An analysis of the effect of reserve participation and training on civilian employment and earnings

“Calhoun

Institutional Archive of the Naval Postgraduate School

Calhoun: The NPS Institutional Archive
DSpace Repository

Theses and Dissertations 1. Thesis and Dissertation Collection, all items

1993-03

An analysis of the effect of reserve
participation and training on civilian
employment and earnings

McGuire, John A.

Monterey, California. Naval Postgraduate School
http://hdl.handle.net/10945/39871

This publication is a work of the U.S. Government as defined in Title 17, United
States Code, Section 101. Copyright protection is not available for this work in the
United States.

Downloaded from NPS Archive: Calhoun

Calhoun is the Naval Postgraduate School's public access digital repository for

\§ D U DL EY research materials and institutional publications created by the NPS community.
«iit Calhoun is named for Professor of Mathematics Guy K. Calhoun, NPS's first

NY KNOX appointed -- and published -- scholarly author.

LIBRARY Dudley Knox Library / Naval Postgraduate School
411 Dyer Road / 1 University Circle

http://www.nps.edu/library Monterey, California USA 93943

31
nt i ii i

NAVAL POSTGRADUATE SCHOOL
Monterey, California

THESIS

AN ANALYSIS OF THE EFFECT OF RESERVE
PARTICIPATION AND TRAINING ON CIVILIAN
EMPLOYMENT AND EARNINGS
by
John A. McGuire

March 1993

Principal Advisor: Stephen L. Mehay
Associate Advisor: Gregory G. Hildebrandt

Approved for public release; distribution is unlimited.

Oo
CO

OTIC
3" 2 7 caf)
2:

| 93-11969
9 ces

Unclassified

Security Classitication of this page
REPORT DOCUMENTATION PAGE

Ja Report Security Classification: Unclassified Ib Restrictive Markings

3 Distribution/Availability of Report
Approved for public release; distribution is unlimited.

5 Monitoring Organization Report Number(s)

6a Name of Pertorming Organization 6b Office Symbol Ta Name of Monitoring Organization
¥ Naval Postgraduate School (if applicable) 36 Naval Postgraduate School

6c Address (city, state, and ZIP code) To Address (city, state, and ZIP code}
Monterey CA 93943-5000 Monterey CA 93943-5000

8a Name of Funding/Sponsoring Organization 6b Office Symbol 9 Procurement Instrument Identification Number
(if applicable)

10 Source of Funding Numbers
Program Element No Work Unit Accession No

11 Title finclude security classification) AN ANALYSIS OF THE EFFECT OF RESERVE PARTICIPATION AND TRAINING ON
CIVILIAN EMPLOYMENT AND EARNINGS

12 Personal Author(s} John A. McGuire

13a Type of Report 13b Time Covered 14 Date of Report fyear, month, day) 15 Page Count
Master's Thesis From To 1993, March 64

16 Supplementary Notation The views expressed in this thesis are those of the author and do not reflect the official policy or position
of the Department of Defense or the U.S. Government.

17 Cosati Codes 18 Subject Terms (continue on reverse if necessary and idenufy by block number)
Subgroup Reserve Training, Reserve Affiliation, Two-Stage Least Squares, Reserve Components
aT

19 Abstract (continue on reverse if necessary and identify by block number)

Utilizing data from the 1986 Reserve Components Surveys, this thesis implements a test of the hypothesis that a positive
relationship exists between occupational training received in the reserves and increased benefits and wages on reservists’ civilian
jobs. The null hypothesis was that no such relationship exists, or that it exists for relatively few reservists, so that reserve
participation is mainly a form of moonlighting with few spillover benefits to the individual or society in the form of increased
worker productivity. Log-earnings regression equations were specified to test the basic hypothesis. The two-stage least squares
(2SLS) estimating technique was utilized to estimate the models due to the existence of simultaneity bias in the regression
equations. It was determined that affiliating with the reserves to receive training results in an increase in civilian benefits and
wages. Therefore the null hypothesis was rejected, leading to the conicusion that reserve training does appear to provide important
benefits to some enlistees, namely those who are motivated to seek skill training that can be used on their civilian job or used to
find a better civilian job.

20 Distribution/Availability of Abstract 21 Abstract Security Classification
a _X_ unclassified/unlimited __ same as report _. DTIC users | Unclassified
228 Name of Responsible Individual 22b Telephone (include Area Code) 22¢ Office Symbol
Stephen L. Mehay (408) 656-2643 AS/MP
DD FORM 1473,84 MAR 83 APR edition may be used until exhausic ‘ security classification of this page
Al! other editions are obsolete Unclassified

L

Approved for public release; distribution is unlimited.
AN ANALYSIS OF THE EFFECT OF RESERVE

PARTICIPATION AND TRAINING ON CIVILIAN .
EMPLOYMENT AND EARNINGS

by
John A. McGuire
Lieutenant, United States Navy
B.S.E., Florida State University, 1987

Submitted in partial fulfillment
of the requirements for the degree of

MASTER OF SCIENCE IN MANAGEMENT

from the

NAVAL POSTGRADUATE SCHOOL
March 1993

Author:

John A. McGuire

Approved by:

David R. Whipple, Chaiyman
Department of Administrativé Sciences

ii

ABSTRACT

Utilizing data from the 1986 Reserve Components Surveys,
this thesis implements a test of the hypothesis that a
positive relationship exists between occupational training
received in the reserves and increased benefits and wages on
reservists’ civilian jobs. The null hypothesis was that no
such relationship exists, or that it exists for relatively few
reservists, so that reserve participation is mainly a form of
moonlighting with few spillover benefits to the individual or
society in the form of increased worker productivity. Log-
earnings regression equations were specified to test the basic
hypothesis. The two-stage least squares (2SLS) estimating
technique was utilized to estimate the models due to the
existence of simultaneity bias in the regression eyuations.
It was determined that affiliating with the reserves to
receive training results in an increase in civilian benefits
and wages. Therefore the null hypothesis was rejected,
leading to the conclusion that reserve training does appear to
provide important benefits to some enlistees, namely those who
are motivated to seek skill training that can be used on their

civilian job or used to find a better civilian job.

Accession For

a =
e -

Il.

IIt.

Iv.

TABLE OF CONTENTS

INTRODUCTION . . . -.- 2...) eee ee eee
A. RESERVE HISTORY ... .

B. PRESENT DAY RESERVES

C. TRAINING

D. PURPOSE

LITERATURE REVIEW ....

A. TRAINING TO OBTAIN A CIVILIAN JOB

B, REQUIREMENTS OF RESERVE PARTICIPATION

C. SECONDARY LABOR MARKET PARTICIPATION THEORY
METHODOLOGY ........

SAMPLE POPULATION

SURVEY DEVELOPMENT ..

RESPONSE RATES .....

MODEL SPECIFICATION . .

wow O90 AQ wD p

VARIABLE DEFINITION .... .
EMPIRICAL RESULTS .....

A. DESCRIPTIVE STATISTICS

B. REGRESSION RESULTS ... .
CONCLUSIONS AND RECOMMENDATIONS
A. CONCLUSIONS .... .

B. RECOMMENDATIONS ... .
REFERENCES . . . . . 2 e- sw ee

INITIAL DISTRIBUTION LIST

58

59

I. INTRODUCTION

A. RESERVE HISTORY

The United States military reserve system is deeply
rooted in American history. From early colonial days, up to
and including the Spanish American War of 1898, the reserves
coexisted with their professional counterparts, the
continental or standing army. There were relatively few
changes made to the reserve system during this period
primarily due to the success of the militia. Those changes
that were made were done so with no regard for the resultant
efficiency of the militia. Military reformers at the time
of the Spanish American War were well aware of the
conscription systems in use by Furopean powers. These
systems drafted men into the active army and then
involuntarily assigned them to reserve mobilization billets.
This provided a readily available pool of trained soldiers
for periods of mobilization. (Sullivan, 1985]

The success of the reserves in the Spanish American War
made changing the reserve system a difficult task. America
was fast becoming a world superpower and took on a more
internationally political role during the first few years of
the nineteenth century. This active role in international
politics required a larger, more effective military force.

Subsequently, this also required a more centralized and

better trained reserve force than the pre-1898 reserve
system could provide. Beginning in 1903 Congress enacted
three acts that provided for a large, voluntary standing
force, a reserve force to be used primarily for support
missions, and a national guard for combat and civil
disturbance missions. These three acts, the Dick Act of
1903, and the National Defense Acts of 1916 and 1920, while
amended several times over the years, have provided the
structural framework found in America’s military system

today.

B. PRESENT DAY RESERVES

There are basically two classifications of reservists:
those who belong to the Selected Reserves (SELRES) and those
who belong to the Individual Ready Reserves (IRR). The IRR
is made up of individuals who have served less than six
years in the active or selected reserve forces and have
residual military service obligation (MSO). Most first term
enlistments require the merber to obligate him/herself for
eight years from the date of enlistment. The IRR are not
organized into units and members do not receive periodic
training or pay.

The Selected Reserves are organized into specific units
whose primary mission is to provide combat, combat support,

and combat service support units that can be mobilized

quickly in wartime. The same reserve units may be used for

civil disturbance missions during peacetime. These "weekend

warriors" are divided into the Army and Air National Guard
and the reserve components of the Army, Navy, Air Force, and
Marines. The Selected Reserves are required to drill one
weekend per month and one two-week period annually.
Selected reservists are paid for both their monthly drills
and the annual training concurrent with their paygrade.
Analysts have often referred to selected reservists as
"moonlighters" since they normally hold a primary full- or
part-time job in addition to their reserve duty.
Moonlighters are characterized as individuals who
participate in the secondary labor market in an effort to
increase their earnings. The amount of moonlighting hours a
worker provides is directly related to the worker’s primary
hours. If a worker is unable to work the amount of hours at
his/her primary occupation that he/she desires, then the
worker is "underemployed." Disregarding any additional
costs of securing a second job, an underemployed worker will
moonlight to enhance his/her total earnings. Since reserve
pay is essentially a fixed wage per training day, reservists
are not able to choose the amount of moonlighting hours they
wish to work. This study will focus primarily on the
Selected Reserves since there is no moonlighting benefit
derived from being a member of the Individual Ready

Reserves.

C. TRAINING

Members of the Selected Reserves are afforded the same
initial training opportunities as their active duty
counterparts upon enlistment. A member is offered a choice
of several military occupational specialties (MOS) depending
on his/her scores on the Armed Services Vocational Aptitude
Battery (ASVAB). The reservist first goes to boot camp or
basic training and, if qualified for a technical occupation,
moves on to advanced training in a formal service school to
learn his/her military occupational specialty. There is no
difference in the training provided the reservist and the

active duty member at the time of enlistment.

D. PURPOSE

Analysts have often argued that the reserves benefit
from enlisting members wicse primary motivation for
participating in the reserves is to moonlight and to earn
extra income. Some of these reservists are already trained
in what eventually becomes their military occupational
specialty. Both those with and those without prior training
participate in the reserves as a way of supplementing their
earnings.

On the other hand, there may be members, mostly non-
prior service, who enlist in the reserves in order to
receive training in an occupational skill to enhance their

opportunities to obtain civilian employment and to boost

their long-run civilian earnings potential. Those who use
the reserves in this way become better trained, more
productive, and more employable civilian workers. As a
result, there is a spillover benefit to society. If,
however, the primary reason for enlistment in the reserves
is to augment one’s civilian income, reserve participation
is simply a second job or a moonlighting activity with few
indirect, or secondary, benefits to the individual and to
society. Of coures, the increase in earnings for those who
moonlight in the reserves may initially exceed the increase
in earnings for those who participate for traininy. But
moonlighting tends to be a temporary phenomenon and ends
when one’s enlistment ends. For those who are trained in
the reserves, benefits on their primary civilian job may
continue throughout their worklife.

This thesis will attempt to test the alternative
hypothesis that a positive relationship exists between
undergoing reserve training and benefits and wages on one’s
primary civilian job for some reservists. The null
hypothesis is that no such relationship exists, or that it
exists for relatively few reservists so that reserve
participation is mainly a form of moonlighting with few
derivative benefits to the individual or to society. If a
positive relationship is observed between reserve training

and civilian occupations and compensation, this research

will also measure the magnitude of the benefits of joining

the reserves to non-py.1-c¢ service high school graduates

entering the labor force, which should provide valuable

information to manpower planners and recruiters.

II. LITERATURE REVIEW

A. TRAINING TO OBTAIN A CIVILIAN JOB

Very few prior studies have treated reserve
participation as a means of obtaining a civilian job or the
training needed to obtain a civilian job. It is generally
accepted that the reserves benefit by enlisting members from
the civilian sector who are already trained in an
occupational specialty. But, what about those reservists
who had no prior skill training and were employed in an
unskilled occupation or were unemployed when they enlisted?

In the 1986 Reserve Components Survey (RCS) over 18,000
(27 percent) of the more than 65,000 respondents stated that
they joined the reserves to obtain skill training to help
them get a civilian job. Presumably, respondents that
either were not employed, or were employed only part-time,
would be most likely to cite this reason for enlistment;
they also would be the most likely to obtain a "Spillover"
benefit from reserve participation. However, others who
were employed in an occupation in which they did not intend
to remain would also stand to gain by receiving training
that would allow them to switch to a better occupation. In
the absence of previous studies in this area, this study
examines the hypothesis that a positive relationship exists

between training obtained in the reserves and increased

wages and benefits to reservists in their primary civilian

occupation.

B. REQUIREMENTS OF RESERVE PARTICIPATION

There are virtually no other second jobs which come to
mind that parallel the characteristics of reserve
participation. Burright, Grissmer and Doering (1982) found
three requirements of reserve participation that set it
apart from other second jobs and voluntary activities.
First, during ennual training, reservists must spend 14 days
of full-time work during the summer. This requires their
absence from home and from their primary civilian full-time
job. Non-prior service rese:vists must, upon entry, train
full time for four months in their occupational specialty.
Additionally, during periods of national crisis, such as
Operation Desert Storm, or civil emergencies, reservists may
be called-up to full-time duty.

The obligation to serve full time during summer ACDUTRA
does not necessarily represent a cost to reservists,
especia’’y if their full-time military pay exceeds their
civilian pay. David Grissmer, Richard Buddin, and Sheila
Nataraj Kirby (1989) found that over 50 percent of the
respondents in the RCS would face moderate or serious
decreases in total income if mobilized for thirty days or
more. If the training recaived in the reserves is
transferable to a job opportunity in the civilian sector,
then any costs associated with full-time duty are reduced.

8

Second, the reservist is legally obligated for up to
eight years of service. Civilian second jobs do not
normally require such an employment contract. This
requirement provides job security to some reservists, while
for others it represents an opportunity cost because it
reduces the possibility of holding other secondary jobs.

Finally, Burright et al. determined the inflexible work
schedule of reserve participation differs significantly from
the work schedules of most moonlighting jobs. In 1982
veservists were paid for either 8 or 16 hours per month with
no opportunity for increasing paid hours. Most mandatory
drills are scheduled on weekends with no flexibility for
alternative schedules to accomodate civilian employer
concerns. Burright et al. found that civilian employer
attitudes toward the reservist’s participation were major

factors in the reenlistment,enlistment decision.

C. SECONDARY LABOR MARKET PARTICIPATION THEORY
Moonlighting has traditionally been treated as a
decision to participate in the secondary labor market as a
means of supplementing one’s primary job income. Most of
the prior studies on reserve participation have treated it
as a labor force decision similar to civilian moonlighting
{Mehay 1990]. Linda Gorman and George Thomas (1991)
hypothesized that reserve membership is part-time employment
that competes with leisure time and will usually have lower
priority than the member's primary occupation. Stephen L.

9

Mehay (1990) hypothesized that, contrary to the assumptions

of prior studies that reserve participation and moonlighting
are influenced by similar economic factors, different
criteria are used in the decisions of reservists and
civilians. If the decision to participate in the reserves
is distinctly different from the decision to moonlight, then
previous studies that have treated them the same will be
affected by specification bias [Mehay 1990}. Robert Shishko
and Bernard Rostker (1976) simply defined anyone who holds
two or more jobs as a moonlighter and thus participates in
the secondary labor market. They estimated the moonlighting
supply curve with data from the Panel Study of Income
Dynamics using the Tobit technique for estimating
relationships with limited dependent variables. Applying
their definition of moonlighting to a person who holds two
part-time jobs would seem to violate the principles of
secondary labor market participation. Which job would be
considered the primary occupation?

Grissmer et al. found that approximately three-quarters
of Army reservists hold full-time civilian jobs in addition
to participating in the reserves. They observed that
reservists are drawn from the competitive labor market and,
as such, the reserves compete with other employers who

provide more flexible hours, perhaps a better wage, and

occasional overtime. If an individual is able to obtain

overtime hours on his/her primary job, he/she will be less
likely to either moonlight or participate in the reserves.

Grissmer et al. outlined four components every
prospective reservist must consider in the decision to
participate: the present and future monetary benefits from
reserve service; the non-monetary benefits of reserve
service; the monetary opportunity costs from reserve
service; and the non-monetary opportunity costs of reserve
service. An implicit assumption is that the prospective
reservist is able to differentiate between monetary and non-
monetary benefits as well as other available alternatives.
This may not necessarily be true.

The decision to moonlight is based on several economic
factors. Shishko and Rostker theorized that an individual's
decision to moonlight is based on whether he/she can work
enough hours at his/her primary wage rate to satisfy desired
income goals. They identified hours worked on the primary
job, the primary wage, the secondary wage, and non-labor
income as the most important variables in the moonlighting
decision. They proposed that "changes in the primary wage
alter the minimum wage necessary to induce people to take a
second job." An increase in the primary wage may result in
an increase or a decrease in the minimum acceptable
secondary wage rate (i.e., the second job reservation wage).

Due to substitution and income effects, an increase in

the secondary wage could result in an increase or decrease

11

in moonlighting hours worked. This is especially true when

the secondary wage is greater than the primary wage. If

there are constraints on the number of hours a worker may

work on the primary job, secondary jobs may be accepted even -
if the secondary wage is less than the primary wage.

Non-labor income only affects hours worked if the
desired hours of employment fall below the actual hours. If
an individual is working more hours than he desires, a small
increase in non-labor income will result in an attempt to
reduce the number of hours worked in either the primary or
secondary job.

Gorman and Thomas proposed that along with economic
factors such as extra income and additional training for
future income, there are also psychic factors such as
patriotism and camaraderie that are associated with reserve
participation. Mehay also challenged the traditional
"moonlighting hypothesis" stating that, along with extra
income, reserve participation offers the member dynamic
training and learning experiences, extensive fringe
benefits, camaraderie and other unique features not normally
found in civilian moonlighting jobs. Like Mehay, and Gorman
and Thomas, Burright, et al. found reserve participation
provides non-pecuniary rewards such as camaraderie and a
sense of team accomplishment. Burright, et al. also found
fringe benefits such as health care, life insurance,

educational benefits, tax benefits, and a cost-of-living-

12

adjusted pension at the age of 60 after 20 years of service
to be major attractions to reserve participation.
In an empirical moonlighting study conducted by Compton
’ (1979), it was concluded that the supply of labor for second
jobs will increase if:
1. Wages in the second job increase significantly.
2. The person is black.
3. The person is non-urban.
4. Wages in the primary job decrease
significantly.
5. The person is not a high school graduate.

6. The number of hours required for the primary
job decreases.

7. The person’s spouse is not working or quits
working.

8. The person’s non-labor income (interest,
dividends, etc.) decreases or is nonexistent.

SOURCE: Motivation For First Term Reserve
Reenlistment, Naval Postgraduate School Master’s
Thesis by James S. Sullivan Jr., 1985.

In his empirical analysis, Mehay constructed a choice-
based sample consisting of reservists and civilians who were
already working full-time in a primary occupation and who
chose to participate either in the reserves or the secondary
labor market. He developed a trichotomous model whereby an
individual could moonlight, participate in the reserves, or
hold one primary job. He hypothesized that if the model

collapsed to a dichotomous one then the decisions to

13

participate in the reserves or to moonlight would be assumed
to be the same and the individual would be indifferent
between reserve affiliation and moonlighting. Mehay
concluded that the two decisions are not equivalent. His
results did support previous research in that he found
participation to be strongly influenced by members’ economic
status and job market factors such as unemployment rates and
prevailing wage rates in the local geographic area. Mehay
also concluded that individuals join the reserves to
Supplement their income, obtain skill training, receive
fringe benefits and serve their country.

Gorman and Thomas observed many college students who
join the U.S. Army Reserve (U.S.A.R.) to help finance their
education. The U.S.A.R. is capable of accommodating people
who pursue goals not fully compatible with service in the
active Army [Gorman, et al. 1991].

Gorman and Thomas were surprised to find that almost 25
percent of their sample joined the Army Reserve intending to
transfer to the active Army. It seemed that many high
school graduates joined the reserves in order to obtain an
"education" in Army life. Without obligating themselves to
a full-time active Army job they could see what it would be
like and decide if they wanted to transfer to the active
Army or remain in the reserves until their commitment was

up. This behavior provides an opportunity to increase the

14

pool of active Army members by increasing the pool of Army

reservists.

Gorman and Thomas utilized log-linear models to estimate
the probability that a person 18 years of age or younger
will choose one of three (author-established) "motives" for
enlisting in the reserves. These motives were "earn more
money", "self-improvement", and "serve." The serve category
included those members who responded they wanted to serve
their country or that family tradition warranted serving.

Gorman and Thomas found that college-age recruits in
mental categories 1 and 2 joined the reserves to earn more
money. Almost 70 percent of those seeking more money did so
to pay college expenses.

Gorman and Thomas concluded that the Army Reserve may be
an invaluable source of high-quality recruits for the active
Army. More than half of their respondents were in mental
categories 1 and 2 and, of these, 25 percent stated that
they planned to transfer to the active Army. Gorman and
Thomas’ findings suggest that reservists enlist fora
variety of reasons including self-improvement, the
opportunity to earn extra money, and patriotism. Of these,
a large percentage enlist with the intent to transfer to the
active Army.

Burright, et al. identified five variables that
influence the reenlistment decision. For the purpose of

this study these variables also could be applied to the

a5

initial enlistment decision. The variables included net
reserve pay, net required days of reserve service, civilian
wage rate, number of hours worked on the civilian job, and
frequency of overtime opportunities on the civilian job.

Utilizing a logistic regression model, Burright, et al.
defined the reenlistment decision by a dichotomous variable
assuming the value of one for reenlistment and zero for
separation. As expected, higher net reserve wages and fewer
net reserve days would increase reenlistment rates.
Presumably, these same factors would increase initial
enlistment rates. Likewise, higher civilian wages, longer
civilian hours, and increased civilian overtime
opportunities decreased the probability of reenlistment.
Presumably, these factors would also decrea:-: initial
enlistment rates.

Finally, Regets (1990) argues that reserve participation
may be simply a form of "compensated leisure." He theorized
that "moonlighting and compensated leisure models of reserve
service generate very different predictions of labor supply
behavior". Utilizing two sets of data, a Naval Reserve data
set created from personnel records, and an all-service
extract of reservists from the Survey of Income and Program
Participation, Regets empirically tested both models. He
found the compensated leisure model predicted a positive
income effect (i.e., increases in non-labor income increase

the quantity of labor supplied to Reserve service). The

16

moonlighting model however, predicts a negative income
effect. With the exception of low-income reservists,
Regets’ empirical results supported the compensated leisure
model.

In summary, most prior studies have utilized the
economic labor market theory of moonlighting to explain
reserve participation and reenlistment decisions. However,
based on significant differences in the characteristics of
reserve jobs and civilian moonlighting jobs, other analysts
have questioned whether the formal economic model of
moonlighting applies to reserve decisions, or whether
reserve participation is a labor market activity at all.

One purpose of this study is to turn the question around and
to inguire whether reserve participation is motivated by a
desire to upgrade earnings capacity on one’s primary job.

If so, economic factors, such as the level of reserve pay,
would not have a strong impact on participations decisions.
Rather, the opportunity for skill training, and the prospect
that such training would augment future long-run earnings on
the primary job, would be the principal motivating factor.
The next section of the thesis proposes an empirical test of

this hypothesis.

17

III. METHODOLOGY

In January, 1983, the Deputy Secretary of Defense
mandated a survey of military families, active and reserve,
who were increasingly recognized as important to the
retention and preparedness of the United States’ armed
forces. Each of the services had previously conducted
small-scale studies of its own member families. However, no
Single consistent cross-service data set was available to
permit the study of emerging DoD family issues. The DoD
also needed to assess the various impacts of numerous
personnel policies that had been implemented in the early
1980’s.

The Assistant Secretary of Defense for Manpower, Reserve
Affairs and Logistics established a DoD-wide committee, the
Family Survey Coordinating Committee, to assess the
information requirements and data sources needed to survey
both the active and reserve components of the military
(Hunt, et al. 1986]. Due to the complexity of surveying
both components, the Committee initiated active force
surveys in 1985 but temporarily postponed the reserve
surveys. The Reserve Components Surveys (RCS) were not

completed until 1986. The Defense Manpower Data Center

(DMDC) was contracted to conduct both surveys.

The data provided by the 1986 Reserve Components Surveys
made possible xesearch on patterns of previous active and
reserve service, financial issues that would face Guard and
Reserve families during periods of mobilization, reserve
compensation and career intentions, the relationships
between civilian and military occupations for reserve
members, and numerous other topics. The RCS is also the

primary data source for this thesis.

A. SAMPLE POPULATION

The Reserve Components Common Personnel Data System
(RCCPDS) as of 30 October 1985 was used to initially define
the population on which the samples were based. This data
system contained administrative information on all members
of the reserves. The 1986 RCS consisted of Selected Reserve
trained officers and enlisted personnel who had already
completed training. Members in the training pipeline were
not included. As a result, the target population was 9
percent smaller than the total population of the Selected
Reserve. [Hunt, et al. 1986]

Survey packages containing questionnaires and related
materials were mailed directly to approximately 15,000
reserve units in the United States and Puerto Rico. Each
unit had, on average, 7-10 survey participants. The number
of survey participants per unit ranged from one or two to 50

or more.

19

The basic sample selected for the RCS consisted of a

total of 109,067 officer and enlisted personnel.
Individuals who participated in the 1979 Reserve Force (RF)
Follow-up Survey were included which brought the total
number of participants to 120,787.

Data collection began in February, 1986 with the mailing
of the initial notification letters to units containing
sampled individuals. The last questionnaires were not
received for processing until June 1986. Most of the

questionnaires were received in March and April 1986.

B. SURVEY DEVELOPMENT

The Family Survey Coordinating Committee consisted of
representatives from each of the reserve components, the
office of the Deputy Assistant Secretary of Defense
(Guard/Reserve Manpower and Personnel) and technical experts
from DMDC. The Committee identified various subject areas
from previous studies which would be important to reevaluate
as well as new areas for which survey data would be helpful.

After the Committee reached agreement on the content of
the survey questionnaires, DMDC prepared draft
questionnaires. Numerous pretests were conducted in
iterative fashion. These tests were administered to
selected officers, enlisted personnel and spouses. Any
problems or deficiencies found in previous tests were

corrected or modified prior to the next test.

20

As a result of numerous pretests, the questionnaire
underwent considerable refinements. In final form it
contained questions pertaining to military background,

{(l.e., reserve component, paygrade, number of active duty
years, etc.), "future" military plans, military training,
benefits and programs, individual and family
characteristics, civilian work, and military life. All of
the survey respondents were provided with the opportunity to
make additional comments or recommendations on all topics,
whether or not the topic was included in the survey

questionnaires. {Hunt, et al. 1986]

C. RESPONSE RATFS

The most logical approach to assess response rates would
be to compare the number of questionnaires mailed out with
the final numbers received. Table 3.1 provides a breakdown
of response rates by reserve component. This table is
adapted from the 1986 RCS User’s Manual and Codebook. In
the table, the column labeled "Frame Count" refers to the
number of reservists in the population, the column labeled
"Selected" is the number chosen to participate using the
RCCPDS, "Eligible" is the number of reservists still
assigned to the same unit they were assigned in 1985, and
"Responding" is the actual number of reservists who
responded to the survey. The unadjusted rates do not
account for the fact that some individuals who had been
selected for participation from the 30 October 1985

21

TABLE 3.1
1986 RESERVE COMPONENTS SURVEYS RESPONSE RATES
FOR MILITARY MEMBERS, BY RESERVE COMPONENTS
Unadjusted Adjusted
Reserve Frame Response Response
Component Court Selected Eligible Responding Rate Rate

Rank Group «= Officer

USAR 53567 6006 5056 3608 60.1 71.4
USAFR 15710 1809 1611 1331 73.6 82.6
ARNG 42139 4421 3922 2810 63.6 71.6
ANG 13027 1393 1333 1124 80.7 84.3
USMCR 3279 1363 1225 965 70.8 78.8
USNR 22838 2456 2126 1685 68.6 79.3
Subtotal 150560 17448 15273 11523 66.0 75.4

Rank Group = Enlisted

USAR 204321 25391 19704 9640 38.0 48.9
USAFR $7955 5783 4960 3565 61.6 71.9
ARNG 356982 42300 36636 21034 49.7 57.4
ANG 92574 9251 8593 6991 75.6 61.4
USMCR 32853 6562 5414 3333 50.8 61.6
USNR 100653 9898 8132 4893 49 .4 69.2
Subtotal 845338 99185 83439 49456 43.9 59.3

Reserve Components

USAR 257888 31397 24760 13248 42.2 53.5
USAFR 73665 7592 6571 4896 64.5 74.5
ARNG 399121 46721 40558 23844 51.0 58.8
ANG 105601 10644 9926 8115 76.2 81.8
USMCR 36132 7925 6639 4298 54.2 64.7
USNR 123491 12354 10258 6578 §3.2 64.1
Subtotal 995898 116633 98712 60979 $2.3 61.8

Source: Description of Officers and Enlisted Personnel in
the U.S. Selected Reserve: 1986, Defense Manpower Data
Center, Washington, D.C.

22

RCCPDS were no longer members of the unit to which the

questionnaires were sent at the time of actual data

collection. Individuals may have separated, transferred to

an active component, or transferred to another reserve unit.

Upon examination of Table 3.1 it can be seen that the

unadjusted response rates for all components except the Army

are over 50 percent. Since the Army comprises 65 percent of

the total DoD sample selected, its response rate lowers the

overall unadjusted DoD response rate to approximately 52

percent.

Of the 120,787 individuals originally selected, only

102,267 were still in an active drilling status in the

reserves. After adjusting for the "missing" members, the

adjusted response rates, shown in Table 3.1, were

substantially higher tha. the unadjusted response rates.

The unadjusted response rate is calculated by dividing the

responding members by the number of members selected for the

survey. The adjusted response rate is calculated by

dividing the responding members by the number of eligible

members (i.e., those still in a drilling status). The

overall DoD response rate increased to 62 percent. For this

reason, the dataset available for analysis consisted of

60,120 observations from the officer and enlisted

communities of guard and reserve units from the Army, Air

Force, Navy, Marine Corps and Coast Guard. For the purposes

of this study all members of the Coast Guard were dropped

23

from the analysis. The Army Reserve response rates were
considerably lower than the other components perhaps due to

the greater mobility of Army reservists.

D. MODEL SPECIPICATION

In order to eliminate those members who may have been
motivated to enlist in the reserves to avoid the draft, only
members who joined after the end of conscription in 1972
were included in this study. In addition, two other
limitations were placed on the sample. To study the effects
of reserve training on civilian earnings, only non-prior
service members holding full-time civilian jobs were
included. Prior service members were omitted because it is
less likely that they will view the reserves as a source of
valuable training because they will have already been
trained in an occupation through their active duty service.
After imposing these limitations on the models, and allowing
for observations with missing variables, the sample size
decreased to 7,377 observations of enlisted members only.
Finally, officers were deleted from this study because most,
if not all, are college graduates and have far different
earnings capacities than enlisted personnel.

A standard human capital earnings equation was specified
and estimated using Two-Stage Least Squares (2SLS)
techniques. The natural log of the individual's annual
income was used as the dependent variable. The coefficients

of the independent variables could then be interpreted as

24

the percentage change in the income of the individual for a
unit change in the independent variable. The models were

specified as:

LNENGS=a+ZB ,X,+B,+e€ (1)

LNENGS=a+L6B ,X;+B +e (2)
where,

LNENGS = log of annual earnings
= vector of explanatory variables summarized in
Table 3.2
= dummy variable representing those who cite
reserve training as reason for reserve
participation
S = dummy variable representing those who cite
supplementing their income as reason for
reserve participation
e = a random error term that is normally
distributed with with mean zero and a constant
variance

xX;
R

Because members who enlist in the reserves do so
voluntarily, and for various reasons, regression models
estimated by Ordinary Least Squares (OLS) may be affected by
selectivity bias. Members may "self select" themselves to
obtain reserve training to increase their civilian earnings.
Previous studies have shown that the probability of reserve
participation decreases as civilian earnings increase.
People with lower civilian earnings may be more likely to
join the reserves to receive training in order to increase
their potential to obtain better jobs and to increase their
civilian earnings. On the other hand, others may join the
reserves with the intent of supplementing their income. OLS

25

models assume one-way causality in which acquisition of

reserve training causes an increase in civilian earnings.

If two-way causality exists, using OLS to estimate an
earnings model would violate the classical assumption of the
regression model which states that all explanatory variables
are uncorrelated with the error term. However, the models,
as specified, may involve correlation of "reserve training"
(RECTNG) or "Supplement income" (SUPPINC) with the error
term because of the two-way causality between LNENGS and
RECTNG or SUPPINC. This interaction between earnings and
the reason for reserve participation creates a simultaneity
bias. If LNENGS and RECTNG or SUPPINC are simultaneously
determined, the expected values of the OLS-estimated
structural coefficients are not equal to the true
coefficients (b’s). Two-way causality may occur because
RECTNG is a function of all the other explanatory variables
including LNENGS. It is likely the lower one’s annual
income, the higher the probability that one will choose to
receive training in order to increase human capital and
subsequently annual income. This relationship is shown in
equation (3). On the other hand one’s annual income is a
function of many other explanatory variables including the
extent of one’s training. Those who have had formal
training are more likely to have a higher annual income than
those who have not. This relationship is shown in equation

(4).

26

In the empirical section, a test for simultaneity is
performed.

When simultaneity bias is present, the Two-Stage Least
Squares (2SLS) technique can be used to generate efficient
estimates of the parameters in the models. The 2SLS method
eliminates simultaneity bias by substituting an instrumental
variable for the endogenous variable that is correlated with
the error term, in this case, RECTNG or SUPPINC. An
endogenous variable is any variable that is simultaneously
determined with any otier variable. The instrumental
variable must be a good proxy for the endogenous variable
and be independent of the error term.

One’s earnings on the civilian job are hypothesized to
be affected by the desire to receive training in the
reserves, or one’s desire to supplement his or her income,
as displayed in equations (1) and (2). However, it may also
be true that the desire to obtain training or supplement
one’s income will be greater the lower one’s annual income.
This simultaneity suggests the following two-equation system

(written for RECTNG) :

RECTNG=0.+C,LNENGS+2D,Z,+€, (3)

LNENGS=6 5+ZB ,X;,+GRECTNG+E, (4)

The instrumental variables approach involves estimating each

endogenous variable, RECTNG and LNENGS, as a function of the

27

exogenous variables in the system, Z; and X; and
substituting the fitted values of RECTNG and LNENGS back
into equations (3) and (4). In order to test the hypothesis
that reservists join to supplement their income, the same
approach is taken and the variable SUPPINC is substituted in

place of RECTNG in equations (3) and (4).

E. VARIABLE DEFINITION
Table 3.2 contains the definitions of all of the

variables used in the model as well as the occupational
variables utilized. Two dummy variables were created for
the purpose of this study to test the hypothesized effect of
reserve training. These variables were "participate to
receive training" (RECTNG) and "participate to supplement
income” {SUPPINC). Question 26 of the RCS specifically
asks:

People participate in the Guard/Reserve for many reasons.

How much have each of the following contributed to your

most recent decision to stay in the Guard/Reserve?
The survey participants were then given 14 choices of
reasons for staying in the Guard/Reserve including
"obtaining training in a skill that would help get a
civilian job" (26C), "needed the money for basic family
expenses" (261), and “wanted extra money to use now" (26d).
The respondents evaluated each reason on a scale from wajor
contribution to no contribution. Table 3.3 lists the

various responses to question 26 and the percentages of

28

TABLE 3.2
VARIABLE DEFINITIONS

INDIVIDUAL CHARACTERISTICS

AGE
CHILD

EDUC

MARRIED

NONWHITE

Range 17 to 57 years

0 if no dependents

1-10 or more if dependents

years of education completed

range gixth grade through 8+ years of college
1 if married

0 otherwise

1 if Black or Hispanic

OQ if Caucasian

WORK CHARACTERISTICS

WORKFTC

SELFEMPL

1 if working full-time in civilian job
0 otherwise

1 if self employed

0 otherwise

DEPENDENT VARIABLE

LNENGS

VARTABLE
ADMIN

CRAFT
MANAGEP

MINEFM
OPLABOR
OPMACHIN
OPMOVG
PROFESS

SERVICE

Natural logarithm of respondent’s annual
income

CENSUS O PATION TEGORIES
Q PATIONS INCLUDED *

Administrative Support, Clerical excluding
Postal

Construction Workers, Mechanics and Engineers
Administrative, Managerial and Management
related

Mine and Farm Workers

Other Handlers, Helpers and Laborers
Precision Production Workers, Machine
Operators, Assemblers and Inspectors

Motor Vehicle Operators, Other Transportation
and Material Moving Occupations

Professional, Scientific, Specialty, Teachers
Education Administration, Technicians
Protective Services, Postal and Food Services

29

VARIABLE

AGRIMIN

MANUFAC
TRANS P

WSALE
RETAIL
FINANCE
REPSERV
PERSERV
PROSERV
ENTREC
PUBADM

TABLE 3.2 (cont)
CENSUS INDUSTRY CATEGORIES
INDUSTRIES INCLUDED *

Agriculture, Forestry, Fisheries, Mining and
Construction

Manufacturing

Transportation, Communication and other Public
Utilities

Wholesale trade

Retail trade

Finance, Insurance, Real Estate, Business
Repair services

Personal services

Professional services

Entertainment and Recreation

Public Administration

*EBach is coded 1 if the respondent is employed full-time in
that category, 0 otherwise.

30

TABLE 3.3
PERCENT YES RESPONSES
TO QUESTION 26 ON RCS

% MENTIONING
REASON FOR STAYING

QUESTION RESPONSE IN RESERVES
26A Serving the country 88.7
268 Using educational benefits 38.4
26C Obtaining training in a skill 41.9

that would help get a
civilian job

26D Serving with people in the unit 68.8

26E Getting credit towards Guard/ 55.5
Reserve retirement

26F Promotion opportunities 63.4

26G Opportunity to use military B3<3
equipment

26H Challenge of military training 70.7

261* Needed the money for basic 57.0
family expenses

26T* Wanted extra money to use now 60.4

26K Saving income for the future 43.7

26L Travel/"get away" opportunities 53.9

26M Just enjoyed the Guard/Reserve 59.6

26N Pride in my accomplishments in 77.5

the Guard/Reserve
*These were combined to form the variable SUPPINC.
NOTE: Responses were coded as "yes" if the respondent
marked the reason as a major or moderate contribution to
his/her decision to stay in the Guard/Reserve.

Sample = 7,377

31

respondents who stated whether each was a major or moderate

contribution to their decision to stay and those who stated

these were a minor or no contribution to their decision to

stay. For example, if respondents stated answer (26C) was a
major or moderate contribution then the RECTNG variable was
coded 1, otherwise it was coded 0. If the respondent stated
that either answers 26I or 267 were a major or moderate
contribution to their decision to stay then the SUPPINC
variable was coded 1, otherwise it was coded 0. In this way
respondents could be sorted according to whether they
desired skill training or simply wanted to supplement their
income.

A major problem with this type of questionnaire is that
there was no "either/or" answer. A respondent was free to
mark all answers that applied. Many reservists who stated
they desired training to obtain a civilian job also stated
that they wished to supplement their income. For this
reason, another dummy variable "Receive training only"
(RECONLY), was created. RECONLY was coded 1 when RECTNG
equalled 1 and SUPPINC equalled 0; otherwise RECONLY was
coded 0. Alternatively, there may be members who wish to
supplement their income but not receive training to obtain a
civilian job. Another dummy variable, "Supplement income
only" (SUPONLY), was created. SUPONLY was coded 1 when
SUPPINC equalled 1 and RECTNG equalled 0; otherwise SUPONLY

was coded 0. RECONLY and SUPONLY were then substituted in

32

place of RECTNG in equations (3) and (4) to test these two
hypotheses. It was anticipated that this would provide a
clearer classification of those members who affiliated to
receive training but not to supplement their income and
those who affiliated to supplement their income but not to
receive training.

The major focus of this thesis is to specify and
estimate equation (4). The specification followed the
conventional human capital earnings model in the literature
{Berndt, 1991]. Thus, the log of annual earnings (LNENGS)
was specified as a function of age (AGE), the square of age
(AGESQ), number of children (CHILD), marital status
(MARRIED), race (NONWHITE), education (EDUCATION), and
several occupational dummy variables as shown in equation

(5).

«+P, AGE+B,AGESQ+B,CHILD+B ,MARRIED+B ,NONWHITE*+
§ ,.SELFEMPL+$,ADMIN+P ,CRAFT+B,MANAGER+B , .MINEFM+
B,, OPLABOR+B,,OPMACHIN+B,,OPMOVG+B,,PROFESS+ '9)
B,,SERVICE+e

LNENGS=

The next chapter presents 2SLS estimates of the specified

model.

33

Iv. EMPIRICAL RESULTS

A. DESCRIPTIVE STATISTICS

Table 4.1 presents the descriptive statistics for the
entire sample of enlisted respondents who reported positive
annual income (N=32,482) as well as the descriptive
statistics for the restricted sample used to estimate the
basic model (N=7,377). The average annual income for the
full sample is $23,284, while the average annual income for
the restricted sample is $16,807. This large disparity in
annual income can be attributed to the large difference in
the mean ages of the two samples: The mean age of those in
the restricted sample is approximately ten years younger
than those in the full sample. This age difference results
from limiting the sample to reservists who affiliated after
1972 and who have no prior active duty. Human capital
theory hypothesizes that the older one is, the higher his or
her position on the earnings curve. The longer one has been
in the labor market, the more experience he or she will have

and the higher his or her income will be, provided there are

few prolonged periods of unemployment or numerous

occupational shifts.
Members in the full sample have approximately six months
more education than those in the basic model, a difference

which also is attributable to their disparity in ages. of

34

TABLE 4.1
DESCRIPTIVE STATISTICS FOR
THE FULL SAMPLE AND THE RESTRICTED SAMPLE

FULL SAMPLE RESTRICTED SAMPLE*
N = 32,482 N = 7,377

VARIABLE MEAN STANDARD MEAN STANDARD

DEVIATION DEVIATION
AGE 34.38 9.42 24.84 4.48
CHILD 1.49 1.42 0.84 1.20
EDUC 12.98 1.88 12.42 £252
MARRIED 0.70 0.46 0.48 0.50
NONWHITE 0.24 0.43 0.26 0.44
SELFEMPL 0.07 0.25 0.04 0.19
ADMIN 0.06 0.23 0.05 0.22
CRAFT 0.20 0.40 0.19 0.40
MANAGER 0.10 0.30 0.06 0.23
MINEFM 0.02 0.14 0.03 0.17
OPLABOR 0.06 0.24 0.09 0.29
OPMACHIN 0.14 0.34 0.18 0.38
OPMOVG 0.06 0.24 0.08 0.28
PROFESS 0.14 0.34 0.08 0.27
SERVICE 0.14 0.34 0.15 0.35
INCANN 23283.95 15092.25 16807.05 13277.60

*This sample restricted to non-prior service (NPS), full-
time civilian (FTC) job, and affiliation with reserves
during AVF period (after 1972).

35

the respondents in the full sample, 24 percent were either

black or hispanic compared to 26 percent in the basic model.
There were fewer self employed workers in the restricted
sample, as one would expect given their relative youth.
Also, fewer of the restricted sample members were managers,
or worked in the more skilled occupations such as
administration, craft, manager, or professional.

Conversely, workers in the restricted sample were more
likely to work in semi-skilled or non-skilled occupations
such as mine or farm worker, basic laborer, machine
operator, transportation, or service. These differences
support the theory that younger workers tend to be employed
in less skilled occupations and earn lower wages than older,
more experienced workers who tend to be employed in skilled
occupations.

The three restrictions placed on the basic model--non-
prior service, full-time civilian job, and affiliation with
the reserves during the AVF period--all contributed to the
lower mean age as well as the lower mean annual income of
reservists in the restricted sample. In the full sample, 54
percent or respondents had prior military service before
affiliating with the reserves. Studies have shown that
civilians who were trained in the military will have a
steeper earnings curve than theizr civilian counterparts

following an initial period of lower earnings immediately

36

following separation from the military (Mehay, 1992). The
disparities mentioned in the mean ages, educational levels,
and occupations are significant enough to cause the $6,477
difference in the mean annual income of the two samples.
Tables 4.2 through 4.5 calculate t-tests of differences
in the means of the characteristics of members who were
categorized by the four dummy variables that represent
reasons for reserve participation. These four variables
were "receive training" (RECTNG), "supplement income”
(SUPPINC), "receive training only" (RECONLY), and
"supplement income only" (SUPONLY). The coding of these
variables was described in the previous chapter. The
results in Table 4.2 reveal that members who affiliated to
receive training (41% of the restricted sample) were
younger, less educated and earned $996 less than members who
were classified as affiliating for cther reasons. All of
these differences are statistically significant at the one
percent level. Of those who affiliated to receive training,
32 percent were black or hispanic, versus 22 percent of
those who joined for other reasons. This is statistically
significant at the one percent level. Recall that the
entire sample is 24 percent black or hispanic. Of the
sample of 7,377 reservists, 32 percent of those in the
restricted sample who affiliated to receive training were
black. This suggests that the rate at which minorities join

the reserves to receive training is disproportionate to the

37

TABLE 4.2
T-TEST OF DIFFERENCES IN MEANS

RECTNG = 1 RECTNG = 0

STANDARD STANDARD TEST
VARIABLE MEAN DEVIATION MEAN DEVIATION STAT
AGE 24.55 4.293 25.04 4.601 4.663%
CHILD 0.85 1.211 0.84 1.194 0.343
EDUC 12.23 1.355 12.53 1.613 6.899*
MARRIED 0.45 0.497 0.49 0.500 3.355*
NONWHITE 0.32 0.466 0.22 0.413 -9.750*
SELFEMPL 0.03 0.179 0.04 0.202 2.108**
ADMIN 0.06 0.229 0.05 0.216 -1.242
CRAFT 0.19 0.391 0.20 0.399 1.100
MANAGER 0.04 0.198 0.06 0.247 4.477%
MINEFM 0.03 0.168 0.03 0.173 0.384
OPLABOR 0.09 0.293 0.09 0.290 -0.199
OPMACHIN 0.17 0.374 0.19 0.390 2.005**
OPMOVG 0.10 0.294 0.08 0.265 -2.995*
PROFESS 0.08 0.267 0.08 0.264 -0.332
SERVICE 0.16 0.369 0.14 0.343 -3.104*

INCANN 16212.95 13308.82 17208.82 13234.29 3.173*

N=3, 026 N=4,359

* Significant at the 0.01 level
** Significant at the 0.05 level

SUPPINC = 1
STANDARD
VARIABLE MEAN DEVIATION MEAN
AGE 25.09 4.605 23.91
CHILD 0.89 1.208 0.66
EDUC 12.35 1.489 12.69
MARRIED 0.49 0.500 0.41
NONWHITE 0.27 0.442 0.23
SELFEMPL 0.04 0.192 0.04
ADMIN 0.05 0.220 0.06
CRAFT 0.20 0.399 0.18
MANAGER 0.05 0.224 0.06
MINEFM 0.03 0.174 0.03
OPLABOR 0.09 0.291 0.09
OPMACHIN 0.18 0.386 0.17
OPMOVG 0.09 0.279 0.08
PROFESS 0.07 0.255 0.10
SERVICE 0.15 0.357 0.14
INCANN 16907.70 13370.42 16405.30
N=5, 813
* Significant at the 0.01 level

TABLE 4.3
T-TEST OF DIFFERENCES IN MEANS

belied Significant at the 0.05 level

kkk

39

Significant at the 0.10 level

SUPPINC = 0

STANDARD

DEVIATION
3.866 -9
1.157 -6
1,595 7
0.493 “5
0.424 -2
0.198 0
0.230 0
0.381 -2
0.245 1
0.163 -0
0.293 0
0.376 -1
0.272 -0
0.298 3
0.343 -

12902.97 -1.

N=1,572

TEST
STAT

TABLE 4.4
T-TEST OF DIFFERENCES IN MEANS

RECONLY = 1 RECONLY = 0

STANDARD STANDARD TEST
VARIABLE MEAN DEVIATION MEAN DEVIATION STAT
AGE 23.43 3.316 24.92 4.531 6.560*
CHILD 0.62 1.055 0.86 1.208 3.906*
EDUC 12.54 1.304 12.42 1.529 -1.552
MARRIED 0.37 0.483 0.48 0.500 4.631*
NONWHITE 0.25 0.433 0.26 0.439 0.515
SELFEMPL 0.04 0.199 0.04 0.193 -0.254
ADMIN 0.07 0.260 0.05 0.220 -1.931%**
CRAFT 0.17 0.377 0.20 0.397 1.214
MANAGER 0.04 0.199 0.06 0.230 1.300
MINEFM 0.02 0.146 0.03 0.173 1.049
OPLABOR 0.11 0.309 0.09 0.290 -0.904
OPMACHIN 0.16 0.367 0.18 0.384 1.084
OPMOVG 0.10 0.299 0.08 0.276 Heh 17
PROFESS 0.12 0.326 0.07 0.261 -3.531*
SERVICE 0.12 0.326 0.15 0.356 1.547
INCANN 15363.16 9701.40 16886.14 13451.04 2.269**

N=414 N=6971

* Significant at the 0.01 level
** Significant at the 0.05 level

40

TABLE 4.5
T-TEST OF DIFFERENCES IN MEANS

SUPONLY = 1 SUPONLY = 0
STANDARD STANDARD
VARIABLE MEAN DEVIATION MEAN DEVIATION
AGE 25.39 4.744 24.42 4.227
CHILD 0.90 1,189 0.80 1.209
EDUC 12.45 1.580 12.41 1.468
MARRIED 0.52 0.500 0.45 0.497
NONWHITE 0.21 0.411 0.29 0.456
SELFEMPL 0.04 0.204 0.03 0.184
ADMIN 0.05 0.216 0.05 0.226
CRAFT 0.21 0.405 0.19 0.388
MANAGER 0.06 0.242 0.05 0.217
MINEFM 0.03 0.175 0.03 0.169
OPLABOR 0.09 0.292 0.09 0.291
OPMACHIN 0.139 0.393 0.17 0.376
OPMOVG 0.08 0.267 0.09 0.286
PROFESS 0.07 0.255 0.08 0.273
SERVICE 0.13 0.341 0.16 0.363
INCANN 17364.72 13001.51 16369.30 13462.65
N=3201 N=4184
* Significant at the 0.01 level

kk
Kk

41

Significant at the 0.05 level
Significant at the 0.10 level

TEST
STAT

-9.
-3

275*

-580*
.176
-969*
.739*
-830%**
999
-289**
-380**
535
-085
-395**
-911***
-824ee*
-673*

196*

rate for the overall population of minorities in the
reserves.

Table 4.3 illustrates the differences in the mean
characteristics of those members who affiliated to
supplement their income (79 percent of the restricted
sample) versus those who did not. Average annual income is
$502 greater than for those who did not join to supplement
their income. While this difference is not statistically
significant, the higher earnings of those who joined to
supplement their income was expected. Those who affiliated
to supplement their income were generally older and had more
children; on the other hand they were less educated than
those members who did not affiliate to supplement their
income and 27 percent were non-white. Only three of the
differences in means of the occupation variables were
statistically significant at conventional levels. These
occupations were CRAFT, MANAGER, and PROFESS. Of these,
PROFESS had the largest disparity, with 7 percent of those
who joined to supplement their income coming from a
professional occupation compared with 10 percent who did not
join to supplement their income. It was expected that fewer
professionals would require a supplement to their income.

As stated in Chapter III, an attempt was made to provide
a better breakdown of those reservists who affiliated to
receive training but not to supplement their income

(RECONLY) and those who affiliated to supplement their

42

income but not to receive training (SUPONLY). Table 4.4
presents the results of the t-test of the differences in
means of the characteristics of those members in the RECONLY
category. Only 5.6 percent of the restricted sample
affiliated to receive training only. Those who joined to
receive training only were younger and had much lower annual
income than those who did not join to receive training only.
These differences are significant at the 1 and 5 percent
levels, respectively. Note, however, that members who
affiliated to receive training only were slightly more
educated than those who did not. This was not expected but
it is statistically significant at only the 12 percent
level. A smaller percentage of those who affiliated to
receive training only were non-white or married compared to
those who did not join to receive training. The mean of the
occupational variable PROFESS was unexpected when compared
to its mean in the RECTNG model: 12 percent of those
members who affiliated to receive training only were
professionals compared with only 8 percent in the RECTNG
model (Table 4.2). While this difference is statistically
Significant at the 1 percent level, this only applies to 50
of the respondents.

Table 4.5 presents a t-test of the differences of the
means of the variables in the SUPONLY model. Over 41
percent of the respondents in the sample population stated

they joined the reserves to supplement their income instead

43

of to receive training. Their mean annual income was almost
$1,000 more than those who did not join to supplement tuisir
income. Those who affiliated to supplement their income
only were older, had more children, and had slightly more
education than those who did not join to supplement their
income. More of them were married (52%) while less came
from a minority (21%). In this classification, most of the
occupational variables were statistically significant.

There was no standard breakdown between skilled and non- or
semi-skilled occupations. Those members who wished to
supplement their income were represented in all occupational

categories.

B. REGRESSION RESULTS

Tables 4.6 and 4.7 present the results for four
different OLS regression models, each model using one of the
four constructed dummy variables representing the reason for
reserve participation--RECTNG, SUPPINC, RECONLY, and
SUPONLY. It is apparent that most of the coefficients of
the independent variables in each model are statistically
significant at conventional (1, 5, or 10 percent) levels.
In all four models the coefficient of the NONWHITE variable
is negative. This would suggest that being a minority is
associated with lower annual earnings. This effect was
Statistically significant at the 1 percent level. The
coefficients of AGE and AGESQ indicate that the age-earnings

profile for reservists is fairly steep. Married persons and

44

TABLE 4.6
REGRESSION RESULTS USING
ORDINARY LEAST SQUARES

RECEIVE TRAINING SUPPLEMENT INCOME
MODEL MODEL

VARIABLE COEFFICIENT T-RATIO COEFFICIENT T-RATIO
INTERCEPT 6.850 36.474* 6.818 36.296*
AGE 0.137 10.675* 0.137 10.717*
AGESQ -0.002 ~7,813* -0.002 -7.828*
CHILD 0.005 0.641 0.004 0.596
EDUC 0.024 4.223% 0.025 4.359*
MARRIED 0.120 6.589* 0.122 6.673*
NONWHITE -0.098 ~5.291* -0.104 -5.€40*
SELFEMPL 0.116 2.856* 0.118 2.886*
ADMIN 0.013 0.290 0.611 0.255
CRAFT 0.063 1.955** 0.063 1.977**
MANAGER 0.192 4.479* 0.195 4.564*
MINEFM -0.193 -3.680* -0.192 -3.662*
OPLABOR -0.026 -0.724 -0.026 -0.694
OPMACHIN 0.086 2.652* 0.087 2.699*
OPMOVG 0.061 1.602 0.058 1.541
PROFESS 0.149 3.741* 0.145 3.650*
SERVICE -0.004 -0.121 -0.005 -0.156
RECTNG -0.047 -2.395* a 2%
SUPPINC iia -- -0.014 -0.728

N=7,377 N=7, 377

F STATISTIC 61.721 F STATISTIC 61.178

R- SQUARE 0.1248 R-SQUARE 0.1238

ADJ R-SQUARE 0.1228 ADJ R-SQUARE 0.1217

* Significant at the 0.01 level
** Significant at the 0.05 level

45

VARIABLE

INTERCEPT
AGE
AGESQ
CHILD
EDUC
MARRIED
NONWHITE
SELFEMPL
ADMIN
CRAFT
MANAGER
MINEFM
OPLAP OR
OPMACHIN
OPMOVG
PROFESS
SERVICE
RECONLY
SUPONLY

TABLE 4.7

REGRESSION RESULTS USING
ORDINARY LEAST SQUARES

RECEIVE TRAINING

ONLY MODEL

COEFFICIENT

OO0O000C O90 000 00 on
i ei %
N

N=7,377

F STATISTIC

R-SQUARE

ADJ R-SQUARE

T-RATIO

36.319*
10.698*

61.143
0.1238
0.1217

* Significant at the 0.01 level
** Significant at the 0.05 level
*** Significant at the 0.10 level

46

SUPPLEMENT INCOME

ONLY MODEL
COEFFICIENT T-RATIO
6.81 36.357*
0.14 10.628*
-0.00 ~7.786*
0.00 0.610
0.03 4.441*
0.12 6.596*
-0.10 -5.406*
0.12 2.869*
0.01 0.260
0.06 1.929***
0.20 4.521*
-0.19 -3.692*
-0.03 -0.727
0.09 2.651*
0.06 1.574
0.15 3.717%
-0.00 -0.150
0.04 2.349**
N=7,377
F STATISTIC 61.513
R-SQUARE 0.1244
ADJ R-SQUARE 0.1224

the self employed tend to earn more than their peers. The
coefficient on EDUC indicates that over a 2 percent annual
return on each additional year of education is obtained.
Over half of the occupational dummy variables are
statistically significant.

The negative coefficient on RECTNG would suggest that
receiving training in the reserves would have a negative
impact on annual income. This is a surprising result. Most
training, no matter the source, would be an investment in
human capital, resulting in some improvement in earnings.

In the "supplement income model" age, age squared,
education, marital status, and race/ethnicity were all
significant at the 1 percent level. The negative sign on
the coefficient of SUPPINC again, however, was unexpected.
Taken at face value this suggests that reservists who
affiliate to supplement their income end up earning less
than those who join for other reasons.

The results of the RECONLY and SUPONLY models are
presented in Table 4.7. The coefficients of the variables
are similar to the coefficients of the variables in the
RECTNG and SUPPINC models, with minor differences. This is
probably due to the fact that many of the observations in
the RECTNG model are also found in the RECONLY model. The
same is true of the SUPPINC and SUPONLY models. Of all the
personal characteristics in the four models, NONWHITE is the

only one to have a negative impact on annual income.

47

Using OLS for all four models resulted in estimates of
the coefficients thxt were not supportive of the initial
hypothesis, namely, that those who initially join the
reserves to receive training would experience an increase in
annual income. The negative coefficient on the RECTNG
variabie and the extremely small coefficients on SUPPINC,
RECONLY, and SUPONLY made it obvious that either these
variables are poor measures of actual intent or training
received or the hypothesis is rejected. The methodology
employed in this thesis was to specify and estimate the
models and not to re-estimate them afce: eliminating
variables whose t-ratios were statistically insignificant.
The model was determined to be appropriate as specified and
no attempt was made to alter it by eliminating what were
important variables. However, as mentioned in Chapter III,
the OLS coefficients may be biased. Simultaneity bias may
affect the estimates of the models which would result in
estimated structural coefficients not equal to the true
coefficients (b’s).

In testing the model for simultaneity two procedures
were used, the first procedure involved estimating log-
earnings (LNENGS) as a function of all of the exogenous
variables in the system. A predicted value of LNENGS was
generated from this OLS equation. This predicted value
(YHAT) was then used as an explanatory variable in a logit

model of RECTNG. The test consists of examining the

48

significance of the coefficients of YHAT in the logit model
of RECTNG. Table 4.8 presents the results from the logit
estimation of RECTNG. The negative coefficient on YHAT,
coupled with its significance at the 2 percent level,
reveals that LNENGS is simultaneously determined with the
probability of affiliating with the reserves to receive
training. In other words, the lower one’s annual income,
the more likely one is to join the reserves to receive
training in order to increase their potential to obtain
better jobs and to increase their civilian earnings.

To rid the models of simultaneity bias, 2SLS estimating
techniques were undertaken. Tables 4.9 and 4.10 present the
2SLS estimates for the four models. Age, Education, marital
status, and racial group were still significant at the 1 or
5 percent level in all of the models except SUPPINC. In
both the RECTNG (Table 4.9) and RECONLY (Table 4.10) models
the coefficents on these variables have now become positive,
which indicates that those who joined the reserves for the
training benefits have higher income on their civilian job.
Of major importance is the fact that the RECTNG variable in
the "receive training" model became positive and significant
at the 1 percent level. The RECONLY variable was
significant at only the 17 percent level.

The coefficients on the SUPPINC and SUPONLY variables in
their respective models are both negative with the SUPONLY

variable being significant at the 1 percent level. This

49

VARIABLE

INTERCEPT
EDUC
MARRIED
NONWHITE
SELFEMPL
ADMIN
CRAFT
MANAGER
MINEFM
OPLABOR
OPMACHIN
OPMOVG
PROFESS
SERVICE
YHAT

* Significant at the 0.01 level
** Significant at the 0.05 level
*** Significant at the 0.10 level

TABLE 4.8
LOGIT ESTIMATES ON
RECEIVE TRAINING (RECTNG)

COEFFICIENT

50

STANDARD

ERROR

105
.018
.058
-055
-130

ooooo0oo0o°cneooo00ooorF
ht
a
=)

WALD
CHI-SQUARE

84.26*

OHM wWHOoOOMOOOD
i)
uy

TABLE 4.9

REGRESSION RESULTS FOR RECEIVE TRAINING AND SUPPLEMENT

VARIABLES COEFFICIENT
INTERCEPT 5.774
AGE 0.146
AGESQ -C.002
CH1iLD -0.005
EDUC 0.056
MARRIED Q.155
NONWHITE -0.256
SELFEMPL 0.157
ADMIN -~0.027
CRAFT 0.073
MANAGER 0.293
MINEFM -0.184
OPLABOR -0.010
OPMACHIN 0.124
OPMOVG 0.002
PROFESS 0.069
SERVICE -0.043
RECTNG 1.179
SUPPINC :
N = 7,377
F STATISTIC

INCOME MODELS USING TWO-STAGE LEAST SQUARES

RECEIVE TRAINING

MODEL

R-SQUARE
ADJ R-SQUARE

34.826

T RATIO

12.937*
8.408*
-5.996*
-0.497
4.202*
5.716*
-4.295*
2.783*
-0.448
1.695***
4.382*
-2.633*
-0.206
2.731*
0.039
1.147
=0 5915

2.798*

* Significant at the 0.01 level
** Significant at tne 0.05 level
*** Significant at the 0.10 level

51

SUPPLEMENT INCOME

COEFFICIENT

oo0o0o0o0oo0co0g CcCOoOcCo eo,
is oe
od
a

-0.956

N = 7,377

F STATISTIC
R- SQUARE
ADJ R-SQUARE

T RATIO

14.770*
7.612*
-6.710*
.009
.065
-801*
-567*
.148**
~227
.158**
.524*
.462**
-432
.540**
.296
.261**
-413

'
Moo

‘
ONKFNONWNON W

-437

1
oo

46.482
0.0970
0.0949

TABLE 4.10

REGRESSION RESULTS FOR RECEIVE TRAINING ONLY AND SUPPLEMENT
INCOME ONLY MODELS USING TWO-STAGE LEAST SQUARES

RECEIVE TRAINING

ONLY MODEL
VARIABLE COEFFICIENT T RATIO
INTERCEPT 6.592 24 .966*
AGE 0.147 9.209*
AGESQ -0.002 -7.113*
CHILD 0.004 0.472
EDUC 0.020 2.794*
MARRIED 0.141 5.626*
NONWHITE -Q0.105 -5.087*
SELFEMPL 0.105 2.229**
ADMIN -0.027 -0.477
CRAFT 0.067 1.848***
MANAGER 0.208 4.263*
MINEFM -0.170 -2.777*
OPLABOR -0.038 -0.892
OPMACHIN 0.088 2.418**
OPMOVG 0.032 0.670
PROFESS 0.086 1.356
SERVICE 0.004 0.116
RECONLY 1.482 1.351
SUPONLY =.= “o>
N = 7,377

F STATISTIC 48.926

R-SQUARE 0.1015

ADJ R-SQUARE C.0995
* Significant at the 0.01 level

** Significant at the 0.05 level
*e* Significant at the 0.10 level

52

SUPPLEMENT INCOME
ONLY MODEL

COEFFICIENT T RATIO

!

'

t

4
CODDDDDDDDOOOOGOg
°
OQ
ron

.016
046

‘
Oo

-2.809

N = 7,377

F STATISTIC

R-SQUARE

ADJ R-SQUARE

15.449*
5.258*
-4.146*

'

DOAONDOWHF OF WW bh
oOo
~]
D
*

a

-0.587

12.006
0.0270
0.0247

reinforces the earlier result that joining the reserves to
supplement income does not benefit one in terms of annual
income. Most of the same variables were statistically
Significant at conventional levels in all of the models

specified whether using OLS or 2SLS.

53

V. CONCLUSIONS AND RECOMMENDATIONS

A. CONCLUSIONS

This thesis tests the alternative hypothesis that a
positive relationship exists between participation in the
reserves to receive training and increased benefits and
wages on one’s civilian job for some reservists. The null
hypothesis is that no such relationship exists, or, that it
exists for relatively few reservists. If so, then reserve
participation is mainly a form of moonlighting with few
derivative benefits to the individual or to society.

The mean statistics of this thesis showed that only 5
percent of the sample population of 7,377 reservists
affiliated solely to receive training while over 41 percent
affiliated solely to supplement their income. Those who
affiliated to receive training had a higher annual income
than those who affiliated for other reasons except those
members who affiliated only to supplement their income. In
the multivariate model estimated by OLS the negative
coefficients on the RECTNG and SUPPINC variables would
suggest that affiliating with the reserves to receive
training or to supplement annual income would reduce annual
income. It was hypothesized however, that a single equation
model of the training-earnings relationship was

inappropriate and that the true relationship was

54

simultaneous in nature. Therefore, a test for simultaneity
was conducted with the results shown in Table 4.8. The
results showed that there was simultaneity in the OLS models
and the 2SLS estimating technique ws consequently applied to
the regression equations.

The 2SLS estimates support the hypothesis that receiving
training in the reserves results in an increase in civilian
benefits and wages. The coefficient on the RECTNG variable
became positive and highly significant while the SUPPINC
coefficient remained negative. Therefore, the null
hypothesis was rejected. We can conclude that reserve
training does appear to provide important benefits to some
enlistees, namely those who are motivated to seek skill
training that can be used on their civilian job or used to
find a better civilian job.

One major issue associated with attempting to estimate
the effect of reserve training on civilian wages is the
possibility that the system of equations defining the model
is recursive rather than simultaneous. In Chapter III,
equation (3) defines RECTNG as a function of several
explanatory variables including the log of current annual
income (LNENGS). In actuality, the probability of
affiliating to receive training may be a function of one’s
annual income in the previous year. If so, Equations (3)

and (4) in Chapter III should be rewritten as follows:

55

RECING,,=«,+C,LNENGS,_,+2D,Z,+€, (1)

LNENGS,=B, +2 ,X,+GRECTNG, +p, (2) :

Although this system of equations might seem simultaneous,
it is actually recursive, and the variables RECTNG, and
LNENGS, are sequentially determined. Values for annual
income in the previous year (LNENGS,.,) would allow us to
solve directly for the probability of affiliating to receive
training (RECTNG,). Then, knowing RECTNG, would allow us to
solve for annual income in the current year.

In any recursive system of this sort, Ordinary Least
Squares (OLS) is the appropriate estimation technique
(Pindyck et al. 1991). OLS is appropriate for equation (1)
because LNENGS,., is predetermined and therefore
uncorrelated with the error term in equation (1). OLS is
also appropriate for equation (2) because RECTNG, is
uncorrelated with the error term in equation (2). However,
the 1986 Reserve Components Surveys did not contain data on
members’ annual incomes prior to affiliation with the
reserves. This prevented specification of the system of
equations as recursive and made it necessary to specify
equations that were simultaneously determined. This
required the application of Two-Stage Least Squares

estimating techniques to estimate the model.

56

Another problem encountered in this thesis was the fact

that reserve affiliation is voluntary and members self-
select themselves to join. People who have a lowe: annual
income than their cohorts are more likely to affiliate with
the reserves to receive training to increase their human
capital and subsequently their annual income. The effects
of reserve training on annual income may be biased downward
by a member's already depressed annual income below that of

his cohorts.

B. RECOMMENDATIONS

Future Reserve Components Surveys should include
questions pertaining to a members status prior to
affiliation with the reserves. Annual income, type of
occupation, marital status, number of children, and years of
completed education, and unemployment and employment status
prior to the member’s enlistment in the reserves, would be
valuable information in measuring the effects of reserve
training on civilian annual income. The data available in
the 1986 Reserve Components Surveys simply were not
sufficient enough to support an empirical analysis of the

economic motives for reserve participation.

57

REFERENCES

Berndt, Ernst R., (1991) The Practice of Econometrics:
Classic and Contemporary, Massachusetts Institute of
Technology and The National Bureau of Economic Research.

Burright, B., Grissmer, D., Doering, Z.,(1982) A Model of
Reenlistment Decisions of Army National Guardsmen, Santa
Monica, CA: Rand Corporation.

Gorman, L., Thomas, G., (1991) "Enlistment Motivations of
Army Reservists: Money, Self-Improvement, or
Patriotism?" Armed Forces and Society, Vol 17 No.4,
Summer 1991.

Grissmer, D., Buddin, R., Kirby, S.,(1989) Improving Reserve
Compensation: A Review of Current Compensation and
Related Personnel and Training Readiness Issues, Santa
Monica CA: Rand Corporation.

Hunt, P., Sparks, M., Simpson, J., Doering, Z., Mahoney, B.,
1986 Reserve Components Survey User’s Manual and
Codebook, prepared for Defense Manpower Data Center,
Arlington, VA.

Mehay, S., (1992) “Post-Service Earnings of Volunteer-Era
Veterans: Evidence From The Reserves" U.S. Naval
Postgraduate School, Monterey, CA.

Mehay, S., (1991) “Reserve Participation Versus Moonlighting:
Are They The Same?" Defence Economics, 1991, Vol.2.

Mehay, S.,{1990) "Determinants of Enlistments in the U.S.
Army Reserve" Armed Forces & Society, Vol.16 No.3,
Spring 1990.

Pindyck, R., Rubinfeld, D., (1991) Econometric Models &
Economic Forecasts, 3rd Ed., McGraw-Hill Inc.

Regets, M., (1990) "Military Reserves as Compensated Leisure:
A Peculiar Case of Labor Supply" U.S. Commission on
Civil Rights, Washington, D.C.

Shishko, R., Rostker, B., (1976) "The Economics of Multiple
Job Holding", The American Economic Review, June 1976.

Sullivan, J., (1985) Motivations for First Term Reserve
Reenlistment, M.S. Thesis, Naval Postgraduate School,
Monterey, CA.

58

INITIAL DISTRIBUTION LIST

Defense Technical Information Center
Cameron Station
Alexandria, Virginia 22304-6145

Library, Code 0142
Naval Postgraduate School
Monterey, California 93943-5002

LT John A. McGuire
207 Via Del Aqua
Clewiston, Florida 33440

Professor Stephen L. Mehay, Code AS/MP
Department of Administrative Sciences
Naval Postgraduate School

Monterey, California 93943-5100

Professor Gregory G. Hildebrandt, Code AS/HI
Department of Administrative Sciences

Naval Postgraduate School

Monterey, California 93943-5100

59