Waivered recruits: An evaluation of their performance and attrition risk

Waivered recruits: An evaluation of

their performance and attrition risk

Lauren Malone • Neil Carey

with contributions by

Yevgeniya Pinelis • Dave Gregory

CRM D0023955.A4/Final

March 2011

Public release version

This document represents the best opinion of CNA at the time of issue.

It does not necessarily represent the opinion of the Department of the Navy.

Approved for public release. Specific authority: N00014-05-D-0500.

Copies of this document can be obtained through the Defense Technical Information Center at www.dtic.mil

or contact CNA Document Control and Distribution Section at 703-824-2123.

This work was created in the performance of Federal Government Contract Number N00014-05-D-0500. Any copyright

in this work is subject to the Government's Unlimited Rights license as defined in DFARS 252.227-7013 and/or DFARS

252.227-7014.

Approved for distribution: January 2011

Anita Hattiangadi

Marine Corps Manpower Team

Resource Analysis Division

Photo credit line: Company C recruits conquer Basilone's Challenge, one of the last obstacles they must

overcome before calling themselves U.S. Marines. Photographer’s Name: Pfc. Michael Ito

Date Shot: 8/25/2010

Contents

Executive summary . . . . . . . . . . . . . . . . . . . . . . . . 1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Background. . . . . . . . . . . . . . . . . . . . . . . . . . 5

This study . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Existing literature . . . . . . . . . . . . . . . . . . . . . . . . . 9

Differences in Services’ waiver regulations . . . . . . . . . 9

Historical trends in the number of waivers . . . . . . . . . 10

Performance of waivered recruits . . . . . . . . . . . . . . 11

Army . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Navy . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Marine Corps . . . . . . . . . . . . . . . . . . . . . . 14

A DoD-wide study . . . . . . . . . . . . . . . . . . . . 16

Implications of previous research for this study. . . . 17

Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Characterizing waivered and nonwaivered recruits . . . . . . . 23

Determinants of attrition . . . . . . . . . . . . . . . . . . . . . 29

Average attrition rates: how do waivered groups

compare? . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Are recruits with multiple waivers riskier accessions? . . . 34

Does attrition vary by waiver type?. . . . . . . . . . . . . . 36

Army . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Navy . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Marine Corps . . . . . . . . . . . . . . . . . . . . . . 40

Air Force. . . . . . . . . . . . . . . . . . . . . . . . . 41

How might the Services reduce attrition probabilities

for waivered recruits? . . . . . . . . . . . . . . . . . . . . 43

Army . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Navy . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Marine Corps . . . . . . . . . . . . . . . . . . . . . . 53

Air Force. . . . . . . . . . . . . . . . . . . . . . . . . 57

Time to E5 promotion: waivered recruits vs. their

nonwaivered counterparts . . . . . . . . . . . . . . . . . . . . 63

Policy recommendations . . . . . . . . . . . . . . . . . . . . . 67

Identify objectives and determine the most

relevant risks . . . . . . . . . . . . . . . . . . . . . . . . 67

Minimize risks . . . . . . . . . . . . . . . . . . . . . . . . 70

Eliminate the use of DAT waivers . . . . . . . . . . . 70

Consider providing commanders of recently

accessed Servicemembers with waiver

information . . . . . . . . . . . . . . . . . . . . . . 71

More carefully screen those with “risky”

waiver combinations. . . . . . . . . . . . . . . . . . 71

Regularly identify those waiver populations

with additional screening potential. . . . . . . . . . 72

A way to reduce the size of the waivered population . . . . 74

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Appendix A: Demographic characteristics of

waivered and nonwaivered populations, by Service . . . . . 81

Appendix B: Geographic distributions of waivered and

nonwaivered recruits . . . . . . . . . . . . . . . . . . . . . . . 83

Appendix C: Analysis of those with multiple waivers . . . . . . 89

Comparing populations . . . . . . . . . . . . . . . . . . . 89

Do particular waiver combinations result in

higher risk? . . . . . . . . . . . . . . . . . . . . . . . . . 91

Appendix D: “Fast to E5” occupations and results . . . . . . . 97

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

List of figures . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

List of tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

Executive summary

In this report, we examine the characteristics, performance, and attri-

tion risk of waivered recruits in each of the Services and compare

them with their nonwaivered counterparts. Our goal is to first identify

what, if any, risks the accession of waivered recruits creates for the Ser-

vices and then identify ways to mitigate these risks. The Services have

historically relied on enlistment waivers to increase the size of their

potential applicant pool because there are deserving people with a

desire to serve who have made mistakes or do not meet all of the stan-

dard qualifications.

The current recruiting climate is good—unemployment rates are

high and overall propensity to serve appears to be rising—causing

some to question the need and rationale for waivers. The Department

of Defense (DoD), however, believes that waivers should continue to

be a part of the recruiting and accession processes; DoD’s concern is

to minimize the risks taken by the Services in doing so. In this light,

OSD-Accession Policy asked CNA to evaluate whether the Services’

current use of enlistment waivers leads to recruitment of quality per-

sonnel and whether, once accessed, waivered recruits impose any

additional risks to the Services.

Our comparison of the demographic characteristics of recruits

reveals that waivered recruits, in all Services, are more likely than

their nonwaivered counterparts to be male, older, and Tier II (i.e.,

holders of nontraditional high school degrees). The waivered popu-

lation also has a greater proportion of whites (thus, a smaller propor-

tion of minorities) and is more likely to be married. When comparing

the military characteristics at accession of these two populations, we

find that waivered recruits, on average, spend less time in the delayed

entry program than their nonwaivered counterparts (highly corre-

lated with the fact that waivered recruits tend to be direct ships) and,

with the exception of the U.S. Air Force (USAF), tend to access at

lower paygrades. Finally, we find that waivered recruits are more likely

to come from the East North Central region and less likely to come

from the Pacific and Mid-Atlantic regions than their nonwaivered

counterparts. These geographic differences likely highlight the varia-

tion in recruiting difficulties that exist in different areas of the United

States. We use these demographic and geographic differences

between waivered and nonwaivered recruits in each of the Services to

inform our analysis of attrition and military performance.

Our comprehensive analysis of the attrition risk of waivered recruits

reveals that, in most cases, waivered recruits attrite at lower rates than

Tier II/III recruits, suggesting that they are, in fact, not the riskiest

accessions. We identify which waivered groups have the highest inher-

ent attrition risk, after controlling for a variety of demographic and

military characteristics. These findings tell us which waivered groups,

in each Service, have higher risk based on behavioral and unobserv-

able characteristics—characteristics that the Services have little

power to influence. For each Service, we choose a few waivered

groups in which there is potential to screen on observable character-

istics, and we identify ways in which the Services could reduce attri-

tion within these populations. All of these findings are highly Service-

specific and vary depending on whether the aim is to reduce 6-, 24-,

or 48-month attrition.

Finally, we evaluate the performance of waivered recruits relative to

nonwaivered recruits. Using time to E5 promotion as our principal

metric, we compare the prevalence of “fast” promoters in a select

number of occupational specialties for each of the Services, and then

compare this with the nonwaivered population. When separating the

populations by waiver type, those without waivers are in no cases

among the Services’ top performers, as measured by time to E5 pro-

motion. This reveals that many waivered recruits become high-quality

Servicemembers and, therefore, may not be the Services’ greatest

accession “risks.”

Overall, we find that waivered recruits are not inherently risky and are

often better performers, with lower attrition risk, than Tier II/III

recruits. There are, however, still ways in which the Services could

minimize the “riskiness” of the waivered population. For example,

each Service has a number of waiver combinations that are most likely

to lead to early attrition; additional screening or mentoring of these

recruits could potentially decrease their attrition risk.

Our analyses have allowed us to identify, within each Service, the

types of waivered recruits that impose the greatest risk, although these

findings depend on how the Services choose to define such risk.

Thus, moving forward, each Service must identify its objectives,

expectations, and most appropriate risk measures before determin-

ing how best to manage its waivered recruits.

This page intentionally left blank.

Introduction

Background

Each year, the Services require thousands of new recruits to support

the military's capabilities and long-term force health. The Services’

primary recruiting pool consists of 17- to 24-year-olds from a cross sec-

tion of society. The ideal recruit would be physically and mentally fit,

would meet height and weight standards, and would have no history

of serious legal problems or family issues. But the Services waive some

of these criteria because, among these youth, there are some who war-

rant the opportunity to serve. Roughly one in four young Americans

is too overweight to serve, and another third have other health prob-

lems that keep them from qualifying for service, such as diabetes,

asthma, or hearing impairments [1]. About 25 percent of young

adults lack a high school diploma or do not score high enough on the

Armed Services Vocational Aptitude Battery (ASVAB) to qualify for

enlistment. Roughly 10 percent of young Americans have had legal

problems, such as a felony or serious misdemeanor offenses. Others

are ineligible for medical reasons or because they are single custodial

parents.

All told, it is estimated that only a quarter of American 17- to 24-year-

olds meet military qualification standards [1]. This figure, however,

masks considerable variation by Service, gender, and ethnicity. Taking

into consideration education, test score, weight, dependents, drug,

and legal qualification standards, one study estimates that less than 15

percent of black men could meet USAF standards, but 32 percent

could meet USMC standards; the corresponding numbers for white

men would be 31 percent for the Air Force and 46 percent for the

Marine Corps, respectively [2].

To compensate for the small percentage of recruits who meet all

entry criteria, the Services allow some portion to enter with marks on

their record that would otherwise disqualify them for service. Making

these exceptions requires an enlistment waiver.

For example, Ser-

vices can waive their usual requirements in such areas as physical fit-

ness (e.g., weight waiver), family status (e.g., a dependent waiver), or

legal matters (e.g., having a misdemeanor or felony waiver). Over the

years, each Service has developed and maintained its own waiver cat-

egories and standards. The Services have adjusted the categories and

their definitions over time in response to recruiting conditions and

policy decisions.

The selective use of waivers helps the Services to meet their recruiting

missions and, it is hoped, allows them to recruit deserving young

Americans who will make good military personnel. Past research has

shown that in some cases waivered recruits do as well as, or better

than, those who enter without waivers [3, 4]. Using waivers saves the

Services money by enlarging the supply of those potentially eligible

for service, but it costs the Services when waivered recruits attrite and

don't fulfill their service obligations. Reference [5], for example, esti-

mated that the annual cost of first-term non-end-of-active-service

(non-EAS) attrition in the Marine Corps exceeded $100 million in

1993.

This study

OSD-Accession Policy asked CNA to investigate whether there are

ways that the Services can minimize the risk of early attrition associ-

ated with waivered recruits. This study comes at a time of high unem-

ployment and a rising propensity to serve in the military, making it

the first time in 35 years when all the Services have been able to meet

their recruiting missions [6]. During this good recruiting climate,

policy-makers are asking whether the considerable resources that

DoD allocates to recruiting and retention programs are really neces-

sary. Annual funding for these programs more than doubled from

2004 to 2008, increasing from $3.4 billion to $7.7 billion [7].

1. Others refer to these as moral or moral conduct waivers. We use the

terms waivers or enlistment waivers throughout this document.

If there are categories of waivered recruits that perform as well as

nonwaivered recruits, selective waiver use could compensate for

potential budget cuts in recruiting and retention programs. Further-

more, even in today’s environment, there are a few specialties that

still face shortages of available, qualified Servicemembers to fill them

(e.g., ones that require high technical aptitude or specialized lan-

guage skills). In addition, the Services cannot expect current recruit-

ing success to continue when the economic situation improves. As a

result, OSD would like answers to such long-term questions as the fol-

lowing:

• In each Service, are there accession-related characteristics asso-

ciated with lower risk of early attrition for those accessed with

enlistment waivers?

• Could changes in the way that recruits with waivers are accessed

in each Service reduce the risk of early attrition?

• Are there other personal characteristics, such as high Armed

Forces Qualification Test (AFQT) scores or Tier I status, that

are associated with lower risk of early attrition for recruits

accessed with enlistment waivers?

• Could changes in accession policies reduce the risk of early

attrition?

In the past, the Services have had fairly heterogeneous waiver criteria.

These differences have historically made it difficult to make cross-

Service comparisons of the behavior of those accessed with any par-

ticular type of enlistment waiver. DoD, however, has required that the

Services standardize the reporting of enlistment waivers to OSD.

Under the Directive Type Memorandum (DTM), which took effect in

2008, the four major categories of waivers are conduct, dependent,

drug, and medical waivers [8].

This study involves two major tasks, each of which we address in this

research memorandum (inclusive of both waiver and no-waiver data

before the 2008 DTM). Within each Service:

1. We compare the performance of recruits accessed with and

without waivers.

2. We examine whether the presence of certain accession/per-

sonal characteristics can predict success for those with enlist-

ment waivers (e.g., does being a Tier I recruit compensate for

having a misdemeanor conviction?).

Throughout this report, we compare our findings relevant to the

waivered population with those relevant to Tier II/III recruits. We

make these comparisons in an effort to illustrate that, although

public perception is often that waivered recruits are among the poor-

est performers and are the most likely to attrite, Tier II/III recruits

are in fact a “riskier” population.

In this study, we conduct only intra-Service analysis. That is, we evalu-

ate the attrition risk and performance of waivered recruits as com-

pared with their nonwaivered counterparts within each Service, but

we do not make cross-Service comparisons. This is because, before

FY09, each Service had its own criteria regarding the behaviors and

characteristics that would necessitate a waiver. In the Marine Corps,

for example, any recruit who admits to one-time marijuana use

requires a drug waiver. This is not true in any of the other Services. As

a result, the percentage of USMC accessions with drug waivers is

much higher than in the other Services. Simply comparing the popu-

lations with drug waivers in the Marine Corps and the Navy, for exam-

ple, would lead to inaccurate conclusions. In other cases, the Services

were using different waiver codes to represent the same waiver type,

increasing the possibility for data misinterpretation.

In the next section, we review the existing literature in this area and

highlight our contributions. Then, we discuss the data sources used

in our analysis. The remaining sections focus on characterizing

waivered recruits (and how they differ from their nonwaivered coun-

terparts), analyzing their attrition behavior (and how this differs by

waiver type), and comparing their likelihood to be fast promoters to

the rank of E5 with that of nonwaivered recruits. Finally, we provide

policy recommendations and conclusions.

Existing literature

Over the last 20 years, enlistment waivers have been the subject of

many studies, newspaper pieces, and journal articles. This section

provides a high-level overview of these works as they relate to three

issues: (1) differences in the Services’ waiver regulations, (2) number

of waivers granted by Service, and (3) performance of waivered

recruits. The purpose of this overview is to help inform our analyses

of data on the performance of waivered and nonwaivered recruits for

FY99 to FY08.

Differences in Services’ waiver regulations

As mentioned earlier, each Service historically had its own enlistment

waiver standards, which made it difficult to compare the waivered

populations across Services and time. A 1999 Government Account-

ability Office (GAO) study summarizes some of the explicit differ-

ences across the Services’ waiver policies (see table 1) [9].

According to the GAO, as table 1 shows, for each offense, there was

at least one Service that used criteria different from the others. The

Navy (USN) and Air Force (USAF) allowed felony waivers for recruits

having more than one felony, whereas the Army (USA) and Marine

Corps (USMC) did not. But the Army and Marine Corps were differ-

ent from each other, too. The Army, for example, required a waiver

for two or more serious (not minor) offenses, whereas the Marine

Corps required a waiver for even one such offense. In short, no two

Services had exactly the same regulations regarding enlistment

waivers. This suggests that our initial analyses, which include data

only from those years before the implementation of the DTM, must

focus on comparing waiver types within each Service, rather than

across Services.

Historical trends in the number of waivers

The percentage of waivered accessions has varied considerably from

year to year and across the Services. As reviewed in [9], in FY90, across

the four Services, the percentage of recruits issued waivers was high-

est in the Marine Corps, at 60 percent, and lowest in the Air Force, at

2 percent. By 1994, however, the Services began to converge: waivered

recruits made up 29 percent of USMC accessions, 16 percent of USN

accessions, 6 percent of USAF accessions, and 5 percent of USA acces-

Table 1. Number of offenses requiring an enlistment waiver, by Service

Offense USA USN USMC USAF

Felony 1; no waiver allowed for

more than 1.

1 or more. 1; no waiver

allowed for

more than 1.

1 or more.

Serious

mis-

demeanor

2; no waiver allowed for 5

or more.

1 or 2; no waiver

allowed for 3 or

more.

1 to 5; no

waiver allowed

for 6 or more.

1 or more.

Minor

mis-

demeanor

Category not used. 3 to 5; no waiver

allowed for 6 or

more.

Category not

used.

1 or more.

Minor

non-

traffic

3 or more; 3 convictions

for a combination of mis-

demeanors and minor

non-traffic offenses.

3 to 5; no waiver

allowed for 6 or

more.

2 to 9; no

waiver allowed

for 10 or more.

Depending on seriousness

of offense: 1 or more; 2 in

the last 3 years; or 3 or

more in a lifetime.

Serious

traffic

Category not used. Category not used. 2 or more; no

waiver for 6 or

more.

Category not used.

Minor

traffic

6 or more where fine

exceeded $100 per

offense.

Within 3 years

before enlistment,

6 or more in any

12-month period or

10 or more in total.

5 or more. Depending on seriousness

of offense: 2 in last 3

years or 3 or more in life-

time; 6 or more or 5

minor traffic and 1 minor

non-traffic-related offense

in any 1-year period in

past 3 years.

Drug 1 arrest for drug posses-

sion or use, no waivers for

sale of drugs, those with

history of drug use coded

as medical waivers.

11 or more inci-

dents of marijuana

use, prior depen-

dence on any drug,

2 or more drug-

related offenses.

1 or more uses

of marijuana or

other drugs, to

include experi-

mentation.

15 or more incidents of

marijuana use, any illegal

use of amphetamines,

other stimulants, barbitu-

rates, or steroids.

a. Sources: [9] and additional input from the Services’ recruiting commands.

sions. We might guess that the Marine Corps issued more waivers than

the other Services in the early 1990s (because of the first Gulf War)

and fewer during the military drawdown of the mid-1990s. The main

reason for these sizable differences, however, is likely that the Marine

Corps is the only Service to require a drug waiver for admission of

one-time marijuana use. The Army and Navy also decreased the per-

centage of waivers in the early 1990s. The Air Force followed an oppo-

site trend by increasing the number of waivers in the early 1990s

(from 2.0 to 6.7 percent of accessions).

These trends demonstrate that the Air Force might be quite different

from the other Services in its need for, and use of, waivered recruits.

The Air Force is the most selective of the four Services in terms of

aptitude test scores. Based on these scores, the smallest percentage of

the youth population is eligible to serve in the Air Force, in compari-

son to the other Services [2]. The civilian economy was growing fairly

rapidly during the 1990s, so the Air Force likely competed with the

civilian sector for high-aptitude recruits. The other three Services, in

contrast, were trying to shrink during the early and mid-1990s in reac-

tion to the end of the Cold War. These overall trends suggest that

each Service has its own patterns and should be analyzed separately.

Performance of waivered recruits

We briefly summarize past findings on the performance of waivered

recruits in the Army, Navy, and Marine Corps.

We then discuss the

one previous study that considers all four services. This section ends

with conclusions about what this previous work means for our current

analyses.

Army

A 2008 Naval Postgraduate School thesis by Distifeno looked at the

effects of enlistment waivers on Soldiers’ first-term attrition [10]. This

study found that the success of waivered recruits varied depending on

2. Likely because of the relatively small size of the USAF waivered popula-

tion, we found no studies evaluating the performance of waivered

recruits in the Air Force.

when success was measured. For example, Distifeno found that attri-

tion rates for waivered Soldiers were lower than for nonwaivered Sol-

diers at the beginning of the first term, but they were higher at the

end of the first term. This study also found differences in attrition

rates based on the type of offense waivered. Specifically, Distifeno

found that recruits with serious traffic and minor non-traffic-related

waivers had lower attrition rates, especially early attrition rates.

Recruits with serious non-traffic-related and felony waivers (for such

offenses as aggravated assault, breaking and entering, and carjack-

ing) seemed to be driving the general conduct waiver attrition pat-

tern—lower attrition rates at first, but higher attrition rates by the end

of the first term.

Distifeno also reported that waivered Army recruits had more disci-

plinary problems and faced more courts-martial than nonwaivered

recruits. Apparently, the pattern of pre-enlistment offenses mattered

because the study found that recruits with a pattern of minor offenses

were more likely to misbehave than those who had only a single major

offense.

Other Army studies have shown that waivered recruits can have vary-

ing levels of success. A 2008 study, for example, examined the perfor-

mance of 276,231 USA recruits from FY03 to FY06. Nearly 18,000 of

these recruits had enlistment waivers. According to the Associated

Press (in an ArmyTimes article [4]), the study found that Soldiers with

enlistment waivers had higher rates than their nonwaivered counter-

parts of desertion, misconduct, court-martial appearances, and alco-

hol rehabilitation failure:

• Desertion (4.3 vs. 3.6 percent)

• Misconduct (6.0 vs. 3.6 percent)

• Court-martial appearances (1 vs. 0.7 percent)

• Alcohol rehabilitation failure (0.3 vs. 0.1 percent).

However, the article also stated that the Soldiers with waivers:

• Were more likely to reenlist (28.5 vs. 26.8 percent)

• Promoted faster to sergeant (34.7 vs. 39 months, on average)

• Had a lower rate of dismissal for personality disorders (0.9 vs.

1.1 percent)

• Had a lower rate of dismissal for unsatisfactory performance

(0.3 vs. 0.5 percent).

These Army findings have two implications. First, it is important to

look at the attrition rates of waivered recruits at multiple points in

time because short-term and long-term results may be quite different.

Second, a thorough assessment of the performance of waivered

recruits should consider both negative performance indicators (e.g.,

desertion and misconduct rates) and positive indicators (e.g., speed

to promotion).

Navy

There have been many studies on the performance of waivered USN

recruits; we discuss two in this subsection. Reference [11] looked at

the attrition rates of USN waivered recruits from FY95 to FY96. Using

both logistic regression and classification trees in its analyses, the

study found that:

[enlistment] waivers do have a significantly higher rate of

unsuitable attrition than that of recruits without [enlist-

ment] waivers … [and] that recruits who are not high

school graduates and receive [an enlistment] waiver are the

most likely unsuitable attrition losses.

Reference [12] examined 36-month attrition for recruits in all Ser-

vices who completed the National Guard Youth ChalleNGe Program.

The authors found the following:

1. Heavy smokers, recipients of General Education Development

(GED) certificates, and high school dropouts were most likely

to attrite.

2. Although a USN waiver predicted attrition, it was not as strong

a predictor as being a heavy smoker, a GED recipient, or a high

school dropout.

3. Number of months in the delayed entry program (DEP) was a

significant predictor of 36-month attrition—shorter periods in

DEP were associated with greater attrition, whereas longer DEP

periods were associated with lower attrition.

4. The effect of having a waiver varied considerably across the Ser-

vices. The marginal effect of an Army waiver was very small (0.5

percentage point), but the effect for the Navy was considerably

larger (5.6 percentage points).

Marine Corps

In his Naval Postgraduate School thesis, Etcho analyzed the relation-

ship between demographic characteristics, enlistment waivers, and

first-term “unsuitability attrition” in the Marine Corps [13].

Etcho

defines unsuitability attrition as Marines who serve fewer than 4 years

of active duty and are separated for "failure to meet minimum behav-

ioral or performance criteria.” Examples of unsuitability attrition

include fraudulent enlistment, motivational problems, behavior dis-

order, financial irresponsibility, unsatisfactory performance, miscon-

duct, and drug use.

This study used data from the Defense Manpower Data Center

(DMDC) for USMC cohorts FY88 through FY91; each cohort ranged

from 28,000 to 34,000 recruits per year. Waivered recruits included

those who had prior involvement with law agencies (minor and seri-

ous traffic offenses, non-traffic-related offenses, serious offenses, and

felony offenses) or past drug or alcohol abuse. At the time of Etcho's

3. CNA has analyzed the performance of waivered USMC recruits several

times in the past, but results were presented in informal memoranda

and briefings (i.e., they are not citable).

4. Separation codes, however, have been found to be unreliable in gen-

eral. In the case of FY99–FY09 bootcamp separations, for example, 60

different separation codes were used, but 5 codes represented 91 per-

cent of the separations. This suggests that data entry may be inaccurate

or that it may be difficult to distinguish different separation reasons. In

addition, losses often occur for more than one reason, but only one

code can be used, and there is no predefined hierarchy. As a result, it is

unclear which code should be used, and this decision is left to the data

entry clerk [14].

study, about 60 percent of all USMC accessions were receiving enlist-

ment waivers.

The study found that the demographic groups with the highest pre-

dicted probability of unsuitability attrition for the combined USMC

cohorts were:

• Nongraduates of high school

• Recruits from mental categories IV, IIIB, or IIIA (in that

order)

• Black recruits.

Demographic characteristics associated with lower probability of

unsuitability attrition included:

• Hispanic recruits

• Male recruits

• Younger recruits.

The study also analyzed which categories of enlistment waivers were

the most (and least) likely to have unsuitability attrition. The catego-

ries of recruits most likely to attrite for unsuitability were those with:

• Felony waivers

• Drug waivers

• Less than three minor non-traffic-related waivers

• Serious offense waivers

• Alcohol waivers.

However, the study did not find a significant relationship between

other categories of enlistment waivers, such as those for three or

more minor non-traffic-related offenses.

5. These mental categories are based on AFQT scores. Specifically, cate-

gory IV recruits are those who scored between 10 and 30 on the AFQT,

while IIIB includes those scoring between 31 and 49, and IIIA includes

those scoring between 50 and 64.

Based on these findings, the study recommended that the Marine

Corps deny enlistment to anyone with a felony conviction. It also rec-

ommended denial of enlistment to anyone who would require a

waiver for serious offenses, drug/alcohol use, or less than three

minor non-traffic-related offenses. In FY08, however, implementing

all but the minor offense restriction would have reduced accessions

from 36,546 to 23,630 recruits (35 percent)—mostly because of the

Marine Corps’ use of drug waivers.

The study found that the strength of the relationship between waiver

category and attrition varies depending on the type of waiver granted.

Recruits with felony waivers were much more likely to attrite than

those with alcohol waivers, for example. This study suggests that our

work must distinguish the performance of recruits in different waiver

categories and must quantify the strength of the relationship between

waivers status and performance outcomes.

A DoD-wide study

In 2004, Putka et al. of the Human Resources Research Organization

evaluated the effects of moral waivers on attrition and “in-service devi-

ance” in each of the Services [15]. They found that 18-month attri-

tion rates were higher for those with moral conduct waivers, with

Service-by-Service variation in which types of waivered recruits were

most likely to attrite. While Drug/Alcohol Test Positive (DAT) waivers

were highly predictive of attrition behavior in all Services, adult

felony and serious non-traffic-related waivers were important predic-

tors in the Army, and waivers for marijuana use and serious non-

traffic-related waivers were significant in the Navy. The authors also

found marijuana use waivers and serious non-traffic-related waivers to

be significant predictors of 18-month attrition in the Marine Corps.

In the Air Force, however, only adult felony waivers were found to

have a positive impact on 18-month attrition. Their tests of in-service

deviance were conducted only for the Marine Corps and the Army.

Putka et al. found a higher prevalence of deviances among waivered

recruits in the Marine Corps, but no significant difference in the

Army.

Implications of previous research for this study

This brief overview of previous research shows that there are several

important factors to consider in our analyses of the effects of waivers

on military performance. Specifically, our analyses:

1. Should be primarily within Service since waiver policies were

different between Services until 2008

2. Should consider both short-term and long-term effects of

waiver status on performance because groups that perform well

in the short term do not always perform well in the long term

3. Should include both positive and negative effects of waiver sta-

tus, preferably using the same metric, because waivers have

been associated with both positive and negative performance

4. Should compare the performance of waivered recruits with

that of other relevant recruits, such as those who lack a tradi-

tional high school diploma

5. Should consider the effect of multiple waivers on performance

since previous research has not addressed this issue.

Keeping these insights in mind, the next sections describe our

research sample, methodology, and findings.

This page intentionally left blank.

Data

The data in this study are from four source files. DMDC provides

three: the accession file based on Military Entrance Processing Com-

mand (MEPCOM) transactions, the active-duty master file (quarterly

snapshots), and the active-duty transaction file. We supplement the

analysis using data from CNA’s military personnel files.

From the MEPCOM accession file, we extract data for FY99 through

FY08 for each Service. Any information about a Servicemember col-

lected at accession comes from this file, including AFQT score,

months spent in DEP, race/ethnicity, age, and waivers. We include

only non-prior-service (NPS) accessions because it is important to be

able to track Servicemembers’ careers and estimate their service

lengths. In addition, our analysis relies only on whether a recruit has

any waiver or a particular type of waiver, not a count of the number of

waivers he/she has.

Figure 1 illustrates the percentage of waivered

recruits accessed with any type of waiver in each Service. Throughout

the sample period, we see that the percentage of waivered recruits

generally fell in the Marine Corps, the Navy, and the Air Force, while

increasing in the Army.

Figure 2 shows intra-Service time trends by waiver type. We find sig-

nificant variation within each branch. The increase in USA waivered

accessions, for example, can be explained mainly by the increased

prevalence of serious and physical waivers. Figure 2 also shows that

the higher percentage of waivered recruits in the Marine Corps is

mainly a result of its stricter drug policy. Roughly 30 percent of USMC

accessions have required a drug/alcohol waiver over time.

6. The MEPCOM file contains separate DEP and accession waivers, and a

maximum of three data fields for each. MEPCOM advised us that, in the

case of the multiple waivers, some of the waivers that were recorded in

a recruit’s DEP section were carried over to his/her accession waiver

section—in essence, resulting in double-counting for that recruit.

The remaining data components for the study are the active-duty

master file, the active-duty transaction file, and CNA’s military per-

sonnel files. The active-duty master file tracks NPS accessions quar-

terly through time, so we can identify if (and when) a recruit attrited.

The active-duty transaction file provides loss reasons, which were used

to help define attrition. Not all losses are “bad losses” and thus should

not be considered attrites. We did not want to categorize, for exam-

ple, those enlisted members who are “lost” because of a transition to

officer training programs or because they reached their end of active

obligated service (EAOS).

Figure 1. Percentage of waivered recruits by Service: FY99 through FY08

10%

20%

30%

40%

50%

60%

70%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Accession FY

Percentage of recruits with at least one

waiver

USA USAF USMC USN

Figure 2. Trends in waivered accessions FY99 through FY08, by waiver type

a. At the Navy’s request, education waivers are reported separately for that Service (they are included in the “other waiver” category

for all other Services).

Army

10%

12%

14%

16%

18%

20%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of accession

Percentage of recruits with a

waiver

Air Force

10%

12%

14%

16%

18%

20%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of acce s sion

Percentage of recruit s with a

waiver

Marine Corps

10%

15%

20%

25%

30%

35%

40%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of access ion

Percen tag e of recruit s with a

waiver

Navy

10%

12%

14%

16%

18%

20%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of accession

Percentage of recruits with a

waiver

Physical

Ment al

Drug / Alcohol

Test Positive

Dependents

Adult Felony

Juvenile Felony

Ser ious

Edu c at i o n

Other

Army

10%

12%

14%

16%

18%

20%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of accession

Percentage of recruits with a

waiver

Air Force

10%

12%

14%

16%

18%

20%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of acce s sion

Percentage of recruit s with a

waiver

Marine Corps

10%

15%

20%

25%

30%

35%

40%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of access ion

Percen tag e of recruit s with a

waiver

Navy

10%

12%

14%

16%

18%

20%

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

FY of accession

Percentage of recruits with a

waiver

Physical

Ment al

Drug / Alcohol

Test Positive

Dependents

Adult Felony

Juvenile Felony

Ser ious

Edu c at i o n

Other

This page intentionally left blank.

Characterizing waivered and nonwaivered

recruits

In this section, we compare the average characteristics of waivered

and nonwaivered recruits in each Service. We compare how they

differ on a variety of demographic and military characteristics, includ-

ing education, marital status, age, gender, time spent in DEP, and

paygrade at accession. In addition, we compare the geographic areas

from which these populations originate. We find a variety of signifi-

cant differences in the characteristics of waivered and nonwaivered

recruits—differences that we ultimately consider when we compare

performance metrics and attrition behavior across these groups.

We present differences across the groups that have any waiver vice no

waiver, and we do not differentiate by waiver type. As expected, the

groups are systematically different; these findings motivate our analy-

sis of behavioral differences in the two populations. Table 2 summa-

rizes our findings for a number of demographic characteristics.

Across Services, waivered recruits are more likely to be male, married,

older, and white than their nonwaivered counterparts.

As noted in table 2, we also find that waivered recruits are less likely

to hold traditional high school diplomas and more likely to be Tier II

recruits. They also are more likely to be Tier III in the Army, while less

likely in the Navy, and the difference was insignificant in the Air Force

and Marine Corps. Figure 3 displays the percentages of waivered and

nonwaivered recruits in each Service that are not Tier I (i.e., that are

either Tier II or Tier III).

7. Appendix A shows the values for these variables and the magnitude of

differences across the waivered and nonwaivered populations.

The percentage of Tier II/III is higher in the waivered population in

all Services, though only barely so for the Air Force. The main reason

for the large and significant difference between the percentages of

Tier II/III recruits in the waivered and nonwaivered populations in

the Navy is that the Navy grants education waivers to nearly half of its

Tier II/III recruits. Over the sample period, while the Navy granted

Table 2. Demographic characteristics of waivered vice nonwaivered recruits, by Service

Characteristic USA USAF USMC USN Waivered recruits are

Gender More likely

male

More likely

male

More likely

male

More likely

male

More likely male

Age Older Older Older Older Older

Ethnicity More likely

white

More likely

white

More likely

white

More likely

white

More likely white

Marital status More likely

married

More likely

married

More likely

married

More likely

married

More likely married

High school degree

graduate (HSDG)

Less likely Less likely Less likely Less likely Less likely

Nontraditional HSDG

More likely More likely More likely More likely More likely

Tier II More likely More likely More likely More likely More likely

Tier III More likely No diff. No diff. Less likely Inconclusive

a. Nontraditional HSDGs have a homeschool or adult education diploma, or have completed 1 semester of college.

Figure 3. Percentages of waivered and nonwaivered Tier II/III recruits, by Service (FY99–FY08)

Army Navy Marine

Corps

Air Force

Percentage

Waivered

Non-waivered

roughly 15,000 education waivers, the corresponding numbers in the

Army, Marine Corps, and Air Force were 14, 445, and 13, respectively.

Service-specific trends in the percentage of yearly accessions who are

Tier II/III are presented in figure 4. In all years, Tier II/III accessions

were highest in the Army, followed by the Navy, Marine Corps, and

Air Force.

We also find, as shown in figure 5, that waivered recruits, on average,

spend less time in DEP than their nonwaivered counterparts. This dif-

ference, which is consistent and significant in all of the Services, is

likely correlated with the fact that waivered recruits are more likely to

be shipped directly to basic training. That is, in all Services, the per-

centage of waivered recruits that access during the JJAS (June, July,

August, September) trimester is smaller than the corresponding per-

centage in the nonwaivered population. They are less likely to be

recruited as far in advance as traditional high school diploma gradu-

Figure 4. Percentages of Tier II/III recruits, by Service and accession FY (FY99–FY08)

a. The significant decline in the percentage of Tier II/III recruits in the Army from 1999-2000 is as reported in the

Enlisted Master File. We note that the precipitous drop seems unlikely and may indicate data entry errors. Results

specific to FY99 for the Army should thus be interpreted with caution.

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Accession FY

Percentage

Army

Nav y

Marine Corps

Air Force

ates and, therefore, spend significantly less time in DEP. This also

may suggest that, in all Services, if ship dates are approaching and

incoming cohorts are still insufficient, there is more flexibility and an

increasing number of those who would require a waiver are accepted.

Finally, we examine whether waivered recruits are more or less likely

than nonwaivered recruits to come from particular regions of the

country.

We find, overall, that waivered recruits are more likely to

come from the East North Central region and less likely to come from

the Pacific and Mid-Atlantic regions than their nonwaivered counter-

parts. These findings are consistent across all Services. In addition,

for all Services except the Marine Corps, waivered recruits are more

likely to come from the West South Central, West North Central, and

Mountain regions. With the exception of the Marine Corps, waivered

recruits also are underrepresented in the South Atlantic and East

South Central regions. As evidenced by these trends, recruiting is

Figure 5. Average number of months spent in DEP (FY99–FY08)

8. Appendix B contains our methodology and detailed findings.

2.9

5.2

5.6

4.9

1.9

3.9

4.2

4.0

USA USN USMC USA F

Months in DEP

No w aiv er

Waiver

more difficult in some areas of the country than others. As a result,

recruiters in some regions find it necessary to make more “exceptions

to policy” via waivers to meet their recruiting missions.

These findings suggest a variety of theories as to how the performance

of waivered and nonwaivered recruits may differ, which will be

explored in this study. On one hand, the fact that waivered recruits

tend to be older and married, for example, suggests that they may

have more stability in their personal lives, potentially allowing them

to devote more energy to their careers and, hence, be less likely to

attrite. On the other hand, they tend to have nontraditional high

school diplomas, a characteristic that is correlated with poorer per-

formance and higher attrition. In addition, waivered recruits’ higher

likelihood to access at the rank of E1 in the Army and Marine Corps

indicates that they are not among the highest performers at boot-

camp (see appendix A).

Similarly, the fact that waivered recruits

spend less time in DEP suggests that they may have higher attrition

risk.

The fact that these groups have differing demographic, geographic,

and military characteristics suggests that they also may differ in unob-

servable ways. Taste for long-term service or career stability, political

affiliation, and overall personal motivation are but a few examples.

Such differences would likely cause members of these groups to

behave differently. Knowledge of how these groups differ is used to

inform the analysis in the remaining sections, where we estimate attri-

tion probability and performance metrics for both the waivered and

nonwaivered populations.

9. Those recruits with stellar bootcamp performance are often accessed at

the rank of E2. In addition, some in the nonwaivered population are

promised the E2 rank prior to arrival at bootcamp.

10. For a discussion of the relationship between time in DEP and attrition,

see [16].

This page intentionally left blank.

Determinants of attrition

In this section, we compare the attrition behavior of waivered and

nonwaivered recruits. Early attrition—defined as early termination of

a Servicemember’s contract—occurs for a variety of reasons, includ-

ing involuntary discharge, early retirement, and involuntary separa-

tion. As discussed earlier, we do not consider a Servicemember to

have attrited if his or her loss was for a “good” reason, such as transi-

tion to the officer corps, reaching EAOS, or a reduction in force.

The Services view early attrition as problematic and costly because

they fail to receive the full return on their training investments and,

in a time of high demand, must replace attriting Servicemembers. In

this section, we evaluate how the attrition behavior of waivered

recruits differs from that of their nonwaivered counterparts, and we

explore feasible methods for narrowing any gap. We begin by com-

paring average attrition rates across the waivered and nonwaivered

populations. We then present findings on which waiver types are the

most likely to attrite and whether, after controlling for other observ-

able characteristics, recruits with certain waivers are still significantly

more likely to attrite. Finally, we determine the predictors of attrition

given a particular waiver type. This will allow us to advise the Services

as to how to minimize attrition risk within their waivered populations.

Average attrition rates: how do waivered groups compare?

Here, we evaluate the validity of the assumption that waivered recruits

attrite at higher rates than their nonwaivered counterparts. After con-

trolling for demographics and other characteristics, we calculate—at

various points in time—the marginal effect of particular waivers on

attrition. This lets us evaluate whether waivered groups share

unobservable characteristics that affect their attrition risk. We also

11. While a reduction in force may not be considered a “good” thing, it is

certainly legitimate because it is beyond the Servicemember’s control.

estimate the expected probability of attrition for each waivered

group.

Figures 6 through 9 show average attrition rates at various points in

time (from 6 to 48 months) for the waiver subpopulations in each Ser-

vice.

In all Services, and for all waiver groups (including those with-

out waivers), attrition rates increase with time. That is, the percentage

of those who attrite by 48 months is greater than the percentage of

those who attrite by 36 months, which is greater than the percentage

of those who attrite by 24 months, and so on, down to those who

attrite by 3 months (while in bootcamp). This is to be expected; the

likelihood of a Servicemember attriting increases with time in service.

There are differences, however, among the Services, as to which

waivered groups tend to be the most and least likely to attrite. In the

Army, for example, those with a drug/alcohol, DAT, or aptitude

waiver (for recruits with insufficient ASVAB scores) are the most

likely to attrite. Those with adult felony waivers have the lowest attri-

tion, followed by those without waivers (see figure 6). In the Navy,

attrition is highest for DAT and education waivers; attrition is lowest

for those with no waiver, followed by physical and adult felony waivers

(see figure 7). In the Air Force, those with aptitude waivers tend to

have the highest attrition, whereas those with a physical waiver,

dependents waiver, or no waiver are the least likely to attrite (see

figure 8).

Finally, in the Marine Corps, those with DAT, aptitude, or adult

felony waivers are the highest attriters (see figure 9). Those accessed

without waivers are the lowest attriters at all points in time. Note that

recruits accessed without waivers do not necessarily have the lowest

attrition risk in each Service. In addition, Tier II and III recruits, in

most cases, have average attrition rates as high as, or higher than,

those in the waivered population. There are a few exceptions. Those

accessed with DAT waivers, for example, are more likely than Tier II

or III recruits to attrite by 48 months in the Army and Navy, and by 6

months in the Air Force. In addition, those with drug/alcohol waivers

12. For the calculation of 48-month attrition, we include only those whose

contracts were for at least 4 years.

Figure 6. Attrition by waiver category: Army

Figure 7. Attrition by waiver category: Navy

a. Education waivers are included in our analysis for the Navy only, at its request. This waiver was not issued suffi-

ciently in the other Services to warrant its inclusion.

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III with no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

waiver types

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III with no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III with no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

waiver types

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

Education

No waiver

Tier II/III

Tier II/III no waiver

Tier II/III with waiver

6 Months

24 M onths

48 M onths

waiver types

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

Education

No waiver

Tier II/III

Tier II/III no waiver

Tier II/III with waiver

6 Months

24 M onths

48 M onths

waiver types

Figure 8. Attrition by waiver category: Air Force

Figure 9. Attrition by waiver category: Marine Corps

a. As of mid-FY09, the USMC is no longer accepting recruits who require a DAT waiver.

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

waiver types

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

waiver types

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

waiver types

10%

20%

30%

40%

50%

60%

Physical

Aptitude

Drug / Alcohol

DA test positive

Dependents

Adult felony

Juvenile felony

Serious

Other

No waiver

Tier II/III

Tier II/III no waiver

Tier II/III with waiver

6 Months

24 Months

48 Months

waiver types

in the Army have higher 24- and 48-month attrition rates than those

in the Tier II/III population.

Although these attrition differences across waivered groups may pro-

vide insights on quality differences of recruits within each Service,

these trends should not be compared across Services because of sig-

nificant variations in Service waiver policies. It cannot, for example,

be concluded that recruits with a drug/alcohol waiver in the Army are

of worse quality than those with a drug/alcohol waiver in the Marine

Corps. Although tempting to jump to this conclusion (since 48-

month attrition rates for Servicemembers with drug/alcohol waivers

are greater than 50 percent in the Army but 25 percent in the Marine

Corps), this assumes that a recruit accessed into the Marine Corps

with a drug/alcohol waiver would require the same waiver in the

Army. Because of differential waiver policies, this is not the case.

Figure 10 tells a similar story to that of figures 6 through 9. It displays

the difference in average attrition probabilities for the waivered and

nonwaivered populations in each Service. (These values are calcu-

lated as the mean 6-, 24-, and 48-month attrition rates for those with

any waiver minus the corresponding attrition rate for those without,

regardless of the number of waivers or waiver type.)

Figure 10. Difference in attrition probabilities (waivered minus nonwaivered)

a. All differences displayed are significant, except Air Force 6-month attrition and Army 24-month attrition.

-2%

10%

6-month

attrition

24-month

attrition

48-month

attrition

Army

Navy

Marine Corps

Air Force

We find that waivered recruits in the Army have lower attrition rates

at 6 months, but higher at 48 months, than their nonwaivered coun-

terparts. Conversely, waivered recruits attrite more frequently at all

time intervals in both the Navy and Marine Corps. Finally, Air Force

waivered recruits have higher 24- and 48-month attrition rates, but

there is no significant difference at 6 months. These attrition rate dif-

ferences will be explored in greater detail later, when we evaluate

whether variation in attrition behavior can be explained entirely by

individual characteristics (such as gender, race/ethnicity, paygrade at

accession, and AFQT score) or is the result of other observable or

unobservable characteristics.

Are recruits with multiple waivers riskier accessions?

The attrition rates presented thus far focus exclusively on whether a

Servicemember has any waiver and on whether he or she has a partic-

ular type of waiver, without making any distinction between those who

have only one waiver and those who have multiple waivers. We also

evaluate whether those with multiple waivers are more likely to be

either a greater attrition risk or a poorer performer than those with

only one waiver. We conduct both an aggregate analysis—comparing

the average attrition rates and E5 promotion rates of those accessed

with one, two, or three or more waivers—and a more detailed analy-

sis, in which we compare these same metrics across the ten most

common waiver combinations within each Service.

In sum, there are a few waiver combinations (highlighted in italics in

what follows) that we find to be particularly risky in each Service, and

some stand out as not imposing that much additional risk:

13. Appendix C contains a detailed description of our methodology and

findings. As previously mentioned, MEPCOM acknowledges that some

waivers may be double-counted (as both a DEP waiver and an accession

waiver). Because of this, we may have counted some recruits who have

only one waiver as having two. In an effort to avoid these potential

errors, we focus on waiver pairs. If a waiver was double-counted, the

same waiver type should appear twice on the recruit’s record. By focus-

ing on recruits with more than one waiver type, we avoid the potential

for data error.

• Army

— Medical and serious: lower 24- and 48-month attrition, higher

percentage are fast to E5

— DAT and serious: less likely to attrite at 6 months, more likely

at 48

• Navy

— Serious and education: higher 24- and 48-month attrition

— Other and education: higher 6-, 24-, and 48-month attrition

— Dependents and serious: more likely to be fast promoters

• Marine Corps

— Drug/alcohol and medical: higher attrition at all intervals and

less likely to promote fast

— Drug/alcohol and DAT: higher attrition at all intervals

— Medical and serious: higher attrition at all intervals

— Dependents and drug/alcohol: higher attrition at all intervals,

but more likely to promote fast

• Air Force

— Medical and aptitude: higher attrition at all intervals and less

likely to promote fast to E5

— Adult felony and serious: more likely to be fast promoters, but

no significant attrition difference.

The Services should keep this information in mind when determin-

ing which waiver combinations to prioritize (in terms of accessions)

and which should require additional screening.

14. A detailed discussion of our methodology for determining whether a

Servicemember promoted “fast to E5” is discussed in appendix D and

documented in [17].

Does attrition vary by waiver type?

In this subsection, we estimate the effect of being in a certain waiver

group on the probability of attrition (at 6, 24, and 48 months), after

controlling for a variety of other factors that are known to determine

a recruit’s attrition risk. Specifically, we control for a variety of demo-

graphic and military characteristics, including geographic region of

origin, age, race/ethnicity, marital status, number of children, AFQT

score, months spent in DEP, education, gender, and FY of accession.

We also control for whether a recruit has more than one waiver since

we found that those with multiple waivers do behave differently than

those with only one waiver. We then estimate the effect of a recruit

being in a particular waiver group (or having any waiver) on attrition

probabilities, taking all of these factors into consideration.

Army

Figures 11 through 14 present the independent effects of particular

waiver types. The marginal effects presented are the effect from being

in a particular waiver group (or having a particular characteristic) on

the probability of attrition, all else equal. The ultimate question this

estimation strategy answers is: after controlling for all of these observ-

able characteristics, are members of certain waiver groups still more

likely than their nonwaivered counterparts to attrite? That is, do

members of a particular waiver group share unobservable, behavioral

characteristics that influence their attrition probability?

Figure 11 presents Army findings. The effect of waiver status on short-

term, or 6-month, attrition is—in most cases—negative. Soldiers

accessed with an adult felony waiver are, all else equal, 2.4 percentage

points less likely to attrite by 6 months than their nonwaivered coun-

terparts. Similarly, those accessed with a juvenile felony, serious,

dependents, drug/alcohol, or DAT waiver are all significantly less

likely to attrite by 6 months than those in the nonwaivered popula-

tion. Conversely, those with aptitude or physical waivers have higher

short-term attrition rates, all else equal. The most significant contri-

butions of waiver status on attrition probabilities, however, are for the

24- or 48-month attrition rates of those with a dependents waiver, and

for the 48-month attrition rates of those with a DAT or drug/alcohol

waiver. Although those with a dependents waiver are significantly less

likely than their nonwaivered counterparts to attrite at all time inter-

vals, those with a DAT or drug/alcohol waiver are significantly more

likely to attrite by 48 months. For example, all else equal, Soldiers

accessed with a DAT waiver are 12 percentage points more likely to

attrite by 48 months than the nonwaivered. In addition to the mar-

ginal effects of waiver type, these figures also present the effects of

being a Tier II/III recruit (relative to Tier I) and accessing in the

October–January (ONDJ) or February–May (FMAM) trimesters (rel-

ative to JJAS). These are included simply as a comparison, to illustrate

how the marginal effect of waiver status compares with that of other

variables related to attrition. In the Army, the effects of having a

dependents, drug/alcohol, or DAT waiver on 48-month attrition are

significantly greater than these other effects. These findings suggest

that, on average, Army recruits with a dependents waiver have an

inherently low attrition risk, whereas those accessed with a DAT or

drug/alcohol waiver have an inherently high attrition risk.

Note that the findings in figures 6 and 11, although different, are not

inconsistent. These two approaches answer somewhat different ques-

tions. The fact, for example, that the independent effect of a having a

dependents waiver is negative and significant (suggesting that, all else

equal, those in this waiver group have a lower attrition probability) is

not inconsistent with the fact that the average attrition probabilities

for this group (refer back to figure 6) are slightly higher than those

in the nonwaivered population. This simply means that the main

attrition drivers in that population are the observable characteristics

we control for, not unobservable, behavioral characteristics associated

with waiver status. This suggests that this (as well as other groups with

these differential effects) is potentially an area for improved screen-

ing. We will discuss this further in a later section.

Figure 11. Marginal effects of waivers on 6-, 24-, and 48-month attrition rates: Army

a. These marginal effects are the resulting percentage-point changes in the probability of attrition from having a par-

ticular characteristic, all else equal. Marginal effects for each waiver type are the independent effect of that waiver

type relative to accessing without a waiver. Similarly, ONDJ and FMAM marginal effects represent the effect from

accessing in these trimesters relative to JJAS, and the Tier II/III effect is that relative to Tier I. Hatch-marked bars

denote statistical insignificance. All other findings are significant at the 5-percent level or better.

-10 -5 0 5 10 15

ONDJ

FMAM

Tier II/III

Other

Physical

Aptitude

DAT

Drug/Alcohol

Dependents

Serious

Juvenile Felony

Adult Felony

Percentage point impact

48 month

24 month

6 month

Less likely to attrite More likely to attrite

Navy

In the Navy, the marginal effect of a particular waiver type on attrition

probabilities, all else equal, is almost always positive, as revealed in

figure 12. The only negative effects are for 6-month attrition, and the

size of these effects is small, generally 1 percentage point or less. The

most sizable effects of waiver status on attrition occur for those with a

serious or DAT waiver. Those with serious waivers, for example, are

4.4 and 7.2 percentage points more likely to attrite by 24 and 48

months, respectively, than their nonwaivered counterparts. The inde-

pendent effect of a DAT waiver on these medium- and long-term

attrition measures is 9.5 and 15 percentage points, respectively. We

conclude, therefore, that those accessed with DAT waivers have inher-

ently high attrition risk. This also is revealed by the fact that the effect

of being in this waiver category on 24- and 48-month attrition rates is

equal to or greater than the effect of being a Tier II/III recruit.

Figure 12. Marginal effects of waivers on 6-, 24-, and 48-month attrition rates: Navy

a. These marginal effects are the resulting percentage-point changes in the probability of attrition from having a par-

ticular characteristic, all else equal. Marginal effects for each waiver type are the independent effect of that waiver

type relative to accessing without a waiver. Similarly, ONDJ and FMAM marginal effects represent the effect from

accessing in these trimesters relative to JJAS, and the Tier II/III effect is that relative to Tier I. Hatch-marked bars

denote statistical insignificance. All other findings are significant at the 5-percent level or better.

-4-20246810121416

ONDJ

FMAM

Tier II/III

Other

Physical

Aptitude

Education

DAT

Drug/Alcohol

Dependents

Serious

Juvenile Felony

Adult Felony

Percentage point impact

48 month

24 month

6 month

Within the Navy, average attrition rates by waiver type were greatest

for those with education or DAT waivers, as illustrated in figure 7.

These rates are comparable to those of the Tier II/III population

(regardless of waiver status). This suggests that, once accounting for

both observable and unobservable (associated with waiver type) char-

acteristics, those accessed into the Navy with either education or DAT

waivers pose the greatest attrition risk.

Marine Corps

We find similar results within the Marine Corps, presented in figure

13. As with the Navy, DAT waivers have the greatest independent effect

on attrition probabilities. Those accessed with a DAT waiver, for

example, are 5 and 8.4 percentage points more likely to attrite by 24

and 48 months, respectively.

They also have the highest average

attrition rate of all waiver types, as was displayed in figure 9. The aver-

age attrition rates of those accessed with DAT waivers are comparable

to those of the Tier II/III population, suggesting that they are an

equally high attrition risk. With the exclusion of those requiring DAT

waivers from the Marine Corps accession pool, the overall attrition

risk of the waivered population should decline, as the independent

effect of all other waiver types is much smaller, at less than 5 percent.

In addition, there are no other waiver groups whose average attrition

15. As of mid-FY09, the Marine Corps is no longer accessing recruits who

require a DAT waiver.

rates approach those of the Tier II/III population, as was displayed in

figure 9.

Air Force

Finally, figure 14 reveals the independent effects of waiver status on

attrition at 6, 24, and 48 months for the Air Force. The marginal

effects of waiver status in the Air Force are small—less than 5 percent

for all waiver types and all attrition rates. The effect of having a DAT

waiver is large, but insignificant, because the Air Force granted only

13 DAT waivers over the entire sample period. As figure 14 shows, the

independent effect of being a Tier II/III recruit is much greater than

any of the waiver effects: Tier II/III recruits are more likely to attrite

at 48, 24, and 6 months, by 14, 12, and 4 percentage points, respec-

Figure 13. Marginal effect of waivers on 6-, 24-, and 48-month attrition rates: Marine Corps

a. These marginal effects are the resulting percentage-point changes in the probability of attrition from having a par-

ticular characteristic, all else equal. Marginal effects for each waiver type are the independent effect of that waiver

type relative to accessing without a waiver. Similarly, ONDJ and FMAM marginal effects represent the effect from

accessing in these trimesters relative to JJAS, and the Tier II/III effect is that relative to Tier I. Hatch-marked bars

denote statistical insignificance. All other findings are significant at the 5-percent level or better.

-10-5 0 5 1015

ONDJ

FMAM

Tier II/III

Other

Physical

Aptitude

DAT

Drug/Alcohol

Dependents

Serious

Juvenile Felony

Adult Felony

Percentage point impact

48 month

24 month

6 month

Less likely to attrite More likely to attrite

tively, than a comparable Tier I recruit. Due to the small size of the

Air Force DAT population, this suggests that the Tier II and III pop-

ulations are riskiest in terms of 24- and 48-month attrition.

In this subsection, we have identified the waiver statuses that have the

greatest independent effect on attrition probability, holding all else

equal. As mentioned earlier, members of waiver groups that have rel-

atively high average attrition rates but small or insignificant marginal

effects are those that have characteristics on which the Services could

potentially screen. That is, for these groups, higher attrition probabil-

ities are driven by observable characteristics. These characteristics,

once identified, could be restricted in waivered recruits—potentially

decreasing their attrition risk. In the following subsection, we identify

Figure 14. Marginal effect of waivers on 6-, 24-, and 48-month attrition rates: Air Force

a. These marginal effects are the resulting percentage-point changes in the probability of attrition from having a par-

ticular characteristic, all else equal. Marginal effects for each waiver type are the independent effect of that waiver

type relative to accessing without a waiver. Similarly, ONDJ and FMAM marginal effects represent the effect from

accessing in these trimesters relative to JJAS, and the Tier II/III effect is that relative to Tier I. Hatch-marked bars

denote statistical insignificance. All other findings are significant at the 5-percent level or better.

-10-5 0 5 1015

ONDJ

FMAM

Tier II/III

Other

Physical

Aptitude

DAT

Drug/Alcohol

Dependents

Serious

Juvenile Felony

Adult Felony

Percentage point impact

48 month

24 month

6 month

Less likely to attrite More likely to attrite

a few such groups within each Service and provide examples of poten-

tial screening mechanisms.

How might the Services reduce attrition probabilities for

waivered recruits?

In this subsection, we evaluate whether there are ways the Services

could reduce the attrition probabilities of particular groups of

waivered recruits. Given that a recruit has a particular waiver type, we

investigate which characteristics significantly affect his or her attrition

likelihood. These are the characteristics on which the Services could

potentially screen these waivered recruits to reduce their attrition

risk. In each Service, we have tried to identify those waiver groups

where improved screening could have the greatest effect on attrition.

In the previous subsection, we presented both marginal effects and

average attrition rates for each waiver type, in each Service. Although

the former is meant to capture unobservable or behavioral character-

istics shared by those with a particular waiver type, after controlling for

characteristics we can observe, the latter incorporates both observable

and unobservable information.

If the average attrition probability is

positive and the marginal effect is either negative or very small, it sug-

gests that predicted attrition is being driven mainly be observable

characteristics. These are precisely the traits on which the Services

could screen these recruits because this is information available at

accession. Thus, these are the waiver groups where additional screen-

ing has the largest potential for reducing attrition probabilities. Table

3 identifies these groups, which will be the focus of our discussion.

We evaluate the effect of a variety of observable characteristics on attri-

tion probability at 6, 24, and 48 months, given that a recruit was

accessed with one of these waivers. Table 4 defines the variables con-

sidered in this analysis. Not all are included in every estimation within

16. Specifically, the predicted probability incorporates the product of each

characteristic and its marginal effect. We include demographic and mil-

itary characteristics (the observables) and the additional behavioral

effect from accessing with a certain waiver type (the unobservables).

each Service because, in some cases, there are no recruits within that

particular waiver group who have a certain characteristic.

Table 3. Waiver groups included in minimizing attrition risk analysis

Army Navy Marine Corps Air Force

Physical Physical Physical Physical

Dependents Adult felony Dependents Adult felony

DAT Juvenile felony Drug/alcohol Serious

Aptitude Aptitude Aptitude Aptitude

Table 4. Characteristics considered in this analysis

Variable Description

ONDJ Equals 1 if the accession occurred in October, November, December, or January;

0 otherwise

FMAM Equals 1 if the accession occurred in February, March, April, or May; 0 otherwise

CAT IV Equals 1if AFQT < 31; 0 otherwise

a. In our sample, category IV recruits are largely USA recruits. That is, of the roughly 20,000 category IV recruits

accessed into the Services between FY99 and FY08, 15,000 were accessed in the Army, and 4,000 were accessed

into the Marine Corps. The majority of the category IV Army accessions occurred between FY05 and FY08, while

the category IV Marine Corps accessions were mainly concentrated in FY07 and FY08.

CAT IIIB Equals 1 if 30 < AFQT < 50; 0 otherwise

CAT IIIA Equals 1 if 49 < AFQT < 65; 0 otherwise

CAT II Equals 1 if 64 < AFQT < 93; 0 otherwise

Long DEP Equals 1 if a recruit spent more time in DEP than the majority of those in that Service

(1 month + for Army, 3+ for Navy, 3+ for Marine Corps, 4+ for Air Force); 0 otherwise

Tier II/III Equals 1 if a Tier II or Tier III recruit; 0 otherwise

E2 Equals 1 if access as an E2 or achieve paygrade E2 within first quarter; 0 otherwise

E3 Equals 1 if access as an E3 or achieve paygrade E3 within first quarter; 0 otherwise

E4 Equals 1 if access as an E4 or achieve paygrade E4 within first quarter; 0 otherwise

Number of

children

Number of children at accession

Married Equals 1 if married at accession; 0 otherwise

Age Age at accession

Male Equals 1 if male; 0 otherwise

17. For example, no Marines had a physical waiver and accessed at either E3

or E4, so we exclude the E3 and E4 variables in this case.

Army

For the Army, we identify characteristics that have the largest effect

on attrition for those with physical, dependents, DAT, or aptitude

waivers. Figures 15 through 18 present our findings; we discuss the

determinants of attrition for each waivered population in turn.

From FY99 to FY08, there were 47,113 Army recruits with physical

waivers. Figure 15 displays the marginal effect of a number of charac-

teristics on the attrition probabilities. Positive values indicate that Sol-

diers with that characteristic were more likely to attrite by 6, 24, or 48

months; negative values correspond to a lower attrition probability.

We find that physically waivered recruits who are male, who access at

paygrades E2, E3, or E4 (relative to E1), or who spend a long time in

DEP are significantly less likely to attrite at all three intervals. As the

figure shows, being male decreases the probability that a physically

waivered recruit will attrite by 11, 16, and 23 percent for 6-, 24-, and

48-month attrition, respectively.

The effects of paygrade on attrition vary from 3 to 5 percent for E2 to

5 to 13 percent for E4. That is, the higher the paygrade at accession,

the less likely that the recruit will attrite. Spending more than 1

month in DEP also decreases attrition probability, by 5 percent or

less; however, those with AFQT scores less than 93 (AFQT categories

II through IV) and those who are Tier II or III are more likely to attrite.

The most sizable effects occur for those in AFQT categories IIIA

through IV. In addition, note that the independent attrition effects

from being a category IIIA, IIIB, or IV recruit are not substantially dif-

ferent. That is, category IVs do not have drastically higher attrition

risk than category IIIs. In sum, given a pool of recruits requiring a

physical waiver, these results suggest that the Army could reduce the

attrition risk of this population by selecting those who are male,

access in paygrades E2 to E4, have AFQT scores greater than 92 (or at

least greater than 64), and/or are Tier I recruits.

Although a recruit

with all of these characteristics would be ideal, selecting recruits with

only some of these attributes still would reduce attrition risk.

18. Accessions at the rank of E2 or above occur, for example, based on edu-

cation qualifications or for referrals that a poolee provides (provided

that these referrals sign contracts).

The number accessed with a dependents waiver is significantly

smaller (5,251 from FY99 to FY08), but still worth studying. The mar-

ginal effect of having a dependents waiver on attrition probabilities

(the unobservables effect) is negative, as we presented in figure 11. As

figure 16 illustrates, recruits with dependents waivers who are male

and who access during FMAM are significantly less likely to attrite at

all intervals. In addition, Tier II or III recruits are more likely to

attrite. The remainder of the results vary by attrition interval. Spend-

ing 3 or more months in DEP, for example, substantially decreases 48-

month attrition, but not 6- or 24-month attrition.

19. We recognize, for example, that screening recruits on gender will be a

legal impossibility.

Figure 15. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a physical waiver: Army

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-25% -20% -15% -10% -5% 0% 5% 10% 15%

Male

Age

Married

Number Children

T ie r II/III

Long DEP

CAT II

CAT IIIA

CAT IIIB

CAT IV

FMAM

ONDJ

48 Month

24 Month

6 Month

Finally, we evaluate factors that contribute to the attrition probabili-

ties for those accessed with DAT and aptitude waivers. From FY99 to

FY08, there were 9,940 Army accessions with DAT waivers. Within this

population, the probability of attriting at 6, 24, and 48 months is 18

to 26 percentage points lower for men than for women, as illustrated

in figure 17. We also find significant and sizable negative effects from

accessing as an E2, E3, or E4 (as opposed to E1). Conversely, being a

Tier II/III recruit or having an AFQT score in categories IIIA or IIIB

makes a recruit more likely to attrite in the medium and long term.

AFQT scores have no predictive power for short-term attrition. Simi-

larly, spending more than 1 month in DEP reduces only 24-month

attrition, and the effect is relatively small; the likelihood of attriting

decreases by less than 5 percentage points.

Figure 16. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a dependents waiver: Army

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-25% -20% -15% -10% -5% 0% 5% 10% 15% 20%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CA T II

CA T IIIA

CA T IIIB

CA T IV

FMA M

ONDJ

48 Month

24 Month

6 Month

Findings for the population with aptitude waivers are somewhat simi-

lar. From FY99 to FY08, there were 1,217 Army accessions with apti-

tude waivers. As illustrated in figure 18, those with AFQT scores in

categories II through IV are all significantly more likely to attrite at all

intervals—approximately 12, 29, and 40 percent more likely to attrite

than those in the highest range of AFQT scores at 6, 24, and 48

months, respectively. In addition, men with aptitude waivers are

much less likely to attrite than women. The resulting reduction in

attrition probability from being male ranges from 26 percent at 6

months to 11 percent at 48 months.

For these four Army waiver categories, we found that gender, pay-

grade at accession, time in DEP, and AFQT score are all significant

predictors of attrition probabilities, with the size and significance of

these effects varying by waiver type. In general, attrition is less likely

Figure 17. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a DAT waiver: Army

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-30% -25% -20% -15% -10% -5% 0% 5% 10% 15%

Male

Age

Married

Num ber Children

Tier II/III

Long DEP

CAT II

C AT IIIA

C AT IIIB

CAT IV

FMAM

ONDJ

48 Month

24 Month

6 Month

for men, those with higher AFQT scores, those who access at a pay-

grade higher than E1, and those who spend a “long” time in DEP.

Navy

In this subsection, we conduct a similar analysis for the Navy and

focus on those accessed with either a physical, aptitude, juvenile fel-

ony, or adult felony waiver.

In our sample, there are a total of 23,260

Navy recruits with physical waivers. The effects of their demographic

and other characteristics observable at accession on their attrition

probabilities are shown in figure 19. As noted in the figure, we find

that Tier II/III status and having an AFQT score in category IIIA or

IIIB significantly increases the probability of attrition. Similarly, those

Figure 18. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

an aptitude waiver: Army

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

20. These four groups were chosen based on the same selection criteria dis-

cussed in the previous subsection.

-40% -20% 0% 20% 40% 60%

Male

Age

Married

Number Children

Tie r II/III

Long DEP

CA T II

CA T IIIA

CA T IIIB

CA T IV

FMA M

ONDJ

48 Month

24 Month

6 Month

who spend a particularly long time in DEP, access at E2 or E3, and are

male are significantly less likely to attrite at all time intervals. This sug-

gests that AFQT scores, gender, education tier, and time in DEP

would be the most effective screens if aiming to reduce the attrition

risk of this population.

Findings are different for those with aptitude waivers (1,211 recruits

in our dataset), as shown in figure 20. Here, gender and time in DEP

are insignificant attrition predictors, and AFQT scores are significant

only when predicting 6-month attrition. In this case, those with lower

AFQT scores are more likely to attrite. Similarly, Tier II/III status and

accessing at the E3 paygrade are characteristics that only predict 24-

and 48-month attrition. Those in Tier II/III are 10 to 15 percentage

points more likely to attrite by 24 and 48 months than recruits with

aptitude waivers who are Tier I. The predictive power of accessing at

E3 is equally large at 24 and 48 months. In addition, those who access

during the ONDJ or FMAM trimesters are significantly more likely to

Figure 19. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a physical waiver: Navy

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-15% -10% -5% 0% 5% 10% 15% 20%

Male

Age

Marri ed

Number Children

Tier II/III

Long DEP

CAT II

CAT IIIA

CAT IIIB

FMAM

ONDJ

48 Month

24 Month

6 Month

attrite by 48 months than those who access with an aptitude waiver

during JJAS, while lower AFQT scores have a positive effect only on

the likelihood of 48-month attrition. In this instance, there are no

characteristics that predict attrition at all three intervals. The deter-

mination of which characteristics to use as an additional screen will

thus be highly dependent on whether the aim is to decrease the prob-

ability of attrition at 6, 24, or 48 months.

Finally, we estimate the marginal effects of observable characteristics

on attrition probabilities for those with felony waivers, either juvenile

or adult (see figures 21 and 22). Both of these populations are

small—a total of 793 accessions with a juvenile felony waiver and 851

with an adult felony waiver from FY99 to FY08. In the juvenile felon

population, we find that those who access at E2 are roughly 5 percent-

age points less likely to attrite by 6 months than those who access at

E1, and those who access at E3 are 17 percentage points less likely to

attrite by 48 months. In addition, those who spend a long time in DEP

Figure 20. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

an aptitude waiver: Navy

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-15% -10% -5% 0% 5% 10% 15%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CA T II

CA T IIIA

C A T IIIB

FMA M

ONDJ

48 Month

24 Month

6 Month

(3 or more months) are 9 percentage points less likely to attrite by 24

months than those who spend less than 3 months in DEP. Once

again, the appropriate screening mechanism will depend on which

attrition rate the Navy is trying to minimize. If the aim, for example,

is to decrease the 48-month attrition rate in this population, requir-

ing those with a juvenile felony to access as E3s would have the largest

effect based on these results.

In the adult felon population, shown in figure 22, having paygrade E3

at accession greatly reduces the likelihood of both 24- and 48-month

attrition, by roughly 15 and 18 percentage points, respectively. There

are similarly large effects from Tier II/III status: those in this group

are roughly 20 percentage points more likely to attrite by 24 or 48

months than those in Tier I. In the Navy, we find less consistent

results across waiver types than in the Army. This suggests that the

decision to screen waivered recruits according to any additional char-

Figure 21. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a juvenile felony waiver: Navy

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-20% -15% -10% -5% 0% 5% 10% 15%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CA T II

CA T IIIA

CA T IIIB

FMA M

ONDJ

48 Month

24 Month

6 Month

acteristics should be made with care, and the appropriate screening

mechanism will vary greatly depending on the waiver type in

question.

Marine Corps

We now turn to identifying the characteristics most correlated with

attrition probability for those accessed into the Marine Corps with

either a physical, dependents, drug/alcohol, or aptitude waiver.

From FY99 to FY08, there were 42,056 USMC accessions with physical

waivers. Figure 23 shows the effect of their demographic and military

characteristics on attrition likelihood. We find that men with physical

waivers are roughly 7 percentage points less likely to attrite by 24 and

48 months than women, and that having an initial paygrade of E2 (as

opposed to E1) decreases all three attrition rates by 2 to 4 percentage

points. There also are negative, although slightly smaller, effects on

attrition from spending 3 or more months in DEP. We find that attri-

tion rates are higher for those with lower AFQT scores and for those

Figure 22. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

an adult felony waiver: Navy

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-25% -20% -15% -10% -5% 0% 5% 10% 15% 20% 25% 30%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CA T II

CA T IIIA

CA T IIIB

FMAM

ONDJ

48 Month

24 Month

6 Month

in the Tier II/III education category. In this particular waiver group,

gender, Tier II/III status, and AFQT score are the characteristics with

the greatest marginal effect on attrition probability. These are thus

the characteristics that the USMC could potentially use as an addi-

tional screening for those accessed with physical waivers.

Findings are similar for recruits with dependents waivers. Namely, as

displayed in figure 24, they are less likely to attrite at all intervals if

they are male, access at E2, or spend more than 3 months in DEP. In

addition, Tier II/III recruits are significantly more likely to attrite.

AFQT scores, however, appear to have inconsistent effects because

not all categories are significant. In addition, having a lower AFQT

score only increases the probability of attriting by 24 or 48 months in

this population; there is no correlation between 6-month attrition

and AFQT. Using AFQT as an additional screen for this population

Figure 23. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a physical waiver: Marine Corps

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-10% -8% -6% -4% -2% 0% 2% 4% 6% 8% 10%

Male

Age

Mar ried

Number Children

Tier II/III

Long DEP

CA T II

CA T IIIA

CA T IIIB

CA T IV

FMAM

ONDJ

48 Month

24 Month

6 Month

should only be considered when aiming to reduce medium- and long-

term attrition. Tier II/III status, gender, time in DEP, and paygrade

at accession, however, are characteristics that have consistent effects

on attrition probabilities at 6, 24, and 48 months. If the Marine Corps

placed restrictions on these categories, or at least examined physically

waived recruits with these characteristics with extra care, the attrition

risk in this population could likely be reduced.

The same characteristics are found to be significant predictors of

attrition in the population with drug/alcohol waivers. This is a sizable

population in the Marine Corps, mainly because of its policy that

admission to one-time marijuana use necessitates drug waivers. In our

sample period, there are a total of 109,822 USMC accessions with a

drug/alcohol waiver. As shown in figure 25, men are significantly less

likely than women to attrite, as are those who access at E2 or spend 3

or more months in DEP. In addition, recruits in Tier II/III and whose

Figure 24. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a dependents waiver: Marine Corps

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-10% -5% 0% 5% 10% 15%

Male

Age

Married

Number Children

T ie r II/ III

Long DEP

CA T II

C A T IIIA

C A T IIIB

CA T IV

FMA M

ONDJ

48 Month

24 Month

6 Month

AFQT score falls in categories II through IV are more likely to attrite

at all intervals. Any aims to reduce attrition risk in the USMC popula-

tion of those with drug/alcohol waivers should focus on either

restricting the number of accessions with these characteristics or care-

fully monitoring their behavior.

The final USMC waiver population we analyze is the group with apti-

tude waivers. It is a considerably smaller population—only 1,865 in

our sample. Among them, we find that those who are male or access

as E2s are significantly less likely to attrite at all intervals, and those

who spend 3 or more months in DEP are less likely to attrite at 6 and

24 months (see figure 26). In addition, Tier II/III recruits attrite at

significantly higher rates at all three intervals; being in this education

group increases the probability of attrition by 10 to 13 percentage

points.

Figure 25. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a drug or alcohol waiver: Marine Corps

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-15% -10% -5% 0% 5% 10%

Male

Age

Married

Number Children

Tier II/ III

Long DEP

CAT II

CAT IIIA

CAT IIIB

CAT IV

FMAM

ONDJ

48 Month

24 Month

6 Month

In sum, for these four waiver categories in the Marine Corps, we have

found that gender, paygrade at accession, time in DEP, Tier II/III sta-

tus, and AFQT score are all significant predictors of attrition proba-

bilities, with the size and significance of these effects varying by waiver

type. In general, attrition is less likely for men, those with higher

AFQT scores, those who access at a paygrade higher than E1, those in

the Tier I education category, and those who spend a relatively longer

time in DEP.

Air Force

Finally, we address the characteristics that the Air Force might poten-

tially use to identify, in advance, those accessed with physical, apti-

tude, serious, or adult felony waivers who may have higher attrition

risk. From FY99 to FY08, there were 14,066 USAF accessions with

physical waivers. As shown in figure 27, those who were also Tier II/

Figure 26. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

an aptitude waiver: Marine Corps

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-20% -15% -10% -5% 0% 5% 10% 15%

Male

Age

Married

Number Children

Tie r II/III

Long DEP

CA T II

CA T IIIA

CA T IIIB

CA T IV

FMA M

ONDJ

48 Month

24 Month

6 Month

III recruits and/or had low AFQT scores were more likely to attrite.

There are sizable, positive effects from having an AFQT score under

30 (category IV) on 6- and 24-month attrition. Specifically, being in

this AFQT category increases the probability of attriting by 6 or 24

months by 30 and 55 percentage points, respectively. In addition,

men and those who access at paygrades E2 or E3 are less likely to

attrite at all intervals. Limiting the number of recruits who require

physical waivers and have an AFQT score under 30 would have signif-

icant effects on short- and medium-term attrition.

In the case of those with aptitude waivers (a significantly smaller pop-

ulation of 3,794), we find that AFQT and Tier II/III status are the

strongest attrition predictors (see figure 28). It will likely be difficult,

however, to make restrictions on the education tier and AFQT score

required for those receiving aptitude waivers since these waivers are

Figure 27. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a physical waiver: Air Force

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-20% -10% 0% 10% 20% 30% 40% 50% 60%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CAT II

C AT IIIA

C AT IIIB

CAT IV

FMAM

ONDJ

48 Month

24 Month

6 Month

given for failure to meet ASVAB requirements. Those whose ASVAB

scores are below the necessary cutoff likely also have lower overall

AFQT scores and may be in Tier II or III. Imposing AFQT and educa-

tion tier restrictions on this population would likely eliminate most

candidates. In addition to these variables, we also find that those who

access as E3s are 7 to 10 percentage points less likely to attrite at all

intervals, and men are 3 to 5 percentage points less likely to attrite in

the medium and long term. Those with more children also are less

likely to attrite at 24 and 48 months; each additional child decreases

these attrition rates by 8 and 10 percentage points, respectively.

Finally, those who spend 4 or more months in DEP are roughly 5 per-

centage points less likely to attrite by 48 months.

For those accessed into the Air Force with a serious waiver (a total of

12,458 in our sample), the most significant predictors of attrition are

Tier II/III status and having an exceptionally low AFQT score (less

Figure 28. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

an aptitude waiver: Air Force

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-20% -10% 0% 10% 20% 30% 40%

Male

Age

Married

Number Children

T ie r II/III

Long DEP

CA T II

CA T IIIA

C A T IIIB

CA T IV

FMA M

ONDJ

48 Month

24 Month

6 Month

than 30), as shown in figure 29. The latter, however, only serves as a

predictor of 24-month attrition; being in this AFQT category

increases attrition probability by 34 percentage points. Other signifi-

cant effects include accessing as an E2 or E3, being male, and spend-

ing more than 4 months in DEP, although not all are significant

predictors of attrition at all intervals, and most effects are 5 percent-

age points or less. In this population, the Air Force would likely see a

large decrease in attrition risk if restrictions were made on the

number that could be Tier II/III and/or category IV recruits.

The final USAF population we analyze is the group with adult felony

waivers (2,957 from FY99 to FY08). As figure 30 shows, there are few

significant attrition predictors in this group. This suggests that there

is little that additional screening could do to decrease attrition risk.

The only significant effects found are for gender (men are less likely

to attrite by 6 months, but more likely by 24 and 48 months) and pay-

Figure 29. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

a serious waiver: Air Force

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-10% 0% 10% 20% 30% 40%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CA T II

C A T IIIA

CA T IIIB

CA T IV

FMA M

ONDJ

48 Month

24 Month

6 Month

grade E3 at accession (lower 6- and 24-month attrition). The size of

these effects is small—all 5 percentage points or lower.

Overall, for all the waiver types discussed in this section, restricting

waivered recruits to be Tier I or to have an AFQT score greater than

30 would reduce much of the attrition risk associated with these USAF

recruits.

In the remaining sections, we compare the performance of waivered

and nonwaivered recruits. Poor performance, in addition to attrition,

is the other ‘‘risk’’ potentially imposed on the Services from accept-

ing waivered recruits.

Figure 30. Marginal effects of accession characteristics on 6-, 24-, and 48-month attrition, given

an adult felony waiver: Air Force

a. Hatch-marked bars denote statistical insignificance. All other findings are significant at the 5% level or better.

-10% -5% 0% 5% 10% 15%

Male

Age

Married

Number Children

Tier II/III

Long DEP

CA T II

C A T IIIA

CA T IIIB

FMA M

ONDJ

48 Month

24 Month

6 Month

This page intentionally left blank.

Time to E5 promotion: waivered recruits vs.

their nonwaivered counterparts

In this section, we compare the performance of waivered and non-

waivered recruits within each Service. Our principal metric is time in

service before being promoted to E5. For each Service, we have

selected a handful of occupations; for each occupation, we identify a

Servicemember as being fast to E5 if he or she is among the fastest 25

percent of promoters for the corresponding occupation-accession

year cohort.

Because promotion rates are determined not only by a Servicemem-

ber’s relative quality but also by the demand for and supply of person-

nel within that occupation at any point in time, we compare time to

E5 only among those within an occupation-accession year cohort.

Looking at a select number of occupations within each Service, we

identify each Servicemember as being either a “fast” or “not fast” pro-

moter to E5. Here, we compare the percentage of Servicemembers

within the waivered and nonwaivered populations who are fast pro-

moters. As figure 31 shows, there are differences across the Services

as to whether fast promotions to E5 are more common in the

waivered or nonwaivered populations. In both the Army and the

Navy, for example, waivered recruits are more likely to promote fast

to E5 than their nonwaivered counterparts; the converse is true in the

Marine Corps and the Air Force.

In addition, we find that there are significant differences in the like-

lihood that a Servicemember is a fast promoter depending on the

type of waiver received. We summarize these findings, for each Ser-

21. The methodology used in this section was first adopted in [17]. A more

in-depth discussion is available there.

22. See appendix D for a list of included occupations.

vice and waiver type, in table 5. In this table, a plus sign reveals that

Servicemembers in that group were more likely to promote fast to E5

than their nonwaivered counterparts; the converse is true for a minus

sign.

Figure 31. Percentage of Servicemembers who promote “fast” to E5, by Service and waiver

status (FY99-FY05 accessions)

Table 5. Summarized “fast to E5” results, by Service and waiver type

a. See figures 50 and 51 in appendix D for more information on the relative size of these effects.

Army Navy Marine Corps Air Force

Waiver + + - -

Dependents waiver + + + +

Aptitude waiver - Insignificant - -

Medical waiver Insignificant + - Insignificant

Drug/Alcohol waiver Insignificant + Insignificant Insignificant

DAT waiver Insignificant Insignificant - Insignificant

Adult felony waiver + Insignificant Insignificant Insignificant

Juvenile felony waiver Insignificant Insignificant Insignificant Insignificant

Serious waiver + + Insignificant Insignificant

Minor waiver Insignificant + + -

Other waiver + - - +

Education waiver N/A - N/A N/A

Tier II/III - - - Insignificant

USA USN USMC USA F

Service

Percentage

Waivered

Nonw aiv er ed

As illustrated, Army and Navy waivered recruits (as a whole) are more

likely to promote quickly to E5 than their nonwaivered counterparts,

suggesting that they are, in fact, better performers. Conversely,

recruits accessed with a waiver are less likely to promote fast to E5 in

the Air Force and Marine Corps. There is, once again, significant vari-

ation in performance by waiver type. In the Army, for example, those

accessed with aptitude waivers have the slowest promotion rates,

while those accessed with a dependents, adult felony, serious, or other

waiver are among the fastest, and are in fact faster promoters than

those without waivers. In the Navy, those with a dependents, medical,

drug/alcohol, serious, or minor waiver are more likely to promote

faster to E5 than their nonwaivered counterparts. In addition, there

are no waiver groups in the Navy that are less likely, on average, to be

fast promoters than the no waiver group. Although the Navy waivered

group as a whole promotes more slowly than the nonwaivered group,

when we differentiate by waiver type, there are no groups with statis-

tically significantly slower promoters. In the Air Force, those with a

dependents or other waiver are significantly faster promoters than

the no-waiver group, while those with aptitude or minor waivers pro-

mote more slowly. And finally, in the Marine Corps, those accessed

with a dependents or minor waiver tend to promote more quickly,

while those accessed with an aptitude, medical, DAT, or other waiver

are among the slower promoters. Across all Services, those with

dependents waivers appear to be among the least risky accessions (in

the sense that they are not among the poorest performers), while

those with aptitude waivers (for all Services except the Navy) are

among the slowest promoters and hence are recruits of lower quality

and higher risk.

Contrary to public perception, we do not find indisputable evidence

that waivered recruits are slower promoters and hence poorer per-

formers across the Services. With the exception of the USMC, those

without waivers do not appear among the five fastest promoting cate-

gories in each Service. When separating the populations by waiver sta-

tus, as we have done earlier, those without waivers are, in no cases,

among the Services’ top performers (as measured by time to E5 pro-

motion). This reveals that many waivered recruits become high-

quality Servicemembers and may therefore not be the Services’ great-

est accession “risks.” Tier II/III recruits are among the least likely to

be fast promoters in the Army, Navy, and Air Force. The percentage

of Tier II/III recruits in the Navy who promote quickly to E5 is lower

than that of all waiver types, indicating that Tier II/III recruits in the

Navy are of lower quality than waivered recruits. In the Army and the

Air Force, only those with an aptitude waiver are less likely to promote

quickly to E5. Finally, in the Marine Corps, Tier II/III recruits are

more likely than those in a number of waiver groups to promote fast

to E5. This suggests that those accessed, for example, with an apti-

tude, medical, DAT, or other waiver are lower quality recruits, and

therefore a greater “risk” to the Marine Corps, than the Tier II/III

recruits.

Policy recommendations

Based on our study of the characteristics, attrition behavior, and over-

all performance of waivered recruits in each of the Services, we offer

several policy recommendations. These should help the Services to

first identify and then minimize the “riskiness” of waivered recruits.

Identify objectives and determine the most relevant risks

All Services must first identify their expectations for waivered recruits:

are they an extra source of labor only to be relied on when recruits

who don’t need waivers are unavailable? That is, are waivers a policy

response to labor shortages? And, if so, what is the most important

goal for this group? Is it that they complete their first terms? Is it that

a particular number of them reenlist? By identifying such goals and

expectations, the Services can more easily identify which metrics are

most important in assessing the performance and risk of the waivered

population. If, for example, waivered recruits are not, on average,

expected to become careerists, time to E5 promotion may not be the

most appropriate measure of whether the Services achieve a sufficient

return on these investments.

In cases where there are two overall objectives (such as minimizing

48-month attrition and increasing the number of fast promoters), we

recommend that the Services conceptualize these tradeoffs using a

type of grid analysis. In figures 32 through 35, we present a technique

for comparing the “risk” associated with each waiver group. In these

examples, risk is measured by a high likelihood of 48-month attrition

and a low probability of being a fast promoter to E5. With “fast to E5”

on the x-axis and “48-month attrition” on the y-axis, risk is highest in

the upper left quadrant of the figures (this is where those waiver

groups with high attrition probability and low promotion speed are

clustered). In addition to the waiver groups, we include the Tier II/

III population.

Figure 32. Army risk tradeoff: 48-month attrition vs. fast to E5

a. The origin of this graph is not (0,0).

Figure 33. Navy risk tradeoff: 48-month attrition vs. fast to E5

a. The origin of this graph is not (0,0).

Tier II/III

Othe r Waive r

Se r ious Waive r

Juv Felony Waiver

Adult Felony Waiver

DAT Wai ve r

Drug/Alc. Waiver

Me dical Waiver

Aptitude Waiver

Dependents Waiver

No Wai ve r

Any Waiver

30%

35%

40%

45%

50%

55%

15% 20% 25% 30% 35% 40% 45%

Fast to E5

48-Month Attrition

Increasing

Risk

Tier II/III

Othe r Waive r

Se r ious Waive r

Juv Felony Waiver

Adult Felony Waiver

DAT Wai ve r

Drug/Alc. Waiver

Me dical Waiver

Aptitude Waiver

Dependents Waiver

No Wai ve r

Any Waiver

30%

35%

40%

45%

50%

55%

15% 20% 25% 30% 35% 40% 45%

Fast to E5

48-Month Attrition

Increasing

Risk

Waive r

No Waive r

De pe nde nt s Waive r

Aptitude Waiver

Medical Waiver

Drug/Alc. Waiver

DAT Waive r

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Othe r Waive r

Education Waiver

Tier II/III

25%

30%

35%

40%

45%

50%

55%

60%

25% 30% 35% 40% 45%

Fast to E5

48-Month Attrition

Increasing

Ris k

Figure 34. Marine Corps risk tradeoff: 48-month attrition vs. fast to E5

a. The origin of this graph is not (0,0).

Figure 35. Air Force risk tradeoff: 48-month attrition vs. fast to E5

a. The origin of this graph is not (0,0).

Any Waiver

No Waiver

Dependents Waiver

Aptitude Waiver

Medical Waiver

Drug/Alc. Waiver

DAT Waive r

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Othe r Waive r

Tier II/III

15%

20%

25%

30%

35%

40%

15% 20% 25% 30% 35% 40%

Fast to E5

48-Month Attrition

Increasing

Risk

Any Waiver

No Waiver

Dependents Waiver

Aptitude Waiver

Medical Waiver

Drug/Alc. Waiver

DAT Waive r

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Othe r Waive r

Tier II/III

15%

20%

25%

30%

35%

40%

15% 20% 25% 30% 35% 40%

Fast to E5

48-Month Attrition

Increasing

Risk

Any Waiver

No Waiver

Dependents Waiver

Aptitude Wa ive r

Medical Waiver

Drug/Alc. Waiver

DAT W aiver

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Other Waiver

Tier II/III

25%

30%

35%

40%

45%

10% 15% 20% 25% 30% 35% 40% 45%

Fast to E5

48-Month Attrition

Increasing

Risk

Any Waiver

No Waiver

Dependents Waiver

Aptitude Wa ive r

Medical Waiver

Drug/Alc. Waiver

DAT W aiver

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Other Waiver

Tier II/III

25%

30%

35%

40%

45%

10% 15% 20% 25% 30% 35% 40% 45%

Fast to E5

48-Month Attrition

Increasing

Risk

Using these two metrics to define risk, we find that the “riskiest”

waiver groups in each of the Services are as shown in table 6.

As noted above, it is essential for each Service to determine its goals

and expectations for the waivered population and thus to determine

the most appropriate “risk” measures. These figures are meant to pro-

vide a framework and examples that the Services can use to identify

especially risky categories of waivered recruits. We are not suggesting

that a tradeoff between promotion time to E5 and 48-month attrition

is most appropriate for each Service.

Minimize risks

Eliminate the use of DAT waivers

Using a variety of performance metrics, we consistently find that

recruits who access with DAT waivers are among the riskiest acces-

sions in the Army, Navy, and Marine Corps.

In addition, we find that

DAT-waived Marines are at greater risk of bad behavior (misconduct

separations, demotions, courts-martial, and NJPs) and are less likely

to receive awards or be recommended and eligible for reenlistment.

Similarly, we find that DAT-waived Sailors are at greater risk of demo-

tion by 48 months than their nonwaivered counterparts. We there-

fore recommend that the Army follow the lead of the Air Force, Navy,

Table 6. Waiver groups with relatively high 48-month attrition risk and

low probability of promoting “fast to E5”, by Service

Army Navy Marine Corps Air Force

Aptitude waiver Tier II/III Tier II/III Tier II/III

Tier II/III Education waiver DAT waiver Aptitude waiver

Drug/alcohol

waiver

DAT waiver Aptitude waiver

DAT waiver

23. The Air Force assessed only 13 recruits with DAT waivers during the

sample period.

and Marine Corps and discontinue use of DAT waivers.

The Army

has granted roughly 1,000 DAT waivers in recent years.

Consider providing commanders of recently accessed

Servicemembers with waiver information

Anecdotal conversations with military officers suggest that command-

ers currently do not know the waiver status of their personnel.

According to OSD-Accession Policy, however, waiver codes are part of

the official personnel record, although most commanders do not

know they are there. In addition, determining the type and level of

waiver requires investing a great deal of time and energy because

there is no associated table that defines waiver codes. We suggest pro-

viding commanders with waiver information that is easier to under-

stand and ensuring that commanders know that this information is

available.

Doing so would allow commanders to give waivered recruits extra

guidance and leadership. As was revealed in previous sections, many

waivered recruits share unobservable, behavioral characteristics that

make them more likely to attrite. Increased guidance and encourage-

ment could potentially change this behavior. A number of report

findings reveal that many waivered recruits who do not attrite become

above-average performers. More specifically, waivered Marines are

more likely to promote meritoriously than their nonwaivered counter-

parts, and waivered Soldiers and Sailors are more likely than their

respective nonwaivered equivalents to promote fast to E5. This sug-

gests that exposure to directed leadership, discipline, and other

aspects of the military lifestyle may improve the behavior of waivered

recruits who are retained. Any efforts that can decrease the probability

of early attrition for these recruits will increase the probability that

such positive behavioral changes will occur.

More carefully screen those with “risky” waiver combinations

Our analysis examined whether recruits who access with more than

one waiver are particularly risky recruits, and whether certain waiver

combinations matter more than others for performance and attri-

24. This was a relatively recent change in the Marine Corps. It stopped

accepting recruits who required a DAT waiver in the middle of FY09.

tion. We evaluated the ten most common waiver pairs in each Service

and identified those combinations that are most problematic, as well

as those that do not inherently impose more risk (see table 7).

If the Services agree with our definition of riskiness as slow promoters

and likely attriters, we recommend more careful screening of recruits

accessed with the waiver combinations identified in the upper por-

tion of table 7. If the Services think that the riskiness of waivered

recruits should be defined differently, we suggest that they conduct

similar analysis, identify which waiver combinations are the most

problematic, and minimize the number of recruits who require these

combinations or more carefully screen them.

Regularly identify those waiver populations with additional

screening potential

Waiver groups with a small or insignificant marginal effect on attri-

tion probability and a relatively high average attrition rate offer the

greatest screening potential. Specifically, these are the groups where

observable characteristics, as opposed to behavioral characteristics,

drive the attrition rate. Although behavioral tendencies are difficult

to identify, such characteristics as AFQT score, trimester of accession,

education tier, paygrade at accession, marital status, and the number

of dependents are known at the time of accession. We recognize that

the Services will not likely be able to screen on marital status or the

number of dependents, but AFQT scores, accession trimester, educa-

tion tier, and paygrade at accession are all characteristics on which

the Services could screen and potentially restrict the number of acces-

Table 7. Waiver combinations with higher and lower risk

Army Navy Marine Corps Air Force

Higher risk pairs

DAT & serious Serious & education Drug/alcohol & physical Medical & aptitude

Other & education Drug/alcohol & DAT

Physical & serious

Dependents & drug/alcohol

Lower risk pairs

Medical & serious Dependents & serious Adult felony & serious

sions with both a waiver and a particular characteristic. We conducted

this analysis for only a few waiver groups within each Service, particu-

larly those with small or insignificant marginal effects on attrition.

The waiver types included in this analysis are listed in table 8.

Table 9 summarizes the findings from this analysis, in which we iden-

tify the demographic and military characteristics that are most corre-

lated with attrition risk for each Service. If the Services were to place

additional emphasis on screening the waivered population or to

apply different requirements to this group, these are the characteris-

tics on which they should focus their efforts.

Table 8. Waiver types included in minimizing attrition risk analysis

Army Navy Marine Corps Air Force

Physical Physical Physical Physical

Dependents Adult felony Dependents Adult felony

DAT Juvenile felony Drug/alcohol Serious

Aptitude Aptitude Aptitude Aptitude

Table 9. Characteristics that are correlated, on average, with a lower attrition risk (for the

waiver types considered)

Army Navy

a. For the Navy, we find significant variation by waiver type. Men, for example, are less likely to attrite only in the

physical waiver population, while Tier II/III status is a significant predictor for all but the juvenile felony waiver

group. This suggests that any screening decisions for the Navy should be made with great care; screening tech-

niques will vary with the overall objective.

Marine Corps Air Force

Male Male Male Higher AFQT scores

Higher AFQT scores Tier I Higher AFQT scores Tier I

Paygrade of E2 or

greater at accession

Paygrade of E2 or

greater at accession

Paygrade of E2 or

greater at accession

Long time in DEP Long time in DEP

Tier I

A way to reduce the size of the waivered population

It appears, based on the geographic distributions of waivered and

nonwaivered recruits, that certain areas of the country are simply

more challenging recruiting terrain. That is, propensity to serve and

qualifications vary significantly from state to state, and thus from one

recruiting district to another. If, and only if, the Services aim to

reduce the size of the waivered population, it may be worth consider-

ing shifting recruiting missions from the "waiver overrepresented"

areas to the "waiver underrepresented" areas. These findings are

repeated in table 10.

It is possible that a sufficient number of quality recruits with an inter-

est in military service may not exist in some parts of the country. To

the extent that any of the Services have as a primary goal a reduction

in the size of the waivered population, it may be worth evaluating the

geographic distribution of recruiting missions, and whether they

could be updated to reflect the demographic and population

changes that have occurred. We recognize that demographic charac-

teristics of regional populations may change frequently (e.g., as a

result of broader economic conditions), and, if so, constant adjust-

ments of missions to reflect these changes would be unwarranted and

unrealistic. In addition, we understand that the Services recruit based

on the “whole recruit” concept and therefore take a number of char-

Table 10. Geographic distribution of waivered recruits, relative to the nonwaivered population

USA USAF USN USMC Consistent findings

West South Central + + - +

West North Central + + - +

South Atlantic - - + -

Pacific - - - - Less likely from Pacific

New England - - + +

Mountain + + - +

Mid-Atlantic - - - - Less likely from Mid-Atlantic

East South Central - - + -

East North Central + + + + More likely from East North Central

Non-U.S. - + - -

acteristics into consideration—waiver status being only one of them.

We therefore are not suggesting that a realignment of recruiting mis-

sions and the potential decrease in the number of waivered recruits

will necessarily improve the overall quality of the Services’ personnel.

With these caveats in mind, we suggest that a historical evaluation of

how recruiting missions have aligned with the recruitable U.S. popu-

lation may indicate that the size of the waivered population could be

reduced if recruiting missions were redefined.

This page intentionally left blank.

Conclusion

In this paper, we compared the characteristics, performance, and

attrition risk of waivered and nonwaivered recruits in each of the Ser-

vices. Waiver policies varied greatly among the Services through FY08,

restricting our ability to make cross-Service comparisons. As a result,

our analysis has been segmented into two distinct parts: an FY99–FY08

intra-Service analysis (included in this report) and an inter-Service

analysis for FY09–FY10 accessions (in a later report). Our analytical

aim in this paper was twofold:

1. Identify the demographic and military characteristics of

waivered recruits, in an effort to better define the waivered pop-

ulation within each Service and understand how it differs from

the nonwaivered population.

2. Compare the performance of waivered and nonwaivered

recruits within each Service, as measured by attrition risk and

speed of promotion to E5.

These analyses allowed us to identify, within each Service, the types of

waivered recruits that impose the greatest risk, although these find-

ings are largely dependent on how the Services define such risk.

Our demographic results suggest that waivered recruits, in all of the

Services, are more likely than their nonwaivered counterparts to be

male, older, and Tier II (or holders of nontraditional high school

degrees). In addition, the waivered population has a greater propor-

tion of whites, and a correspondingly smaller proportion of minori-

ties, and is more likely to be married. We also compare military

characteristics at accession and find that waivered recruits, on aver-

age, spend less time in DEP than their nonwaivered counterparts

(highly correlated with the fact that waivered recruits tend to be

direct ships) and, with the exception of the Air Force, tend to access

at lower paygrades. Finally, we evaluated whether waivered recruits

were particularly likely to come from certain areas of the country and

found that, when comparing the home of record for the waivered and

the nonwaivered, the East North Central region is overrepresented

(more waivered recruits come from this region than expected), while

the Pacific and Mid-Atlantic regions are underrepresented. These find-

ings highlight the variation in recruiting difficulty throughout the

country, likely due to differences in both recruitability and propensity

to serve from one region to another. Our demographic analysis

informed the estimation strategy we used to evaluate the relative attri-

tion risk and performance of the waivered population.

We conducted a comprehensive analysis of the attrition risk of

waivered recruits in each Service. First, we compared average attrition

rates across waiver types and found that, in most cases, attrition rates

were lower in the waivered population than in the population of Tier

II/III recruits, suggesting that they are, in fact, not the riskiest acces-

sions. We then identified which waiver groups have the highest inher-

ent attrition risk, after controlling for a variety of demographic and

military characteristics. These findings showed us which waiver

groups, within each Service, have higher risk based on behavioral and

unobservable characteristics—characteristics that the Services have

little power to influence. In the Army, Navy, and Marine Corps, those

accessed with a DAT waiver were the most likely to attrite, indepen-

dent of their demographic and military characteristics. In addition,

our analysis of those with multiple waivers revealed particular combi-

nations with high attrition risk and few fast promoters for each Ser-

vice. Finally, combining these results, we chose a few waiver groups

within each Service with low or insignificant behavioral effects on

attrition, and identified the characteristics that serve as significant

attrition predictors, which the Services could therefore use as an addi-

tional screen if they aim to reduce attrition within these populations.

All of these findings are highly Service specific and vary depending on

whether the aim is to reduce 6-, 24-, or 48-month attrition.

Finally, we evaluated the performance of waivered recruits relative to

nonwaivered recruits. Using time to E5 promotion as our principal

metric, we compared the prevalence of “fast” promoters in a select

number of occupational specialties for each Service and then com-

pared this with the nonwaivered population. With the exception of

the Marine Corps, nonwaivered recruits do not appear among the

five fastest promoting categories in each Service. That is, when sepa-

rating the populations by waiver type, those without waivers are, in no

cases, among the Services’ top performers, as measured by time to E5

promotion. This reveals that many waivered recruits become high-

quality Servicemembers and therefore may not be the Services’ great-

est accession “risks.”

This page intentionally left blank.

Appendix A

Appendix A: Demographic characteristics of

waivered and nonwaivered populations, by

Service

Table 11. Mean demographic characteristics of waivered and nonwaivered recruits, by Service

(FY99–FY08)

a. Red text indicates that waivered and nonwaivered recruits in that Service do not differ significantly in terms of that

characteristic. All other findings are statistically significant at the 10-percent level or better.

Waivered Nonwaivered Waivered Nonwaivered Waivered Nonwaivered Waivered Nonwaivered

Male 86

81% 87% 80% 93% 93% 81% 75%

Married 20% 14% 10% 3% 4% 2% 17% 9%

Number of Dependents 0.32 0.15 0.14 0.0 3 0.04 0.00 0.09 0.02

Tier III 4

7% 6% 1% 0% 0% 0% 0%

Tier II 19% 15% 10% 2% 3% 2% 2% 1%

Traditional HSDG 64% 68% 73% 89% 92% 94% 94% 95%

Nontraditional HSDG 6

4% 7% 5% 4% 3% 1% 1%

HSDG with College 7% 5% 3% 3% 1% 1% 3% 3%

Age 21.99 20.34 20.70 19.63 19.52 18.87 20.52 19.59

E1 a t acce ssio n 55% 55% 59% 57% 78% 72% 74% 76%

E2 a t acce ssio n 24% 24% 22% 22% 22% 28% 9% 9%

E3 at accession 15% 17% 19% 21% 0% 0% 16% 15%

E4 a t acce ssio n 5 % 4% 0% 0% 0% 0% 0% 0%

AFQT 60.55 58.81 62.65 59.60 58.34 60.79 63.66 65. 00

Months in DEP 1.91 2.89 3.89 5.22 4.23 5.64 4.04 4.92

ENcentral 17% 13% 15% 13% 1 7% 14% 15% 14%

EScentral 6% 6% 6% 6% 6% 5% 6% 7%

Midatlantic 10% 10% 9% 12% 11% 12% 9% 10%

Mountain 8% 8% 9% 7% 7% 7% 9% 7%

NewEng 3 % 3% 3% 2% 5% 4% 3% 3%

NonUSer 1 % 2% 0% 1% 0% 0% 1% 1%

Pacific 12% 15% 16% 18% 1 4% 17% 13% 15%

SAtlantic 21% 22% 20% 20% 20% 18% 20% 22%

WNCentral 7% 6% 6% 6% 6% 7% 9% 7%

WSCentral 15 % 15% 15% 15% 13% 15% 16% 15%

White 70 % 63% 63% 54% 70% 68% 72% 69%

Black 14% 17% 15% 20% 10% 10% 14% 16%

Hispanic 9% 12% 12% 15% 1 5% 16% 9% 9%

Asian 2% 2%3% 4%2% 2%3% 3%

Native American 1% 1% 5% 4% 1% 1% 1% 1%

Oth er Race 0

0% 1% 1% 0% 1% 1% 1%

Unknown Race 4% 4% 1% 2% 2% 2% 1% 2%

Number of Accessions 111,979 532, 133 119,821 276,396 175,513 141,343 42,185 257,959

US NavyUS Army US Marine Corps US Air Force

Appendix A

This page intentionally left blank.

Appendix B

Appendix B: Geographic distributions of

waivered and nonwaivered recruits

In this appendix, we examine whether waivered recruits are more or

less likely than nonwaivered recruits to come from particular regions

of the country. Using the ten geographic areas identified by the U.S.

Census (see figure 40), we begin by noting the percentage of non-

waivered recruits that come from each region.

Figure 36. U.S. Census map

Appendix B

If there is no difference in the geographic distribution for the two

groups, waivered recruits should access from the ten regions in simi-

lar proportions as nonwaivered recruits. Thus, multiplying the total

number of U.S.-wide waivered recruits by the percentage expected to

come from a particular region yields the number of waivered recruits

expected to access from that area. For example, the expected number

of waivered recruits from the Northeast (NE) region, is calculated as:

Expected number of waivered recruits from Northeast region =

[Percent of nonwaivered recruits from NE]*[total number of waivered recruits (US-wide)]

This is based, of course, on the underlying assumption that waivered

and nonwaivered recruits are distributed evenly across the United

States. A statistically significant deviation in the actual number of

waivered recruits coming from a region from the expected number will

reveal that the waivered population is either overrepresented or

underrepresented in that region. Figures 41 through 44 show the size

and significance of these deviations by region for each Service. Table

16 summarizes these findings.

Figure 37. Percentage differences between observed and expected numbers of USA waivered

recruits by geographic origin

a. Bars to the right of zero indicate that more waivered recruits than expected came from that region of the country.

The converse is true for bars to the left of zero.

27%

-7%

-3%

-4%

-51%

-17%

-6%

27%

-60% -50% -40% -30% -20% -10% 0% 10% 20% 30% 40%

ENCentral

ESCentral

MidAtlantic

Mountain

NewEng

Non-US

Pacific

SAtlantic

WNCentral

WSCentral

Appendix B

Figure 38. Percentage differences between observed and expected numbers of USN waivered

recruits by geographic origin

a. Bars to the right of zero indicate that more waivered recruits than expected came from that region of the country.

The converse is true for bars to the left of zero.

Figure 39. Differences between observed and expected numbers of USMC waivered recruits by

geographic origin

a. Bars to the right of zero indicate that more waivered recruits than expected came from that region of the country.

The converse is true for bars to the left of zero.

18%

-6%

31%

15%

-8%

-3%

16%

-46%

-26%

-50% -40% -30% -20% -10% 0% 10% 20% 30% 40%

ENCentral

ESCentral

MidAtlantic

Mountain

NewEng

Non-US

Pacific

SAtlantic

WNCentral

WSCentral

18%

15%

-2%

-5%

28%

-22%

-8%

-9%

-18%

-30% -20% -10% 0% 10% 20% 30% 40%

ENCentral

ESCentral

M idAtlantic

Mountain

NewEng

Non-US

Pacific

SA tlantic

WNCentral

WSCentral

Appendix B

It appears, based on the geographic distributions of waivered and

nonwaivered recruits, that certain areas of the country are simply

more challenging recruiting terrain. That is, propensity to serve and

qualifications vary significantly from state to state and, thus, from one

recruiting district to another. If the Services aim to reduce the size of

Figure 40. Differences between observed and expected numbers of USAF waivered recruits by

geographic origin

a. Bars to the right of zero indicate that more waivered recruits than expected came from that region of the country.

The converse is true for bars to the left of zero.

Table 12. Geographic distribution of waivered recruits, relative to nonwaivered recruits

USA USAF USN USMC Consistent findings

West South Central + + - +

West North Central + + - +

South Atlantic - - + -

Pacific - - - - Less likely from Pacific

New England - - + +

Mountain + + - +

Mid-Atlantic - - - - Less likely from Mid-Atlantic

East South Central - - + -

East North Central + + + + More likely from East North Central

Non-U.S. - + - -

-5%

-11 %

19%

-18 %

13%

-11 %

-8%

31%

-30% -20% -10% 0% 10% 20% 30% 40%

ENCentral

ESCentral

M idA tlantic

Mountain

NewEng

Non-US

Pacific

SAtlantic

WNCentral

WSCentral

Appendix B

the waivered population, they may consider shifting recruiting mis-

sions from waiver-overrepresented areas to waiver-underrepresented

areas. That may present significant challenges, however, since fre-

quently changing economic conditions are likely to change their rel-

ative distributions over time.

It is possible that a sufficient number of quality recruits with an inter-

est in military service may not exist in some parts of the country. It is

important that recruiting missions be updated sufficiently frequently

to reflect demographic and population changes that occur.

Appendix B

This page intentionally left blank.

Appendix C

Appendix C: Analysis of those with multiple

waivers

Comparing populations

In this appendix, we evaluate whether those with multiple waivers are

more likely to be either a greater attrition risk or a poorer performer

than those who access with only one waiver. We begin by comparing

the average attrition rates at 6, 24, and 48 months among those who

access with one, two, or three or more waivers. We exclude those with-

out waivers from this analysis because the ultimate objective is to eval-

uate if recruits accessed with more than one waiver behave differently

from those accessed with only one waiver. The group of those with one

waiver is therefore the relevant comparison population.

As displayed in figure 45, we find only a few significant differences in

these average “performance metrics,” depending on the number of

waivers that a recruit required. The horizontal, black bars in this

figure represent the average for the one-waiver population. In the

Army, we find that those with two waivers are less likely to attrite by 6

months than those with only one waiver. In addition, in the Marine

Corps, we find that the two-waiver population has higher attrition at

6, 24, and 48 months, and is more likely to promote “fast to E5” than

the one-waiver population. In the Navy and Air Force, we find no sig-

nificant differences at the aggregate level (before taking the type of

waivers into account) in attrition rates or E5 promotion rates between

the multiple-waiver and single-waiver populations.

Appendix C

Figure 41. Performance metrics of multiple- vs. single-waver populations

a. Horizontal black lines represent control group averages. That is, these are the average attrition rates and likelihood of being

fast to E5 for the group with one waiver. Hatch-marked bars denote statistical insignificance. All other findings are significantly

different from the one-waiver population at 5% level or better.

US Navy

10%

15%

20%

25%

30%

35%

40%

45%

50%

6 month attrition 24 month attrition 48 month attrition fast to E5

US M arine Corps

10%

15%

20%

25%

30%

35%

40%

45%

50%

6 month attrition 24 month attrition 48 month attrition fast to E5

US Arm y

10%

15%

20%

25%

30%

35%

40%

45%

50%

6 month attrition 24 month

attrition

48 month

attrition

f a s t to E5

US Air Force

10%

15%

20%

25%

30%

35%

40%

45%

50%

6 month attrition 24 month attrition 48 month attrition fast to E5

2 waivers 3 or m ore waivers

Appendix C

Do particular waiver combinations result in higher risk?

To determine if certain waiver combinations are more likely to lead

to attrition or poorer performance (and thus to highlight any combi-

nations that the Services should be concerned about), we compare

these indicators across the ten most common waiver combinations in

each Service. We restrict this analysis to those with two waivers since

the majority of the multiple-waivered population has only two.

The ten most common waiver combinations for each Service are iden-

tified in table 17, along with the percentage of the multiple-waivered

population that are accounted for by these categories. As reported in

the table, these ten pairs account for a majority of the two-waiver pop-

ulation, with a minimum of 56.8 percent in the Air Force and a max-

imum of 91.6 percent in the Navy.

25. The percentages of those with two waivers are 95 percent (USA), 89 per-

cent (USN), 80 percent (USMC), and 95 percent (USAF).

Table 13. Ten most common waiver pairs, by Service

USA USN USMC USAF

Waiver combination 1 Dependents-

Medical

Dependents-

Serious

Dependents-

Drug/Alcohol

Dependents-

Medical

Waiver combination 2 Dependents-

Serious

Medical- Serious Drug/Alcohol-

Medical

Dependents-

Serious

Waiver combination 3 Medical-DAT Medical-Other Drug/Alcohol-

DAT

Dependents-

Other

Waiver combination 4 Medical-

Adult Felony

Drug/Alcohol-

Serious

Drug/Alcohol-

Serious

Medical-

Aptitude

Waiver combination 5 Medical-Serious Drug/Alcohol-

Other

Medical-Serious Medical-Serious

Waiver combination 6 Medical-Other Serious-Minor Medical-Other Medical-Other

Waiver combination 7 DAT-Serious Serious-Other Drug/Alcohol-

Minor

Adult Felony-

Serious

Waiver combination 8 Adult Felony-

Serious

Serious-

Education

Drug/Alcohol-

Other

Serious-Minor

Waiver combination 9 Serious-Other Minor-Other Serious-Other Serious-Other

Waiver combination 10 Minor-Other Education-Other Minor-Other Minor-Other

Percentage of 2-waiver

population accounted for

72.3% 91.6% 84.3% 56.8%

Appendix C

In figures 46 through 49, we present the average 6-, 24-, and 48-month

attrition rates for the ten most common waiver pairs in each of the

Services, as well as the percentages of these groups that promote fast

to E5. The horizontal, black lines denote the values for waivered

recruits who have only one waiver, regardless of waiver type. The bars

without hatch marks indicate that recruits with that particular waiver

combination have statistically significantly different behavior from

their single-waiver counterparts.

In the Army, for example, we find that those accessed with a DAT

waiver and a serious waiver are less likely to attrite by 6 months than

those accessed with only one waiver, but they are more likely to attrite

by 48 months (see figure 46). This suggests that this is a population

Figure 42. Average time to E5 promotion and attrition rates for the ten most frequent USA

waiver pairs

a. Hatch-marked bars denote a statistically insignificant difference from the population with only one waiver. All

others are significant at the 5-percent level or better.

Dependents & Medical

Dependents & Serious

Medical & DAT

Medical & Adult Felony

Medical & Serious

Medical & Other

DAT & Serious

Adult Felony & Serious

Other & Serious

Other & Minor

Tier II/III, waiver

Tier II/III, no waiver

10%

20%

30%

40%

50%

60%

70%

6 month attrition rate

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

Fast to E5

Dependents & Medical

Dependents & Serious

Medical & DAT

Medical & Adult Felony

Medical & Serious

Medical & Other

DAT & Serious

Adult Felony & Serious

Other & Serious

Other & Minor

Tier II/III, waiver

Tier II/III, no waiver

10%

20%

30%

40%

50%

60%

70%

6 month attrition rate

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

Fast to E5

Appendix C

with particularly high long-term attrition risk. Conversely, those with

a medical waiver and a serious waiver are a relatively less risky popula-

tion, as they are less likely to attrite by 24 and 48 months and more

likely to promote fast to E5.

In the Navy (see figure 47), those with an education and other waiver

are more likely to attrite by 6, 24, and 48 months. Those with an edu-

cation and serious waiver, however, have higher attrition rates only at

24 and 48 months. Recruits accessed with a medical waiver and a seri-

ous waiver are more likely to attrite by 6 and 24 months, but the dif-

ference is insignificant for longer term attrition. Those accessed with

dependents and serious waivers are significantly more likely to be fast

promoters, and they have attrition rates that do not differ signifi-

cantly from the single-waiver population. This suggests that these Sail-

ors are among the least risky of those with multiple waivers.

Figure 43. Average time to E5 promotion and attrition rates for the ten most frequent USN

waiver pairs

a. Hatch-marked bars denote a statistically insignificant difference from the population with only one waiver. All

others are significant at the 5% level or better.

Other & Minor

Other & Education

Tier II/III, waiver

Tier II/III, no waiver

Dependents & Serious

Medical & Serious

Medical & Other

Drug/Alcohol & Serious

Drug/Alcohol & Other

Serious & Minor

Serious & Other

Serious & Education

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

6 month attrition rate

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

Fas t to 345

Other & Minor

Other & Education

Tier II/III, waiver

Tier II/III, no waiver

Dependents & Serious

Medical & Serious

Medical & Other

Drug/Alcohol & Serious

Drug/Alcohol & Other

Serious & Minor

Serious & Other

Serious & Education

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

6 month attrition rate

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

Fas t to 345

Appendix C

Findings for Marines with two waivers are presented in figure 48.

There are four waiver pairs whose attrition rates are consistently sig-

nificantly different from the group with one waiver only. These are

dependents and drug/alcohol, drug/alcohol and medical, drug/

alcohol and DAT, and medical and serious. All of these groups have

higher attrition rates at 6, 24, and 48 months than their one-waiver

counterparts. Those with dependents and drug/alcohol waivers, how-

ever, also are more likely to be fast promoters. This suggest that those

who do not attrite are, on average, better performers than their peers.

Finally, figure 49 contains Air Force findings. Two sets of findings

stand out. First, those accessed with both a medical waiver and an apti-

Figure 44. Average time to E5 promotion and attrition rates for the ten most frequent USMC

waiver pairs

a. Hatch-marked bars denote a statistically insignificant difference from the population with only one waiver. All

others are significant at the 5% level or better.

10%

20%

30%

40%

50%

60%

70%

6 month attrition

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

Fast to 345

Medical & Serious

Medical & Other

Drug/Alcohol & Minor

Dependents & Drug/Alcohol

Drug/Alcohol & Medical

Drug/Alcohol & DAT

Drug/Alcohol & Serious

Other & Drug/Alcohol

Other & Serious

Other & Minor

10%

20%

30%

40%

50%

60%

70%

6 month attrition

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

Fast to 345

Medical & Serious

Medical & Other

Drug/Alcohol & Minor

Dependents & Drug/Alcohol

Drug/Alcohol & Medical

Drug/Alcohol & DAT

Drug/Alcohol & Serious

Other & Drug/Alcohol

Other & Serious

Other & Minor

Medical & Serious

Medical & Other

Drug/Alcohol & Minor

Dependents & Drug/Alcohol

Drug/Alcohol & Medical

Drug/Alcohol & DAT

Drug/Alcohol & Serious

Other & Drug/Alcohol

Other & Serious

Other & Minor

Appendix C

tude waiver tend to be poorer performers than the single-waiver pop-

ulation. Specifically, they are more likely to attrite by 6, 24, and 48

months and are less likely to be fast promoters. In addition, those

accessed with an adult felony waiver and a serious waiver are much

more likely than their one-waiver peers to promote fast to E5. In addi-

tion, their attrition rates are not significantly different. This suggests

that Airmen with this waiver combination, although politically con-

tentious, are not a particularly risky accession group.

This is information that the Services should keep in mind when deter-

mining which waiver combinations to prioritize (in terms of acces-

sions) and which should require additional screening.

Figure 45. Average time to E5 promotion and attrition rates for the ten most frequent USAF

waiver pairs

a. Hatch-marked bars denote a statistically insignificant difference from the population with only one waiver. All

others are significant at the 5% level or better.

Dependents & Medical

Dependents & Serious

Dependents & Other

Medical & Aptitude

Medical & Serious

Medical & Other

Adult Felony & Serious

Serious & Minor

Other & Serious

Other & Minor

Tier II/III, waiver

Tier II/III, no waiver

10%

20%

30%

40%

50%

60%

70%

6 month attrition rate

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

Fas t to 345

Dependents & Medical

Dependents & Serious

Dependents & Other

Medical & Aptitude

Medical & Serious

Medical & Other

Adult Felony & Serious

Serious & Minor

Other & Serious

Other & Minor

Tier II/III, waiver

Tier II/III, no waiver

10%

20%

30%

40%

50%

60%

70%

6 month attrition rate

10%

20%

30%

40%

50%

60%

70%

24 month attrition

10%

20%

30%

40%

50%

60%

70%

48 month attrition

10%

20%

30%

40%

50%

60%

70%

Fas t to 345

Appendix C

This page intentionally left blank.

Appendix D

Appendix D: “Fast to E5” occupations and

results

Table 18 lists, by Service, the fast-to-E5 occupations. Figures 50 and 51

present the percentages of those who promote fast to E5 within

waiver categories.

Table 14. Occupations included in fast-to-E5 calculations

Service Occupation Description Service Occupation Description

Army 11B Infantryman Marine Corps 0121 Personnel Clerk

11C Indirect Fire

Infantryman

0151 Administrative Clerk

19K Armor Crewman 0311 Rifleman

31B Military Police 0621 Field Radio Operator

92A Automated Logisti-

cal Specialist

1371 Combat Engineer

Air Force 2A5X1 Aerospace Mainte-

nance

2844 Ground Communica-

tions Organizational

Repairer

2A6X1 Aerospace Propul-

sion

3521 Automotive Mainte-

nance Technician

2T2X1 Air Transportation 3531 Motor Vehicle Opera-

tor

2W1X1 Aircraft Armament

Systems

Navy AE Aviation Electrician’s

Mate

3A0X1 Knowledge Opera-

tions Management

FC Fire Controlman

3C0X1 Communication-

Computer System

Operations

HM Hospital Corpsman

3E7X1 Fire Protection IT Information Systems

Technician

3P0X1 Security Forces MA Master-at-Arms

4N0X1 Aerospace Medical

Service

OS Operations Specialist

Appendix D

Figure 46. Percentage who promote fast to E5, within a waiver category, USA and USN (for

those in select occupations)

Figure 47. Percentage who promote fast to E5, within a waiver category, USMC and USAF (for

those in select occupations)

10%

15%

20%

25%

30%

35%

40%

45%

Army Navy

Service

Percentage

Waiver

No Waiver

Dependents Waiver

Ap ti tu de Wa ive r

Medical Waiver

Drug/Alc. Waiver

DAT Waiver

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Mino r Wa iver

Other Waiver

No Waiver

Education Waiver

Ti e r II/III

N ote : H a tc h-m ar ke d b a rs de n o te th a t th e d i ffe re n ce of th a t w ai ve r typ e fr om "n o w ai ver" is

statistically insignificant. All other findings are significant at the 5% level or better.

10%

15%

20%

25%

30%

35%

40%

45%

Army Navy

Service

Percentage

Waiver

No Waiver

Dependents Waiver

Ap ti tu de Wa ive r

Medical Waiver

Drug/Alc. Waiver

DAT Waiver

Adult Felony Waiver

Juv Felony Waiver

Serious Waiver

Mino r Wa iver

Other Waiver

No Waiver

Education Waiver

Ti e r II/III

N ote : H a tc h-m ar ke d b a rs de n o te th a t th e d i ffe re n ce of th a t w ai ve r typ e fr om "n o w ai ver" is

statistically insignificant. All other findings are significant at the 5% level or better.

10%

15%

20%

25%

30%

35%

40%

45%

50%

USAF USMC

Service

Percentage

Waive r

No Waiver

Dependents Waiver

Aptitu d e W aiver

Medical Waiver

D rug /Alc. W aiver

DAT Waiver

Adu lt Felo ny Waiver

Juv Felo ny Waive r

Se riou s Waiver

Minor Waiver

Other Waiver

No Waiver

Tier II/III

Note: Hatch-m arked bars denote that the difference of that waiver type from "no waiver" is

statis tically insignifican t. All othe r find ings are sign ificant a t the 5 % le ve l or better.

10%

15%

20%

25%

30%

35%

40%

45%

50%

USAF USMC

Service

Percentage

Waive r

No Waiver

Dependents Waiver

Aptitu d e W aiver

Medical Waiver

D rug /Alc. W aiver

DAT Waiver

Adu lt Felo ny Waiver

Juv Felo ny Waive r

Se riou s Waiver

Minor Waiver

Other Waiver

No Waiver

Tier II/III

Note: Hatch-m arked bars denote that the difference of that waiver type from "no waiver" is

statis tically insignifican t. All othe r find ings are sign ificant a t the 5 % le ve l or better.

References

[1] Mission: Readiness (Military Leaders for Kids). Ready, Willing,

and Unable To Serve, 5 Nov 2009 (available for download at:

http://cdn.missionreadiness.org/NATEE1109.pdf)

[2] Beth J. Asch et al. Military Enlistment of Hispanic Youth: Obstacles

and Opportunities, 2009 (RAND Corporation) (available for

download at http://www.rand.org/pubs/monographs/

2009/RAND_MG773.pdf)

[3] Aline Quester and Jacob Morse. Bibliography of CNA Manpower

Work: Marine Corps Focus, Dec 2007 (CNA Information Memo-

randum D0017311.A1/Final)

[4] Lolita C. Baldor (Associated Press). “Study: Recruits on Waiv-

ers Get Promoted Faster.” Army Times, 1 May 2008 (available

for download at http://www.armytimes.com/news/2008/

04/ap_military waivers_042908/)

[5] Aline Quester. First-Term Attrition in the Marine Corps, Mar 1993

(CNA Research Memorandum 92-200)

[6] Ann Scott Tyson. “A Historic Success in Military Recruiting:

In Midst of Downturn, All Targets Are Met.” Washington Post,

11 Oct 2009

[7] Steve Vogel. “Military Recruiting Faces a Budget Cut,” Wash-

ington Post, 11 May 2009

[8] Under Secretary of Defense, Personnel and Readiness.

“Enlistment Waivers,” 27 Jun 2008 (Directive Type Memoran-

dum 0-8-018)

100

[9] United States General Accounting Office. Military Recruiting:

New Initiatives Could Improve Criminal History Screening, Feb

1999 (GAO/NSIAD-99-53)

[10] Christopher Distifeno. “Effects of Moral Conduct Waivers on

First-Term Attrition of U.S. Army Soldiers,” Mar 2008 (Naval

Postgraduate School thesis)

[11] Lyle D. Hall. “Analyzing Success of Navy Enlistees With Moral

Waivers,” Sep 1999 (Naval Postgraduate School thesis)

[12] Jennie W. Wenger and Apriel K. Hodari. Predictors of Attrition:

Attitudes, Behaviors, and Educational Characteristics, Jul 2004

(CNA Research Memorandum D0010146.A2)

[13] Leonard Etcho. “The Effect of Moral Waivers on First-Term,

Unsuitability Attrition in the Marine Corps,” Mar 1996 (Naval

Postgraduate School thesis)

[14] Aline O. Quester. Marine Corps Recruits: A Historical Look at

Accessions and Bootcamp Performance, Sep 2010 (CNA Anno-

tated Briefing D0023537.A1)

[15] Dan J. Putka et al. Evaluating Moral Character Waiver Policy

Against Servicemember Attrition and In-Service Deviance Through

the First 18 Months of Service, Jan 2004 (Human Resources

Research Organization FR-03-96)

[16] Dana L. Brookshire and Anita U. Hattiangadi. Emerging Issues

in USMC Recruiting: Comparing Relative Attrition Risk Among

Marine Corps Recruits, Aug 2006 (CNA Research Memoran-

dum D0014200.A2)

[17] Peggy Golfin. Manning Under AIP, Jun 2006 (CNA Annotated

Briefing D0014440.A1/Final)

101

List of figures

Figure 1. Percentage of waivered recruits by Service:

FY99 through FY08 . . . . . . . . . . . . . . . . . . 20

Figure 2. Trends in waivered accessions FY99

through FY08, by waiver type . . . . . . . . . . . . . 21

Figure 3. Percentages of waivered and nonwaivered

Tier II/III recruits, by Service (FY99–FY08) . . . . . 24

Figure 4. Percentages of Tier II/III recruits, by Service

and accession FY (FY99–FY08) . . . . . . . . . . . . 25

Figure 5. Average number of months spent in DEP

(FY99–FY08) . . . . . . . . . . . . . . . . . . . . . . 26

Figure 6. Attrition by waiver category: Army . . . . . . . . . . 31

Figure 7. Attrition by waiver category: Navy . . . . . . . . . . 31

Figure 8. Attrition by waiver category: Air Force . . . . . . . . 32

Figure 9. Attrition by waiver category: Marine Corps . . . . . 32

Figure 10. Difference in attrition probabilities

(waivered minus nonwaivered) . . . . . . . . . . . . 33

Figure 11. Marginal effects of waivers on 6-, 24-, and

48-month attrition rates: Army . . . . . . . . . . . . 38

Figure 12. Marginal effects of waivers on 6-, 24-, and

48-month attrition rates: Navy . . . . . . . . . . . . 39

Figure 13. Marginal effect of waivers on 6-, 24-, and

48-month attrition rates: Marine Corps . . . . . . . 41

102

Figure 14. Marginal effect of waivers on 6-, 24-, and

48-month attrition rates: Air Force . . . . . . . . . . 42

Figure 15. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

physical waiver: Army . . . . . . . . . . . . . . . . . 46

Figure 16. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

dependents waiver: Army . . . . . . . . . . . . . . . 47

Figure 17. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

DAT waiver: Army . . . . . . . . . . . . . . . . . . . 48

Figure 18. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given an

aptitude waiver: Army . . . . . . . . . . . . . . . . . 49

Figure 19. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

physical waiver: Navy . . . . . . . . . . . . . . . . . 50

Figure 20. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given an

aptitude waiver: Navy . . . . . . . . . . . . . . . . . 51

Figure 21. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

juvenile felony waiver: Navy. . . . . . . . . . . . . . 52

Figure 22. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given an

adult felony waiver: Navy . . . . . . . . . . . . . . . 53

Figure 23. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

physical waiver: Marine Corps . . . . . . . . . . . . 54

103

Figure 24. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

dependents waiver: Marine Corps . . . . . . . . . . 55

Figure 25. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

drug or alcohol waiver: Marine Corps . . . . . . . . 56

Figure 26. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given an

aptitude waiver: Marine Corps . . . . . . . . . . . . 57

Figure 27. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

physical waiver: Air Force . . . . . . . . . . . . . . . 58

Figure 28. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given an

aptitude waiver: Air Force. . . . . . . . . . . . . . . 59

Figure 29. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given a

serious waiver: Air Force . . . . . . . . . . . . . . . 60

Figure 30. Marginal effects of accession characteristics

on 6-, 24-, and 48-month attrition, given an

adult felony waiver: Air Force. . . . . . . . . . . . . 61

Figure 31. Percentage of Servicemembers who

promote “fast” to E5, by Service and waiver

status (FY99-FY05 accessions) . . . . . . . . . . . . . 64

Figure 32. Army risk tradeoff: 48-month attrition vs.

fast to E5 . . . . . . . . . . . . . . . . . . . . . . . . 68

Figure 33. Navy risk tradeoff: 48-month attrition vs.

fast to E5 . . . . . . . . . . . . . . . . . . . . . . . . 68

Figure 34. Marine Corps risk tradeoff: 48-month

attrition vs. fast to E5 . . . . . . . . . . . . . . . . . 69

104

Figure 35. Air Force risk tradeoff: 48-month attrition

vs. fast to E5 . . . . . . . . . . . . . . . . . . . . . . 69

Figure 36. U.S. Census map. . . . . . . . . . . . . . . . . . . . 83

Figure 37. Percentage differences between observed

and expected numbers of USA waivered

recruits by geographic origin . . . . . . . . . . . . . 84

Figure 38. Percentage differences between observed

and expected numbers of USN waivered

recruits by geographic origin . . . . . . . . . . . . . 85

Figure 39. Differences between observed and expected

numbers of USMC waivered recruits by

geographic origin . . . . . . . . . . . . . . . . . . . 85

Figure 40. Differences between observed and

expected numbers of USAF waivered

recruits by geographic origin . . . . . . . . . . . . . 86

Figure 41. Performance metrics of multiple- vs.

single-waver populations . . . . . . . . . . . . . . . 90

Figure 42. Average time to E5 promotion and attrition

rates for the ten most frequent USA

waiver pairs . . . . . . . . . . . . . . . . . . . . . . 92

Figure 43. Average time to E5 promotion and attrition

rates for the ten most frequent USN

waiver pairs . . . . . . . . . . . . . . . . . . . . . . 93

Figure 44. Average time to E5 promotion and attrition

rates for the ten most frequent USMC

waiver pairs . . . . . . . . . . . . . . . . . . . . . . 94

Figure 45. Average time to E5 promotion and attrition

rates for the ten most frequent USAF waiver

pairs . . . . . . . . . . . . . . . . . . . . . . . . . . 95

105

Figure 46. Percentage who promote fast to E5, within a

waiver category, USA and USN (for those in

select occupations) . . . . . . . . . . . . . . . . . . 98

Figure 47. Percentage who promote fast to E5, within a

waiver category, USMC and USAF (for those

in select occupations) . . . . . . . . . . . . . . . . . 98

106

This page intentionally left blank.

107

List of tables

Table 1. Number of offenses requiring an enlistment

waiver, by Service . . . . . . . . . . . . . . . . . . . 10

Table 2. Demographic characteristics of waivered vice

nonwaivered recruits, by Service . . . . . . . . . . . 24

Table 3. Waiver groups included in minimizing attrition

risk analysis . . . . . . . . . . . . . . . . . . . . . . 44

Table 4. Characteristics considered in this analysis . . . . . . 44

Table 5. Summarized “fast to E5” results, by Service and

waiver type . . . . . . . . . . . . . . . . . . . . . . . 64

Table 6. Waiver groups with relatively high 48-month

attrition risk and low probability of promoting

“fast to E5”, by Service. . . . . . . . . . . . . . . . . 70

Table 7. Waiver combinations with higher and lower risk . . 72

Table 8. Waiver types included in minimizing attrition

risk analysis . . . . . . . . . . . . . . . . . . . . . . 73

Table 9. Characteristics that are correlated, on average,

with a lower attrition risk (for the waiver types

considered) . . . . . . . . . . . . . . . . . . . . . . 73

Table 10. Geographic distribution of waivered recruits,

relative to the nonwaivered population . . . . . . . 74

Table 11. Mean demographic characteristics of waivered

and nonwaivered recruits, by Service

(FY99–FY08) . . . . . . . . . . . . . . . . . . . . . . 81

108

Table 12. Geographic distribution of waivered recruits,

relative to nonwaivered recruits . . . . . . . . . . . 86

Table 13. Ten most common waiver pairs, by Service . . . . . 91

Table 14. Occupations included in fast-to-E5 calculations . . . 97

4825 Mark Center Drive, Alexandria, VA 22311-1850 703-824-2000 www.cna.org

CRM D0023955.A4/Final