Navy Lack Data

Black and male servicemembers were more likely than White or female
servicemembers to be tried in summary courts-martial in the Air Force
and the Marine Corps after controlling for other attributes, as shown in
figure 9 below. 74 We did not identify any statistically significant differences
in summary courts-martial rates for servicemembers who identified as
Hispanic or in the Other race category in either the Air Force or the
Marine Corps. 75 We could not determine whether there were racial or
gender disparities for summary courts-martial in the Army, the Navy, and
the Coast Guard due to data limitations.

74

These analyses, taken alone, should not be used to make conclusions about the
presence or absence of unlawful discrimination. When we merged the military services’
military justice data with their personnel information, we found the data to be sufficiently
reliable for assessing gender and racial disparities. We conducted multivariate regression
analyses, which analyzed the degree to which one racial or gender group was more likely
or less likely than another racial or gender group to be tried in summary courts-martial
while controlling for race, gender, rank, and education. In the Air Force, we also controlled
for years of service among the lower enlisted ranks (E1-E4). A multivariate regression
analysis examines several variables simultaneously to estimate whether each of these
variables are more likely or less likely to be associated with a certain outcome. See
Appendix I for a more detailed explanation of how we conducted our multivariate
regression analysis, and a full explanation of the attributes we used in each service model.
In addition, see Appendix II for the summary statistics and bivariate regression analyses
for the racial and gender groups in each of the services, and see Appendixes VI and VII
for the demographic breakdowns of each of those attributes in the Marine Corps and the
Air Force, respectively.

75
Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case, trial
in summary courts-martial. The Other race category includes servicemembers who
identify as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander,
and multiple races.

Page 48

GAO-19-344 Military Justice

Figure 9: Likelihood of Trial in Summary Courts-Martial in the Air Force and the Marine Corps by Race and Gender, After
Controlling for Rank and Education, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be tried in summary courts-martial after controlling for race, gender,
rank, and education. We also controlled for years of service among the lower enlisted ranks (E1-E4)
in the Air Force. We made all racial comparisons with White servicemembers and all gender
comparisons with female servicemembers as the reference categories. Odds ratios that are
statistically significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that
individuals with that characteristic are more likely or less likely, respectively, to be tried in summary
courts-martial. Not statistically significant means that we could not conclude there was an association
between race and the likelihood of trial in summary courts-martial. The Other race category includes
individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific
Islander, and multiple races. The racial breakdowns in this figure may not sum to 100 percent due to
rounding and because we excluded from this figure data for White servicemembers and those whose
race was unknown.

We could not analyze Coast Guard cases due to the small number of
summary courts-martial adjudicated in the Coast Guard from 2013
through 2017. We could not determine whether disparities existed among
servicemembers tried in summary courts-martial in the Army and the
Navy because the Army and the Navy did not collect complete summary
courts-martial data in their investigations, military justice, or personnel
databases. Specifically, as part of our data reliability checks, we identified
the total number of summary courts-martial that the Army and the Navy
reported in the Court of Appeals for the Armed Forces annual reports for
fiscal years 2013 through 2017, and compared these totals to the number

Page 49

GAO-19-344 Military Justice

of cases we identified in their military justice databases. 76 While our
comparisons are not exact, due to differences in the dates we used to
count the number of cases, we found that approximately 60 percent of the
Army’s reported summary courts-martial cases and less than 50 percent
of the Navy’s reported summary courts-martial cases were included in
their military justice databases. 77
Army and Navy officials cited several reasons why complete summary
courts-martial information was not collected. First, they said that the
services are not required to collect and maintain complete data on
summary courts-martial because these cases result in non-criminal
convictions under the UCMJ. Summary courts-martial are typically used
for minor offenses, and the accused is not guaranteed the right to be
represented by a military attorney. As a result, military attorneys may not
be involved in summary courts-martial. Army and Navy officials said that if
military attorneys are not involved in the case, there is not likely to be a
record of the case in their service’s military justice database. In contrast,
Air Force officials said that they provide a military attorney to represent
the accused in summary courts-martial; as a result, Air Force officials said
their attorneys create records for these cases in the Air Force’s military
justice database. The Marine Corps does not maintain summary courtmartial data in its military justice database but tracks summary courtsmartial in its personnel database.
Officials in the Navy and the Army told us that the lack of complete
summary court-martial data in their military justice databases is also in
part because these systems were not designed to serve as repositories
for complete military justice data. Instead, the officials said that the
military justice databases were primarily created to assist attorneys in
76

According to Army and Navy officials, the total numbers of summary courts-martial
included in the Court of Appeals for the Armed Forces annual reports are taken from their
internal monthly and quarterly staff judge advocate reports that were discussed earlier in
this report.

77

We could not compare the total number of cases that we identified in the military justice
databases precisely against the reported number of cases because we counted cases
based on the date of preferral, whereas the cases reported in the Court of Appeals for the
Armed Forces annual report are based on the judgment date. However, we combined the
total number of cases over a 5-year period, which made differences in which particular
fiscal year a case was counted less important for these purposes. We found that while the
total number of cases were different, the totals we computed provided a basis for
comparison that allowed us to confirm that the military justice databases did not have
complete data about summary courts-martial, as Army and Navy officials had told us.

Page 50

GAO-19-344 Military Justice

generating trial documents, meeting timeframes, and other aspects of
case management. Nevertheless, Army officials said they plan to start
collecting more complete summary court-martial information. Specifically,
Army officials said that the Army is encouraging their judge advocate
general staff to create records for all summary courts-martial in the
service’s military justice database.
The absence of complete summary court-martial data in the military
justice databases of the Army and the Navy limits these services’ visibility
into any disparities that may exist among servicemembers involved in
these types of military justice proceedings. On December 17, 2018, the
General Counsel of the Department of Defense issued the uniform
standards and criteria required by article 140A of the Military Justice Act
of 2016. As part of these uniform standards, the services were directed to
collect certain information about all cases in their military justice
databases, which a DOD official said includes summary courts-martial
cases. The military services are to implement the Secretary’s direction no
later than December 23, 2020.

Page 51

GAO-19-344 Military Justice

Black and Male
Servicemembers Were More
Likely to Be Subject to
Nonjudicial Punishments in the
Air Force and the Marine
Corps, and the Army, Navy,
and Coast Guard Lack Data

Black and male servicemembers were more likely than White or female
servicemembers to be subject to nonjudicial punishments in the Air Force
and the Marine Corps, after controlling for other attributes, as shown in
figure 10 below. 78 In the Air Force, we found that Hispanic
servicemembers were more likely than White servicemembers to receive
nonjudicial punishments, while we observed no statistically significant
differences in nonjudicial punishment rates for Hispanic servicemembers
in the Marine Corps. 79 Servicemembers in the Other race category in the
Marine Corps were less likely to receive nonjudicial punishments, but we
observed no statistically significant differences in nonjudicial punishment
rates for servicemembers in the Other race category in the Air Force. 80

78

These analyses, taken alone, should not be used to make conclusions about the
presence or absence of unlawful discrimination. When we merged the military services’
military justice data with their personnel information, we found the data to be sufficiently
reliable for assessing gender and racial disparities. We conducted multivariate regression
analyses, which analyzed the degree to which one racial or gender group was more likely
or less likely than another racial or gender group to receive nonjudicial punishments while
controlling for race, gender, rank, and education. In the Air Force, we also controlled for
years of service among the lower enlisted ranks (E1-E4). A multivariate regression
analysis examines several variables simultaneously to estimate whether each of these
variables are more likely or less likely to be associated with a certain outcome. See
Appendix I for a more detailed explanation of how we conducted our multivariate
regression analysis, and a full explanation of the attributes we used in each service model.
In addition, see Appendix II for the summary statistics and bivariate regression analyses
for the racial and gender groups in each of the services, and see Appendixes VI and VII
for the demographic breakdowns of each of those attributes in the Marine Corps and the
Air Force, respectively.

79

Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case,
nonjudicial punishment.

80

The Other race category includes servicemembers who identify as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 52

GAO-19-344 Military Justice

Figure 10: Likelihood of Nonjudicial Punishments in the Air Force and the Marine Corps by Race and Gender, After
Controlling for Rank and Education, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be subject to nonjudicial punishments after controlling for race, gender,
rank, and education. We also controlled for years of service among the lower enlisted ranks (E1-E4)
in the Air Force. We made all racial comparisons with White servicemembers and all gender
comparisons with female servicemembers as the reference categories. Odds ratios that are
statistically significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that
individuals with that characteristic are more likely or less likely, respectively, to be subject to
nonjudicial punishment. Not statistically significant means that we could not conclude there was an
association between race and the likelihood of nonjudicial punishment. The Other race category
includes individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other
Pacific Islander, and multiple races. The racial breakdowns in this figure may not sum to 100 percent
due to rounding and because we excluded from this figure data for White servicemembers and those
whose race was unknown.

However, we could not determine whether there were racial or gender
disparities among servicemembers subject to nonjudicial punishments in
the Army, the Navy, and the Coast Guard because these services do not
collect complete nonjudicial punishment data, such as data on the
servicemember’s race, ethnicity, gender, offense, and punishment, in any
of their databases. As part of our data reliability checks, we identified the
total number of nonjudicial punishments that the Army, the Navy, and the
Coast Guard reported in the Court of Appeals for the Armed Forces
annual reports for fiscal years 2013 through 2017, and compared these
totals to the number of cases we identified in their military justice and

Page 53

GAO-19-344 Military Justice

personnel databases. 81 As shown in figure 11 below, we found that 65
percent of the Army’s reported nonjudicial punishments, 8 percent of the
Navy’s reported nonjudicial punishments, and 82 percent of the Coast
Guard’s reported nonjudicial punishments were recorded in their military
justice databases.
Figure 11: Army, Navy, and Coast Guard Reported Nonjudicial Punishments Compared to Nonjudicial Punishments in Military
Justice Databases, Fiscal Years 2013–2017

Note: Nonjudicial punishments are reported as a combined number for the Navy and the Marine
Corps in the Court of Appeals for the Armed Forces annual reports. To calculate this reported figure
for the Navy, we subtracted the number of Marine Corps nonjudicial punishment cases we identified
in the Marine Corps personnel database from the reported totals.

Officials from these services cited several reasons why they did not have
complete information about all nonjudicial punishments. First, they said
that the services are not required to track nonjudicial punishment cases
because they are non-criminal punishments that are typically imposed for
less serious offenses. Army and Navy officials noted that complete
records of these punishments are not recorded at least in part because
81

Nonjudicial punishments are reported as a combined total for the Navy and the Marine
Corps in the Court of Appeals for the Armed Forces annual reports. As a result, to
calculate this reported figure for the Navy, we subtracted the number of Marine Corps
nonjudicial punishment cases that we had identified in the Marine Corps personnel
database from the reported totals.

Page 54

GAO-19-344 Military Justice

nonjudicial punishments are not meant to follow servicemembers
throughout their career, but instead are intended to incentivize
servicemembers to correct their behavior. Because nonjudicial
punishments are not criminal punishments, the process afforded to
servicemembers in nonjudicial punishment proceedings differs as well.
For example, the servicemember is not guaranteed the right to
representation by a military attorney. Army and Navy officials noted that
their military justice databases contain records of nonjudicial punishments
if there was legal involvement by the Judge Advocate General’s Corps in
the case. Similarly, Coast Guard officials said that their military justice
database contains records of nonjudicial punishment if a case originated
as a criminal case involving a judge advocate, for example, if charges
were preferred. According to Air Force and Marine Corps officials, the Air
Force maintains complete nonjudicial punishment data in its military
justice database, and the Marine Corps maintains complete nonjudicial
punishment data in its personnel database. 82
Standards for Internal Control in the Federal Government state that
management should use quality information to achieve an entity’s
objectives. Additionally, management should identify information
requirements; ensure that the data it receives are timely and reliable; and
process the data obtained into quality information. 83 Officials from the
Army, the Navy, and the Coast Guard expressed concerns regarding the
feasibility of collecting and maintaining information about all nonjudicial
punishments. Army officials stated that the collection and maintenance of
all nonjudicial punishment data would be a substantial administrative
burden due to the number of nonjudicial punishments awarded to
servicemembers every week. Navy officials also stated that it would be a
significant challenge to collect and maintain information about all
nonjudicial punishments in either the Navy’s military justice database or
its personnel database. They stated that there are few individuals who
have access and can input data into the military justice database, and to
expand the scope of criminal justice data collected in that manner, more
people would have to be hired or assigned to assist with data entry.
Similarly, Coast Guard officials said that tracking all nonjudicial
punishment cases would be a difficult addition to their current data
82

Marine Corps officials said that commanders fill out a form for all executed
administrative actions, nonjudicial punishments, and all types of courts-martial, and
information from those forms are then recorded in the personnel database.

83

GAO-14-704G.

Page 55

GAO-19-344 Military Justice

collection and maintenance workload. Coast Guard officials further stated
that in addition to providing commanders with an essential means of
providing good order and discipline, nonjudicial punishment also may
promote positive change. Some Coast Guard officials stated concerns
that recording all nonjudicial punishments in a database may inhibit the
rehabilitative component of nonjudicial punishment.
While the Army, Navy, and Coast Guard officials expressed these
concerns, none of these military services had formally assessed the
feasibility of collecting data on nonjudicial punishments. The absence of
complete nonjudicial punishment data limits the military services’ visibility
into the vast majority of legal punishments imposed on servicemembers
under the UCMJ every year. Without such data, these three services will
remain limited in their ability to assess or identify disparities among
populations subject to this type of punishment.

Page 56

GAO-19-344 Military Justice

Few Statistically
Significant Racial or
Gender Disparities Exist in
Likelihood of Conviction or
Severity of Punishment,
but the Coast Guard Does
Not Collect and Maintain
Complete Data
Race Was Not a Statistically
Significant Factor in
Convictions in General and
Special Courts-Martial, but
Gender Was in the Marine
Corps

Among the servicemembers convicted in general and special courtsmartials, we found no statistically significant differences regarding the
likelihood of conviction among racial groups in the Army, the Navy, the
Marine Corps, and the Air Force, while controlling for other attributes, as
shown in figure 12 below. 84 In the Marine Corps, male servicemembers
were more likely to be convicted compared to female servicemembers.
We found no statistically significant differences in the likelihood of
convictions between males and females in the Army, the Air Force, and
the Navy.

84

We conducted multivariate regression analyses, which analyzed the degree to which
one racial or gender group was more likely or less likely than another racial or gender
group to be convicted in general and special courts-martial, while controlling for race,
gender, education, rank, and offense type. In the Air Force, we also controlled for years of
service among the lower enlisted ranks (E1-E4) and composition of the deciding panel. In
the Army, we could not control for education, but we were able to control for age and
composition of the deciding panel. A multivariate regression analysis examines several
variables simultaneously to estimate whether each of these variables are more likely or
less likely to be associated with a certain outcome. Not identifying any statistically
significant findings means that we could not conclude there was an association between
race or gender and the likelihood of an outcome, in this case, conviction in general and
special courts-martial. See Appendix I for a more detailed explanation of how we
conducted our multivariate regression analysis, and a full explanation of the attributes we
used in each service model. In addition, see Appendix II for the summary statistics and
bivariate regression analyses for the racial and gender groups in each of the services, and
see Appendixes IV through VII for the demographic breakdowns of the modeled attributes
in each of the military services.

Page 57

GAO-19-344 Military Justice

Figure 12: Likelihood of Conviction in General and Special Courts-Martial by Race and Gender, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be convicted in general and special courts-martial after controlling for
race, gender, rank, education, and offense type. We also controlled for years of service among the
lower enlisted ranks (E1-E4) and composition of the deciding panel in the Air Force. In the Army, we
could not control for education, but we were able to control for age and composition of the deciding
panel. We made all racial comparisons with White servicemembers and all gender comparisons with
female servicemembers as the reference categories. Odds ratios that are statistically significant (pvalue < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic
are more likely or less likely, respectively, to be convicted in general and special courts-martial. Not
statistically significant means that we could not conclude there was an association between race and
the likelihood of conviction in general and special courts-martial. The Other race category includes
individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific
Islander, and multiple races. The racial breakdowns in this figure do not sum to 100 percent because
we excluded from this figure data for White servicemembers and those whose race was unknown.

Page 58

GAO-19-344 Military Justice

Minority Servicemembers
Were Either Less Likely to
Receive a More Severe
Punishment or There Were No
Differences Among Racial
Groups, but Punishment
Severity Varied by Gender
Among the Services

In the military services that maintained complete punishment data—the
Army, the Navy, the Marine Corps, and the Air Force—we found that
minority servicemembers were either less likely to receive a more severe
punishment in general and special courts-martial compared to White
servicemembers, or there were no statistically significant differences in
punishments among racial groups. 85 Our findings regarding gender varied
among the services. Male servicemembers were more likely to receive a
more severe punishment compared to females in the Marine Corps, the
Army, and the Air Force; for the Navy, we found there were no statistically
significant differences in punishments between males and females. 86
Navy and Marine Corps: Among servicemembers that were convicted in
general and special courts-martial in the Marine Corps, we found no
statistically significant differences regarding minority servicemembers
being more likely or less likely to receive a dismissal or discharge
punishment versus some other punishment, while controlling for other
attributes, as shown in figure 13 below. 87 In the Navy, among
servicemembers that were convicted in general and special courtsmartial, Black servicemembers were less likely than White
servicemembers to receive a discharge or dismissal. We found no
statistically significant differences regarding Hispanic servicemembers or
those of Other races in the Navy. In the Marine Corps, among
servicemembers that were convicted in general and special courts85

Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case,
punishment severity.

86

We measured the severity of punishments in two groups for the Navy and the Marine
Corps, and in three groups for the Air Force and the Army, which are defined in Appendix
I. We could not create a third punishment group for confinement without dismissal or
discharge for the Navy and the Marine Corps because of the small number of cases with
confinement that did not also include some sort of discharge. Based on discussions with
service officials, we determined that a sentence resulting in a dismissal or discharge was
the most severe punishment outcome.

87

We conducted multivariate regression analyses to analyze the degree to which one
racial, ethnic, or gender group was more likely or less likely than another group to receive
a more severe punishment in general and special courts-martial while controlling for race,
gender, education, rank, and offense type. A multivariate regression analysis examines
several variables simultaneously to estimate whether each of these variables are more
likely or less likely to be associated with a certain outcome. See Appendix I for a more
detailed explanation of how we conducted our multivariate regression analysis, and a full
explanation of the attributes we used in each service model. In addition, see Appendix II
for the summary statistics and bivariate regression analyses for the racial and gender
groups in each of the services, and see Appendixes V and VI for the demographic
breakdowns of the modeled attributes in the Navy and the Marine Corps.

Page 59

GAO-19-344 Military Justice

martial, male servicemembers were more likely than female
servicemembers to receive a discharge or dismissal. In the Navy, there
were no statistically significant differences in punishments between males
and females.
Figure 13: Likelihood of Dismissal or Discharge in General and Special Courts-Martial by Race and Gender in the Navy and
the Marine Corps, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to be to be dismissed or discharged after conviction in general and
special courts-martial after controlling for race, gender, rank, education, and offense type. We made
all racial comparisons with White servicemembers and all gender comparisons with female
servicemembers as the reference categories. Odds ratios that are statistically significant (p-value <
0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic are
more likely or less likely, respectively, to be dismissed or discharged after conviction in general and
special courts-martial. Not statistically significant means that we could not conclude there was an
association between race and the likelihood of dismissal or discharge after conviction in general and
special courts-martial. The Other race category includes individuals who identified as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races. The racial
breakdowns in this figure do not sum to 100 percent because we excluded from this figure data for
White servicemembers and those whose race was unknown.

Army and Air Force: We found no statistically significant differences
regarding Black or Hispanic servicemembers being more likely or less
likely to receive a more severe punishment in the Air Force or the Army,

Page 60

GAO-19-344 Military Justice

while controlling for other attributes, as shown in figure 14 below. 88 We
also found that servicemembers in the Other race group were less likely
to receive a more severe punishment compared to White
servicemembers in the Army, but punishment results for servicemembers
in the Other race group in the Air Force were not statistically significant.
Additionally, we found that male servicemembers were more likely to
receive a more severe punishment compared to female servicemembers
in the Army and the Air Force.

88

We conducted ordered logistic regression analyses to analyze the degree to which one
racial, ethnic, or gender group was more likely or less likely than another group to receive
a more severe outcome in general and special courts-martial, while controlling for race,
gender, education, rank, composition of the deciding panel, and offense type. In the Air
Force, we controlled for years of service among the lower enlisted ranks (E1-E4). In the
Army, we could not control for education, but we were able to control for age. Using the
three punishment groups listed in table 8 in Appendix I, based on discussions with service
officials, we determined that a sentence resulting in a dismissal or discharge was the most
severe punishment outcome. An ordered logistic regression is an extension of the logistic
regression model that applies to dependent variables where there are more than two
response categories. See Appendix I for a more detailed explanation of how we
conducted our ordered logistic regression analysis, and a full explanation of the attributes
we used in each service model. In addition, see Appendixes IV and VII for the
demographic breakdowns of the modeled attributes in the Army and the Air Force.

Page 61

GAO-19-344 Military Justice

Figure 14: Likelihood of More Severe Punishment in General and Special Courts-Martial by Race and Gender in the Army and
the Air Force, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial ordered logistic regression
analysis results demonstrate the degree to which a racial or gender group is more likely or less likely
than the reference category to receive a more severe punishment after conviction in general and
special courts-martial after controlling for race, gender, offense type, and composition of the deciding
panel. We also controlled for education and years of service among the lower enlisted ranks (E1-E4)
in the Air Force. In the Army, we also controlled for age and rank. We made all racial comparisons
with White servicemembers and all gender comparisons with female servicemembers as the
reference categories. Odds ratios that are statistically significant (p<0.05) and greater than 1.00 or
lower than 1.00 indicate the likelihood that individuals with that characteristic would receive a more
severe or less severe punishment, respectively, than the reference category. Not statistically
significant means that we could not conclude there was an association between race and the
likelihood of a more severe punishment after conviction in general and special courts-martial.
Punishment severity in the Air Force, ordered from most to least severe, was (3) any type of dismissal
or discharge (regardless of any confinement); (2) confinement without dismissal or discharge, and (1)
all other possible sentencing options. In the Army, it was (3) any type of dismissal or discharge or
confinement of more than 2 years, (2) confinement of less than 2 years without dismissal or
discharge, and (1) all other possible sentencing options. The Other race category includes individuals
who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. The racial breakdowns in this figure do not sum to 100 percent because we excluded
from this figure data for White servicemembers and those whose race was unknown.

Page 62

GAO-19-344 Military Justice

Coast Guard Did Not Collect
and Maintain Complete Case
Outcome Data

We could not determine disparities in case outcomes—convictions and
punishment severity—in the Coast Guard’s general and special courtsmartial for fiscal years 2013 through 2017 because the Coast Guard did
not collect and maintain complete conviction and punishment data in its
military justice database. 89 Specifically, 16 percent of all Coast Guard
cases were missing conviction and punishment data. When broken down
by court-martial type, 20 percent of general court-martial cases, 15
percent of special court-martial cases, and 4 percent of summary courtmartial cases were missing conviction and punishment data. Coast Guard
officials acknowledged that incomplete conviction and punishment data
entry is a consistent problem. They said that data entry had improved
recently. On December 17, 2018, the General Counsel of the Department
of Defense issued the uniform standards and criteria required by article
140a of the Military Justice Act of 2016. 90 As part of these uniform
standards, the services were directed to collect information about the
findings for each offense charged, and the sentence or punishment
imposed. The military services are to implement the Secretary’s direction
no later than December 23, 2020.

89

Although we could not analyze Coast Guard cases due to the small number of general
and special courts-martial adjudicated in the Coast Guard from fiscal years 2013 through
2017, case outcomes could potentially be analyzed in the Coast Guard using a longer
period of time than what we used in our review.

90

The Coast Guard is a voting member of the Joint Service Committee on Military Justice,
and according to Coast Guard officials, they participated in the Joint Service Committee’s
subcommittee that developed the recommendations leading to the issuance of these
standards. A Coast Guard official told us that they consider these standards to be binding
on the Coast Guard.

Page 63

GAO-19-344 Military Justice

DOD and the Military
Services Have Conducted
Some Assessments of
Military Justice Disparities,
but Have Not Studied the
Causes of Disparities

DOD and the military services have conducted some assessments of
disparities in the military justice system. We previously reported in 1995
on DOD studies on discrimination and equal opportunity, and found DOD
and the services conducted seven reviews of racial disparities in
discipline rates between 1974 and 1993. 91 Since our 1995 report through
2016, DOD and service assessments of military justice disparities have
been limited. Officials in the Office of Diversity, Equity and Inclusion
(ODEI) noted DOD has not conducted any department-wide assessments
of racial or gender disparities in military justice during this period. The
military services’ diversity offices also were not able to identify any
service-specific reviews of disparities in military justice.
However, the military services have some initiatives to examine and
address disparities in military justice. For example, Air Force officials said
that in May 2016, the Air Force conducted a servicewide data call to
solicit information about cases involving a challenge to a member of a
court-martial based on race or a motion for selective prosecution. The
officials said that a thorough review revealed no evidence of selective
prosecution in Air Force courts-martial. 92 In addition, the Air Force has
conducted analyses of its own military justice data. Specifically, the Air
Force routinely analyzes military justice data using a rates-per-thousand
analysis to identify whether certain demographic groups are tried by
court-martial or subject to nonjudicial punishments at higher rates than
others. 93 These Air Force analyses found that Black and male
servicemembers were more likely than White and female
servicemembers to be subject to courts-martial and nonjudicial
punishments from fiscal years 2013 through 2017, which is consistent
91

GAO/NSIAD-95-103. For example, studies conducted in the 1970s and 1980s showed
no disparities in discipline rates between Black and White servicemembers and found no
evidence that minority groups received courts-martial or nonjudicial punishment out of
proportion to certain types of violations. Studies published by the Navy and the Defense
Equal Opportunity Management Institute in the 1990s found that Black servicemembers
were overrepresented in the populations of servicemembers receiving judicial and
nonjudicial punishments. See Appendix I of GAO/NSIAD-95-103 for a summary of each of
the studies’ findings and recommendations.

92

A claim of selective prosecution is one that alleges that the decision to prosecute was
based, at least in part, on an unjustifiable standard, such as race, gender, religion, sexual
orientation, or other arbitrary classification. See, for example, United States v. Armstrong,
517 U.S. 456, 464 (1996).

93

A rates-per-thousand analysis computes the number of servicemembers within a
demographic group that are subject to a particular military justice action, divided by the
total number of servicemembers of that demographic group, multiplied by 1,000.

Page 64

GAO-19-344 Military Justice

with what we found. 94 However, the other services do not routinely
conduct such analyses.
Moreover, DOD has conducted climate surveys to address
servicemembers’ perceptions of bias. In 2013, for example, DOD
conducted service-wide equal opportunity surveys that queried
servicemembers on whether they believed they received nonjudicial
punishment or a court martial they should not have, and whether they
believed their race or ethnicity was a factor. The survey responses
showed that 1.3 percent of servicemembers indicated experiencing a
perceived undue punishment, a result that was unchanged from the 2009
survey. 95 Minority members were more likely to indicate experiencing
perceived undue punishment than White members, but there were no
significant differences between racial or ethnic groups who indicated
experiencing undue punishment. ODEI officials told us that their office did
not make any recommendations related to military justice as a result of
these 2013 survey results because the findings were too small to warrant
such steps. Moreover, ODEI officials said that while they have not
completed their analysis of the 2017 survey data, the question about
receiving nonjudicial punishment or court-martial had been removed from
the 2017 survey. ODEI officials explained that the question was removed
because the perception of unfair punishment was not the goal of the
survey, although they said that the question could be reinstated for future
surveys if the goals for the survey change.
In June 2017, ODEI initiated a review of the military justice system
following the publication of a report by a non-profit organization that found
racial disparities in military justice actions. 96 According to ODEI officials,
their review assesses disparities in the military justice system using a
similar analysis to that in the non-profit organization’s report, which
analyzed rates of military justice actions per thousand servicemembers.
ODEI officials told us they also observed racial and gender disparities
among servicemembers involved in the military justice system in their
94

In addition, in 2017, the Air Force assembled a working group called the Disciplinary
Actions Analysis Team to examine the barriers certain demographic groups face to career
success, including barriers to training opportunities, promotion, and retention. The working
group is in the early stages of organizing and has not yet published any findings or
recommendations for service leadership.

95

Defense Manpower Data Center, 2013 Workplace and Equal Opportunity Survey of
Active Duty Members Overview Report (October 2014).

96

Protect Our Defenders, Racial Disparities in Military Justice (May 2017).

Page 65

GAO-19-344 Military Justice

own analysis of the service data. The officials said that the report on the
results of their review will not directly address the issue of whether bias
exists in the military justice process or the causes of any disparities, but
will serve as a precursor to a future research study that looks more
comprehensively into the issue of whether bias exists in the military
justice system. ODEI officials said that their report should be issued in
2019.
Standards for Internal Control in the Federal Government state that
management uses quality information to make informed decisions and
evaluate the entity’s performance in achieving key objectives and
addressing risks. The standards further provide that management should
evaluate issues identified through monitoring activities and determine
appropriate corrective actions. 97 Officials from DOD and the military
services acknowledged that they do not know the cause of the racial and
gender disparities that have been identified in the military justice system.
This is because they have not conducted a comprehensive evaluation to
identify potential causes of these disparities and make recommendations
about any appropriate corrective actions to remediate the cause(s) of the
disparities. By conducting a comprehensive analysis into the causes of
disparities in the military justice system, DOD and the military services
would be better positioned to identify actions to address disparities, and
thus help ensure that the military justice system is fair and just, a key
principle of the UCMJ.

Conclusions

The single overarching principle of the UCMJ is that a system of military
law can foster a highly disciplined force if it is fair and just, and is
recognized as such by both members of the armed forces and by the
American public. DOD and the military services collect and maintain data
on the race, ethnicity, and gender of all servicemembers. However, these
data vary within and across the services, limiting the ability to collectively
or comparatively assess military justice data to identify any disparities.
DOD has recently taken steps to address this issue by directing the
military services to, no later than December 23, 2020: collect uniform race
and ethnicity data in their military justice databases, or aggregate any
expanded ethnic or racial categories to the categories listed in the
standards; collect either the social security number or DOD identification
number in their military justice databases; and collect complete summary
97

GAO-14-704G.

Page 66

GAO-19-344 Military Justice

courts-martial information. It will be important for the military services to
complete these actions to allow for efficient analysis and reporting of
consistent military justice data.
However, the newly issued standards apply only to the military justice
databases and not to the investigations and personnel databases. The
ability to query and report on the gender of servicemembers in its military
justice database would provide the Coast Guard with more readily
available data to identify or assess any gender disparities that may exist
in the investigation and trial of military justice cases without merging data
from multiple databases. Moreover, taking steps to develop the capability
to present the race and ethnicity data from the military services’ personnel
and investigations databases using the same categories included in the
December 2018 standards for the military justice databases would enable
DOD and the military services to more easily and efficiently assess the
extent to which there are any racial or ethnic disparities throughout the
military justice process.
Further, DOD’s annual reports about the number and status of pending
military justice cases do not include demographic information, such as
breakdowns by race or gender, about servicemembers who experienced
a military justice action. Reporting this information would provide
servicemembers and the public with greater visibility into potential
disparities and help build confidence that DOD is committed to a military
justice system that is fair and just. Moreover, DOD does not have
guidance that establishes criteria to determine when data indicating
possible disparities among racial, ethnic, or gender groups in the
investigations, trials, or outcomes of cases in the military justice system
should be further reviewed, or describes the steps that should be taken to
conduct such further review. By establishing such criteria, DOD and the
services would be better positioned to monitor the military justice system
to help ensure that it is fair and just, a key principle of the UCMJ.
Our analysis of available data identified racial and gender disparities in all
of the military services for servicemembers with recorded investigations,
and for four of the military services for trials in special and general courtsmartial, but these disparities generally were not present in the convictions
or punishments of cases. 98 These findings suggest disparities may be
98

Our findings of racial and gender disparities, 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.

Page 67

GAO-19-344 Military Justice

limited to particular stages of the military justice process for the period
covered by our analysis. However, we were unable to determine whether
there were disparities among servicemembers subject to nonjudicial
punishments in the Army, the Navy, and the Coast Guard because these
services do not collect complete nonjudicial punishment data, such as
data on the servicemember’s race, ethnicity, gender, offense, and
punishment for all nonjudicial punishments, in any of their databases. The
absence of complete nonjudicial punishment data in the Army, the Navy,
and the Coast Guard limits their visibility into the vast majority of legal
punishments imposed on servicemembers under the UCMJ every year.
Without such data, these three services will remain limited in their ability
to assess or identify disparities among populations subject to this type of
punishment.
Finally, DOD recently conducted a study of racial and gender disparities
in the military justice system, and expects to complete its report in 2019.
However, this study will not assess the causes of the racial and gender
disparities identified in the military justice system. Our findings of racial
and gender disparities, 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. By
conducting a comprehensive evaluation of the causes of these disparities,
DOD and the military services would be better positioned to identify
actions to address disparities, and thus help ensure that the military
justice system is fair and just, a key principle of the UCMJ.

Recommendations for
Executive Action

We are making a total of 11 recommendations, including 3 to the
Secretary of Homeland Security, 3 to the Secretary of Defense, 2 to the
Secretary of the Army, 2 to the Secretary of the Navy, and 1 to the
Secretary of the Air Force.
The Secretary of Homeland Security should ensure that the Commandant
of the Coast Guard modifies the Coast Guard’s military justice database
so that it can query and report on gender information. (Recommendation
1)
The Secretary of the Army should develop the capability to present
servicemembers’ race and ethnicity data in its investigations and
personnel databases using the same categories of race and ethnicity
established in the December 2018 uniform standards for the military
justice databases, either by (1) modifying the Army’s investigations and
personnel databases to collect and maintain the data in accordance with

Page 68

GAO-19-344 Military Justice

the uniform standards, (2) developing the capability to aggregate the data
into the race and ethnicity categories included in the uniform standards,
or (3) implementing another method identified by the Army.
(Recommendation 2)
The Secretary of the Air Force should develop the capability to present
servicemembers’ race and ethnicity data in its investigations and
personnel databases using the same categories of race and ethnicity
established in the December 2018 uniform standards for the military
justice databases, either by (1) modifying the Air Force’s investigations
and personnel databases to collect and maintain the data in accordance
with the uniform standards, (2) developing the capability to aggregate the
data into the race and ethnicity categories included in the uniform
standards, or (3) implementing another method identified by the Air
Force. (Recommendation 3)
The Secretary of the Navy should develop the capability to present
servicemembers’ race and ethnicity data in its investigations and
personnel databases using the same categories of race and ethnicity
established in the December 2018 uniform standards for the military
justice databases, either by (1) modifying the Navy’s investigations and
personnel databases to collect and maintain the data in accordance with
the uniform standards, (2) developing the capability to aggregate the data
into the race and ethnicity categories included in the uniform standards,
or (3) implementing another method identified by the Navy.
(Recommendation 4)
The Secretary of Homeland Security should ensure that the Commandant
of the Coast Guard develops the capability to present servicemembers’
race and ethnicity data in its investigations and personnel databases
using the same categories of race and ethnicity established in the
December 2018 uniform standards for the military justice databases,
either by (1) modifying the Coast Guard’s investigations and personnel
databases to collect and maintain the data in accordance with the uniform
standards, (2) developing the capability to aggregate the data into the
race and ethnicity categories included in the uniform standards, or (3)
implementing another method identified by the Coast Guard.
(Recommendation 5)
The Secretary of Defense should ensure that the Joint Service Committee
on Military Justice, in its annual review of the UCMJ, considers an
amendment to the UCMJ’s annual military justice reporting requirements
to require the military services to include demographic information,

Page 69

GAO-19-344 Military Justice

including race, ethnicity, and gender, for all types of courts-martial.
(Recommendation 6)
The Secretary of Defense, in collaboration with the Secretaries of the
military services and the Secretary of Homeland Security, should issue
guidance that establishes criteria to specify when data indicating possible
racial, ethnic, or gender disparities in the military justice process should
be further reviewed, and that describes the steps that should be taken to
conduct such a review. (Recommendation 7)
The Secretary of the Army should consider the feasibility, to include the
benefits and drawbacks, of collecting and maintaining complete
information for all nonjudicial punishment cases in one of the Army’s
databases, such as information on the servicemembers’ race, ethnicity,
gender, offense, and punishment imposed. (Recommendation 8)
The Secretary of the Navy should consider the feasibility, to include the
benefits and drawbacks, of collecting and maintaining complete
information for all nonjudicial punishment cases in one of the Navy’s
databases, such as information on the servicemembers’ race, ethnicity,
gender, offense, and punishment imposed. (Recommendation 9)
The Secretary of Homeland Security should ensure that the Commandant
of the Coast Guard considers the feasibility, to include the benefits and
drawbacks, of collecting and maintaining complete information for all
nonjudicial punishment cases in one of the Coast Guard’s databases,
such as information on the servicemembers’ race, ethnicity, gender,
offense, and punishment imposed. (Recommendation 10)
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. (Recommendation 11)

Page 70

GAO-19-344 Military Justice

Agency Comments
and Our Evaluation

We provided a draft of this report to DOD and the Department of
Homeland Security for review and comment. Written comments from
DOD and the Department of Homeland Security are reprinted in their
entirety in appendixes X and XI, respectively. DOD and the Department of
Homeland Security provided additional technical comments, which we
incorporated in the report, as appropriate. In written comments, DOD
concurred with six recommendations, and partially concurred with two
recommendations that were directed to the Secretary of Defense. The
Department of Homeland Security concurred with the three
recommendations directed to the Secretary of Homeland Security.
DOD concurred with our six recommendations to present
servicemembers’ race and ethnicity data in each of the military services’
respective investigations and personnel databases using the same
categories of race and ethnicity established for their military justice
databases; consider an amendment to the UCMJ’s annual military justice
reporting requirements to require the military services to include
demographic information for all types of courts-martial; and consider the
feasibility of collecting and maintaining complete information for all
nonjudicial punishment cases.
DOD partially concurred with two of our recommendations, agreeing with
the content, but requesting that we modify the recommendations to direct
them to more appropriate entities. Specifically, DOD concurred with our
recommendations that guidance should be issued to establish criteria
specifying when data indicating possible racial, ethnic, or gender
disparities require further review and the steps that will be taken to
conduct the review; and to conduct an evaluation to identify the causes of
any racial or gender disparities in the military justice system and, if
necessary, take remedial steps to address the causes of these
disparities. For both recommendations, DOD suggested that the
Secretary of Homeland Security be added, and that we remove the DOD
Office for Diversity, Equity and Inclusion and the Commandant of the
Coast Guard, as they fall under the Secretary of Defense and the
Secretary of Homeland Security, respectively. We agree with DOD’s
suggestions, and we have modified both recommendations accordingly.
In an email correspondence, the Department of Homeland Security and
the Coast Guard concurred with the updates.
In its written comments, the Department of Homeland Security concurred
with our three recommendations to modify the Coast Guard’s military
justice database so that it can query and report on gender information, to
present servicemembers’ race and ethnicity data in its investigations and

Page 71

GAO-19-344 Military Justice

personnel databases using the same categories of race and ethnicity
established for the military justice database, and to consider the feasibility
of collecting and maintaining complete information for all nonjudicial
punishment cases.
We are sending copies of this report to the appropriate congressional
committees, the Acting Secretary of Defense, and the Acting Secretary of
Homeland Security. In addition, this report will also be available at no
charge on the GAO website at http://www.gao.gov.
If you or your members of your staff have any questions regarding this
report, please contact me at (202) 512-3604 or [email protected]. Contact
points for our Offices of Congressional Relations and Public Affairs may
be found on the last page of this report. GAO staff who made significant
contributions to this report are listed in Appendix XII.

Brenda S. Farrell
Director
Defense Capabilities and Management

Page 72

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology
Appendix I: Objectives, Scope, and
Methodology

The objectives of this report were to assess the extent to which (1) the
military services collect and maintain information about the race, ethnicity,
and gender of servicemembers investigated and disciplined for violations
of the Uniform Code of Military Justice (UCMJ) that can be used to
assess disparities; and (2) there are racial and gender disparities in
investigations, disciplinary actions, and case outcomes in the military
justice system, and whether the Department of Defense (DOD) and the
military services have taken steps to study any identified disparities.

Methods Used to Address
Both Objectives

To address both of our objectives, we analyzed data collection, data
maintenance, and military justice disciplinary actions involving active-duty
servicemembers in the Army, the Navy, the Marine Corps, the Air Force,
and the Coast Guard. Although the Coast Guard is part of the Department
of Homeland Security, the Coast Guard is a military service and a branch
of the armed forces at all times.
•

We analyzed military justice actions initiated and recorded in service
investigations and military justice databases between fiscal years
2013 through 2017. We chose this time period because it provided the
most recent history of available military justice data at the time of our
review.

•

We requested record-level data from each of the military services’
personnel, investigations, and military justice databases, which
resulted in a total of 15 data requests.

Table 6 below provides an overview of the databases included in our
review, broken out by database type.

Page 73

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Table 6: Military Service Personnel, Investigations, and Military Justice Databases
Military service Personnel database

Investigations database

Military justice database

Army

Total Army Personnel Database
(TAPDB)

Army Law Enforcement Reporting and Army Court-Martial Information System
Tracking System (ALERTS)
(ACMIS)

Navy

Navy Personnel Database
(NPDB)

Consolidated Law Enforcement
Operations Center (CLEOC)

Case Management System (CMS)

Marine Corps

Marine Corps Total Force System Consolidated Law Enforcement
(MCTFS)
Operations Center (CLEOC)

Case Management System (CMS)

Air Force

Military Personnel Data System
(MilPDS)

Investigative Information Management Automated Military Justice Analysis and
Systems (I2MS)
Management System (AMJAMS)

Coast Guard

Direct Access

Field Activity Case Tracking System
(FACTS)

Military Justice Online (MJO)

Law Manager

Source: GAO analysis of DOD and military service database information. | GAO-19-344

We sent individual data requests that were tailored based on our
conversations with service officials and our own analysis of the availability
of data. In addition to requesting the race, ethnicity, and gender of
servicemembers subject to military justice actions, we also requested
other demographic and administrative attribute data—such as rank, age,
years of service, duty station, and occupation—from the services’
personnel databases to include in our statistical models. We identified
these attributes by reviewing relevant literature and interviewing agency
officials.
Personnel databases. We requested and received monthly snapshots
with record-level data on all active-duty servicemembers in each of the
military services from fiscal years 2013 through 2017. Specifically, we
requested demographic and administrative data, including race, ethnicity,
gender, rank, education, age or date of birth, years of service, occupation,
location or duty station, deployed status, administrative or disciplinary
actions and dates, character of service separation, and servicemembers’
unique identifiers (social security number and employee identification
number).
Investigations databases. We requested and received record-level data
on all investigations recorded in a military service military criminal
investigative organization (MCIO) database that were initiated from fiscal
years 2013 through 2017, where the subject of the investigation was an
active-duty servicemember. For each case, we requested certain attribute
data on the investigation subject, including race, ethnicity, gender, rank,
age or date of birth, service and component, offense(s) investigated, case

Page 74

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

initiation date, investigation source, investigating entity, investigation
outcome and date, incident location, and the subject’s unique identifier,
such as social security number or employee identification number. In
some services not all of these attributes were available or requested. For
example, since the Air Force database only included investigations
conducted by the Air Force Office of Special Investigations, we did not
request information about the investigating entity. In addition, the Navy
Criminal Investigative Service provided us with data about and we
analyzed closed cases only, whereas the Army and the Air Force MCIOs
provided us with data about and we analyzed all cases in their database
during the period of our review.
Military justice databases. We requested and received record-level data
on all cases where a servicemember was subject to disciplinary
proceedings under the Uniform Code of Military Justice (UCMJ) from
fiscal years 2013 through 2017. For each case where charges were
preferred against a servicemember during this period, we requested
demographic and administrative data on the servicemember as well as
key information related to their case, including race, ethnicity, gender,
rank, age or date of birth, component, case type and forum, offense(s)
charged, case disposition and date, appeals status, case outcome or
sentence, disciplinary action taken, date charges were first preferred, and
the servicemember’s unique identifier, such as social security number or
employee identification number. 1 We received general and special courtsmartial data from all of the services from their military justice databases.
For the Army, in addition to data from their military justice database,
Military Justice Online, we also received courts-martial data from a
separate database, called the Army Court-Martial Information System
(ACMIS), which is used by the service’s trial judiciary to track courtsmartial.
For summary courts-martial and nonjudicial punishments, the services
varied in the extent that and the location where they collected and
maintained complete data for these two military justice actions, as is
discussed further earlier in this report.
•

In the Air Force, summary courts-martial and nonjudicial punishment
data is maintained in the service’s military justice database, the
Automated Military Justice Analysis and Management System.

1

The preferral date is the date when an accused servicemember was first charged with a
violation of the UCMJ.

Page 75

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Methods Used to Evaluate
Collection and
Maintenance of Data

•

The Marine Corps did not collect and maintain complete data about
summary courts-martial or nonjudicial punishments in its military
justice database, however, its personnel database included
information about all summary courts-martial and nonjudicial
punishments imposed on servicemembers during the period of our
review.

•

The Army and the Navy did not collect and maintain complete data
about summary courts-martial or nonjudicial punishments in their
military justice databases, or other databases. In these services,
summary courts-martial and nonjudicial punishments were recorded in
their military justice databases if these actions had involvement by the
services’ legal offices. Further, summary courts-martial and
nonjudicial punishments were recorded in the personnel databases
used by these services only if these actions resulted in an
administrative action against the accused, such as a forfeiture of pay
or reduction in grade.

•

The Coast Guard did not collect and maintain complete data about
nonjudicial punishments in its military justice database or other
databases; nonjudicial punishments were recorded in its military
justice database if a legal office was involved in the action. Further,
nonjudicial punishments were recorded in the Coast Guard’s
personnel database if they resulted in an administrative action against
the accused, such as a forfeiture of pay or reduction in grade.

To evaluate the extent to which the military services collect and maintain
race, ethnicity, and gender data about servicemembers investigated and
disciplined for violations of the UCMJ, we first reviewed service guidance,
user manuals, and other documents related to the services’
investigations, military justice, and personnel databases. We reviewed
these documents to determine:
•

the types of data officials are required to collect and maintain; and

•

the internal procedures the services follow in inputting information
about race, ethnicity, and gender data into each type of database.

For example, we determined whether collection of this information was
mandatory, and how this information was entered into and recorded in
each database. Specifically, we determined whether information about
race, ethnicity, and gender was entered into each database manually,
using a drop-down menu, or was auto-populated from another database.
Further, we identified the number of possible response options that each
database contained for each of these demographic fields.

Page 76

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Second, we interviewed service officials who manage and use the military
justice, investigations, and personnel databases to discuss:
•

which fields in each database track the race, ethnicity, and gender of
servicemembers; and

•

how these data are input and their insights regarding the reliability of
these data.

Specifically, we interviewed officials from the legal branches of the
military services, including the Army Office of the Judge Advocate
General, the Navy Judge Advocate General’s Corps, the Marine Corps’
Judge Advocate Division, the Air Force Judge Advocate General’s Corps,
and the Coast Guard Office of the Judge Advocate General. In addition,
we spoke with officials in the military criminal investigative organizations
(MCIO), including the Army Criminal Investigation Command, the Naval
Criminal Investigative Service, the Air Force Office of Special
Investigations, and the Coast Guard Investigative Service. We also
interviewed officials from the manpower and personnel offices of the
services with responsibility for the services’ personnel databases,
including the Army’s Human Resources Command and the Office of the
Deputy Chief of Staff; the Navy’s Personnel Command; the Marine Corps
Manpower and Reserve Affairs Manpower Information Systems Branch;
the Air Force Personnel Center; and the Coast Guard’s Personnel
Service Center.
Finally, we analyzed the data we received from the investigations, military
justice, and personnel databases to determine the completeness of the
race, ethnicity, and gender information that was recorded in each of the
databases. We assessed the military services’ systems and procedures
for collecting data against DOD and service guidance and relevant federal
internal control standards. 2

Methods Used to Evaluate
Racial, Ethnic, and
Gender Disparities

To evaluate the extent to which there are racial, ethnic, and gender
disparities in investigations, disciplinary actions, and case outcomes, we
analyzed data from the military services’ investigations, military justice,
and personnel databases to determine summary statistics and we then
conducted bivariate and multivariate regression analyses.
2

GAO, Standards for Internal Control in the Federal Government, GAO-14-704G
(Washington, D.C.: September 2014).

Page 77

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Investigations. We focused on alleged violations of the UCMJ that were
recorded in databases used by service-specific MCIOs. Investigations are
recorded in the MCIO databases when a servicemember is the subject of
a criminal allegation made by another person; for purposes of this report,
we say the servicemember had a “recorded investigation” to describe
these cases. We analyzed investigation information from the databases
used by each of the military services’ MCIOs. Specifically, we analyzed
data from the Army’s Criminal Investigation Command, which included
cases investigated by military police and Criminal Investigation
Command; the Navy and Marine Corps’ Naval Criminal Investigative
Service, which included cases investigated by the Naval Criminal
Investigative Service and military police; the Air Force’s Office of Special
Investigations, which included only Office of Special Investigations cases;
and the Coast Guard Investigative Service, which included only Coast
Guard Investigative Service cases. Our analysis of recorded
investigations data did not include investigations conducted by a
servicemember’s command, because those investigations are not
recorded in the MCIO databases.
Military Justice Discipline. We included in our definition of
servicemembers disciplined for a violation of the UCMJ those
servicemembers with cases that resulted in a trial in any type of courtmartial (general, special, and summary), or servicemembers who were
subject to a nonjudicial punishment from fiscal years 2013 through 2017.
We analyzed data for trials in general and special courts-martial
separately from trials in summary courts-martial because general and
special courts-martial result in a criminal conviction if the servicemember
is found guilty, while summary courts-martial are not a criminal forum and
do not result in a criminal conviction. We analyzed general and special
courts-martial cases together due to the small number of cases for some
racial or gender groups. In addition, we also separated general and
special courts-martial into cases that either were or were not preceded by
an investigation recorded in an MCIO database. Our analysis of general
and special courts-martial cases without a recorded investigation included
those general and special courts-martial that were investigated by a
servicemember’s command or other law enforcement entities.
We used the preferral date, or the date when an accused servicemember
was first charged with a violation, to count the number of courts-martial
that occurred in a given fiscal year. However, each military service uses
the date in which the court-martial judgment was given when reporting the
number of each type of court-martial in their annual reports to the Court of
Appeals for the Armed Forces. As a result, the number of court-martial

Page 78

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

cases in a given year analyzed for our review differs from what was
reported in the annual reports. In discussions with officials after we had
completed our preliminary analyses, they recommended that we use the
referral date instead of the preferral date, so that our total number of
cases would be more consistent with the number of cases that they
reported. However, changing the date for grouping cases would have
required us to request new military justice data from each of the military
services, and conduct additional work. Above all, using the preferral date
would not impact the findings of racial and gender disparities. In addition,
our analyses only counted cases that were ultimately tried at general,
special, or summary courts-martial, and excluded those cases where
charges were dismissed, withdrawn, or subject to some alternate
resolution. For nonjudicial punishments, we used the date that the
punishment was imposed.
To prepare the data for our analyses and ensure that we had consistent
profiles for the race, ethnicity, and gender of the servicemembers, we
merged records from the military services’ investigations, military justice,
and personnel databases. We merged records using servicemembers’
unique identifiers, such as social security number or employee
identification number, that were common among a particular service’s
databases. In some instances—a small proportion of cases—we could
not match personnel records with military justice records because the
social security number or employee identification number in the military
justice database did not match the information in the personnel database.
In other instances, we could not match personnel records with military
justice records because the military justice records did not contain a
social security number or employee identification number to match with
information found in their personnel record. We first tried to match these
cases using the servicemembers’ name and date of birth; however, in
some cases we were unable to match personnel records with
investigations or military justice cases. As a result, we compiled lists of
those cases we were unable to match, and we provided the services with
lists of these cases. Service officials manually looked up this data and
provided us with the missing social security numbers or employee
identification numbers for these cases so that we could complete our
analyses. These manual look up efforts increased our match rates so that
we had a data set that we determined was sufficiently complete to
perform our analyses.
For servicemembers who were the subjects of military justice actions, we
used the attribute data that was available in the personnel database at
the time an investigation or disciplinary action was initiated (the preferral

Page 79

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

date for courts-martial). For our total service populations, which included
servicemembers who were not the subject of a military justice action, we
used their attribute data from the “median” snapshot of the five fiscal
years of personnel data we received. Based on discussions with service
officials, we treated the personnel databases as the authoritative sources
for servicemembers’ demographic and administrative data. For some
services when needed, if we identified a discrepancy in the race or
gender value for a servicemember between the data in the personnel and
military justice databases, we used the value recorded in the personnel
database because service officials had told us that the personnel
databases were the official sources for demographic data such as race
and gender, and would be more likely to contain more reliable data for
these fields than the investigations or military justice databases. For some
services where there were cases where an attribute value was missing in
the personnel database, we used the military justice or investigative
database as a secondary source for this information. In merging the
records from the personnel, military justice, and investigations databases,
we created a single data file for each service that contained attribute data
for all active-duty servicemembers, as well as complete information on
the investigation and discipline of servicemembers who were the subject
of a military justice action from fiscal years 2013 through 2017.
In using this methodology to merge the records, the total number of
servicemembers we use in our report when discussing the total service
populations for each service is greater than the total active-duty force end
strength of that service in any given fiscal year. This is because our total
service populations represent the number of unique individuals who
served on active duty from fiscal years 2013 through 2017.
In addition, as part of our data preparation, we consolidated the various
race and ethnicity values in the service personnel databases to the five
groups for race and the two groups for ethnicity established by Office of
Management and Budget (OMB) standards for maintaining, collecting,
and presenting data on race and ethnicity for all federal reporting
purposes. 3 The five race groups in the standards are American Indian or
3

Office of Management and Budget, Revisions to the Standards for the Classification of
Federal Data on Race and Ethnicity, 62 Fed. Reg. 58,782 (Oct. 30, 1997). In 2016, the
Office of Management and Budget issued a proposed revision to the standards. See
Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity,
81 Fed. Reg. 67,398 (Sept. 30, 2016). As of May 2019, the Office of Management and
Budget had not issued the revised standards.

Page 80

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Alaska Native; Asian; Black or African American; Native Hawaiian or
Other Pacific Islander; and White. The two ethnic groups are Hispanic or
Latino and Not Hispanic or Latino. First, we collapsed race and ethnicity
data into a single combined field. Specifically, we grouped individuals of
Hispanic ethnicity together, regardless of their racial identification, so that
we could compare those of Hispanic ethnicity to other racial groups. We
did this in part because of the ways in which some of the services record
these data in their databases. For example, the Navy’s and the Marine
Corps’ military justice databases do not have separate fields for race and
ethnicity; instead, the values are tracked in a single field. Throughout the
discussion for objective 2 of this report, we refer to the combined race
and ethnicity values as race.
We then consolidated races to the five racial groups in the OMB
standards. When military service personnel databases included different
or additional possible options for race and ethnicity than the groups
established by the OMB standards, we consolidated the options in
accordance with the definitions for each race and ethnicity listed in the
OMB standards. Given the small number of cases in some racial groups,
we collapsed certain racial groups into an “Other” group in order to report
statistically reliable results. The “Other” group includes individuals who
identified as Asian, Native Hawaiian/Other Pacific Islander, American
Indian/Alaska Native, and multiple races.
Summary statistics. We analyzed data from the military services’
investigations, military justice, and personnel databases to determine the
extent to which racial and gender groups were the subjects of recorded
investigations, tried in courts-martial, and subject to nonjudicial
punishments (for Army and Marine Corps, services for which we had
complete data) at higher rates or lower rates than each racial and gender
group’s proportion of the overall service populations. Other than our
analysis of recorded investigations, we did not analyze Coast Guard
cases due to the small number of general and special courts-martial
adjudicated in the Coast Guard from fiscal years 2013 through 2017.
To conduct this analysis, we used data on all active-duty servicemembers
to identify what proportion each racial group (White, Black, Hispanic, and
Other) and gender group (male, female) made up of the overall service
population from fiscal years 2013 through 2017. We then used data from
the services’ military justice or personnel databases to calculate the
representation of each racial and gender group as a percent of the
population subjected to each type of military justice action.

Page 81

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

We also examined the rates at which certain racial and gender groups
were charged with drug offenses (Article 112a) and sexual assault
offenses (Article 120) compared to their proportions of the overall service
populations. 4 See Appendix III for information regarding recorded
investigations and general and special courts-martial of drug and sexual
assault offenses. We analyzed these two specific UCMJ offenses
because officials from some services told us that an investigation into
these offenses may frequently be mandatory, and thus could potentially
mitigate the risk of bias. 5 To conduct this analysis, we used offense data
from the services’ military justice databases to determine each racial and
gender group’s representation in the population that was the subject of a
military justice action for a drug, sexual assault, or other offense type.
Bivariate and Multivariate Regression Analyses. We developed a
logistic regression model using the data we received from the services’
investigations and military justice databases to determine the extent that
certain attributes were associated with higher rates of investigation or
discipline of servicemembers. We conducted bivariate logit analyses to
estimate the association between select attribute factors (or independent
variables) and the outcome variables (the dependent variable) in a binary
format, except for the two offense outcome variables. Table 7 below lists
all of the dependent and independent variables we used in our analyses.

4

The drug offenses we analyzed were charges under UCMJ Article 112a, wrongful use,
possession, distribution, etc. of a controlled substance. In addition, the sexual assault
offenses we analyzed were any charges under UCMJ Article 120, which includes rape and
sexual assault generally, rape and sexual assault of a child, and other sexual misconduct.
5

Service officials stated that some drug (Article 112a) offenses are initiated as a result of
random urinalysis tests, and in those cases a positive result will trigger an investigation
regardless of the servicemember’s race, ethnicity, or gender. According to Department of
Defense Instruction 5505.18, all allegations of adult sexual assault are immediately
reported to the appropriate MCIO, and that MCIO will initiate a criminal investigation into
that allegation if the offense occurred within its jurisdiction.

Page 82

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Table 7: Independent and Dependent Variables Included in GAO’s Regression Analyses
Independent variables
Age
Education
Gender
Offense
Race/Ethnicity
Rank
Years of service
Outcome (dependent) variables
Recorded investigations
Recorded investigations for drug, sexual assault, and all other offenses
General and special courts-martial
General and special courts-martial for drug, sexual assault, and all other offenses
General and special courts-martial following a recorded investigation
General and special courts-martial without a recorded investigation
Case outcome (conviction or acquittal in general and special courts-martial)
Punishment severity
Summary courts-martial
Nonjudicial punishment
Source: GAO summary of variables analyzed from services’ investigations, military justice, and personnel databases. | GAO-19-344

To conduct our statistical analyses, we created groups for each
demographic and administrative attribute (independent variable) that we
tested in our regression model. We created these groups based on input
and guidance from service officials. While the modeling subgroups we
created are largely consistent across services, some values are different
for certain services. Table 8 summarizes the modeling groups we
constructed for each service for each attribute included in our regression
analyses.

Page 83

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Table 8: Modeling Groups Used in Regression Analyses for Each Military Service
Attribute

Army

Navy

Marine Corps

Air Force

Coast Guard

Race

White
Black
Hispanic
Other
Unknown

White
Black
Hispanic
Other
Unknown

White
Black
Hispanic
Other
Unknown

White
Black
Hispanic
Other
Unknown

White
Black
Hispanic
Other
Unknown

Gender

Male
Female

Male
Female

Male
Female

Male
Female

Male
Female

Age

< 25 years
25-30 years
30-40 years
40 or more years

< 21 years
21-25 years
26-30 years
30-40 years
> 40 years

< 21 years
21-25 years
26-30 years
30-40 years
> 40 years

< 21 years
21-25 years
26-30 years
31-35 years
> 35 years

< 25 years
25-30 years
30-40 years
> 40 years

Rank

E1-E4
E5-E9
Officers

E1-E4
E5-E9
Officers

E1-E4
E5-E9
Officers

E1-E4
E5-E6
E7-E9
Officers

E1-E4
E5-E9
Officers

Years of service

0-4 years
4-8 years
8-12 years
> 12 years

< 2 years
3-4 years
5-6 years
7-10 years
11-15 years
> 15 years

< 2 years
3-4 years
5-6 years
7-10 years
11-15 years
> 15 years

0-4 years
4-6 years
> 6 years

0-2 years
2-4 years
4-6 years
6-10 years
10-15 years
> 15 years

Education

High school or less
High school or less
More than high school More than high
school
Unknown
Unknown

High school or less
More than high
school
Unknown

High school and some
college
Associates degree
Bachelor’s degree
Post-bachelor’s
degree
Unknown

High school or
less
More than high
school
Unknown

Outcome

Conviction
Acquittal

Conviction
Acquittal

Conviction
Acquittal

Conviction
Acquittal

Punishment
severity

Any type of dismissal
or discharge or
confinement > 2 years
Confinement < 2
years without
dismissal or discharge
All other possible
sentencing options

Dismissal or any
kind of discharge
All other possible
sentencing options

Dismissal or any
kind of discharge
All other possible
sentencing options

Any type of dismissal
or discharge
(regardless of any
confinement)
Confinement without
dismissal or discharge
All other possible
sentencing options

Page 84

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

Attribute

Army

Navy

Marine Corps

Air Force

Offenses

Drug offenses (Article
112a)
Sexual assault
offenses (Article 120)
All other offenses

Drug offenses
(Article 112a)
Sexual assault
offenses (Article
120)
All other offenses

Drug offenses
(Article 112a)
Sexual assault
offenses (Article
120)
All other offenses

Drug offenses (Article
112a)
Sexual assault
offenses (Article 120)
All other offenses

Coast Guard

Source: GAO analysis. | GAO 19-344

When analyzing the severity of punishments, we developed two groups
for the Navy and the Marine Corps, and three groups for the Air Force
and the Army, as shown in table 9 below. We did not create a third
punishment group for confinement without dismissal or discharge for the
Navy and the Marine Corps because of the small number of cases with
confinement that did not also include some sort of discharge. Based on
discussions with service officials, we determined that a sentence resulting
in a dismissal or discharge was the most severe punishment outcome.
Table 9: Groups Used to Measure Punishment Severity
Navy and Marine Corps

Army

Air Force

1.
2.

1.

1.

dismissal or any kind of discharge
all other possible sentencing options

2.
3.

any type of dismissal or discharge or
confinement of more than 2 years
confinement of less than 2 years
without dismissal or discharge
all other possible sentencing options

2.
3.

any type of dismissal or discharge
(regardless of any confinement)
confinement without dismissal or
discharge
all other possible sentencing options

Source: GAO analysis. | GAO-19-344

Typically, a logistic regression model is appropriate when the model
outcome is a binary (yes/no) response. Because the punishment groups
for the Army and the Air Force were not binary, they could not be
analyzed using a multivariate logistic regression. Instead, we used an
ordered logit model, also called an ordered logistic regression model, to
analyze punishment severity in the Army and the Air Force. An ordered
logistic regression is an extension of the logistic regression model that
applies to dependent variables where there are more than two response
categories. This model allowed us to examine the degree to which a
racial or gender group was more likely or less likely than another group to
receive a more severe punishment in general and special courts-martial,
while controlling for other attributes, such as gender, education, rank,
composition of panel, and offense type. To conduct this analysis, we

Page 85

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

reviewed outcome data from the services’ personnel, investigations, and
military justice databases.
Based on our bivariate analyses, we determined which variables were
significantly associated with military justice actions, and that appeared to
be statistically significant predictors of an individual’s likelihood to be
subject to a military justice action. 6 Appendix IX includes a summary of
those indicators for each of the services. We also examined correlation
matrices of the independent variables to determine where there were high
correlations between two variables. Where variables were highly
correlated, we chose one variable over the others or created a hybrid
variable combining those two variables. Specifically, we excluded age
and years of service for most of the military services, due to high
correlation with the rank variable. Based on our discussions with service
officials, they indicated that rank would be the preferred variable to
include in our analyses if selecting only one variable among rank, age,
and years of service. However, for the Air Force, based on discussion
with Air Force officials, we did control for years of service among the
lower enlisted ranks (E1-E4). In addition, we could not include education
for the Army due to variability and overlapping values in the data. Further,
we chose not to model attributes such as occupation and location due to
the great variability in these data and the difficulty in creating groups and
reaching agreement about those groups with service officials.
Based on these results, we then conducted a series of multivariate
logistic regression models. Multivariate logistic regression modeling is a
statistical method that examines several variables simultaneously to
estimate whether each of these variables are more likely or less likely to
be associated with a certain outcome. A multivariate regression analysis
analyzes the potential influence of each individual factor on the likelihood
of a binary outcome (e.g., a specific military justice action) while
simultaneously accounting for the potential influence of the other factors.
This type of modeling allowed us to test the association between
servicemember characteristics, such as race or gender, and the odds of a
military justice action (shown as the outcome variables in table 7 above),
while holding other servicemember attributes constant (such as gender,
6
For purposes of this report, we use the term “likelihood” when discussing the odds ratios
from the results of our regression analyses. Odds ratios that are statistically significant
and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic
are more likely or less likely, respectively, to be subject to a particular military justice
action.

Page 86

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

rank, and education, shown as the independent variables in table 7
above). We conducted a separate regression for each of the military
justice actions listed as an outcome variable. We selected this type of
model because it could account for the attributes simultaneously. For the
purposes of consistency, in our multivariate regression analyses, we
made all racial comparisons with White servicemembers as the reference
category. Similarly, we made all gender comparisons with female
servicemembers as the reference category.
A logistic regression model provides an estimated odds ratio, where a
value greater than one indicates a higher or positive association; in this
case, between the race, ethnicity, or gender of a servicemember (the
independent variables) and the likelihood of being the subject of a military
justice action (the dependent, or outcome, variable). An estimated odds
ratio less than one indicates lower odds or likelihood of being the subject
of a military justice action when a factor—here, a specific demographic or
administrative attribute—is present. The statistical significance of the
logistic regression model results is determined by a p-value of less than
0.05. As a result, in our report we state that odds ratios that are
statistically significant and greater than 1.00 or lower than 1.00 indicate
that individuals with that characteristic are more likely or less likely,
respectively, to be the subject of a particular outcome or military justice
action. In cases where the p-value was greater than 0.05, we report that
we could not identify any statistically significant differences, which means
that we could not conclude that there was an association between race or
gender and the likelihood of a military justice action.
We report the results from our regression models as odds ratios. We
generally report multivariate results from testing associations between
key attributes—including race, ethnicity, gender, rank, and education—on
a servicemember’s likelihood of being investigated and disciplined for a
UCMJ violation. In the body of this report, we focused on race and gender
disparities among servicemembers investigated and disciplined for
violations of the UCMJ, while holding other factors constant; however, our
analyses of recorded investigations and general and special courtsmartial for drug and sexual assault offenses are discussed in Appendix III.
In all of these analyses for the Air Force, we also controlled for years of
service among the lower enlisted ranks (E1-E4). In the analyses we
conducted for the Army, we could not control for education, but we were
able to control for age.
All regression models are subject to limitations. For our analyses, the
limitations included:

Page 87

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

•

Results of our analyses are associational and do not imply a causal
relationship. 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. Our analyses of
these data in finding the presence or absence of racial or gender
disparities, taken alone, do not establish the presence or absence of
unlawful discrimination, as that is a legal determination that would
involve other corroborating information along with supporting
statistics.

•

We could not assess some attributes that potentially could be related
to a servicemember’s likelihood of facing a military justice action in the
data analyzed for this review. For example, a servicemember’s
socioeconomic background or receipt of a waiver upon entering the
service could potentially be related to the likelihood of being
investigated, tried in a court-martial, or subject to a nonjudicial
punishment. However, we were unable to test these associations
because most services indicated they did not have information about
socioeconomic status or waivers in the databases that we requested
data from. Furthermore, while some other attributes may have been
available—such as marital status of the subject or the number of
dependent children—we did not include these attributes in our data
requests because we prioritized analyzing other demographic factors
based on our background research and conversations with service
officials.

•

As outlined above, we incorporated input from service officials to the
extent possible as we prepared our modeling groups for the
demographic and administrative attributes we tested, such as rank,
education, and years in service. However, this process was
necessarily imprecise. Our modeling results may have been impacted
by our discretionary decisions to include certain values in the groups
we created for these variables.

Data reliability. We conducted data reliability assessments on the
datasets we received from the databases in our review. We examined the
documentation officials provided to us on each database and conducted
electronic tests on the data we received to check for completeness and
accuracy. We also sent data reliability questionnaires to database
managers about how the data are collected and their appropriate uses,
and had discussions with database managers to discuss the reliability of
the data in their databases. When we determined that particular fields
were not sufficiently reliable, we excluded them from our analysis. For
example, we did not use data in our analysis where a substantial number

Page 88

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

of values were missing. We also checked to see that the values for
variables were internally consistent and that results were not affected
unduly by outlier values that might suggest miscoded values. For the
purposes of our analysis, we found the variables we ultimately reported
on to be sufficiently reliable. Furthermore, due to the sensitivity of the
information analyzed in this report, we did not include information in
instances where the number of servicemembers subjected to a particular
military justice action was fewer than 20, to protect privacy.
Literature review. To assess the extent to which disparities in the military
justice system and the civilian justice system had been previously
assessed, we conducted a literature review. To identify relevant
publications about disparities in the military justice system and the civilian
justice system, we performed a literature search of a number of
bibliographic databases, including ProQuest Academic, ProQuest Dialog,
Scopus, EBSCO, and HeinOnline. We also searched two think tank
search engines: Policy File and the Think Tank Search (from the Harvard
Kennedy School). We received the following types of publications:
scholarly/peer reviewed material, dissertations, and association/think
tank/nonprofit publications. To identify publications by DOD and the
services related to the military justice system, we reviewed prior GAO
reports and asked officials at the DOD Office of Diversity, Equity and
Inclusion, and in the services’ respective diversity and inclusion offices to
identify relevant publications. We concluded our searches in October
2018. We also asked the service Judge Advocate General offices for
publications relevant to disparities in military justice. We also identified
publications in our own background information search. We reviewed
those publications that assessed racial, ethnic, or gender disparities
among servicemembers in the military justice system. While the civilian
and military justice systems differ from each other, we selected a few
nationwide studies examining disparities in the civilian justice system to
summarize in the background section of our report, in order to enhance
our understanding of the complexities of the issues, including how others
have attempted to measure disparities. We did not assess the
methodologies used in any of these studies or the reliability of the data
cited in the studies; the studies related to the civilian justice system are
discussed in our report to provide broader context for the discussion
about racial and gender disparities in the military justice system.
We conducted this performance audit from November 2017 to May 2019
in accordance with generally accepted government auditing standards.
Those standards require that we plan and perform an audit to obtain
sufficient, appropriate evidence to provide a reasonable basis for our

Page 89

GAO-19-344 Military Justice

Appendix I: Objectives, Scope, and
Methodology

findings and conclusions based on our audit objectives. We believe that
the evidence obtained provides a reasonable basis for our findings and
conclusions based on our audit objectives.

Page 90

GAO-19-344 Military Justice

Appendix II: Summary Statistics and
Bivariate Results for Regression Analyses
Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

This appendix contains several figures that show the underlying data and
analyses used throughout the report relating to investigations and military
justice disciplinary actions from fiscal years 2013 through 2017. Our
analyses of the services’ investigations, military justice, and personnel
databases, as reflected in these tables, taken alone, do not establish the
presence or absence of unlawful discrimination. We could not analyze
Coast Guard general, special, and summary courts-martial due to the
small number of cases adjudicated in the Coast Guard from fiscal years
2013 through 2017. In addition, we could not analyze nonjudicial
punishments in the Army, the Navy, and the Coast Guard because these
services do not collect complete nonjudicial punishment information. The
following figures and information are included in this appendix:
•

Figure 15: Rate and Likelihood of Recorded Investigations for Alleged
Violations of the Uniform Code of Military Justice by Race and
Gender, without Controlling for Any Other Attributes, Fiscal Years
2013–2017 1

•

Figure 16: Rate and Likelihood of Trial in General and Special CourtsMartial by Race and Gender, without Controlling for Any Other
Attributes, Fiscal Years 2013–2017

•

Figure 17: Rate and Likelihood of Trial in General and Special CourtsMartial Following a Recorded Investigation by Race and Gender,
without Controlling for Any Other Attributes, Fiscal Years 2013–2017

•

Figure 18: Rate and Likelihood of Trial in General and Special CourtsMartial without a Recorded Investigation by Race and Gender, without
Controlling for Any Other Attributes, Fiscal Years 2013–2017

•

Figure 19: Rate and Likelihood of Trial in Summary Courts-Martial in
the Air Force and the Marine Corps by Race and Gender, without
Controlling for Any Other Attributes, Fiscal Years 2013–2017

•

Figure 20: Rate and Likelihood of Nonjudicial Punishments in the Air
Force and the Marine Corps by Race and Gender, without Controlling
for Any Other Attributes, Fiscal Years 2013–2017

1
For purposes of this report, we use the term “likelihood” when discussing the odds ratios
from the results of our regression analyses. Odds ratios that are statistically significant
and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic
are more likely or less likely, respectively, to be subject to a particular military justice
action.

Page 91

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Rate and Likelihood of
Recorded Investigations
by Race and Gender

•

Figure 21: Rate and Likelihood of Conviction in General and Special
Courts-Martial by Race and Gender, without Controlling for Any Other
Attributes, Fiscal Years 2013–2017

•

Figure 22: Rate and Likelihood of Dismissal or Discharge in General
and Special Courts-Martial in Navy and Marine Corps by Race and
Gender, without Controlling for Any Other Attributes, Fiscal Years
2013–2017

•

Figure 23: Rate and Likelihood of More Severe Punishment in
General and Special Courts-Martial in Army and Air Force by Race
and Gender, without Controlling for Any Other Attributes, Fiscal Years
2013–2017

As shown in figure 15 below, our analysis of data contained in the military
services’ military criminal investigations databases found that Black
servicemembers were subjects of recorded investigations at a higher rate
compared to their proportion of the overall service population in all of the
military services. 2 Hispanic servicemembers were the subjects of
recorded investigations at a higher rate compared to their proportion of
the overall service population in the Navy and the Air Force, at a lower
rate in the Marine Corps, and at the same rate in the Army. 3 Additionally,
we found that males were the subjects of recorded investigations at
higher rates than their share of the general service population in all of the
military services.

2

Our analysis focused on violations of the UCMJ that were recorded in databases used by
service specific investigative entities known as military criminal investigative organizations
(MCIO). MCIOs conduct criminal investigations in cases with a DOD nexus, such as if a
crime occurred on a DOD installation, or the subject of the investigation is currently
affiliated with DOD or was subject to the UCMJ at the time of the offense. Investigations
are recorded in the MCIO databases when a servicemember is the subject of a criminal
allegation made by another; for purposes of this report, we say the servicemember had a
“recorded investigation” to describe these cases.
3

The degree to which a racial or gender group was determined to have a higher or lower
rate was calculated through a comparison between the racial and gender compositions of
those who were the subjects of recorded investigations and the racial and gender
compositions of the military services’ total populations.

Page 92

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Figure 15: Rate and Likelihood of Recorded Investigations for Alleged Violations of the Uniform Code of Military Justice by
Race and Gender, without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely than all other
groups to be the subject of an investigation recorded in the services’ military criminal investigative
organizations databases for alleged violations of the Uniform Code of Military Justice. Odds ratios
that are statistically significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that
individuals with that characteristic are more likely or less likely, respectively, to be the subject of a
recorded investigation. Not statistically significant means that we could not conclude there was an

Page 93

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

association between race and the likelihood of a recorded investigation. The Other race category
includes individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other
Pacific Islander, and multiple races. Percentages in this figure may not sum to 100 percent due to
rounding and/or exclusion of data for those with an unknown race.

In addition, figure 15 above also shows the results of our bivariate
analyses, which calculated the degree to which one racial or gender
group was more likely or less likely than another racial or gender group to
be the subject of recorded investigations. 4 Our bivariate analyses found
that Black and male servicemembers in all of the military services were
statistically significantly more likely to be the subjects of recorded
investigations for alleged UCMJ violations than servicemembers of all
other races or females. Hispanic servicemembers were statistically
significantly more likely in the Navy, the Air Force, and the Coast Guard,
and were statistically significantly less likely in the Army to be the subjects
of recorded investigations than servicemembers of all other races.
Servicemembers in the Other race category were statistically significantly
less likely than servicemembers of all other races to be the subjects of
recorded investigations in the Army and the Marine Corps. Our bivariate
analyses did not show any statistically significant differences for
servicemembers in the Other race category in the Navy, the Air Force, or
the Coast Guard, or Hispanic servicemembers in the Marine Corps. 5

Rate and Likelihood of
Trial in General and
Special Courts-Martial

As shown in figure 16 below, Black, Hispanic, and male servicemembers
in all of the military services included in this analysis were represented at
a higher rate than their proportions of the overall service population. 6
White and female servicemembers in all of the military services were
represented at a lower rate than their proportions of the overall service
population. Servicemembers in the Other race category were represented
at a higher rate in the Navy, at a lower rate in the Army and the Air Force,
4
We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.
5
Not identifying any statistically significant findings means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case, a
recorded investigation.
6

The degree to which a racial or gender group was determined to have a higher or lower
rate was calculated through a comparison between the racial and gender compositions of
those tried in general and special courts-martial and the racial and gender compositions of
the military services’ total populations.

Page 94

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

and at the same rate in the Marine Corps compared to their proportion of
the overall service population. 7 We could not analyze Coast Guard cases
due to the small number of general and special courts-martial adjudicated
in the Coast Guard from fiscal years 2013 through 2017.
Figure 16: Rate and Likelihood of Trial in General and Special Courts-Martial by Race and Gender, without Controlling for Any
Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely than all other
groups to be tried in general and special courts-martial. Odds ratios that are statistically significant (pvalue < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic
are more likely or less likely, respectively, to be tried in general and special courts-martial. Not
statistically significant means that we could not conclude there was an association between race and
the likelihood of trial in general and special courts-martial. The Other race category includes
7

The Other race category includes individuals who identified as American Indian/Alaska
Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 95

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific
Islander, and multiple races. Percentages in this figure may not sum to 100 percent due to rounding
and/or exclusion of data for those with an unknown race.

The bivariate regression analysis results in figure 16 above calculate the
degree to which one racial or gender group was more likely or less likely
than servicemembers of all other races and genders to be tried in general
and special courts-martial. 8 We found that Black and male
servicemembers in all of the military services were more likely to be tried
in general and special courts-martial than servicemembers of all other
races or females. Our bivariate analyses found that Hispanic
servicemembers in the Army were more likely to be tried in general and
special courts-martial than servicemembers of all other races. We found
no statistically significant differences in the likelihood of Hispanic
servicemembers to be tried in general and special courts-martial
compared to servicemembers of all other races in the Navy, the Marine
Corps, and the Air Force. White and female servicemembers in all of the
military services were less likely to be tried in general and special courtsmartial than servicemembers of other races or males. Furthermore,
servicemembers in the Other race category were more likely in the Navy
and less likely in the Army to be tried in general and special courts-martial
than servicemembers of other races. We found no statistically significant
differences in the likelihood of servicemembers in the Other race category
to be tried in general and special courts-martial in the Marine Corps and
the Air Force compared to servicemembers of other races.

Rate and Likelihood of
Trial in General and
Special Courts-Martial
Following a Recorded
Investigation

As shown in figure 17 below, for trials in general and special courtsmartial that followed a recorded investigation, Black servicemembers
were represented at a lower rate in the Army, the Navy, and the Marine
Corps, and at the same rate in the Air Force compared to their
proportions of the service population that had recorded investigations. 9
Hispanic servicemembers in trials of general and special courts-martial
8
We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.
9
The degree to which a racial or gender group was determined to have a higher or lower
rate was calculated through a comparison between the racial and gender compositions of
those tried in general and special courts-martial following a recorded investigation and the
racial and gender compositions of the populations with recorded investigations in the
military services.

Page 96

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

following a recorded investigation were represented at a higher rate than
their proportion of the overall service population that had recorded
investigations in the Army and the Marine Corps, and at the same rate in
the Navy and the Air Force. White servicemembers were represented at a
lower rate in the Army, the Navy, and the Marine Corps, and at the same
rate in the Air Force compared to their proportions of the service
population with recorded investigations. Servicemembers in the Other
race category were represented at a higher rate in the Army, the Navy,
and the Marine Corps, and at the same rate in the Air Force compared to
their proportions of the overall service population with recorded
investigations. 10 We could not analyze Coast Guard cases due to the
small number of general and special courts-martial adjudicated in the
Coast Guard from fiscal years 2013 through 2017.
Male servicemembers with trials in general and special courts-martial that
followed a recorded investigation were represented at a higher rate in all
of the military services compared to their proportions of the service
population that had recorded investigations. Females were represented at
a lower rate in all of the military services compared to their proportions of
the service population that had recorded investigations.

10

The Other race category includes individuals who identified as American Indian/Alaska
Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 97

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Figure 17: Rate and Likelihood of Trial in General and Special Courts-Martial Following a Recorded Investigation by Race and
Gender, without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely than all other
groups to be tried in general and special courts-martial following an investigation recorded in the
services’ military criminal investigative organizations databases. Odds ratios that are statistically
significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that
characteristic are more likely or less likely, respectively, to be tried in general and special courtsmartial following a recorded investigation. Not statistically significant means that we could not
conclude there was an association between race and the likelihood of trial in general and special
courts-martial following a recorded investigation. The Other race category includes individuals who
identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. Percentages in this figure may not sum to 100 percent due to rounding and/or
exclusion of data for those with an unknown race.

Page 98

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

As shown in figure 17 above, our bivariate regression analyses showed
that, in the Army, White servicemembers were statistically significantly
less likely to be tried in general and special courts-martial following a
recorded investigation than servicemembers of all other races, whereas
Hispanic servicemembers were statistically significantly more likely to be
tried following a recorded investigation. 11 In the Navy, servicemembers in
the Other race category were statistically significantly more likely to be
tried in general and special courts-martial following a recorded
investigation than servicemembers of all other races. Males were more
likely, and females were less likely, to be tried in general and special
courts-martial following a recorded investigation in the Army and the Air
Force. The remaining odds ratios shown in figure 17 above were not
statistically significant. 12

Rate and Likelihood of
Trial in General and
Special Courts-Martial
without Recorded
Investigation

We identified racial and gender disparities in the rate and likelihood of trial
in general and special courts-martial in cases without a recorded
investigation in all of the military services. Specifically, as shown in figure
18 below, for trials in general and special courts-martial without a
recorded investigation, Black and male servicemembers in all of the
military services were represented at a higher rate than their proportion of
the service population that did not have a recorded investigation. 13
Hispanic servicemembers were represented at a higher rate in the Army
and the Marine Corps, and at the same rate in the Navy and the Air Force
compared to their proportions of the service population that did not have a
recorded investigation. Servicemembers in the Other race category were
represented at a lower rate in the Marine Corps and the Air Force, and at
the same rate in the Army and the Navy compared to their proportion of

11

We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.

12

Not identifying any statistically significant differences means that we could not conclude
there was an association between race and the likelihood of an outcome, in this case, trial
in general and special courts-martial without a recorded investigation.
13

The degree to which a racial or gender group was determined to have a higher or lower
rate was calculated through a comparison between the racial and gender compositions of
those tried in general and special courts-martial without a recorded investigation and the
racial and gender compositions of the overall population without recorded investigations in
the military services.

Page 99

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

the overall service population that did not have a recorded investigation. 14
White and female servicemembers in all of the military services were
represented at a lower rate than their proportions of the overall service
population without a recorded investigation. We could not analyze Coast
Guard cases due to the small number of general and special courtsmartial adjudicated in the Coast Guard from fiscal years 2013 through
2017.
Figure 18: Rate and Likelihood of Trial in General and Special Courts-Martial without a Recorded Investigation by Race and
Gender, without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely than all other
14

The Other race category includes individuals who identified as American Indian/Alaska
Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 100

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

groups to be tried in general and special courts-martial without an investigation recorded in the
services’ military criminal investigative organizations databases. Odds ratios that are statistically
significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that
characteristic are more likely or less likely, respectively, to be tried in general and special courtsmartial without a recorded investigation. Not statistically significant means that we could not conclude
there was an association between race and the likelihood of trial in general and special courts-martial
without a recorded investigation. The Other race category includes individuals who identified as
American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.
Percentages in this figure may not sum to 100 percent due to rounding and/or exclusion of data for
those with an unknown race.

The bivariate regression analysis results in figure 18 above calculate the
degree to which one racial or gender group was more likely or less likely
than servicemembers of all other races and genders to be tried in general
and special courts-martial without a recorded investigation. 15 We found
that Black and male servicemembers in all of the military services were
more likely to be tried at special and general courts-martial that were not
preceded by a recorded investigation than servicemembers of all other
races or females. White and female servicemembers in all of the military
services were less likely to be tried at special and general courts-martial
that were not preceded by a recorded investigation than servicemembers
of all other races and males. We found no statistically significant
differences in the likelihood of Hispanic servicemembers or
servicemembers in the Other race category in any of the military services
being tried in general and special courts-martial without a recorded
investigation compared to servicemembers of all other races.

Rate and Likelihood of
Trial in Summary CourtsMartial in the Air Force
and the Marine Corps

We identified racial and gender disparities in the rate and likelihood of trial
in summary courts-martial in the Air Force and the Marine Corps.
Specifically, as shown in figure 19 below, Black and male
servicemembers were tried in summary courts-martial for UCMJ
violations at higher rates than their share of the overall service population
in the Air Force and the Marine Corps. 16 White and Hispanic
servicemembers were tried in summary courts-martial at lower rates than
15

We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.

16

The degree to which a racial, ethnic, or gender group was determined to have a higher
or lower rate was calculated through a comparison between the racial, ethnic, and gender
compositions of those tried in summary courts-martial and the racial, ethnic, and gender
compositions of the military services’ total populations.

Page 101

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

their share of the overall service population in both services.
Servicemembers that were included in the Other race category were tried
at higher rates in the Air Force, and at lower rates in the Marine Corps. 17
We could not determine whether there were any racial or gender
disparities for summary courts-martial in the Army and the Navy because
these services did not collect complete summary court-martial data—
information about all summary court-martial cases, to include
demographic information about the subject—in their investigative, military
justice, or personnel databases, as discussed above in the report. We
could not analyze Coast Guard cases due to the small number of
summary courts-martial adjudicated in the Coast Guard from fiscal years
2013 through 2017.
Figure 19: Rate and Likelihood of Trial in Summary Courts-Martial in the Air Force and the Marine Corps by Race and Gender,
without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely than all other
groups to be tried in summary courts-martial. Odds ratios that are statistically significant (p-value <
0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic are
more likely or less likely, respectively, to be tried in summary courts-martial. Not statistically
significant means that we could not conclude there was an association between race and the
likelihood of trial in summary courts-martial. The Other race category includes individuals who
identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. Percentages in this figure may not sum to 100 percent due to rounding, exclusion of
data for those with an unknown race, and/or to ensure protection of sensitive statistical information.
17

The Other race category includes servicemembers that identify as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 102

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

The bivariate regression analysis results in figure 19 above calculate the
degree to which one racial or gender group was more likely or less likely
than servicemembers of all other races and genders to be tried in
summary courts-martial. 18 We found that Black servicemembers in the
Marine Corps and the Air Force were more likely to be tried in summary
courts-martial than servicemembers of all other races. We also found that
male servicemembers were more likely than their female counterparts to
be tried in summary courts-martial in the Marine Corps and the Air Force.
We observed no statistically significant differences in summary courtmartial rates for servicemembers in the Other race category in either the
Marine Corps or the Air Force, or for Hispanic servicemembers in the
Marine Corps. 19

Rate and Likelihood of
Nonjudicial Punishments
in the Air Force and the
Marine Corps

As shown in figure 20 below, we found that Black and male
servicemembers were subject to nonjudicial punishment for UCMJ
violations at a higher rate than their share of the overall service
population in the Marine Corps and the Air Force. White servicemembers
were subject to nonjudicial punishments at lower rates than their share of
the overall service population in both services, and Hispanic
servicemembers were subject to nonjudicial punishments in a proportion
equal to their share of the general service population in both services.
Servicemembers that were included in the Other race category were
subject to nonjudicial punishment at lower rates than their share of the
overall service population in the Marine Corps and the Air Force. 20 We
could not analyze nonjudicial punishments in the Army, the Navy, and the
Coast Guard because these services do not collect complete nonjudicial
punishment information.

18

We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.

19

The Other race category includes servicemembers who identify as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.
Statistical insignificance indicates that we could not conclude there was an association
between race and the likelihood of an outcome, in this case, trial in summary courtsmartial.

20

The Other race category includes servicemembers that identify as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 103

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Figure 20: Rate and Likelihood of Nonjudicial Punishments in the Air Force and the Marine Corps by Race and Gender,
without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely than all other
groups to be subject to nonjudicial punishments. Odds ratios that are statistically significant (p-value <
0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic are
more likely or less likely, respectively, to be subject to nonjudicial punishment. Not statistically
significant means that we could not conclude there was an association between race and the
likelihood of nonjudicial punishment. The Other race category includes individuals who identified as
American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.
Percentages in this figure may not sum to 100 percent due to rounding and/or exclusion of data for
those with an unknown race.

The bivariate regression analyses in figure 20 above calculate the degree
to which one racial or gender group was more likely or less likely than
another racial or gender group to be subject to nonjudicial punishment. 21
We found that Black and male servicemembers were more likely than
servicemembers of all other races or female servicemembers to receive
nonjudicial punishments in the Marine Corps and the Air Force. We also
found that Hispanic servicemembers in the Air Force were less likely to
be subject to nonjudicial punishment, but we observed no statistically
significant difference for Hispanic servicemembers in the Marine Corps.
Servicemembers in the Other race category were less likely to be subject
21

We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.

Page 104

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

to nonjudicial punishment than servicemembers of all other races in the
Marine Corps and the Air Force. 22

Rate and Likelihood of
Conviction in General and
Special Courts-Martial

As shown in figure 21 below, we found that Black servicemembers were
convicted in general and special courts-martial at a lower rate in the Army
and the Air Force, and at an equal rate in the Navy and the Marine Corps
compared to their proportion of the overall general and special courtsmartial population. 23 In the Army, the Navy, and the Marine Corps,
Hispanic servicemembers were convicted in general and special courtsmartial at an equal rate compared to their proportion of the overall general
and special courts-martial population. Compared to their proportion of the
overall general and special courts-martial population, Hispanic
servicemembers were convicted at a lower rate in the Air Force. We could
not analyze Coast Guard cases due to the small number of general and
special courts-martial adjudicated in the Coast Guard from fiscal years
2013 through 2017.

22

The Other race category includes servicemembers that identify as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

23

The degree to which a racial or gender group was determined to have a higher or lower
rate was calculated through a comparison between the racial and gender compositions of
those convicted in general and special courts-martial and the racial and gender
compositions of the service population that were tried in general and special courtsmartial.

Page 105

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Figure 21: Rate and Likelihood of Conviction in General and Special Courts-Martial by Race and Gender, without Controlling
for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely to be convicted
in general and special courts-martial compared to all other racial or gender groups. Odds ratios that
are statistically significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that
individuals with that characteristic are more likely or less likely, respectively, to be convicted in
general and special courts-martial. Not statistically significant means that we could not conclude there
was an association between race and the likelihood of conviction in general and special courtsmartial. The Other race category includes individuals who identified as American Indian/Alaska
Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races. Percentages in this figure
may not sum to 100 percent due to rounding and/or exclusion of data for those with an unknown race.

As shown in figure 21 above, bivariate regression analyses found that, in
the Army, White servicemembers were statistically significantly more
likely to be convicted, whereas Black servicemembers were statistically

Page 106

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

significantly less likely to be convicted in general and special courtsmartial compared to all other servicemembers. 24 White servicemembers
in the Air Force were also statistically significantly more likely to be
convicted in general and special courts-martial compared to all other
servicemembers. In the Marine Corps, we found that males were more
likely to be convicted than females, whereas in the Air Force, males were
less likely to be convicted than females. The remaining odds ratios shown
in figure 21 above were not statistically significant. 25

Rate and Likelihood of
More Severe Punishment

As shown in figures 22 and 23 below, we found that Black
servicemembers received a more severe punishment at a lower rate
compared to their share of the convicted service population in the Army,
the Navy, and the Air Force. 26 We also found that Hispanic
servicemembers received a more severe punishment at a lower rate
compared to their share of the convicted service population in the Air
Force, but at a higher rate in the Marine Corps. We found that male
servicemembers in the Marine Corps and the Air Force received a more
severe punishment at a higher rate, and at the same rate in the Army and
the Navy, compared to their share of the convicted service population.
Females received a more severe punishment at a lower rate in the Air
Force and the Marine Corps, and at the same rate in the Army and the
Navy, compared to their share of the convicted service population. We
could not analyze Coast Guard cases due to the small number of general
and special courts-martial adjudicated in the Coast Guard from fiscal
years 2013 through 2017.

24

We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.

25

Not identifying any statistically significant findings means that we could not conclude
there was an association between race or gender and the likelihood of an outcome, in this
case, conviction in general and special courts-martial.

26

We measured the severity of punishments in two groups for the Navy and the Marine
Corps, and in three groups for the Air Force and the Army, which are defined in Appendix
I. We did not create a third punishment group for confinement without dismissal or
discharge for the Navy and the Marine Corps because of the small number of cases with
confinement that did not also include some sort of discharge. Based on discussions with
service officials, we determined that a sentence resulting in a dismissal or discharge was
the most severe punishment outcome.

Page 107

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Figure 22: Rate and Likelihood of Dismissal or Discharge in General and Special Courts-Martial in Navy and Marine Corps by
Race and Gender, without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These bivariate regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely to be dismissed
or discharged after conviction in general and special courts-martial compared to all other racial or
gender groups. Odds ratios that are statistically significant (p-value < 0.05) and greater than 1.00 or
lower than 1.00 indicate that individuals with that characteristic are more likely or less likely,
respectively, to be dismissed or discharged after conviction in general and special courts-martial. Not
statistically significant means that we could not conclude there was an association between race and
the likelihood of dismissal or discharge after conviction in general and special courts-martial. The
Other race category includes individuals who identified as American Indian/Alaska Native, Asian,
Native Hawaiian/Other Pacific Islander, and multiple races. Percentages in this figure may not sum to
100 percent due to rounding, exclusion of data for those with an unknown race, and/or to ensure
protection of sensitive statistical information.

Page 108

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

Figure 23: Rate and Likelihood of More Severe Punishment in General and Special Courts-Martial in Army and Air Force by
Race and Gender, without Controlling for Any Other Attributes, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These regression analysis results
demonstrate the degree to which a racial or gender group is more likely or less likely to receive a
more severe punishment after conviction in general and special courts-martial compared to all other
racial or gender groups. Because the punishment groups for the Army and the Air Force consisted of
3 outcome categories, we used an ordered logit regression model to analyze punishment severity in
the Army and the Air Force. Odds ratios that are statistically significant (p-value < 0.05) and greater
than 1.00 or lower than 1.00 indicate that individuals with that characteristic are more likely or less
likely, respectively, to receive a more severe punishment. Not statistically significant means that we
could not conclude there was an association between race and the likelihood of a more severe
punishment. Punishment severity in the Air Force, ordered from most to least severe, was (3) any
type of dismissal or discharge (regardless of any confinement); (2) confinement without dismissal or
discharge, and (1) all other possible sentencing options. In the Army, it was (3) any type of dismissal
or discharge or confinement of more than 2 years, (2) confinement of less than 2 years without
dismissal or discharge, and (1) all other possible sentencing options. The Other race category
includes individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other
Pacific Islander, and multiple races. Percentages in this figure may not sum to 100 percent due to

Page 109

GAO-19-344 Military Justice

Appendix II: Summary Statistics and Bivariate
Results for Regression Analyses

rounding, exclusion of data for those with an unknown race, and/or to ensure protection of sensitive
statistical information.

The bivariate regression analyses in Figures 22 and 23 above calculated
the degree to which one racial or gender group was more likely or less
likely than another racial or gender group to be dismissed or discharged
after a conviction in general and special courts-martial. 27 In the Navy, we
found that Black servicemembers were statistically significantly less likely
to be dismissed or discharged after conviction in general and special
courts-martial compared to all other servicemembers. We found no
statistically significant differences regarding minority servicemembers
being more likely or less likely to be dismissed or discharged after
conviction in general and special courts-martial in the Marine Corps, or to
receive a more severe punishment in the Army or the Air Force. 28 We
found that males in the Marine Corps and the Air Force were more likely
to be dismissed or discharged or receive a more severe punishment after
conviction than females, but we did not find any statistically significant
differences regarding male servicemembers in the Army or the Navy.

27
We conducted bivariate logit analyses (which we refer to as bivariate analyses) to
estimate the association between the attribute factors (or independent variables) and the
outcome variables (the dependent variable) in a binary format. For additional explanation
of how we conducted our bivariate analyses, see Appendix I.
28

Not identifying any statistically significant findings means that we could not conclude
there was an association between race or gender and the likelihood of an outcome, in this
case, punishment severity.

Page 110

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses
Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

This appendix contains several figures that show the underlying data
related to drug and sexual assault offenses from fiscal years 2013
through 2017 for the Army, the Navy, the Marine Corps, and the Air
Force. 1 Across most military services, Black, Hispanic, and male
servicemembers were the subjects of recorded investigations and tried in
general and special courts-martial at higher rates than their shares of the
overall service population for drug offenses, sexual assault offenses, and
all other offenses. 2 We found that the likelihood of conviction varied
among the services for these two offenses. 3 We analyzed these two
specific Uniform Code of Military Justice (UCMJ) offenses separately from
all other offenses because service officials told us that an investigation
into these offenses may frequently be mandatory, and thus could
potentially mitigate the risk of bias. 4 We analyzed data for these offenses
for recorded investigations, trials in general and special courts-martial,
and convictions from fiscal years 2013 through 2017 to assess the extent
to which racial and gender disparities may exist. 5 Our analyses of the
1

We could not analyze Coast Guard cases due to the small number of general and special
courts-martial adjudicated in the Coast Guard from fiscal years 2013 through 2017.

2

The drug offenses we analyzed were charges under UCMJ Article 112a, wrongful use,
possession, or distribution of a controlled substance. In addition, the sexual assault
offenses we analyzed were any charges under UCMJ Article 120, which includes rape and
sexual assault generally, rape and sexual assault of a child, and other sexual misconduct.
3

For purposes of this report, we use the term “likelihood” when discussing the odds ratios
from the results of our regression analyses. Odds ratios that are statistically significant
and greater than 1.00 or lower than 1.00 indicate that individuals with that characteristic
are more likely or less likely, respectively, to be subject to a particular military justice
action.
4
Service officials stated that some drug (Article 112a) offenses are initiated as a result of
random urinalysis tests, and in those cases a positive result will trigger an investigation
regardless of the servicemember’s race, ethnicity, or gender. According to Department of
Defense Instruction 5505.18, all allegations of adult sexual assault are immediately
reported to the appropriate military criminal investigative organization (MCIO), and that
MCIO will initiate a criminal investigation into that allegation if the offense occurred within
its jurisdiction.
5

We were unable to present an analysis of UCMJ offenses tried in summary courts-martial
due to data limitations. The Air Force was the only service that maintained offense data for
summary courts-martial for fiscal years 2013 through 2017. The data for summary courtsmartial in the Marine Corps personnel database did not include information about the
offense type. See Appendix VII for our analysis of the Air Force summary courts-martial
offense data. Furthermore, all analyses presented in this section examine the number of
instances in which an offense was investigated or tried in general and special courtsmartial. Multiple offenses may be involved in an investigation or court-martial of a
servicemember; if a case involved both a drug offense and a sexual assault offense, it
was counted in both groups for purposes of this analysis. See Appendixes II through VI for
the demographic breakdowns of each of those offenses in each of the military services.

Page 111

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

services’ investigation, military justice, and personnel databases, as
reflected in these figures, taken alone, do not establish the presence or
absence of unlawful discrimination.
The following figures and information are included in this appendix:

Recorded Investigations of
Drug and Sexual Assault
Offenses

•

Figure 24: Recorded Investigation Rates for Drug Offenses, Sexual
Assault Offenses, and All Other Offenses by Race and Gender, Fiscal
Years 2013–2017

•

Figure 25: General and Special Courts-Martial Trial Rates for Drug
Offenses, Sexual Assault Offenses, and All Other Offenses by Race
and Gender, Fiscal Years 2013–2017

•

Figure 26: Likelihood that Charges of Drug Offenses and Sexual
Assault Offenses Resulted in Convictions in General and Special
Courts-Martial, After Controlling for Race, Gender, Rank, and
Education, Fiscal Years 2013–2017

We identified racial and gender differences in recorded investigation rates
for drug offenses, sexual assault offenses, and all other offenses
compared with the total service populations. 6 Our analysis focused on
alleged UCMJ violations for these offenses that were recorded in the
Military Criminal Investigative Organization (MCIO) investigations
databases. Other investigations conducted within the military, such as
command investigations, were not considered in this analysis. For
example, as shown in figure 24 below, Black servicemembers were the
subjects of recorded investigations for drug offenses, sexual assault
offenses, and all other offenses at a higher rate than their share of the
overall service population across all military services. 7 Hispanic
6
Investigations are recorded in the MCIO databases when a servicemember is the subject
of a criminal allegation made by another; for purposes of this report, we say the
servicemember had a “recorded investigation” to describe these cases. The remaining
general and special courts-martial cases would have been investigated by other sources,
such as local civilian law enforcement, command investigations, or in the case of the Air
Force, their military law enforcement security forces, and thus would not be recorded in
the MCIO databases. For additional explanation of the databases we used to analyze
investigations, please see Appendix I.
7
The degree to which a racial or gender group was determined to have a higher or lower
rate was calculated through a comparison between the racial and gender compositions of
those investigated for drug or sexual assault offenses and the racial and gender
compositions of the military services’ total populations.

Page 112

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

servicemembers were the subjects of recorded investigations for drug
offenses, sexual assault offenses, and all other offenses at a higher rate
than their share of the overall service population in the Air Force, but
were the subjects of recorded investigations for drug offenses at a lower
rate than their share of the overall service population in both the Army
and the Marine Corps. Male servicemembers were the subjects of
recorded investigations for drug offenses and sexual assault offenses at a
higher rate than their share of the overall service population across all of
the military services.
Figure 24: Recorded Investigation Rates for Drug Offenses, Sexual Assault Offenses, and All Other Offenses by Race and
Gender, Fiscal Years 2013–2017

Note: These analyses, taken alone, should not be used to make conclusions about the presence or
absence of unlawful discrimination. The degree to which the representation of a racial or gender
group is determined to be higher or lower in the population subject to an investigation recorded in the

Page 113

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

military services’ military criminal investigative organizations databases is calculated by subtracting
the group’s percentage of the military service’s total population from the percentage that group
represents within the population that was the subject of a recorded investigated. The Other race
category includes individuals who identified as American Indian/Alaska Native, Asian, Native
Hawaiian/Other Pacific Islander, and multiple races. Percentages in this figure may not sum to 100
percent due to rounding and/or exclusion of data for those with an unknown race.

General and Special
Courts-Martial Trials for
Drug and Sexual Assault
Offenses

We found that White servicemembers were tried for drug offenses, sexual
assault offenses, and all other offenses in general and special courtsmartial at lower rates than their share of the overall service population
across all of the military services. Black servicemembers were tried for
drug offenses, sexual assault offenses, and all other offenses in general
and special courts-martial at a higher rate than their share of the overall
service population in all of the military services. Hispanic servicemembers
were tried for drug offenses in general and special courts-martial at a
lower rate in the Navy and the Marine Corps, and at a higher rate in the
Air Force, compared to their share of the overall service population.
Hispanic servicemembers were tried for sexual assault offenses at a
higher rate than their proportion of the overall service population in all of
the military services. Female servicemembers were tried for drug
offenses, sexual assault offenses, and all other offenses in general and
special courts-martial at lower rates than their share of the general
service population in the Army, the Navy, and the Air Force, and were
tried for sexual assault offenses and all other offenses at lower rates than
their share of the overall service population in the Marine Corps. Figure
25 below shows the gender and racial composition of general and special
court-martial trials for drug offenses, sexual assault offenses, and all
other offenses. We could not analyze Coast Guard cases due to the small
number of general and special courts-martial adjudicated in the Coast
Guard from fiscal years 2013 through 2017.

Page 114

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

Figure 25: General and Special Courts-Martial Trial Rates for Drug Offenses, Sexual Assault Offenses, and All Other Offenses
by Race and Gender, Fiscal Years 2013–2017

Note: These analyses, taken alone, should not be used to make conclusions about the presence or
absence of unlawful discrimination. The degree to which the representation of a racial or gender
group is determined to be higher or lower in the population tried in special or general courts-martial is
calculated by subtracting the group’s percentage of the military service’s total population from the
percentage that group represents within the population tried. The Other race category includes
individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific
Islander, and multiple races. Percentages in this figure may not add up to 100 percent due to
rounding, exclusion of data for those with an unknown race, and/or to protect privacy.

Likelihood of Conviction
for Drug and Sexual
Assault Offenses

We conducted multivariate regression analyses to calculate the degree to
which servicemembers charged with drug offenses and sexual assault
offenses were more likely or less likely than a composite variable
comprised of all other offenses to be convicted in general and special
courts-martial, while controlling for other attributes, such as race, gender,

Page 115

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

education, and rank. 8 As shown in figure 26 below, we did not identify any
statistically significant difference in conviction rates for drug offenses
compared to all other offenses in the Army, the Navy, the Marine Corps,
and the Air Force. Sexual assault offenses were less likely to result in a
conviction in the Army, the Navy, and the Air Force, and there was no
statistically significant difference for the Marine Corps. We could not
analyze Coast Guard cases due to the small number of general and
special courts-martial adjudicated in the Coast Guard from fiscal years
2013 through 2017.
Figure 26: Likelihood that Charges of Drug Offenses and Sexual Assault Offenses
Resulted in Convictions in General and Special Courts-Martial, After Controlling for
Race, Gender, Rank, and Education, Fiscal Years 2013–2017

Note: The information presented in this figure, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These partial multivariate regression
analysis results demonstrate the degree to which a servicemember was more likely or less likely than
8

A multivariate regression analysis examines several variables simultaneously to estimate
whether each of these variables are more likely or less likely to be associated with a
certain outcome. See Appendix I for a more detailed explanation of how we conducted our
multivariate regression analysis, and a full explanation of the attributes we used in each
service model. In addition, see Appendixes IV through VIII for the demographic
breakdowns of the modeled attributes in each of the military services.

Page 116

GAO-19-344 Military Justice

Appendix III: Analysis of Drug Offenses,
Sexual Assault Offenses, and All Other
Offenses

the reference category to be convicted in general and special courts-martial after being charged with
drug offenses or sexual assault offenses after controlling for race, gender, rank, and education. We
also controlled for years of service among the lower enlisted ranks (E1-E4) in the Air Force. In the
Army, we could not control for education, but we were able to control for age. We made all offense
comparisons to a composite variable that contains all other offenses. Odds ratios that are statistically
significant (p-value < 0.05) and greater than 1.00 or lower than 1.00 indicate that individuals with that
characteristic are more likely or less likely, respectively, to be convicted. Not statistically significant
means that we could not conclude there was an association between race and the likelihood of a
recorded investigation.

Page 117

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses
Appendix IV: Army Data and Analyses

This appendix contains several tables that show the underlying data and
analyses used throughout this report relating to Army personnel and
military justice disciplinary actions from fiscal years 2013 through 2017.
We did not include populations that contained fewer than 20
servicemembers in the total populations presented in these tables to
ensure the protection of sensitive information. As a result, the total
populations presented in this appendix may vary among the different
tables and may vary from the total populations presented in the body of
the report. Our analyses of the Army’s investigations, military justice, and
personnel databases, as reflected in these tables, taken alone, do not
establish the presence or absence of unlawful discrimination.
The following tables and information are included in this appendix:
•

Table 10: Total Population of the Army by Race, Fiscal Years 2013–
2017

•

Table 11: Summary Statistics by Race for Army Military Justice
Actions, Fiscal Years 2013–2017

•

Table 12: Summary Statistics by Gender for Army Military Justice
Actions, Fiscal Years 2013–2017

•

Table 13: Summary Statistics by Rank for Army Military Justice
Actions, Fiscal Years 2013–2017

•

Table 14: Summary Statistics by Age for Army Military Justice
Actions, Fiscal Years 2013–2017

•

Table 15: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Army Military Justice Actions, Fiscal Years
2013–2017

•

Table 16: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Army Military Justice Actions, Fiscal
Years 2013–2017

•

Table 17: Odds Ratios for Army Multivariate Regression Analyses

Page 118

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 10: Total Population of the Army by Race, Fiscal Years 2013–2017
Population
Race

Number

Percent

7,873

1

American Indian or Alaskan Native
Asian or Pacific Islander

47,644

5

Black

183,379

21

Hispanic

117,413

13

Other

8,434

1

White

521,820

59

Total population

886,563

100%

Source: GAO analysis of Army personnel data. | GAO-19-344

Note: The total populations presented in this table represent the number of unique active-duty
servicemembers who served during fiscal years 2013 through 2017.

Table 11: Summary Statistics by Race for Army Military Justice Actions, Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

N

N

%

N

%

N

%

N

%

886,563

521,820

59%

183,379

21%

117,413

13%

63,951

7%

50,547

24,819

49%

16,648

33%

6,547

13%

2,533

5%

General and special
courts-martial

3,129

1,488

48%

972

31%

473

15%

196

6%

General and special
courts-martial with a
recorded investigation

2,107

988

47%

668

32%

327

16%

124

6%

General and special
courts-martial without a
recorded investigation

1,022

500

49%

304

30%

146

14%

72

7%

Acquittals (general and
special courts-martial)

383

159

42%

136

36%

56

15%

32

8%

2,746

1,329

48%

836

30%

417

15%

164

6%

Total population
Recorded investigations

Convictions (general and
special courts-martial)

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The Other race group includes individuals
who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. The total populations presented in this table represent the number of unique activeduty servicemembers during fiscal years 2013 through 2017. The term “recorded investigation” refers
to an investigation where a servicemember was the subject of a criminal investigation that was
recorded in the Army’s military criminal investigative organization’s database. Percentages in this
table may not add up to 100 due to rounding.

Page 119

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 12: Summary Statistics by Gender for Army Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded investigations

Male

Female

N

N

%

N

%

866,563

755,862

85%

130,701

15%

50,547

46,092

91%

4,455

9%

General and special courts-martial

3,129

3,028

97%

101

3%

General and special courts-martial
with a recorded investigation

2,107

2,059

98%

48

2%

General and special courts-martial
without a recorded investigation

1,022

969

95%

53

5%

Convictions (general and special
courts-martial)

2,746

2,652

97%

94

3%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total populations presented in this
table represent the number of unique active-duty servicemembers during fiscal years 2013 through
2017. The term “recorded investigation” refers to an investigation where a servicemember was the
subject of a criminal investigation that was recorded in the Army’s military criminal investigative
organization’s database.

Page 120

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 13: Summary Statistics by Rank for Army Military Justice Actions, Fiscal Years 2013–2017
Population

Rank E1-E4

Rank E5-E9

Officers

N

N

%

N

%

N

%

866,563

507,498

57%

251,488

28%

127,577

14%

50,547

35,669

71%

12,952

26%

1,926

4%

General and special
courts-martial

3,129

1,777

57%

1,172

37%

180

6%

General and special
courts-martial with a
recorded
investigation

2,107

1,203

57%

776

37%

128

6%

General and special
courts-martial without
a recorded
investigation

1,022

574

56%

396

39%

52

5%

383

177

46%

174

45%

32

8%

2,746

1,600

58%

998

36%

148

5%

Total population
Recorded
investigations

Acquittals (general
and special courtsmartial)
Convictions (general
and special courtsmartial)

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total populations presented in this
table represent the number of unique active-duty servicemembers during fiscal years 2013 through
2017. The term “recorded investigation” refers to an investigation where a servicemember was the
subject of a criminal investigation that was recorded in the Army’s military criminal investigative
organization’s database. Percentages in this table may not add up to 100 due to rounding.

Page 121

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 14: Summary Statistics by Age for Army Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded investigations

< 25 years

25-30 years

30-40 years

> 40 years

N

N

%

N

%

N

%

N

%

866,563

457,494

52%

172,427

19%

178,014

20%

78,628

9%

50,547

31,744

63%

9,498

19%

7,646

15%

1,659

3%

General and special
courts-martial

3,129

1,330

43%

767

25%

802

26%

230

7%

General and special
courts-martial with a
recorded investigation

2,107

972

46%

467

22%

533

25%

135

6%

General and special
courts-martial without a
recorded investigation

1,022

358

35%

300

29%

269

26%

95

9%

Acquittals (general and
special courts-martial)

383

145

38%

83

22%

121

32%

34

9%

2,746

1,185

43%

684

25%

681

25%

196

7%

Convictions (general and
special courts-martial)

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total populations presented in this
table represent the number of unique active-duty servicemembers during fiscal years 2013 through
2017. The term “recorded investigation” refers to where a servicemember was the subject of a
criminal investigation that was recorded in the Army’s criminal investigative organization’s database.
Percentages in this table may not add up to 100 due to rounding.

Page 122

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 15: Offenses Investigated and Tried in General and Special Courts-Martial by Race for Army Military Justice Actions,
Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

N

N

%

N

%

N

%

N

%

19,703

10,464

53%

6,401

32%

2,108

11%

730

4%

Recorded investigations for
sexual assault offenses (Article
120)

8,679

4,237

49%

2,541

29%

1,398

16%

503

6%

Recorded investigations for all
other offenses

39,352

18,503

47%

13,231

34%

5,390

14%

2,228

6%

404

240

57%

110

26%

54

13%

<20

General and special courtsmartial for sexual assault
offenses (Article 120)

1,442

665

46%

410

28%

269

19%

98

7%

General and special courtsmartial for all other offenses

1,997

925

46%

693

35%

260

13%

119

6%

Recorded investigations for drug
offenses (Article 112a)

General and special courtsmartial for drug offenses (Article
112a)

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The Other race group includes individuals
who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. This table summarizes the number of instances in which an offense was investigated
or tried in a general and special courts-martial. Multiple offenses may be incorporated into the
investigation or court-martial of a single servicemember. As such, a single case could be included in
multiple offense groups in this table. The term “recorded investigation” refers to where a
servicemember was the subject of a criminal investigation that was recorded in the Army’s criminal
investigative organization’s database. The summary statistics for servicemembers tried in general
and special courts-martial for drug offenses (Article 112a) were omitted from this table to protect
privacy because a gender group had fewer than 20 servicemembers. Percentages in this table may
not add up to 100 due to rounding.

Page 123

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 16: Offenses Investigated and Tried in General and Special Courts-Martial by Gender for Army Military Justice Actions,
Fiscal Years 2013–2017
Population

Male

Female

N

N

%

N

%

Recorded investigations for drug offenses
(Article 112a)

19,703

18,085

92%

1,618

8%

Recorded investigations for sexual assault
offenses (Article 120)

8,679

8,412

97%

267

3%

Recorded investigations for all other offenses

1,531

1,401

92%

130

8%

General and special courts- martial for all
other offenses

1,997

1,873

94%

124

6%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. This table summarizes the number of
instances in which an offense was investigated or tried in a general and special courts-martial.
Multiple offenses may be incorporated into the investigation or court-martial of a single
servicemember. As such, a single case could be included in multiple offense groups in this table. The
term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Army’s criminal investigative organization’s database. The
summary statistics for servicemembers tried in general and special courts-martial for drug offenses
(Article 112a) and sexual assault offenses (Article 120) were omitted from this table to protect privacy,
because a gender group had fewer than 20 servicemembers.

Multivariate Regression
Analyses of Army Data

The multivariate results listed below in table 17 show the odds ratios for
the multivariate regression analyses of the Army data. We used logistic
regression to assess the relationship between the independent variables,
such as race, education, rank, or gender, with the probability of being
subject to a military justice action. Logistic regression allows for the
coefficients to be converted into odds ratios. Odds ratios that are
statistically significant and greater than 1.00 indicate that individuals with
that characteristic are more likely to be subject to a military justice action.
For example, an odds ratio of 1.55 for Black servicemembers would mean
that they are 1.55 times more likely to be subject to a military justice
action compared to White servicemembers. Odds ratios that are
statistically significant and lower than 1.00 indicate that individuals with
that characteristic are less likely to be subject to a military justice action.
We excluded years of service from the Army analyses due to high
correlation with the rank variable.

Page 124

GAO-19-344 Military Justice

Appendix IV: Army Data and Analyses

Table 17: Odds Ratios for Army Multivariate Regression Analyses
Black

Hispanic

Other

Male

E5-E9

Officers

Likelihood of being subject
of recorded investigations

2.11**

1.13**

0.92**

2.14**

0.91**

0.30**

Likelihood of trial in general
and special courts-martial

2.00**

1.41**

1.12

5.75**

0.93

0.34**

Likelihood of trial in general
and special courts-martial
with a recorded
investigation

1.16**

1.37**

1.26*

3.92**

1.28**

1.32*

Likelihood of trial in general
and special courts-martial
without a recorded
investigation

1.85**

1.29**

1.16

3.53**

0.74**

0.22**

Likelihood of conviction in
general and special courtsmartial

0.8

1.11

0.74

0.91

0.76

0.58*

Likelihood of receiving a
more severe punishment
when convicted in general
and special courts-martial

0.84

0.99

0.56**

1.58*

0.34**

0.76

Legend: ** next to the odds ratio indicate that the finding has a strong degree of statistical significance with a p-value less than 0.01.
* next to the odds ratio indicates that the finding is statistically significant with a p-value between 0.01 and 0.05.
If the space next to the odds ratio is blank, then the finding was not statistically significant.
“N” refers to the population size for each group.
Source: GAO analysis of Army investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These multivariate regression analysis
results demonstrate the degree to which a racial, gender, or rank group is more likely than the
reference category to be subject of a recorded investigation, tried in general and special courtsmartial, and convicted in general and special courts-martial. The term “recorded investigation” refers
to where a servicemember was the subject of a criminal investigation that was recorded in the Army’s
military criminal investigative organization’s database. We used an ordered logistic regression
analysis to calculate the likelihood of receiving a more severe punishment as a result of being
convicted in general and special courts-martial. All racial categories listed are in reference to White
servicemembers, all gender groups listed are in reference to female servicemembers, and all rank
groups are in reference to servicemembers between ranks E1-E4. The Other race group includes
individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific
Islander, and multiple races.

Page 125

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses
Appendix V: Navy Data and Analyses

This appendix contains several tables that show the underlying data and
analyses used throughout this report relating to Navy personnel and
military justice disciplinary actions from fiscal years 2013 through 2017.
We did not include populations that contained fewer than 20
servicemembers in the populations presented in these tables to ensure
the protection of sensitive information. As a result, the populations
presented in this appendix may vary among the different tables and may
vary from the populations presented in other places in this report. Our
analyses of the Navy’s investigations, military justice, and personnel
databases, as reflected in these tables, taken alone, do not establish the
presence or absence of unlawful discrimination.
The following tables and information are included in this appendix:
•

Table 18: Navy Population of the Navy by Race, Fiscal Years 2013–
2017

•

Table 19: Summary Statistics by Race for Navy Military Justice
Actions, Fiscal Years 2013–2017

•

Table 20: Summary Statistics by Gender for Navy Military Justice
Actions, Fiscal Years 2013–2017

•

Table 21: Summary Statistics by Rank for Navy Military Justice
Actions, Fiscal Years 2013–2017

•

Table 22: Summary Statistics by Education for Navy Military Justice
Actions, Fiscal Years 2013–2017

•

Table 23: Summary Statistics by Age for Navy Military Justice Actions,
Fiscal Years 2013–2017

•

Table 24 Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Navy Military Justice Actions, Fiscal Years
2013–2017

•

Table 25: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Navy Military Justice Actions, Fiscal
Years 2013–2017

•

Table 26: Odds Ratios for Navy Multivariate Regression Analyses

Page 126

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 18: Total Population of the Navy by Race, Fiscal Years 2013–2017
Population
Number

Percent

American Indian or Alaska Native

Race

11,938

2

Asian

26,780

5

Black or African American

87,415

16

Hispanic

83,609

15

Mixed

38,282

7

4,699

1

12,571

2

Native Hawaiian or Pacific Islander
Unknown
White

287,094

52

Total population

552,388

100

Source: GAO analysis of Navy personnel data. | GAO-19-344

Note: The total populations presented in this table represent the number of unique active-duty
servicemembers who served during fiscal years 2013 through 2017.

Page 127

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 19: Summary Statistics by Race for Navy Military Justice Actions, Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

N

N

Percent

N

Percent

N

Percent

N

Percent

539,817

287,094

52%

87,415

16%

83,609

15%

81,699

15%

Recorded
investigations

7,193

2,954

41%

1,816

25%

1,316

18%

1,107

15%

General and special
courts-martial

1,018

417

40%

247

24%

177

17%

177

17%

General and special
courts-martial with a
recorded investigation

759

306

40%

178

23%

137

18%

138

18%

General and special
courts-martial without
a recorded
investigation

259

111

42%

69

26%

40

15%

39

15%

Acquittals (general
and special courtsmartial)

151

61

40%

36

24%

27

18%

27

18%

Convictions (general
and special courtsmartial)

867

356

40%

211

24%

150

17%

150

17%

Total population

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with an unknown race, were not included in the
populations presented in this table to protect privacy. As a result, the populations presented in this
table may vary from the populations presented in the body of the report or from other tables in this
appendix. The Other race group includes individuals who identified as American Indian/Alaska Native,
Asian, Native Hawaiian/Other Pacific Islander, and multiple races. The total population presented in
this table represents the number of unique active-duty servicemembers during fiscal years 2013
through 2017. The term “recorded investigation” refers to where a servicemember was the subject of
a criminal investigation that was recorded in the Navy’s military criminal investigative organization’s
database. Percentages in this table may not add up to 100 due to rounding, the exclusion of
information identified as unknown or missing, and/or to protect privacy.

Page 128

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 20: Summary Statistics by Gender for Navy Military Justice Actions, Fiscal Years 2013–2017
Population

Male

Female

N

N

%

N

%

552,388

442,184

80%

110,204

20%

Recorded investigations

7,293

6,681

92%

612

8%

General and special courts-martial

1,034

975

94%

59

6%

General and special courts-martial
with a recorded investigation

770

739

96%

31

4%

General and special courts-martial
without a recorded investigation

264

236

89%

28

11%

Convictions (general and special
courts-martial)

881

828

94%

53

6%

Total population

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Navy’s military criminal investigative organization’s database.

Table 21: Summary Statistics by Rank for Navy Military Justice Actions, Fiscal Years 2013–2017
Population
Total population

Rank E1-E4

Rank E5-E9 and officers

N

N

%

N

%

552,388

273,247

49%

279,141

51%

Recorded investigations

7,293

4,170

57%

3,123

43%

General and special courts-martial

1,034

523

51%

511

49%

General and special courts-martial
with a recorded investigation

770

390

51%

380

49%

General and special courts-martial
without a recorded investigation

264

133

50%

131

50%

Acquittals (general and special
courts-martial)

153

68

44%

85

56%

Convictions (general and special
courts-martial)

881

455

52%

426

48%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Navy’s military criminal investigative organization’s database.
Although we analyzed officers separately from enlisted servicemembers, for reporting purposes we
combined servicemembers in rank categories E5-E9 with officers to protect privacy in instances when
the number of servicemembers was fewer than 20.

Page 129

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 22: Summary Statistics by Education for Navy Military Justice Actions, Fiscal Years 2013–2017
Population

High school or less

More than high school

Unknown education

N

N

%

N

%

N

%

552,388

403,581

73%

122,824

22%

25,983

5%

Recorded investigations

7,293

6,221

85%

931

13%

141

2%

General and special courtsmartial

1,034

858

83%

152

15%

24

2%

General and special courtsmartial with a recorded
investigation

754

639

83%

115

15%

<20

General and special courtsmartial without a recorded
investigation

256

219

83%

37

14%

<20

Acquittals (general and
special courts-martial)

149

112

73%

37

24%

<20

Convictions (general and
special courts-martial)

881

746

85%

115

13%

20

Total population

2%

Legend: <20 refers to education groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown education level, were not included in the
populations presented in this table to protect privacy. As a result, the populations presented in this
table may vary from the populations presented in the body of the report or from other tables in this
appendix. The total population presented in this table represents the number of unique active-duty
servicemembers during fiscal year 2013 through 2017. The term “recorded investigation” refers to
where a servicemember was the subject of a criminal investigation that was recorded in the Navy’s
military criminal investigative organization’s database. Percentages in this table may not add up to
100 due to rounding, the exclusion of information identified as unknown or missing, and/or to ensure
protection of privacy.

Table 23: Summary Statistics by Age for Navy Military Justice Actions, Fiscal Years 2013–2017
Population
Total population

< 21 Years

21-25 years

26-30 years

> 30 years

N

N

%

N

%

N

%

N

%

552,388

82,562

15%

180,939

33%

115,945

21%

172,869

31%

Recorded investigations

7,293

979

13%

3,041

42%

1,535

21%

1,671

23%

General and special
courts-martial

1,034

93

9%

391

38%

220

21%

330

32%

770

63

8%

292

38%

170

22%

245

32%

General and special
courts-martial with a
recorded investigation

Page 130

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

< 21 Years

21-25 years

26-30 years

General and special
courts-martial without a
recorded investigation

Population
264

30

11%

99

38%

50

19%

85

> 30 years
32%

Acquittals (general and
special courts-martial)

148

<20

—

56

37%

44

29%

48

31%

Convictions (general and
special courts-martial)

881

88

10%

335

38%

176

20%

282

32%

Legend: <20 refers to age groups that have zero or fewer than 20 servicemembers
“N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicememberswere not included in this table to protect privacy. As a result, the populations
presented in this table may vary from the populations presented in the body of the report or from
other tables in this appendix. The total population presented in this table represents the number of
unique active-duty servicemembers during fiscal years 2013 through 2017. The term “recorded
investigation” refers to where a servicemember was the subject of a criminal investigation that was
recorded in the Navy’s military criminal investigative organization’s database. Although we analyzed
age categories of 30-40 years separately from greater than 40 years, for reporting purposes, we
combined servicemembers in these two age categories to protect privacy during instances when the
number of servicemembers in an age category was fewer than 20. Percentages in this table may not
add up to 100 due to rounding or the exclusion of information to ensure protection of privacy.

Page 131

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 24: Offenses Investigated and Tried in General and Special Courts-Martial by Race for Navy Military Justice Actions,
Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

N

N

%

N

%

N

%

N

%

Recorded investigations for
drug offenses (Article 112a)

1,394

648

46%

303

21%

213

15%

230

16%

Recorded investigations for
sexual assault offenses
(Article 120)

3,196

1,259

39%

816

25%

638

20%

483

15%

Recorded investigations for
all other offenses

2,603

1,047

40%

697

26%

465

18%

394

15%

General and special courtsmartial for drug offenses
(Article 112a)

186

79

42%

48

25%

21

11%

38

20%

General and special courtsmartial for sexual assault
offenses (Article 120)

451

166

36%

115

25%

89

20%

81

18%

General and special courtsmartial for all other offenses

364

165

44%

83

22%

64

17%

52

14%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown race, were not included in the populations
presented in this table to protect privacy. As a result, the populations presented in this table may vary
from the populations presented in the body of the report or from other tables in this appendix. The
Other race group includes individuals who identified as American Indian/Alaska Native, Asian, Native
Hawaiian/Other Pacific Islander, and multiple races. The term “recorded investigation” refers to where
a servicemember was the subject of a criminal investigation that was recorded in the Navy’s criminal
investigative organization’s database. This table summarizes the number of instances in which an
offense was investigated or tried in general and special courts-martial. Multiple offenses may be
incorporated into the investigation or court-martial of a single servicemember. As such, a single case
could be included in multiple offense groups in this table. Percentages in this table may not add up to
100 due to rounding, the exclusion of information identified as unknown or missing, and/or to protect
privacy.

Page 132

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 25: Offenses Investigated and Tried in General and Special Courts-Martial by Gender for Navy Military Justice Actions,
Fiscal Years 2013–2017
Population

Male

Female

N

N

%

N

%

Recorded investigations for drug
offenses (Article 112a)

1,410

1,257

89%

153

11%

Recorded investigations for
sexual assault offenses (Article
120)

3,246

3,140

97%

106

3%

Recorded investigations for all
other offenses

2,637

2,284

87%

353

13%

General and special courts-martial
for drug offenses (Article 112a)

190

166

87%

24

13%

General and special courts-martial
for all other offenses

371

340

92%

31

8%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The term “recorded investigation” refers to
where a servicemember was the subject of a criminal investigation that was recorded in the Navy’s
criminal investigative organization’s database. This table summarizes the number of instances in
which an offense was investigated or tried in general and special courts-martial. Multiple offenses
may be incorporated into the investigation or court-martial of a single servicemember. As such, a
single case could be included in multiple offense groups in this table. The summary statistics for
servicemembers tried in general and special courts-martial for sexual assault offenses (Article 120)
were omitted from this table to protect privacy because a gender group had fewer than 20
servicemembers.

Multivariate Regression
Analyses of Navy Data

The multivariate results listed below in table 26 show the odds ratios for
the multivariate regression analyses of Navy data. We used logistic
regression to assess the relationship between the independent variables,
such as race, education, rank, or gender, with the probability of being
subject to a military justice action. Logistic regression allows for the
coefficients to be converted into odds ratios. Odds ratios that are
statistically significant and greater than 1.00 indicate that individuals with
that characteristic are more likely to be subject to a military justice action.
For example, an odds ratio of 1.55 for Black servicemembers would mean
that they are 1.55 times more likely to be subject to a military justice
action compared to White servicemembers. Odds ratios that are
statistically significant and lower than 1.00 indicate that individuals with
that characteristic are less likely to be subject to a military justice action.
We excluded age and years of service from the Navy multivariate
regression analyses due to high correlation with the rank variable.

Page 133

GAO-19-344 Military Justice

Appendix V: Navy Data and Analyses

Table 26: Odds Ratios for Navy Multivariate Regression Analyses
Black

Hispanic

Other

Unknown
race

Male

High
school
or less

Unknown
education

Likelihood of
being subject of
recorded
investigations

2.06**

1.47**

1.27**

0.75**

3.03**

1.44**

1.07

1.13**

0.51**

Likelihood of trial
in general and
special courtsmartial

2.01**

1.42**

1.45**

0.91

4.42**

1.23*

1.18

0.88

0.41**

Likelihood of trial
in general and
special courtsmartial with a
recorded
investigation

1.45

1.13

1.71

1.13

1.16

0.93

1.63

0.97

2.94

Likelihood of trial
in general and
special courtsmartial without a
recorded
investigation

2.07**

1.19

1.19

1.05

2.26**

1.35

1.6

0.86

0.42*

1.04

1.08

0.95

1.08

1.06

1.83*

1.58

1.02

0.82

0.55**

0.88

1.24

0.9

1.16

1.48

0.7

1.21

1.34

Likelihood of
conviction in
general and
special courtsmartial
Likelihood of
dismissal or
discharge when
convicted in
general and
special courtsmartial

Rank E1-E4 Officers

Legend: ** next to the odds ratio indicate that the finding has a strong degree of statistical significance with a p-value less than 0.01.
* next to the odds ratio indicates that the finding is statistically significant with a p-value between 0.01 and 0.05.
If the space next to the odds ratio is blank, then the finding was not statistically significant.
Source: GAO analysis of Navy investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These multivariate regression analysis
results demonstrate the degree to which a racial, gender, education, or rank group is more likely than
the reference category to be subject of a recorded investigation, tried in general and special courtsmartial, convicted in general and special courts-martial, and receive a more severe punishment
following a conviction. All racial categories listed are in reference to White servicemembers, all
gender groups listed are in reference to female servicemembers, all education groups listed are in
reference to servicemembers with more than a high school education, and all rank groups are in
reference to servicemembers between ranks E5 and E9. The Other race group includes individuals
who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races.

Page 134

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and
Analyses
Appendix VI: Marine Corps Data and Analyses

This appendix contains several tables that show the underlying data and
analyses used throughout this report relating to Marine Corps personnel
and military justice disciplinary actions from fiscal years 2013 through
2017. We did not include populations that contained fewer than 20
servicemembers in the populations presented in these tables to ensure
the protection of sensitive information. As a result, the populations
presented in this appendix may vary among the different tables and may
vary from the populations presented in other places in this report. Our
analyses of the Marine Corps investigations, military justice, and
personnel databases, as reflected in these tables, taken alone, do not
establish the presence or absence of unlawful discrimination.
The following tables and information are included in this appendix:
•

Table 27: Total Population of the Marine Corps by Race, Fiscal Years
2013–2017

•

Table 28: Summary Statistics by Race for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017

•

Table 29: Summary Statistics by Gender for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017

•

Table 30: Summary Statistics by Rank for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017

•

Table 31: Summary Statistics by Education for Marine Corps Military
Justice Actions, Fiscal Years 2013–2017

•

Table 32: Summary Statistics by Age for Marine Corps Military Justice
Actions, Fiscal Years 2013–2017

•

Table 33: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Marine Corps Military Justice Actions,
Fiscal Years 2013–2017

•

Table 34: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Marine Corps Military Justice Actions,
Fiscal Years 2013–2017

•

Table 35 Odds Ratios for Marine Corps Multivariate Regression
Analyses

Page 135

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 27: Total Population of the Marine Corps by Race, Fiscal Years 2013–2017
Population
Race
American Indian or Alaska Native
Asian

Number

Percent

3,451

1%

9,580

3%

Black or African American

36,529

10%

Hispanic

63,044

18%

Native Hawaiian or Pacific Islander

3,426

1%

Unknown

4,680

1%

White

232,083

66%

Total population

352,793

100%

Source: GAO analysis of Marine Corps personnel data. | GAO-19-344

Note: The total populations presented in this table represent the number of unique active-duty
servicemembers who served during fiscal years 2013 through 2017.

Page 136

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 28: Summary Statistics by Race for Marine Corps Military Justice Actions, Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

Unknown

N

N

%

N

%

N

%

N

%

N

%

352,793

232,083

66%

36,529

10%

63,044

18%

16,457

5%

4,680

1%

Recorded investigations

6,833

3,985

58%

1,286

19%

1,193

17%

258

4%

111

2%

General and special courtsmartial

1,354

749

55%

247

18%

268

20%

63

5%

27

2%

General and special courtsmartial with a recorded
investigation

848

479

56%

152

18%

175

20%

42

5%

< 20

General and special courtsmartial without a recorded
investigation

479

270

55%

95

19%

93

19%

21

4%

< 20

Acquittals (general and special
courts-martial)

190

115

57%

37

18%

38

19%

< 20

Convictions (general and
special courts-martial)

1,152

634

55%

210

18%

230

20%

55

5%

23

Summary courts-martial

1,389

886

63%

221

16%

230

16%

52

4%

< 20

Nonjudicial punishments

49,184

30,853

63%

6,815

14%

8,656

18%

2,081

4%

779

Total population

< 20
2%

2%

Legend: <20 refers to racial groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown race, were not included in the populations
presented in this table to protect privacy. As a result, the populations presented in this table may vary
from the populations presented in the body of the report or from other tables in this appendix. The
Other race group includes individuals who identified as American Indian/Alaska Native, Asian, Native
Hawaiian/Other Pacific Islander, and multiple races. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to an investigation where a servicemember was the subject
of a criminal investigation that was recorded in the Navy’s military criminal investigative organization’s
database. Percentages in this table may not add up to 100 due to rounding, the exclusion of
information identified as unknown or missing, and/or to protect privacy.

Page 137

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 29: Summary Statistics by Gender for Marine Corps Military Justice Actions, Fiscal Years 2013–2017
Population

Male

Female

N

N

%

N

%

352,793

323,491

92%

29,302

8%

Recorded investigations

6,833

6,539

96%

294

4%

General and special courts-martial

1,354

1,306

96%

48

4%

General and special courts-martial with a
recorded investigation

862

842

98%

20

2%

General and special courts-martial
without a recorded investigation

492

464

94%

28

6%

Convictions (general and special courtsmartial)

1,152

1,117

97%

35

3%

Summary courts-martial

1,406

1,359

97%

47

3%

Nonjudicial punishments

49,184

45,828

93%

3,356

7%

Total population

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to an investigation where a servicemember was the subject
of a criminal investigation that was recorded in the Navy’s military criminal investigative organization’s
database.

Page 138

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 30: Summary Statistics by Rank for Marine Corps Military Justice Actions, Fiscal Years 2013–2017
Population

Rank E1-E4

Rank E5-E9 and officers

N

N

%

N

%

352,800

247,195

70%

105,605

30%

Recorded investigations

6,833

5,127

75%

1,706

25%

General and special courts-martial

1,354

859

63%

495

37%

General and special courts-martial with a
recorded investigation

862

568

66%

294

34%

General and special courts-martial
without a recorded investigation

492

291

59%

201

41%

Acquittals (general and special courtsmartial)

202

110

54%

92

46%

1,152

749

65%

403

35%

Total population

Convictions (general and special courtsmartial)
Summary courts-martial

1,408

1,098

78%

310

22%

Nonjudicial punishments

49,184

34,362

70%

14,822

30%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to an investigation where a servicemember was the subject
of a criminal investigation that was recorded in the Navy’s military criminal investigative organization’s
database. Although we analyzed officers separately from enlisted servicemembers, for reporting
purposes we combined servicemembers in rank categories E5-E9 with officers to protect privacy in
those instances when the number of officers was fewer than 20.

Page 139

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 31: Summary Statistics by Education for Marine Corps Military Justice Actions, Fiscal Years 2013–2017
Population

High school or less

More than high school

Unknown education

N

N

%

N

%

N

%

352,793

307,568

87%

41,678

12%

3,547

1%

Recorded investigations

6,833

6,452

94%

312

5%

69

1%

General and special courtsmartial

1,342

1,265

93%

77

6%

< 20

General and special courtsmartial with a recorded
investigation

853

806

94%

47

5%

< 20

General and special courtsmartial without a recorded
investigation

489

459

93%

30

6%

< 20

Acquittals (general and special
courts-martial)

187

187

93%

< 20

Convictions (general and
special courts-martial)

1,142

1,078

94%

64

6%

< 20

Summary courts-martial

1,394

1,357

97%

37

3%

< 20

Nonjudicial punishments

49,184

46,182

94%

2,509

5%

493

Total population

< 20

1%

Legend: <20 refers to education groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown education level, were not included in the
populations presented in this table to protect privacy. As a result, the populations presented in this
table may vary from the populations presented in the body of the report or in other tables in this
appendix. The total population presented in this table represents the number of unique active-duty
servicemembers during fiscal years 2013 through 2017. The term “recorded investigation” refers to an
investigation where a servicemember was the subject of a criminal investigation that was recorded in
the Navy’s military criminal investigative organization’s database. Percentages in this table may not
add up to 100 due to rounding, the exclusion of information identified as unknown or missing, and/or
to protect privacy.

Page 140

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 32: Summary Statistics by Age for Marine Corps Military Justice Actions, Fiscal Years 2013–2017
Population

< 21 years

21-25 years

26-30 years

> 30 years

N

N

%

N

%

N

%

N

%

352,793

114,326

32%

141,427

40%

48,688

14%

48,352

14%

Recorded investigations

6,833

1,986

29%

3,229

47%

954

14%

664

10%

General and special courtsmartial

1,354

323

24%

568

42%

248

18%

215

16%

General and special courtsmartial with a recorded
investigation

862

210

24%

389

45%

139

16%

124

14%

General and special courtsmartial without a recorded
investigation

492

113

23%

179

36%

109

22%

91

18%

Acquittals (general and special
courts-martial)

202

22

11%

91

45%

46

23%

43

21%

Convictions (general and special
courts-martial)

1,152

301

26%

477

41%

202

18%

172

15%

Summary courts-martial

1,406

496

35%

613

44%

179

13%

118

8%

Nonjudicial punishments

49,184

11,414

23%

22,288

45%

8,166

17%

7,316

15%

Total population

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to an investigation where a servicemember was the subject
of a criminal investigation that was recorded in the Navy’s military criminal investigative organization’s
database.

Page 141

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 33: Offenses Investigated and Tried in General and Special Courts-Martial by Race for Marine Corps Military Justice
Actions, Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

Unknown race

N

N

%

N

%

N

%

N

%

N

%

Recorded
investigations for drug
offenses (Article 112a)

2,290

1,471

64%

403

18%

327

14%

64

3%

25

1%

Recorded
investigations for
sexual assault offenses
(Article 120)

1,661

881

53%

295

18%

390

23%

65

4%

30

2%

Recorded
investigations for all
other offenses

2,882

1,633

57%

588

20%

476

17%

129

4%

56

2%

General and special
courts-martial for drug
offenses (Article 112a)

193

132

64%

36

18%

25

12%

< 20

< 20

General and special
courts-martial for
sexual assault offenses
(Article 120)

319

185

55%

53

16%

81

24%

< 20

< 20

General and special
courts-martial for all
other offenses

560

307

53%

103

18%

117

20%

33

6%

< 20

Legend: <20 refers to racial groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown race, were not included in the populations
presented in this table to protect privacy. As a result, the populations presented in this table may vary
from the populations presented in the body of the report or from other tables in this appendix. The
Other race group includes individuals who identified as American Indian/Alaska Native, Asian, Native
Hawaiian/Other Pacific Islander, and multiple races. This table summarizes the number of instances
in which an offense was investigated or tried in general and special courts-martial. Multiple offenses
may be incorporated into the investigation or court-martial of a single servicemember. As such, a
single case could be included in multiple offense groups in this table. The term “recorded
investigation” refers to an investigation where a servicemember was the subject of a criminal
investigation that was recorded in the Navy’s military criminal investigative organization’s database.
Percentages in this table may not add up to 100 due to rounding, the exclusion of information
identified as unknown or missing, and/or to protect privacy.

Page 142

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 34: Offenses Investigated and Tried in General and Special Courts-Martial by Gender for Marine Corps Military Justice
Actions, Fiscal Years 2013–2017
Population

Male

N

N

%

N

%

2,290

2,181

95%

109

5%

Recorded investigations for sexual assault 1,661
offenses (Article 120)

1,639

99%

22

1%

Recorded investigations for all other
offenses

2,882

2,719

94%

163

6%

General and special courts-martial for all
other offenses

575

549

95%

26

5%

Recorded investigations for drug offenses
(Article 112a)

Female

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. This table summarizes the number of
instances in which an offense was investigated or tried in general and special courts-martial. Multiple
offenses may be incorporated into the investigation or court-martial of a single servicemember. As
such, a single case could be included in multiple offense groups in this table. The term “recorded
investigation” refers to an investigation where a servicemember was the subject of a criminal
investigation that was recorded in the Navy’s military criminal investigative organization’s database.
Although we analyzed gender data for general and special courts-martial for drug offenses and
sexual assault offenses, we omitted those data from this table to protect privacy because a gender
group had fewer than 20 servicemembers.

Multivariate Regression
Analyses of Marine Corps
Data

The multivariate results listed below in table 35 show the odds ratios for
the multivariate regression analyses of Marine Corps data. We used
logistic regression to assess the relationship between the independent
variables, such as race, education, rank, or gender, with the probability of
being subject to a military justice action. Logistic regression allows for the
coefficients to be converted into odds ratios. Odds ratios that are
statistically significant and greater than 1.00 indicate that individuals with
that characteristic are more likely to be subject to a military justice action.
For example, an odds ratio of 1.55 for Black servicemembers would mean
that they are 1.55 times more likely to be subject to a military justice
action compared to White servicemembers. Odds ratios that are
statistically significant and lower than 1.00 indicate that individuals with
that characteristic are less likely to be subject to a military justice action.
We excluded age and years of service from the Marine Corps multivariate
regression analyses due to high correlation with the rank variable.

Page 143

GAO-19-344 Military Justice

Appendix VI: Marine Corps Data and Analyses

Table 35: Odds Ratios for Marine Corps Multivariate Regression Analyses
Black

Hispanic

Other

Unknown
race

Male

High
school or
less

Likelihood of being
subject of recorded
investigations

2.07**

1.07*

0.92

1.6**

2.15**

1.68**

1.89**

1.02

0.35**

Likelihood of trial in
general and special
courts-martial

1.99**

1.29**

1.2

1.7**

2.56**

1.4*

NA

0.57**

0.17**

0.84

2**

0.86

NA

0.81

0.91

NA

0.58**

NA

Likelihood of trial in
general and special
courts-martial without a
recorded investigation

2.09**

1.23

1.09

NA

1.56*

1.35

NA

0.47**

NA

Likelihood of conviction
in general and special
courts-martial

1.06

1.11

1.27

1.18

2.47**

0.98

NA

1.52**

NA

Likelihood of dismissal
or discharge when
convicted in general
and special courtsmartial

0.96

1.30

1.81

2.48

2.65**

2.00*

NA

3.03**

NA

Likelihood of trial in
summary courtsmartial

1.57**

0.92

0.84

1.17

2.69**

1.9**

NA

1.09

NA

Likelihood of receiving
nonjudicial
punishments

1.41**

0.99

0.94**

1.35**

1.29**

1.52**

1.71**

0.7**

0.16**

Likelihood of trial in
general and special
courts-martial with a
recorded investigation

Unknown Rank E1education
E4

Officers

Legend: ** next to the odds ratio indicates that the finding has a strong degree of statistical significance with a p-value less than 0.01.
* next to the odds ratio indicates that the finding is statistically significant with a p-value between 0.01 and 0.05.
If the space next to the odds ratio is blank, then the finding was not statistically significant.
NA indicates that the odds ratio for that group is not available because the number of servicemembers was too small to produce reliable findings.
Source: GAO analysis of Marine Corps investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These multivariate regression analysis
results demonstrate the degree to which a racial, gender, education, or rank group is more likely than
the reference category to be the subject of a recorded investigation, tried in general and special
courts-martial, convicted in general and special courts-martial, receive a more severe punishment
following a conviction, tried in summary courts-martial, or subject to nonjudicial punishments. All
racial categories listed are in reference to White servicemembers, all gender groups listed are in
reference to female servicemembers, all education groups listed are in reference to servicemembers
with more than a high school education, and all rank groups are in reference to servicemembers
between ranks E5 and E9. The Other race group includes individuals who identified as American
Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and multiple races.

Page 144

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses
Appendix VII: Air Force Data and Analyses

This appendix contains several tables that show the underlying data and
analyses used throughout this report relating to Air Force personnel and
military justice disciplinary actions from fiscal years 2013 through 2017.
We did not include populations that contained fewer than 20
servicemembers in the populations presented in these tables to ensure
the protection of sensitive information. As a result, the populations
presented in this appendix may vary among the different tables and may
vary from the populations presented in other places in this report. Our
analyses of the Air Force’s investigations, military justice, and personnel
databases, as reflected in these tables, taken alone, do not establish the
presence or absence of unlawful discrimination.
The following tables and information are included in this appendix:
•

Table 36: Total Population of the Air Force by Race, Fiscal Years
2013–2017

•

Table 37: Summary Statistics by Race for Air Force Military Justice
Actions, Fiscal Years 2013–2017

•

Table 38: Summary Statistics by Gender for Air Force Military Justice
Actions, Fiscal Years 2013–2017

•

Table 39: Summary Statistics by Rank for Air Force Military Justice
Actions, Fiscal Years 2013–2017

•

Table 40: Summary Statistics by Education for Air Force Military
Justice Actions, Fiscal Years 2013–2017

•

Table 41: Summary Statistics by Age for Air Force Military Justice
Actions, Fiscal Years 2013–2017

•

Table 42: Summary Statistics by Rank and Years of Service Hybrid
Variable for Air Force Military Justice Actions, Fiscal Years 2013–
2017

•

Table 43: Offenses Investigated and Tried in General and Special
Courts-Martial by Race for Air Force Military Justice Actions, Fiscal
Years 2013–2017

•

Table 44: Offenses Investigated and Tried in General and Special
Courts-Martial by Gender for Air Force Military Justice Actions, Fiscal
Years 2013–2017

•

Table 45: Odds Ratios for Air Force Multivariate Regression Analyses

Page 145

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 36: Total Population of the Air Force by Race, Fiscal Years 2013–2017
Total population
Race
American Indian/Alaska Native

Number

Percent

2,802

1%

Asian

16,273

3%

Black

73,836

15%

5,000

1%

Hispanic

16,844

3%

Multiple

13,165

3%

Unknown

12,891

3%

Hawaiian/Pacific Islander

White

343,655

71%

Total population

484,466

100%

Source: GAO analysis of Air Force personnel data. | GAO-19-344

Note: The total populations presented in this table represent the number of unique active-duty
servicemembers who served during fiscal years 2013 through 2017.

Page 146

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 37: Summary Statistics by Race for Air Force Military Justice Actions, Fiscal Years 2013–2017
Population

White

Black

Hispanic

Other

Unknown

N

N

%

N

%

N

%

N

%

N

%

484,466

343,655

71%

73,836

15%

16,844

3%

37,243

8%

12,888

3%

Recorded
investigations

6,630

4,232

64%

1,470

22%

268

4%

476

7%

184

3%

General and special
courts-martial

2,359

1,518

64%

534

23%

86

4%

165

7%

56

2%

General and special
courts-martial with a
recorded investigation

1,259

809

64%

274

22%

54

4%

85

7%

37

3%

General and special
courts-martial without
a recorded
investigation

1,081

709

64%

260

24%

32

3%

80

7%

<20

—

404

265

61%

111

26%

<20

—

28

6%

<20

—

1,887

1,235

65%

416

22%

66

3%

130

7%

40

2%

473

297

61%

133

27%

<20

—

43

9%

<20

—

20,899

13,117

63%

5,274

25%

596

3%

1,470

7%

442

2%

Total population

Acquittals (general
and special courtsmartial)
Convictions (general
and special courtsmartial)
Summary courtsmartial
Nonjudicial
punishments

Legend: <20 refers to racial groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown race, were not included in the populations
presented in this table to protect privacy. As a result, the populations presented in this table may vary
from the populations presented in the body of the report or from other tables in this appendix. The
Other race group includes individuals who identified as American Indian/Alaska Native, Asian, Native
Hawaiian/Other Pacific Islander, and multiple races. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Air Force’s military criminal investigative organization’s
database. Percentages in this table may not add up to 100 due to rounding, the exclusion of
information identified as unknown or missing, and/or to protect privacy.

Page 147

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 38: Summary Statistics by Gender for Air Force Military Justice Actions, Fiscal Years 2013–2017
Population

Male

Female

N

N

%

N

%

484,466

387,970

80%

96,496

20%

Recorded investigations

6,630

6,006

91%

624

9%

General and special courts-martial

2,359

2,188

93%

171

7%

General and special courts-martial
with a recorded investigation

1,259

1,179

94%

80

6%

General and special courts-martial
without a recorded investigation

1,100

1,009

92%

91

8%

Convictions (general and special
courts-martial)

1,887

1,738

92%

149

8%

Summary courts-martial

490

441

90%

49

10%

Nonjudicial punishments

20,899

17,991

86%

2,908

14%

Total population

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Air Force’s military criminal investigative organization’s
database.

Page 148

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 39: Summary Statistics by Rank for Air Force Military Justice Actions, Fiscal Years 2013–2017
Population

Rank E1-E4

Rank E5-E6

Rank E7-E9 and officers

N

N

%

N

%

N

%

484,466

224,739

46%

127,516

26%

132,211

23%

Recorded investigations

6,630

4,248

64%

1,556

23%

826

12%

General and special courtsmartial

2,359

1,764

75%

396

17%

199

8%

General and special courtsmartial with a recorded
investigation

1,259

1,034

82%

135

11%

90

7%

General and special courtsmartial without a recorded
investigation

1,100

730

66%

261

24%

109

9%

435

221

51%

167

38%

47

10%

Convictions (general and
special courts-martial)

1,887

1,531

81%

216

11%

140

8%

Nonjudicial punishments

20,899

16,151

77%

3,707

18%

1,041

5%

Total population

Acquittals (general and
special courts-martial)

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Air Force’s military criminal investigative organization’s
database. Although we analyzed officers separately from enlisted servicemembers, for reporting
purposes we combined servicemembers in rank categories E7-E9 with officers in this table to protect
privacy in those instances when the number of servicemembers was fewer than 20. The summary
statistics for servicemembers tried in summary courts-martial were omitted from this table to protect
privacy because a rank group had fewer than 20 servicemembers. Percentages in this table may not
add up to 100 due to rounding.

Page 149

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 40: Summary Statistics by Education for Air Force Military Justice Actions, Fiscal Years 2013–2017
Population

Total population

High school and
some college

Associates
degree

Bachelor’s
degree

Post-bachelor’s
degree

Unknown
education

N

N

%

N

%

N

%

N

%

N

%

484,466

206,029

43%

156,762

32%

54,551

11%

59,161

12%

7,963

2%

Recorded
investigations

6,630

3,511

53%

2,268

34%

448

7%

377

6%

26

1%

General and special
courts-martial

2,345

1,392

59%

724

31%

138

6%

91

4%

<20

General and special
courts-martial with a
recorded investigation

1,256

692

55%

437

35%

78

6%

49

4%

<20

General and special
courts-martial without
a recorded
investigation

1,089

700

64%

287

26%

60

5%

42

4%

<20

418

210

48%

168

39%

40

9%

<20

1,876

1,167

62%

547

29%

94

5%

68

481

417

85%

64

13%

<20

20,899

15,205

73%

4,475

21%

654

Acquittals (general
and special courtsmartial)
Convictions (general
and special courtsmartial)
Summary courtsmartial
Nonjudicial
punishments

<20
4%

<20
3%

377

<20

<20
2%

188

1%

Legend: <20 refers to education groups that have zero or fewer than 20 servicemembers, to protect privacy. “N” refers to the population size for each
group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown education level, were not included in the
populations presented in this table to protect privacy. As a result, the populations presented in this
table may vary from the populations presented in the body of the report or from other tables in this
appendix. The total population presented in this table represents the number of unique active-duty
servicemembers during fiscal years 2013 through 2017. The term “recorded investigation” refers to
where a servicemember was the subject of a criminal investigation that was recorded in the Air
Force’s military criminal investigative organization’s database. Percentages in this table may not add
up to 100 due to rounding, the exclusion of information identified as unknown or missing, and/or to
protect privacy.

Page 150

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 41: Summary Statistics by Age for Air Force Military Justice Actions, Fiscal Years 2013–2017
Population

< 21 years

21-25 years

26-30 years

31-35 years

> 35 years

N

N

%

N

%

N

%

N

%

N

%

484,466

70,019

14%

119,557

25%

114,941

24%

73,072

15%

106,877

22%

Recorded
investigations

6,630

512

8%

2,214

33%

1,906

29%

989

15%

1,009

15%

General and special
courts-martial

2,359

282

12%

759

32%

676

29%

335

14%

307

13%

General and special
courts-martial with a
recorded investigation

1,259

49

4%

414

33%

420

33%

198

16%

178

14%

General and special
courts-martial without
a recorded
investigation

1,100

233

21%

345

31%

256

23%

137

12%

129

12%

435

32

7%

126

29%

134

31%

79

18%

64

15%

1,887

251

13%

627

33%

523

28%

253

13%

233

12%

471

160

33%

242

49%

69

14%

<20

20,899

6,078

29%

7,614

36%

4,090

20%

1,811

Total population

Acquittals (general
and special courtsmartial)
Convictions (general
and special courtsmartial)
Summary courtsmartial
Nonjudicial
punishments

<20
9%

1,306

6%

Legend: <20 refers to age groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers in certain age groups, were not included in the
populations presented in this table to protect privacy. As a result, the populations presented in this
table may vary from the populations presented in the body of the report or from other tables in this
appendix. The total population presented in this table represents the number of unique active-duty
servicemembers during fiscal years 2013 through 2017. The term “recorded investigation” refers to
where a servicemember was the subject of a criminal investigation that was recorded in the Air
Force’s military criminal investigative organization’s database. Percentages in this table may not add
up to 100 due to rounding, the exclusion of information identified as unknown or missing, and/or to
protect privacy.

Page 151

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 42: Summary Statistics by Rank and Years of Service Hybrid Variable for Air Force Military Justice Actions, Fiscal Years
2013–2017
Population Rank E1-E4 and 0-4 years Rank E1-E4 and 4-6 years
of service
of service

Rank E1-E4 and > 6 years
of service

N

N

%

N

%

N

%

224,726

180,899

37%

35,495

7%

8,332

2%

Recorded investigations

4,248

2,189

33%

1,222

18%

837

13%

General and special courtsmartial

1,764

925

39%

362

15%

477

20%

General and special courtsmartial with a recorded
investigation

1,034

375

30%

272

22%

387

31%

General and special courtsmartial without a recorded
investigation

730

550

50%

90

8%

90

8%

Acquittals (general and
special courts-martial)

221

123

28%

67

15%

31

7%

Convictions (general and
special courts-martial)

1,531

794

42%

296

16%

441

23%

Nonjudicial punishments

16,151

12,964

62%

2,205

11%

982

5%

Total population

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Rank and years of service were highly
correlated variables, which usually results in selecting just one of the variables to analyze in our
multivariate analyses. Based on discussion with Air Force officials, we developed this hybrid rank and
years of service variable that controlled for years of service among the lower enlisted ranks (E-1
through E-4). The total population presented in this table represents the number of servicemembers
in ranks E1 through E4 that also had the reported number of years of service among unique activeduty servicemembers during fiscal years 2013 through 2017. The term “recorded investigation” refers
to where a servicemember was the subject of a criminal investigation that was recorded in the Air
Force’s military criminal investigative organization’s database. The summary statistics for
servicemembers tried in summary courts-martial were omitted from this table to protect privacy
because a rank group had fewer than 20 servicemembers. Percentages in this table do not add up to
100 because the table only shows populations in lower enlisted ranks, but the percentages were
computed based on the rank group’s proportion of the total Air Force population.

Page 152

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 43: Offenses Investigated and Tried in General and Special Courts-Martial by Race for Air Force Military Justice
Actions, Fiscal Years 2013–2017
Population
N

White

Black

Hispanic

Other

Unknown race

N

%

N

%

N

%

N

%

N

%

Recorded
investigations for
drug offenses
(Article 112a)

2,226 1,360

61%

558

25%

126

6%

79

4%

103

5%

Recorded
investigations for
sexual assault
offenses (Article
120)

2,522 1,623

64%

504

20%

218

9%

109

4%

68

3%

Recorded
investigations for all
other offenses

1,652 1,074

65%

336

20%

105

6%

68

4%

69

4%

General and special
courts-martial for
drug offenses
(Article 112a)

925

599

65%

164

18%

85

9%

77

8%

<20

—

General and special
courts-martial for
sexual assault
offenses (Article
120)

718

407

57%

155

22%

102

14%

54

8%

<20

—

General and special
courts-martial for all
other offenses

670

380

57%

167

25%

71

11%

52

8%

<20

—

Legend: <20 refers to race groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown race, were not included in the populations
presented in this table to protect privacy. As a result, the populations presented in this table may vary
from the populations presented in the body of the report or from other tables in this appendix. The
Other race group includes individuals who identified as American Indian/Alaska Native, Asian, Native
Hawaiian/Other Pacific Islander, and multiple races. This table summarizes the number of instances
in which an offense was investigated or tried in general and special courts-martial. Multiple offenses
may be incorporated into the investigation or court-martial of a single servicemember. As such, a
single case could be included in multiple offense groups in this table. The term “recorded
investigation” refers to where a servicemember was the subject of a criminal investigation that was
recorded in the Air Force’s military criminal investigative organization’s database. Percentages in this
table may not add up to 100 due to rounding, the exclusion of information identified as unknown or
missing, and/or to protect privacy.

Page 153

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 44: Offenses Investigated and Tried in General and Special Courts-Martial by Gender for Air Force Military Justice
Actions, Fiscal Years 2013–2017
Population

Male

Female

N

N

%

N

%

Recorded investigations for drug
offenses (Article 112a)

2,226

1,903

85%

323

15%

Recorded investigations for sexual
assault offenses (Article 120)

2,519

2,414

96%

105

4%

Recorded investigations for all other
offenses

1,651

1,472

89%

179

11%

General and special courts-martial for
drug offenses (Article 112a)

938

837

89%

101

11%

General and special courts-martial for
all other offenses

682

617

90%

65

10%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. This table summarizes the number of
instances in which an offense was investigated or tried in general and special courts-martial. Multiple
offenses may be incorporated into the investigation or court-martial of a single servicemember. As
such, a single case could be included in multiple offense groups in this table. The term “recorded
investigation” refers to where a servicemember was the subject of a criminal investigation that was
recorded in the Air Force’s criminal investigative organization’s database. The summary statistics for
servicemembers tried in general and special courts-martial for sexual assault offenses (Article 120)
were omitted from this table to protect privacy because a gender group had fewer than 20
servicemembers. There were 3 cases with missing gender among recorded investigations for sexual
assault offenses (Article 120) and there was 1 case with missing gender among recorded
investigations for all other offenses.

Page 154

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Multivariate Regression
Analyses of Air Force Data

The multivariate results listed below in table 45 show the odds ratios for
the multivariate regression analyses of Air Force data. We used logistic
regression to assess the relationship between the independent variables,
such as race, education, rank, or gender, with the probability of being
subject to a military justice action. Logistic regression allows for the
coefficients to be converted into odds ratios. Odds ratios that are
statistically significant and greater than 1.00 indicate that individuals with
that characteristic are more likely to be subject to a military justice action.
For example, an odds ratio of 1.55 for Black servicemembers would mean
that they are 1.55 times more likely to be subject to a military justice
action compared to White servicemembers. Odds ratios that are
statistically significant and lower than 1.00 indicate that individuals with
that characteristic are less likely to be subject to a military justice action.
We controlled for years of service among the lower enlisted ranks (E1E4), but excluded age from the Air Force multivariate regression analyses
due to high correlation with the rank and years of service variables.

Page 155

GAO-19-344 Military Justice

Appendix VII: Air Force Data and Analyses

Table 45: Odds Ratios for Air Force Multivariate Regression Analyses
Black Hispanic Other

Likelihood of being subject of recorded
investigations

Male

High Associates
Rank
Rank
Rank
school
degree E1-E4
E1-E4
E1-E4
and 0-4 and 4-6 and > 6
or
years
years
years
some
of
of
of
college
service service service

1.58**

1.36**

1.05 2.41**

1.55**

1.54**

1.14**

3.22**

9.74**

Likelihood of trial in general and special courts- 1.51**
martial

1.34**

1.01 3.14**

1.41**

1.40**

1.98**

3.88**

21.60**

0.89

1.06

0.90 1.64**

0.54*

0.65*

2.57**

3.54**

9.58**

Likelihood of trial in general and special courts- 1.64**
martial without a recorded investigation

1.17

1.06 2.84**

2.66**

1.71**

1.31**

1.14

5.17**

Likelihood of conviction in a general and
special courts-martial

0.87

1.38

0.92

0.97

0.69

0.63*

2.77**

2.65**

11.15**

0.80*

0.87

1.00 1.70**

0.51**

0.41**

1.95**

4.12**

3.43**

1.87 **

NA

1.25 2.31 **

5.72 **

3.09 *

7.64 **

8.02 **

8.51 **

Likelihood of receiving nonjudicial punishments 1.77 **

1.42 **

0.99 1.62 **

3.95 **

2.34 **

2.34 **

2.00 **

4.05 **

Likelihood of trial in general and special courtsmartial with a recorded investigation

Likelihood of receiving a more severe
punishment when convicted in general and
special courts-martial
Likelihood of trial in summary courts-martial

Legend: ** next to the odds ratio indicates that the finding has a strong degree of statistical significance with a p-value less than 0.01.
* next to the odds ratio indicates that the finding is statistically significant with a p-value between 0.01 and 0.05.
If the space next to the odds ratio is blank, then the finding was not statistically significant.
“N” refers to the population size for each group.
“NA” indicates that the odds ratio for that group were not available because the number of servicemembers was too small to produce reliable findings.
The double line separates the multivariate analyses from the ordered logistic regression analysis.
Source: GAO analysis of Air Force investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These multivariate regression analysis
results demonstrate the degree to which a racial, gender, or rank group combined with length of
service is more likely than the reference category to be subject of a recorded investigation, tried in
general and special courts-martial, convicted in general and special courts-martial, tried in summary
courts-martial, and receive nonjudicial punishments. We used an ordered logistic regression analysis
to calculate the likelihood of receiving a more severe punishment as a result of being convicted in
general and special courts-martial. All racial categories listed are in reference to White
servicemembers, all gender groups listed are in reference to female servicemembers, all education
groups listed are in reference to servicemembers with more than a high school education, and all
rank groups are in reference to servicemembers between ranks E5 and E9. The Other race group
includes individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other
Pacific Islander, and multiple races. The term “recorded investigation” refers to where a
servicemember was the subject of a criminal investigation that was recorded in the Air Force’s military
criminal investigative organization’s database.

Page 156

GAO-19-344 Military Justice

Appendix VIII: Coast Guard Data and
Analyses
Appendix VIII: Coast Guard Data and Analyses

This appendix contains several tables that show the underlying data and
analyses used throughout this report relating to Coast Guard personnel
and military justice disciplinary actions from fiscal years 2013 through
2017. We did not include populations that contained fewer than 20
servicemembers in the populations presented in these tables to ensure
the protection of sensitive information. As a result, the populations
presented in this appendix may vary among the different tables and may
vary from the populations presented in other places in this report. Our
analyses of the Coast Guard’s investigations, military justice, and
personnel databases, as reflected in these tables, taken alone, do not
establish the presence or absence of unlawful discrimination.
The following tables and information are included in this appendix:
•

Table 46: Total Population of the Coast Guard by Race, Fiscal Years
2013–2017

•

Table 47: Summary Statistics by Race for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017

•

Table 48: Summary Statistics by Gender for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017

•

Table 49: Summary Statistics by Rank for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017

•

Table 50: Summary Statistics by Education for Coast Guard Military
Justice Actions, Fiscal Years 2013–2017

•

Table 51: Summary Statistics by Age for Coast Guard Military Justice
Actions, Fiscal Years 2013–2017

•

Table 52: Odds Ratios for Coast Guard Multivariate Regression
Analyses

Page 157

GAO-19-344 Military Justice

Appendix VIII: Coast Guard Data and Analyses

Table 46: Total Population of the Coast Guard by Race, Fiscal Years 2013–2017
Race

Population

American Indian/Alaskan Native
Asian

Number

Percent

1,234

2%

879

1%

Black or African American

3,404

5%

Hispanic

8,534

13%

Multiple races

3,253

5%

Unknown race

4,357

7%

White

45,043

68%

Total population

66,704

100%

Source: GAO analysis of Coast Guard personnel data. | GAO-19-344

Note: The total populations presented in this table represent the number of unique active-duty
servicemembers who served during fiscal years 2013 through 2017.

Table 47: Summary Statistics by Race for Coast Guard Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded
investigations
General and special
courts-martial

White

Black

N

N

%

N

66,704

45,043

68%

1,437

845

59%

175

133

64%

< 20

Hispanic

Other

Unknown race

%

N

%

N

%

N

%

3,404

5%

8,534

13%

5,366

8%

4,357

7%

144

10%

253

18%

114

8%

81

6%

42

20%

< 20

< 20

Legend: <20 refers to race groups that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Coast Guard investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers were not included in the total populations presented in this table to protect privacy.
As a result, the populations presented in this table may vary from the populations presented in the
body of the report or from other tables in this appendix. The Other race group includes individuals
who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and
multiple races. The total population presented in this table represents the number of unique activeduty servicemembers during fiscal years 2013 through 2017. The term “recorded investigation” refers
to where a servicemember was the subject of a criminal investigation that was recorded in the Coast
Guard’s criminal investigative organization’s database or was recorded in the Coast Guard’s military
justice database as an investigation. Percentages in this table may not add up to 100 due to
rounding, the exclusion of information identified as unknown or missing, and/or to protect privacy.

Page 158

GAO-19-344 Military Justice

Appendix VIII: Coast Guard Data and Analyses

Table 48: Summary Statistics by Gender for Coast Guard Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded investigations

Male

Female

N

N

Percent

N

Percent

66,704

56,117

84%

10,587

16%

1,437

1,278

89%

159

11%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Coast Guard investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Coast Guard’s criminal investigative organization’s database or
was recorded in the Coast Guard’s military justice database as an investigation. The summary
statistics for servicemembers tried in general and special and summary courts-martial were omitted
from this table protect privacy because a gender group had fewer than 20 servicemembers.

Table 49: Summary Statistics by Rank for Coast Guard Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded investigations

Rank E1-E4

Rank E5-E9 and officers

N

N

Percent

N

Percent

66,704

28,939

43%

37,765

57%

1,437

622

43%

815

57%

209

81

39%

128

61%

General and special courtsmartial
Legend: “N” refers to the population size for each group.

Source: GAO analysis of Coast Guard investigations, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The populations presented in this table
represent the unique records of active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Coast Guard’s criminal investigative organization’s database or
was recorded in the Coast Guard’s military justice database as an investigation. Although we
analyzed officers separately from enlisted servicemembers, for reporting purposes we combined
servicemembers in rank categories E5-E9 with officers in this table to protect privacy in those
instances when the number of officers was fewer than 20.

Page 159

GAO-19-344 Military Justice

Appendix VIII: Coast Guard Data and Analyses

Table 50: Summary Statistics by Education for Coast Guard Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded
investigations

High school or less

More than high school

Unknown education

N

N

Percent

N

Percent

N

Percent

66,704

43,985

66%

20,931

31%

1,788

3%

1,437

1,073

75%

325

23%

39

3%

202

171

82%

31

15%

< 20

General and special
courts-martial

Legend: <20 refers to education levels that have zero or fewer than 20 servicemembers, to protect privacy.
“N” refers to the population size for each group.
Source: GAO analysis of Coast Guard investigation, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. Populations that contained fewer than 20
servicemembers, including servicemembers with unknown education level, were not included in the
populations presented in this table to protect privacy. As a result, the populations presented in this
table may vary from the populations presented in the body of the report or from other tables in this
appendix. The total population presented in this table represents the number of unique active-duty
servicemembers during fiscal year 2013 through 2017. The term “recorded investigation” refers to
where a servicemember was the subject of a criminal investigation that was recorded in the Coast
Guard’s criminal investigative organization’s database or was recorded in the Coast Guard’s military
justice database as an investigation. Percentages in this table may not add up to 100 due to
rounding, the exclusion of information identified as unknown or missing, and/or to protect privacy.

Table 51: Summary Statistics by Age for Coast Guard Military Justice Actions, Fiscal Years 2013–2017
Population
Total population
Recorded
investigations
General and special
courts-martial

<25 years

25-30 years

30-40 years

≥40 years

N

N

Percent

N

Percent

N

Percent

N

Percent

66,704

19,108

29%

15,524

23%

20,927

31%

11,145

17%

1,437

399

28%

357

25%

510

35%

171

12%

209

44

21%

61

29%

82

39%

22

11%

Legend: “N” refers to the population size for each group.
Source: GAO analysis of Coast Guard investigation, military justice, and personnel data. | GAO-19-344

Note: The information presented in this table, taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. The total population presented in this table
represents the number of unique active-duty servicemembers during fiscal years 2013 through 2017.
The term “recorded investigation” refers to where a servicemember was the subject of a criminal
investigation that was recorded in the Coast Guard’s criminal investigative organization’s database or
was recorded in the Coast Guard’s military justice database as an investigation.

Page 160

GAO-19-344 Military Justice

Appendix VIII: Coast Guard Data and Analyses

Multivariate Regression
Analyses of Coast Guard
Data

The multivariate results listed below in table 52 show the odds ratios for
the multivariate regression analyses of Coast Guard data. We used
logistic regression to assess the relationship between the independent
variables, such as race, education, rank, or gender, with the probability of
being subject to a military justice action. Logistic regression allows for the
coefficients to be converted into odds ratios. Odds ratios that are
statistically significant and greater than 1.00 indicate that individuals with
that characteristic are more likely to be subject to a military justice action.
For example, an odds ratio of 1.55 for Black servicemembers would mean
that they are 1.55 times more likely to be subject to a military justice
action compared to White servicemembers. Odds ratios that are
statistically significant and lower than 1.00 indicate that individuals with
that characteristic are less likely to be subject to a military justice action.
We excluded age and years of service from the Coast Guard analyses
due to high correlation with the rank variable.

Table 52: Odds Ratios for Coast Guard Multivariate Regression Analyses

Recorded
investigations

Black

Hispanic

Unknown
race

Other

Male

High
school
or less

Unknown
education

Rank E5E9

Officers

2.36**

1.54**

1.03

1.13

1.43**

1.45**

1.52*

1.22**

0.71**

Legend: ** next to the odds ratio indicate that the finding has a strong degree of statistical significance with a p-value less than 0.01.
* next to the odds ratio indicates that the finding is statistically significant with a p-value between 0.01 and 0.05.
If the space next to the odds ratio is blank, then the finding was not statistically significant.
Source: GAO analysis of Coast Guard investigations and personnel data. | GAO-19-344

Note: The information presented in this table taken alone, should not be used to make conclusions
about the presence or absence of unlawful discrimination. These multivariate regression analysis
results demonstrate the degree to which a racial, gender, education, or rank group is more likely than
the reference category to be subject of a recorded investigation. The term “recorded investigation”
refers to where a servicemember was the subject of a criminal investigation that was recorded in the
Coast Guard’s criminal investigative organization’s database or was recorded in the Coast Guard’s
military justice database as an investigation. All racial categories listed are in reference to White
servicemembers, all gender groups listed are in reference to female servicemembers, all education
groups listed are in reference to servicemembers with more than a high school education, and all
rank groups are in reference to servicemembers between ranks E1 and E4. The Other race group
includes individuals who identified as American Indian/Alaska Native, Asian, Native Hawaiian/Other
Pacific Islander, and multiple races. We could not perform a multivariate regression analysis on
general and special courts-martial due to the small number of adjudications in the Coast Guard during
2013 through 2017.

Page 161

GAO-19-344 Military Justice

Appendix IX: Key Indicators for Military
Justice Actions
Appendix IX: Key Indicators for Military Justice
Actions

We found that age, rank, length of service, and education were indicators
of a servicemember’s likelihood of being the subject of a recorded
investigation, court-martial, or nonjudicial punishment across the military
services. 1 To analyze age, rank, length of service, and education, we
used bivariate regression analyses to determine which sub-population of
each attribute was most likely to be subject to a recorded investigation,
court-martial, or nonjudicial punishment. This appendix contains several
tables that show the rank, education, length of service, and age groups
most likely to be subject to a recorded investigation, tried in general and
special courts-martial, tried in summary court-martial, and receive a
nonjudicial punishment for all services from fiscal years 2013 through
2017. For the Coast Guard, we could not analyze age, rank, length of
service, and education as indicators for courts-martial or nonjudicial
punishment due to the small number of recorded military justice cases
from fiscal years 2013 through 2017. Our analyses of the services’
investigations, military justice, and personnel databases, as reflected in
these tables, taken alone, do not establish the presence or absence of
unlawful discrimination.
The following tables and information are included in this appendix:
•

Table 53: Servicemember Rank Groups Most Likely to Be Subject to
Investigations, Courts-Martial, and Nonjudicial Punishments when
Compared with All Other Rank Groups

•

Table 54: Overview of Servicemember Education Groups Most Likely
to Be Subject to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Education Groups

•

Table 55: Servicemember Length of Service Groups Most Likely to Be
Subject to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Length of Service
Groups

1

Investigations are recorded in the military services’ criminal investigative organizations’
databases when a servicemember is the subject of a criminal allegation made by another.
For purposes of this report, we state that the servicemember had a “recorded
investigation” to describe these cases. For additional explanation of the databases we
used to analyze investigations, please see Appendix I. For purposes of this report, we use
the term “likelihood” when discussing the odds ratios from the results of our regression
analyses. Odds ratios that are statistically significant and greater than 1.00 or lower than
1.00 indicate that individuals with that characteristic are more likely or less likely,
respectively, to be subject to a particular military justice action.

Page 162

GAO-19-344 Military Justice

Appendix IX: Key Indicators for Military Justice
Actions

•

Table 56: Servicemember Age Groups Most Likely to Be Subject to
Investigations, Courts-Martial, and Nonjudicial Punishments when
Compared with All Other Age Groups

Table 53: Servicemember Rank Groups Most Likely to Be Subject to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Rank Groups
Service

Investigations

Summary courts-martial

General and special
courts-martial

Nonjudicial punishments
(NJP)

Army

E1-E4
1.85 times more likely

No results due to incomplete
summary courts-martial data

E5-E9
1.51 times more likely

No results due to incomplete
NJP data.

Navy

E1-E4
1.37 times more likely

No results due to incomplete
summary courts-martial data

E5-E9
1.39 times more likely

No results due to incomplete
NJP data

Marine Corps

E1-E4
1.29 times more likely

E1-E4
1.52 times more likely

E5-E9
1.96 times more likely

E5-E9
1.56 times more likely

Air Force

E1-E4
2.08 times more likely

E1-E4
16.04 times more likely

E1-E4
3.42 times more likely

E1-E4
4.16 times more likely

Coast Guard

E5-E9
1.36 times more likely

No results due to small
number of cases

No results due to small
number of cases

No results due to small
number of cases

Source: GAO analysis of each military service’s personnel, investigations, and military justice database information. | GAO-19-344

Note: The results presented in this table were calculated through a bivariate regression model that
compared servicemembers in one rank group with a composite variable of servicemembers in all
other rank groups. We grouped the ranks in the Army, Navy, Marine Corps, and Coast Guard as; rank
group 1: E1-E4, rank group 2: E5-E9, rank group 3: officers and warrant officers. We grouped the
ranks in the Air Force as; rank group 1: E1-E4, rank group 2: E5-E6, rank group 3: E7-E9, rank group
4: officers and warrant officers.

Table 54: Servicemember Education Groups Most Likely to Be Subject to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Education Groups
Service

Investigations

Summary courts-martial

General and special
courts-martial

Nonjudicial punishments
(NJP)

Army

No results due to data
reliability issues

No results due to data
reliability issues

No results due to data
reliability issues

No results due to data
reliability issues

Navy

High school or less
2.16 times more likely

No results due to incomplete High school or less
summary courts-martial data 1.80 times more likely

No results due to incomplete
NJP data

Marine
Corps

High school or less
2.52 times more likely

High School or Less
3.93 times more likely

High school or less
2.09 times more likely

High school or less
2.49 times more likely

Air Force

High school/Some college
1.53 times more likely

High school/Some college
7.73 times more likely

High school/Some college
1.95 times more likely

High school/Some college
3.82 times more likely

No results due to small
number of cases

No results due to small
number of cases

No results due to small
number of cases

Coast Guard High school or less
1.56 times more likely

Source: GAO analysis of each military service’s personnel, investigations, and military justice database information. | GAO-19-344

Page 163

GAO-19-344 Military Justice

Appendix IX: Key Indicators for Military Justice
Actions

Table 55: Servicemember Length of Service Groups Most Likely to Be Subject to Investigations, Courts-Martial, and
Nonjudicial Punishments when Compared with All Other Length of Service Groups
Service

Investigations

Summary courts-martial

General and special
courts-martial

Nonjudicial punishments
(NJP)

Army

0-4 years
1.36 times more likely

No results due to incomplete 4-8 years
summary courts-martial data 1.90 times more likely

No results due to
incomplete NJP data

Navy

3-4 years
1.62 times more likely

No results due to incomplete 3-4 years
summary courts-martial data 1.37 times more likely

No results due to
incomplete NJP data

Marine Corps 5-6 years
1.63 times more likely

5-6 years
1.34 times more likely

5-6 years
1.96 times more likely

7-10 years
1.69 times more likely

Air Force

Over 6 years
9.07 times more likely

0-4 years
5.02 times more likely

Over 6 years
15.08 times more likely

Over 6 years
3.06 times more likely

Coast Guard

6-10 years
1.27 times more likely

No results due to small
number of cases

No results due to small
number of cases

No results due to small
number of cases

Source: GAO analysis of each military service’s personnel, investigations, and military justice database information. | GAO-19-344

Table 56: Servicemember Age Groups Most Likely to Be Subject to Investigations, Courts-Martial, and Nonjudicial
Punishments when Compared with All Other Age Groups
Service

Investigations

Summary courts-martial

General and special courtsmartial

Nonjudicial punishments
(NJP)

Army

Under 25 years old
1.63 times more likely

No results due to incomplete
summary courts-martial data

30-40 years old
1.37 times more likely

No results due to
incomplete NJP data

Navy

21-25 years old
1.48 times more

No results due to incomplete
summary courts-martial data

21-25 years old
1.25 times more likely

No results due to
incomplete NJP data

Marine Corps

21-25 years old
1.35 times more likely

21-25 years old
1.15 times more likely

26-30 years old
1.4 times more likely

31-40 years old
1.32 times more likely

Air Force

21-25 years old
1.54 times more likely

21-25 years old
2.98 times more likely

21-25 years old
1.44 times more likely

Less than 21 years old
2.56 times more likely

Coast Guard

30-40 years old
1.12 times more likely

No results due to small
number of cases

No results due to small number
of cases

No results due to small
number of cases

Source: GAO analysis of each military service’s personnel, investigations, and military justice database information. | GAO-19-344

Page 164

GAO-19-344 Military Justice

Appendix X: Comments from the Department
of Defense
Appendix X: Comments from the Department
of Defense

Page 165

GAO-19-344 Military Justice

Appendix X: Comments from the Department
of Defense

Page 166

GAO-19-344 Military Justice

Appendix X: Comments from the Department
of Defense

Page 167

GAO-19-344 Military Justice

Appendix X: Comments from the Department
of Defense

Page 168

GAO-19-344 Military Justice

Appendix XI: Comments from the
Department of Homeland Security
Appendix XI: Comments from the Department
of Homeland Security

Page 169

GAO-19-344 Military Justice

Appendix XI: Comments from the Department
of Homeland Security

Page 170

GAO-19-344 Military Justice

Appendix XII: GAO Contact and Staff
Acknowledgments
Appendix XII: GAO Contact and Staff
Acknowledgments

GAO Contact

Brenda S. Farrell, (202) 512-3604 or [email protected].

Staff
Acknowledgments

In addition to the contact named above, key contributors to this report
were Kimberly C. Seay, Assistant Director; Parul Aggarwal; Christopher
Allison; Renee S. Brown; Vincent M. Buquicchio; Won (Danny) Lee; Amie
M. Lesser; Serena C. Lo; Dae B. Park; Samuel J. Portnow; Clarice
Ransom; Christy D. Smith; Preston Timms; and Schuyler Vanorsdale.

(102463)

Page 171

GAO-19-344 Military Justice

GAO’s Mission

The Government Accountability Office, the audit, evaluation, and investigative
arm of Congress, exists to support Congress in meeting its constitutional
responsibilities and to help improve the performance and accountability of the
federal government for the American people. GAO examines the use of public
funds; evaluates federal programs and policies; and provides analyses,
recommendations, and other assistance to help Congress make informed
oversight, policy, and funding decisions. GAO’s commitment to good government
is reflected in its core values of accountability, integrity, and reliability.

Obtaining Copies of
GAO Reports and
Testimony

The fastest and easiest way to obtain copies of GAO documents at no cost is
through GAO’s website (https://www.gao.gov). Each weekday afternoon, GAO
posts on its website newly released reports, testimony, and correspondence. To
have GAO e-mail you a list of newly posted products, go to https://www.gao.gov
and select “E-mail Updates.”

Order by Phone

The price of each GAO publication reflects GAO’s actual cost of production and
distribution and depends on the number of pages in the publication and whether
the publication is printed in color or black and white. Pricing and ordering
information is posted on GAO’s website, https://www.gao.gov/ordering.htm.
Place orders by calling (202) 512-6000, toll free (866) 801-7077, or
TDD (202) 512-2537.
Orders may be paid for using American Express, Discover Card, MasterCard,
Visa, check, or money order. Call for additional information.

Connect with GAO

Connect with GAO on Facebook, Flickr, Twitter, and YouTube.
Subscribe to our RSS Feeds or E-mail Updates. Listen to our Podcasts.
Visit GAO on the web at https://www.gao.gov.

To Report Fraud,
Waste, and Abuse in
Federal Programs

Contact FraudNet:

Congressional
Relations

Orice Williams Brown, Managing Director, [email protected], (202) 512-4400,
U.S. Government Accountability Office, 441 G Street NW, Room 7125,
Washington, DC 20548

Public Affairs

Chuck Young, Managing Director, [email protected], (202) 512-4800
U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, DC 20548

Strategic Planning and
External Liaison

James-Christian Blockwood, Managing Director, [email protected], (202) 512-4707
U.S. Government Accountability Office, 441 G Street NW, Room 7814,
Washington, DC 20548

Website: https://www.gao.gov/fraudnet/fraudnet.htm
Automated answering system: (800) 424-5454 or (202) 512-7700

Please Print on Recycled Paper.