Military Equal Opportunity: Certain Trends in Racial and Gender Data May
Warrant Further Analysis (Letter Report, 11/17/95, GAO/NSIAD-96-17).

Pursuant to a congressional request, GAO reviewed the services' Military
Equal Opportunity Assessments (MEOA) to determine whether certain
active-duty personnel data reflect racial and gender disparities within
the services.

GAO found that: (1) MEOA do not consistently identify and assess the
significance of possible racial and gender disparities within the
services; (2) MEOA categories are reported differently by each service,
since the services are not required to report data on low rank
promotions and voluntary separations; (3) the Deputy Assistant Secretary
of Defense for Equal Opportunity (DASD EO) has not analyzed the
services' MEOA or DOD summaries for fiscal year 1993; (4) there are
significant statistical disparities in the number of minorities
considered for accessioning, career enhancement, and promotion; and (5)
some of these disparities can be attributed to job-related and societal
factors.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  NSIAD-96-17
     TITLE:  Military Equal Opportunity: Certain Trends in Racial and 
             Gender Data May Warrant Further Analysis
      DATE:  11/17/95
   SUBJECT:  Racial discrimination
             Sex discrimination
             Military promotions
             Military recruiting
             Combat readiness
             Minorities
             Monitoring
             Military personnel
             Fair employment programs
IDENTIFIER:  DOD Equal Opportunity Program
             
**************************************************************************
* This file contains an ASCII representation of the text of a GAO        *
* report.  Delineations within the text indicating chapter titles,       *
* headings, and bullets are preserved.  Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are           *
* identified by double and single lines.  The numbers on the right end   *
* of these lines indicate the position of each of the subsections in the *
* document outline.  These numbers do NOT correspond with the page       *
* numbers of the printed product.                                        *
*                                                                        *
* No attempt has been made to display graphic images, although figure    *
* captions are reproduced. Tables are included, but may not resemble     *
* those in the printed version.                                          *
*                                                                        *
* A printed copy of this report may be obtained from the GAO Document    *
* Distribution Facility by calling (202) 512-6000, by faxing your        *
* request to (301) 258-4066, or by writing to P.O. Box 6015,             *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time.                                    *
**************************************************************************


Cover
================================================================ COVER


Report to the Ranking Minority Member, Committee on National
Security, House of Representatives

November 1995

MILITARY EQUAL OPPORTUNITY -
CERTAIN TRENDS IN RACIAL AND
GENDER DATA MAY WARRANT FURTHER
ANALYSIS

GAO/NSIAD-96-17

Military Equal Opportunity

(703054)


Abbreviations
=============================================================== ABBREV

  DOD - Department of Defense
  DASD(EO) - Deputy Assistant Secretary of Defense for Equal
     Opportunity
  MEOA - Military Equal Opportunity Assessment

Letter
=============================================================== LETTER


B-261469

November 17, 1995

The Honorable Ronald V.  Dellums
Ranking Minority Member
Committee on National Security
House of Representatives

Dear Mr.  Dellums: 

Department of Defense (DOD) policy holds that discrimination that
adversely affects persons based on race, national origin, gender, or
other factors is counterproductive to combat readiness and mission
accomplishment and contrary to good order and discipline.  In
response to your request expressing concern that inequality of
treatment and opportunity could be affecting some of the more than
1.5 million active-duty military servicemembers, we initiated three
separate reviews.  In April 1995, we issued a report that identified
studies done by DOD during the past 20 years on discrimination in the
military.\1 In another review, we are examining the services' systems
for handling discrimination complaints from active-duty
servicemembers and will report on this issue later.  For this report,
we

  reviewed Military Equal Opportunity Assessments (MEOA) prepared by
     the military services and the use of those assessments by the
     Office of the Secretary of Defense in monitoring the services'
     equal opportunity programs and

  examined certain active-duty personnel data to determine whether
     possible racial\2 or gender disparities in selection rates
     existed. 

The methodology we used is one of several approaches that can be used
to identify possible racial or gender disparities in selection rates
to help DOD identify areas for further analysis.  We did not evaluate
whether any disparities identified were evidence that discrimination
had occurred.  Discrimination 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 DOD's personnel management practices. 


--------------------
\1 Equal Opportunity:  DOD Studies on Discrimination in the Military
(GAO/NSIAD-95-103, Apr.  7, 1995). 

\2 DOD defines five racial categories as (1) American Indian or
Alaskan Native, (2) Asian or Pacific Islander, (3) black (not of
Hispanic origin), (4) Hispanic, and (5) white (not of Hispanic
origin).  For purposes of this report, we refer to these groups as
American Indian, Asian, black, Hispanic, and white, respectively. 


   BACKGROUND
------------------------------------------------------------ Letter :1

Active-duty military personnel are not covered by Title VII of the
Civil Rights Act of 1964, as amended, or the implementing
governmentwide equal employment opportunity and affirmative action
regulations and guidelines of the Equal Employment Opportunity
Commission.  However, the Secretary of Defense has established a
separate equal opportunity program with similar requirements for
these personnel. 

In 1969, the Secretary of Defense issued a Human Goals Charter that
remains the basis for DOD's equal opportunity program.  It states
that DOD is to strive to provide everyone in the military the
opportunity to rise to as high a level of responsibility as possible
based only on individual talent and diligence.  The charter also
states that DOD should strive to ensure that equal opportunity
programs are an integral part of readiness and to make the military a
model of equal opportunity for all, regardless of race, color, sex,
religion, or national origin. 

To help ensure equal opportunity in the services, a 1988 DOD
directive and related instruction require that the services prepare
annual MEOAs.\3 In preparing their MEOAs, the services collect,
assess, and report racial and gender data in 10 categories.\4 The
Deputy Assistant Secretary of Defense for Equal Opportunity
(DASD(EO)) is primarily responsible for monitoring the services'
equal opportunity programs, including preparing written analyses of
the services' MEOAs and a DOD summary.\5

As recently as March 1994, the Secretary of Defense reaffirmed DOD's
equal opportunity goals, stating that equal opportunity is a military
and an economic necessity.  While noting that DOD has been a leader
in equal opportunity, the Secretary stated that it can and should do
better.  He initiated several measures, including a major DOD study
looking at ways to improve the flow of minorities and women into the
officer ranks from recruitment through high-level promotions. 


--------------------
\3 DOD Directive 1350.2, DOD Military Equal Opportunity Program,
dated December 23, 1988, and DOD Instruction 1350.3, Affirmative
Action Planning and Assessment Process, dated February 29, 1988. 

\4 The 10 MEOA reporting categories are (1) recruiting/accessions,
(2) force composition, (3) promotions, (4) professional military
education, (5) separations, (6) augmentation (reserve officers
transferring to an active-duty component) and retention, (7)
assignments (those considered career enhancing), (8) discrimination
or sexual harassment complaints, (9) utilization of skills (skill
categories with high or low concentrations of minorities or women),
and (10) discipline. 

\5 DASD(EO) is within the Office of the Assistant Secretary of
Defense for Force Management Policy, Office of the Under Secretary of
Defense for Personnel and Readiness.  DASD(EO) was created in 1994. 
Previously, the Assistant Secretary of Defense for Personnel and
Readiness was tasked with these responsibilities. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :2

MEOAs are not as useful as they could be for monitoring the services'
equal opportunity programs.  Overall, the analytical approach did not
consistently identify and assess the significance of possible racial
or gender disparities, the services reported on most of the MEOA
categories differently, and some data that would be helpful in
conducting more complete analyses was not required to be reported. 
In addition, DASD(EO) and his predecessors have not prepared the
required annual analyses of the services' MEOAs or the DOD summaries. 

Using MEOA and other data for fiscal years 1989 through 1993, we
analyzed accessions, assignments considered career enhancing, and
promotions to identify possible racial or gender disparities in
selection rates.  Our analysis showed statistically significant
disparities in each of the three categories, although the number of
disparities varied by category and service and by race and gender.\6
It is important to note that the existence of statistically
significant disparities does not necessarily mean they are the result
of unwarranted or prohibited discrimination.  Many job-related or
societal factors can contribute to racial or gender disparities. 
Further analyses would be required to determine the cause(s) of these
disparities. 

We are making several recommendations to the Secretary of Defense for
improving the usefulness of the services' MEOAs. 


--------------------
\6 Throughout this report, we use the term "statistically
significant" to denote those instances where the likelihood of the
outcome having occurred randomly is less than 5 percent. 


   MEOAS CAN BE IMPROVED
------------------------------------------------------------ Letter :3

According to DASD(EO)'s Director of Military Equal Opportunity, MEOAs
are the primary source of information for monitoring the services'
equal opportunity programs.  While MEOAs provide some useful
information, the analyses of this information did not consistently
identify and assess the significance of possible racial or gender
disparities.  In addition, data for
9 of the 10 MEOA reporting categories was reported inconsistently
among the services.  For the promotion and separation categories,
some key data that would be helpful in understanding the progression
of minorities and women through the ranks was not required to be
reported. 


      DOD IS NOT CONSISTENTLY
      IDENTIFYING THE SIGNIFICANCE
      OF POSSIBLE DISPARITIES
---------------------------------------------------------- Letter :3.1

In analyzing the outcomes of an organization's personnel actions for
possible racial or gender disparities, Equal Employment Opportunity
Commission guidance recommends using the racial and gender
composition of the eligible pool as a basis for comparison.  All
other things being equal, the racial and gender makeup of persons
selected for a particular action should--over time--reflect the
racial and gender composition of the eligible pool.  In other words,
the likelihood or odds of a particular outcome occurring for a
minority group should be about the same as for the majority or
dominant group in the long run.\7 When the actual odds are less and
the difference is statistically significant, and patterns or trends
are identified, further analysis would be necessary to determine the
cause(s) of the disparity. 

Seven of the 10 MEOA reporting categories lend themselves to
comparing the odds of a minority group member being selected to the
odds of a dominant group member being selected.  However, the DOD
directive and the related instruction do not require such an
analysis, and none was done by the services.  The services did make
some comparisons to the group average; that is, they compared a
minority group selection rate to the overall selection rate for all
groups (minority and majority).  But because the minority group was
usually so small compared to the total group, disparities in the
minority group selection rate compared to the overall group rate
often were not detected or appeared insignificant.  Also, this
approach is not helpful in identifying trends or patterns. 

Statistical significance testing can provide a basis to determine if
a disparity in the odds of being selected for a minority group
compared to the odds of the majority group is due to random chance. 
Statistical significance testing, over time, can also assist in
identifying trends or patterns in equal opportunity data that may
warrant further analysis.  In the fiscal year 1993 MEOAs (the latest
available), only the Army routinely reported statistical significance
testing results.  The Marine Corps and the Navy reported some
statistical significance testing.  The Air Force did not report any
statistical significance testing. 

While the DOD instruction on preparing MEOAs encourages the use of
statistical significance testing, its use is not required, and
instructions on how to conduct such tests are not provided.  All four
of the officials responsible for preparing the MEOAs for their
respective service said they did not have prior experience in
analyzing equal opportunity data and that DOD's instruction was not
particularly helpful. 


--------------------
\7 The odds of a group member being selected for an outcome is
determined by dividing the number of individuals selected by the
number not selected. 


      SERVICES REPORTED MEOA DATA
      DIFFERENTLY
---------------------------------------------------------- Letter :3.2

In analyzing the services' 1993 MEOAs, we found that the MEOA
reporting requirements were addressed differently by one or more of
the services in 9 of the 10 categories.  Only the promotion category
appeared to be consistently reported.  In most instances, definitions
and interpretations of what is called for were not consistent among
the services.  In some cases, one or more of the services did not
comply with the DOD instruction.  Following are examples of some of
the inconsistencies we found: 

  The Army specifically reported accessions for its professional
     branches, such as legal, chaplain, and medical.  The other
     services did not. 

  The Air Force, the Army, and the Navy reported on officers who had
     been separated involuntarily.  But the Army did not separately
     report officers who had been separated under other than
     honorable conditions or for bad conduct.  The Marine Corps did
     not report any separation data for officers. 

  The Air Force, the Army, and the Navy provided enlisted and officer
     assignment data by race and gender.  The Marine Corps combined
     into one figure its data on selections to career-enhancing
     assignments for its O-2 through O-6 officers for each racial and
     gender category and did not provide any information on its
     enlisted members. 

  The Air Force, the Marine Corps, and the Navy reported
     discrimination or sexual harassment complaints by race and
     gender.  The Army did not identify complainants by race and
     gender. 

  The Army reported utilization of skills data by each racial
     category and for women.  The Air Force reported skills data for
     blacks, Hispanics, and women.  The Marine Corps and the Navy
     combined the racial categories into one figure for each skill
     reported and did not report on women. 

  The Air Force and the Army included officers in their reports on
     discipline.  The Marine Corps and the Navy did not. 


      CERTAIN USEFUL DATA IS NOT
      REQUIRED FOR TWO MEOA
      CATEGORIES
---------------------------------------------------------- Letter :3.3

Two important factors in analyzing the progression of minorities and
women in the services are how competitive they are for promotions and
whether they are leaving the services at disproportionate rates. 
These factors have been of concern in the officer ranks.  In March
1994, the Secretary of Defense directed that a study of the officer
"pipeline" be conducted.  This study is still underway but is
addressing ways to improve the flow of minorities and women through
the officer ranks.  Although DOD's MEOA guidance requires reporting
on promotions and separations, it does not require the services to
report racial and gender data for all promotions or voluntary
separations. 

The guidance requires the services to report racial and gender data
in their MEOAs for promotions that result from a centralized
servicewide selection process.  For enlisted members, this includes
promotions to E-7, E-8, and E-9; for officers, this includes
promotions to O-4, O-5, and O-6.  For the most part, promotions at
the lower ranks are not routinely assessed.  In addition, the MEOA
data for officers in each of the services and enlisted members in the
Marine Corps is limited to those promotions that occurred "in the
zone."\8 We noted that about 900, or about 8 percent, of the
services' officer promotions and about 500, or about 19 percent, of
Marine Corps enlisted promotions in fiscal year 1993 were not
reported and were from either below or above the zone.  Without
routinely assessing promotions in the lower ranks and in each of the
promotion zones for possible racial or gender disparities, the
services' ability to identify areas warranting further analysis is
limited. 

The services are also required to report in their MEOAs racial and
gender data on involuntary separations, such as for reduction in
force or medical reasons, but are not required to report on the great
majority of separations that are for voluntary reasons.  In fiscal
year 1993, about 163,500 enlisted members and about 16,400 officers
voluntarily left the services for reasons other than retirement. 
Analyzing this data for racial or gender disparities could increase
the services' understanding of who is leaving the services and help
focus their efforts in determining why. 


--------------------
\8 Officer promotion selection boards consider three cohort groups
known as "below the zone," "in the zone," and "above the zone." Most
promotions are in the zone, which is considered the normal range or
length of service for promotion for that cohort group.  However, a
relatively small number of promotions go to officers who have
demonstrated outstanding leadership potential and are promoted ahead
of their cohort group, or below the zone.  Similarly, a small number
of promotions go to officers who are promoted after their cohort
group, or above the zone.  A similar system is used for enlisted
promotions in the Marine Corps.  The other services have different
procedures, but each told us that all enlisted promotions were
reported in MEOA. 


   DASD(EO) HAS NOT ANALYZED THE
   MEOAS
------------------------------------------------------------ Letter :4

DASD(EO) and his predecessors have not provided the services with
analyses of their MEOAs and have prepared a DOD summary only on 1990
data, even though both have been required annually since fiscal year
1988.  Although one Marine Corps official recalled receiving the
summary, she said that it was not helpful or constructive.  In
addition, some of the service officials responsible for their
service's MEOAs said the assessments were done primarily to satisfy
the DOD requirement.  They noted that, except for the promotion
category, MEOAs generally received little attention outside the
services' equal opportunity offices. 

Although DASD(EO) acknowledges these problems, they continue.  The
DOD instruction calls for the services to submit their MEOAs for the
prior fiscal year by February 1 each year and for DASD(EO) to
complete its analyses within 90 days.  The 1993 MEOAs were not all
received by DASD(EO) until May 1994.  As of the end of June 1995,
DASD(EO) had not provided its 1993 MEOA analyses to the services, and
the 1994 MEOAs have not been completed by all the services. 


   ANALYSIS SHOWS SOME
   STATISTICALLY SIGNIFICANT
   DISPARITIES
------------------------------------------------------------ Letter :5

To identify possible disparities, we analyzed three MEOA
categories--accessions, assignments, and promotions--for fiscal years
1989 through 1993.  We compared each minority group--American Indian,
Asian, black, and Hispanic--to the dominant white group and compared
females to males.  The analytical approach we used is one of several
methods for analyzing and identifying trends in equal opportunity
data.  It compares the odds of selection from a particular racial or
gender group to the odds of selection from the dominant group for a
particular outcome.  Used as a managerial tool, this methodology is
especially well suited to analyzing various outcomes for racial and
gender groups of very different sizes and selection rates.  Appendix
I contains a more detailed explanation of our methodology, including
our rationale for using this approach rather than alternative
approaches. 

Our analysis showed some racial or gender disparities, although the
number of disparities varied considerably among the MEOA categories,
across the services, and by race and gender.  Appendix II presents
our detailed results. 

Conclusions about DOD's personnel management practices cannot be
based solely on the existence of statistically significant
disparities.  Further analysis would be necessary to determine why
the disparities occurred.  Certain job criteria or selection
procedures may have an adverse impact on one or more groups, but if
the criteria or procedure can be shown to accurately measure required
job skills, the impact could be warranted.  Additionally, a group's
social characteristics may lead to disparities; for example, a
group's low interest or propensity to serve in the military could
help explain its lower odds of entering the services. 


      ACCESSIONS
---------------------------------------------------------- Letter :5.1

MEOAs did not report information on the eligible pools for
accessions.  At the suggestion of the DOD Office of Accession Policy,
we used certain data from the Defense Manpower Data Center for the
eligible pools.  For enlisted accessions, we used the gender and
racial makeup of persons who had taken the Armed Forces Qualification
Test.  This meant the individual had expressed interest in the
military and had made the time and effort to take the initial tests
for entrance into the services.  Because comparable eligible pool
data for officers was not available, the DOD Office of Accession
Policy suggested we use civilian labor force data for college
graduates between 21 and 35 years old as the eligible pool.  This
data provides a comparison to the overall racial and gender
composition of this portion of the U.S.  population but does not
account for an individual's interest or propensity to serve in the
military, which may vary by race and gender. 

Using these eligible pools, we found statistically significant racial
and gender disparities that may warrant further analysis.  For
example, in all the services, Asians had statistically significant
lower odds of entering as either an enlisted member or officer in
nearly all the years examined; the odds of blacks and Hispanics
entering the Air Force as either an enlisted member or officer were
statistically significantly lower than whites in most of the years we
examined; and in the Army, Hispanics had statistically significantly
lower odds than whites of entering the officer corps. 


      ASSIGNMENTS
---------------------------------------------------------- Letter :5.2

For the eligible pool for career-enhancing assignments, we used the
numbers of enlisted members and officers eligible for such
assignments reported in each of the services' MEOAs.  In the three
services we examined, we found that the odds of enlisted and officer
minorities being selected for these assignments were not
statistically significantly different from whites in most
instances.\9 An exception, however, was Asian officers in the Navy. 
As a group, they had statistically significant lower odds than whites
of being selected for most assignments.  In addition, the odds of Air
Force and Navy women officers being selected for many of the
assignments in the years we examined were statistically significantly
lower than the odds of selection for their male counterparts. 


--------------------
\9 Our analysis was limited to the Air Force, the Army, and the Navy
because the Marine Corps did not provide sufficient data in its MEOA
for a meaningful analysis. 


      PROMOTIONS
---------------------------------------------------------- Letter :5.3

Like assignments, we used the eligible pool data for promotions
reported in the services' MEOAs.  In about 37 percent of the enlisted
(E-7, E-8, and E-9) and officer (O-4, O-5, and O-6) promotion boards
we examined, one or more minority groups had statistically
significant lower odds of being promoted than whites.  We found
statistically significant lower odds of minorities being promoted
compared to whites most often (1) for blacks, (2) at the E-7 and O-4
levels, and (3) in the Air Force.  On the other hand, the odds of
females being promoted were not statistically significantly different
or were greater than the odds for males in nearly all the enlisted
and officer boards we examined. 


   RECOMMENDATIONS
------------------------------------------------------------ Letter :6

To help make the services' MEOAs more useful in monitoring the
services' equal opportunity programs, we recommend that the Secretary
of Defense direct DASD(EO) to do the following: 

  Devise methodologies for analyzing MEOA data that would more
     readily identify possible racial and gender disparities than
     current methods permit and establish criteria for determining
     when disparities warrant more in-depth analyses.  The Secretary
     may wish to consider the methodology we used in this report, but
     other methods are available and may suit the purposes of MEOAs. 

  Ensure that the services (1) use comparable definitions and
     interpretations in addressing the MEOA categories and (2)
     provide complete information for each of the MEOA categories. 

  Prepare the analyses of the services' annual MEOAs and the DOD
     summary, as required. 


   AGENCY COMMENTS AND OUR
   EVALUATION
------------------------------------------------------------ Letter :7

In commenting on a draft of this report, DOD concurred with the
report and stated that it has already initiated several efforts to
make the recommended improvements.  DOD's comments are reproduced in
appendix III. 


   SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :8

To evaluate whether MEOAs provided DASD(EO) with sufficient
information to effectively monitor the services' equal opportunity
programs, we reviewed the services' MEOAs for fiscal years 1989
through 1993.  In addition, we analyzed the services' fiscal year
1993 MEOA--the latest available at the time of our review--for
reporting completeness and consistency.  We reviewed the DOD
directive and instruction governing the military's equal opportunity
program.  We discussed preparation of MEOAs with cognizant officials
in the services and DASD(EO)'s Office of Military Equal Opportunity. 

To determine whether possible racial or gender disparities in
selection rates existed, we analyzed military accessions,
assignments, and promotions for active-duty enlisted members and
officers.  We chose to analyze these categories because relatively
large numbers of servicemembers were involved and, for the most part,
the necessary data was readily available.  For accessions, we used
data from the Defense Manpower Data Center.  For assignments and
promotions, we used data from the services' MEOAs.  We did not
independently verify the accuracy of the data.  We performed our
review from January 1994 to April 1995 in accordance with generally
accepted government auditing standards. 


---------------------------------------------------------- Letter :8.1

We are sending copies of this report to the Chairmen and Ranking
Minority Members of the Senate Armed Services Committee and the
Senate and House Committees on Appropriations; the Chairman, House
Committee on National Security; the Secretaries of Defense, the Air
Force, the Army, and the Navy; and the Commandant of the Marine
Corps.  Copies will also be made available to others upon request. 

Please contact me at (202) 512-5140 if you or your staff have any
questions concerning this report.  Major contributors to this report
are listed in appendix IV. 

Sincerely yours,

Mark E.  Gebicke
Director, Military Operations and
 Capabilities Issues


ODDS RATIO METHODOLOGY
=========================================================== Appendix I

The Equal Employment Opportunity Commission has established policies
and procedures for federal agencies to collect and analyze data on
civilian personnel actions such as hiring, assignments, and
promotions to determine whether selection procedures adversely affect
any race, sex, or ethnic group.  Although these policies and
procedures do not apply to active-duty military personnel, the
Department of Defense (DOD) directive and instruction related to its
military equal opportunity program set forth similar requirements. 

We chose not to use the "four-fifths" rule described in the
Commission's guidance for determining whether adverse impact may have
occurred.\1 As pointed out by the Commission, the four-fifths rule is
a "rule of thumb" and has limitations.  For example, when the
relevant groups are very large--as in the military--differences in
the ratio of the two selection rates greater than four-fifths may be
statistically significant; that is, areas of possible adverse impact
may not be detected if just the four-fifths rule is used. 

Therefore, to determine whether possible racial or gender disparities
existed in the military services' personnel actions that we examined,
we used an "odds ratio" methodology.  This methodology is especially
well suited to analyzing various outcomes for racial and gender
groups of very different sizes and selection rates.  Use of this
methodology also enabled us to do analyses that are more sensitive to
changes in the relative numbers of women and minorities than the more
traditional method, which compares selection rates (the number
selected divided by the total number eligible). 

The odds of a particular group member being selected for an outcome
is determined by dividing the number of individuals selected by the
number not selected.  An "odds ratio" is the odds of one group member
being selected divided by the odds of another group member being
selected for that same outcome.  If the odds of being selected for
both group members are equal, the ratio will be one.  When the ratio
is not equal to one, the methodology allows us to determine whether
the difference is statistically significant, that is, whether it is
likely due to random chance or not.\2 For purposes of this report, we
use the term statistically significant to denote those instances
where the likelihood of the outcome having occurred randomly is less
than 5 percent. 


--------------------
\1 According to the guidance, a selection rate for any group that is
less than four-fifths of the rate for the group with the highest rate
will generally be regarded by federal enforcement agencies as
evidence of adverse impact.  If a procedure is determined to have
adverse impact, agencies must revise the selection procedure to
eliminate the adverse impact or provide evidence that the procedure
accurately measures required job skills.  A number of factors,
however, must be considered in arriving at a final determination. 
For example, affirmative action efforts may occasionally affect other
groups' selection rates and lead to a preliminary finding that
adverse impact occurred.  However, instances of adverse impact
related to bona fide affirmative action efforts are permitted under
the guidelines. 

\2 For enlisted accessions and officer and enlisted assignments and
promotions, we used the chi-square distribution to determine the
statistical significance of the observed differences between race and
gender categories.  Basically, this method tests the hypothesis that
the mean (or average) frequencies of a particular occurrence are
distributed in the same proportions from category to category.  For
officer accessions, the number of persons in the eligible pool was so
large (millions of people) that small variations in the odds ratios
using the chi-square method showed up as significant.  To help make
sure the population sizes were not distorting the results, we used
the Bayesian approach, which is considered a better test for
statistical significance when large sample sizes are involved. 


   REDUCING THE NUMBER OF
   CALCULATIONS
--------------------------------------------------------- Appendix I:1

The odds ratio methodology is relatively straightforward but can
involve a large number of calculations and comparisons.  If we had
calculated odds ratios for each racial and gender group for each
personnel action outcome in the three Military Equal Opportunity
Assessment (MEOA) categories we examined--accessions,
career-enhancing assignments, and promotions--almost 3,000 odds
ratios would have been needed.  Instead of performing all these
calculations, we used "modeling" techniques to determine how race and
gender affected the reported outcomes for the three sets of data. 
Once we understood the effect race and gender had on the outcomes, we
had to calculate and analyze only the odds ratios that significantly
affected the actual outcomes.  For each personnel action, we
considered five different models, as follows: 

  Model one assumed that race and gender had no effect on the outcome
     of accessions, assignments, or promotions. 

  Model two assumed that only gender had an effect--that is, all
     racial groups would have equal odds of being selected for the
     outcome, but males and females would not. 

  Model three assumed just the opposite--males and females would have
     equal odds of being selected, but the racial groups would not. 

  Model four assumed that both race and gender affect the odds of
     selection independently of one another.  In other words, the
     odds ratios indicating the difference between males and females
     in one racial group would be the same as the corresponding
     ratios in the other groups. 

  Model five assumed that both race and gender had an effect and that
     the two factors operated jointly.  That is, the odds ratios
     describing racial differences varied by gender, and the odds
     ratios describing gender differences varied by racial group. 

Determining which model to use required two steps.  First, using
statistical software, we created a hypothetical database for each
model essentially identical to the actual data but modified to
reflect the assumptions we made.  For example, the hypothetical
database created for the third model assumed that the odds of males
and females being selected would be equal (that is, the odds ratio
would be 1.0).  Second, the hypothetical odds ratios were compared to
the actual odds ratios for each of the personnel actions.  If there
were significant differences, we rejected the model's assumptions.\3

In virtually all instances, model four was the most appropriate and
preferred way to present the results.  Its overall results were not
significantly improved upon by any of the other models.  This meant
that for the personnel actions we analyzed, we only needed to
calculate the odds ratios for each racial and gender group compared
to whites and males, respectively (see app.  II).  We did not have to
calculate the odds ratios for males and females within each racial
group because, according to the model, the gender difference was the
same across racial groups. 


--------------------
\3 Davis, James A., "Hierarchical Models for Significance Tests in
Multivariate Contingency Tables:  An Exegesis of Goodman's Recent
Papers," from Analyzing Qualitative/Categorical Data:  Log-Linear
Models and Latent Structure Analysis, by Leo Goodman, University
Press of America, 1978. 


RESULTS OF RACIAL AND GENDER
DISPARITY ANALYSIS
========================================================== Appendix II

This appendix presents the odds ratios we calculated for each of the
three MEOA categories we examined--accessions, assignments, and
promotions.  Some ratios are much less than 1 (less than three
one-thousandths, for example) or much greater (over 16,000, for
example).  Such extremes occurred when the percentage of persons
selected from a small-sized group was proportionately very low or
very high compared to the percentage selected from the dominant
group. 

Our tests of statistical significance, however, took group size into
account.  Therefore, although many odds ratios were less than one
(some much less), the disparity was not necessarily statistically
significant.  In the tables in this appendix, we have shaded the odds
ratios that indicate possible adverse impact; that is, the ratios are
less than one and statistically significant.  A more in-depth
analysis would be warranted to determine the cause(s) of these
disparities. 

As discussed in appendix I, we compared the odds for females with
those for males and the odds of minority racial groups with those for
whites.  To help the reader remember the relationships in our tables,
we have labeled the top of each column listing odds ratios with the
gender or racial group and symbols of what the proper comparison is. 
For example, F:M means the ratio compares the odds of females to
males and B:W means the ratio compares the odds of blacks to whites
for the particular outcome being analyzed. 

The odds ratios can also be used to make certain comparisons within
and among the services and identify trends whether they are
statistically significant or not.  If the objective, for example, is
to increase the representation of a particular minority group
vis-a-vis whites, the odds ratio should be greater than one.  When it
is not, it means whites are being selected in proportionately greater
numbers than the minority group. 


   ACCESSIONS
-------------------------------------------------------- Appendix II:1

Tables II.1 and II.2 present the odds ratios for enlisted and officer
accessions, respectively.  We compared gender and racial data for
those entering the military to the gender and racial composition of
selected eligible pools.  In determining what to use for the eligible
pool, we conferred with officials in DOD's Office of Accession
Policy. 

For the enlisted member eligible pool, we used those men and women
who had taken the Armed Forces Qualification Test and scored in the
top three mental categories during the respective fiscal year.\1
These were generally high school graduates who had been initially
screened by the recruiter for certain disqualifying factors such as a
criminal record or obvious physical disabilities.  Using test takers
as the eligible pool also took into account the propensity to serve
in the military, since the men and women taking the test had to make
the time and effort to do so.  Moreover, this data was readily
available from the Defense Manpower Data Center. 

For officers, determining a relevant eligible pool was not as
precise.  Officers primarily come from Reserve Officers' Training
Corps programs, officer candidate schools, and the military
academies, but no information was reported on the racial and gender
makeup of the programs' applicants in the services' MEOAs, nor was it
available from the Defense Manpower Data Center.  At the suggestion
of DOD's Office of Accession Policy, we used national civilian labor
force gender and racial statistics for college graduates 21 to 35
years old as the eligible pool.  This data was readily available from
the Defense Manpower Data Center, and nearly all officers have
college bachelor's degrees and are in this age group when they enter
the service. 

We could not account for an individual's propensity or desire to
serve as a military officer using civilian labor force data.  While
our analyses highlight those racial groups that entered the services'
officer corps at lower rates or odds compared to whites based on
their representation in the civilian labor force, further analyses
would be necessary to determine why this occurred. 

In both tables we present the odds ratios for females compared to
males.  In each of the 5 years we reviewed and across the services,
the odds of women entering the services were statistically
significantly lower than for men.  This fact is not surprising
considering that women's roles in the military are limited and they
may, as a group, have less interest or propensity to serve in the
military than men.  Even in recent years when the restrictions have
been loosened, the services have not reported accessing more than
about 14 percent of women for the enlisted ranks and about 19 percent
for the officer ranks, compared to over 50 percent representation in
the civilian labor force.  Nevertheless, we present the data to
illustrate the disparities among the services.  For example, in
fiscal year 1993, the odds of women in our eligible pool entering the
Marine Corps as officers were less than one-tenth the odds for men. 
In contrast, for the same year, the odds of women entering the Air
Force as officers were about one-third the odds for men. 

   Table II.1:  Odds Ratios for
   Enlisted Accessions, Fiscal
   Years 1989-93

   (See figure in printed
   edition.)

   Table II.2:  Odds Ratios for
   Officer Accessions, Fiscal
   Years 1989-93

   (See figure in printed
   edition.)


--------------------
\1 The categories are labeled I, II, and III.  Category III is
divided between IIIA and IIIB at the 50-percentile level.  Around 99
percent of those entering the military services in recent years
scored in test categories IIIB or higher. 


   ASSIGNMENTS
-------------------------------------------------------- Appendix II:2

Tables II.3 through II.6 present the odds ratios for enlisted and
officer career-enhancing assignments as identified by the services in
their respective MEOAs.  For the gender and racial makeup of the
eligible pools and of who was selected, we used data reported in the
MEOAs.  As previously noted, the Marine Corps data for officer
assignments is an accumulation of all its officers in the ranks O-2
through O-6.  Although we calculated the odds ratios for this data
and they are presented in table II.5, more detailed analysis by more
specific assignments may be appropriate before any conclusions are
drawn.  In addition, the Marine Corps did not report any assignment
data for its enlisted personnel. 

For several of the assignments, the MEOA data was insufficient for
our analysis; these instances are indicated as "no data." In others,
no minority candidates were in the eligible pool, and these instances
are indicated as "none" in the appropriate odds ratio column. 
Finally, in the Navy, combat exclusion laws prohibit women from
serving aboard submarines, and this is so noted in the chief of the
boat assignment for E-9s. 

   Table II.3:  Odds Ratios for
   Air Force Enlisted and Officer
   Career-Enhancing Assignments,
   Fiscal Years 1989-93

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)

   Table II.4:  Odds Ratios for
   Army Enlisted and Officer
   Career-Enhancing Assignments,
   Fiscal Years 1989-93

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)

   Table II.5:  Odds Ratios for
   Marine Corps Officer
   Career-Enhancing Assignments,
   Fiscal Years 1989-93

   (See figure in printed
   edition.)

   Table II.6:  Odds Ratios for
   Navy Enlisted and Officer
   Career-Enhancing Assignments,
   Fiscal Years 1989-93

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)


   PROMOTIONS
-------------------------------------------------------- Appendix II:3

Tables II.7 and II.8 present the odds ratios for enlisted and officer
promotion boards, respectively, for each of the services.  For the
gender and racial makeup of the eligible pools and of who was
selected, we used data reported in the MEOAs.  In several instances,
no promotion boards were held, or data was not reported in the
service's MEOA for a particular rank, service, and year; these are
noted as appropriate.  In other instances, no minority group
candidates were in the eligible pool for promotion to a particular
rank; we have indicated these as "none" in the appropriate ratio
column. 

   Table II.7:  Odds Ratios for
   the Services' Enlisted
   Promotion Boards, Fiscal Years
   1989-93

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)



   (See figure in printed
   edition.)

   Table II.8:  Odds Ratios for
   the Services' Officer Promotion
   Boards, Fiscal Years 1989-93

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)



   (See figure in printed
   edition.)




(See figure in printed edition.)Appendix III
COMMENTS FROM THE DEPARTMENT OF
DEFENSE
========================================================== Appendix II



(See figure in printed edition.)

Now on p.  9. 

Now on p.  9. 



(See figure in printed edition.)

Now on p.  9. 


MAJOR CONTRIBUTORS TO THIS REPORT
========================================================== Appendix IV

NATIONAL SECURITY AND
INTERNATIONAL AFFAIRS DIVISION,
WASHINGTON, D.C. 

Sharon A.  Cekala, Associate Director
A.H.  Huntington, III, Assistant Director
Laura L.  Talbott, Evaluator-in-Charge
Frank Bowen, Senior Evaluator
Kerry A.  O'Brien, Evaluator

GENERAL GOVERNMENT DIVISION,
WASHINGTON, D.C. 

Douglas M.  Sloane, Social Science Analyst

*** End of document. ***