chapter three differences in retention and promotion …€¦ · there has been little change in...

23
25 Chapter Three DIFFERENCES IN RETENTION AND PROMOTION FOR MINORITY AND FEMALE LINE OFFICERS The purpose of our empirical analysis was to assess whether there are differences in career progression between officers in dif- ferent racial, ethnic, and gender groups. Specifically, we wanted to compare: how officers in different minority/gender groups enter military service, whether they choose to stay in service, and whether they are selected for promotion. These three career outcomes mix choices made by the officer and choices made by the officer’s service. Before they are commissioned, officer candidates choose among the entry programs (com- missioning sources) for which they qualify. Retention is a voluntary decision, although it is important to note that some individuals leave in anticipation of promotion failure and subsequent involuntary separation. On the other hand, promotion is an involuntary out- come from the officer’s perspective. A particular goal of our analysis was to separate the voluntary outcome—retention—from the in- voluntary outcome—promotion. THE OFFICER COHORT DATA FILES To carry out this analysis, we used a specially created data file pre- pared by the Defense Manpower Data Center (DMDC); this file was

Upload: others

Post on 22-Sep-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

25

Chapter Three

DIFFERENCES IN RETENTION AND PROMOTION FORMINORITY AND FEMALE LINE OFFICERS

The purpose of our empirical analysis was to assess whether thereare differences in career progression between officers in dif-ferent racial, ethnic, and gender groups. Specifically, we wanted tocompare:

• how officers in different minority/gender groups enter militaryservice,

• whether they choose to stay in service, and

• whether they are selected for promotion.

These three career outcomes mix choices made by the officer andchoices made by the officer’s service. Before they are commissioned,officer candidates choose among the entry programs (com-missioning sources) for which they qualify. Retention is a voluntarydecision, although it is important to note that some individuals leavein anticipation of promotion failure and subsequent involuntaryseparation. On the other hand, promotion is an involuntary out-come from the officer’s perspective. A particular goal of our analysiswas to separate the voluntary outcome—retention—from the in-voluntary outcome—promotion.

THE OFFICER COHORT DATA FILES

To carry out this analysis, we used a specially created data file pre-pared by the Defense Manpower Data Center (DMDC); this file was

Page 2: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

26 Minority and Gender Differences in Officer Career Progression

originally prepared for researchers at the Naval Postgraduate School.The file included an extract of the Officer Master Record created atthe end of each fiscal year served by every officer in seven cohorts:1967, 1970, 1977, 1980, 1983, 1987, and 1991.1 We also received first-year records for the 1994 cohort, which we used to compare enteringcharacteristics over time. This chapter begins with a brief descrip-tion of these data, followed by descriptions of changes in the minor-ity and gender composition of the entering cohorts, the commission-ing sources through which the officers prepared for military service,and the military occupations they entered. This chapter concludeswith an analysis of the career progression of the different racial andgender officer groups in these cohorts.

The DMDC was only able to provide records beginning in 1977 forthose officers still in service from the two earliest cohorts in our datafile—1967 and 1970. For all cohorts, the last annual record was for1993. Consistent with our focus, we deleted the records for officersin professional occupations (principally medical, legal, and religiousprofessionals). Substantial proportions of women officers are repre-sented in these fields, particularly the medical professions. Thus, therestriction eliminated slightly more than half of women officers fromour study, but only one-fifth of the men. The restriction eliminatedroughly the same proportion of whites, blacks, and other minoritiesfrom the study.2

These data provide limited information. For each person, therecords indicate:

• Entry path: commissioning source, prior federal military service(used to determine prior service entrants), and military occupa-tion;

______________ 1The cohorts are separated by three- or four-year intervals. In keeping with this pat-tern, the original data files included the 1973 cohort. However, the data for this cohortturned out to be unusable.2In the 1977 cohort, 24 percent of male officers and 52 percent of female officers werein the professional occupations. By the 1991 cohort, 20 percent of male officers and 56percent of female officers were in the professional occupations. In the 1977 cohort, 28percent of white officers, 21 percent of black officers, and 50 percent of other minorityofficers were in the professional occupations. By the 1991 cohort, 27 percent of whiteofficers, 26 percent of black officers, and 27 percent of other minority officers were inthe professional occupations.

Page 3: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 27

• Retention: inferred from the number of annual records;

• Promotion: inferred from pay grade in each year of active com-missioned service;

• Personal characteristics: marital status and number of depen-dents (in each year), race/ethnicity, gender.

Defining Race and Ethnicity

Our preliminary analysis of these data uncovered inconsistencies inthe recording of race and ethnicity across cohorts. Race and ethnic-ity were recorded in a single data element with a limited number ofcategories before 1980, but the records were revised to separate raceand ethnicity after this date. In exploring these data to determinehow best to handle this change, we discovered a pattern of changingracial and ethnic identification for Hispanic and Native Americanservice members. In the later cohorts we studied, more membersidentified themselves as belonging to these ethnic groups, and moreself-identified Hispanics and Native Americans listed their race as“other.”3 Similarly, some Hispanics and Native Americans in theearlier cohorts changed their racial designation over time.

To carry out this analysis of career progression, we needed to defineracial and ethnic groups in a way that remained consistent over time.The most consistent classification possible over the long period oftime covered by our data defines only three racial groups: white,black, and other. With this classification, some of the growth in the“other” group over time is misleading because it merely reflects theincreasing tendency for Hispanics to list their race as other; the im-pact is negligible on the white group and virtually nonexistent for theblack group.

______________ 3See Appendix C for additional details on racial and ethnic identification in the mili-tary. Similar patterns are emerging in other data systems as well, including the U.S.Census, and they are described in three volumes: (1) National Research Council(1995); (2) Bean and Tienda (1987); and (3) Kissam et al. (1993).

Page 4: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

28 Minority and Gender Differences in Officer Career Progression

DIVERSITY OF OFFICER ACCESSIONS OVER TIME

Table 3 shows how the racial and gender mix of officer accessionschanged in the 1977–1991 cohorts, the years for which we have data.4

Like all the data reported in this chapter, the table excludes officersin the scientific, legal, religious, and medical professions. After 1977,there was little change in diversity in the accession cohorts. Therewas an increase in the fraction of officers from other minority groupsin the 1987 and 1991 cohorts. Some of this increase is illusory since itreflects the change in racial identification of Hispanics; un-fortunately, no data exist that would allow us to quantify thesechanges in self-identification in the military. The fraction who werefemale increased from 9 percent in 1977 to 11 percent in 1991; sincethis figure varied over the time period, there is no discernible trend.

We are unable to show the mix for the 1967 and 1970 cohorts be-cause we have no records for individuals in these cohorts who leftservice before 1977. Limited data on the fraction of blacks, and sepa-rately on the fraction of women, in earlier officer accession cohortsare reported in DoD (1998). This data series begins in 1973 with theadvent of the All Volunteer Force (AVF). Since it does not report databy occupational group, it is not possible to determine racial and gen-der diversity among nonprofessionals. However, these data do pro-vide a broad indication of the trends in the early years of the AVF.Between 1973 and 1977, there was a dramatic increase in the diver-sity of new officer accessions. The fraction of new officers who wereblack increased from 3 percent to 7 percent, the fraction of Hispanicsincreased from just over 0 percent to 1 percent, and the fraction ofwomen increased from 8 percent to 14 percent.

As Table 3 shows, however, there are noticeably different levels of di-versity and patterns of change in diversity across the military ser-vices. The Army and Air Force are more diverse than are the othertwo services, with each having about one-quarter of their new

______________ 4Unfortunately, the data used for our analysis do not allow us to examine very recentchanges in the diversity of the officer corps. Other published data indicate that theproportion of recent officer accessions who are black has risen from 7.4 percent in1993 to 8.5 percent in 1997, and the proportion who are female has risen from 18.5percent in 1993 to 19.4 percent in 1997 (Office of the Assistant Secretary of Defense:1998).

Page 5: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 29

Table 3

Percent of Officer Nonprofessional Accessions by Race and Gender:1977–91 Cohorts, by Service, and for All Services

1977 1980 1983 1987 1991DoD TotalWhite Male 82.2% 79.6% 81.2% 82.0% 78.3%Black Male 6.6 5.3 6.7 5.5 6.1Other Minority 1.9 2.1 2.0 2.8 4.4White Female 7.9 11.1 8.0 8.0 9.1Black Female 1.2 1.6 1.8 1.4 1.4Other Minority 0.2 0.3 0.3 0.4 0.7

ArmyWhite Male 76.1% 76.9% 75.3% 73.9% 74.9%Black Male 8.9 6.3 10.0 10.1 9.2Other Male 2.7 2.8 1.9 2.5 3.0White Female 10.3 11.6 9.7 10.1 9.6Black Female 1.7 1.9 2.8 2.9 2.9Other Female 0.3 0.4 0.3 0.5 0.5

NavyWhite Male 90.4% 86.3% 87.3% 89.2% 80.4%Black Male 3.6 3.0 3.1 3.6 5.1Other Male 1.1 2.1 2.5 2.5 5.4White Female 4.6 8.1 6.2 4.3 7.6Black Female 0.3 0.3 0.6 0.3 0.9Other Female 0.0 0.2 0.3 0.1 0.7

Air ForceWhite Male 80.8% 75.8% 81.0% 81.5% 76.7%Black Male 7.0 6.1 6.2 3.2 4.5Other Male 1.8 1.8 1.9 3.0 4.6White Female 8.6 13.5 8.4 10.3 12.2Black Female 1.6 2.3 2.1 1.3 0.9Other Female 0.3 0.4 0.4 0.8 1.1

Marine CorpsWhite Male 91.6% 91.4% 87.4% 85.7% 86.3%Black Male 3.9 3.2 6.3 5.6 4.1Other Male 0.5 0.4 1.8 3.6 5.1White Female 3.8 4.5 4.3 4.6 3.9Black Female 0.2 0.4 0.2 0.3 0.2Other Female 0.0 0.0 0.0 0.1 0.3

NOTES: Excludes officers in professional occupations. Totals may not equal 100because of rounding error.

Page 6: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

30 Minority and Gender Differences in Officer Career Progression

1977

0

10

20

30

40

50

60

Academy ROTCScholarship

ROTCRegular

OCS/OTS DirectAppt.

Unknown

Academy ROTCScholarship

ROTCRegular

OCS/OTS DirectAppt.

Unknown

White MalesMinority MalesWhite FemalesMinority Females

1991

0

10

20

30

40

50

60

Per

cent

age

Per

cent

age

RAND MR1184-2

Figure 2—Commissioning Source of Entering Officersby Race and Gender, 1977 and 1991

(Percent of Minority/Gender Group, Excludes Professional Occupations)

officers belonging to a nonwhite-male group in the 1991 cohort.There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was an anomalous recruitingyear for this service.5 In contrast, the Air Force has experienced adecline in black officers, both male and female, and an increase in

______________ 5At this time, low interest in military service due to Vietnam and inadequate paycaused the services, especially the Army, to have recruiting problems.

Page 7: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 31

the other minority groups. Diversity increased sharply in the Navybetween the 1987 and 1991 entering cohorts to just below the Armyand Air Force levels; every minority/gender group showed an in-crease. The least diverse service is the Marine Corps, but the numberof other minority males there has increased rapidly in the last tenyears.

COMMISSIONING SOURCE AND OCCUPATIONALASSIGNMENT

The modest change in racial and gender mix for minority andwomen officers after 1977 was accompanied by more dramaticchanges in the commissioning sources and initial occupations. Mostanalyses of officer progression, including ours, have found thatcommissioning source and occupation are factors in retention andpromotion.6 Later in this chapter, when we estimate differences inthese important determinants of career progression, we will controlfor minority/gender differences in the way officers entered intoservice.

Most of the cohorts we studied entered service during the 1970s and1980s, when the military was expanding, and early 1990s, when amajor drawdown began. As the total number of officer accessionsvaried—between 12,000 and 20,000—the services changed their useof the different commissioning sources. The services took a dispro-portionate share of the cut in officer accessions from the regular(nonscholarship) ROTC programs and, to a lesser extent, from theOCS/OTS. The OCS/OTS have produced more or fewer officers ascommissioning requirements increased or decreased because of theshort lead time between program entry and commissioning.

Figure 2 shows the change in the mix of commissioning sources byrace and gender between the earliest cohort we observed at entry,1977, and the latest cohort whose career progression we studied,1991. White male officers’ commissioning sources shifted to theacademies and ROTC scholarship programs and away fromOCS/OTS. Before and during the drawdown, minority male officerswere considerably more likely than white males to enter through

______________ 6See Bowman (1990), Mehay (1995), and North et al. (1995).

Page 8: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

32 Minority and Gender Differences in Officer Career Progression

regular ROTC programs, but the overall decline in these programsnarrowed this difference. Much more dramatic changes are seen forwomen officers because the first women graduated from the militaryacademies or with ROTC scholarships in 1980. The differencesamong women are also more pronounced. In both years, a higherfraction of minority women entered through the regular ROTCprogram, whereas white women were more likely to enter throughOCS/OTS in 1977 and from one of the academies or with an ROTCscholarship in 1991.

Figure 3 examines changes between 1977 and 1991 in the fraction ofofficers entering with prior service. Between 15 and 25 percent ofmale officers come from the enlisted ranks, and the racial differenceamong men in this characteristic is modest (Figure 3). This fractionis smaller for white women, but minority women are as likely as mento have had prior service. In recent years, the commissioningprograms for enlisted personnel have increasingly been a source ofminority officers.

All the services assign their officers to an occupation before or soonafter they are commissioned. However, the personnel records do notrecord this occupation until the officers complete their occupationaltraining, so we identified the officers’ eventual occupations from thefourth-year records. White and minority male officers showed gen-erally similar changes between the 1977 and 1991 cohorts, as therewas a shift to tactical, intelligence, and supply occupations and awayfrom maintenance and administrative occupations (Figure 4).Among male officers, minorities were less likely to be in tactical(combat) occupations and more likely to be in supply occupations,and this difference grew somewhat in the later cohort. This differ-ence was greater for black males than for other minority males (notshown). Of course, female officers are considerably less likely to beassigned to tactical occupations, but the easing of the occupationrestrictions described in Chapter Two has enabled far more womento enter these fields in recent cohorts. Although the proportion ofwomen in administration has declined over time, so has the propor-tion of men in administration, and the female-male “gap” has actu-ally grown a little over these cohorts. Finally, the differencesbetween white women and minority women largely mirror thedifferences between white men and minority men.

Page 9: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 33

0

5

10

15

20

25

White males Minority males White females Minority females

1977

1991

RAND MR1184-3P

erce

ntag

e

Figure 3—Percent of Entering Officers Who Have Prior Enlisted Service,1977 and 1991

(Percent of Minority/Gender Group, Excludes Professional Occupations)

MARITAL STATUS

Women officers are considerably less likely than men to be marriedat all career stages. In our data set, one-half of male officers weremarried at entry, compared to only one-third for women officers. Inthe senior ranks (O-5 and O-6), 90 percent of the men and 55 percentof the women were married. Black women were the least likelyto marry if they stayed in service. This difference in marital status,and its potential effects on careers, is discussed further in the nextchapter.

In conducting our empirical analysis, we had hoped to be able to de-termine how getting married and having children affect retentionand promotion, especially for women officers. However, our datadid not allow us to separate the effects of marriage and children from

Page 10: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

34 Minority and Gender Differences in Officer Career Progression

1977

0

10

20

30

40

50

60

Per

cent

age

0

10

20

30

40

50

60

Per

cent

age

1991

Tactical Intelligence Mainte-nance

Adminis-tration

Supply None/other

Tactical Intelligence Mainte-nance

Adminis-tration

Supply None/other

RAND MR1184-4

White MalesMinority MalesWhite FemalesMinority Females

Figure 4—Distribution of Military Occupations, 1977 and 1991 Cohorts(Percent of Minority/Gender Group, Excludes Professional Occupations)

the gender effects beyond the very early stages of the career.7 Todetermine the effects of marriage and children on men versuswomen, we needed to be able to estimate the differences in promo-tion and retention at each stage for all possible combinations of gen-der and marital status (or parental status). We could not do this for

______________ 7Married personnel are more likely to be promoted early on, but this effect is smallerfor women. Married men are also more likely to be retained, but married women aremore likely to leave. See Appendix C for more information on marital status.

Page 11: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 35

several reasons. We could not measure outcomes for single men be-yond the early career stages because almost all men marry early intheir careers. Even if we could have done this, the data do not tell uswho left service to get married or have children. Consequently, wewould have treated an officer leaving to get married as the loss of asingle officer and failed to detect the connection to marriage.Further, we expect that, even for those whose marital status was cor-rectly identified, retention and marriage decisions are not made in-dependently. Unless these decisions are fully independent, it is notappropriate to include marital status as an explanatory variable inthe retention analysis. For all these reasons, we decided to drop anyconsideration of marital status and number of children in our finaldata analysis. To the extent that these factors affect women differ-ently than men, they are included in the differences in promotionand retention estimated for women versus men.

DIFFERENCES IN RETENTION AND PROMOTION

We analyzed career progression as a series of retention and promo-tion outcomes, each conditional on its predecessor. The outcomesare: retention at O-1, promotion to O-2, retention at O-2, promotionto O-3, and so on. So, for example, we analyze promotion to O-3 onlyfor officers who stay through O-1, are promoted at O-2, and staythrough O-2. We analyzed nine outcomes:

1. retention at O-1 and promotion to O-2

2. retention at O-2

3. promotion to O-3

4. retention at O-3

5. promotion to O-4

6. retention at O-4

7. promotion to O-5

8. retention at O-5

9. promotion to O-6.

Page 12: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

36 Minority and Gender Differences in Officer Career Progression

We could not establish separate retention and promotion outcomesfor O-1 to O-2 because these happen too quickly to observe throughannual records. We include this period in the “promotion” categorybecause the departures often result from failure to perform duringtraining.

The sample used for each of the nine outcomes included all individ-uals who reached the relevant career stage. An officer who left ser-vice at the O-3 retention point would be in the samples for the firstfour outcomes but not the last five. Since our data were for the years1977 through 1993, they covered different periods of service for dif-ferent cohorts, and the retention and promotion outcomes weremeasured for different subsets of these cohorts. For each outcome,Table 4 shows which cohorts were used for each retention and pro-motion point.

Determining Retention Outcomes

Earlier, we indicated that determining promotion outcomes was dif-ficult. That is because the data do not record who was considered forpromotion. To infer whether individuals were considered for a par-ticular promotion, we looked at whether they stayed in service into

Table 4

Cohorts Used in Analysis of Retention and Promotion at Each Pay Grade

CohortWindow 1967 1970 1977 1980 1983 1987 1991

RetentionO-2 √ √ √ √O-3 √ √O-4 √ √ √O-5 √ √

PromotionO-1–O-2 √ √ √ √ √O-2–O-3 √ √ √ √O-3–O-4 √ √O-4–O-5 √ √O-5–O-6 √ √

Year of Servicein Data 10–27 7–24 1–17 1–14 1–11 1–7 1–3

Page 13: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 37

the time interval when others in the same cohort were promoted.These promotions occur over a period of up to three years, whichvary by service and cohort. Therefore, we identified a three-yearpromotion window for each grade, cohort, and service based on thedistribution of promotions in the data. As an example of how thisworked, we defined an individual Army officer as being in the O-4promotion window if he or she was in service at least the first of thethree years others in the same 1977 cohort of Army officers werepromoted. This same method was used to identify the other promo-tion windows for the 1977 Army officer cohort. We then replicatedthis analysis for the other Army cohorts and the cohorts in each otherservice. For each cohort and service, we evaluated promotion ratesonly to those grades for which our data covered the full three-yearpromotion window.

The years between promotion windows were defined as retentionwindows. Individuals not promoted to a particular grade were notconsidered to have ever entered the succeeding retention window.For example, officers who were not selected for promotion to O-5were not included in the sample for the O-5 retention window. Table5 shows the sample sizes and promotion/retention rates in the sam-ple that resulted from this analysis. Note that, although we havecombined the charts and services in Table 5, separate windows weredefined for each cohort in each service.

Table 5

Sample Size and Promotion/Retention Rate by Pay Grade

Window Sample SizePercent

Retained/Promoted

RetentionO-2 61,837 92.1%O-3 25,028 64.6O-4 17,556 89.3O-5 8,465 68.5

PromotionO-1–O-2 76,337 95.3%O-2–O-3 56,926 87.9O-3–O-4 16,176 74.6O-4–O-5 10,619 73.6O-5–O-6 5,800 50.2

Page 14: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

38 Minority and Gender Differences in Officer Career Progression

Figure 5 again combines the data for all cohorts and services to showthe average timing of career windows and the average fraction of theentering cohorts we studied who remained in the military at the endof each window. Since officers who fail to make a promotion are notcounted in the next window even if they stay, Figure 5 looks some-what different than the usual year-of-service (YOS) cohort profile,which includes these individuals until they actually leave.

Method for Estimating Differences

For each of the nine windows, we estimated the difference in thefraction retained or promoted between white male officers and up tofive minority/gender groups, depending on sample sizes: black men,other minority men, white women, black women, and other minoritywomen. This is different from other studies we are aware of, whichtypically have estimated differences for whites versus minorities andfor men versus women, but not for groups defined by race/ethnicityand gender.

We used logistic regressions to separate the effects of race and gen-der from the other factors measured in the cohort data file, such as

RAND MR1184-5

0

25

50

75

0 10 20

O 1

O 2

O 3

O 3

O 4

O 4

O 5

O 5

O 6

Retention

Promotion

Years of Service

Per

cent

left

in S

ervi

ce

100

5 15

Figure 5—Average Profile of Officer Cohorts Studied

Page 15: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 39

commissioning source, prior enlisted service, and military occupa-tion. We should note here that the effects of other factors that mightbe correlated with the outcomes we analyzed were not measured inthe cohort data file (e.g., individual aptitude, quality of undergradu-ate education, and physical fitness are not controlled for in ourmodel). To the extent that these unmeasured factors systematicallydiffer for minority and female officers, their effects will be includedin the retention and promotion differences we estimated for thesegroups of officers.

Estimation Results

Figures 6–8 display the estimated differences in the completion ratesfor black and other minority males, white females, and black andother minority females while holding constant the other variables inthe model. These other variables are:

• whether the person had prior enlisted service;

• military service;

• accession source: academy, ROTC scholarship, ROTC regular,OCS/OTS, direct appointment, unknown accession source;

• occupation: executive, intelligence, engineering and mainte-nance, administration, supply/procurement, other; and

• cohort: varies by promotion or retention window.

The figures show the differences between the groups of minorityand/or female officers and white male officers in the fractions whocomplete the retention and promotion windows.8 For example, the+7 percent estimate in Figure 6 means that the percentage of blackmales who complete the O-3 retention window is 7 points higherthan the percentage of white males who complete the same window.Again, these results are for officers who are not in a profession suchas medicine.

______________ 8These estimates are transformed from the logistic regression coefficients by evaluat-ing the expression ρ ρ βi i(1 ) i− ˆ , where ρi is the average fraction in all relevant cohortsprogressing through the ith window and β̂i is the coefficient from the logisticregression that measures the difference in the fraction for a minority/gender group.

Page 16: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

40 Minority and Gender Differences in Officer Career Progression

Figure 6 presents the results for black men in all career windows andfor other minority men through the O-4 retention window (beyondthis point, the sample size was too small to estimate their outcomes).Generally, minority men are more likely than white men to stay in

RAND MR1184–6

Retention Windows

0%

7%*

–3%*

11%*

0% 0%

5%

–10–8–6–4–2

02468

1012

O–2 O–3 O–4 O–5

Black males

Other minoritymales

Promotion Windows

–2%* –2%*–4%*

–6%*

5%

–2%*–3%*

–8%*–10–8–6–4–2

02468

1012

O–1 to O–2 O–2 to O–3 O–3 to O–4 O–4 to O–5 O–5 to O–6

Per

cent

age

Per

cent

age

*Differences that are significant at .05 level.

Figure 6—Differences in Completion Rates for Minority Male vs. WhiteMale Officers

Page 17: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 41

service between promotion periods.9 The one exception to this is af-ter promotion to O-4, when black men are slightly less likely to re-main. Some of the officers who leave at this point have had prior en-listed service and have reached the 20-year retirement point.Although slightly more black officers have prior enlisted service andmight leave at this point, our analysis controls for this difference. Wewere not able to determine what other factors might be contributingas well.

In contrast, all groups of minority men are consistently less likelythan white men to successfully pass through promotion windows.The pattern of negative differences is broken, however, at promotionto O-6. By this point, the remaining officers have all compiledrecords of consistently superior performance. Also, the sample foranalyzing the O-6 promotion window was very small since it in-cluded officers who were commissioned in 1967 and 1970, when en-tering cohorts were far less diverse.

The lower promotion rates largely counteract the higher retentionrates for black men. Thus, combining promotion and retention dif-ferences, we find that there is almost no difference between blackand white male officers in the likelihood of getting from O-1 to O-4,the first field-grade rank. Of those who are commissioned, only 1percent fewer black men make O-4—a difference that is not statisti-cally significant. The same cannot be said for other minority men; 6percent fewer of them than white men reach O-4 rank.

In the early stages of their careers (before O-4), white women are lesslikely to stay during retention periods and somewhat less likely thanwhite men to be promoted (Figure 7). The differences generally be-come positive at the O-4 point. There are two possible explanationsfor the contrast between the pre–O-4 and post–O-4 results for whitewomen. First, there could have been a shift in the behavior and per-formance of women who entered in the cohorts after 1980 sincethese cohorts are included only in the pre–O-4 analysis. Second, itmay be the case that women who do not choose to actively pursue amilitary career fall out before the field grades and the women whoremain are more committed. Since we were unable to detect any dif-

______________ 9The lower retention rates estimated for other minorities at O-3 are statistically sig-nificant only in the uncontrolled model.

Page 18: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

42 Minority and Gender Differences in Officer Career Progression

RAND MR1184–7

Retention Windows

–10–8–6–4–2

02468

1012

O–2 O–3 O–4 O–5 O–6

Promotion Windows

–10–8–6–4–2

02468

1012

O–1 to O–2 O–2 to O–3 O–3 to O–4 O–4 to O–5 O–5 to O–6

Per

cent

age

Per

cent

age

–3%*–4%

(Insufficientdata)%

3%

3%

–7%*

–2%*0%

5%*

–3%*

*Differences that are significant at .05 level.

Figure 7—Differences in Completion Rates for WhiteFemale vs. White Male Officers

ferences in the cohorts that would explain this result, we are left withthe second explanation.

Career progression for minority women generally resembles careerprogression for minority men (Figure 8), with a higher retention ratefor black women at the rank of O-3 and lower completion rates in allpromotion windows through O-4. The sample of minority womenabove the O-4 promotion point is increasingly small, so we were notable to obtain results for the later career windows.

Page 19: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 43

(Insufficient data)–2%–4%

5%

–1%

RAND MR1184–8

Retention Windows

–10–8–6–4–2

02468

1012

O–2 O–3 O–4 O–5

Promotion Windows

–10–8–6–4–2

02468

1012

O–2 to O–3 O–3 to O–4 O–4 to O–5 O–5 to O–6

Per

cent

age

Per

cent

age

(Insufficientdata)–3%

–3% –3% –4%–6% –6%

Black

O–ther minority females

*Differences from rate of white males that are significant at .05 level.

Figure 8—Differences in Completion Rates for Minority Femalevs. White Male Officers

Minority/Gender Differences Not Controlling for Entry Path

Source of commission, prior service, and occupation significantlyaffected the probability of progressing through most of the retentionand promotion windows (see regression results in Appendix C). Theresults we presented above control for these other factors, so we canconclude that the racial and gender differences we found are not at-tributable to racial and gender differences in commissioning source,prior service, or occupation. It is possible that there are further

Page 20: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

44 Minority and Gender Differences in Officer Career Progression

racial/gender differences in career progression that are caused by thedifferences in these entry attributes, but that these differences aremasked by the multivariate approach we used for this analysis. Tosee whether there are indirect effects due to source of commission,prior service, and occupation, we reanalyzed the data and omittedthese attributes. We will call these the “unadjusted” results.

We did not find any evidence of additional indirect effects due tothese factors. Where the estimated differences between minorities orwomen and white males change at all, the change is very modest.For example, the gaps in completion of the promotion windows forblack males are slightly larger in the unadjusted results (–5.6 vs. –4.2percentage points for the O-4 promotion window). Thus, our find-ings cannot be explained by the differences in source of commission,prior service, and occupation for minority and female officers com-pared to white male officers.

Minority and Gender Differences by Service and Cohort

In carrying out our analysis, we looked for evidence that career pro-gression for minorities and women has differed across services orcohorts. Does one service have a better or worse record than theother services? Is there evidence of a trend toward more equal pro-gression over time?

The results of our service- and cohort-specific analyses are summa-rized in Appendix C. Our results did vary across services and cohorts.However, there was no consistent pattern to the variations, a conclu-sion supported by the formal statistical tests we carried out.

Summary of Retention and Promotion Results

To summarize the racial and gender differences we found, we calcu-lated how many officers out of an entering group of 100 would be lostfrom the pipeline and whether they were lost in a retention or pro-motion window (Table 6). We focused on the results for whites andblacks because these are based on larger samples than the results forother minorities. The bottom line is that women are less likely toreach field grade than are men, but within the genders there are noracial differences in this outcome. How officers leave does vary by

Page 21: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 45

Table 6

Comparison of Losses Between Commissioning and O-4by Race and Gender

WhiteMen

BlackMen

WhiteWomen

BlackWomen

Initial Count 100 100 100 100Number who leave during retention 35 28 40 30Number who leave during promotion 28 36 30 39Number who remain to O-4 37 36 30 31

race, however. White men and white women are more likely to leavethan are black men and women during periods when their departureis likely to be voluntary. Black men and women are considerablymore likely to leave at a promotion point.

The differences are even clearer if we look at the relative likelihoodsof leaving for the two reasons, voluntary departure versus promotionfailure. We will illustrate the calculation of these relative odds usingthe example of black men versus white men. During retention win-dows before O-4, Table 6 shows that, out of an initial count of 100 inboth groups, 28 black men will leave and 35 white men will leave. Wecalculate that the likelihood of a black man leaving is 20 percent lessthan the likelihood of a white man leaving because (.28–.35)/.35 =–.20. Figure 9 plots the relative differences in retention and promo-tion for black men, white women, and black women. In simpleterms, compared to white men:

• Black men are more likely to stay in the service between promo-tions but are less likely to be promoted.

• White women are less likely to stay in the service between pro-motions and are slightly less likely to be promoted.

• Black women are more likely to stay in the service betweenpromotions but are less likely to be promoted.

When all three groups we describe here are compared, black men areleast likely to choose to leave the service and white women are mostlikely to choose to leave. All three groups are less likely to be

Page 22: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

46 Minority and Gender Differences in Officer Career Progression

RAND MR1184–9

Chances of failingpromotion relativeto white men

Chances of choosingto leave relativeto white men

Black men

29%

39%

7%

–20%

–14%

14%

Black womenWhite women

Figure 9—Retention vs. Promotion of Black and Female OfficersThrough O-4, Relative to White Male Officers

promoted than are white men, but this discrepancy is smaller forwhite women than it is for black men and women.

Our promotion results are consistent with other studies conducted inrecent years of officer promotion in the Navy and Marine Corps.Mehay (1995) has found that black naval surface officers and Marineofficers were less likely to be selected for O-4 promotion between1985 and 1990. North et al. (1995) have estimated that, between 1987and 1993, black Marine officers scored lower at TBS, were less likelyto be augmented to a regular commission, and were less likely to beselected for promotion to O-3 and O-4. Hispanic and female officersdid as well as white officers on most measures. An older study ofNaval Academy graduates (Bowman, 1990) has concluded that blackgraduates from 1976 to 1980 were more likely than white graduatesto stay beyond their initial obligation and less likely to rate in the tophalf of all graduates on their early performance evaluations.Therefore, in turning to our interviews and focus groups, we weremost interested in obtaining officers’ opinions about why black men

Page 23: Chapter Three DIFFERENCES IN RETENTION AND PROMOTION …€¦ · There has been little change in the diversity of the Army’s officer ac-cessions, if we disregard 1980, which was

Differences in Retention and Promotion for Minority and Female Line Officers 47

and women are less likely to complete promotion windows andwhite women are less likely to complete retention windows than arewhite men. Our estimates of promotion and retention rates adjustfor the effects of commissioning source, prior service, and militaryoccupation as well as service and cohort. These were the factorsmeasured in our data. The other research cited above has used moredetailed service data and found that other factors also are important.The most extensive set of factors have been evaluated by Mehay; hehas found that officers who graduated from more competitive insti-tutions of higher learning, majored in technical subjects, and hadhigher grades received more positive performance evaluations thatled directly to higher promotion rates to lieutenant commander(O-4). Although fewer minority officers than white officers camefrom competitive colleges or had good grades, controlling for thesedifferences did not explain the lower evaluation ratings andpromotion rates for black officers.