ABSTRACT 
 
 
 
Title of Dissertation:          THE WORLD WAR II VETERAN ADVANTAGE?  
A LIFETIME CROSS-SECTIONAL STUDY OF SOCIAL 
STATUS ATTAINMENT 
  
    Irving Smith III, Doctor of Philosophy, 2007 
 
 
 
Dissertation directed by:  Professor David R. Segal  
    Department of Sociology  
 
 
 
   The impact of military service on the social status attainment of 
World War II veterans has been studied since the 1950s; however, the research has failed 
to come to any consensus with regard to the level of their attainment. Analyses have 
generally focused on cross-sectional data or longitudinal data without considering the 
effects of military service over the life course. In this study I argue that World War II 
veterans had greater social attainment over their lifetimes; that black World War II 
veterans attained more than white World War II veterans relative to their non-veteran 
peers; that veterans who served in the latter years of the World War II mobilization 
attained more than those who served in the earlier years; and that veterans born in cohorts 
with large proportions of veterans attained more than veterans born to cohorts with 
smaller proportions of veterans. Social status is measured in terms of education, income, 
and Duncan Socio-Economic Index. In order to test these hypotheses I use data from the 
1950 through 2000 Public Use Microdata Sample. Military service clearly afforded 
veterans significant advantages through their early and middle working years; however, 
their non-veteran peers eventually did catch up. Black veterans attained more social 
 
   
status than their non veteran peers throughout their lives. Furthermore, the magnitude of 
the difference in social status attainment is greater for black veterans relative to their 
non-veteran peers than the difference for white-veterans relative to their non-veteran 
peers until very late in the life course. Additionally, peak mobilization phase veterans 
receive advantage although it is relatively short lived.
 
   
THE WORLD WAR II VETERAN ADVANTAGE? 
A LIFETIME CROSS-SECTIONAL STUDY OF SOCIAL STATUS ATTAINMENT 
 
by 
 
 
Irving Smith III 
 
 
 
 
 
Dissertation submitted to the Faculty of the Graduate School of the 
University of Maryland, College Park in partial fulfillment 
of the requirements for the degree of 
Doctor of Philosophy  
2007 
 
 
 
 
 
Advisory Committee: 
 
 Professor David R. Segal, Chair 
 Professor William W. Falk 
 Professor Meyer Kestnbaum 
 Professor Seth Sanders 
 Professor Mady W. Segal
 
   
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
?Copyright by 
Irving Smith III 
20007
ACKNOWLEDGEMENTS 
 
 
 I would like to acknowledge the members of my dissertation committee for their 
guidance and expertise on this project.  I would especially like to thank David Segal, my 
coach, advisor, teacher, cheerleader, and most importantly friend.  Through his 
inspiration, leadership, and mentorship I have learned to be a better student, scholar, and 
teacher. Mady Segal and Meyer Kestnbaum and have also had a profound affect on me, 
this research, and other work during my graduate career.  They selflessly dedicated hours 
of their time in my personal and professional development and their investment in my 
training and scholarship are clearly reflected in this work.  In addition, Bill Falk and Seth 
Sanders provided me with sound advice and timely recommendations that ultimately 
made this dissertation stronger than it otherwise would have been. 
  The development of this dissertation came in several phases. There were 
many professors including David Segal, Mady Segal, Myer Kestnbaum, John Iceland, 
and Bart Landry who encouraged and sparked my interest in stratification and social 
status attainment. Initial drafts of this study benefited from critiques by Roberto Patricio 
Korzeniewicz and my classmates in SOCY 701, the Work-in-Progress seminar organized 
and lead by Meyer Kestnbaum, and the military sociology seminar group. Within this 
special group folks including Brian Reed, Kirby Bowling, Ryan Kelty, Alex Bierman, 
Kimberly Bonner, Rachel Nobel, and Yuko Whitestone provided me with a great deal of 
feedback and support. Finally, the staff and faculty at West Point including COL Thomas 
Kolditz, Morten Ender, and Ryan Kelty who provided me with support and kept me 
focused.  Each group helped me gain a deeper understanding of the complexities 
ii 
 
   
surrounding the issues and provided constructive criticism which I have tried my best to 
include in this work. 
 In addition to those listed above, my family and friends provided me with the 
support necessary to bring this dissertation to fruition.  This work is only possible 
because I have had the benefit of a wonderful, supportive family.  I wish to thank my 
mother and father, Diane and Irving Smith Jr., who have always believed in me even 
when I did not believe in myself. My two boys, Ty and Malcolm Smith, allowed me to 
focus on the future. They understood why dad could not make every game or school 
event and never complained. There was not a day during this process that I did not think 
about them and their futures. Most of all, I am indebted to my wife, Mavis, who has been 
a selfless supporter, sounding board, copy maker, motivator, and friend.  Without her 
love and support I would have never completed this journey.  
 
 
 
 
 
 
 
 
 
 
 
 
iii 
 
   
TABLE OF CONTENTS 
 
 
List of Tables ..................................................................................................................... vi 
List of Figures.................................................................................................................. viii 
CHAPTER 1: INTRODUCTION ........................................................................................1 
Overview..............................................................................................................................1 
Research Questions..............................................................................................................6 
Contributions to the Field ....................................................................................................8 
CHAPTER 2:  STATUS ATTAINMENT AND MILITARY SERVICE ...........................9 
Early works on Status Attainment .......................................................................................9 
Theoretical Models ............................................................................................................11 
Contemporary Research.....................................................................................................14 
General...............................................................................................................................14 
Race....................................................................................................................................16 
Consistency Across Temporal and Socio-Demographic Groups ......................................22 
Critique of Veteran Social Status Attainment Studies.......................................................24 
Hypotheses.........................................................................................................................28 
CHAPTER 3:  METHODS................................................................................................31 
Data....................................................................................................................................31 
Data Analysis.....................................................................................................................32 
Variables ............................................................................................................................34 
Dependent Variables..........................................................................................................35 
Independent Variables .......................................................................................................38 
Control Measures...............................................................................................................39 
Interaction Variables..........................................................................................................40 
Statistical Methodology .....................................................................................................41 
Selection Bias ....................................................................................................................46 
CHAPTER 4: WORLD WAR II VETERANS AND THE 1950 CENSUS ......................48 
Enumeration Problems with the 1950 Census ...................................................................48 
Underreporting...................................................................................................................49 
G.I.s at College ..................................................................................................................52 
Descriptive data .................................................................................................................58 
Background Descriptive Statistics and Discussion............................................................61 
Education Models and Regression Analysis......................................................................83 
Income Models and Regression Analysis..........................................................................87 
Duncan SEI Models and Regression Analysis ..................................................................93 
Summary............................................................................................................................97 
CHAPTER 5: WORLD WAR II VETERANS AND THE 1960 CENSUS ......................99 
Descriptive Data ..............................................................................................................100 
Background Descriptive Statistics and Discussion..........................................................100 
Models and Multivariate Regression Analysis ................................................................126 
Education Models and Regression Analysis....................................................................127 
Income Models and Regression Analysis........................................................................131 
Duncan SEI Models and Regression Analysis ................................................................136 
Summary..........................................................................................................................140 
CHAPTER 6: WORLD WAR II VETERANS AND THE 1970 CENSUS ....................142 
iv 
 
   
Descriptive Data ..............................................................................................................143 
Background Descriptive Statistics and Discussion..........................................................146 
Models and Multivariate Regression Analysis ................................................................164 
Education Models and Regression Analysis....................................................................165 
Income Models and Regression Analysis........................................................................170 
Duncan SEI Models and Regression Analysis ................................................................176 
Summary..........................................................................................................................181 
CHAPTER 7: WORLD WAR II VETERANS AND THE 1980 CENSUS ....................183 
Descriptive Data ..............................................................................................................184 
Background Descriptive Statistics and Discussion..........................................................189 
Models and Multivariate Regression Analysis ................................................................207 
Education Models and Regression Analysis....................................................................208 
Income Models and Regression Analysis........................................................................213 
Duncan SEI Models and Regression Analysis ................................................................220 
Summary..........................................................................................................................225 
CHAPTER 8: WORLD WAR II VETERANS AND THE 1990 CENSUS ....................228 
Descriptive Data ..............................................................................................................228 
Background Descriptive Statistics and Discussion..........................................................234 
Models and Multivariate Regression Analysis ................................................................251 
Education Models and Regression Analysis....................................................................252 
Income Models and Regression Analysis........................................................................256 
Duncan SEI Models and Regression Analysis ................................................................260 
Summary..........................................................................................................................265 
CHAPTER 9: WORLD WAR II VETERANS AND THE 2000 CENSUS ....................268 
Descriptive Data ..............................................................................................................268 
Background Descriptive Statistics and Discussion..........................................................274 
Models and Multivariate Regression Analysis ................................................................291 
Education Models and Regression Analysis....................................................................291 
Income Models and Regression Analysis........................................................................295 
Duncan SEI Models and Regression Analysis ................................................................299 
Summary..........................................................................................................................302 
CHAPTER 10: SUMMARY AND CONCLUSIONS.....................................................305 
 
 
v 
LIST OF TABLES 
 
Table 3.1:  Education to Grade Level Conversion Chart................................................36 
Table 4.1:  Veterans in 1950-2000 IPUMS Data Sets ....................................................48 
Table 4.2:  Veteran Age by Year of Entry by Age in the 1950 Census..........................62 
Table 4.3:  Descriptive Characteristics of the 1950 Census Sample ..............................64 
Table 4.4:  Significance Tests for Age............................................................................65 
Table 4.5:  Birth Year Comparison of Veterans and Non-Veterans ...............................66 
Table 4.6:  Regional Significance Tests..........................................................................69 
Table 4.7:  Marital Significance Tests ............................................................................72 
Table 4.8:  Education Significance Test .........................................................................74 
Table 4.9:  Income Significance Tests ............................................................................79 
Table 4.10:  Duncan SEI Significance Tests.....................................................................82 
Table 4.11:  1960 Education Models and Regressions .....................................................85 
Table 4.12:  1960 Income Models and Regressions .........................................................89 
Table 4.13:  1960 Duncan SEI Models and Regressions..................................................94 
Table 5.1:  1960 Census Descriptive Summary............................................................102 
Table 5.2:  Veteran Expected Ages in 1960 by Entry Age ...........................................104 
Table 5.3:  Age Significance Tests ...............................................................................104 
Table 5.4:  Region Significance Tests ..........................................................................108 
Table 5.5:  Marital Status Significance Tests ...............................................................110 
Table 5.6:  1960 Education Significance Tests.............................................................113 
Table 5.7:  Education Significance Test by Age Category...........................................114 
Table 5.8:  Income Significance Test by Age Categories.............................................123 
Table 5.9:  Significance Tests for Duncan SEI.............................................................125 
Table 5.10:  1960 Education Models and Regressions ...................................................130 
Table 5.11:  1960 Income Models and Regressions .......................................................132 
Table 5.12:  1960 Duncan SEI Models and Regressions................................................137 
Table 6.1:   1970 Sample Descriptive Characteristics ...................................................144 
Table 6.2:  Veteran Expected Ages in 1970 by Entry Age ...........................................147 
Table 6.3:  1970 Census Age Significance Tests..........................................................148 
Table 6.4:  1970 Region Significance Tests .................................................................149 
Table 6.5:  1970 Marital Status Significance Tests ......................................................152 
Table 6.6:  1970 Education Significance Tests.............................................................154 
Table 6.7:  1970 Education Significance Test by Age Category..................................155 
Table 6.8:  1970 Income Significance Test by Age Categories....................................161 
Table 6.9:  1970 Significance Tests for Duncan SEI....................................................163 
Table 6.10:  1970 Education Models and Regressions ...................................................168 
Table 6.11:  1970 Income Models and Regressions .......................................................171 
Table 6.12:  1970 Duncan SEI Models and Regressions................................................180 
Table 7.1:  1980 Census Descriptive Summary............................................................186 
Table 7.2:  Veteran Expected Ages in 1980 by Entry Age ...........................................189 
Table 7.3:  1980 Census Age Significance Tests..........................................................190 
Table 7.4:  1980 Region Significance Tests .................................................................192 
Table 7.5:  1980 Marital Status Significance Tests ......................................................194 
Table 7.6:  1980 Education Significance Tests.............................................................196 
Table 7.7:  1980 Education Significance Test by Age Category..................................197 
Table 7.8:  1980 Income Significance Test by Age Categories....................................203 
vi 
 
   
Table 7.9:  1980 Significance Tests for Duncan SEI....................................................206 
Table 7.10:  1980 Census Education Models and Regressions.......................................209 
Table 7.11:  1980 Census Income Models and Regressions...........................................215 
Table 7.12:  1980 Duncan SEI Models and Regressions................................................222 
Table 8.1:  1990 Census Descriptive Summary............................................................231 
Table 8.2:  Veteran Expected Ages in 1990 by Entry Age ...........................................235 
Table 8.3:  1990 Census Age Significance Tests..........................................................235 
Table 8.4:  1990 Region Significance Tests .................................................................237 
Table 8.5:  1990 Marital Status Significance Tests ......................................................238 
Table 8.6:  1990 Education Significance Tests.............................................................241 
Table 8.7:  1990 Education Significance Test by Age Category..................................242 
Table 8.8:  1990 Income Significance Test by Age Categories....................................248 
Table 8.9:  1990 Significance Tests for Duncan SEI....................................................250 
Table 8.10:  1990 Education Models and Regressions ...................................................254 
Table 8.11:  1990 Income Models and Regressions .......................................................257 
Table 8.12:  1990 Duncan SEI Models and Regressions................................................262 
Table 9.1:  2000 Census Descriptive Summary............................................................271 
Table 9.2:  Veteran Expected Ages in 1990 by Entry Age ...........................................274 
Table 9.3:  2000 Census Age Significance Tests..........................................................275 
Table 9.4:  2000 Region Significance Tests .................................................................277 
Table 9.5:  2000 Marital Status Significance Tests ......................................................279 
Table 9.6:  2000 Education Significance Tests.............................................................281 
Table 9.7:  2000 Education Significance Test by Age Category..................................282 
Table 9.8:  2000 Income Significance Test by Age Categories....................................288 
Table 9.9:  2000 Significance Tests for Duncan SEI....................................................290 
Table 9.10:  2000 Education Models and Regressions ...................................................293 
Table 9.11:  2000 Income Models and Regressions .......................................................296 
Table 9.12:  2000 Duncan SEI Models and Regressions................................................301 
vii 
 
   
LIST OF FIGURES 
 
Figure 1.1:  Veterans by Period of Service ........................................................................3 
Figure 3.1:  Regression Models........................................................................................42 
Figure 3.2:  Selection Bias ..............................................................................................44 
Figure 4.1:  Census Form 5 "Overseas Census Report.....................................................50 
Figure 4.2:  Census For P4: "Crew of Vessels Report .....................................................51 
Figure 4.3:  WW II Cohort Percent Change in Some College or College Degrees ........53 
Figure 4.4:  Earnings Income for Veterans and Non-Veterans ....................................... 57 
Figure 4.5:  Natural Log of Earnings Income for Veterans and Non-Veterans ..............57 
Figure 4.6  1950 Census WW II Cohort Original .......................................................... 60 
Figure 4.7:  1950 Census Cohort with Changes...............................................................60 
Figure 4.8:  1960 Census WW II Birth Cohort ................................................................61 
Figure 4.9:  1960 Census WW II Cohort Mobilization Phases ........................................67 
Figure 4.10:  1950 Census WW II Cohort Education Levels.............................................73 
Figure 4.11:  1950 Census WW II Veteran Versus -Veteran Educational Attainment......75 
Figure 4.12:  Log Income by Birth Year ............................................................................77 
Figure 4.13:  Black Veteran versus Non-Veteran Log Income .........................................78 
Figure 4.14:  White Veteran versus Non-Veteran Log Income..........................................78 
Figure 4.15:  1950 Census World War II Cohort Duncan SEI by Birth Year....................81 
Figure 5.1:  1960 Census Veterans versus Non-Veteran................................................103 
Figure 5.2:  1960 Census World War II Cohort Mobilization Phases ...........................106 
Figure 5.3:  1960 Cohort Education Levels ...................................................................112 
Figure 5.4:  Veteran versus Non-Veteran Education 
 (College and High School Graduates Only.................................................115 
Figure 5.5:  Black Veteran versus Black Non-Veteran Education  
 (College and High School Graduates Only.................................................116 
Figure 5.6:  White Veteran versus White Non-Veteran Education 
  (College and High School Graduates Only.................................................117 
Figure 5.7:  1960 Veteran versus Non-Veteran Log Income .........................................119 
Figure 5.8:  Black Veteran versus Non-Veteran Log Income........................................120 
Figure 5.9:  White Veteran versus Non-Veteran Log Income .......................................121 
Figure 5.10:  1960 Census Combined Black and White Veteran versus 
 Non-Veteran Log Income ...........................................................................122 
Figure 5.11:  1960 Veteran Versus Non-Veteran Duncan Median SEI   
 score by Birth Year .....................................................................................124 
Figure 5.12:  960 Black and White Veteran Versus Non-Veteran Duncan  
 Median SEI score by Birth Year.................................................................126 
Figure 6.1:  1970 Census Veterans versus Non-Veteran................................................146 
Figure 6.2:  1970 Cohort Education Levels ...................................................................153 
Figure 6.3:  Veteran versus Non-Veteran Education 
 (College and High School Graduates Only.................................................156 
Figure 6.4:  Black Veteran versus Black Non-Veteran Education  
 (College and High School Graduates Only.................................................157 
Figure 6.5:  White Veteran versus White Non-Veteran Education  
 (College and High School Graduates Only.................................................158 
Figure 6.6:  1970 Veteran versus Non-Veteran Log Income .........................................159 
Figure 6.7:  1970 Census by Race and Veteran Status Log Income ..............................160 
viii 
 
   
Figure 6.8:  1970 Veteran versus Non-Veteran Duncan Median SEI  
 score by Birth Year .....................................................................................162 
Figure 6.9:  1970 Black and White Veteran versus Non-Veteran Duncan  
 Median SEI score by Birth Year.................................................................164 
Figure 7.1:  1980 Census Veterans versus Non-Veteran................................................187 
Figure 7.2:  1980 Cohort Education Levels ...................................................................195 
Figure 7.3:  Veteran versus Non-Veteran Education 
 (College and High School Graduates Only.................................................198 
Figure 7.4:  Black Veteran versus Black Non-Veteran Education  
 (College and High School Graduates Only.................................................198 
Figure 7.5:  White Veteran versus White Non-Veteran Education 
 (College and High School Graduates Only.................................................199 
Figure 7.6:  1980 Veteran versus Non-Veteran Log Income .........................................201 
Figure 7.7:  1980 Black and White Veteran versus Non-Veteran Log Income .............202 
Figure 7.8:  1980 Veteran versus Non-Veteran Duncan Median SEI score by  
 Birth Year ...................................................................................................204 
Figure 7.9:  1980 Black and White Veteran Versus Non-Veteran Duncan Median SEI    
 score by Birth Year ................................................................................. ...207 
Figure 8.1:  1990 Census Veterans versus Non-Veteran............................................... 234 
Figure 8.2:  1990 Cohort Education Levels ...................................................................240 
Figure 8.3:  Veteran versus Non-Veteran Education  
 (College and High School Graduates Only.................................................243 
Figure 8.4:  Black Veteran versus Black Non-Veteran Education  
 (College and High School Graduates Only.................................................243 
Figure 8.5:  White Veteran versus White Non-Veteran Education  
 (College and High School Graduates Only.................................................244 
Figure 8.6:  1990 Veteran versus Non-Veteran Log Income .........................................246 
Figure 8.7:  1990 Census Black and White Veteran versus Non-Veteran Log Income.247 
Figure 8.8:  1990 Veteran versus Non-Veteran Duncan Median SEI score by  
 Birth Year ...................................................................................................249 
Figure 8.9:  1990 Black and White Veteran Versus Non-Veteran Duncan Median SEI  
 score by Birth Year .....................................................................................251 
Figure 9.1:  1990 Census Veterans versus Non-Veteran............................................... 272 
Figure 9.2:  2000 Cohort Education Levels ...................................................................280 
Figure 9.3:  Veteran versus Non-Veteran Education 
  (College and High School Graduates Only.................................................283 
Figure 9.4:  Black Veteran versus Black Non-Veteran Education 
 (College and High School Graduates Only.................................................283 
Figure 9.5:  White Veteran versus White Non-Veteran Education 
 (College and High School Graduates Only.................................................284 
Figure 9.6:  2000 Veteran versus Non-Veteran Log Income .........................................286 
Figure 9.7:  2000 Census Black and White Veteran versus Non-Veteran Log Income.287 
Figure 9.8:  1990 Veteran versus Non-Veteran Duncan Median SEI score by  
 Birth Year ...................................................................................................289 
Figure 9.9:  2000 Black and White Veteran versus Non-Veteran Duncan  
 Median SEI score by Birth Year.................................................................291 
Figure 10.1:  Veteran versus Non-Veteran Education .....................................................310 
Figure 10.2: Black and White Veteran versus Non-Veteran Education..........................310 
ix 
 
   
Figure 10.3:  Veteran versus Non-Veteran Income..........................................................313 
Figure 10.4: Black and White Veteran versus Non-Veteran Income..............................313 
Figure 10.5:  Veteran versus Non-Veteran SEI................................................................314 
Figure 10.6:  Black and White Veteran versus Non-Veteran SEI....................................315 
 
 
 
 
 
 
 
 
 
x 
Chapter 1: Introduction 
 
Overview 
  
 Military service represents a pervasive and varied experience in men's [and 
women?s] lives across successive birth cohorts in this century (Sider and Cole 1984). 
Veterans have always represented the spectrum of demographic categories in terms of 
race, ethnicities, gender, geographic region of origin, socioeconomic background, and 
age.  The military has often been used as the subject of inquiry to understand better a 
broad range of sociological theories, issues, and problems that cut across several areas of 
intellectual investigation and social inquiry including attitude research, small groups, 
race relations, social change, the family, professions, and political economy.     
 As of September 2003 there were approximately 26 million armed forces veterans 
in the United States, representing approximately 12.4 percent of the U.S population over 
age eighteen.  Of those, 4.3 million served during World War II; 3.5 million during the 
Korean War; 8.1 million during the Vietnam Conflict: 3.8 million during the Gulf War; 
and 6.3 million during the interwar periods (see figure 1.1).  Approximately 16.5 percent 
of all veterans were World War II veterans.  In 1990 there were 27.3 million veterans of 
whom 27.1 percent were World War II veterans and in 2000 there were approximately 
26.4 million veterans of whom 21.7 percent were World War II percent veterans (United 
States Census Bureau 2005).  Although these numbers represent a decreasing trend in the 
number of World War II veterans, veterans are still a significant part of the United States 
population.  
 In 2003 there were approximately 22.6 million white, 2.6 million black, 1.2 
million Latino, and 284,000 Asian veterans.  Of those 3,271,668 were white, 213,324 
1 
 
   
were black, 79,483 were Latino, and 26,546 were Asian/Pacific Islander World War II 
veterans.  
 With the exception of Latinos, all other racial and ethnic categories of veterans 
are expected to decrease substantially by 2020.  Most notably the White veteran 
population is expected to decrease to approximately 13.6 million. There has been a great 
deal of research on veteran social status attainment over the past forty years; however, 
the research generally has been specific to a particular era, cohort, or demographic group.  
The vast majority of the research focuses on how veterans of a specific cohort compare to 
their non-veteran peers in terms of income, education, criminal history, social status, and 
occupational outcomes or how veterans of a specific demographic category have done 
compared to their veteran peers along these same dimensions.  Relatively little of the 
existing research attempts to compare veterans to their non-veteran peers across specific 
demographic groups, across different eras, and birth cohorts simultaneously.  This 
research attempts to bridge this gap. 
 I use census data from the 1950 through 2000 decennial censuses to test the 
bridging hypothesis using the life course perspective and human capital theory as guides.  
The bridging hypothesis, first articulated by Leonard Broom and J.H. Smith (1963) and 
subsequently adapted to veteran attainment studies by Browning, Lopreato, and Poston 
(1973), contends that service members learn valuable skills while serving in the military 
that translate into opportunities in the civilian sector.  The census data will allow me to 
capture veterans? attainment (income, education, and occupation) and compare it not only 
across birth cohorts but also across race boundaries.  These data allow me to provide 
information about World War II veterans and make comparisons among them and their 
non-veteran peers at six distinct points in time. 
2 
 
   
3 
 
V
e
t
e
ra
ns
  B
y
 P
e
ri
o
d
 o
f
 S
e
rv
i
c
e
(
i
n
 m
i
llio
n
s
)
Vi
e
t
n
a
m
 W
a
r
,
 8
.
1
Gu
l
f
 W
a
r
 
, 3
.
8
I
n
t
e
rw
a
r
 P
e
ri
o
d
s
,
 
6.
3
W
o
rl
d
 W
a
r I
I
,
 4
.
3
K
o
r
e
an
 W
a
r
,
 3.
5
Wo
r
l
d
 Wa
r
 I
I
Ko
r
e
a
n
 W
a
r
Vi
e
t
n
a
m
 W
a
r
Gu
l
f
 W
a
r 
I
n
t
e
rw
a
r
 P
e
ri
o
d
s
SOURCE: VetPop2001Adjusted to Census 2000 
Figure 1.1 Veterans by Period of Service 
 
   
  We know that some of today?s service members are recruited into the service 
through promises of future education via the Montgomery GI Bill, learning a skill, and 
gaining experience for future use in the civilian sector.  However, this has not always 
been the case.  The World War II era forces were, for the most part, conscripted forces 
with a small cohort of volunteers.  In fact, the original G.I. Bill was introduced at the end 
of the war to speed the nation?s transition from war time to peace time production and 
compensate veterans whose educational plans were interrupted by military service.  
Experimental studies have provided evidence that veterans? educational benefits increase 
enlistments.  Although the enlistment effects of veteran?s benefits suggest recruits value 
these benefits, few studies have directly considered what federal subsidies to post-service 
education and training are worth to veterans? (Angrist 1993:1).   
Samuel Stouffer et al. (1949) conducted one of the first studies on veteran 
attainment.  Among other things they found that World War II veterans were generally 
optimistic about their personal prospects for increased social status after the war but 
pessimistic about their prospects as a group. The fact that servicemen tended to be more 
optimistic about their personal chances of employment than about the chances of 
employment for veterans in general is interesting because most of the research on World 
War II veterans shows that they did well compared to their non-veteran peers (Martindale 
and Poston 1979; Villemez and Kasarda 1976).  If they do better than their non-veteran 
peers is their success a function of the period of service or some demographic 
characteristic?  
 To understand the number of people who are affected by World War II military 
service it is necessary to take a cursory look at the Department of Defense (DOD) profile.  
Some have arguably referred to the United States DOD as the United States? oldest, 
4 
 
   
largest, busiest, and most successful business in the United States.  The DOD is among 
the largest employers in the United States, employing over 3,250,000 personnel including 
more than 1.4 million active duty military, 650,000 DOD civilians, and 1.2 million 
Reserve and National Guard personnel.  This number is small when compared to the 10.1 
million who served during World War II at any one time.  The more than 1.4 million 
service members on active duty in the United States Armed Forces (see table 1.1) 
represent approximately sixty seven percent of the total DOD labor force.  Each year 
approximately 200,000 service members leave the service and enter the civilian labor 
market (Segal and Segal 2004).  These employment numbers are not an anomaly and in 
fact, they pale in comparison to the World War II numbers. Since World War II the 
Department of Defense (earlier the War Department) always has been one of the largest 
employers in the United States.  
  What happens to service members when they end their term of service (ETS)?  
There has been a great deal of research documenting World War II veteran attainment 
since the 1970s; however, the research has generally looked at veterans as a part of 
specific groups.  These groups include minorities such as blacks, Asians, and women or 
cohort groups such as World War II, Korean, Vietnam, and AVF veterans.  The research 
has shed light on particular groups but has failed to provide an analysis of what happens 
to veterans of a specific cohort over time.  The failure to link the data across time and 
demographic groups has created a jigsaw puzzle of information. In contrast to other 
sociological studies, this dissertation focuses on bringing the pieces of the jigsaw puzzle 
together to create a more complete picture of veteran attainment.  This study attempts to 
determine how World War II veterans have fared compared to their non-veteran peers 
5 
 
   
over the life course. Additionally, this study uses multiple criteria/variables including 
income, education, and occupational status to make these comparisons.     
There are some limitations as to how much one can generalize from this type of 
analysis. My findings can only be generalized to the World War II veteran population 
and any attempt to bridge these findings to other veteran populations might be specious at 
best. Furthermore, not only are the findings limited to the World War II veteran 
population they are also limited by several other factors to include race, gender, and 
employment status (see Chapter 3 Methods).  
The most evident reason that one should not generalize to populations other than 
the World War II veteran population is that the pattern of veteran effects is more complex 
than the literature on veteran social status attainment suggests.  Some of these 
complexities include but are not limited to how we count veterans, who counts as a 
veteran, macro-economic and historical events, and who the comparison groups are. 
While I make every attempt to explain how I treat these important issues (see Chapter 3 
Methods), others have treated them differently in some cases which could have the effect 
of clouding the interpretation of veteran social status attainment.  
Moreover, as a result of looking at this study I would expect that the reader gain 
an appreciation of the following: does being a veteran matter; does being a black veteran 
matter, and does being in a particular birth cohort matter? Furthermore, the reader should 
understand that the answers to this question with regard to my analysis apply only to the 
World War II veterans with respect to their non-veteran peers.    
 
Research Questions 
Many have viewed service in the armed forces as a clear cut way to achieve full 
6 
 
   
first class citizenship (Segal 1989), where full first class citizenship rights include, but 
are not limited to, equal representation in government, the opportunity to participate in all 
areas of community life, and the opportunity for social and economic equality.  It is 
generally understood that social status is the standing, honor, or prestige attached to one's 
position in society.  This research will essentially answer the question, did service in the 
armed forces function to enhance or detract from World War II veterans life chances, as 
measured by social status achievement?  Specifically, how are military service and the 
military as an institution valued by society as indicated by the social status that veterans 
achieve when they leave the service?  Many have described the United States military as 
a meritocracy.  In a meritocracy or any society where one is judged by his or her 
contributions one should be able to change their social position by hard work and 
achievement.  However, can one?s gains in human capital and achievement overcome 
ascribed status, like race?  In the Forward to Moskos and Butler?s All That We Can Be: 
Black Leadership and Racial Integration the Army Way (1997), Richard C. Leone, the 
President of the Century Foundation states,  
?The armed services today are almost certainly 
the most popular public institutions in the nation.  Amid 
certain post-Cold War cutbacks, the military retains a 
core of high quality leadership; more than ninety percent 
of enlisted men are high school graduates; a retired Afro-
American general is perhaps the most popular public 
figure in the nation.  Moreover, and perhaps most 
significantly, at a time of stark tensions and continuing 
separation between the races, not only is the Army a 
thoroughly integrated institution, its members seem at 
peace with the idea? (Moskos and Butler 1997).    
 
There is an opportunity cost associated with the decision to enter the military.  
Every person who enters the armed forces has forgone something that they may have had 
the opportunity to do had they not chosen to serve.  These opportunities include 
7 
 
   
furthering their education, increasing their time in civilian employment positions, and 
increased earnings.  What do veterans receive, if anything, for forgoing these 
opportunities and how does service affect their life course trajectories?  If they do earn 
human capital or social status as a result of their service how long does it take for them to 
realize these gains and how long do they last?      
 
Contributions to the Field 
 
 This study will make a number of contributions to the fields of military sociology 
and social stratification.  First, this research will allow us to understand better how 
veterans have fared historically over several decades.  This research, unlike most of the 
research to date, describes how the World War II cohort fared compared to their non-
veteran peers over their entire adult lives.  By looking at veterans through this particular 
lens the field should gain a better appreciation of how this particular veteran cohort fits 
into the stratification system.  Additionally, this study shows how military service 
affected those who serve in World War II in terms of occupational status, economics, and 
education.   
 This study will also help identify the effects of race, age, decade, birth year, and 
life changing events (World War II) on the earnings, occupational prestige, and 
educational outcomes of those who served during the World War II era.  If we can 
understand why World War II veterans did comparatively better or worse than their non-
veteran peers it might provide us with some insight as to how society has changed and 
more importantly where we might be headed.   
   
 
8 
 
   
Chapter 2:  Status Attainment and Military Service  
 
Early works on Status Attainment  
 The study of status attainment and mobility in the United States has become a 
staple of American sociologists since the middle of the twentieth century.  The social 
upheaval of World War II caused many scholars to focus on social changes that the war 
had brought about.  One of the first studies on social mobility, conducted by Lipset and 
Bendix (1959), found that social mobility is an enduring and important part the 
industrialization process.  They also disputed claims that the U.S. had more mobility than 
Western European nations and that social mobility tends to decline as industrial societies 
mature (Lipset and Bendix 1959).  
Samuel Stouffer et al. (1949) examined some of the issues associated with veteran 
attainment in a chapter titled ?The Soldier Becomes a Veteran.?  Although this chapter 
did not analyze the relationship between service and actual attainment it served as the 
basis for the study of veteran attainment.  The author?s examine soldiers? postwar 
expectations and attitudes near the end of World War II and then used data collected by 
the United States Army Research Branch and the Division of Program Surveys of the 
Bureau of Agricultural Economics to examine the extent to which veterans carried out 
their post war occupational and educational plans after the war.  They concluded that the 
majority of Army soldiers believed that their service was more of a liability than an asset 
and that approximately eighty-five percent were ?pretty well settled? as to what they 
would do (Stouffer, Suchman, DeVinney, Star, and Williams 1949).   
 Blau and Duncan (1969) used path analysis to analyze the occupational structure 
and mobility processes in the United States.  They considered five variables in their 
9 
 
   
study, father?s educational attainment, father?s occupational status, respondent?s 
educational attainment, status of respondent?s first job and status of respondent?s 
occupation in 1962.  Their findings suggested that family background independently 
explained seven percent of the variation in attainment, family background via education 
explained fourteen percent of the variation in attainment, and education independently 
explained twenty-one percent of the variance in attainment.  All of  factors together 
accounted for forty-three percent of the variance in attainment and fifty-seven percent of 
the variation was unexplained (Blau and Duncan 1967b).  
 Researchers using the Wisconsin Model of Status Attainment attempted to build 
on Blau and Duncan?s findings by adding social psychological and peer influence 
variables to the model.  In the study, the Wisconsin model showed that the main 
contributors to status attainment were parental socio-economic status, peer relations, 
educational aspirations, and educational attainment.  The Wisconsin Model of Status 
Attainment subsequently became a template for research on the life course.  Much of the 
classic literature on veteran status attainment stems from the classic social stratification 
literature described above. 
Service in the military is a life course event because it has the potential to alter 
the trajectory of one?s later life outcomes.  The study of veteran attainment seeks to 
understand how the transition into and out of the military affects veterans in terms of 
economics, education, and occupational status.  Some of the questions that have surfaced 
since the end of World War II are as follows: What is the level of income, education, 
class, and status attainment that veterans achieve compared to their peers who have never 
served?  Furthermore, does military service tend to increase veterans? opportunities by 
providing them with skills and education that they might not have gained had they not 
10 
 
   
served in the armed forces; or on the contrary does military service inhibit the 
opportunities for income, education, class, and status attainment? If veterans tend to lag 
behind their peers, do they ever catch up, and if so how long does it take?  Is there any 
demographic group of veterans that does particularly well, in that they earn a premium 
for their service?  Conversely, are there any groups of veterans who pay a penalty for 
their service compared to both their veteran and non-veteran peer groups?  
 
Theoretical Models  
 Two of the hypotheses that researchers have used in the past are the bridging 
hypothesis and the interruption hypotheses.  The bridging hypothesis states that service 
members learn valuable skills while serving in the military that translate into 
opportunities in the civilian sector (Browning, Lopreato, and Poston 1973). According to 
this theory the bridge can be created because the military provides opportunities for both 
training and education. In many cases this training is directly applicable to civilian sector 
work.  This is consistent with Broom and Smith?s (1963) original conception of a 
bridging occupation, in which they describe a bridging occupation as one ?that provides, 
through work experience, the conditions and opportunities for for movement from one 
occupation to another.?  Broom and Smith identify the five attributes that contribute to 
successful bridging including resocialization, independency, health and physical bearing, 
access to information, and financial competence.  Interestingly, in their original work 
they describe the work of a soldier as a bridging occupation.  
Additionally, the military can increase veterans? independence by knifing off or 
severing the service member?s ties to their origins and past.  The independence that 
service members gain from knifing off their pasts may also provide them with the human 
11 
 
   
and cultural capital required to relocate after their service (Sampson and Laub 1996; Xie 
1992).  Furthermore, service in the military exposes veterans, especially minorities, to 
mainstream achievement values, working with diverse racial and ethnic groups, and 
manipulating large-scale bureaucracies (Browning, Lopreato, and Poston 1973; Cooney, 
Segal, Segal, and Falk 2003).  A.R. Hollingshead surmises that people enter the service 
with a set of values that do not match military objectives.  This requires that the new 
recruit be reeducated to learn military norms and the military lifestyle (Hollingshead 
1946).  These norms include but are not limited to an understanding of bureaucracies and 
bureaucratic procedures, discipline, and placing organizational goals in context with 
personal goals and needs.   
 The interruption hypothesis states that entrance into the service stifles service 
members? post military civilian careers making them forgo years of education and labor 
force experience and thereby exposes them to lost opportunities that may have been more 
beneficial than military service (Cutright 1974).    The interruption hypothesis is akin to 
the economic theory of opportunity cost.  Opportunity cost can be defined as follows: 
 The true cost of something is what you give 
up to get it.  This includes not only the money spent 
in buying (or doing) the something, but also the 
economic benefits utility that you did without 
because you bought (or did) that particular 
something and thus can no longer buy (or do) 
something else.  For example, the opportunity cost 
of choosing to train as a lawyer is not merely the 
tuition fees, price of books, and so on, but also the 
fact that you are no longer able to spend your time 
holding down a salaried job or developing your 
skills as a footballer.  These lost opportunities may 
represent a significant loss of utility or the cost of 
something in terms of an opportunity foregone (and 
the benefits that could be received from that 
opportunity), or the most valuable foregone 
alternative (Bishop 2004:1).  
 
12 
 
   
Viewed from this perspective the opportunity costs involved with serving in the military 
center around lost time that could be used to further civilian job opportunities or getting 
more education as described above. 
The literature that revolves around these two diametrically opposed hypotheses 
(bridging and interruption) surfaces many of the issues surrounding veterans? post-
service attainment patterns.  Additionally, some of the more recent work has examined 
veteran attainment through the lens of a life course alteration whereby military service is 
viewed as a life altering experiencing that affects ?the timing and sequencing of events in 
the transition to adulthood, especially when they occurred early in life? (Elder and 
Meguro 1987:439).  Furthermore, many have identified service in the armed forces as a 
turning point in the life course because it can alter trajectories in either an advantageous 
or a detrimental manner.  Elder, Modell, and Parke (1993) suggest that service in the 
military increases the likelihood of redirection in a person?s life and provides 
opportunities that may not be available in civilian life. 
Human Capital Theory (HCT) is integrally tied to both the bridging hypothesis 
and the life course perspective in that it too attempts to explain occupation, wage, and 
education differentials in individuals.  One of the foundations of HCT is that both 
education and training are costly in terms of time and money and should therefore be 
considered as investments since they are undertaken for the purpose of increasing 
personal incomes (Becker 1993).  For service members the costs are defined in terms of 
time that could be spent in civilian jobs that could potentially increase seniority, on the 
job training, increased earnings, and networking opportunities.  The time could also be 
spent on higher education that could increase occupational status and earnings. 
Conversely, Elder (1986) argues that the military helps develop human capital from a life 
13 
 
   
course perspective.  His argument revolves around the belief that service in the armed 
forces can serve as a ?mechanism by which unpromising beginnings lead to opportunity 
and fulfillment instead of to failure.?  The dynamics of this mechanism are represented 
by military service as a turning point that manifests itself in ?later transitions to 
adulthood, as expressed in school leaving, first marriage, and first child? (Elder 1986) 
and ultimately allows veterans to achieve occupational status equal to non-veterans. 
 
Contemporary Research  
 Over the past half century several patterns have been identified regarding the 
impact of service on post-service lives.  These patterns are salient when illuminated 
under the lenses of race, gender, and period of service; however, the overall results of 
veteran attainment are mixed in that they vary by these same factors.  For example, 
World War II veterans did well compared to their non-veteran peers (Browning, 
Lopreato, and Poston 1973), however, if we add an additional variable of age, some have 
found that older World War II veterans did not do as well as their non-veteran or younger 
veteran peers (Elder and Meguro 1987; Sampson and Laub 1996).  What is the extent to 
which service effects have been consistent across temporal and socio-demographic 
groups, and why does variation exist among these groups. 
 
General  
 The first studies of veteran attainment were focused on the attainment and post-
service status of conscripted white men (Bailey and Cargill 1969; Davis and Palomba 
1968; Hansen and Weisbrod 1967; Miller and Tollison 1971; Oi 1967; Willett 1968).  
The impetus behind the studies was the widespread belief that there were essentially 
14 
 
   
three types of inductees: true volunteers, reluctant volunteers (volunteered only to get a 
better position than if they were drafted), and draftees (Davis and Palomba 1968; Oi 
1967; Segal 1989).  Furthermore, there was public concern that the last two categories 
bore the burden of the draft.  The early studies focused on the reluctant volunteers and 
draftee portions of the veteran population to test this hypothesis.  In general, these studies 
showed that military service resulted in income penalties to veterans.  The early literature 
tends to cite two reasons for white veteran income penalties.  The first is the differential 
pay scale between military and civilian occupations.  It was widely held in the late 1960?s 
and early 1970?s that lost lifetime earnings for veterans were significant and the veteran 
never caught up to his non-veteran peers (Browning, Lopreato, and Poston 1973:75; 
Miller and Tollison 1971).  The second hypothesis was that serving in the military 
interrupted the life course.  These interruptions included interruptions to higher 
education, on the job training, apprenticeships, and advancement on the basis of 
seniority.   
 These two hypotheses are still widely held today, although some might concede 
that the all volunteer force has required the Department of Defense (DOD) to increase 
pay and benefits to attract and retain the contemporary service member (Moskos and 
Wood 1988; Segal 1989) and thus reduced  the gap between civilian and military pay 
differentials.  Similarly, much of the research suggests that white men with 
disadvantaged or criminal backgrounds tend to do better in post-service life than their 
non-veteran peers with similar backgrounds.  The literature suggest that this occurs 
because of the change of environment and the ability to ?knife off past experiences? 
(Bouffard 2005; Sampson and Laub 1996; Teachman and Tedrow 2004).  However, it 
15 
 
   
should be noted that these results tend to vary by period, research method, data, and 
control variables (see ?Critique of Literature? section).  
 The research generally shows that white World War II  veterans tends to do better 
than their non-veteran peers in terms of income and education (Cooney, Segal, Segal, and 
Falk 2003; Elder and Meguro 1987; Fredland and Little 1985; Martindale and Poston 
1979; Sampson and Laub 1996; Teachman and Tedrow 2004; Villemez and Kasarda 
1976).  The research to date has not provided any consistent answers to what happens to 
white veterans upon their return to civilian labor markets although there are clearly some 
temporal consistencies in the literature.  World War II and Korean War veterans seem to 
have earned a premium for their service, while Vietnam and AVF era veterans appear to 
have gained no appreciable premium in social status as measured in terms of income, 
education, or occupational status. 
 
Race  
 African Americans have served in every American conflict from the 
Revolutionary War to the current conflict in Iraq.  Crispus Attucks, a black American, 
became the first casualty of the American Revolution in 1770 after being shot during the 
Boston Massacre.  Shoshana Johnson of the 507
th
 Maintenance Company was received as 
a hero after surviving being taken as a prisoner of war in the most recent conflict in Iraq.  
Black participation in American wars came only as a result of long struggles by blacks 
and benevolent whites who wished to see blacks have full citizenship in the United 
States.   
 There were several arguments against black participation in the armed forces.  
The cohesion argument stated that blacks would ruin the social cohesion that already 
16 
 
   
existed among white soldiers.  Furthermore, white soldiers would never get along with, 
much less serve as an equal with, a black soldier.  The medical argument claimed that 
blacks were carriers of disease and would spread their diseases to the whites serving in 
the armed forces.  The effectiveness argument essentially stated that blacks could not 
fight.  Blacks were not intelligent enough, or physically and emotionally strong enough 
to bear the burden of war.  Military labor shortages, changes in the law, and 
accomplishments in combat allowed blacks to overcome these arguments and earn the 
obligation, and the rights that came with these obligations, to serve their country.  Blacks 
often endured the tough conditions of service life because life in the military was often a 
better alternative than employment in the civilian sector (Segal 1989).  
 Today, some would argue that the military is a labor utopia for blacks, in that they 
have opportunities similar to those of whites (Moskos and Butler 1997).  In 2002, blacks 
made up approximately twenty-two percent of the enlisted force across all branches of 
service although they comprised only thirteen percent of the eighteen to forty-four year 
old population at large.   
 In spite of the obstacles placed in the path of black service in the military it is 
clear that blacks have made great strides within the military.  However, the question 
becomes, has their success within the military translated to post-service educational, 
income, and status attainment?  Much like the literature on whites, the literature on black 
attainment is mixed: however, some general patterns do appear.  The results of the 
analysis differ based on the period of analysis, control variables, and definition of the 
term veteran that the researcher uses.   
Browning et al. (1973) conducted some of the first research on African-American 
post-service attainment. Using a one percent Public Use Microdata Series (PUMS) of the 
17 
 
   
1960 Decennial Census they examined the relationship between military service and the 
incomes of ?Mexican Americans, Blacks, and Anglos? in five Southwestern states.  Most 
of the veterans in their study would have served during the Korean War.  Prior to 
Browning et al.?s study, several studies had determined that military service resulted in 
earnings penalties for veterans (Browning, Lopreato, and Poston 1973:75).  They found 
that white non-veterans had higher reported incomes than white veterans, but Black and 
Mexican American veterans had higher incomes than Black and Mexican American non-
veterans.  In their research, they credit Leonard Broom and J.H. Smith in ?Bridging 
Occupations? (1963) with coining the term bridging environment (Browning, Lopreato, 
and Poston 1973, page 76).  Browning and his colleagues attributed the income and 
social mobility gained by blacks and Mexican Americans to several factors including the 
independency effect, the acquisition of education and job skills, and exposure to 
bureaucracy (see above for detailed explanation of the bridging hypothesis).  Although 
Browning et al.?s study made startling new revelations about income attainment, it was 
criticized for several reasons.  The foremost criticism was that it was not a national study.  
Some argued that because the focus was restricted to five states, the data were not 
generalizable.  Others argued that they did not control for age and that minorities were 
under-represented in their sample because of low enumeration in the 1960 Census 
(Cutright 1974). 
The following year Phillips Cutright (1974) found that the bridging hypothesis 
was not supported using samples taken from a Selective Service Survey and Social 
Security Administration earnings? records.  In his study he controlled for age, race, 
region of employment, mental aptitude (as measured by the Armed Forces Qualification 
Test [AFQT]), and years of education.  He linked the Selective Service files and Social 
18 
 
   
Security earnings records in an attempt to correct what he saw as two of Browning et 
al.?s critical flaws: the low number of blacks in the Census and the low reliability of 
income reporting (Cutright 1974:320).  Although Cutright?s findings contradict 
Browning et al.?s., in the final analysis it is not that surprising because Cutright only 
considered draftees in his study.  This means that he did not consider those who entered 
the military through the service academies, ROTC, or voluntary enlistment.  Second, 
Cutright categorized his non-veteran population into four categories, which included 
AFQT failures, medically deferred, other deferments, and medical rejects.  These four 
categories are not exhaustive of the non-veteran population because they do not account 
for those who would qualify and did not receive deferments but nevertheless were not 
called up for service. 
Fredland and Little (1982) also conducted a test of Browning et al.?s bridging 
environment hypothesis.  They analyzed data from the National Longitudinal Study 
(NLS) to test the bridging hypothesis as it applied to World War II veterans who had 
been discharged from service for about twenty years.  Their sample included 5020 men 
who ranged in age from forty-five to fifty-nine years.  This study controlled for several 
factors including age, race, education, health, civilian training, military training, 
independence (Rotter Scale), employment in the government, geographic mobility, and 
work commitment.  They found that only some of the elements in Browning et al.?s 
understanding of the bridging environment are important in increasing income 
attainment.  Specifically, Fredland and Little found that for white males education, 
training, and personal independence made a difference in increasing earnings and 
socioeconomic status while only education and independence acquired while serving 
made a difference for blacks (Fredland and Little 1985).  This was contradictory to the 
19 
 
   
findings of Browning et al. because Fredland and Little suggested that whites males 
benefited from service more than blacks because of bridging experiences, but blacks 
gained more than their non-veteran peers overall because of post service employment in 
the government (Fredland and Little 1985:533).  They argued that the magnitude of the 
change was greater for blacks than for whites.  
Phillips et al. (1992) studied the impact of service on black, Hispanic and white 
youth in the all-volunteer era using a special military subsample of the National 
Longitudinal Survey of Youth (NLSY).  Their findings suggest that although all veterans 
have significant in-service earnings advantages, minorities obtain an even larger 
premium for their service than their civilian counterparts.  Contrary to other AVF era 
studies, they found that whites gain a significant post-service earnings premium, but 
blacks and Hispanics do not.  They attribute these findings to discrimination in the 
civilian labor market.  This finding lends credence to sociologists? long standing 
contention that blacks have historically remained in the service under less than ideal 
conditions because the conditions in service are better than those in the civilian sector 
(Moskos and Butler 1997; Segal 1989).  
Teachman and Call?s (1996) study significantly adds to the body of literature on 
veterans? and post-service income attainment.  Their study is significant because its 
findings explicitly state that income attainment studies must be measured in the context 
of the era that a veteran served.  Their study, using the High School and Beyond (HSB), 
the NLS, and the Career Development Studies, shows that Vietnam veterans tend to 
acquire less education, income, and occupational status than their non-veteran peers.  The 
authors attribute this not to the unpopularity of the Vietnam War, but to the erosion of 
government GI Bill incentives.  More specifically, the authors state, ?The relative value 
20 
 
   
of the GI Bill was seriously eroded because of the massive expansion of educational 
benefits available to non-veterans? (Teachman and Call 1996, page 27).  
 Leana Bouffard (2005) hypothesizes that the effects of military service might 
depend on race.  Using the National Longitudinal Survey of Youth (NLSY) to test the 
effects of military service on African Americans, she finds that military service reduces 
the risk of violence for African-American veterans compared to the their non-veteran 
peers.  She attributes this to the fact that the military may provide opportunities for these 
veterans that they otherwise would not have had.  This finding supports the bridging 
hypothesis and supports the notion that the military may serve as a transition that alters 
the life course trajectory for African-Americans.  These findings are consistent with other 
sociological and criminology studies on the effects of military service on life course 
trajectories (Sampson and Laub 1996). 
 The research on post-service African American veteran attainment in general has 
yielded mixed results but one can appreciate some general patterns.  Minorities who 
served in the World War II and Korean eras appear to have earned an income premium 
for their service.  The premiums earned may have been a result of increased opportunities 
for training and education that came about as result of military service.  Furthermore, 
military service may allow minorities to understand more clearly mainstream 
achievement values, work within a structured environment, and appreciate the value 
systems of others to whom they may otherwise not have been exposed were it not for 
military service.  Finally, the military may serve as a transition point in the lives of black 
youth that allows them to change the trajectory of their life course, thus severing them 
from their past and allowing them to move on to new opportunities.   
 
21 
 
   
Consistency Across Temporal and Socio-Demographic Groups 
 Although no hard and fast rules can be established by the research cited above, 
there are some general patterns that have emerged.  World War II and Korean veterans 
(to a smaller extent) seem to have earned a premium for their service in the armed forces 
(Charmarette and Thomas 1982; Detray 1982; Fredland and Little 1985; Martindale and 
Poston 1979; Rosen and Taubman 1982; Villemez and Kasarda 1976).  The research 
points to the expansive benefits of the GI bill, which created educational opportunities, 
heath factors (the military selects healthy people), and the expanding economy after 
World War II and Korea as significant  factors in this analysis.  This premium extended 
to both white men and minorities in World War II and Korea.  In fact, some authors 
suggest that the premium to blacks was greater than that of other veterans during the 
World War II and Korean eras (Detray 1982; Martindale and Poston 1979; Villemez and 
Kasarda 1976).  Fredland and Little (1985) suggest that the black veteran premium arose 
because that population was afforded preferential status in government employment.  
One notable exception to the group of researchers who found that Korean veterans earned 
a premium for their service was Schwartz (1986).  Using 1967 and 1969 Current 
Population data to compare the earnings of Korean and Vietnam War veterans he found 
that Korean veterans were ?indistinguishable? from non-veterans after controlling for 
education, race, marital status, and age.  
 Furthermore, veterans who served in the Vietnam era paid a penalty for their 
service.  This stems from the facts that the Vietnam War was generally unpopular; 
veterans? GI bill benefits eroded in that educational benefits had proliferated to the rest of 
society; and the generally poor economy restricted employment opportunities (Berger 
and Hirsch 1983; Charmarette and Thomas 1982; Detray 1982; Fredland and Little 1985; 
22 
 
   
Martindale and Poston 1979; Villemez and Kasarda 1976).  Martindale and Poston 
(1979) found that black Vietnam veterans earned a premium but the premium was 
smaller that earned by either World War II or Korean era veterans.  Rosen and Taubman 
(1982) provide an interesting twist to the research on Vietnam era veterans attainment.  
They found that those studies that use longitudinal data and the life course perspective 
tend to find veteran post-service earnings premiums while those that use cross sectional 
data tend to find earnings penalties.  They suggest that the data used to examine Vietnam 
veterans may not have extended far enough to paint an accurate picture of how these 
veterans fare compared to their non-veteran peers.  
 The bulk of the literature on veterans of the AVF can be best described as mixed.  
Cohen, Warner, and Segal (1995) suggest that these veterans suffered an educational 
disadvantage.  The educational disadvantage for AVF veterans extended to minorities 
and women as well.  On the other hand, Teachman and Call (1996) found that white 
males received an education premium but no discernable premium in income or 
occupational prestige.  They also found that black male AVF veterans gained an 
educational premium, but no premium in income or occupational prestige.  Philips et al. 
(1992) found that white AVF veterans received an advantage in earnings compared to 
their non-veteran peers; however, black males did not receive this same premium.  
 Poston, Segal, and Butler (1984) found that women who served between World 
War II and the AVF earned more than their non-veteran peers; however, they did not 
disaggregate the women veterans by their period of service.  Warner (1985) suggested 
that woman of the AVF do not earn an advantage compared to their non-veteran cohort in 
terms of earnings; however, her sample included women between the ages of 17 and 24 
23 
 
   
who may not have had time realize earnings increases because only a small amount of 
time had elapsed since their transition out of the military. 
 
Critique of Veteran Social Status Attainment Studies  
Earlier studies have allowed us to understand better the dynamics of veteran 
social status attainment as well as the mechanisms that serve to make the dynamics work.  
However, as much as these studies have added to our knowledge there are some 
methodological, data, and theoretical issues that have served to confound these studies 
that must be addressed.   
In general, relatively few measures of social status attainment have been used in 
veteran studies.  Most of the studies to date have used only earnings income (Browning, 
Lopreato, and Poston 1973; Cutright 1974; Martindale and Poston 1979) or educational 
attainment (Cohen, Segal, and Temme 1986) or both (Detray 1982; Little and Fredland 
1979; Xie 1992) as measures of social status attainment.  Some studies have added a 
measure of occupational prestige, usually Duncan?s Socioeconomic Index of the 
respondent?s job (Cohen, Segal, and Temme 1992; Fredland and Little 1985). The 
Duncan Socioeconomic Index is a measure of occupational status that is calculated using 
the income level and educational attainment associated with each occupation in 1950.  
Otis Dudley Duncan derived scores for each occupation using median income and 
education levels for men in 1950 to predict prestige assessments based on a 1947 survey 
(of a select group of occupations).  Duncan?s resulting statistical model was used to 
generate scores for the entire range of 1950 occupations and is still widely used to 
compare occupations today (Reiss 1962).  Furthermore, these studies have generally 
relied on using earnings income and have not used family or household income, 
24 
 
   
retirement income, disability income, or benefits in kind to measure veterans or 
comparison group?s earnings.  Adding these sources of income and benefits might change 
the way we understand veteran social status attainment. 
A great number of the studies have used longitudinal data to examine veteran social 
status attainment (Cohen, Segal, and Temme 1986; Defleur and Warner 1985; Goldberg 
and Warner 1986; Little and Fredland 1979; Mangum and Ball 1989); however, several 
have also relied on cross sectional data (Browning, Lopreato, and Poston 1973; 
Martindale and Poston 1979; Prokos 1996; Villemez and Kasarda 1976), either census or 
Social Security Records, that may not tell the whole story behind veteran attainment.  Is 
it possible that the bridging hypothesis may not work immediately upon discharge from 
the armed forces and that there may be some period of time  before the bridge becomes 
apparent?  Moreover, is it possible that bridging effects could work differently for people 
of different ages?  If so, this would suggest that analysis across several periods, either 
cross sectional or longitudinal, might be used to better inform the answers to these 
questions.  
Most of the previous studies on veteran attainment do not distinguish among various 
types of veterans.  In fact, the literature reviewed in this dissertation does not contain a 
single study that distinguishes between the differences in social status attainment of 
officers and enlisted service members.  Furthermore, very little of the research has 
attempted to make distinctions between volunteers and draftees.  All of the studies that 
were conducted prior to the advent of the all volunteer force fall into this category.  
Moreover, only a scant portion of the literature has distinguished among service 
member?s military occupational specialties (MOS).  The bridging theory presumes that 
military training among other things creates a bridge for future status attainment.  If this 
25 
 
   
is true then one might assume that different military specialties might produce very 
different status attainment outcomes.  Service members might also be distinguished by 
whether they served as Active Duty, National Guard, Reserve or some combination of the 
three.  Most of the studies to date have not distinguished among these three types of 
service.  Rather, Active Duty, Reserve, and National Guard service has been treated 
equally, as if the bridge would work the same for each.  Most of the work on social status 
attainment has focused on either veterans who served on active duty or makes no 
distinction at all in terms of veteran status. To date there have been relatively few studies 
have reflected on the social status attainment of Guard and Reserve soldiers (Lakhani 
1998).  
The research has generally excluded veterans and comparison groups with no or 
negative income.  Much of the logic behind eliminating non-income producing 
respondents is based on simple mathematics: you cannot take the log of zero or a 
negative number.  Moreover, it is difficult to ascertain an appreciation of the differences 
in social status attainment if the respondents are not working and no occupational score 
can be derived.  These exclusions present several problems for the study of veteran 
attainment that is illustrated in the previous literature.  Although the Census Bureau 
reports that 54.7 percent of veterans were employed in 2000
1
 and that poverty rates were 
low among veterans regardless of service period (Richardson and Waldrop 2003), the 
Veterans Administration reports that approximately twenty-three percent of all homeless 
                                                 
1
 ?The majority of U.S. veterans (54.7 percent) were employed in 2000, as shown in Table 1.7 reflecting 
the ties between age and period of service, and age and employment,veterans who served most recently 
were most likely to be employed in 2000. Among veterans serving in August 1990 or later, 81.4 percent 
were employed, while 82.7 percent of those who served from September 1980 to July 1990 were 
employed. They were loosely followed by veterans who served from May 1975 to August 1980 (78.0 
percent). More than three quarters (75.4 percent) of veterans of the Vietnam era were employed in 2000, 
and more than half (51.4 percent) of those who served from February 1955 to July 1964 were employed. 
26 
 
   
adults have served their country in the armed services.  Furthermore, the VA has 
suggested that as many as more than 250,000 veterans may be homeless on any given 
night and that twice as many veterans experience homelessness over the course of a year.  
Therefore, a small but important part of the veteran population has been systematically 
excluded from the study of veteran social status attainment.  
Several studies have failed to control for period of service (Lopreato and Poston 
1977; Poston, Segal, and Butler 1984; Prokos 1996; Xie 1992).  The findings in these 
studies are generally attributed to veterans of all periods when they might in fact be more 
attributable to veterans of specific service periods, ages, or some combination of the two.  
The studies that have disaggregated the data by period of service have shown this to have 
a great effect on findings.  For example several studies, (Martindale and Poston 1979; 
Villemez and Kasarda 1976) found that white veterans of the World War II and the 
Korean War eras earned more than their non-veteran peers; however, Vietnam War 
veterans tended to do worse than their non-veteran peers. Similarly, some studies that 
have considered only a particular period (Angrist 1993) may not be generalizable to all 
periods of veteran service. 
Although some studies have controlled for geographic region, most have used 
national data and not made any attempt to understand how geographic region of origin 
and place of residence have affected veteran social status attainment.  These studies 
assume that income in one area is equal to income in another area when regional and 
local labor market differences may play a significant difference on veteran attainment.  
These regional and local differences may make a difference from several perspectives.  
For example, Booth et al. (2000) showed that women employed in labor markets with a 
                                                                                                                                                 
The percentage employed was low for Korean War veterans (24.6 percent) and World War II veterans 
27 
 
   
high military presence suffer an earnings penalty compared to their peers working in non 
military labor markets.  Moreover, it has been show that those labor markets with high 
military presence are less segregated in terms of residence and employment (Segal and 
Segal 2004:7).  Additionally, some states and regions of the country historically have 
provided more military manpower than others and thus may be more veteran friendly.  
These factors may have significant bearing on veteran social status attainment.   
 
Hypotheses 
 
On the basis of the literature on veteran status attainment and the theory 
developed with regard to human capital, the life course, and bridging the following 
hypotheses reflect my expectations regarding the effects of comparisons between male 
World War II veterans and non-veterans over the six decennial census periods since the 
war.  Although research findings on various aspects of veteran status attainment have 
been inconsistent, the bridging environment hypothesis suggests that, overall, veterans 
(particularly those who are more disadvantaged from the outset) should achieve greater 
socioeconomic status than their non-veteran counterparts.  Using earnings income, 
education, and occupational prestige as indicators, I hypothesize the following: 
1. As HCT and previous research suggest controlling for background factors 
such as age, race, regional residence, marital status, and education level, 
World War II veterans attain greater social status than their non-veteran peers. 
More specifically, veterans will earn more income, education, and 
occupational prestige than their non-veteran peers due to the tangible and 
intangible skills learned while in service and the subsequent government and 
societal benefits (e.g. the GI Bill) provided after service
                                                                                                                                                 
(11.6 percent), many of whom were likely to be retired? (Richardson and Waldrop 2003: 4). 
28 
 
2.  As the bridging hypothesis, HCT, and previous research suggest, black 
veterans receive more of a social status attainment premium relative to black 
non-veterans than white veterans relative to white non-veterans. The black 
veteran advantage will manifest itself in increased income and educational 
opportunities. As Sampson and Laub (1993) suggest, military service allows 
disadvantaged groups to knife off their pasts, redirect their life courses, and 
inhibit deviant behavior thereby enhancing opportunities for increased social 
status attainment. 
3. As the life course perspective suggests, those whose lives are disrupted the 
most will be affected the most by military service. Older and presumably 
more established persons who served during World War II had their lives 
disrupted to a greater degree than younger ones.  Therefore, veterans whose 
birth cohorts were drafted during the demobilization phases of World War II 
will have greater status attainment than those whose birth cohorts were 
drafted during the initial and peak mobilization phases of World War II. This 
suggests that the salience of social status opportunities are diminished for 
those born to the earliest World War II birth cohorts and that social status 
attainment outcomes and life trajectories are profoundly affected by timing in 
the life course.  
4.  Because the military selected only those who met specific health and 
intellectual requirements there is a chance of seeing the effects of selection 
bias as opposed to veteran effects.  However, selectivity bias will be lowest 
where the largest proportions of veterans serve in a cohort. Furthermore, when 
29 
 
   
selection bias is low, if the bridging hypothesis and HCT perspectives hold 
prevail, one would expect that World War II veterans born to cohorts with 
larger proportions of veterans should achieve more social status attainment 
than those with a lower proportion of veterans. 
 
 
30 
 
   
Chapter 3:  Methods 
 
Data 
 Given the limitations of other data sources for studying the full careers of World 
War II veterans and their non-veteran peers I have determined that using census data is 
the most effective and efficient way to conduct this study of veteran attainment. My main 
struggle with this data set initially was that did not provide me with a starting point for 
comparing veterans.  Most of the previous literature has relied on longitudinal panel 
studies.  However, I have reasoned that the starting point for a veteran group could 
feasibly be a particular census.  In my case I will start with the 1950 census and use the 
1960, 1970, 1980, 1990, and 2000 censuses as points of comparison. 
The data for this study were obtained from the Integrated Public Use Microdata 
Sample (IPUMS) that are subsets of the 1950 through 2000 Decennial Censuses.  The 
IPUMS data Series (IPUMS-98) was developed at the University of Minnesota in 
October 1997.  The IPUMS consists of thirty-seven representative samples drawn from 
the American population over fifteen federal censuses.  The Census PUMS data that I 
will use contain a 1-in-100 weighted, national, random sample of the population.  The 
smallest identifiable geographic unit of the PUMS data is the Super-Public Use 
Microdata Area (PUMA) which contains at least 400,000 persons.  However, PUMS data 
are set up to allow researchers to have access to random sub-samples of both the 1 
percent and 5 percent samples.  These sub-samples can be ?tiny,? as described by the 
Census Bureau with as few as 5,000 adults and households.  For this study I will use a 
?one percent? which contains approximately 744,000 household and 2,267,000 person 
records (Ruggles, Sobek, Alexander, Fitch, Goeken, Hall, King, and Ronnander 2004).  
31 
 
   
All of the IPUMS samples are cluster samples.  These samples are based on households 
or dwellings.  The IPUMS data allow for analysis of individual records on topics such as 
fertility, household composition, and nuptiality about multiple individuals within the 
same household.  The IPUMS samples are also stratified in that ?they divide the 
population into strata based on key characteristics, and then sample separately from each 
stratum.  This ensures that each stratum is proportionately represented in the final 
sample?(Ruggles et al. 2004:2.1).  The 1960 and subsequent samples, which include the 
1970, 1980, 1990, and 2000 samples that will be used in this study, employ more 
elaborate stratification schemes than earlier samples.  The latter samples are based not 
only on geography but also on such characteristics as household size, race, and group 
quarters membership.  Additionally, the 2000 decennial Census allowed for multiple race 
categories to be entered as responses.  The IPUMS allows for integration of these 
variables to ensure a seamless comparison across the six decades. 
 
Data Analysis 
 
 Data analysis for this dissertation will focus on the central question of what are 
the differences in attainment between World War II veterans and their non-veteran peers  
This analysis will have several parts.  First, analysis will integrate descriptive statistics 
that define the characteristics of the sample and the distributions of the variables to be 
analyzed.  Second, a correlation analysis that includes all of the model variables and 
establishes the basic zero-order relationships among the variables will be conducted.  
Finally, a multivariate regression analysis (OLS) will be employed to determine the 
effect of veteran status, background factors, (age, race, marital status, regional residence, 
32 
 
   
education) birth year, and percentage of veterans in the year in predicting each of the 
dependent variables of income, education, and occupational status.  Effects by race 
(black vs. white) are also analyzed using multivariate regression techniques.   
As a result of my decision to exclude several demographic groups from the 
sample I have limited the scope and application of this to males who were of service age 
during World War II.  
I limited the sample to men because women were not asked their veteran status 
until the 1980 Decennial Census. If I had included them I would not have been able to 
capture the same data that are available for men over the same period.  
I limited the data to black and white veterans because the data is available to 
conduct a study of these two groups over a long period of time. The racial and ethnic 
categories for Hispanics and other minority groups have changed multiple times between 
the 1950 and 2000 Censuses and would therefore be difficult to disentangle in this study. 
As a result it would be very difficult, given the data at hand, to truly understand the 
nature of the relationships between veteran social status attainment and other minority 
groups.    
I limited the study to those who had some positive income and were working 
because it is impossible to take the log of a negative number or zero. As a result this 
study does not capture the social status attainment of those with no income, the homeless, 
those who dropped out of the labor force for any other reason, and those who only 
received benefits in kind. 
This study also excludes any respondent born prior to 1900 or after 1930. The 
purpose of eliminating these respondents was twofold.  The first purpose was to prevent 
those born in the 19
th
 century from skewing the data. Moreover, if these respondents were 
33 
 
   
left in the sample the comparisons would have been a comparison of veterans to the 
general population rather than the intended comparison of veterans to their non-veteran 
peers. The second reason was to allow my final analysis of the 2000 Census to contain no 
one older than 100 years old. 
This decision to make these exclusions was by no means arbitrary and in fact was 
done in consideration of the data available and the scope of the problem. However, the 
exclusions by race, gender employment status, and positive income all come at the cost 
of degrees of freedom, reduction in sample size, and a clearer understanding of veteran 
social status attainment in the aggregate. In each chapter I provide a brief overview of the 
exclusions and the resulting number of respondents that remain as a result of the 
exclusions. 
 
Variables 
 I use fifteen variables in this study to test the stated hypothesis.  They are income, 
education, Duncan Socioeconomic Index score, veteran status, race, age
2
, marital status, 
regional residence, education attainment, mobilization phase, birth year, percent of 
veterans in a birth year, and three interaction terms that account for the interaction of 
veteran status and race and veteran status and mobilization phase.  My main dependent 
variable is wages, income, and salary earnings.  Additionally, I use education and 
occupational status as dependent variables, recognizing that education can affect income 
and occupational status as well.  
   
 
 
34 
 
   
 
Dependent Variables 
 
Earnings Income and Ln (Earnings) - Indicates each respondent's total pre-tax wage 
and salary income (money received as an employee) for the previous calendar year.  This 
includes wages, salaries, commissions, cash bonuses, tips, and other money income 
received from an employer.  It does not include any payments-in-kind or reimbursements 
for business expenses, or retirement income. In order to normalize this variable and 
reduce its skewness I use the natural logarithm of income for the descriptive as well as 
the regression analysis.  
Given that most of the relationships between earnings and veteran status will be 
statistically significant, because of the large size of the sample, I have created criteria for 
distinguishing between statistically significant and substantially important relationships.  
For comparisons between veterans and non-veterans and white veterans and white non 
veterans I used a one thousand dollar difference and for blacks I use a seven hundred 
dollar difference. The typical annual raise for service members is about three percent per 
year. This figure allows military pay to keep pace with inflation and to a lesser degree 
with civilian pay. I used the same three percent figure as evidence of a substantially 
important difference in earnings. The one thousand dollar (ln $6.91) figure represents 
approximately three percent of the median income ($31,267) for all male householders in 
the 2000 decennial census.  Additionally the 2000 decennial census revealed that black 
householders earned slightly less than seventy percent of what white householders 
earned. Therefore, I use a difference of seven hundred dollars (ln $6.55)  as the criteria 
for distinguishing between statistically significant and substantially important 
35 
 
   
relationships for blacks.   
 
Education ? Indicates the respondent's highest completed level of formal education. I 
used the original IPUMS coding of this variable.  The IPUMS coding scheme includes 
nine categories of education (see table 3.1).   
 
 
 
 
 
 
Code Grade Level 
0N/A
1 None or preschool
2 Grade 1, 2, 3, or 4
3 Grade 5, 6, 7, or 8
4G 9
5G 10
6Grade1
7G 12
8 1 to 3 years of college
9 4+ years of college
Table 3.1:  IPUMS Coding Scheme 
 
 
 
 
The IPUMS education coding scheme is a combination of two separate IPUMS 
variables that measure educational attainment in different ways. One measure that was 
available for 1940 through 1980 reports the respondent's highest grade of school or year 
of college completed. The other, available for the 1990-2000 censuses reports the 
respondent's highest grade of school completed through 11th grade, but classifies high 
school graduates according to their highest diploma or degree earned. The IPUMS 
education coding scheme used throughout this dissertation is essentially a compromise to 
bridge census periods and make education comparable over the 1950-2000 census period. 
According to IPUMS, ?EDUCREC was created to facilitate analysis of data from the 
1990-2000 censuses and the ACS (EDUC99) in conjunction with data from earlier years 
contained in HIGRADE? (Ruggles et al. 2004). 
36 
 
   
Furthermore, I use an increase of one as indicated by the IPUMS coding scheme 
(table 3.1) to differentiate between substantive differences and statistically significant 
relationships. 
 
Duncan Socioeconomic Index (SEI) Score ? This is an interval variable that has a 
minimum value of one and a maximum value of ninety-six.?  
The SEI is based on the 1950 occupational 
classification system, is a measure of occupational 
status based upon the income level and educational 
attainment associated with each occupation in 1950. 
The score was derived by using median income and 
education levels for men in 1950 to predict prestige 
assessments from a 1947 survey (of a select group 
of occupations). The resulting statistical model was 
used to generate scores for the entire range of 1950 
occupations. Occupation Prestige - SEI is a 
constructed variable that assigns a Duncan 
Socioeconomic Index (SEI) score to each 
occupation.? (Ruggles et al. 2004: 127-128).    
 
As with the other dependent variables it is very likely that most of the 
relationships between SEI and veteran status will be statistically significant because of 
the large size of the sample. As is the case with education I have determined to use an 
increase of one full point on the Duncan scale as evidence of a substantive relationship 
between SEI and veteran status.  The reason for using one full point is that the Duncan 
scale does not easily lend itself to being broken at any particular point.  
37 
 
   
Independent Variables  
 
Veteran Status ? This is a dichotomous variable that describes whether or not the 
respondent identified himself as a World War II veteran, where World War II veteran is 
defined as those persons who were engaged in active-duty military service in the armed 
forces of the United States during the World War II era (between September 1940 and 
July 1947)  (Ruggles et al. 2004). Non-veterans are coded as 0 and veterans are coded as 
1. In this data set, National Guard and Reserve service is included only if individuals 
were called to active duty. Additionally, the census defines service as follows: 
 
?Service" is defined as active duty in the United States Army, 
Air Force, Navy, Marine Corps, or Coast Guard for any length 
of time and at any place at home or abroad. The following 
types of service were not to be reported as military service: (1) 
persons whose only service was as a civilian employee of the 
armed forces or volunteer for the Red Cross, USO, Public 
Health Service, or War or Defense Department; (2) those 
whose service was limited to National Guard units (except 
when these were called to active duty, as during World War II 
and the Korean War); (3) those whose service consisted only of 
reserve training such as duty for 2 weeks during the year or 
attendance at weekly reserve meetings; and (4) those who 
served only in the armed forces of a foreign country. Service as 
a Merchant Marine Seaman was not considered active duty 
until the 1990 census, when World War II Merchant Marine 
service was defined as active duty for purposes of defining a 
"veteran." No other Merchant Marine service was defined as 
active duty.  
 
Age
2
 - This is a continuous variable that identifies the respondent?s age in years 
(squared) as of their last birthday prior to or on the day of enumeration. I use age
2
 as 
opposed to age in this study, as is convention, to control for the curvilinear effect of age 
and income. This helps to control for the fact that wages typically peak some time during 
middle age and then decrease with time.  
38 
Race - This is a dichotomous variable that is used to compare social status attainment of 
blacks and whites. Whites are coded as zero and blacks are coded as 1.  
 
Mobilization Phase 2 (peak) ? This is a dichotomous variable that is used to compare 
respondents serving during the World War II peak mobilization period (1916 through 
1926) versus those serving during other periods.  
 
Mobilization Phase 3 (demobilization) ? This is a dichotomous variable that is used to 
compare respondents serving during the World War II demobilization period (1927 
through 1930) versus those serving during other periods of World War II.  
 
Control Measures 
Geographic region ? These are dichotomous variables that identify the respondents 
current region of residence.  I broke the data into three distinct regions: North, South and 
West to control for any regional differences in income, education, or SEI differences.   
 
 Marital Status: is a dichotomous variable that indicates each person?s marital status.  
Although there are several categories of marital status available in the data the only 
applicable statuses to this study are the married, divorced, widowed and single.  If a 
respondent has a particular marital status he is coded as 1 otherwise he is coded as 0. 
 
Education ? This is a categorical variable that describes a respondents highest level of 
education (see above). In the regression analyses only high school graduate and some 
college categories are considered (see above for discussion of other categories).   
39 
 
   
Interaction Variables  
 
 In order to incorporate the joint effect of veteran status and some of the other 
variables, I created three interaction terms.  They are veteran*race, veteran*mobilization 
phase 2, and the veteran*mobilization phase 3 interaction terms. The race*veteran 
interaction term was designed to measure the effects of being a black veteran in the 
model. Hypothesis number two, which is rooted in the bridging hypothesis, HCT, and the 
previous research states that black veterans receive more of a social status attainment 
premium relative to black non-veterans than white veterans relative to white non-
veterans. This interaction term serves to provide a measure for the test of this hypothesis  
  The veteran*mobilization phase interaction terms were created to capture the 
effects of being a veteran as well as being from either the peak mobilization phase or 
demobilization phase of World War II.  Previous literature and the life course theory 
suggest that those whose lives are disrupted the most will be affected the most by 
military service.  These variables allow me to separate older and presumably more 
established cohorts from younger cohorts that served during World War II to test 
hypothesis three.  
 
 
40 
 
   
Statistical Methodology 
 
 In order to test the above hypotheses I conduct several phases of statistical 
analysis. The first phase I describe the sample disaggregating it by veteran status and 
race. I calculated the means or proportions and standard deviations for all of the central 
variables in the data. I then compared veterans to non-veterans, blacks to whites, black-
veterans to black non-veterans and white veterans to white non-veterans using 
significance tests for the difference of means or proportions as appropriate.  
 While the bivariate comparisons provide valuable insight as to the social status 
attainment of the aforementioned group they do not allow me to conduct an analysis of 
the veterans advantage or disadvantage while simultaneously controlling for other 
background, birth year, or interaction variables. Therefore I used multivariate regression 
(OLS) to conduct an analysis of the impact of veteran status net of the effects of the 
variables described above.  
 The multivariate regressions are conducted using five models that are rooted in 
the theory and research discussed in Chapter 2 and designed to test my hypotheses (see 
above).  
Furthermore, the regression models control for factors associated with earnings, 
education and Duncan SEI outcomes and determines the net premium or penalty to 
veterans and non-veterans (see Figure 3.1).  
  
 
41 
 
   
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42 
 
   
The first model consists of a dependent variable (income, education, or SEI) regressed on 
World War II veteran status, age
2
, and race. In this and all subsequent models I use age 
squared as opposed to age in order to control for the fact that income typically rises with 
age and then levels out. Additionally, I use the natural logarithm of wages and earnings 
income as opposed to income because it more closely approximates the normal 
distribution (see figure 4.4 and discussion above).  
 The second model consists of the dependent variables regressed on World War II 
veteran status, age, race, and background variables, including dichotomized marital status 
variables (married and single), and dichotomized regional variables (north and south). I 
removed the widowed and west variables in all of the regression equations because they 
caused a dependency problem among the independent variables in the proposed model. 
Additionally, in all of the regressions that include education as a dependent variable I 
removed the education independent variables.  
 The third model adds two dichotomized mobilization phase variables 
(mobilization phase number two (peak), and mobilization phase number three 
(demobilization) and an interaction term for race and veteran to the regression equation. 
The interaction term is coded so that it measures the joint effect of being black and a 
veteran. The fourth model includes all of the model two variables, the interaction term 
and adds a variable that controls for the percent of veterans in a particular birth year.  
Table 5 (above) shows the percentage of veterans born in each birth year.  
 The fifth and final model regresses a dependent variable on World War II veteran 
status, age, and race, dichotomized martial status variables (married and single), 
dichotomized regional variables (north and south), the interaction term for race and 
veteran, and two interaction terms for World War II veteran status and mobilization 
43 
 
   
phase. The veteran* mobilization phase interaction term is coded so that it measures the 
joint effect of being a veteran and a part of a birth cohort that was a part of a particular 
mobilization phase.  
 Throughout this dissertation when comparing between models within census a 
year I make the comparisons using standardized coefficients. Furthermore, throughout 
the dissertation I identify standardized coefficients by using an uppercase ?B.? I make 
comparisons across census years using unstandardized coefficients and I identify 
unstandardized coefficients using a lowercase ?b.?  
 
Selection Bias 
X
Service
Status 
Attainment
Service as a moderating / intervening variable 
X Service
Status 
Attainment
Service as a antecedent variable 
Service as a spurious variable 
X
Service
Status 
Attainment
Figure 3.2: Selection Bias 
One of the well documented problems with conducting analyses of  this type is 
the potential for selection bias (Cohen, Segal, 
and Temme 1986; Cohen, Segal, and Temme 
1992; Cooney 1997; Heckman 1979).   
Heckman (1979) stated that there are two types 
of selection bias that are often found in 
sociological work.  The first is found when the 
persons or units being investigated self-select 
themselves or are selected by someone else in 
or out of the sample being studied.  This can also be described as an antecedent variable.  
An antecedent variable is one that comes before the dependent variable and may need to 
be controlled (see Figure 1.2).  For example, research has shown that during the World 
War II era as well as the Vietnam era many men from the highest socioeconomic groups 
self-selected out of military service through educational, occupational, and medical 
44 
 
   
deferments; others applied for conscientious objector status.  Simultaneously, those from 
the lower social strata were then and are now selectively excluded from service because 
the armed forces only accept healthy members of the population who possess a minimum 
aptitude as determined by the services.   
One might argue that those who ultimately chose and were accepted into the 
armed forces would have done well even without veteran status, and that the attributes of 
being healthy and obtaining a certain level of education were the causes of their 
attainment, not their military status.  However, some have concluded that during 
conscription the opposite type of selection bias was in effect.  In conducting their 
research on Vietnam veteran attainment, Cohen et al. (1986) surmised that the 
continuation of educational deferments produced the century?s most educationally 
unrepresentative wartime force.  They believed that higher education allowed some to 
avoid the draft, which may have made the relationship between military service and 
educational attainment more negative than it otherwise may have been had the draft been 
more equal.   
It is very difficult to control for selection bias; however, one could set up the ideal 
type study to ascertain veteran social status attainment by finding two subjects with the 
same backgrounds.  For example, one would look for others with the same class, 
educational background, intelligence, parental situation, parental occupations, and 
parental education.  In fact, the ideal subjects might be twins who were raised in the same 
house, attended the same schools with similar outcomes, and had similar health 
conditions.  Moreover, it would be ideal if one of these twins finished high school and 
went into the service while the other finished high school and went to work in the civilian 
sector.  The researcher would then test to see if there were differences in the social status 
45 
 
   
attainment of these two subjects. In fact, the National Academy of Sciences-National 
Research Council has a twin registry program. It was designed to study the effects of 
heredity and environment on twins who were veterans of World War II and was based on 
the work of Francis Galton in the late nineteenth century. This program has collected 
information on white males since the 1950s (Institute of Medicine 2006). However, even 
these types of studies still suffer from selection bias and generalization problems.   
The second type of selection bias occurs when the analyst or data processor omits 
some of the sample based on selection decisions (Heckman 1979). An example of this 
can be illustrated by comparing World War II veterans to their non-veteran peers over 
some period of time.  One could argue that morbidity rates for non-veterans might be 
higher than those of their veteran peers, excluding combat deaths, because only healthy 
members of the population were conscripted.  Fewer veterans would die off and their 
earnings at later periods of time would conceivably be higher because they were healthier 
and outlasted the non-veteran population during the studied time period.  
 In this dissertation both of these problems, if not carefully controlled, have the 
potential to invalidate the findings.  Veteran social status attainment research has 
generally employed a methodology that involves selecting a random sample; dividing 
that random sample by veteran status; conducting an analysis of the two populations 
while controlling for certain variables; and attributing any differences to veteran status.  
However, using this method to determine the effects that veteran status may have on 
socioeconomic attainment leaves room for selection bias as well as drawing the 
conclusion that military service leads to outcomes that are in reality based upon 
background factors that the military uses to selectively select its members (Cohen, Segal, 
46 
 
   
and Temme 1986; Cohen, Segal, and Temme 1992; Cooney, Segal, Segal, and Falk 2003; 
Heckman 1979; Xie 1992). 
 Several researchers have taken steps to mitigate the problems associated with 
selection bias by creating a probit model of the selection process and then constructing a 
hazard rate that represents the instantaneous probability of being excluded from the 
sample, conditional upon the agent in the pool at risk.  The hazard rate is treated as a new 
variable and included in all substantive equations. However many researchers (Cohen, 
Segal, and Temme 1986), who used this method to determine the effects of veteran status 
on attainment found that using the hazard rate as a control for selection bias did not 
change the magnitude, direction, or significance of the regression coefficients and thus 
the relationship between military service and attainment was not due to selection bias. 
 The prudent researcher must take the necessary steps to control for selection bias 
when comparing veterans to non-veterans.  As Cooney (1997) pointed out, ?the bridging 
hypothesis would predict that black veterans would gain more than white veterans from 
military service, but the same results might be achieved through selection bias alone in 
the AVF era.?   
 
 
47 
 
   
Chapter 4: World War II Veterans and the 1950 Census 
 
 
Enumeration Problems with the 1950 Census  
 
 This study of World War II veteran attainment begins with the 1950 Decennial 
Census because it was the first census following World War II in which World War II 
veterans could have claimed veteran status. However, this study does not use the 1950 
Decennial Census to draw baseline descriptions about the World War II veteran 
population for several reasons. The first and most important of these reasons is the fact 
that the veteran population in general and the World War II veteran population in 
particular were underenumerated in the 1950 Decennial Census.  Table 4.1 shows the 
changes in the number of veterans who reported themselves as such between the 1950 
and 2000 Decennial Censuses in the IPUMS dataset.  The 1960 census provides a more 
realistic estimate of the numbers of veterans in the population, while subsequent censuses 
reflect the mortality of veterans. 
 
Table 4.1: Veterans in 1950-2000 IPUMS Data Sets 
 
 
Census Year 1950 1960 1970 1980 1990 2000
Number of WWII 
Veterans in Sample
43,020 129,916 129,148 92,505 50,207 11,445
Annual % Change 201.99% -0.59% -28.37% -45.73% -77.20%
 
 
 There are several hypotheses as to why veterans were underenumerated in the 
1950 Census. The United States Census Bureau cites three reasons for the 
underenumeration: Inconsistent reporting, underreporting, and the location of the 
question on the 1950 census form.   
48 
 
   
Data on veteran status in the 1940 and 1950 censuses were not 
satisfactory. Age and type-of-service entries were sometimes 
inconsistent, and nonreporting was relatively high. Furthermore, 
the totals did not correspond well with figures compiled by the 
Veterans Administration. The location of the question on the 
schedule may have been responsible for some of the errors. 
Because of these problems, the results of these original inquiries 
were not published (Ruggles et al. 2004: 249; U.S. Bureau of the 
Census 1955: 64). 
 
Underreporting 
 
 Inconsistent reporting and underreporting could have been due to the fact that the 
data was collected, edit, coded, and tabulated manually by enumerators who were given 
vague instructions. For example the instructions for enumerators with regard to ?How to 
Enumerate Special Types of Living Quarters,? which included military installations, 
stated ?Do not make any entries on the Population and Housing Schedule. Persons will 
be enumerated on Individual Census Reports (ICR) by special assignment? (U.S. Bureau 
of the Census 1955: 21).  Moreover, the Census Bureau never inspected the entries on 
military service during either the screening or the editing processes. The rational used to 
forego inspection of these data was the fact that the respondent?s answers were limited to 
yes, no, and blank and therefore not subject to the same errors as other types of data (U.S. 
Bureau of the Census 1955: 64). 
 Additionally, the ICR (see figure 4.1) that was administered to service members 
serving overseas never asked the veteran questions administered on the census forms in 
the continental United States. Interestingly, question number twenty-two on the 1950 
Form P4 that was administered to service members serving aboard ?Vessels at Sea? did 
ask the veteran questions (see figure 4.2). Moreover, there is no record of what the 
Department of Defense, which was engaged in a War in Korea, did to ensure a high 
response rate.      
49 
 
   
50 
Figure 4.1: Census Form 5 "Overseas Census Report "
 
   
51 
Figure 4.2: Census for P4: "Crew of Vessels Report"
 
   
 
G.I.s at College 
 
 Another plausible reason for the underenumeration could have been the large 
number of veterans who were in institutions of higher education as result of the 
Servicemen?s Readjustment Act of 1944.  President Franklin D. Roosevelt signed the bill 
into law on June 22, 1944. Its primary purpose was to ease World War II veterans? 
transitions back into civilian life by providing education, training, guaranteed home 
loans, unemployment compensation, and job finding assistance.  
 The impetus behind this bill was twofold. It was designed as a social welfare 
program to reduce the economic burden caused by the demobilization of millions of 
service members (Segal 1989: 87-88). Secondly, it was designed to alleviate some of the 
problems that resulted from the World War I demobilization.  Returning World War I 
veterans received a sixty dollar allowance and a train ticket home making their return to 
civilian status a ?rude and bitter? experience, punctuated by a recession, and followed by 
the Great Depression of the 1930?s. These conditions led to several protests in the 1930?s 
(Diehl and Ward 1975). One of the goals of the G.I Bill was to prevent this from 
occurring again. 
  In the peak year of the program, 1947, veterans accounted for forty-nine percent 
of college enrollment. Of the 15.4 million veterans in the population approximately 7.8 
million were trained, including: 2,230,000 in college, 3,480,000 in other schools, 
1,400,000 in on-job training, and 690,000 in farm training (United States Department of 
Veterans Affairs 2005). Some estimate that the influx of veterans into the education 
system doubled the number of college students at that time (Blair 1999). In fact, at some 
institutions of higher education veterans comprised the majority of the student body. For 
52 
 
   
example, in 1945 veterans made up eighty-nine percent of the New York University?s 
student body (Gambone 2005: 69).  These trends are clearly reflected in the percent 
change in the number of respondents who stated that they had some college or a college 
degree between 1950 and 1980 (see figure 4.3). 
 
248.89%
4.95%
-11.90%
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
Percent Change
% change 1950 - 1960 % change 1960 - 1970 % change 1970 - 1980
Census Period 
World War II Cohort Percent Change in Some College or College Degrees 1950-1980
Figure 4.3: 
 It is therefore a reasonable hypothesis that veterans could have been 
underenumerated because they were in educational institutions at the time of the 1950 
census.  Data from the 1950 through 2000 Decennial Censuses seem to bear this out. 
Figure 4.3 illustrates the dramatic rise in percent change of World War II veterans with 
some college or college degrees between 1950 and 1960 (248.89 percent) which dropped 
significantly thereafter. The drop in the percent change of World War II veterans with 
some college or college degrees after the 1960 Decennial Census could also reflect a 
natural progression of the life course that includes traditional ages at which people tend 
to enroll in colleges and universities as well as mortality rates. The 248.89 percent 
53 
 
   
change in World War II veterans with some college or college graduates also matches the 
201.99 percent change in the number of respondents reporting that they were World War 
II veterans between the 1950 and 1960 Decennial Censuses. 
 Clearly the underenumeration of veterans in the 1950 Census mandates the use of 
the 1960 Census for baseline comparisons and measurement of change over the life 
course. However, the 1950 Decennial Census is good place to start from a life course 
perspective because it is at this point that the life trajectories of many World War II 
veterans began to change. As the preceding paragraphs illustrate many veterans left the 
service and went to college; some joined the labor force; and others remained in the 
service.  
 In the remainder of this chapter I will briefly describe the general political, social, 
and economic landscape of the period after World War II and leading up to the 1950 
Decennial Census, describe the sample, compare the dependent measures, and examine 
the dependent variables through multiple regressions.           
 August 15, 1945, V-J Day, officially ended World War II for the United States. 
At the end of World War 16 million veterans turned their thoughts to home. Much had 
changed politically, socially, and economically over the course of the war in the United 
States.  
 
Social Changes 
 Many social changes occurred from the early to late forties that impacted the lives 
of veterans. The United States? population grew from 132,122,446 in 1942 to 
149,188,130 in 1949 representing a 12.92 percent population increase.  
54 
 
   
  In 1940, males who were over twenty-five years of age had 8.3 years of 
education completed (median) and by 1947 this number had increased to 8.9 years 
(median). In 1942 the total number of students enrolled in institutions of higher education 
(universities, colleges, and professional schools) was 1,403,990. By 1947 that number 
had increased by thirty-four percent to 1,882,505. The median wage or salary income in 
1940 for white males with ages between 25 and 64 was $1,217, which is equal to 
$8695.47 in year 2000 dollars. Black male?s median wage or salary income in 1940 was 
$520 or $3715 in year 2000 dollars. By 1947 these numbers had increased to $2357 and 
$1279 respectively (United States Census Bureau 1945; United States Census Bureau 
1949).   
 One of the biggest social changes was the fact that women were a larger part of 
the American workforce than at any other time in history. Rosie the Riveter was 
emblematic of the American woman working on the home front to free a man for combat.  
Although many of the women who worked during the war lost or voluntarily gave up 
their jobs at the end of the war; women?s post war employment rates were much higher 
than their prewar employment rates (Gambone 2005).  In 1933, President Franklin D. 
Roosevelt began implementing the New Deal, a broad set of sweeping social and 
economic reforms designed to bring the United States out of the depression. By the end 
of World War II most of these programs were in place and being implemented at both the 
state and federal levels.  
 However, the most important piece of legislation for veterans during this period 
was arguably the Servicemen?s Readjustment Act of 1944, more commonly referred to as 
the G.I. Bill of Rights. It was designed to help veterans transition back into civilian life 
by helping them in three areas including educational training, unemployment 
55 
 
   
compensation, and loan guarantees for homes, farms and businesses. Any veteran with at 
least ninety days of service and a discharge from the service that was characterized as 
anything other dishonorable could qualify. Veterans used these benefits to the fullest 
extent and in 1946 the Veterans? Administration reported that 6.6 million veterans had 
applied for school or job training (Gambone 2005). 
 With all of the benefits that veterans received black veterans returned home to a 
country that many perceived as not having changed a great deal. Black veterans had 
fought in a segregated army and returned home to a segregated country, where much of 
their social life was ruled by the phrase ?separate and unequal.? World War II veterans 
served in segregated units and the policy of integration did not begin until 1950 during 
the Korean War. 
 
Economic Changes   
  The United States emerged from World War II as an economic superpower. By 
the end of World War II it was one of the only countries that could boast great economic 
growth. During the war the United States was producing more in one day than it had 
produced in a year prior to the war. The Gross Domestic Product grew from $101.4 billon 
in 1940 to $293 billion by 1950. Americans had saved a great deal because of rationing 
as well as the shortage of supplies. As a result, some Americans were saving as much as 
25 percent of their income and at the end of the war Americans had total personal savings 
in excess of 140 billion dollars (Gambone 2005: 22 - 23). This resulted in a personal 
consumption spending increase of $120 billion between 1940 and 1950.  Both imports 
and exports increased dramatically from $3.4 billion to $11.6 billion and $4.9 billion to 
$12.4 billion (United States Bureau of Economic Analysis 2005).  
56 
 
   
57 
 Although the economic news was generally good there was some bad new as 
well. ?In the first ten days of peace, 1.8 million Americans lost their jobs, and 620,000 
filed for unemployment insurance ?. By February, 1946 American economic 
productivity had fallen 31 percent below its peak in the previous June? (Gambone 2005: 
30).  However, even with these job losses, the unemployment rate in 1949 was a meager 
3.8 percent, down 284 percent from the 14.8 percent unemployment rate in 1940. 
 Figure 4.4: Earnings Income for Veterans and Non-
Veterans 
 
0
2.
0e-
05
4.
0
e
-
0
5
6.
0e-
05
0 20000 40000 60000 80000 0 20000 40000 60000 80000
No or N/A Yes
Density
normal incinf
# of
 
Res
p
ondent
s
Earnings Income
Graphs by Veteran, served WWII era
 
 
 
 
 
 
 
 
Figure 4.5: Natural Logarithm of Earnings Income for 
Veterans and Non-Veterans 
 
0
.5
1
1.
5
6 8 10 12 6 8 10 12
No or N/A Yes
Density
normal loginc
# of
 
Res
p
ondent
s
Earnings Income
Graphs by Veteran, served WWII era
 
 
 
 
   
Descriptive data 
  The first step in this analysis is to calculate the descriptive statistics of the 
sample by comparing the means or medians and standard deviations of the key variables. 
This sample initially had 579,233 respondents.  After excluding respondents who 
answered ?N/A? to the following question, ?Last year (1949), how much money did he 
earn working as an employee for wages or salary,? I was left with 153,157 respondents. 
None of the 426,076 respondents who were excluded for having ?N/A? wages in the 
1950 Census were World War II veterans. In order to ensure that the distributions were as 
close to normal as possible I took the natural logarithm of earnings. One of the 
constraints in conducting the analysis in this way is that all of the respondents with non-
positive and zero earnings and wages income had to be removed from the sample because 
it is impossible to take the log of zero or a negative number.  Figures 4.4 and 4.5 show 
the effects of transforming the income variable to the natural logarithm of income. 
  Of the 153,157 respondents, 44,985 of them had zero or no income; 3, 890 of the 
no income respondents were black and of those 658 were veterans. Of the original sample 
(restricted by age and ?N/A? earnings) the zero earnings veterans represented 21.31 
percent of the total black veterans, 17.76 percent of the white veterans, 30.55 percent of 
the black non-veterans, and 34.16 percent of the white non-veterans. After removing 
those with zero income and zero SEI scores, 101, 343 respondents remained. 33,375 were 
veterans and 67, 968 were non veterans. In the final step, I eliminated any respondent 
born in the 19
th
 century. The purpose of eliminating these respondents was twofold.  The 
first purpose was to prevent those born in the 19
th
 century from skewing the data. 
Moreover, if these respondents were left in the sample the comparisons would have been 
a comparison of veterans to the general population rather than the intended comparison 
58 
 
   
of veterans to their non-veteran peers. The second reason was to allow my final analysis 
of the 2000 Census to contain no one older than 100 years old. By removing those born 
prior to the year 1900 I was left with 74,363 total respondents.  
 Of the 74,363 total respondents 32,712 are veterans who represent approximately 
43.48 percent of the sample and black veterans are approximately 6.62 percent of the 
veterans. Estimates are that in 1945 there were approximately sixty-six million men in 
the United States and approximately fifteen million men or 22.72 percent served. Blacks 
represented about 6.85 percent of the total force at the their greatest strength including 
8.7 percent of the Army, 4 percent of the Navy, and 2.5 percent of the Marine Corps 
(Binkin and Eitelberg 1982: 24).  Figures 4.6 and 4.7 illustrate the before and after effects 
of shaping the data. Although the data is more comparable to the actual numbers in 1950 
after excluding certain demographic categories, the 1960 data still more closely 
approximates the actual ratio of veterans to non-veterans in 1950. Figure 4.8 illustrates 
the 1960 census sample ratio of veterans to non-veterans after making the same 
adjustments that were made for the 1950 data.     
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
59 
 
   
Figure 4.6 1950 Census World War II Cohort Original 
0
2000
4000
6000
8000
10000
12000
14000
16000
1
900
1
901
1
902
1
903
1
904
1
905
1
906
1
907
1
908
1
909
1
910
1
911
1
912
1
913
1
914
1
915
1
916
1
917
1
918
1
919
1
920
1
921
1
922
1
923
1
924
1
925
1
926
1
927
1
928
1
929
1
930
Birth Year
Num
b
er
Non-Veterans
Veterans
Figure 4.7: 1950 Census Cohort with Changes 
0
500
1000
1500
2000
2500
3000
3500
19
00
19
01
19
02
19
03
19
04
19
05
19
06
19
07
19
08
19
09
19
10
19
11
19
12
19
13
19
14
19
15
19
16
19
17
19
18
19
19
19
20
19
21
19
22
19
23
19
24
19
25
19
26
19
27
19
28
19
29
19
30
Birth Year
Number
Non-Veteran 
Veteran
60 
 
   
 
 
 
 
 
 
 
 
 
 
 
Figure 4.8: 1960 Census World War II Birth Cohort 
 
Background Descriptive Statistics and Discussion 
 
Age   
 Sample respondent?s average age is 34.22 with a standard deviation of 8.60 and a 
median age of 33.  On average the veterans in the sample are 30 years old (median). They 
are significantly younger than their non-veteran peers by 6.32 years. Although this 
difference is significant it is also expected.  If one assumes that those who fought in 
World War II were between the ages of 18 and 26 then they would have been between 
the ages of 21 and 34 in the 1950 census (see Table 4.2).  
61 
 
   
  
Table 4.2: Veteran Age by Year of Entry in the 1950 Census  
Year of Entrance 18 19 20 21 22 23 24 25 26
1942 26 27 28 29 30 31 32 33 34
1943 25 26 27 28 29 30 31 32 33
1944 24 25 26 27 28 29 30 31 32
1945 23 24 25 26 27 28 29 30 31
1946 22 23 24 25 26 27 28 29 30
1947 21 22 23 24 25 26 27 28 29
1948 20 21 22 23 24 25 26 27 28
1949 19 20 21 22 23 24 25 26 27
1950 18 19 20 21 22 23 24 25 26
Age of Entrance 
 
 As table 4.3 illustrates, most of the veterans fall into the 25 to 34 year old age 
category and the largest category of non-veterans is the twenty-five to thirty-four year old 
age category. Moreover, it would be relatively rare to find people serving in the armed 
forces over the age of 50; however, a 50 year old serving between 1942 and 1947 would 
have been between the ages of 53 and 58 during the 1950 census. One would expect that 
the non-veterans would be both younger and older than the veterans because they had no 
age restrictions. The sample does reflect the above characteristics with two notable 
exceptions. The exceptions are that some of the veterans are younger than twenty-one 
years old and some are older than thirty-four (see table 4.2). There are several possible 
reasons for this, the most probable being data entry errors (see above discussion of 
underreporting) and the fact that some may have legitimately not known their actual age 
and guessed.  
 White respondents have an average age of 34.21 and they are older than black 
respondents, but not significantly so. Black veterans have mean age of 30.87 years and 
they are significantly younger than their non-veteran black peers, who have a mean age 
of 35.81 years. Similarly white World War II veterans have a mean age of 30.63 years 
and they are significantly younger than their non veteran peers who have a mean age of 
37.12 years.  Table 4.3 provides the median ages for all groups with percentages that fall 
62 
 
   
into four distinct categories by veteran status and race. Table 4.4 provides the results of 
the significance tests described above. 
Table 4.3: Descriptive Characteristics of the 1950 Census Sample 
 
 
63 
 
   
 
Table 4.4: Significance Tests for Age 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
 
Another method for examining age is to use birth cohorts.   The use of birth 
cohorts is a particularly effective way to examine how groups of different ages are 
affected by military service. Using birth cohorts allows us to test for veteran effects while 
simultaneously reducing the effects of selection bias (see chapter 3). If the bridging 
hypothesis holds and military service provides social and human capital that members 
might not otherwise receive, then one would expect that birth cohorts with smaller 
proportions of veterans will receive less status attainment (education, income, and 
occupation) than those with higher proportions of veterans. This use of birth cohorts to 
conduct this type of analysis is often referred to as regression discontinuity design 
(Bound and Turner 2003; Campbell and Stanley 1966).  In the 1960 sample, which is the 
base year (see discussion of underenumeration above) both veterans and non-veterans are 
represented by birth years 1900 - 1930 (see table 4.5).
64 
 
   
 
Table 4.5: Birth Year Comparison of Veterans and Non-Veterans 
 
  
Birth Year Non-Veterans Veterans Percent Veterans
1900 6,147 460 6.96%
1901 7,399 600 7.50%
1902 6,075 507 7.70%
1903 6,788 643 8.65%
1904 7,002 686 8.92%
1905 7,355 890 10.79%
1906 7,182 1,265 14.98%
1907 6,767 1,633 19.44%
1908 6,888 1,935 21.93%
1909 6,896 2,221 24.36%
1910 6,783 2,432 26.39%
1911 6,962 2,696 27.91%
1912 6,764 2,884 29.89%
1913 6,757 3,268 32.60%
1914 6,466 3,680 36.27%
1915 6,328 4,154 39.63%
1916 5,298 4,824 47.66%
1917 4,891 5,472 52.80%
1918 4,428 6,162 58.19%
1919 3,744 7,213 65.83%
1920 3,405 7,646 69.19%
1921 3,111 8,259 72.64%
1922 2,877 8,563 74.85%
1923 2,961 8,448 74.05%
1924 3,057 8,264 73.00%
1925 3,222 8,197 71.78%
1926 3,165 7,919 71.45%
1927 3,280 7,518 69.62%
1928 5,053 6,165 54.96%
1929 7,989 2,535 24.09%
1930 9,535 1,161 10.85%
1931 10,561 0 0.00%
1932 9,815 0 0.00%
1933 9,800 0 0.00%
1934 9,282 0 0.00%
1935 9,730 0 0.00%
1960 Census Data 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
In order to facilitate the regression discontinuity design I have broken the 1960 birth 
cohorts in three distinct categories of analysis that denote the beginning, middle, and end 
of the World War II mobilization. The beginning phase begins with the 1906 birth year 
cohort and ends with the 1915 birth year cohort. I selected 1906 as the beginning birth 
65 
 
   
year cohort because the draft began in 1940 and included those between the ages of 
twenty-one and thirty-six. The thirty-six year olds are a part of the 1906 birth year cohort.  
The middle phase begins with the 1916 birth year cohort and ends with the 1926 birth 
year cohort. I selected 1916 as the next break point for several reasons. First, this is 
where the curve begins to rise markedly (see figure 4.9).   
 
 
 
Figure 4.9: 1960 Census World War II Cohort Mobilization Phases 
 
 
 
 
 
 
 
 
 
 
 Secondly, by 1942 the United States was engaged in World War II and those who 
were 26 years of age would have been prime candidates for the draft. The 26 year olds in 
1942 were a part of the 1916 cohort. The end phase beings with the 1927 birth year 
66 
 
   
cohort and ends with the 1930 birth year cohort. Table 4.5 illustrates how the proportion 
of veterans to non-veterans declines after 1926.  
 
Race  
 As one might expect the majority of the respondents in this sample are white. 
About ninety percent (67,303) of the sample are white and the remainder is black. Blacks 
are approximately 6.62 percent (2,141) of the veteran group and 11.70 percent (4,919) of 
the non-veteran group, while white veterans comprise 93.38 percent (30,191) of the 
veterans and 88.30 percent (37,112) of the non-veteran groups.  These numbers are 
representative of the demographics of both the population at large as well as the armed 
forces demographics in 1950.  Although the number of blacks in the veteran category 
might be viewed as abnormal today, the ratio?s are close to what they would have been in 
the 1950?s (see discussion above). During the period that these veterans served (1942 - 
1947) blacks were limited in the percent that could serve in each branch of  service by 
statute and  the number of units in which they could serve because of segregation. 
 
Region  
 Just over half of the sample (58.34 percent) maintains their residence in the 
South, 27.82 percent in North, and 13.84 percent in the West. The veteran and non?
veteran proportions for region of residency are very similar.  28.32 percent of the 
veterans reside in the North versus 27.43 percent of the non-veterans; 56.50 percent of 
the veterans and 59.76 percent of the non-veterans reside in the South; and 15.18 percent 
of the veterans versus 12.81 percent of the non-veterans live in the West.  
67 
 
   
  I conducted a significance test assuming unequal variance to determine if the 
population proportions for veterans and non-veterans were different in terms of region. In 
this test I combined all of the regions and tested whether there was a significant 
difference between veterans and non-veterans. The results show that the population 
proportions are significantly different, with the proportion for veterans being higher. 
These results were significant at the .001 level. In essence this test points out that 
veterans tend be more from the South and West than non-veterans in this sample.  I also 
conducted a t-test with the disaggregated regional data to compare veterans and non-
veterans by specific region and found that the population means of all of regions have 
significant differences for veterans versus non-veterans (see table 4.6).  
Table 4.6: Regional Significance Tests  
 
  
  
 
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001 
68 
 
   
 
Blacks are more likely than whites to live in the South, with 78.53 percent of the 
blacks living in that region versus 56.23 percent of the whites.  116.13 percent of the 
black respondents lived in the North versus 29.04 percent of the whites. The West has the 
greatest disparity between black and white residents with 5.34 percent and 14.73 percent 
respectively. The differences in the proportions for all regions are significant at an alpha 
level of .05.  
  Overall, there are significant differences in the regional residences of black 
veterans and non-veterans. Black veterans are significantly more likely to be from North 
and West than non-veterans and black non-veterans are significantly more likely to be 
from the South than black veterans. The proportions of black veterans are 20.50 percent, 
71.42 percent and 8.08 percent from the North, South and West regions respectively and 
black non-veterans were 14.23 percent, 81.62 percent and 4.15 percent from the North, 
South, and West respectively.  Whites also have significant differences in their overall 
regional residences.  Moreover, White veterans represent 28.87 percent, 55.44 percent, 
and 15.68 percent versus 29.18 percent, 56.87 percent, and 13.96 percent for white non-
veterans with residences in the North, South, and West respectively. These differences 
are significant at the .05 level. 
  
Marital Status 
 Almost seventy-four percent of the overall sample is married. Whites are 
significantly more likely than blacks to be married; however, blacks are significantly 
more likely to be either divorced or widowed than whites. There was no statistical 
69 
 
   
difference in the proportion of single whites and blacks in the overall sample population 
(see table 4.7).  
 Veterans are significantly more likely to be married than non veterans; however, 
they are significantly less likely to be widowed or single than their non-veteran peers. 
There is no significant difference in the proportion of divorced veterans and non-
veterans. In many respects the census data prior to 1960 was the only real data on 
marriage and divorce rates in the United States.   
The United States has lagged behind most nations of the 
Western World in the development of a nationwide 
organization for registering, collecting, and analyzing 
data about marriages, and divorces. In the 1958 issue of 
the Demographic Yearbook of the United Nations, the 
data on marriages from the United States represented 
coverage of less than 90 percent of all marriages 
occurring in that year (Ortmeyer 1962: 741) 
 
Table 4.7: Marital Significance Tests 
 
 
 
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
70 
 
   
The fact that non-veterans are more likely to be divorced is interesting in light of 
the research on the general conditions in which marriages tend to last and Segal?s (1986) 
argument that the military is a ?greedy institution.? The research has generally found that 
?couples in their first marriage, in higher socio-economic groups, who live in middle-
sized population areas and have children, show longer durations than the rest? (Monahan 
1962: 625).  Most of the veterans who served during World War II were single, because 
married men were often the last to be called up by local draft boards.  This fact would 
make it more likely that veterans were on their first marriage if they reported being 
married in the 1950 Decennial Census.  Moreover, very few military personnel were or 
are members of higher socioeconomic groups, although it is possible that some become a 
part of the higher socioeconomic groups after they separate from the military.  
 Segal (1986) contends that the military is a greedy institution because of the 
demands that it makes upon service members. We know that most of those who served in 
World War II were single, so the fact that veterans are married in similar or higher 
proportions than non-veterans would suggest that upon leaving the military veterans 
caught up to non veterans. It might even be suggested that they married as veterans 
because they were discouraged from marrying as active duty service members.  
  Black veterans are significantly more likely to be married than their black non-
veteran peers and a significantly larger proportion of black non-veterans were widowed 
or single than black veterans. White veterans were also significantly more likely to be 
married than their non-veteran peers and significantly less like to be widowed or single 
than white non-veterans (see table 4.7 above).  
 
71 
Education 
 I begin the description as well as the regression analysis of the dependent 
variables with education because past research has highlighted the effects of the G.I. Bill 
in increasing both income and education (see chapter 2). Furthermore, because education 
is used in the analysis of both income and SEI it is important to examine its effects 
independent of income and SEI in order to know its true contribution to these other 
variables.  
 The average person in the sample attained a tenth grade education level (see 
figure 4.10). Black respondents attained a little more than an eighth grade education and 
whites attained a tenth grade education.  
 
Figure 4.10: 1950 Census World War II Cohort Education Levels 
0
200
400
600
800
1,000
1,200
1,400
1
9
0
0
19
0
2
1
9
0
4
1
9
06
19
0
8
19
1
0
19
1
2
19
1
4
19
1
6
19
1
8
19
2
0
19
2
2
19
2
4
1
9
2
6
19
2
8
1
9
3
0
Birth Year
Tot
a
l
Grade School
Ninth to Eleventh 
High School Graduate
Some College
 
72 
 
   
When the data are disaggregated by veteran and racial category we find that veterans 
significantly attained almost a full grade of education more than non-veterans and that 
both black and white veterans achieved a grade more than their non-veteran peers with 
mean differences being significant at the .05 level (see table 4.8).    
 Figure 4.11 illustrates the percent of veterans and non-veterans by birth year who 
were high school graduates or had some college. Veterans attained significantly more 
education than non-veterans. Figure 4.12 also illustrates the veteran premium for serving 
during the peak mobilization period of World War II. 
Table 4.8: Education Significance Test 
 
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
20-24 5544 5.758117 1.945803 *** 6388 5.26941 2.034821
25-34 18694 5.891249 2.056599 *** 8382 5.064543 2.25173
35-44 7117 5.467472 2.23367 *** 16563 4.841756 2.22576
5-50 977 5.099284 2.351179
***
10698 4.431109 2.207556
Tests of 
Si
4
gnificance 
White
NMeanSD p NMeanSD
20-24 1167 4.098543 1.929652 *** 10765 5.648026 1.958112
25-34 2450 4.080816 1.982982 *** 24626 5.789978 2.107504
35-44 2291 3.479267 1.850028 *** 21389 5.195895 2.221648
45-50 1152 3.126736 1.726146
***
10523 4.63594 2.22585
Tests of 
Significance 
NMeanSD p NMeanSD
20-24 322 4.515528 1.975254 *** 845 3.939645 1.889081
25-34 1219 4.59639 1.996595 *** 1231 3.570268 1.832336
35-44 552 4.132246 2.045466 *** 1739 3.271995 1.733437
45-50 48 4 2.278484
**
1104 3.088768 1.689201
Tests of 
Significance 
NMeanSD p NMeanSD
20-24 5222 5.834738 1.917988 *** 5543 5.472127 1.979298
25-34 17475 5.981574 2.03019 *** 7151 5.321773 2.216991
35-44 6565 5.579741 2.21249 *** 14824 5.025904 2.204514
45-50 929 5.156082 2.342074
***
9594 4.585574 2.207902
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  
  
 
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
73 
 
   
Although one might expect veteran?s educational attainment to lag behind that of 
non-veterans, because of interruption to veteran?s education during service, there is a 
very real possibility that the data in this census reflects the selection bias discussed 
earlier. 
 
 
 
 
 
 
Figure 4.11: 1950 Census World War II Veteran versus  
Non-Veteran Educational Attainment 
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
19
00
1
90
2
1
90
4
1
90
6
1
90
8
1
91
0
1
9
12
1
91
4
1
91
6
1
91
8
19
20
19
22
1
92
4
1
92
6
1
92
8
19
30
Birth Year
P
e
rcen
t o
f
 Vet
e
ran
s
 in
 B
i
rt
h Year
Non-Veteran High School Graduate 
Non-Veteran Some College
Veteran High School Graduate 
Veteran Some College
 
 
 There is a chance that educational attainment had less to do with service than the 
preexisting conditions of those who did and did not serve. However, based on prior 
research, it is also clear that veterans used the G.I. Bill extensively and attained a great 
deal of education between 1945 and 1950 and the differences in education level as well 
as income and SEI could be a manifestation of this process.  
74 
 
   
 Blacks have a mean education attainment of 3.733 which corresponds to a mid 
elementary school level and is significantly less than whites who averaged a little more 
than a ninth grade education attainment. The difference between black and white 
educational attainments is statistically and substantively significant at the .001 level.  
Black veterans achieved significantly and substantially more education than black non-
veterans. The black veteran mean was 4.451 (ninth grade) versus black non-veterans who 
had a mean of 3.420 (mid elementary).  Black veterans attained almost twice as much 
education as black non-veterans. White veterans also attained more education than white 
non-veterans, and although the differences are not nearly as large as the differences 
between black veterans and non veterans they are statistically significant but not 
substantively different.  All of the mean comparison tests for level of education yielded 
significant differences between veterans and non-veterans (see table 4.8 above); however 
the only substantial differences that occur are between blacks and whites of all age 
categories and black veterans versus non veterans in the 25 -34 year old age group.   
 
Income  
 
 The average (median) income in the sample is $9.8487 log dollars ($18,934.25)
2
.  
Figure 4.9 illustrates the median log income by birth year. The figure illustrates the fact 
that younger birth cohorts had lower median incomes than older cohorts. The decline in 
earnings begins with either the 1921 or 1922 birth cohort, both of which were identified 
above as being a part of the middle World War II mobilization phase. However, one  
                                                 
2
 All income figures are expressed in log dollars. Log dollars were computed by taking 
the natural logarithm of income adjusted in year 2000 dollars. The inflation factor was 
computed by multiplying the 1950 dollar amount by a factor of 7.145.   
75 
 
   
would expect that younger cohorts would earn less than older cohorts because older 
cohorts have had more opportunity for education, work experience, and networking. 
Figure 4.12 also dramatically illustrates the premium that veterans of the peak 
mobilization phase received compared to non veterans and veterans of other mobilization 
periods. Veterans in the sample have a significant and substantially larger mean income 
of $9.69 (ln) than their non-veterans peers who have a mean income $9.62 (ln).  Blacks 
have a mean income of $9.15 (ln) that is significantly and substantially less than whites 
who have a mean income of $9.70 (ln).  Additionally, both black and white veterans have 
significantly higher incomes than their non-veteran peers; however, the difference in the 
mean is only substantial for black veterans versus non veterans. (see figures 4.13 and 
4.14). 
 
Figure 4.12: Log Income by Birth Year 
9
9.2
9.4
9.6
9.8
10
10.2
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
L
og Inco
me 
Veteran 
Non-Veteran
 
 
76 
 
   
 
Figure 4.13: Black Veteran Versus Non-Veteran Log Income 
8.5
8.7
8.9
9.1
9.3
9.5
9.7
9.9
10.1
19
00
19
01
19
02
19
03
19
04
19
05
19
06
19
07
19
08
19
09
19
10
19
11
19
12
19
13
19
14
19
15
19
16
19
17
19
18
19
19
19
20
19
21
19
22
19
23
19
24
19
25
19
26
19
27
19
28
19
29
19
30
Birth Year
Y
e
a
r
 
20
00
 Lo
g I
n
c
o
me
Veteran
Non-Veteran
 
Figure 4.14: White Veteran Versus Non-Veteran Log Income 
9
9.2
9.4
9.6
9.8
10
10.2
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
Year
 2000 Log I
n
come
Veteran 
Non-Veteran
 
 
77 
 
   
Based on this information alone, one might conclude that World War II veterans are 
significantly more likely to have higher incomes than non-veterans; however, when these 
groups are further disaggregated by age category some of the significant differences 
remain, but others are no longer significant (see table 4.9).  
Table 4.9: Income Significance Tests 
 
  
  
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
 
 
 
 
 
 
 
 
 
 
 
 
 World War II veterans aged twenty to thirty-four are significantly and 
substantially more likely to have a larger income than their non-veteran peers. When the 
veteran population is further disaggregated by race we find that veterans, both black and 
78 
 
   
white are significantly more likely to earn more than their non-veteran peers in every age 
category except the oldest (forty-five to fifty) (see table 4.9 and figures 4.13 and 4.14).  
However, the differences for black veterans are substantial in all of the age categories 
(difference greater than $700) except the 35 to 44 category). Furthermore, the only age 
group of white veterans with a substantial difference in income is the 20 ? 24 year old 
age group.   
    
Socio-Economic Indicator (SEI)   
 
 The Duncan SEI scale ranges from three to ninety-six. A three might typically 
include occupations such as ?Coal mine operatives and laborers; operatives, yarn, thread 
and fabric mills; porters; laborers, saw mills, planning mills and mill work? (Blau and 
Duncan 1967a:122-123). An SEI score of ninety-six would include occupations and 
professions such as architects, dentists, chemical engineers, lawyers and judges, 
physicians, and surgeons. The mean SEI for the sample is 31.71 with a median of twenty-
two (see figure 4.15). Figure 4.14 also illustrates the veteran premium for SEI 
particularly for veterans of the peak mobilization period. 
  
 
 
 
 
 
 
 
 
 
 
 
Figure 4.15: 1950 Census World War II Cohort Duncan SEI by Birth Year 
79 
 
   
15
20
25
30
35
40
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
D
u
ncan S
E
I
 S
c
or
e
Veteran
 Non-Veteran 
 
 
 An SEI of 31 on the 1950 Duncan scale would include occupations such as 
building managers and superintendents, self employed proprietors, boilermakers, 
machinists, sheriffs, and plumbers. A nineteen would include occupations such as 
blacksmiths, carpenters, painters, and laundry and dry cleaning operatives (Blau and 
Duncan 1967a).  One would expect that the mean SEI differences between veterans and 
non-veterans, blacks and whites, black veterans and non-veterans, and white veterans and 
non-veterans would be similar to the mean differences in income because one of the 
variables used to create SEI is income. In fact, the sample dramatically illustrates this. 
Veterans have a significantly and substantially higher mean and median SEI than non-
veterans, whites have a significantly and substantially higher mean SEI than blacks, black 
veterans have a significantly higher mean SEI than black non-veterans that are 
substantially higher in all but the 40- 55 age category, and white veterans have a 
80 
 
   
significantly and substantially higher SEI than white non-veterans in all age categories 
(see table 4.10).  
 
 
Table 4.10: Duncan SEI Significance Tests 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
 
 
 
 
 
 
81 
 
   
Models and Multivariate Regression Analysis 
 In this section I make use of the five multivariate regression models described in 
Chapter 3 to control for factors associated with earnings, education, and Duncan SEI 
outcomes to determine the net premium or penalty to veterans and non-veterans.  
 
 
Education Models and Regression Analysis 
 
Model 1  
 As the bivariate descriptions above suggested, veterans tend to achieve higher 
education levels than their non-veteran peers, whites attained higher levels of education 
than blacks, and mean education levels increase up to a certain age and then begin to 
decline (see table 4.8 above).   Model 1 generally confirms the descriptions above (see 
table 4.11). When education is regressed on World War II veteran status, age
2
, and race 
we see that the model explains eight percent of the variance.  We also find that for 
veterans the predicted educational attainment is higher than for non-veterans (B=.1270). 
In fact, at no other point in this analysis of 1950 World War II veteran educational 
attainment does the veteran advantage reach this level again.  The coefficient for age is -
.1280 indicating older respondents had achieved less education than younger 
respondents. Blacks also paid a penalty (B= -.1840).  
 
 
 
 
 
 
 
 
82 
 
   
Model 2  
  
 The veteran relative to non-veteran advantage in this model remains large and 
significant, while the black race penalty decreases slightly but is still large and 
significant. Both married and single respondents have significant educational attainment 
premiums relative to unmarried and non-single respondents respectively net of the effects 
of the other variables. Those residing in the both the North and the South paid an 
educational attainment penalty relative to those not living in those regions respectively. 
The penalty was much higher for those living in the South than those in the North.  The 
addition of marital status and region of residence background variables in model 2 
modestly increases the proportion of variance in education that can be predicted from the 
independent variables from eight to nine percent.
83 
 
   
 
 
T
a
bl
e 4.11: 1950 E
ducati
o
n Model
s
 and Regressi
ons 
84 
 
   
Model 3  
 
 The addition of the mobilization phase variables in model 3 increased the adjusted 
R-square from .0924 to .0931. The highest premiums for educational attainment in this 
model come from World War II veteran status and marital status and the largest penalties 
come from race and age
2
.   
 The mobilization phase additions also serve to decrease the veteran educational 
premium and increase the educational penalty for being black. The other variables were 
for the most part unaffected.  The coefficient for mobilization phase 2 (peak period) is -
.00010, which tells us that those who were a part of birth cohorts in the peak mobilization 
paid a penalty in educational attainment compared to those who were in birth cohorts of 
other phases. Those who were a part of mobilization phase 3 also paid a small penalty 
(coefficient = -.0341) albeit greater than the phase 2 penalty.    
 
Model 4 
 
 Although model 4 is significant overall in that the independent variables reliably 
predict the dependent variable (p<.001), there is not much change in the amount of 
variance explained in model 3 versus model 4.  However, this model does show that there 
is premium in educational attainment for being a part of birth cohorts with large 
proportions of veterans (Beta=.0207).  This finding would seem to be counterintuitive to 
in light of the regressions in model 3 which show that those in birth cohorts associated 
with mobilization phase 2 received educational attainment premiums relative to those in 
other mobilization phases.  
 
85 
 
   
 
Model 5 
 
 Model 5 adds two veteran mobilization interaction terms to the regression 
equation. This addition slightly increases our ability to predict the variance in educational 
attainment. In model 4 we could predict 9.26 percent of the variance and in model 5 we 
can only predict 9.36 percent of the variance.  In model 5 the veteran educational 
premium decreases slightly and is significant. This model clearly shows the educational 
advantage for being a veteran associated with mobilization phase 2 (B = .0167) and the 
educational disadvantage for being a veteran associated with mobilization phase 3 (B = -
.028). 
 
Income Models and Regression Analysis 
 
Model 1 
 
Table 4.12 illustrates the results of all five of the multivariate models for income. 
In model 1 income is regressed on World War II veteran status, age
2
, and race and we see 
that this model explains about seven percent of the variance in earnings income. This 
model reliably predicts the dependent variable: income (F = 1991.02 and p ? .001). All 
three of the independent variables are significant. Age
2
 is the strongest predictor of 
income in this model followed by race, and then veteran status. The coefficients are 
.2057, -.1870, and .1038 for age
2
, race and veteran status respectively.  
For veterans the predicted income is higher than for non-veterans, blacks pay a 
large penalty relative to whites and older respondents are predicted to have more income 
86 
 
   
than younger ones. The results of this model are generally consistent with the hypothesis 
that all things being equal veterans will earn more than their non-veteran peers. 
Moreover, it comes as no surprise that blacks in 1950 paid an earnings penalty because of 
the color of their skin. 
87 
 
   
 
 
 
 
 
 
T
a
bl
e 4.12: 1950 Income Model
s
 and Regressi
ons 
 
88 
 
   
Model 2 
  When the marital status, region, and education variables are added to model 2 the 
overall model is still statistically significant with an F value of 1392.17 and p  .001. 
Most of the additional power in the model comes from the addition of the education 
variables. As suggested by the education regression models (above) a large part of the 
veteran effect operates through education. As expected those with some college 
(B=.1366) and high school graduates (B=.1351) earn substantial premiums for their 
education  The addition of the marital, region, and education variables decreased the 
magnitude of the World War II veteran status variable from .1038 in model 1 to .0620 in 
model 2. Although the direction of the race variable was unaffected, in that blacks still 
earned less, the magnitude decreased from -.187 in model 1 to -.145 in model 2. Being 
married increases earnings (B= .1377) and being single decreases earnings (B=-.073).  
Residing in the South tends to substantially decrease earnings income while living in the 
North tends to increase income. However, this could be due to a number of factors 
including a lower cost of living in the South versus the North, as well as the agrarian 
versus industrial economies in the South and North regions respectively in the 1950s.  
Approximately fourteen percent of the variance in earnings income is explained by this 
model. 
?
 
Model 3   
The addition of the mobilization phase variables and an interaction term for being 
a black veteran model 3 had relatively little influence on the control variables used in the 
previous models, but it did have an effect on the World War II veteran status and age
2
 
variables.  
89 
 
   
The premium for World War II veterans decreased from .0620 to .0350. The 
introduction of the race*veteran interaction term captures some of the variance in the 
veteran status variable. The introduction of this interaction term is also responsible for 
increasing the penalty that blacks pay in this model.  
The premium for age
2
 decreased in this model from .1565 to .0763 because some 
of the variance is captured in mobilization phases 2 and 3, which contain only the 
younger respondents in the sample.  
Those who were in cohorts that were a part of the peak and demobilization phases 
of World War II paid penalties. The coefficients are -.035and -.123 for the peak and 
demobilization phases respectively. This should not be confused with veterans who were 
a part of these periods and the associated veteran*mobilization phase interaction terms 
(see Model 5 of Table 4.12). 
 As the bridging hypothesis and human capital models suggest black veterans 
earned premiums for the service (Beta=.0275). These premiums are maintained at the 
same magnitudes throughout the rest of the income regressions in this census period. The 
adjusted R-squared value increased from 14.41 percent to 15.11 percent. 
  
Model 4  
The only control variable affected by the removal of the mobilization phase 
variables and the addition of the percent of a birth cohort that served variables was age
2
. 
The coefficient for age
2
 increased from .0763 to .2303.  
However, there is a big premium for being a part of a cohort with a large 
proportion of veterans. This is interesting in light of the fact that in model three 
mobilization phase two had a large proportion of veterans and yet those in cohorts 
90 
 
   
associated with that phase paid an income penalty. This finding supports the notion 
developed above that there is an earnings premium for veterans born to cohorts that 
served during the peak mobilization phase. During that phase of the mobilization the ratio 
of those who served to those who did not serve was much higher than the other two 
phases (see table 4.5 above). The removal of the mobilization phase variables and the 
addition of the percent of a birth cohort that served decrease the explanation of the 
variance from 15.11 in model 3 to 14.79 percent in model four. 
 
Model 5  
The removal of the percent of birth year variable and the addition of the veteran 
and mobilization phase interaction terms in model 5 shows that veterans who were a part 
of the peak mobilization period earned a small premium (coefficient = .0149) for their 
service. Additionally, veterans who served during phase 3 (demobilization) paid a 
penalty for their service (coefficient = .0-.041).  The adjustment from model 4 to model 5 
also increased the World War II veteran premium; however, this combination does not 
appreciably change any of the other control variable coefficients. The coefficient for the 
race*veteran interaction term decreased from .0297 to .0278: however the black veteran 
premium remains.  The removal of the percent of birth year variable and the addition of 
the veteran and mobilization phase interaction terms in model 5 decreases the explained 
variance from 14.79 percent to 14.66 percent. 
 
 
 
91 
 
   
Duncan SEI Models and Regression Analysis 
 
 The descriptive section above showed very clearly that veterans generally have 
higher SEIs than non-veterans, whites generally have higher SEIs than blacks, and that 
both black and white veterans generally have higher SEIs than their non-veteran peers.  
As in the previous sections, I use OLS regression to simultaneously control for several 
factors that might be associated with SEI to determine if veterans groups receive an 
advantage or premium for their service. The results of the models are shown in table 
4.13.  
92 
 
   
 
 
 
T
a
bl
e  4.13: SE
I Model
 and OL
S Regressi
on Anal
ysi
s
 
93 
 
   
Model 1 
 In model 1, I regressed SEI on World War II veteran status, age
2
, and race. This 
model explains almost seven percent of the variance in SEI.  More specifically, we find 
that World War II veterans receive an SEI premium for their service (Beta =.0841).      
Additionally, older respondents also received SEI premiums (Beta =.1069). The strongest 
predictor of SEI is race (Beta= -.2190). The coefficient for race is almost twice as large 
as the coefficient for age
2
 or veteran status.  
 
Model 2 
 The addition of marital, regional, and educational background variables to the 
regression equation increases the proportion of the variance that can be explained from 
six percent to thirty percent.  Veterans continue to earn SEI premiums even though the 
coefficient decreases from .0841 to .0245. There is little difference in the magnitude of 
the age
2
 coefficient from the previous model; however, it is still positive and significant.  
 The penalty for being black decreases when one controls for marital status, 
region, and education; however, the coefficient remains large and negative. Married 
respondents receive an SEI premium while single respondents receive a penalty net of the 
effects of the other variables. Residing in the North provides an increase on the SEI scale 
versus not being from the North as does residing in the South versus not being from the 
South.  These numbers are not so unexpected for the North, but I would have expected 
the coefficient for the South to be much lower or even negative in light of the negative 
coefficients in the income and education regressions for that region. 
   Clearly, the largest coefficients come from the education independent variables. 
This is not so surprising in light of the fact that the Duncan SEI was formulated in large 
94 
 
   
part based on the education required for jobs and professions. Those jobs that required 
little education received relatively low numbers and those that required many years of 
education received higher scores. Therefore, it is not surprising that those with some 
college (Beta =.4966) and those with a high school degree (Beta =.2341) earn very large 
SEI premiums. 
 
Model 3  
 In this model the addition of the race*veteran interaction term and the 
mobilization phase variables reduces the veteran premium despite the fact that the 
interaction term was not significant. The coefficient for the veteran premium decreases 
from .0245 to .0158; however, it remains significant. The age
2
 variable decreased from 
.1144 to .0871 and also remains significant.  
 The demobilization phase variable produced a negative coefficient indicating that 
respondents with birth cohorts associated with that phase paid an SEI penalty. In this 
model the addition of the race*veteran interaction term and the mobilization phase 
variables modestly increases the adjusted R-squared from 30.00 to 30.09 percent. 
  
Model 4  
 In this model, I once again remove the mobilization phase variables and add in a 
variable that controls for the percent of veterans in a particular birth cohort. Making these 
changes in the model did not affect most of the coefficients of the other variable or the 
adjusted R-squared.  
95 
 
   
 The veteran premium for SEI decreased slightly but the premium for married 
respondents increased slightly.  Furthermore, the coefficient for the percent of veterans in 
a birth cohort is .0388. 
 
Model 5  
 In the final SEI model I removed the percent of veterans in a birth year variable and 
added in the veteran*mobilization phase interaction terms. Making these changes yields 
only slight changes in the adjusted R-squared. More specifically, about thirty percent of 
the variance in SEI can be predicted from the variables in the model, which is the same as 
the variance that can be predicted from background factors alone. The coefficient for 
World War II veteran status increased slightly; however, the coefficients for age
2
, race, 
married, North and South residence, high school graduate and some college remained 
virtually the same in magnitude, direction, and significance.  
 Interestingly, although there was an SEI premium for being a part of the peak 
mobilization phase, there is a penalty for being a veteran in the demobilization phase. 
The coefficient for the peak mobilization phase is .0180, which suggests that veterans for 
the peak World War II mobilization period received a statistically significant premium of 
half an SEI point for their service.   Veterans who were a part of the demobilization phase 
suffered an SEI penalty; however their penalty was substantially less than the penalty 
suffered by entire phase 3 mobilization group in model 3.  
 
Summary 
  In the 1950 census there was clearly a premium for veteran status with regard to 
education, income, and Duncan SEI score. Therefore we could conclude that hypothesis 
96 
 
   
one is supported for the 1950 census period.  Furthermore, this premium was significant 
for both black and white veterans versus their non-veteran peers with regard to all three 
dependent variables. Moreover, these premiums remain large and significant even when 
we control for background factors including age, race, marital status, regional residency 
and education. Additionally, the differences are for black veterans than white veterans 
versus their non-veteran peers. As such we could reasonably conclude that hypothesis 
two is valid for the 1950 census period. 
 Although there were penalties in the aggregate for being a part of both the World 
War II peak mobilization and demobilization periods, veterans of the peak mobilization 
phase received a premium, albeit a small one, for their service. This finding supports the 
notion in hypothesis three that those serving during the peak mobilization periods will 
achieve more social status attainment than those serving in the beginning mobilization 
phase cohorts of World War II.  However, it does not support the notion that those of the 
World War II demobilization phase will achieve more social status than those of the peak 
phase.  
 Finally, for all three of the dependent variables it is obvious that cohorts with 
large proportions of veterans earned a premium. This supports the notion espoused in 
hypothesis four that one would expect that veterans born to cohorts with larger 
proportions of veterans should achieve more social status attainment than those with a 
lower proportion of veterans.
97 
 
   
Chapter 5: World War II Veterans and the 1960 Census
 
 
 The 1950s through the 1960?s was period of great change in the United States. At 
the end of World War II millions of service-members returned home. American industry 
quickly shifted from war time production to peace time production to meet peacetime 
needs. The availability of goods and the huge amount of savings that Americans had 
accumulated during the war created corporate expansion and jobs.   
 As service-members returned home and became reacquainted with their families 
they created the ?baby boom?. Although veterans were busy making babies, they were 
also attending colleges and universities in record numbers. As result, the veteran 
population raised its education level and increased its prospects for social status 
attainment. Moreover, many of the World War II veterans were in their prime working 
years (thirty ? forty five years old) during the 1960 census. Veterans also received 
employment hiring preferences in Civil Service employment and this may have 
contributed to them receiving jobs that their non-veteran peers were not eligible for.  
  The country was still segregated and many blacks and particularly black veterans 
were disenchanted with life in America. They had fought for freedom that they did not 
have at home. Until 1954, an official policy of separate but equal existed in the United 
States. In that year the Supreme Court wrote in Brown v. the Board of Education of 
Topeka, Kansas that separate facilities for blacks were unconstitutional and the slow 
process of integration began across the nation.   
 
 
 
 
 
98 
 
   
Descriptive data 
I aged the 1950 sample ten years by using 1960 census data. In order to age the 
World War II cohort by ten years I excluded all of the respondents whose birth years 
were not between 1900 and 1930 and those who had zero or not applicable income and 
SEI scores. After making these adjustments I was left with a total sample of 247, 253 (see 
table 5.1). 
Of the 247,253 total respondents 109, 213 are veterans, who represent 
approximately forty-four percent of the sample. Figure 5.1 illustrates the 1960 census 
sample ratio of veterans to non-veterans after making the same adjustments that were 
made for the 1950 data. The small number of veterans in the older birth cohorts explains 
much of the age difference between veterans and non-veterans. Black veterans make up 
approximately seven percent (7,386) of the veteran sample population. Interestingly, in 
the 1950 census sample there were only 32,712 veterans. The number of veterans 
increased nearly threefold between the 1950 and 1960 Censuses, lending credence to the 
underenumeration theory presented in Chapter 4. Moreover, as one would expect, the 
ratio of black veterans to the total sample population is consistent across the samples: 
6.66 percent in 1950 and 6.76 percent in 1960.  
 
Background Descriptive Statistics and Discussion 
Age   
 The average age in the sample is 43.29 years with a standard deviation of 8.58 
and a median age of 43.  In 1950 the average sample respondent was 33.35 years old with 
a 9.12 standard deviation and median age of 33.  On average the veterans in the 1960 
census sample are 39 years old (median). They are significantly younger than their non-
99 
 
   
veteran peers by 5.8 years.  Although this difference is significant it is also expected.  If 
one assumes that those who fought  
Table 5.1: 1960 Census Descriptive Summary 
Ca
t
e
g
o
r
y
Su
b-
C
a
teg
o
r
y
T
ot
al
T
o
ta
l
B
la
ck W
h
it
e
T
ot
al
B
T
o
ta
l Sam
p
le
 Siz
e
247,253
109
,213
7,386
101
,827
138,040
Ag
e
M
edi
a
n
 Age, y
e
a
r
s
 
43
39
39
39
47
#
 age 30
-34
1
9
.
34%
20.59
%
1
9.42%
20.67
%
1
8.35%
#
 age 35
-44
3
7
.
41%
57.24
%
5
6.72%
57.28
%
2
1.72%
#
 age 45
-54
2
9
.
93%
19.50
%
2
1.31%
19.37
%
3
8.18%
#
 age 55
-60
1
3
.
33%
2.67%
2
.
56%
2.68%
2
1
.75%
Age
 Ran
ge (
Y
oun
ges
t -
 O
l
de
st
)
30 -
 60
30
 - 60
30 -
 60
30
 - 60
30 -
 60
30
R
ace Blac
k
8
.90%
N/A
1
0
0
.00%
N/A
N
/A
W
h
ite
9
1
.
10%
N/A
100.00
%
N
/A
R
e
gi
on
N
o
r
t
h
2
7
.
54%
28.43
%
2
1.27%
28.95
%
2
6.83%
Sou
t
h
5
6
.
68%
54.95
%
7
0.43%
53.83
%
5
8.05%
W
e
s
t
15
.78%
16.62
%
8
.30%
17.22
%
1
5.11%
Ma
r
i
t
a
l
 S
t
a
t
u
s
 
M
a
rr
ied
8
8
.
06%
88.60
%
8
0.42%
89.19
%
8
7.63%
D
i
vor
c
ed 3
.
98%
3.93%
1
0
.36%
3.46%
4
.
02%
W
i
d
o
w
ed 1
.
20%
0.67%
1
.
49%
0.61%
1
.
62%
Single Ne
v
e
r
 Mar
r
i
ed
6
.
76%
6.81%
7
.
73%
6.74%
6
.
73%
In
co
m
e
 
M
edi
a
n
 Ear
n
i
n
g
s
 Inc
o
m
e
 
$5,050
.00
$
5
,
450.00
$3,550
.00
$
5
,
550.00
$4,85
0.00
$
2
V
e
te
r
a
ns
 N
on-
W
o
rld
 W
a
r II
 1960 C
e
n
s
us
 
l
ack 
W
h
ite
14,
616
123,424
45
4
7
20.16
%
1
8.13%
27.96
%
2
0.98%
34.00
%
3
8.68%
17.88
%
2
2.21%
 - 6
0
30 -
 60
100.00
%
N
/A
N/A
1
0
0
.00%
17.17
%
2
7.98%
76.86
%
5
5.82%
5.97
%
1
6.20%
79.08
%
8
8.64%
9.76
%
3
.34%
3.25
%
1
.43%
7.91
%
6
.59%
,650.00
$5,05
0.00
Adj
u
s
t
e
d
 to 20
00 Do
l
l
ars
$29,38
0.90
$3
1,708.10
$20,65
3.90
$3
2,289.90
$28,21
7.30
$1
5,417.7
0
$29,38
0.90
L
n
 of
 In
c
o
m
e
10.2881
10.3
643
9.9357
10.3
825
10.2477
9.6
433
10.2881
Ed
uc
atio
n
N
one-
 G
r
ade 8
3
4
.
86%
23.89
%
4
5.29%
22.33
%
4
3.54%
70.18
%
4
0.38%
G
r
ade 9-1
1
21
.33%
23.26
%
2
5.68%
23.08
%
1
9.81%
15.19
%
2
0.35%
H
i
g
h
 Sc
hool G
r
a
duate
2
4
.
01%
28.45
%
1
7.10%
29.28
%
2
0.50%
9.07
%
2
1.86%
Som
e
 
Coll
eg
e or
 Coll
e
ge G
r
ad
19
.80%
24.40
%
1
1.93%
25.31
%
1
6.15%
5.57
%
1
7.41%
SEI (
m
edia
n)
27
37
16
42
24
15
2
7
V
e
te
r
a
ns
 
Figure 5.1: 1960 Census Veterans versus Non-Veteran 
100 
 
   
0
2000
4000
6000
8000
10000
12000
1
9
0
0
1
9
02
1
9
0
4
1906 1
9
0
8
19
1
0
1
9
1
2
1
9
1
4
1
9
1
6
1
9
1
8
1920 1
9
2
2
1924 1
9
2
6
19
2
8
1
9
3
0
Birth Year
Num
b
e
r
Non-Veteran 
Veteran
 
in World War II were between the ages of 18 and 26 then they would have been between 
the ages of 31 and 44 in the 1960 census (see Table 5.1).  Additionally figure 5.1 
illustrates the fact that there are far more non-veterans than veterans in the older cohorts 
who drive the average veteran age down and the average non-veteran age up.  
 The majority of the veterans in the 1960 sample fall into the 35 to 44 year old age 
category and the largest category of non-veterans is the 45 to 54 year old age category 
(see table 5.1 and table 5.2). As in the 1950 Census, one would expect that non-veterans 
would be both younger and older than veterans because they had no age restrictions. 
101 
 
   
Table 5.2: Veteran Expected Ages in 1960 by Entry Age 
1960
Year of Entrance 18 19 20 21 22 23 24 25 26
1942 36 37 38 39 40 41 42 43 44
1943 35 36 37 38 39 40 41 42 43
1944 34 35 36 37 38 39 40 41 42
1945 33 34 35 36 37 38 39 40 41
1946 32 33 34 35 36 37 38 39 40
1947 31 32 33 34 35 36 37 38 39
Age of Entrance 
 
  
Table 5.3: Age Significance Tests 
 
Tests of 
Significance 
N Mean SD p NMeanSD
109213 40.05628 6.2849 *** 138040 45.8447 9.276074
Tests of 
Significance 
White
N Mean SD p NMeanSD
22002 43.09204 8.578152 *** 225251 43.30706 8.588414
Tests of 
Significance 
N Mean SD p NMeanSD
7386 40.19523 6.351245 *** 14616 44.5559 9.165444
Tests of 
Significance 
N Mean SD p NMeanSD
101827 40.04621 6.280002 *** 123424 45.99732 9.277275
White Veteran White Non-Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
102 
 
   
White respondents have an average age of 43.31 and they are significantly older 
than black respondents (43.10). However, the magnitude of the difference in mean age 
between Black and White respondents is very small and the significance is a reflection of 
sample size. Black veterans have a mean age of 40.19 years and they are significantly 
younger than their non-veteran black peers by 4.4 years.   Similarly white World War II 
veterans have a mean age of 40.05 years and they are significantly younger than their non 
veteran peers who have a mean age of 45.99 years. The median ages for black veterans, 
black non-veterans, white veterans, and white non-veterans are 39, 45, 39 and 47 
respectively. Table 5.3 provides the results of the significance tests described above. 
As in the previous chapter I disaggregate the data by birth cohorts to examine 
how groups of different ages are affected in terms of attainment by military service as 
well as phase of military mobilization. The logic for using birth cohorts as well the 
beginning and ending years for the establishment of the mobilization phase years remains 
the same as the previous chapter: if military service truly affects social status attainment, 
then birth cohorts that served in lower proportions should receive less social status 
attainment than those with higher proportions of service. Some birth cohorts might have 
lower proportions serving because they were at the beginning or end of the war?s 
mobilization. In the 1960 sample, the beginning mobilization phase (phase1) begins with 
the 1906 birth year cohort and ends with the 1915 birth year cohort; the middle or peak 
mobilization phase (phase 2) beings with the 1916 birth year cohort and ends with the 
1926 birth year cohort; and the demobilization phase (phase 3) begins with the 1927 birth 
year cohort and ends with the 1930 birth year cohort (see figure 5.2).  
  
 
103 
 
   
Figure 5.2: 1960 Census World War II Cohort Mobilization Phases 
 
  
Race  
 As one might expect the majority of the respondents in this sample are white. 
91.10 percent (225,251) of the sample are white and the remainder is black. Blacks are 
approximately 6.76 percent (7,386) of the veteran group and 10.59 percent (14,616) of 
the non-veteran group, while white veterans comprise 93.24 percent (101,827) of the 
veterans and 89.41 percent (123,424) of the non-veteran groups.  According to the United 
States Census Bureau (2006) the total population in the United States was 179,323,175 
and 88,331,494 of that was male. Furthermore, they estimated that the number of white 
males was 78,367,149 (88.72 percent) and the number of black males was 9, 113,408 
(10.32 percent) (United States Census Bureau 1964).  As one would expect in a random 
104 
 
   
sample the proportion of blacks and whites in the sample population are representative of 
the United States? demographics in 1960.   
 
Region  
 Just over half of the sample, (56.68 percent), maintains their residence in the 
South, 27.54 percent in the North, and 15.78 percent in the West. In the 1950 58.34 
percent lived in the South, 27.82 percent in North, and 13.84 percent in the West. The 
differences in proportions would lead us to believe that there was a general westward 
shift in the population over the decade.   
 The veteran and non?veteran proportions for region of residency are very similar. 
28.43 percent of the veterans reside in the North versus 26.83 percent of the non-
veterans; 54.95 percent of the veterans and 58.05 percent of the non-veterans reside in the 
South; and 16.62 percent of the veterans versus 15.11 percent of the non-veterans live in 
the West.  I conducted a significance test assuming unequal variance to determine if the 
population proportions for veterans and non-veterans were different in terms of region.  
In this test I combined all of the regions and tested whether there was a significant 
difference between veterans and non-veterans. The results show that the population 
proportions are significantly different, with the proportion of veterans from the South and 
West being higher than non-veterans in this sample. These results were significant at the 
.05 level.   
 I also conducted a significance test with the disaggregated regional data to 
compare veterans and non-veterans by specific region and found that all of regions have 
significant differences for veterans versus non-veterans (see table 5.4).  
Table 5.4: Region Significance Tests 
 
105 
 
   
Tests of 
Significance 
N Proportion p N Proportion
North 31,047 0.2843 *** 37,043 0.2683
South 60,015 0.5495 *** 80,135 0.5805
West 18,151 0.1662
***
20,862 0.1511
Tests of 
Significance 
N Proportion p N Proportion
North 4081 0.1855 *** 64,009 0.2842
South 16436 0.7470 *** 123,714 0.5492
West 1485 0.0675
***
37,528 0.1666
Tests of 
Significance 
N Proportion p N Proportion
North 1,571 0.2127 *** 2,510 0.1717
South 5,202 0.7043 *** 11,234 0.7686
West 613 0.0830
***
872 0.0597
Tests of 
Significance 
N Proportion p N Proportion
North 29,476 0.2895 NS 34,533 0.2798
South 54,813 0.5383 *** 68,901 0.5582
West 17,538 0.1722
***
19,990 0.1620
Black Veteran Black Non-Veteran
White Veteran White Non-Veteran
Veteran Non-Veteran
Black White
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
 Blacks are more likely than whites to live in the South, with 74.70 percent of the 
blacks living in that region versus 54.92 percent of the whites.  About 18.5 percent of the 
black respondents lived in the North versus 28.42 percent of the whites. The West has the 
greatest disparity between black and white residents with 6.75 percent and 16.77 percent 
respectively. The differences in the proportions for all regions are significant at an alpha 
level of .001.  
  Overall, there are significant differences in the regional residences of black 
veterans and non-veterans. Black veterans are significantly more likely to reside in the 
North and West than non-veterans and non-veterans are significantly more likely to be 
from the South than veterans. The proportions of black veterans are 21.27 percent, 70.43 
106 
 
   
percent and 8.30 percent from the North, South and West regions respectively and black 
non-veterans were 17.17 percent, 76.86 percent and 5.97 percent from the North, South, 
and West respectively.  Whites also have significant differences in their overall regional 
residences.  Moreover, white veterans represent 28.95 percent, 53.83 percent, and 17.22 
percent versus 27.98 percent, 55.82 percent, and 16.20 percent for white non-veterans 
with residences in the North, South, and West respectively. These differences are 
significant at the .001 level for the South and West regions (see table 5.4 above). 
  
Marital Status 
 More than eighty-eight percent of the overall sample is married versus 73.84 in 
the 1950 census. Whites are significantly more likely than blacks to be married or single; 
however, blacks are significantly more likely to be either divorced or widowed than 
whites. Although there was a small difference in the proportions in the 1950 census the 
relationships remain the same. Table 5.5 illustrates the significance tests described above.  
 Veterans are significantly more likely to be married than non-veterans; however, 
non-veterans are significantly more likely to be widowed than their veteran peers. The 
proportion of married veterans increased from 76.31 percent in 1950 to 88.60 in 1960, 
while the proportions for non-veterans increased from 72.08 percent to 87.63 percent.  
 
107 
Table 5.5: Marital Status Significance Tests 
 
Tests of 
Significance 
N Proportion p N Proportion
Maried 96,761 0.8860
***
120,958 0.8763
Divorced 4,287 0.0393 NS 5,553 0.0401
Widowed 731 0.0067 *** 2,241 0.0106
Single Never Married 7,434 0.0681
NS
9,288 0.0673
Tests of 
Significance 
N Proportion p N Proportion
Maried 17498 0.7953 *** 200,221 0.8889
Divorced 2192 0.0996 *** 7,648 0.034
Widowed 585 0.0266 *** 2,387 0.0106
Single Never Married 1727 0.0666
***
14,995 0.0785
Tests of 
Significance 
N Proportion p N Proportion
Maried 5,940 0.8042 * 1 0.7908
Divorced 765 0.1036 NS 1,427 0.0976
Widowed 110 0.0149 *** 475 0.0325
Single Never Married 571 0.0773
NS
1,156 0.0791
Tests of 
Significance 
N Proportion p N Proportion
Maried 90,821 0.8919 *** 109,400 0.8864
Divorced 3,522 0.0346 NS 4,126 0.0334
Widowed 621 0.0061 *** 1,766 0.0143
Single Never Married 6,863 0.0674
NS
8,132 0.0659
Black Veteran Black Non-Veteran
White Veteran White Non-Veteran
Veteran Non-Veteran
Black White
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
  
 Black veterans are significantly more likely to be married than their black non-
veteran peers and a significantly larger proportion of black non-veterans are widowed 
than black veterans. One explanation for this is that black non-veterans are older than 
black veterans. The proportion of married black veterans increased from 69.89 percent in 
1950 to 80.42 percent in 1960 which was on par with increase for their non-veteran 
peers. White veterans were also significantly more likely to be married than their non-
veteran peers and significantly less like to be widowed than white non-veterans (see table 
108 
 
   
5.5 above). The proportion of married white veterans increased from 76.76 percent in 
1950 to 89.19 percent in 1960. The increase in the proportion of married white veterans 
is similar to the increase in the proportion of married white non-veterans.   
 
Education 
 The average person in the sample attained a tenth grade education level. Black 
respondents attained a little more than a ninth grade education and whites attained almost 
an eleventh grade education. Figure 5.3 displays 1960 census World War II cohort 
education levels by birth cohort. This graph shows that older cohorts generally had more 
people with grade school educations than younger cohorts and the steady increase in the 
number of college graduates from cohort to cohort. 
109 
Figure 5.3: 1960 Cohort Education Levels 
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
1
900
1
902
1
904
1
906
1
9
0
8
1
9
1
0
1
9
1
2
1
9
1
4
1
9
1
6
1
9
1
8
1
9
2
0
1
9
2
2
1
924
19
2
6
19
2
8
19
3
0
Birth Year
Nu
mb
er
Grade School
Ninth to Eleventh 
High School Graduate
 Some College
 
 When the data are disaggregated by veteran and racial category we find that 
veterans significantly attained almost a full grade of education more than non-veterans 
and that both black and white veterans achieved a grade more than their non-veteran 
peers with mean differences being significant at the .05 level for both black and white 
veterans versus their non-veteran peers (see table 5.6 and 5.7).   The net advantage for 
black veterans versus black non-veterans (1.129) is greater than the net advantage for 
white veterans versus white non-veterans.  So, although whites have higher educational 
attainment the magnitude of the difference for black veterans is greater than that of white 
veterans. Furthermore, the difference between black veterans and non veterans is 
substantively significant with the primary differences deriving from the 25 ? 35 year old 
age group that has a difference of means equal to 1.128 (see table 5.7). 
110 
 
   
Table 5.6: 1960 Education Significance Tests 
 
 
Tests of 
Significance 
NMeanSD p N Mean SD
Aggregate 109,213 5.927738 2.154631 *** 138,040 4.992111 2.294166
Tests of 
Significance 
White
NMeanSD p N Mean SD
Aggregate 22002 3.983183 2.094966 *** 225,251 5.544299 2.251158
Tests of 
Significance 
NMeanSD p N Mean SD
Aggregate 7,386 4.733144 2.135588 *** 14,616 3.604201 1.968366
Tests of 
Significance 
NMeanSD p N Mean SD
Aggregate 101,827 6.014387 2.130114
***
123,424 5.156469 2.274325
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001 
 
111 
 
   
 
 Table 5.7: Education Significance Test by Age Category 
 
 
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
30-34 22,482 6.221688 2.074275 *** 25,326 5.777778 2.27986
35-44 62,512 6.016285 2.104657 *** 29,977 5.192448 2.296122
45-54 21,299 5.467064 2.253675 *** 52,709 4.842645 2.236369
55-60 2,920 5.12911 2.370043
***
30,028 4.391834 2.193018
Tests of 
Significance 
White
Age Group N Mean SD p N Mean SD
30-34 4380 4.669863 2.140456 *** 43428 6.119324 2.158254
35-44 8276 4.242992 2.127204 *** 84213 5.897296 2.153727
45-54 6543 3.554639 1.940266 *** 67465 5.164693 2.237026
55-60 2803 3.143418 1.754198
***
30145 4.579333 2.218362
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
30-34 1434 5.199442 2.046044 *** 2946 4.412084 2.138318
35-44 4189 4.800191 2.132533 *** 4087 3.671886 1.964255
45-54 1574 4.228717 2.105251 *** 4969 3.341115 1.834256
55-60 189 3.910053 2.069884
***
2614 3.087988 1.716376
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
30-34 21048 6.291334 2.057836 *** 22,380 5.957551 2.236614
35-44 58323 6.10363 2.075409 *** 25,890 5.432484 2.252311
45-54 19725 5.565881 2.235792 *** 47,740 4.998932 2.216459
55-60 2731 5.213475 2.366629
***
27,414 4.51616 2.19307
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
  
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
Figure 5.4 illustrates the educational attainment of veterans and non-veterans by 
birth year. As described above, veterans attained significantly more education than non-
veterans. Additionally, this figure dramatically illustrates the veteran premium for 
serving during the peak mobilization period of World War II.  Figures 5.5 and 5.6 display 
the net education advantage for veterans, both black and white, as well as the peak 
mobilization phase advantage. These figures also display the net education disadvantage 
for veterans in the demobilization phase. 
 
 
112 
 
   
 
Figure 5.4: Veteran versus Non-Veteran Education  
(College and High School Graduates Only) 
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
1
90
0
1
90
2
1
90
4
1
90
6
1
90
8
1
91
0
1
91
2
1
91
4
1
91
6
1
9
18
1
92
0
1
92
2
1
9
24
19
26
19
28
1
9
30
Bith Year
P
e
r
c
e
n
t
 of
 V
e
t
e
r
a
n
 
a
n
d N
o
n-
V
e
t
e
r
a
n P
o
p
u
l
a
t
i
on
 
Non Veteran High School Graduate 
Non Veteran Some College
Veteran High School Graduate
Veteran Some College
113 
Figure 5.5: Black Veteran versus Black Non-Veteran Education  
                    (College and High School Graduates Only) 
0
0.05
0.1
0.15
0.2
0.25
190
0
190
2
19
0
4
190
6
190
8
19
10
191
2
191
4
1
91
6
191
8
192
0
1
92
2
192
4
192
6
192
8
193
0
Birth Year
P
e
rc
ent o
f
 Ve
ter
a
n and
 N
o
n-Ve
ter
a
n 
P
o
pul
a
t
io
ns
 
Black Non-Veteran High Shool Graduate 
Black Non-Veteran Some College
Black Veteran High Shool Graduate 
Black Veteran Some College
 
Figure 5.6: White Veteran versus White Non-Veteran Education  
                                  (College and High School Graduates Only) 
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
19
00
19
02
19
04
19
06
19
08
19
10
19
12
19
14
1
916
19
18
19
20
19
22
19
24
19
26
19
28
19
30
Birth Year
Pere
cnt
 of
 V
e
t
e
ra
n an
d N
o
n-
veter
an
 P
o
pu
latio
n
White Non-Veteran High Shool Graduate 
White Non-Veteran Some College
White Veteran High Shool Graduate 
White Veteran Some College
 
114 
 
   
 Research on the 1940s and 1950s supports the notion that veterans used the G.I 
Bill extensively and the data from the 1960 census continues to support that notion. In 
1950 19.35 percent of veterans reported that they had at least some college and in 1960 
24.40 reported that they had at least some college. The proportion of non-veterans with at 
least some college was 13.02 percent in the 1950 census and it grew to 16.15 percent in 
the 1960 census. This also represents an increase but not as much as the veteran increase 
during the same period (see tables 4.3 and 5.1). 
 
 
Income  
 The average (median) income in the sample is $10.2881 log dollars ($29,380.90)
3
, 
which is higher than the $9.8103 log dollars ($18,219.75) median in 1950. Incomes for 
veterans and non-veterans as well as blacks and whites rose between the 1950 and 1960 
Censuses. This was an expected trend because these workers aged into their primary 
work years and gained ten years of education, work experience, and networking. 
Furthermore there was significant inflation and growth in the GDP during this period 
which served to produce increased wages. Figure 5.7 illustrates the median log income 
for veterans versus non-veterans by birth year. In the 1950s younger cohorts for both 
veterans and non-veterans had lower median earnings than the older cohorts.  
 In the 1960s we see the younger veteran cohorts earning more than the older 
veteran cohorts; however, the younger non-veteran cohorts are still lagging behind the 
older non-veteran cohorts. Figure 3 also dramatically illustrates the premium that 
115 
 
   
veterans of the peak mobilization phase received compared to non veterans and veterans 
of other mobilization periods. Although the veteran peak mobilization cohort effect was 
pronounced in the 1950 cenus, the differences in income for veterans who served during 
this peak phase in 1960 are even more so.  
 Veterans in the sample have significantly and substantively higher mean incomes 
of $10.26 (ln) than their non-veterans peers who have mean incomes of $10.07 (ln).  
Blacks have a mean income of $9.53 (ln) that is significantly and substantively less than 
whites who have a mean income of $10.26 (ln).  Black veterans have a significantly and 
substantively higher mean log income of $9.72 (ln) than black non-veterans ($9.43 (ln)) 
and white veterans have a significantly and substantively higher mean log income of $ 
10.30 (ln) than their white non-veteran peers ($10.14291 (ln))(see figures 5.7, 5.8, and 
5.9). 
                                                                                                                                                 
3
 All income figures are expressed in log dollars. Log dollars were computed by taking 
the natural logarithm of income adjusted in year 2000 dollars. The inflation factor was 
computed by multiplying the 1960 dollar amount by a factor of 5.818.   
116 
Figure 5.7: 1960 Veteran versus Non-Veteran Log Income 
9.9
10
10.1
10.2
10.3
10.4
10.5
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
Log I
N
c
o
m
e
Non-Veteran
Veteran 
 
Figure 5.8: Black Veteran versus Non-Veteran Log Income 
 
9.1
9.2
9.3
9.4
9.5
9.6
9.7
9.8
9.9
10
10.1
10.2
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
Year
 2
000 Log 
I
n
com
e
 
Dol
l
ar
s
Non-Veterans
Veterans
 
 
 
 
 
 
 
 
 
 
117 
 
   
Figure 5.9: White Veteran versus Non-Veteran Log 
Income
10.05
10.1
10.15
10.2
10.25
10.3
10.35
10.4
10.45
10.5
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
L
og Inco
me
Non-Veteran
Veteran
 
  
The difference in the means is greater for black veterans (.2856) than for white veterans 
(.1614).  In essence, this indicates that even though both black and white veterans receive 
an earnings premium for their service, the magnitude of the black veteran premium is 
greater. Moreover, the black veteran advantage in some ways serves to reduce the 
differences in the disadvantages that blacks have versus whites in income attainment (see 
figure 5.10).      
 In order to gain a full appreciation of how the dynamics of income work I 
disaggregated the main demographic groups (veterans, non-veterans, blacks and white) 
by age category (see table 5.8).  
 World War II veterans aged thirty to fifty-four are significantly more likely to 
have a larger income than their non-veteran peers. These differences are also 
118 
 
   
substantively significant. When the veteran population is further disaggregated by race 
we find that veterans, both black and white, earn significantly more than their non-
veteran peers in every age category except in the oldest category (fifty-five to sixty) for 
white veterans. Moreover, the differences are substantively significant for all age and 
demographic categories of veterans versus non-veterans with exception of white veterans 
versus their non-veteran peers in the 45-60 age categories.  
 
Figure 5.10: 1960 Census Combined Black and White Veteran versus Non-Veteran Log Income 
9.4
9.6
9.8
10
10.2
10.4
10.6
190
0
190
2
190
4
190
6
190
8
191
0
191
2
191
4
191
6
191
8
192
0
192
2
19
24
192
6
192
8
193
0
Birth Year
Y
e
a
r
 20
00 Lo
g I
n
co
m
e
Black Non-Veterans
Black Veterans
White Non-Veteran
White Veteran
 
 
119 
Table 5.8: Income Significance Test by Age Categories 
 
 
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
30-34 22,482 10.27252 0.638241 *** 25,326 10.04653 0.741192
35-44 62,512 10.29743 0.661559 *** 29,977 10.04808 0.834117
45-54 21,299 10.1916 0.754546 *** 52,709 10.11082 0.831562
55-60 2,920 10.03305 0.88612
NS
30,028 10.02916 0.877834
Tests of 
Significance 
White
Age Group N Mean SD p N Mean SD
30-34 4380 9.556932 0.835714 *** 43428 10.2129 0.65985
35-44 8276 9.612839 0.830827 *** 84213 10.27595 0.693051
45-54 6543 9.480592 0.916152 *** 67465 10.19745 0.771123
55-60 2803 9.342204 0.990993
***
30145 10.09341 0.83925
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
30-34 1434 9.682726 0.784697 *** 2946 9.4957 0.852892
35-44 4189 9.75572 0.769954 *** 4087 9.466392 0.864833
45-54 1574 9.657619 0.828052 *** 4969 9.424516 0.935459
55-60 189 9.639017 0.700191
***
2614 9.320744 1.005482
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
30-34 21048 10.3127 0.60652 *** 22,380 10.11904 0.693367
35-44 58323 10.33633 0.635559 *** 25,890 10.13991 0.791008
45-54 19725 10.23421 0.731793 *** 47,740 10.18226 0.786305
55-60 2731 10.06032 0.017054
*
27,414 10.09671 0.833839
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
 
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001 
 
Socio-Economic Indicator (SEI)   
 The mean Duncan SEI for the sample is 35.56, which represents over a four point 
increase from the 1950 Census. The median SEI jumped from nineteen in 1950 to 
twenty-seven in 1960. Figure 5.11 illustrates the veteran premium for SEI particularly for 
veterans of the peak mobilization period.  
120 
Figure 5.11: 1960 Veteran Versus Non-Veteran Duncan Median SEI score by Birth Year 
0
5
10
15
20
25
30
35
40
45
50
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
Nu
m
b
er
Non-Veteran
Veteran
 
   As in the 1950 census one would expect that the mean SEI differences between 
veterans and non-veterans, blacks and whites, black veterans and non-veterans, and white 
veterans and non-veterans would be similar to the mean differences in income because 
one of the variables used to create SEI is income. In fact, the sample dramatically 
illustrates this. Veterans have a significantly higher mean SEI than non-veterans; whites 
have a significantly higher mean SEI than blacks, black veterans have a significantly 
higher mean SEI than black non-veterans; and white veterans have a significantly higher 
SEI than white non-veterans. These differences apply in all age groups (see Table 5.9). 
 
 
 
 
 
 
 
 
 
121 
 
   
Table 5.9: Significance Tests for Duncan SEI 
 
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
30-34 22,482 40.36509 23.46299 *** 25,326 33.98744 23.36944
35-44 62,512 39.10574 23.17892 *** 29,977 30.26974 21.72476
45-54 21,299 37.26673 23.31551 *** 52,709 33.56197 22.28667
55-60 2,920 35.74795
23.8652 ***
30,028 33.44319 22.40892
Tests of 
Significance 
White
Age Group N Mean SD p NMeanSD
30-34 4380 19.76416 15.99203 *** 43428 38.72357 23.57757
35-44 8276 18.81126 15.331 *** 84213 37.95485 23.01539
45-54 6543 17.36069 14.60754 *** 67465 36.30284 22.5908
55-60 2803 15.95219 13.51911
***
30145 35.29282 22.51709
Tests of 
Age Group
30-34
35-44
45-54
55-60
Age Group
30-34
35-44
45-54
55-60
Non-Veteran Veteran 
Black
 
 The net advantage for black veterans is higher than the net advantage for white 
veterans, although white veterans enjoy a significantly higher m
advantage for black and white veterans increased between 1950 and 1960, but m
black veterans. Figure 5.12 illustrates the differences for black and white veterans versus 
their non-veteran peers. It also highlights the peak m
as the dem
may have reduced the gap for blacks versus whites in occupational status attainm
particularly for those born to the peak m
Significance 
p
Black Veteran Black Non-Veteran
N Mean SD NMeanSD
1434 21.7106 0.448987 *** 2946 18.8167 15.39055
4189 21.20793 16.70729 *** 4087 16.35478 13.34083
1574 19.81131 16.56866 *** 4969 16.58442 13.84048
189 17.01587 0.984384
NS
2614 15.87529 13.51745
Tests of 
Significance 
N Mean SD p NMeanSD
21048 41.63602 23.30264 *** 22,380 35.98445 23.50675
58323 40.39124 40.20419 *** 25,890 32.46636 21.97705
19725 38.65962 23.21305 *** 47,740 35.32907 22.25615
2731 37.04431 23.88289
***
27,414 35.11833 22.36941
White Non-VeteranWhite Veteran
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
ean SEI. The net 
ore for 
obilization period premium as well 
obilization period penalty. Finally figure 5.12 illustrates how veteran status 
ent, 
obilization cohorts. 
122 
 
   
 
Figure 5.12: 960 Black and White Veteran versus Non-Veteran Duncan Median SEI score by Birth 
Year 
0
5
10
15
20
25
30
35
40
45
50
190
0
190
2
1
90
4
190
6
190
8
1
91
0
191
2
191
4
191
6
191
8
1
92
0
192
2
192
4
19
26
192
8
193
0
Birth Year
Median
 Duncan
 S
E
I
White Non-Veteran
White Veteran
Black Non-Veteran
Black Veteran 
 
 
Models and Multivariate Regression Analysis 
 In this section I follow the same conventions used in Chapter 4 by making use of 
the same five multivariate regression models to control for factors associated with 
earnings, education, and Duncan SEI outcomes to determine the net premium or penalty 
to veterans and non-veterans.  
 
 
 
 
123 
 
   
Education Models and Regression Analysis 
 
Model 1  
 As the bivariate descriptions above suggest, veterans tend to achieve higher 
education levels than their non-veteran peers, whites attained higher levels of education 
than blacks, and older birth cohorts generally had less education than younger cohorts 
(see table 5.6 above).   Model 1 generally confirms the descriptions above (see table 
5.10). When education attainment is regressed on World War II veteran status, age
2
, and 
race we see that the model explains nine percent of the variance. We also find that for 
veterans the predicted educational attainment is higher than for non-veterans (B =.1139).  
The coefficient for age
2
 is negative; however, it is the strongest predictor of education in 
this model (B = -.19).  In the 1950 census age
2
 was a strong predictor but it was less 
influential than race. As expected, this model suggests that older respondents had less 
education than younger respondents and that black respondents had significantly less 
education than whites. 
 
Model 2  
 The addition of marital status and region of residence background variables in 
model 2 modestly increases the proportion of variance in education that can be predicted 
from the independent variables from eight to nine percent. The veteran relative to non-
veteran advantage is still large and significant, while the black race penalty decreases 
slightly but is still large and significant. Both married and single respondents have 
significant educational attainment premiums relative to unmarried and non-single 
respondents respectively net of the effects of the other variables. Those residing in the 
124 
 
   
both the North and the South paid an educational attainment penalty relative to those not 
living in those regions respectively. The penalty was much higher for those living in the 
South than those in the North.  All of these findings are consistent with the 1950 census. 
 
Model 3  
 The addition of the mobilization phase variables in model 3 increases the adjusted 
R-square from .1024 to .1028. The highest educational attainment premiums in this 
model come from World War II veteran status (B = .1141) and marital status (B = .0513) 
and the biggest penalties come from race (B = -.162) and age (B = -.164). The other 
variables were for the most part unaffected.  The coefficient for mobilization phase 2 
(peak period) is .0118, which tells us that those who were a part of birth cohorts in the 
peak mobilization phase received a premium in educational attainment compared to those 
who were in birth cohorts of other phases. In the 1950 census this variable produced a 
negative coefficient (b=-.2057). Those who were a part of mobilization phase 3 also 
received an educational attainment premium (coefficient = .0309). This variable also 
produced a negative coefficient in the 1950 census (b = -.034). Finally, model 3 shows 
that black veterans received a significant educational premium (B = .0115).  
 
Model 4 
 Although model 4 is significant overall in that the independent variables reliably 
predict the dependent variable (p<.001), there is not much change in the amount of 
variance explained in model 3 (10.28 percent) versus model 4 (10.32 percent).  However, 
this model does show that there is a penalty in educational attainment for being a part of 
birth cohorts with large proportions of veterans.  This is curious for two reasons. The first 
125 
 
   
is the fact that this model produced a premium in 1950 (b= .0932). The second is that the 
peak mobilization phase variable in model 3 produced a premium and it has the largest 
concentrations of birth years with large proportions of veterans. The premium for black 
veterans increased slightly and remained significant. 
 
Model 5 
 Model 5 adds two veteran mobilization interaction terms to the regression 
equation. This addition slightly decreases the ability to predict the variance in educational 
attainment. In model 4 we could predict 10.32 percent of the variance and in model 5 we 
can predict 10.25 percent of the variance.  In model 5 the veteran educational premium 
decreases from .1212 to .1056 and the age
2
 penalty decreases from -.213 to -.188. 
However there is very little change, if any, in the other variables. As in the 1950 census 
this model clearly shows the educational advantage for being a veteran associated with 
the peak mobilization phase (coefficient =.0007).  
126 
 
   
Table 5.10: 1960 Education Models and Regressions 
 
 
 
 
 
 
 
 
127 
 
   
Income Models and Regression Analysis 
 
Model 1 
Table 5.11 illustrates the results of all five of the multivariate models for income. 
In model 1 income is regressed on World War II veteran status, age
2
, and race and we see 
that the model reliably predicts the dependent variable: income (F = 6861.69 and p ? 
.001). All three of the independent variables are significant.  The strongest predictor is 
race (Beta = -.2470), which indicates that blacks pay a substantial penalty in this census 
period. In 1950 the strongest predictor was age (beta = .0003) indicating that older 
respondents received a premium. Furthermore, in the 1950 census blacks paid a penalty, 
but not nearly as large as the 1960 penalty. The unstandardized race coefficients are -
.53307 in the 1950 census versus -.6705 in the 1960 census.  
For veterans the predicted income is higher than for non-veterans (B= .0992). In 
1950 the unstandardized coefficient was .1753 versus .1547 in 1960. The results of this 
model are generally consistent with the hypothesis that all things being equal veterans 
will earn more than their non-veteran peers. Moreover, it comes as no surprise that in 
1960 blacks paid an earnings penalty because of the color of their skin. One of the most 
interesting changes is that the age
2
 variable, that was so dominant in the 1950 census, is 
the weakest and has turned from positive to negative (b = .00029 in 1950 versus -.00003 
in 1960 unstandardized). This model explains 7.69 percent of the variance in earnings 
income versus 7.44 percent of the variance in 1950. 
 
 
 
128 
 
   
Table 5.11: 1960 Income Models and Regressions 
 
 
 
 
129 
 
   
Model 2 
  When marital status, region, and education are added to model 2 the overall 
model is still statistically significant with an F value of 6624.21 and p  .001. The 
addition of the marital, region, and education variables decreased the magnitude of the 
World War II veteran status variable from .1547 in model 1 to .1084 in model 2. 
Although the direction of the race variable was unaffected, in that blacks still earned less, 
the magnitude decreased from -.6705 to in model 1 to -.5038 in model 2. Being married 
increases and in fact the coefficient for married (.1633) is the third strongest in the 
model. Residing in the South tends to substantially decrease earnings income while living 
in the North tends to increase income.  
?
 As expected those with some college earned a large premium that was greater 
than the premium they earned in this model in 1950.  The unstandardized coefficients in 
1950 and 1960 were .3134 and .5243 respectively. High school graduates also earned a 
large premium (B = .1590) which is larger than the 1950 premium. Approximately 19.43 
percent of the variance in earnings income is explained by this model. 
 
Model 3   
The addition of the mobilization phase variables and an interaction term for being 
a black veteran in model 3 had relatively little influence on the variables used in previous 
models however, it did affect the World War II veteran status and race variables. This is 
somewhat expected since some of the variance is captured in the interaction term. 
Additionally, by controlling for mobilization phase the premium for World War II 
veterans decreased. The coefficient for the veteran status variable changed from .0695 to 
.0581, which is a smaller change than in the 1950 census when the coefficients changed 
130 
 
   
from .062 to .035; however, it is still positive.  The race variable increased in magnitude 
from -.5038 to -.5465, indicating that controlling for mobilization phase and background 
factors that blacks continued to pay a large income penalty. In contrast black veterans 
receive a premium for their service (B = .0265). Additionally, being a part of a birth 
cohort in the midst of the peak mobilization or the demobilization phases of World War 
II yields a penalty. The penalty was less than the penalty in 1950 for both phases, and the 
penalty was more substantial for the demobilization phase in both samples. The negative 
mobilization phase coefficients reflect the overall decline in earnings during this period 
as opposed to the veteran*mobilization interaction effects (see model 5). The addition of 
the mobilization variables increased the adjusted R-squared value to 19.52 percent. 
 
Model 4  
The removal of the mobilization phase variables and the addition of the percent of 
a birth cohort that served in model 4 had very little effect on the veteran status or veteran 
race interaction term. The strongest predictor in this model, as in the previous models, 
and in the 1950 analysis is having some college (B = .2703). The next two most 
influential predictors are race (B= -.2010) and being a high school graduate (.1589), 
which had virtually the same coefficients in the 1950 census.     
The magnitude of the veteran advantage remained the same as in model 3, which 
is consistent with the lack of change in the 1950 census. However, we find that those in 
cohorts with large proportions of veterans earn a small premium (B=.0183). The 
unstandardized coefficient is .0576 which is much smaller than the unstandardized 
coefficient in 1950 (.3434). This finding supports the notion developed above that there 
is an earnings premium for veterans born to cohorts that served during the peak 
131 
 
   
mobilization phase. During that phase of the mobilization the ratio of those who served to 
those who did not serve was much higher than the other two phases (see figure 5.2 
above). The difference in the magnitudes of the coefficients in 1950 and 1960 might be 
explained by the fact that the 1950 percentages of veterans in a cohort were inflated by 
using the 1960 percentages because of the underenumeration problems in the 1950 
census (see chapter 4). The explanation of the variance decreases from 19.52 percent to 
19.48 percent between models. 
 
Model 5  
  This model clearly illustrates the veteran advantage for being a part of the peak 
mobilization period. Veterans who were a part of the peak mobilization period earned a 
premium (B =.0365) for their service. Although the magnitude of this variable is small it 
still provides more influence than all except race (-.2000), married (.1625), high school 
(.1586), some college (.2699), and veteran status (.0371).  
 Additionally, veterans who served during the demobilization phase received 
premium for their service (B = .0109). In the 1950 census they paid a penalty (b = .0410).  
The adjustment from model 4 to model 5 decreased the World War II veteran premium 
from .0578 to .0371; however, this combination does not appreciably change any of the 
other control variable coefficients. Additionally, model 5 shows that black veterans 
earned a premium for their service controlling for all other variables in the model that 
was about the same as in model 4. Moreover the change between model 4 and 5 was 
similar to the changes in the 1950 census.  The removal of the percent of birth year 
variable and the addition of the veteran and mobilization phase interaction terms in 
model 5 nominally increases the explained variance from 19.48 percent to 19.51 percent. 
132 
 
   
Duncan SEI Models and Regression Analysis 
 The descriptive section above showed very clearly that veterans generally have 
higher SEIs than non-veterans, whites generally have higher SEIs than blacks, and that 
both black and white veterans generally have higher SEIs than their non-veteran peers 
These relationships held for both the 1950 and the 1960 Censuses. The results of the SEI 
regression models are shown in table 5.12.  
 
Model 1 
 In model 1, I regressed SEI on World War II veteran status, age
2
, and race.  More 
specifically, we find that World War II veterans receive an SEI premium for their service 
(B = .1097) and that older respondents (B = -.014) as well as blacks (B = -228) paid an 
SEI penalty.  Interestingly, in 1950 there was a premium associated with age.  The 
magnitudes and the directions of the coefficients for veteran status and race were similar 
to the 1950 census coefficients. This model explains .0687 of the variance in SEI versus 
.0624 in the 1950 census. 
133 
 
   
Table 5.12: 1960 Duncan SEI Models and Regressions 
 
 
 
 
134 
 
   
Model 2 
 The addition of marital, regional, and educational background variables to the 
regression equation decreases the coefficient for the veteran status variable from .1097 to 
.0614 and the magnitude of the age
2
 coefficient changes from negative to positive and 
gets larger as a result of controlling for background factors. In the 1950 census, the 
model 1 coefficient for age
2
 was positive and increased slightly in model 2.  
 The penalty for being black decreases when one controls for marital status, 
region, and education; however, the coefficient remains large and negative. This is 
consistent with the results of the 1950 census.  
 Married respondents (B = .0698), those residing in the North (B = .0631), and 
those residing in the South (B = .0440) all receive SEI premiums as they did in the 1950 
census.   
 Also consistent with 1950 census analysis, is the fact that the largest coefficients 
come from the education independent variables. Those with some college (B = .5583) 
and those with high school degrees (B = .2341) continued to earn substantial SEI 
premiums. 
 
Model 3  
 The addition of the race*veteran interaction term and the mobilization phase 
variables also increases the veteran premium by almost half an SEI unit. The age
2
 
variable decreased marginally but remains positive and significant.  
 There was little change in the magnitudes of the married or the education 
variables. The peak mobilization phase produced negative coefficients indicating that 
respondents associated with the peak mobilization cohorts paid an SEI penalty (B = -
135 
 
   
.019).  However, respondents associated with demobilization phase cohorts received a 
small premium (B = .0035). These findings are not surprising in light of the data 
presented in figure 5.11, which shows a substantial dip in the non-veteran portion of 
mobilization phase 2 SEI curve.  In this model the addition of the race*veteran 
interaction term and the mobilization phase variables modestly increases the adjusted R-
squared from 35.10 to 35.14 percent. 
 
Model 4  
 In this model, I once again remove the mobilization phase variables and add in a 
variable that controls for the percent of veterans in a particular birth cohort. The veteran 
premium for SEI decreased slightly as did the premium for married respondents.  
Furthermore, the coefficient for the percent of veterans in a birth cohort is -.031. 
Essentially, this tells us that for every one standard deviation increase in the percentage 
of veterans in birth year we can expect a .031 standard deviation decrease in SEI. This is 
much different than in the 1950 census when cohorts with a large proportion of veterans 
earned a premium (b = 3.4798). Making these changes in the model did not affect most of 
the coefficients of the other variables or the adjusted R-squared. 
 
Model 5  
  In the final SEI model I removed the percent of veterans in a birth year variable 
and added in the veteran*mobilization phase interaction terms. Making these changes 
makes no difference in the adjusted R-squared.  The coefficient for World War II veteran 
status decreases from .0725 to .0363 and the coefficients for age
2
 increases from .0427 to 
136 
 
   
.073. The coefficients for race, married, North and South residence, high school graduate 
and some college remained virtually the same in magnitude, direction, and significance.  
 Interestingly, although there was an SEI penalty for being a part of the peak 
mobilization phase in model 3, there is a premium for being a veteran in the peak 
mobilization phase in this model.  The coefficient for this variable is .0332, which 
suggests that veterans for the peak World War II mobilization period received a 
statistically significant premium for their service. Veterans who were a part of the 
demobilization phase received a smaller SEI premium of .0277.  Past research would 
suggest that the reasons for this increase are that they received the benefits of being 
veteran in terms of hiring preferences during their prime working years (30-34) and they 
gained educational advantages through the G.I. Bill.  
 
Summary 
 The results of this chapter have shed additional light on the social status 
attainment of World War II veterans. First and foremost the veteran advantage continued 
in this census period. Veterans received significant education, earnings, and SEI 
premiums in all of the regression models which would substantiate hypothesis one.  
 Black and white veterans continued to benefit from their service compared to 
their non-veteran peers in terms of education, income, and occupational status. Moreover, 
the magnitude of the differences between mean income for black veterans and black non-
veterans is greater than the differences between white veterans and non-veterans. In 
essence the service effect for black veterans decreases the race gap in individual earnings, 
education, and occupational status. This finding supports hypothesis number two. 
137 
 
   
 Additionally, veterans who are a part of a birth cohort associated with the peak 
mobilization phase continued receiving a premium from the 1950 Census for all three 
dependent variables. However, demobilization phase veterans earned premiums only for 
income and SEI and suffered a penalty in terms of educational attainment. Furthermore, 
although the income coefficients for the peak phase veterans were larger than coefficients 
for the demobilization phase veterans, the demobilization phase veterans had larger SEI 
coefficients. These findings generally support hypothesis three that veterans serving 
during the peak mobilization period will earn more than those serving during the 
beginning phase and less than those serving during the demobilization phase of World 
War II.  
 The 1960 census was somewhat different than the 1950 census with respect to 
percent of veterans in a cohort. In the 1950 census cohorts with large percentages of 
veterans received education, income, and SEI premiums. However, in the 1960 census 
those in cohorts with a large proportion of veterans earned an income premium, but paid 
educational and SEI penalties. These findings would suggest that hypothesis four does 
not hold for the 1960 census period.     
  As expected the proportion of married respondents increased between the 1950 
and 1960 censuses. Furthermore, those who were married received significant earnings, 
education, and SEI premiums compared to those who were not married. Those who 
resided in the North received income, and SEI premiums but not educational premiums, 
while those with residence in the South paid penalties in terms of all three dependent 
variables. As expected, those with high school or some college received education 
premiums that manifested themselves in terms of higher incomes and occupational status. 
Past research makes it clear that throughout the 1950s veterans, both black and white, 
138 
 
   
took advantage of the G.I. bill. The findings in this chapter generally support those 
findings. Both groups and veterans in general had higher mean education attainment than 
their non?veteran peers.     
     Chapter 6: World War II Veterans and the 1970 Census 
 
 
 In the 1970 census the World War II cohort is between the ages of forty and 
seventy and it has been twenty-three years since the end of World War II. The majority of 
the cohort is in its prime working years and earning at or near its peak income. 
Additionally, the United States has just emerged from one of the most turbulent periods 
in its history. The most prominent event in the decade between the 1960 and 1970 census 
was the Civil Rights Movement which by law brought an official end to segregation and 
disenfranchisement of African-Americans even if inequality persisted in practice. 
Furthermore, the 1960s brought increased racial tensions with large race riots in Watts 
(Los Angeles) in 1966, and Detroit, Cleveland, and Newark in 1967.  This analysis gives 
us an opportunity to evaluate whether these events impacted the racial gap in attainment 
of World War II veterans.   
 In 1962 the Cuban Missile Crisis put the country on edge, brought the world to 
the brink of a nuclear war, and demonstrated the fact that Cold War was alive and well. 
The Vietnam War was ongoing and the center of a great deal of controversy. One of the 
more vivid events that illustrates the passion on both sides of the Vietnam War debate 
was the Kent State University shootings in May, 1970. 
 Finally, the 1960s has been described as a period of counter culture revolution 
because the new generation was determined to reject a pre-World War II conformist 
lifestyle. John Macionis (1999: 80) defines counterculture as ?cultural patterns that 
139 
 
   
strongly oppose those widely held accepted within a society.? In describing the 1960s he 
refers to hippies and other counterculturalists favoring a cooperative lifestyle in which 
?being? took precedence over ?doing??. Such differences led some people at that time 
to ?drop out? of the larger society.  
 
Descriptive Data 
In this chapter I use the same analytical methods utilized in chapters four and five 
to describe the sample, compare the dependent measures, and examine the dependent 
variables while simultaneously making comparisons to the 1960 census. As in the 
previous chapters I shaped the data to age the World War II cohort by ten years.  After 
making all of the necessary adjustments (see chapters four and five) I was left with a total 
sample of 220,335, which is 26,918 respondents less than the total in the 1960 census.    
The median age for the sample is 51 years old with an age range of forty to 
seventy. In the 1960 census the median age was 43. Additionally, all of the respondents 
have birth years between 1900 and 1930 as in the 1960 census.  The plurality of the 
sample (41.82 percent) is in the forty-five to fifty-four year old age category.  As 
expected the majority of the sample is white (91.26 percent), from the south (56.66 
percent), and married (87.07 percent).  In the 1960 census the sample was 91.10 white, 
56.68 percent from the South, and 88.06 percent married.  Moreover, the average income 
adjusted to year 2000 dollars is $35,725.90, the majority of the respondents had less than 
a ninth-grade education, and had a median SEI of thirty-three  (see table 6.1). 
Of the 220,335 total respondents 109,539 are veterans, who represent 
approximately 40.71 percent of the sample. Figure 6.1 illustrates the 1970 census 
proportion of veterans to non-veterans. Black veterans make up approximately seven 
140 
 
   
percent (7870) of the veteran population and white veterans comprise the other ninety-
three percent (101,669).  
Table 3.1:  1970 Sample Descriptive Characteristics 
 
Cat
e
gor
y
Su
b
-
C
a
t
e
g
o
r
y
T
o
t
al
T
o
t
a
l
B
lack W
h
it
e
T
ot
al
Black W
h
ite
T
o
t
a
l S
a
mp
le S
i
ze
220,335
109,539
7,870
101,66
9
1
10,796
11,379
99,417
Ag
e
Medi
an Age, 
y
ear
s
 
51
49
48
49
55
5
3
56
# age 40-
44
22.25%
22.02%
22.72%
2
1
.97%
22.48%
24.40%
22.26%
# age 45-
54
41.82%
60.31%
59.67%
6
0
.35%
23.55%
29.38%
22.88%
# age 55-
64
29.24%
16.63%
16.58%
1
6
.64%
41.69%
35.87%
42.36%
# age 65-
70
6.69%
1.04%
1.03%
1
.
04%
12.27%
10.35%
12.49%
Age Range (
Y
ounges
t -
 Ol
d
e
st)
40 -
 70
40 -
 70
40 -
 70
40 -
 70
40 -
 70
40 - 70
40 - 
70
Race 
Bl
a
c
k
8
.74%
N/A
100.00
%
N
/A
N/A
100.00%
N/A
W
h
i
t
e
91.26%
N/A
N
/A
1
00.00%
N/A
N
/A
100.00%
Re
gi
on
Nor
t
h
27.03%
27.44%
22.34%
2
7
.84%
26.62%
18.35%
27.57%
South
56.66%
55.31%
68.20%
5
4
.32%
57.99%
75.60%
55.97%
W
e
s
t
16.31%
17.24%
9.47%
1
7
.84%
15.39%
6.05%
16.46%
Ma
r
i
t
a
l
 S
t
a
t
u
s
 
Mar
r
i
ed
87.07%
87.48%
74.68%
8
8
.47%
86.65%
76.25%
87.84%
Di
vorc
ed 5.17%
5.45%
13.62%
4
.
82%
4.89%
10.97%
4.20%
W
i
dow
ed 2.34%
1.64%
3.71%
1
.
48%
3.03%
5.98%
2.70%
Si
n
g
l
e
 Nev
e
r
 Mar
r
i
ed
5.43%
5.43%
7.99%
5
.
23%
5.42%
6.80%
5.26%
In
co
me 
Medi
an Earn
i
ngs 
I
n
com
e
 
$8,050.00
$8,850.00
$6,050.00
$9,050.00
$7,250.0
0
$4,5
50.00
$7,650.00
Adj
u
s
t
ed to 2000 Dol
l
a
r
s
$35,725.90
$39,276.30
$26,849.90
$40,163.90
$32,175.50
$20,19
2.90
$
33,950.70
Natur
a
l
 Logar
i
t
hm
 Inc
o
m
e
 i
n
 Y
ear
 $2000
10.4836
10.5784
10.1980
10.600
7
1
0.3790
9.9131
10.4327
Ed
u
cat
ion
None-
 Gr
ade 8
29.45%
20.68%
41.07%
1
9
.10%
38.13%
62.91%
35.29%
G
r
ad
e 9-11
20.92%
22.10%
26.49%
2
1
.76%
19.75%
18.62%
19.88%
Hi
gh Sc
hool
 
Gr
aduate
27.11%
30.76%
20.18%
3
1
.58%
23.50%
11.95%
24.83%
Som
e
 Col
l
ege or Coll
ege G
r
a
d
22.52%
26.46%
12.26%
2
7
.56%
18.62%
6.51%
20.00%
SEI
 (
m
ed
i
a
n
)
33
44
18
44
27
1
5
32
V
e
te
r
a
ns
 Non-V
e
te
r
a
ns
W
o
rld
 W
a
r II 1970 Cen
s
u
s
 
 
 
141 
 
   
The majority of the veterans in the 1970 sample fall into the forty-five to fifty 
year old age category with an average age of forty-nine years. Most are from the South 
(55.31 percent) and are married (87.48 percent). 
Furthermore, they have a median income of $39,276.30 (adjusted to year 2000 
dollars), are high school graduates, and have a median SEI of 44.  Their non-veteran 
peers are on average fifty-three years old (median) and the majority of them are a part of 
the fifty-five to sixty-four year old age category.  The majority of the non-veteran group 
is from the South (57.99 percent) and married (86.65 percent). They have a median 
income of $32,175.50 (adjusted to year 2000 dollars), average less than a ninth-grade 
education, and have an average SEI score of twenty-seven (see table 6.1). 
As one might expect white veterans look much like the general veteran 
population.  Like the general veteran population they have a median age of forty-nine 
years, are in the forty-five to fifty-four year old age category (60.35 percent), are married, 
and reside in the South (88.47 percent).  Their average income adjusted to year 2000 
dollars is $40,163.90, over fifty percent of them are high school graduates or have some 
college, and their average SEI is forty-four (median) (see table 6.1).   
Black veterans in this sample are on average forty-eight years old (median) and 
most of them are in the forty-five to fifty-four year old age category (59.67 percent).   In 
contrast their non-veteran peers they have a median age of fifty-three years and the 
largest proportion of them are in the fifty-five to sixty-four year old age group. Black 
veterans like their non-veteran peers are predominantly married and from the South. They 
have an adjusted income of $26,849.90, the majority have less than a ninth grade 
education (41.07%) and they have an eighteen mean SEI (see table 6.1). 
 
142 
 
   
 
Figure 6.1: 1970 Census Veterans versus Non-Veteran 
 
Background Descriptive Statistics and Discussion 
Age   
 The first comparison between veterans and non-veterans in this, preceding, and 
succeeding chapters is with age.  Although the general descriptive characteristics are 
stated above, in this section I use significance tests to compare veterans to their non-
veteran peers then disaggregate them by black and white veterans and make the same 
comparisons. On average the veterans are significantly younger than their non veteran 
peers by 4.67 years. In the 1960 sample they were 5.8 years younger. Based on their 
entry age into World War II, veterans should be between the ages of forty-one and fifty-
four if they entered between the ages of eighteen and twenty-six (see table 6.2).  
 
 
143 
 
   
 
Table 6.2: Veteran Expected Ages in 1970 by Entry Age 
1970
Year of Entrance 18 19 20 21 22 23 24 25 26
1942 46 47 48 49 50 51 52 53 54
1943 45 46 47 48 49 50 51 52 53
1944 44 45 46 47 48 49 50 51 52
1945 43 44 45 46 47 48 49 50 51
1946 42 43 44 45 46 47 48 49 50
1947 41 42 43 44 45 46 47 48 49
Age of Entrance 
 
  
 White respondents have an average age of 51.73 and they are significantly older 
than black respondents by only .399 years.  However, the magnitude of the difference in 
mean age between black and white respondents is very small and the significance is more 
a reflection of sample size than difference in mean age.  Black veterans have a mean age 
of 49.19 years and they are significantly younger than their non-veteran peers by 3.68 
years. Similarly white World War II veterans have a mean age of 49.36 years and they 
are significantly younger than their non-veteran peers who have a mean age of 54.15 
years. The median ages for black veterans, black non-veterans, white veterans, and white 
non-veterans are forty-eight, fifty-three, forty-nine, and fifty-six respectively. Table 6.3 
provides the results of the significance tests described above. 
 As in previous chapters I disaggregate the data by birth cohorts to examine the 
effects of military service and mobilization phase on social status attainment. The logic 
for using birth cohorts and the cutoff years for the establishment of mobilization phases 
remains the same as in previous chapters. Figure 6.1 illustrates the 1970 census World 
War II cohort identified by mobilization phase.  
 
 
144 
Table 6.3: 1970 Census Age Significance Tests 
 
Tests of 
Significance 
NMeanSD p NMeanSD
109,539 49.34758 5.683032
***
110,796 54.01679 8.857428
Tests of 
Significance 
White
NMeanSD p NMeanSD
19,249 51.36776 7.823939
***
201,086 51.72688 7.805223
Tests of 
Significance 
NMeanSD p NMeanSD
7,870 49.1939 5.732816 *** 11,379 52.87125 8.677329
Tests of 
Significance 
NMeanSD p NMeanSD
101,669 49.35948 5.679015
***
99,417 54.1479 8.868419
White Veteran White Non-Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
Region  
 Just over half of the sample, (56.66 percent), maintains their residence in the 
South, 27.03 percent in North, and 16.31 percent in the West. The regional distribution of 
respondents looks very much like the geographical distribution in the 1960s. The veteran 
and non?veteran proportions for region of residency are very similar. Just over twenty-
seven percent of the veterans reside in the North versus 26.62 percent of the non-
veterans; 55.31 percent of the veterans and 57.99 percent of the non-veterans reside in the 
South; and 17.24 percent of the veterans versus 15.39 percent of the non-veterans live in 
the West.  I conducted a significance test assuming unequal variance to determine if the 
population proportions for veterans and non-veterans were different in terms of region.  
In this test I combined all of the regions and tested whether there was a significant 
difference between veterans and non-veterans. Moreover, the proportion significance 
tests that disaggregate each demographic category by region are all significant with the 
145 
 
   
exception of the test that compares the proportion of white veterans to non-veterans in the 
North (see table 6.4).  
Table 6.4: 1970 Region Significance Tests 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
  
 Blacks are significantly less likely than whites to live in the South, with 55.13 
percent of the blacks living in that region versus 56.66 percent of the whites. In the 1960 
blacks were more likely than whites to live in the South, with 74.70 percent of the blacks 
living in that region versus 54.92 percent of the whites. Some might argue that this 
represents the migration of blacks to the North. Although this is not a new finding it does 
confirm studies on this topic.  
 
 
146 
 
   
 
During the early part of the 20th century, black Americans left 
the American South in large numbers. Several factors 
precipitated their ?Great Migration? to northern cities.1 
First, the mechanization of southern agriculture rendered many 
farm workers, including blacks, redundant. Second, the 
industrialization of the Northeast and Midwest created millions 
of manufacturing jobs for unskilled workers. And not least in 
importance, the generally oppressive racial climate in the South 
acted as a ?push? factor for many decades as blacks sought out 
more tolerant communities in other regions. Even as whites 
migrated to the Sunbelt in large numbers at mid-century, black 
migration out of the South exceeded black in-migration as late as 
the period 1965?70 (Cowper, Longino, Kubal, Manheim, 
Dienstfrey, and Palmer 2000: 2). 
 
 A significantly higher proportion of blacks live in the North and West than whites 
(see table 6.4).  In the 1960 census the West had the greatest disparity between black and 
white residents with 6.75 percent and 16.77 percent respectively. In the 1970s the 
proportion of blacks living in the West was slightly but significantly greater than the 
proportion of whites living in the West. The differences in the proportions for all regions 
are significant at an alpha level of .001.  
  Overall, there are significant differences in the regional residences of black 
veterans and non-veterans. Black veterans are significantly more likely to reside in the 
North and West than non-veterans and non-veterans are significantly more likely to be 
from the South than veterans (see table 6.4). Military service might have been one of the 
dynamics of black migration from the South.  White veterans and non-veterans also have 
significant differences in their overall regional residences. White veterans are 
significantly more like to reside in the West than white non-veterans and white non-
veterans are significantly more like reside in the South than White veterans. These 
differences are significant at the .001 level (see table 6.4 above). 
 Marital Status 
147 
 
   
 Married respondents represent approximately 87.07 percent of the 1970 sample, 
which indicates a leveling of the marital status for the World War II cohort. The rates 
increased between 1950 and 1960 from 73.84 to 88.06 percent.  As in the 1960 census 
whites are significantly more likely than blacks to be married. However, blacks are 
significantly more likely to be divorced, widowed, or single than whites (see table 6.5).  
Veterans are significantly more likely to be married or divorced than non-veterans; 
however, non-veterans are significantly more likely to be widowed than their veteran 
peers. The proportion of married veterans decreased slightly from 88.60 percent in 1960 
to 87.07 percent in 1970, while the proportions for non-veterans decreased from 87.63 
percent to 86.65 percent over the same period.  
  Black veterans are significantly more likely to be divorced or single and 
significantly less likely to be married or widowed than their non-veteran peers. White 
veterans are significantly more likely to be married, divorced or widowed than their non-
veteran peers and significantly less like to be widowed than white non-veterans (see table 
6.5 below). 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 6.5: 1970 Marital Status Significance Tests 
148 
 
   
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001 
Single Never Married 5,946 0.0543
NS
6,008 0.0542
Tests of 
Significance 
N Proportion p N Proportion
Maried 14554 0.7561 *** 177,281 0.8816
Divorced 2320 0.1205 *** 9,068 0.0451
Widowed 972 0.0505 *** 4,186 0.0208
Single Never Married 1403 0.0729
***
10,551 0.0525
Tests of 
Significance 
N Proportion p N Proportion
Maried 5,877 0.7468 * 8,677 0.7625
Divorced 1,072 0.1362 *** 1,248 0.1097
Widowed 292 0.0371 *** 680 0.0598
Single Never Married .0680
Tests of 
Si
629 0.0799
**
774 0
gnificance 
N Proportion p N Proportion
Maried 89,950 0.8847 *** 87,331 0.8784
Divorced 4,896 0.0482 *** 4,172 0.0420
Widowed 1,506 0.0148 *** 2,680 0.0270
Single Never Married 5,317 0.0523
NS
5,234 0.0526
Black White
Black Veteran Black Non-Veteran
White Veteran White Non-Veteran
 
 
 
 
 
 
 
 
 
 
 
Education 
 In every age and demographic category there was an increase in education 
between 1960 and 1970. Although this is not a longitudinal survey, we can infer from the 
data that the population either continued to formally educate itself or that the more 
educated portions stayed in the labor force longer in the decade between 1960 and 1970. 
For example a comparison of the thirty-five to forty-four year old veteran age group in 
1960 should look similar to the forty-five to fifty-four year old age group in 1970.  
However, we find that in 1960 this group had a mean education of 6.016 and in 1970 this 
level increased to 6.166. The average person in the sample attained more than a tenth 
grade education level which is a full grade higher than the average in the 1960 census. 
Figure 6.2 illustrates the educational attainment by birth year for all respondents in the 
149 
 
   
1970 census. Black respondents attained a little more than a ninth grade education on 
average and whites attained almost an eleventh grade education.  
 
Figure 6.2: 1970 Cohort Education Levels 
0
500
1000
1500
2000
2500
3000
3500
4000
19
00
1
9
0
2
1
9
0
4
19
06
19
08
19
10
19
12
1
9
1
4
19
16
19
18
19
20
19
22
1
9
2
4
1
9
2
6
19
28
19
30
Birth Year
Nu
m
b
er
Grade School
Ninth to Eleventh 
 High School Graduate
 Some College
 When the data are disaggregated by veteran status, race, and age categories we 
find that veterans significantly attained almost a full grade of education more than non-
veterans. Both black and white veterans achieved a grade more education than their non-
veteran peers with mean differences being significant at the .001 level (see table 6.6); 
however, these differences are only substantively different for black veterans versus non 
veterans. Moreover, although there are statistically significant differences for every age 
group and demographic category; only the white versus black (all age categories) and 
black veterans versus non veterans in the 45 -54 year old age categories yield substantive 
differences (see table 6.7).  Overall, the difference in the means is greater for black 
veterans versus black non-veterans (1.00) than for white veterans versus non-veterans 
150 
 
   
(.791). Compared to 1960 both black and white veterans relative to their non-veteran 
peers lost a small bit of their veteran premium. The differences were 1.128 and .857 for 
blacks and whites respectively. 
 
Table 6.6 1970 Education Significance Tests 
Tests of 
Significance 
N Mean SD p N Mean SD
Aggregate 109,539 6.123454 2.113354
***
110,796 5.27029 2.318069
Tests of 
Significance 
White
N Mean SD p N Mean SD
Aggregate 19249 4.316484 2.145599
***
201,086 5.826343 2.225576
Tests of 
Significance 
N Mean SD p N Mean SD
Aggregate 7,870 4.911055 2.137471
***
11,379 3.905264 2.05291
Tests of 
Significance 
N Mean SD p N Mean SD
Aggregate 101,669 6.217303 2.082254
***
99,417 5.426527 2.295313
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
151 
 
   
Table 6.7: 1970 Education Significance Test by Age Category 
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
40-44 24,122 6.332228 2.051607 *** 24,907 5.980206 2.262232
45-54 66,058 6.165885 2.077142 *** 26,094 5.312831 2.306036
55-64 18,220 5.73101 2.242608 *** 46,196 5.040263 2.262726
65-70 1,139 5.518876 2.421409 *** 13,599 4.669829 2.311738
Tests of 
Significance 
White
Age Group N Mean SD p N Mean SD
40-44 4564 4.975679 2.152844 *** 44465 6.274283 2.133429
45-54 8039 4.461749 2.131413 *** 84113 6.064116 2.131143
55-64 5387 3.768702 1.986204 *** 59029 5.369513 2.256207
65-70 1259 3.343129 1.940237 *** 13479 4.865494 2.322416
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
40-44 1788 5.401007 2.07194 *** 2776 4.701729 2.159832
45-54 4696 4.878833 2.130265 *** 3343 3.87586 1.990814
55-64 1305 4.396935 2.096665 *** 4082 3.567859 1.906649
65-70 81 4.246914 2.31047 ** 1178 3.280985 1.897544
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
40-44 22334 6.406779 2.031647 *** 22,131 6.140572 2.2235
45-54 61362 6.264382 2.039856 *** 22,751 5.523977 2.273543
55-64 16915 5.833934 2.220474 *** 42,114 5.18298 2.243526
65-70 1058 5.616257 2.403128
***
12,421 4.801546 2.304232
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
Figure 6.3 illustrates the educational attainment of veterans and non-veterans by 
birth year. As described above, veterans attained significantly more education than non-
veterans. Additionally, this figure dramatically illustrates the veteran premium for 
serving during the peak mobilization period of World War II. Figures 6.4 and 6.5 display 
the net education advantage for veterans, both black and white, as well as the peak 
mobilization phase advantage. These figures also display the net education disadvantage 
for veterans in the demobilization phase.  
 
 
 
 
 
 
 
 
152 
 
   
Figure 6.3: Veteran versus Non-Veteran Education (College and High School Graduates Only) 
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
1
900
19
0
2
1
9
0
4
1
906
1
9
0
8
1
910
19
1
2
1
9
1
4
1
916
1
918
1
9
2
0
1
9
2
2
1
9
2
4
1
926
1
928
1
9
3
0
Birth Year
P
e
r
c
nt
 of 
V
e
t
erans or N
o
n
-
V
e
t
eran
s
 in
 
Cohort
Non-Veteran High School Graduate
Non-Veteran Some College
Veteran High School Graduate
Veteran Some College
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
153 
 
   
 
Figure 6.4: Black Veteran versus Black Non-Veteran Education (College and High School Graduates 
Only) 
0
0.05
0.1
0.15
0.2
0.25
0.3
1
900
1
902
1
9
0
4
1
9
0
6
1
908
1
9
1
0
1
9
1
2
1
914
1
916
1
9
1
8
1
9
2
0
1
922
1
9
2
4
1
9
2
6
1
928
19
3
0
Birth Year 
Percen
t o
f
 B
l
ack 
V
e
te
r
a
n
s
 
o
r
 No
n
-
V
e
teran
s
Black Non-Veteran High School Graduate
Black Non-Veteran Some College
Black Veteran High School Graduate
Black  Veteran Some College
 
 
Figure 6.5: White Veteran versus White Non-Veteran Education (College and High School 
Graduates Only) 
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
19
00
1
9
02
1
9
04
1
9
06
1
9
08
1
91
0
1
91
2
1
91
4
1
91
6
1
91
8
1
92
0
19
22
19
24
19
26
19
28
1
9
30
Birth Year
P
e
r
c
e
n
t
 W
h
i
t
e
 V
e
t
e
r
a
n or
 
N
on-
V
e
t
e
r
a
n 
White Non-Veteran High School Graduate
White Non-Veteran Some College
White Veteran High School Graduate
White Veteran Some College
 
154 
 
   
 
 Income  
 The average (median) income in the sample is $10.48 log dollars ($35,725.90)
4
, 
which is higher than the $10.29 log dollars ($29,380.90) median in 1960. As one would 
expect, incomes for veterans and non-veterans as well as blacks and whites rose between 
the 1960 and 1970 censuses. This was an expected trend because these workers continued 
to gain work experience and networking opportunities.  Figure 6.6 illustrates the median 
log income for veterans versus non-veterans by birth year.  
 In the 1960s we saw younger veteran cohorts earning more than older veteran 
cohorts and this trend continues in the 1970s. Moreover, we see the precipitous drop in 
income for those in cohorts 1906 and prior. Those serving during in the 1906 cohort 
would have been 64 years old in the 1970 census.  Figure 6.6 also illustrates the fact that 
the veteran advantage for the peak mobilization phase continues in the 1970?s. Although 
the difference is not as pronounced as it was in the 1960?s the advantage is still very 
salient.  
 Veterans in the sample have significantly and substantively higher mean incomes 
of $10.48 (ln) than their non-veterans peers who have mean incomes of $ 10.19 (ln).  
Blacks have a mean income of $9.81 (ln) that is significantly and substantively less than 
whites who have a mean income of $ 10.39 (ln).  Additionally, both black and white 
veterans have significantly and substantively higher incomes than their non-veteran peers 
(see figure 6.7). Black and white veterans have significantly and substantively higher 
mean incomes than black and white non-veterans (see figure 6.7 and table 6.8). 
155 
 
   
 
Figure 6.6: 1970 Veteran versus Non-Veteran Log Income 
9
9.2
9.4
9.6
9.8
10
10.2
10.4
10.6
10.8
19
00
19
01
19
02
19
03
19
04
19
05
1
9
06
1
9
0
7
19
0
8
19
09
19
10
19
11
19
12
19
13
19
14
19
15
1
9
16
1
9
1
7
1
9
1
8
19
19
19
20
19
21
19
22
19
23
19
24
19
25
19
26
1
9
27
1
9
2
8
19
2
9
19
30
Birth Year
Loga
ri
thmi
c
 I
n
c
o
me
 i
n
 Y
e
a
r
 
$2000
Non-Veteran
Veteran
 
 In the 1960s the difference in the log income means for black veterans versus 
black non-veterans was .2856419. In 1970 the difference grew to .2992489. White 
veterans and non-veterans had a significant difference of .1613541 in 1960 and the 
difference grew to .2696017 in 1970.  The difference in log income between black 
veterans and non-veterans is greater than the difference between white veterans and non 
veterans in the 1970s, although the difference in change from the 1960s to the 1970s was 
greater for white veterans versus non-veterans than black veterans versus non-veterans. 
As stated in previous chapters the black veteran advantage continues to reduce the 
difference in the disadvantage that blacks have versus whites in income attainment (see 
figure 6.7).      
  
                                                                                                                                                 
4
 All income figures are expressed in log dollars. Log dollars were computed by taking 
the natural logarithm of income adjusted in year 2000 dollars. The inflation factor was 
computed by multiplying the 1960 dollar amount by a factor of 4.438.   
156 
 
   
Figure 6.7: 1970 Census by Race and Veteran Status Log Income 
8
8.5
9
9.5
10
10.5
11
1900 1
9
0
2
1
9
0
4
190
6
1
9
0
8
1
910
191
2
1
9
1
4
1916 1
9
1
8
1
9
2
0
1922 1
9
2
4
1
9
2
6
1928 1
9
3
0
Birth Year
L
o
g
a
ri
th
m
i
c I
n
co
m
e
 
i
n
 Y
ea 
$2000
Black Veteran 
Black Non-Veteran 
White Veteran 
White Non-Veteran 
 When one looks at income disaggregated by age category, one finds that for each 
age category veterans have significantly higher incomes than non-veterans (see table 
6.8). Moreover, whites have significantly and substantively higher incomes than blacks, 
black veterans have significantly and substantively higher incomes than black non-
veterans, and white veterans have significantly and substantively higher incomes that 
white non-veterans. 
157 
Table 6.8: 1970 Income Significance Test by Age Categories  
 
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
40-44 24,122 10.55086 0.699201 *** 24,907 10.4261 0.748884
45-54 66,058 10.51083 0.723398 *** 26,094 10.28223 0.845062
55-64 18,220 10.34127 0.820937 *** 46,196 10.22614 0.876816
65-70 1,139 9.70124 1.229982
***
13,599 9.488688 1.220936
Tests of 
Significance 
White
Age Group N Mean SD p N Mean SD
40-44 4564 9.972364 0.807306 *** 44465 10.54035 0.697658
45-54 8039 9.888424 0.85208 *** 84113 10.4994 0.736332
55-64 5387 9.713218 0.930948 *** 59029 10.30848 0.838985
65-70 1259 9.089537 1.190316
***
13479 9.543932 1.218742
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
40-44 1788 10.02457 0.849062 *** 2776 9.938737 0.777529
45-54 4696 9.99864 0.788769 *** 3343 9.7336 0.911675
55-64 1305 9.904787 0.869551 *** 4082 9.651974 0.941653
65-70 81 9.510156 1.073284
***
1178 9.060615 1.192916
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
40-44 22334 10.59299 0.668123 *** 22,131 10.48723 0.722384
45-54 61362 10.55003 0.702949 *** 22,751 10.36284 0.803892
55-64 16915 10.37495 0.807346 *** 42,114 10.28179 0.84991
65-70 1058 9.71587 1.2404
***
12,421 9.529286 1.215807
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
Socio-Economic Indicator (SEI)   
 The mean Duncan SEI for the sample is 37.86521 versus 35.55713 in 1960. The 
median SEI jumped from twenty-seven in 1960 to thirty-three in 1970 representing a six 
point increase.  The increase in SEI for the preceding decade was only four points. Figure 
6.8 illustrates the veteran premium for SEI particularly for veterans of the peak 
mobilization period. In the 1960s the veteran SEI advantage began with the 1909 cohort 
and extended to the 1930 cohort. In the 1970s the advantage begins with the 1900 cohort 
158 
 
   
and extends to the 1928 cohort before the veteran and non-veteran SEIs reach parity for 
the 1929 cohort (see figure 5.11 of chapter 5 and figure 6.8 below).   
 
Figure 6.8: 1970 Veteran versus Non-Veteran Duncan Median SEI score by Birth Year 
0
5
10
15
20
25
30
35
40
45
50
19
00
1
9
0
2
1
9
0
4
19
06
19
08
19
10
19
12
1
9
1
4
19
16
19
18
19
20
19
22
1
9
2
4
19
26
19
28
19
30
Birth Year
Du
an
can
 S
E
I
Veteran SEI
Non-Veteran SEI
 
   As in the preceding decades, one would expect that the mean SEI differences 
between veterans and non-veterans, blacks and whites, and black and white veterans and 
non-veterans would be similar to the mean differences in income. In fact, the sample 
dramatically illustrates this. Veterans have a significantly and substantively higher mean 
SEI than non-veterans; whites have a significantly and substantively higher mean SEI 
than blacks, black veterans have a significantly and substantively higher mean SEI than 
black non-veterans; and white veterans have a significantly and substantively higher SEI 
than white non-veterans. These differences apply in all age groups (see table 6.9). 
 
159 
Table 6.9: 1970 Significance Tests for Duncan SEI 
 
  Veteran  
Tests of 
Significance 
Non-Veteran  
Age Group N Mean SD p N Mean SD 
40-44 24,122 42.15397 23.57186 *** 24,907 37.86943 23.97212
45-54 66,058 40.83451 23.53081 *** 26,094 33.43351 22.96095
55-64 18,220 38.56937 23.76209 *** 46,196 34.78689 22.86624
65-70 1,139 38.88323 25.15948 *** 13,599 33.75844 23.19576
                
  Black 
Tests of 
Significance 
  White  
Age Group N Mean SD p N Mean SD 
40-44 4564 24.16082 18.61505 *** 44465 41.60085 23.75915
45-54 8039 21.93046 17.08672 *** 84113 40.34526 23.5179 
55-64 5387 19.55504 16.41086 *** 59029 37.34446 23.14291
65-70 1259 17.80143 15.16796 *** 13479 35.68195 23.44228
                
  Black Veteran  
Tests of 
Significance 
Black Non-Veteran 
Age Group N Mean SD p N Mean SD 
40-44 1788 26.30425 19.6731 *** 2776 22.78026 17.76769
45-54 4696 23.93782 18.13148 *** 3343 19.11068 15.05779
55-64 1305 22.23525 18.14032 *** 4082 18.69819 15.72443
65-70 81 21.66667 17.50429 * 1178 17.53565 14.96581
                
  White Veteran 
Tests of 
Significance 
White Non-Veteran 
Age Group N Mean SD p N Mean SD 
40-44 22334 43.42285 23.39713 *** 22,131 39.76214 23.97956
45-54 61362 42.1276 23.39655 *** 22,751 35.53809 23.16879
55-64 16915 39.82956 23.67822 *** 42,114 36.34632 22.84871
65-70 1058 40.20132 25.17665 *** 12,421 35.297 23.24909
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
  
One might recall that in the 1960 census black veterans enjoyed a greater 
premium than white veterans relative to their non-veteran peers.  However, in the 1970 
census the net advantage for white veterans (5.219) is higher than the net advantage for 
black veterans (4.475). Therefore, in the 1970 census white veterans enjoy not only a 
significantly higher mean SEI than their non-veteran peers, but more of net advantage 
160 
 
   
compared to black veterans and their non-veteran peers. Figure 6.9 illustrates the 
differences for black and white veterans versus their non-veteran peers. It also highlights 
the peak mobilization period premium as well as the demobilization period penalty. 
Finally figure 6.9 illustrates how veteran status may have been an equalizer for blacks 
relative to whites in occupational status attainment.  
 
Figure 6.9: 1970 Black and White Veteran versus Non-Veteran Duncan Median SEI score by Birth 
Year 
0
5
10
15
20
25
30
35
40
45
50
19
00
1
9
0
2
1
9
0
4
19
06
19
08
19
10
19
12
1
9
1
4
19
16
19
18
19
20
19
22
1
9
2
4
1
9
2
6
19
28
19
30
Birth Year
SEI 
Black Non-Veteran 
Black Veteran
White Non-Veteran 
White Veteran
 
Models and Multivariate Regression Analysis 
 In this section I follow the same conventions used in previous chapters by making 
use of the same five multivariate regression models to control for factors associated with 
earnings, education, and Duncan SEI outcomes to determine the net premium or penalty 
to veterans and non-veterans.  
 
161 
 
   
Education Models and Regression Analysis 
 
Model 1  
 As the bivariate descriptions above suggest, veterans tend to achieve higher 
education levels than their non-veteran peers, whites attained higher levels of education 
than blacks, and older birth cohorts generally had less education than younger cohorts 
(see tables 6.6 and 6.7 above).   Model 1 generally confirms the descriptions above (see 
table 6.10). When education attainment is regressed on World War II veteran status, age
2
, 
and race we see that race is the strongest predictor (Beta = -.1692), followed by age (B=-
.1572), and then veteran status (B=.1189). This is different than in the 1960 census when 
the strongest predictor was age, followed by race then veteran status.  
  For veterans the predicted educational attainment is higher than for non-veterans 
(b=.2695) and the coefficient is very close to the unstandardized coefficient in 1960 
(b=.2613).  The coefficient for age is .1572 and is second in magnitude to race. This 
indicates that older respondents had less education than younger respondents and that 
black respondents had significantly less education than whites. When education 
attainment is regressed on World War II veteran status, age
2
, and race we see that the 
model explains nine percent of the variance. 
 
Model 2  
 When marital status and region of residence background variables are added in 
model 2 the veteran relative to non-veteran advantage is still large and significant, while 
the black race penalty decreases slightly but is still large and significant. Both married 
and single respondents have significant educational attainment premiums relative to 
162 
 
   
unmarried and non-single respondents respectively net of the effects of the other 
variables. However, the magnitudes of the coefficients for the marital variables were 
higher in the 1960?s. Those residing in the both the North and the South paid an 
educational attainment penalty relative to those not living in those regions respectively. 
Interestingly, if the model included the West as a variable those in the West would have 
received a premium. As in previous decades the penalty was much higher for those living 
in the South than those in the North.   
 
Model 3  
 With the addition of the mobilization phase variables in model 3 the highest 
premiums for educational attainment come from World War II veteran status and marital 
status.  The mobilization phase additions also serve to increase the veteran educational 
premium and increase the educational penalty for being black. The other variables were 
for the most part unaffected.  The coefficient for mobilization phase 2 (peak period) was 
not significant; however, those that were a part of the demobilization phase received a 
small educational premium (B=.0336). Those who were a part of mobilization phase 3 in 
1960 also received an educational attainment premium (unstandardized coefficient = 
.0975), but not as great as those in the 1970s (unstandardized coefficient = .0994). 
Finally, model 3 shows that black veterans received a significant educational premium in 
1970 (unstandardized coefficient = .0521) versus an educational premium in 1960 
(unstandardized coefficient = .0772) representing a decrease in the black veteran 
advantage over the decade controlling for World War II Veteran status, age race, marital 
status, geographic region, and mobilization phase. The addition of the mobilization phase 
variables in this model increases the adjusted R-square from .0922 to .0927. 
163 
 
   
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 6.10: 
1970 Education 
Models 
and 
Regressions 
  
164 
 
 
   
Model 4 
 Model 4 is significant overall in that the independent variables reliably predict the 
dependent variable (p<.001); however, as in the 1960s census, this model does show that 
there is an educational attainment penalty for being a part of birth cohorts with large 
proportions of veterans.  For every one unit increase in the percentage of veterans in a 
birth cohort, one would expect a .2193 (unstandardized) unit decrease in educational 
attainment versus .1895 (unstandardized) unit decrease in educational attainment in the 
1960 census, holding all other variables constant. This finding would seem to substantiate 
the regressions in model 3 which show that those in birth cohorts associated with 
mobilization phase 2 received less of educational attainment premium than those in 
mobilization phase 3. The premium for black veterans remained about the same. 
 
Model 5 
 Model 5 adds two veteran mobilization interaction terms to the regression 
equation. In Model 5 the veteran educational premium decreases slightly but is still 
significant. This model also shows that there is an education penalty for being a veteran 
associated with mobilization phase 2. This is the first decade in which the education 
penalty for mobilization phase 2 veterans presents itself. Much like the income analysis 
this is interesting in light of the fact that in the bivariate case veterans that were a part of 
the peak mobilization phase clearly have an educational advantage. The addition of these 
variables slightly decreases the ability to predict the variance in educational attainment. 
In model 4 we could predict 9.33 percent of the variance and in model 5 we can predict 
9.24 percent of the variance.  Moreover, in the 1960 census we could predict 10.26 
percent of the variance in education. 
165 
 
   
Income Models and Regression Analysis 
 
Model 1 
Table 6.10 illustrates the results of all five multivariate models for income. In 
model 1 income is regressed on World War II veteran status, age
2
, and race and all three 
of the independent variables are significant. This model and the remaining 4 all reliably 
predict the dependent variables in that the F values are all significant (p ? .001). The 
results of this analysis clearly show that age is the strongest predictor of log income with 
a beta weight of -.2580. This finding indicates that age is the prime determinant of log 
income holding race and veteran status constant. This is a change from the 1960 census 
when race was the primary determinate of income for the World War II cohort.  This is 
not surprising as some of the veterans were seventy years old in this sample.  
Race contributes second with a -.1888 coefficient followed by veteran status with 
a coefficient of .0877. In the 1960 model the unstandardized race coefficient was .6705 
versus .5751 in 1970. Veteran received a premium and had a higher unstandardized 
coefficient in the 1970 model than in the 1960 model (.1508 versus .1547). Although the 
differences are not large they are significant in each census year. 
The results of this model are generally consistent with the hypothesis that all 
things being equal veterans will earn more than their non-veteran peers. Moreover, it is 
not surprising that blacks might have gained ground in terms of income in the 1970s 
because of the passage of Civil Rights Acts in the mid-1960s. However, one might expect 
that measurable differences might not become apparent until the 1980s, when enough 
time has passed for procedure to catch up with policy.  
166 
 
   
Table 6.11: 1970 Income Models and Regressions 
 
  
 
 
167 
 
   
Model 2 
  The addition of the marital, region, and education variables decreased the 
magnitude of the World War II veteran status variable from .0877 in model 1 to .0618 in 
model 2. Although the direction of the race variable was unaffected, in that blacks still 
earned less, the magnitude decreased from -.1888 to in model 1 to -.1286 in model 2. As 
in the 1960 data model 2 produces the lowest race coefficients of any of the five income 
models.  
 Being married increased earnings more in the 1960 census than it did in the 1970 
census. The unstandardized coefficients in the 1960 and 1970 censuses were .3898 and 
.3345 respectively. Marriage was the fourth strongest variable after college graduate, age, 
and high school graduate. Furthermore being single decreases earnings income by about 
the same amount as it did in the 1960 census.  Residing in the South tends to substantially 
decrease earnings income while living in the North tends to increase income.  
The age
2
 coefficient in this model continues to be a prime determinate of the 
explained income and is second in strength after having some college. As stated in model 
1 above, age was not much of a factor in the 1960 census but is one of the most 
influential in the 1970 census. Furthermore, this is the first census period in which we see 
the effects of age remaining negative throughout all the models. 
 As expected those with some college and high school graduates earned very large 
income premiums for their education. The income premiums grew for both high school 
graduates as well as those with some college over the decade. The unstandardized 1960 
coefficients for high school graduates and those with some college were .2882 and .5243 
respectively and in 1970 they were .2684 and .5753 respectively.   
 
168 
 
   
Model 3   
The addition of the mobilization variables and the interaction term had an 
influence on World War II veteran status, age2, and race. The fact that the premium for 
being a veteran was reduced in this model from .0618 to .0509 was expected since some 
of the variance is captured in the interaction term and the previous decades of analysis 
produced the same effects. The disadvantage for race increased between models 2 and 3 
(.1296 to .1386) as it did in the 1960 census; however, it decreased over the decade. The 
unstandardized coefficient in 1960 was -.5465 versus -.4220 in 1970. This is due in large 
part to the addition of the black*veteran interaction term. Black veterans receive a 
considerably smaller premium in the 1970 census than they did in the 1960 census. The 
unstandardized coefficients in 1960 and 1970 were .1203 and .0627 in respectively. As 
stated earlier this could be due in large part to the fact that all blacks made gains during 
this time.  Additionally, being a part of a birth cohort in the midst of either the World 
War II peak mobilization period or the demobilization period yields an earnings penalty. 
The coefficients are -.1440 and -.1986 respectively. The negative mobilization phase 
coefficients reflect the overall decline in earnings during this period as opposed to the 
veteran*mobilization interaction effects (see model 5). Of all of the periods examined up 
to this point these are the largest penalties paid by respondents in these mobilization 
phases. Some of this may be accounted for by the fact that some of the World War II 
veterans were reaching retirement age. 
 
 
 
169 
Model 4  
The removal of the mobilization phase variables and the addition of the percent of 
a birth cohort that served did not substantially change the veteran status variable. The 
veteran advantage remains positive and significant. The same is true for married 
respondents, those from the North, high school graduates and, those with at least some 
college. As in model 3 those who are single, black, or from the South pay significant 
earnings penalties. The magnitudes of these variables are all very similar to those 
described in model three.  
Additionally the black veteran advantage continues in this model. Moreover, if 
one compares the black veteran advantage in the 1960s to that of the 1970s it is clear that 
the black veteran advantage decreased over the decade. This might be an indication that 
either Civil Rights benefits were not tied to service or that Civil Rights benefits had not 
been realized by 1970.   
 Furthermore, we find that those who were in cohorts with larger percentages of 
World War II veterans received a premium. Although those who were in cohorts with 
larger percentages of World War II veterans in the1960 census also received a premium it 
was half as much as they received in the 1970 census. The unstandardized coefficients 
were .0576 and .1055 in 1960 and 1970 respectively. The removal of the mobilization 
phase variables and the addition of the percent of a birth cohort that served reduced the 
amount of explained variance from 21.57 percent in model 3 to 20.88 percent of the 
variance in model 4. 
 
 
 
170 
 
   
Model 5  
The removal of the percent of birth year variable and the addition of the veteran 
and mobilization phase interaction terms in model 5 increases the veteran premium from 
.0483 to .0959. The veteran premium decreased between models in 4 and 5 in 1960 and 
the magnitude of the coefficients is much greater in the 1970 census than in the 1960 
census. 
 Most importantly this is the first time that we see a penalty for peak mobilization 
phase veterans and demobilization phase veterans.  This is interesting because in the 
bivariate case veterans that were a part of the peak mobilization phase and black veterans 
that were a part of the peak mobilization phase earned a premium for the service. 
However, when we control for background variables veterans of the peak mobilization 
phase pay an earnings penalty. This may be due to a combination of factors that are not 
controlled for in these models including unrest in the United States over the Vietnam War 
and the subsequent backlash on active service members and veterans. Alternatively it 
may be that the penalties that applied to the overall mobilization phase 2 respondents 
were extended to the veterans. The removal of the percent of birth year variable and the 
addition of the veteran and mobilization phase interaction terms in model 5 nominally 
increases the explained variance from 20.88 percent to 20.97 percent. This is consistent 
with change in the amount of explained variance from the 1950to to the 1960s when the 
change was 19.49 percent to 19.52 percent. 
 
 
 
 
 
 
Duncan SEI Models and Regression Analysis 
 
171 
 
   
 
 The descriptive section above showed very clearly that veterans generally have 
higher SEIs than non-veterans, whites generally have higher SEIs than blacks, and that 
both black and white veterans generally have higher SEIs than their non-veteran peers 
These relationships held for the 1950 through the 1970 Census. The results of the SEI 
regression models are shown in table 6.12.  
 
Model 1 
 In model 1, I regressed SEI on World War II veteran status, age
2
, and race. Race 
is the strongest predictor, followed by veteran status, then age
2
.  More specifically, we 
find that World War II veterans receive an SEI premium for their service (B=.1017) and 
that older respondents (B=-.0529) as well as blacks (B=-.2095) paid an SEI penalty. 
These coefficients are consistent with the SEI regressions in the 1960s analysis although 
the age
2
 coefficient is higher in the 1960 census (b=-.0004) than in the 1970 census (b=-
.0002). This model explains six percent of the variance in SEI. 
 
Model 2 
 The addition of marital, regional, and educational background variables to the 
regression equation decreases the veteran status variable from .1017 to .0437. Although 
this represents a decrease from model one to model two the veteran premium remains. 
However, much like in the 1960 census, the strength of this predictor pales in comparison 
to the education and race predictors. Those with some college or a high school degree 
continue to earn very large SEI premiums.  
172 
 
   
 The penalty for being black decreases when one controls for marital status, 
region, and education; however, the coefficient remains large and negative. Married 
respondents receive an SEI premium of almost four SEI points, which is the same amount 
of the premium in 1960.  
 The magnitude of the age
2
 coefficient changes from negative to positive and gets 
larger as a result of controlling for background factors. In 1960 this model produced the 
same effects with regard to age
2
. Residing in the North provides an increase on the SEI 
scale (B=.0345) versus not being from the North as does residing in the South (B=.0208) 
versus not being from the South.   
 
Model 3  
 In this model the addition of the race*veteran interaction term and the 
mobilization phase variables does not affect the adjusted R-squared of the model.  
However there is a modest increase in the veteran SEI premium. The age
2
 variable 
increased marginally and also remains significant. There was little change in the 
magnitudes of the married or the education variables. Mobilization phase 2 produced 
negative coefficients indicating that respondents associated with the peak mobilization 
cohorts received an SEI penalty. However, respondents associated with demobilization 
phase cohorts received a premium. These findings are not surprising in light of the data 
presented in figure 6.10 and the fact that the data presented itself in the same way in 
1960.  
 
 
 
173 
 
   
 Model 4  
 In this model, I once again remove the mobilization phase variables and add in a 
variable that controls for the percent of veterans in a particular birth cohort. Making these 
changes in the model did not affect most of the coefficients; however, the veteran and 
married premiums decreased slightly. Furthermore, the coefficient for the percent of 
veterans in a birth cohort is  
-.0169. Essentially, this tells us that for every one standard deviation increase in the 
percent of veterans in a birth cohort we can expect a .0169 standard deviation decrease in 
SEI score.  This penalty is smaller than the 1960 penalty. The race*veteran interaction 
term was not significant in this model. 
 
Model 5  
  In the final SEI model I removed the percent of veterans in a birth year variable 
and added in the veteran*mobilization phase interaction terms. Making these changes 
makes no difference in the adjusted R-squared.  The coefficient for World War II veteran 
status decreases from .0503 to .0252 and the coefficient for age
2
 decreases slightly. The 
coefficients for race, married, North and South residence, high school graduate and some 
college remained virtually the same in magnitude, direction, and significance, as they did 
in 1960. 
 Although there was an SEI penalty for being a part of the peak mobilization phase 
in model 3, there is a premium for being a veteran in the peak mobilization phase in this 
model.  Moreover, this becomes even more interesting in light of the fact that veterans in 
the peak phase of mobilization received penalties in terms of income and education. The 
coefficient for this variable is .0221, which suggests that veterans for the peak World 
174 
 
   
War II mobilization period received a statistically significant premium of an SEI point for 
their service. Veterans who were a part of the demobilization phase received also 
received an SEI premium (B= .0199) for their service. This is consistent with the 
premium that they received in 1960.  
175 
 
   
 
Table 6.12: 1970 Duncan SEI Models and Regressions 
176 
 
   
   
Summary 
 
 In general the analysis revealed that incomes generally went up. However those in 
cohorts older than 1906 began to suffer decreases in earnings. The veteran earnings 
premium remained in effect in this decade for all three dependent variables, although it 
was not as salient as it was in the 1960s. Therefore, hypothesis one is supported for the 
1970s census.  
 Furthermore, both black and white veterans held income and education 
advantages over their non-veteran peers and the difference in the black veteran advantage 
was greater. Although black veterans earned an SEI premium relative to their non-
veteran peers in the bivariate, when I controlled for background variables and other 
factors the race*veteran interaction term was negative and insignificant. Based on these 
findings we can conclude that hypothesis two generally holds with the caveat that black 
veterans earn a larger premium than white veterans only for income and education in the 
1970 census. 
Veterans of both the peak and demobilization phases of World War II earned SEI 
premiums and suffered income and education penalties in the 1970 census. In the 1950 
and 1960 censuses veterans of the peak mobilization period earned premiums in terms of 
all three dependent variables. The demobilization phase veterans earned premiums in 
terms of all three dependent variables in 1950 and only income and SEI premiums in 
1960. Based on these findings we can conclude that hypothesis three does not hold for 
the 1970 census. More specifically, veterans who were serving during the peak 
177 
 
   
mobilization period did not attain more than those serving during the beginning phase of 
World War II nor did those serving during the demobilization phase. 
 Furthermore, we find that those who were in cohorts with larger percentages of 
World War II veterans received an earnings income premium. Although those who were 
in cohorts with larger percentages of World War II veterans in the1960 census also 
received an income premium it was half as much as they received in the 1970 census. 
However, those who were in cohorts with large proportions of veterans paid education 
and SEI penalties. Therefore, we must conclude that hypothesis four hold only with 
respect to income in the 1970 census.    
178 
 
   
Chapter 7: World War II Veterans and the 1980 Census 
 
 
 The 1970s was a turbulent decade in the history of the United States. The 
Vietnam War ended, affirmative action became law, and the Cold War intensified.  One 
might argue that myriad of social changes that occurred during this decade did not 
necessarily affect World War II veterans. However, some of the social changes did in fact 
affect the military institution and its relationship to the greater society. As such, World 
War II veterans may have been directly or indirectly affected by some of the social 
changes. An example of this is Executive Order 11246 and its affirmative action 
provisions. This policy although not directed towards the World War II cohort may have 
allowed minorities to break through glass ceilings or gain employment where they 
otherwise might not have.  Furthermore, the continuation of the Cold War provided a 
number of federal jobs for which veterans received preference in hiring,  
  In the 1980 census the World War II cohort was between the ages of fifty and 
eighty and generally near the end of their primary working careers. This does not mean 
that they did not work Although many of the World War II cohort were eligible for and 
did draw retirement pay in this census it is possible and probable that some of them drew 
retirement pay and continued to work thus having earnings income from employment.  
This analysis allows us to distinguish between age and cohort effects by the use of 
aspirate age and percent of veterans in a birth year variables.  
  
 
 
 
 
179 
 
   
Descriptive data 
  
In this chapter I use the same analytical methods utilized in chapters four through 
six to describe the sample, compare the dependent measures, and examine the dependent 
variables while simultaneously making comparisons to the 1970 census. As in the 
previous chapters I shaped the data to age the World War II cohort by ten years.  After 
making all of the adjustments that were made in previous chapters I was left with a total 
sample of 220, 335, which is 91,963 respondents less than the total in 1970.    
The median age for the sample is fifty-seven years with an age range of fifty to 
eighty and all of the respondents have birth years between 1900 and 1930. The only 
exception is black veterans whose birth years range from 1903 to 1930.  As expected the 
majority of the sample is white (91.96 percent), from the south (57.82 percent), and 
married (86.08 percent).  In 1970 the sample was 91.26 percent white, 56.66 percent from 
the south, and 87.07 percent married. Moreover, the average income adjusted to year 
2000 dollars is $33,450 versus $35,725.90 in 1970. The majority of the respondents were 
high school graduates in this census versus less than a ninth-grade education in 1970 and 
had a median SEI of forty?four versus thirty-three in 1970 (see table 7.1). 
Of the 128,132 total respondents 71,104 are veterans, who represent 
approximately 53.39 percent of the sample which is an increase from the 1970 census 
(40.71 percent). Moreover, this is the first period of analysis in which veterans 
outnumber non-veterans and would lead one to believe that either the mortality rate for 
non-veterans was higher than that of veterans or that non-veterans left the labor force 
earlier than veterans in the decade between 1970 and 1980.  Figure 7.1 illustrates the 
1980 census proportion of veterans to non-veterans. Black veterans make up 
180 
 
   
approximately 5.57 percent (3,958) of the veteran population and white veterans 
comprise the other 94.43 percent (71,104). In 1970 black veterans comprised 
approximately seven percent of the veteran population.  
181 
 
   
Table 7.1: 1980 Census Descriptive Summary 
 
  
 
 
 
 
 
 
 
 
 
 
182 
 
   
  
Figure 7.1: 1980 Census Veterans versus Non-Veteran 
 
 
The majority of the veterans in the 1980 sample fall into the fifty-five to sixty-
four year old age category as do their non-veteran peers and both veterans and non-
veterans have a median age of fifty-seven. Most of the veterans are from the South (55.40 
percent) and are married (87.72 percent). This represents almost no change from the 1970 
census.  Furthermore, they have a median income of $37,630.45 versus $39,276.30 in 
1970 (adjusted to year 2000 dollars), are high school graduates (30.44 percent) or have 
some college (29.74 percent), and have a median SEI of 44.  There was little change in 
the percentage of high school graduates or SEI between 1970 and 1980; however, there 
was an increase in the percent that are college graduates. This does not necessarily mean 
183 
 
   
that more veterans went to college during the 1970s, it may be a reflection of higher 
mortality rates for less educated veterans or less educated veterans leaving the labor 
force.   
The majority of the non-veterans are from the South (59.60 percent) and married 
(84.05 percent). They have a median income of $28,225.45 (adjusted to year 2000 
dollars), average a high school education (median), and have an average SEI score of 
thirty-two (see table 1). Non-veterans, like veterans, earned a little less on average and 
higher percentages are in the upper education categories in the 1980 census relative to the 
1970 census.  
White veterans look much like the general veteran population.  Like the general 
veteran population they have a median age of fifty-seven years, are predominantly in the 
fifty-five to sixty-four year old age category (64.46 percent), are married, and reside in 
the South (88.48 percent).  Their average income adjusted to year 2000 dollars is 
$37,630.45 versus $40,163.90 in 1970, over fifty percent of them are high school 
graduates (33.25 percent) or have some college (34.50 percent), and their average SEI is 
forty-four (median) (see table 1).   
Black veterans in this sample are on average fifty-seven years old (median) and 
most of them are in the fifty-five to sixty-four year old age category (65.03 percent).  
Black veterans like their non-veteran peers are predominantly married (74.81 percent) 
and from the South (69.38 percent). They have an adjusted income of $27,180.45 versus 
$26,849.90 in 1970 representing a 1.23 percent increase. They average about a ninth 
grade education and they have an eighteen mean SEI (see table 1). It is interesting that 
black veterans raised their incomes over the decade, yet their (median) SEI score 
remained exactly the same.  White veterans? SEI scores remained the same; however, 
184 
 
   
they did not have as much room on the upper end of the scale for improvement as black 
veterans. Black non-veteran also raised their incomes over the decade from $20,192.90 to 
$20,659.65 which represents a 2.3 percent increase. Additionally, black non-veterans 
raised their (median) SEI score from 15 to 18 over the decade.   
   
Background Descriptive Statistics and Discussion 
Age   
 On average veterans are significantly younger than their nonveteran peers by less 
than a year (.9509). In the 1970 sample they were 4.76 years younger. Based on their 
entry age into World War II, veterans should be between the ages of fifty-one and sixty-
four if they entered between the ages of eighteen and twenty-six (see table 7.2).  
Table 7.2: Veteran Expected Ages in 1980 by Entry Age 
 
 White respondents have an average age of 58.72 years versus 51.73 years in 1970 
and they are significantly older than black respondents by only .1595 years versus .399 
years in 1970.  The magnitudes of the differences in mean age between veterans and non-
veterans and black and white respondents is very small and the significance is more a 
reflection of sample size than difference in mean age.  Black veterans have a mean age of 
57.93 years versus 49.19 years in 1970 and they are significantly younger than their non-
veteran peers by .1941 years, a decrease from the 3.68 year difference in 1970. Similarly 
white World War II veterans have a mean age of 57.76 years versus 49.36 years in 1970 
and they are significantly younger than their non-veteran peers by 1.04 years versus a 
185 
 
   
4.79 year difference in 1970. The median ages for all of the demographic categories in 
this study is fifty-seven. Table 7.3 provides the results of the significance tests described 
above. 
 
 
 
Table 7.3: 1980 Census Age Significance Tests 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
 The logic for using birth cohorts and the cutoff years for the establishment of 
mobilization phases remains the same as in previous chapters and figure 7.1 illustrates 
the 1980 census World War II cohort identified by mobilization phase.  
 
Region  
 Almost fifty-eight percent of the sample maintains their residence in the South, 
24.78 percent in the North, and 17.40 percent in the West. In the 1970s the sample 
resided 56.66, 27.03, and 16.31 percent in the South, North, and West respectively. The 
veteran and non?veteran proportions for region of residency are very similar. 24.46 
186 
 
   
percent of the veterans reside in the North versus 23.93 percent of the non-veterans; 
56.40 percent of the veterans and 59.60 percent of the non-veterans reside in the South; 
and 18.15 percent of the veterans versus 16.48 percent of the non-veterans live in the 
West. In the 1970 census 27.44 percent of the veterans lived in the North versus 26.62 
percent of the non-veterans; 55.31 percent of the veterans and 57.99 percent of the non-
veterans lived in the South; and 17.24 percent of the veterans versus 15.39 percent of the 
non-veterans lived in the West.  
  I conducted a significance test assuming unequal variance combining all of the 
regions and tested whether there was a significant difference between veterans and non-
veterans. Moreover, the proportion significance tests that disaggregate each demographic 
category by region are all significant with the exception of the test that compares the 
proportion of black veterans to non-veterans in the North (see table 7.4).  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
187 
 
   
 
 
 
Table 7.4: 1980 Region Significance Tests 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
  
 The results of these tests provide evidence that veterans in general, as well as black and 
white veterans, are significantly more likely to live in the North and West than their non-
veteran peers and that non-veterans are significantly more likely to live in the South. The 
difference in the proportion of black veterans relative to black non-veterans living in the 
North is not statistically significant.  
 
Marital Status 
188 
 
   
 Married respondents represent approximately 86.08 of the sample versus 87.07 of 
the 1970 sample. The highest percentage of married respondents was 88.06 percent in 
1960.  As in the 1970 census, whites are significantly more likely than blacks to be 
married. However, blacks are significantly more likely to be divorced, widowed, or 
single than whites (see table 7.5).  
 As in the 1970 data veterans are significantly more likely to be married or 
divorced than non-veterans; however, non-veterans are significantly more likely to be 
widowed than their veteran peers. The proportions of married, divorced, widowed, and 
single veterans and non-veterans were stable over the decade. The proportion of married 
veterans increased from 87.07 percent in 1970 to 87.72 percent in 1980, while the 
proportions for non-veterans decreased from 88.65 percent to 84.05 percent over thee 
same period.  
  Black veterans are significantly more likely to be married than their non-veteran 
peers and significantly less likely to be widowed or single than their non-veteran peers. 
This is a substantial change from the 1970?s when black veterans were significantly more 
likely to be divorced or single and significantly less likely to be married or widowed than 
their non-veteran peers. White veterans are significantly more likely to be married, than 
their non-veteran peers and significantly less like to be widowed or single than white 
non-veterans (see table 7.5). In the 1970 census white veterans were significantly more 
likely to be married, divorced, or widowed than their non-veteran peers and significantly 
less likely to be widowed than white non-veterans. 
 
 
 
189 
 
   
 
 
Table 7.5: 1980 Marital Status Significance Tests 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
 
Education 
 In every age and demographic category there was an increase in education 
between 1970 and 1980.  For example comparisons of the veteran forty-five to fifty-four 
year old age group in 1970 should look similar to the veteran fifty-five to sixty-four year 
old age group in 1980.  However, we find that in 1970 this group had a mean education 
190 
 
   
of 6.166 and in 1980 this level increased to 6.519. There are at least two explanations for 
this increase. The first is that this particular group continued to formally educate itself 
over the life course and the second is that more of the less educated respondents died or 
left the labor force during the decade increasing the overall average. 
 The average person in the sample attained an eleventh grade education level 
(6.203) which is about the same as in 1970. Figure 7.2 illustrates educational attainment 
for the entire sample by birth year. Black respondents attained a little more than a ninth 
grade education (4.856) and whites attained a little more than an eleventh grade 
education (6.321). Both of these are slight increases from 1970 when black respondents 
attained slightly more than a ninth grade education and whites attained more than a tenth 
grade education.  
 
Figure 7.2: 1980 Cohort Education Levels 
0
500
1000
1500
2000
2500
3000
3500
1
900
19
02
19
04
19
06
19
08
1
910
19
12
19
14
19
16
19
18
19
20
19
22
19
24
1
9
26
1
928
19
30
Birth Year
Nu
mb
e
r
Grade School
Ninth to Eleventh 
 High School Graduate
 Some College
 
191 
 
   
 When the data are disaggregated by veteran status, race, and age categories we 
find that there are statistically significant differences but no substantive differences in 
these categories. Veterans significantly attained almost a full grade of education more 
than non-veterans. Both black and white veterans achieved a grade more education than 
their non-veteran peers with mean differences being significant at the .001 level (see 
table 7.6), although the differences are not substantively significant in two categories. 
These two categories include the 55 ? 64 year old age categories for veterans versus non 
veterans and 55 ? 64 age category for black veterans versus non veterans (see table 7.7).  
The difference in the means is greater for black veterans versus black non-veterans (.930) 
than for white veterans versus non-veterans (.697). The differences were 1.005 and .791 
respectively in 1970. Compared to 1970 the black veteran educational premium 
decreased slightly while the white veteran premium increased slightly. 
Table 7.6: 1980 Education Significance Tests 
Tests of 
Significance 
N Mean SD p NMeanSD
Aggregate 71,104 6.555201 2.014522
***
57,268 5.766432 2.341211
Tests of 
Significance 
White
N Mean SD p NMeanSD
Aggregate 10326 4.856576 2.311747
***
118,046 6.321129 2.151929
Tests of 
Significance 
N Mean SD p NMeanSD
Aggregate 3,958 5.430268 2.24595
***
6,368 4.5 2.280268
Tests of 
Significance 
N Mean SD p NMeanSD
Aggregate 67,146 6.621511 1.980217
***
50,900 5.924872 2.300184
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
 
 
192 
 
   
 
 
 
 
 
 
 
 
Table 7.7: 1980 Education Significance Test by Age Category 
 
Tests of 
Significance 
Age
50-
 Group N Mean SD p N Mean SD
54 19,514 6.719381 1.939976 *** 23,062 6.206097 2.240044
 55-64 45,859 6.519549 2.005636 *** 20,229 5.511741 2.340043
 65-74 5,507 6.293263 2.265736 *** 11,757 5.45828 2.372915
 75-80 224 5.991071 2.464076
***
2,220 5.151802 2.466161
Tests of 
Significance 
White
Age Group N Mean SD p N Mean SD
50-54 3545 5.322426 2.262566 *** 39031 6.542979 2.08058
 55-64 5220 4.778161 2.283387 *** 60868 6.333952 2.108656
 65-74 1376 4.06686 2.258535 *** 15888 5.868202 2.325986
 75-80 185 4.016216 2.355557
***
2259 5.328021 2.461208
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
50-54 1024 5.914063 2.115538 *** 2521 5.08211 2.276519
 55-64 2574 5.327117 2.246435 *** 2646 4.244142 2.190915
 65-74 350 4.774286 2.332755 *** 1026 3.825536 2.181915
 75-80 10 5.4 2.988868
NS
175 3.937143 2.299811
Tests of 
Significance 
Age Group N Mean SD p N Mean SD
50-54 18490 6.763981 1.920002 *** 20,541 6.344044 2.196302
 55-64 43285 6.590459 1.967795 *** 17,583 5.702497 2.302076
 65-74 5157 6.396354 2.22406 *** 10,731 5.614388 2.331324
 75-80 214 6.018692 2.441739
***
2,045 5.255746 2.452608
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
  
 
  
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
Figure 7.3 illustrates the educational attainment of veterans and non-veterans by 
birth year. As described above, veterans attained significantly more education than non-
veterans. Additionally, this figure dramatically illustrates the veteran premium for 
serving during the peak mobilization period of World War II. Figures 7.4 and 7.5 display 
the net education advantage for veterans, both black and white, as well as the peak 
mobilization phase advantage. These figures also display the net education disadvantage 
for veterans in the demobilization phase.  
193 
 
   
 
 
 
 
 
 
       Figure 7.3: Veteran versus Non-Veteran Education  
             (College and High School Graduates Only) 
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1
900
1
9
0
2
1
9
0
4
1
906
1
9
0
8
1
9
1
0
1
912
1
914
1
9
1
6
1
918
1
920
1
9
2
2
1
9
2
4
1
926
1
9
2
8
19
3
0
Birth Year 
Perc
en
t
 V
e
t
e
ra
ns
 an
d No
n-
Ve
t
e
ran
s
 Non-Veteran High School Graduate
Non-Veteran  Some College
Veteran  High School Graduate
 Veteran Some College
 
         Figure 7.4: Black Veteran versus Black Non-Veteran Education  
             (College and 
High School Graduates 
Only) 
70.00%
 
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
1
9
0
0
19
0
2
1
9
04
19
0
6
19
0
8
19
1
0
1
9
12
19
1
4
19
1
6
1
9
18
19
2
0
19
2
2
19
2
4
1
9
26
19
2
8
19
3
0
Birth Year
Percen
t
 B
l
ack 
V
e
teran
s a
n
d
 N
o
n
-
V
e
t
eran
s
 
  
Percent Black Non-Veteran High School
Graduate
Percent Black Non-Veteran Some College
Percent Black Veteran High School Graduate
 Percent Black Veteran Some College
 
 
194 
 
   
  
 
Figure 7.5: White Veteran versus White Non-Veteran Education  
      (College and High School Graduates Only) 
 
 
 
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
40.00%
45.00%
50.00%
19
0
0
19
0
2
1
9
04
19
0
6
19
0
8
19
1
0
1
9
1
2
19
1
4
19
1
6
1
9
18
1
9
2
0
19
2
2
19
2
4
1
9
2
6
19
2
8
19
3
0
Birth Year
P
r
ecen
t
 W
h
i
t
e 
V
e
t
e
r
a
n
s
 an
d
 
N
o
n
-
V
e
t
e
r
a
n
s
Percent  White Non-Veteran High School
Graduate
 Percent White Non-Veteran Some College
 Percent White Veteran High School Graduate 
Percent  White Veteran Some College 
 
 
 
 
 
 
 
 
 
 
 
 
There is a veteran advantage in the aggregate as well as when the group is 
disaggregated by race. Veterans in the aggregate receive a .789 educational unit 
advantage for their service which is less than the .853 that they received in 1970. Black 
veterans enjoyed a .930 mean difference premium over their non-veteran peers and white 
veterans enjoyed a .697 educational unit advantage over their non-veteran peers. These 
differences are not substantively significant; however, they are smaller than the 
differences 1970 differences which were 1.005 and .791 for black and white veterans 
respectively.   
195 
 
   
 
 
 Income  
 The average (median) income in the sample is $10.41 log dollars ($33,450.45)
5
 
which is less than the 1970 census median income of $10.48 log dollars ($35,725.90). 
Whereas incomes for veterans and non-veterans as well as blacks and whites rose 
between the 1960 and 1970 censuses, in the 1980 census incomes decreased for white 
veterans and increased for blacks. The decrease in earnings income was somewhat 
expected, because many of the World War II veterans would be moving from work to 
retirement after age 65. My assumption is that there are many in the sample who have 
retired, but have taken up lower paying jobs to supplement their incomes, thus keeping 
them in the sample.  Figure 7.6 illustrates the median log income for veterans versus non-
veterans by birth year.  
 In the 1970s we saw younger veteran cohorts earning more than older veteran 
cohorts and this trend continues in the 1980s. Those born in the 1915 cohort are 64 years 
old in the 1980 census and we can see lower earnings income for cohort born before them 
and higher for those born after. Figure 3 also illustrates the fact that the veteran 
advantage for the peak mobilization phase continues in 1980; however, it is not nearly as 
salient as it was in the 1950, 1960, and 1970 censuses.  
 
 
 
 
 
 
 
                                                 
5
 All income figures are expressed in log dollars. Log dollars were computed by taking 
the natural logarithm of income adjusted in year 2000 dollars. The inflation factor was 
computed by multiplying the 1980 dollar amount by a factor of 2.09.   
196 
 
   
 
 
 
 
Figure 7.6: 1980 Veteran versus Non-Veteran Log Income 
7
7.5
8
8.5
9
9.5
10
10.5
11
190
0
190
1
190
2
190
3
190
4
190
5
190
6
190
7
190
8
190
9
191
0
191
1
191
2
191
3
191
4
191
5
191
6
191
7
191
8
191
9
192
0
192
1
192
2
192
3
192
4
192
5
192
6
192
7
192
8
192
9
193
0
Birth Year
Lo
g In
com
e
Non-Veteran
Veteran 
 
 Veterans in the sample have significantly and substantively higher mean incomes 
of $10.31 (ln) than their non-veterans peers who have mean incomes of $9.95 (ln). In 
year 2000 dollars this equates to a difference of more than $8,000. In 1970 veterans had 
log incomes of $10.36 (ln) versus $10.07 (ln) for their non-veterans peers. Blacks have a 
mean income of $ 9.77 (ln) that is significantly and substantively less than whites who 
have a mean income of $10.18325 (ln). In 1970 blacks had a mean income of $9.81 (ln) 
that was significantly less than whites who had a mean income of $10.39 (ln).  
Additionally, both black and white veterans have significantly and substantively higher 
incomes than their non-veteran peers (see figure 7.7 and table 7.8). For black veterans 
this difference is more than $5,800 and for white veterans this difference is more than 
197 
 
   
$7,500. Figure 7.7 also shows that in several early cohorts black veterans? incomes 
surpassed those of white veterans and non-veterans alike.    
 
 In the 1970s the difference in the mean log incomes between black veterans and 
black non-veterans was .2992489.  In 1980 the difference grew to .3227736. In 1970 
white veterans and non-veterans had a significant difference of .2696017 in mean log 
income and the difference grew to .3518936 in 1980.  The difference in log income 
between black veterans and non-veterans is less than the difference between white 
veterans and non-veterans in the 1980s.  This is different than in the 1970s and earlier 
periods when the difference in log income between black veterans and non-veterans was 
greater than the difference between white veterans and non veterans.  
Figure 7.7: 1980 Black and White Veteran versus Non-Veteran Log Income 
7
7.5
8
8.5
9
9.5
10
10.5
11
19
00
19
02
19
04
19
06
1
908
19
10
19
12
1
9
14
19
16
19
18
1
920
19
22
19
24
19
26
19
28
19
30
Birth Year
Lo
g In
com
e
 in
 2000 
Do
llars
White Non-Veterans
White Veterans
Black Non-Veterans
Black Veterans
 
 When one looks at income disaggregated by age category, one finds that for each 
age category veterans have significantly and substantively (every difference is over 
198 
 
   
$3,500) higher incomes than non-veterans (see table 7.6). Moreover, whites have 
significantly higher incomes than blacks, black veterans have significantly and 
substantively higher incomes than black nonveterans, and white veterans have 
significantly higher incomes that white nonveterans. 
 
 
Table 7.8: 1980 Income Significance Test by Age Categories  
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001 
 
199 
 
   
Socio-Economic Indicator (SEI)   
 The mean Duncan SEI for the sample is 40.81 versus 37.87 in 1970. The median 
SEI jumped from thirty-three in 1970 to forty-four in 1980 representing an eleven point 
increase.  The increase in SEI for the preceding decade was only six points. Figure 7.8 
illustrates the veteran premium for SEI particularly for veterans of not only the peak 
mobilization period, but mobilization phase 1 as well. In the 1970s the SEI advantage 
began with the 1900 cohort and extends to the 1928 cohort before the veteran and non-
veteran SEIs reach parity for the 1929 cohort. In the 1980 data every veteran cohort with 
exception of the 1903 and 1929 cohorts have an SEI advantage (see figure 6.8 of chapter 
6 and figure 7.8 below).   
 
Figure 7.8: 1980 Veteran versus Non-Veteran Duncan Median SEI score by Birth Year 
0
5
10
15
20
25
30
35
40
45
50
190
0
190
1
190
2
190
3
190
4
190
5
190
6
190
7
190
8
190
9
191
0
191
1
191
2
191
3
191
4
191
5
191
6
191
7
191
8
191
9
192
0
192
1
192
2
192
3
192
4
192
5
192
6
192
7
192
8
192
9
193
0
Birth Year
D
u
cnan
 SE
I
Non-Veteran 
Veteran
 
   As in the preceding decades, one would expect that the mean SEI differences 
between veterans and non-veterans, blacks and whites, and black and white veterans and 
200 
 
   
non-veterans would be similar to the mean differences in income. As in the 1950s 
through the 1970s, the sample dramatically illustrates this. Veterans have a significantly 
higher mean SEI than non-veterans; whites have a significantly higher mean SEI than 
blacks, black veterans have a significantly higher mean SEI than black non-veterans; and 
white veterans have a significantly and substantively higher SEI than white non-veterans. 
These differences apply in all age groups (see table 7.9). 
201 
 
   
 
Table 7.9: 1980 Significance Tests for Duncan SEI 
  Veteran  
Tests of 
Significance 
Non-Veteran  
Age Group N Mean SD p N Mean SD 
50-54 19,514 44.84616 23.95109 *** 23,062 40.21347 24.66804
 55-64 45,859 43.16921 24.14835 *** 20,229 35.26734 23.93919
 65-74 5,507 41.9684 25.23089 *** 11,757 36.02815 24.29677
 75-80 224 41.44643 25.9839 ** 2,220 36.21261 24.39276
                
  Black 
Tests of 
Significance 
  White  
Age Group N Mean SD p N Mean SD 
50-54 3545 27.33061 20.95291 *** 39031 43.69973 24.28944
 55-64 5220 25.13238 20.09756 *** 60868 42.08992 24.22509
 65-74 1376 21.9891 18.7471 *** 15888 39.303 24.72807
 75-80 185 19.14595 16.50149 *** 2259 38.12926 24.5845 
                
  Black Veteran  
Tests of 
Significance 
Black Non-Veteran 
Age Group N Mean SD p N Mean SD 
50-54 1024 30.65918 21.74827 *** 2521 25.97858 20.47146
 55-64 2574 28.01593 21.29734 *** 2646 22.32729 18.43335
 65-74 350 24.64286 19.94594 ** 1026 21.08382 18.24219
 75-80 10 31.5 24.33219 NS 175 18.44 15.74968
                
  White Veteran 
Tests of 
Significance 
White Non-Veteran 
Age Group N Mean SD p N Mean SD 
50-54 18490 45.63186 23.82208 *** 20,541 41.96052 24.57356
 55-64 43285 44.07032 24.00815 *** 17,583 37.21464 24.06703
 65-74 5157 43.14427 25.12216 *** 10,731 37.45699 24.32272
 75-80 214 41.91121 26.01889 * 2,045 37.7335 24.40237
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001 
 
  
 In the 1970 census the net advantage for white veterans (5.219) was higher than 
the net advantage for black veterans (4.47).  This trend continued in the 1980s with the 
net advantage for white veterans increasing to 5.22 and the net advantage for black 
veterans increasing to 4.94. However, the percent change in the difference was higher for 
202 
 
   
black veterans (10.5 percent versus .027 percent). Figure 7.9 illustrates the differences for 
black and white veterans versus their non-veteran peers. It makes it quite apparent that 
peak mobilization premium for veterans applied mostly to white veterans. Furthermore 
unlike white veterans, black veterans enjoyed SEI premiums in both beginning and 
demobilization phases. Finally, figure 7.9 also illustrates how veteran status may have 
served to decrease the SEI gap for blacks relative to whites in the early cohorts.  
 
Figure 7.9: 1980 Black and White Veteran Versus Non-Veteran Duncan Median SEI score by Birth 
Year 
0
5
10
15
20
25
30
35
40
45
50
1
90
0
1
9
02
1
90
4
19
06
1
90
8
1
91
0
1
9
12
1
91
4
19
16
1
91
8
1
92
0
1
9
22
1
92
4
19
26
1
92
8
1
93
0
Birth Year
D
u
ncan
 SE
I
Black Nonveteran SEI
Black Veteran SEI
White Nonveteran SEI
White Veteran SEI
 
Models and Multivariate Regression Analysis 
 In this section I follow the same conventions used in previous chapters by making 
use of the same five multivariate regression models to control for factors associated with 
earnings, education, and Duncan SEI outcomes to determine the net premium or penalty 
to veterans and non-veterans. 
203 
 
   
  
Education Models and Regression Analysis 
 
Model 1  
 As the bivariate descriptions above suggest, veterans tend to achieve higher 
education levels than their non-veteran peers, whites attained higher levels of education 
than blacks, and older birth cohorts generally had less education than younger cohorts 
(see tables 7.6 and 7.7 above).   Model 1 generally confirms the descriptions above (see 
table 7.10). When education attainment is regressed on World War II veteran status, age
2
, 
and race we see that race has the most strength holding age
2
 and veteran status constant. 
Blacks pay a considerable penalty for the color of their skin, which is consistent with 
what we saw in the 1970s census. In fact, the penalty for blacks decreased over the 
decade from -.6790187 (unstandardized) in the 1970 census to -.6286 (unstandardized) in 
the 1980 census.  
 We also find that for veterans the predicted educational attainment is higher than 
for non-veterans. The 1970 census model 2 yielded a veteran status coefficient of .2695 
(unstandardized) versus .3188 (unstandardized) in the 1980 census. As in 1970, the 
standardized coefficient for age
2
 is both negative and small when compared to veteran 
status and race. However, it is significant and negative indicating that older respondents 
had less education and that in any prediction model using model 2 variables education 
will tend to decrease with increased age. When education attainment is regressed on 
World War II veteran status, age
2
, and race we see that the model explains six percent of 
the variance. The same model in 1970 explained approximately nine percent of the 
variance.  
204 
 
   
 
Table 7.10: 1980 Census Education Models and Regressions 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
205 
 
   
 
Model 2  
 As in previous analysis periods model 2 adds marital status and region of 
residence background variables. The results of theses additions are that the veteran 
relative to non-veteran advantage is still large and significant (B=.1395), while the race 
penalty decreases slightly from -.1534 (unstandardized) to -.1399 (unstandardized); 
however, it is still large, significant, and has more strength than any of the other variables 
in the model. As they did in the 1970s both married and single respondents have 
significant educational attainment premiums relative to unmarried and non-single 
respondents respectively net of the effects of the other variables. Those residing in the 
North paid an educational penalty (B=-.0773) that was about half as much as those living 
in the South (B=.1426). As in previous decades the penalty was much higher for those 
living in the South than those in the North.  The addition of marital status and region of 
residence background variables in model 2 modestly increases the proportion of variance 
in education that can be predicted from the independent variables from 6.31 to 7.51 
percent versus eight to 9.22 percent in 1970. 
 
Model 3  
 The highest premiums for educational attainment in this model come from World 
War II veteran status and marital status.  The mobilization phase additions also serve to 
increase the veteran educational coefficient from .1395 to .1539 and increase the 
educational penalty for being black from -.1399 to -.1457. Veteran status and race are the 
two strongest predictors of education in this model followed by residence in the South 
(B=-.1429). The other variables were for the most part unaffected. 
206 
 
   
  The coefficient for mobilization phase 2 (peak period) is significant and negative. Those 
who were a part of the peak mobilization phase, both veteran and non-veteran, paid an 
education penalty (B=-.0385). In the 1970s model the peak mobilization variable was not 
significant. Finally, model 3 shows that black veterans received a significant .0777 
(unstandardized) educational unit premium in 1980 versus a .0521 (unstandardized) 
educational unit premium in 1970 representing an increase in the black veteran advantage 
over the decade controlling for World War II Veteran status, age race, marital status, 
geographic region, and mobilization phase. Although the black veteran premium is 
significant it is one of the weakest variables in the model (B=.0120). The addition of the 
mobilization phase variables in model 3 increases the adjusted R-square from .0752 to 
.0775 which is consistent with the small change (.0922 to .0927) from model 2 to model 
3 in 1970. 
  
Model 4 
 As in the 1970s census, this model shows that there is a substantial educational 
attainment penalty for being a part of birth cohorts with large proportions of veterans.  
This was expected because in model three we saw that mobilization period 2 produced 
negative coefficients.  For every one unit standard deviation increase in the percentage of 
veterans in a birth cohort, one would expect a .0788 unit decrease in educational 
attainment, holding all other variables constant.  The premium for veterans grew with 
addition of the percent of veterans in a birth cohort variable from .1539 to .1715 between 
models 3 and 4. This also represents a substantial increase over the decade as well .3846 
(unstandardized) versus .2969 (unstandardized). 
207 
 
   
 The premium for black veterans decreased marginally from .0120 to .0098, which 
is consistent with change between model 3 and model 4 in the 1970 census.  
 
Model 5 
 Model 5 adds two veteran mobilization interaction terms to the regression 
equation. The premium for being a veteran in this model remains large and significant (B 
= .1783). In fact, veteran status is the strongest predictor of education in this model 
followed by race (B = -.1486), age
2
 (B = -.1152). 
  As in 1970 this model also shows that there is an education penalty for being a 
veteran associated with mobilization phase 2. This is the second decade in which the 
education penalty for mobilization phase 2 veterans presents itself. Much like the income 
analysis this is interesting in light of the fact that in the bivariate case veterans that were a 
part of the peak mobilization phase clearly have an educational advantage. However, this 
finding is consistent with the negative percent veteran effect. The demobilization phase 
veterans pay an educational penalty which is less than the peak mobilization phase 
veteran penalty.     
 Black veterans continue to earn a premium for their service and the coefficient 
increases form .0630876 to .0868941. Black veterans earned educational premiums in 
every education model in the 1980s.   The addition of these variables slightly decreases 
the ability to predict the variance in educational attainment. In model 4 we could predict 
7.99 percent of the variance and in model 5 we can predict 7.55 percent of the variance in 
education. This compares with being able to predict 9.33 percent of the variance in model 
4 and 9.24 percent of the variance in model 5 in 1970. 
 
208 
 
   
Income Models and Regression Analysis 
 
Model 1 
Table 7.11 illustrates the results of all five multivariate models for income. In 
model 1 income is regressed on World War II veteran status, age
2
, and race and we see 
that all three of the independent variables are significant.  
The results of this analysis clearly show that age is the strongest predictor of log 
income with a beta weight of -.2580. This finding indicates that age is the prime 
determinant of log income holding race and veteran status constant. As in other periods 
of analysis the beta weight for race is second in strength (b= -.1162), indicating that 
blacks pay a substantial penalty in log income compared to whites. Furthermore, for a 
standard deviation increase in race (.2719), the expected change in log income equals 
.1162
 
standard deviations, holding all other variables constant. We also see that veteran 
status as in other census periods has a positive effect on log income after holding race 
and age constant.  These findings are consistent with the 1970 census results, at least in 
terms of the strength of the predictors, where the unstandardized coefficients for veteran 
status, age, and race were .15077, -.0002283, and -.5751 respectively versus .2708, -
.0005, and -.3727 respectively.  However, in 1980 veteran status provides more of an 
advantage for log income than in 1970 as opposed to race which provided more of 
disadvantage in the 1970 census. The age coefficient did not change much in magnitude 
across census periods and the direction is negative throughout both the 1970 and 1980 
censuses.   
209 
 
   
Furthermore, this model explains 16.67 percent of the variance in earnings versus 
twelve percent in the same model in 1970. This model and the remaining 4 all reliably 
predict the dependent variables in that the F values are all significant (p ? .001).  
210 
 
   
Table 7.11: 1980 Census Income Models and Regressions 
 
 
 
 
 
211 
 
   
Model 2 
 The addition of the marital, region, and education variables did not affect the 
relative importance of age
2
, in that age
2
  it still has the strongest beta weight (-.3459); 
however, it did reverse the relative importance between race and veteran status. In model 
1 race was second in strength to age
2
 followed by veteran status albeit in different 
directions. In model 2, we see that veteran status has a beta weight of .0964 and race has 
beta weight of -.0557.   
 Although veteran status contributes more to log income in model 2 than in model 
1 the variable with the second strongest strength in model 2 is some college education 
(B=.2144) followed closely by high school education (B=.0983). This is expected in that 
it is well known in human capital and stratification theory that education is a prime 
determinant of income. 
 This is interesting in that in 1970 having some college (b=.5753) was the 
strongest predictor of log income followed by age
2
 (b=.5095). This would lead one to 
assume that the 1980 census was the first period in which age truly began to affect the 
earnings of the World War II cohort. Furthermore, although veterans and non-veterans 
were affected equally by the age penalty veterans could offset this penalty with a veteran 
premium.  
 In this model race is not nearly as important as it was in model 1. In fact it has 
less strength as a variable than age squared, veteran status, being married, having high 
school degree, or having some college. This is consistent with the 1970 model 2, 
although the magnitude of the unstandardized coefficient was higher in 1970 (-.3346) 
versus -.2222 in the 1980 census. 
212 
 
   
 As in the 1970 census, married respondents earned a premium in every model in 
the 1980 census. The premium was smaller in the 1980s than in the 1970s in every 
model. A typical example of this is model 2 which yielded a .2455 unstandardized 
coefficient in 1980 versus a .3346 unstandardized coefficient in 1970.     
 Although they are significant, the variables for regional residence are not as 
influential as one might expect. In fact, they prove to be among the least important of all 
of the variables in this model and remain so in the remainder of the models.  
 
Model 3   
The addition of the mobilization variables and the interaction term had the most 
influence on the World War II veteran status, age
2
, and race variables.  As in models 1 
and 2 age
2
 has the most strength of any of the variables. In fact, the unstandardized 
coefficient decreases from -.3459 to -.2501 and the associated penalty for age
2
 is reduced. 
Veterans continue to receive an earnings premium in this model although it is reduced 
from the previous model.  The fact that the premium for being a veteran was reduced in 
this model was expected since some of the variance is captured in the race*veteran 
interaction term and the previous decades of analysis produced the same effects. The 
disadvantage for race increased slightly in this model and has more strength than the 
veteran status variable. Race has a beta weight of -.0618 versus .0608 for veteran status. 
This is due in large part to the addition of the black*veteran interaction term which is not 
significant in this model. 
Additionally, being a part of a birth cohort in the midst of the peak or 
demobilization periods of World War II yields an earnings premium. In fact, the 
mobilization periods 2 and 3 variables are second and third in terms of strength in this 
213 
 
   
model with betas of .2169 and .1555 respectively. This is expected because these two 
phases are comprised of the younger respondents in the sample and we know that there is 
a substantial penalty for age in this model holding all other variables constant. In 1970 
the coefficients for the mobilization phases were negative indicating that there was a 
penalty for being a part of these two phases. The difference between the 1970 and 1980 
censuses appears to be that age played a much more significant and negative role in 
earnings. Those who were younger in the 1970 census (mobilization phase 3) paid a 
penalty because of their young age, which as theory suggests is also tied to labor force 
experience, networking, and salary history. In the 1980 census they earned a premium 
because of the fact that they were younger. As the descriptive section of this chapter 
shows this was the first period in which median incomes went down for the World War II 
cohort.      
 
Model 4  
In model 4 the removal of the mobilization phase variables and the addition of the 
percent of a birth cohort that served did not create a substantial change in the veteran 
advantage although it remains positive and significant. Furthermore, age remains the 
strongest variable in this model (B=-.3286), and the magnitude increases from model 3; a 
trend that continues from the 1970 census.   
Married respondents, those from the North, high school graduates and, those with 
at least some college all earn premiums. The beta weights show that the educational 
variables continue to make highly significant contributions to earnings income. As in 
model 3 those who are single, black, or from the South pay significant earnings penalties. 
The magnitudes of these variables are all very similar to those described in model three.  
214 
 
   
Additionally, the black veteran advantage continues in this model. We find that 
although the contribution of this variable is less than other variables (B=.0063) black 
veterans continue to enjoy a premium for their service. It is clear that the black veteran 
advantage decreased over the decade; however, in 1970 the race*veteran interaction term 
was not significant.   
Furthermore, we find that those who were in cohorts with larger percentages of 
World War II veterans received a significant premium. The premium for those who were 
in cohorts with larger percentages of World War II veterans increased from the 1970 
census almost four fold, from .1055 (unstandardized) in 1970 to .4222 (unstandardized) 
in 1980. This finding is expected in that the peak mobilization phase variable in model 3 
had a positive coefficient and it is built on the percentage of veterans in the cohort. The 
removal of the mobilization phase variables and the addition of the percent of a birth 
cohort that served reduced the amount of explained variance from 22.82 percent in model 
3 to 22.12 percent of the variance in this model. 
 
Model 5  
The removal of the percent of birth year variable and the addition of the veteran 
and mobilization phase interaction terms in model 5 nominally decreases the explained 
variance from 22.13 percent to 22.08 percent. This magnitude of the change in the 
amount of explained variance from the 1960s to the 1970s was about the same as the 
magnitude of the change from the 1970s to the 1980s; however, the 1970s model 5 
explained 20.96 percent of the variance while the 1980s model explained 22.08 percent. 
Although the magnitudes and variances of most the variables remain very close to those 
in model 4, the veteran premium decreased from .0574 to a veteran penalty of -.0095. 
215 
 
   
Furthermore, in four decades of analysis thus far, this is the first model in which we see a 
veteran penalty.  Although the veteran premium decreased in the 1960s and 1970s 
between models 4 and 5, the magnitude of the change was the greatest in the 1980s. 
However, the peak mobilization phase veteran advantage returns. In fact, not only is there 
a veteran peak mobilization phase advantage there is also a veteran demobilization phase 
advantage.  Moreover, the fact that the veteran*mobilization phase interaction premiums 
exist partially explains why there is a decrease in the veteran premium in this model. 
Once we partial out the effects of being a veteran in mobilization phases 2 and 3 from the 
veteran status variable we are left with the older veterans of the beginning phases of the 
mobilization who receive a significant penalty for their age.  Additionally, the bivariate 
cases would seem to support this finding (see figure 7.3 above).  
 
Duncan SEI Models and Regression Analysis 
  
Model 1 
 In model 1, I regressed SEI on World War II veteran status, age
2
, and race. In this 
model race is the strongest SEI predictor holding veteran status and age
2
 constant. The 
beta coefficient (-.1770) indicates that blacks pay a penalty relative to whites. The 
coefficient for the race penalty is less in 1980 (-15.9348 (unstandardized)) than it was in 
1970 (-17.5417 (unstandardized)). World War II veterans receive an SEI premium in this 
model, which is consistent with 1970. The strength of this variable is less than the race 
variable; however, it is a substantial .0996 premium.  Older respondents pay a penalty 
that is less substantial than the race penalty and slightly more substantial than the 1970 
penalty for age
2
. Both the fact that the penalty is more substantial in 1980 as well as the 
216 
 
   
fact that the magnitude of the penalty is small is expected. The veterans in 1980 are 
between the ages of fifty and eighty and have in all probability achieved their maximum 
SEIs. Unlike education it is possible to reduce one?s SEI score, by taking on a new job or 
profession that generates a lower SEI score than one?s old profession or job. This model 
explains almost five percent of the variance in SEI versus six percent in the same model 
in 1970. Table 7.12 shows all of the Duncan SEI models and regressions. 
217 
 
   
Table 7.12: 1980 Duncan SEI Models and Regressions 
 
218 
 
   
Model 2 
 Model 2 adds marital, regional, and educational background variables to the 
regression equation. Clearly the three most influential variables in this model are college 
(B=.5776), high school graduate (B=.2093), and race (B=-.1074). This is expected and 
carries over from preceding decades.  
 The coefficient for the veteran status variable is 1.994 (unstandardized) versus 
2.064 (unstandardized) in 1970. This tells us that veterans earn fewer points on the 
Duncan SEI scale for their service in 1980 versus 1970. There was not only a decrease 
between decades there was also a decrease between models; the coefficients changed 
from .0996 to .0405.  
   The magnitude of the age squared coefficient decreases and remains negative as 
a result of controlling for background factors. In 1970 the age
2
 coefficient changed from 
negative to positive and the magnitude stayed about the same. The penalty for being 
black decreases from -.1770 to -.10749 when one controls for marital status, region, and 
education. Married respondents receive an SEI premium of 3.352 points versus almost 
four SEI points in 1970. Single respondents paid a 1.637 SEI unit penalty.   
 Residing in the North provides almost a one unit increase versus a 1.835 SEI unit 
increase in 1970 while residing in the South provides a half a unit increase versus almost 
a one unit (.9903) increase on the SEI scale in 1970. As in the previous two decades of 
study, the largest coefficients come from the education independent variables. As 
expected, those with some college (.5776) or a high school degree (.2093) continue to 
earn SEI premiums. The addition of marital, regional, and educational background 
variables to the regression equation increases the proportion of the variance that can be 
explained from five percent to thirty-one percent. 
219 
 
   
Model 3  
 In this model the addition of the race*veteran interaction term and the 
mobilization phase variables does not affect the adjusted R-squared of the model.  The 
education variables continue to provide the strongest predictors of SEI attainment. 
However, there is a modest increase in the veteran premium from .0405 to .0440. The 
associated change in 1970 was about one third of an SEI unit.  The age
2
 variable 
increased marginally and but remains negative and significant. There was little change in 
the magnitudes of the married or the education variables. The mobilization phase and 
race*veteran interaction were not significantly different from zero in this model.  
 
Model 4  
 In this model, I once again remove the mobilization phase variables and add in a 
variable that controls for the percent of veterans in a particular birth cohort. Making these 
changes in the model increased the veteran premium for SEI slightly and decreased the 
premium for married respondents slightly.  Furthermore, the coefficient for the percent of 
veterans in a birth cohort is -.9430 (unstandardized) which is much is smaller than the -
1.704 (unstandardized) penalty in the 1970 census. Essentially, this tells us that for every 
one unit increase in the percent of veterans in a birth cohort we can expect to see a .4008 
decrease in SEI score.  Additionally, in this model black veterans paid a significant .9430 
(unstandardized) SEI unit penalty in contrast to the .3353 (unstandardized) SEI unit 
penalty for serving in 1970.  
 
 
 
220 
 
   
Model 5  
  In the final SEI model I removed the percent of veterans in a birth year variable 
and added in the veteran*mobilization phase interaction terms.  The coefficient for World 
War II veteran status decreases from .0452 to .0375, which is consistent with the 
decrease in 1970. The coefficient for age
2
 decreases slightly, while the coefficients for 
race, married, North and South residence, high school graduate and some college 
remained virtually the same in magnitude, direction, and significance, as they did in 
1970. Both veteran mobilization phase variables were not significant in this model. 
Making these changes makes no difference in the adjusted R-squared. 
  
Summary 
 Several findings arise from this analysis of the 1980s census. This is the first 
period of analysis where the number of veterans is  larger than the number of non-
veterans, which would lead on to believe that veterans? mortality rates were lower than 
their non-veteran-peers or that veterans were less likely to leave the labor force during 
the intervening decade. 
 Veterans continued to earn significant premiums in terms of all three dependent 
variables in the 1980 census with the exception the model 5 income OLS regression. 
These findings support hypothesis one that controlling for background factors such as 
age, race, regional residence, marital status, and education level, World War II veterans 
attain greater social status than their non-veteran peers. 
 In general the analysis reveals that incomes generally decreased for whites and 
white veterans and increased slightly for blacks and black veterans.  Furthermore, 
although both black and white veterans held income advantages over their non-veteran 
221 
 
   
peers, for the first time in any of the censuses the difference in mean incomes was greater 
for white veterans versus their non-veteran peers than for black veterans versus their non-
veteran peers. Furthermore, model 4 was the only income model in 1980 to produce a 
significant coefficient for the race*veteran interaction term. This coefficient was positive 
but its magnitude was very small compared to previous census years.  Additionally, black 
veterans received significant education premiums in the 1980 census that were larger 
than the premiums that white veterans received relative to their non-veteran peers. 
Finally with respect to Duncan SEI score both black and white veterans earned 
significant premiums for their service; however, the difference for white veterans was 
greater than that of black veterans relative to their non-veteran peers. Based on this 
evidence we can conclude that for the 1980 census hypothesis 2 holds only for education.     
 In both the bivariate case as well as when we controlled for background factors 
the mobilization phase 2 veterans as well as the demobilization phase veterans regained 
the advantage that they lost in the 1970s in terms of SEI; however, both groups suffered 
education penalties in the 1980 census and the differences were not significant for 
income. This would lead us to conclude that hypothesis two holds only for education and 
that the premium for education was higher for peak phase veterans than for 
demobilization phase veterans.  
 Unlike the 1970s when they paid a penalty, cohorts with larger percentages of 
veterans gained earnings advantages. However, they paid education and SEI penalties in 
1980. This would lead us to conclude that hypothesis four holds only for income.  
 Much like other census periods being married or from the North generally tends 
to increase earnings income and being single or from the South tends to decrease 
earnings. Furthermore those with high school and some college continued to earn income 
222 
 
   
premiums for the education. As in the last decade of analysis older respondents generally 
paid an income penalty.  
   
223 
 
   
Chapter 8: World War II Veterans and the 1990 Census 
 
 
 In the 1990 census the youngest members of the World War II cohort are sixty 
and the oldest members are ninety years old.  They have lived through the turbulent 
sixties and seventies and most are past what would be considered by most to be their 
prime earning years. By the time of the 1990 census they have witnessed profound 
change not only in the United States, but in the World as well. They witnessed the fall of 
communism, the Berlin Wall, and the end of the Cold War, which had dominated most of 
their adult lives.  They witnessed globalization in an increasingly technical world which 
had become more connected through the growth of the internet and communications 
technology like the personal computer and pagers. This period also produced two 
recessions. The first in the early 1980s was caused by tight monetary policy to control 
inflation and correct overproduction problems. The second was in the late 1980s which 
was generally attributed to the collapse of the stock market in 1987 (Alcaly 2003: 
Chapter 1).  
 
Descriptive data: 
 In this chapter I use the same analytical methods utilized in chapters four through 
seven to describe the sample, compare the dependent measures, and examine the 
dependent variables while simultaneously making comparisons to the 1980 census. As in 
the previous chapters I shaped the data to age the World War II cohort by ten years 
ageing them from a range of fifty to eighty to sixty to ninety.  After making all of the 
adjustments that were made in previous chapters I was left with a total sample of 55,779, 
which is 164,556 respondents less than the total in 1980.    
224 
 
   
The median age for the sample is sixty-four years with an age range of sixty to 
ninety years. As in the 1960 through 1980 samples all of the respondents have birth years 
between 1900 and 1930. The only exception is that black veterans had no respondents 
born in 1903 or 1904. As expected the majority of the sample is white (93.50 percent), 
from the South (58.17 percent), and married (84.06 percent).  In the 1980 census 91.96 
percent were white, 57.82 percent were from the South, and 86.08 percent were married. 
The average income adjusted to year 2000 dollars is $21,088, which represents the 
second decade of declining incomes for the overall sample. The majority of the 
respondents are high school graduates (30.81 percent) or have some college (40.16 
percent) versus 30.44 percent high school graduates and 29.74 percent with some college 
in 1980.  This would lead one to believe that those with less education left the work 
force, one way the other, faster than those with more education. The median SEI for the 
sample is forty?four which is the same as it was in 1980 (see Table 8.1). 
 Of the 55,779 total respondents 28,910 are veterans, who represent approximately 
51.83 percent of the sample which is a decrease from the 1980 census (53.39 percent). 
This is the second period of analysis in which veterans outnumber non-veterans.  Figure 
8.1 illustrates the 1990 census proportion of veterans to non-veterans. Black veterans 
make up approximately 4.56 percent (1,318) of the veteran population versus 5.57 
percent (3,958) of the population in 1980.  In 1970 black veterans comprised seven 
approximately seven percent of the veteran population, therefore this represents two 
periods of decline for black veterans which indicates that they most likely left the labor 
force at higher rates than white veterans during the 1970s and 1980s.  
The majority of the veterans in the 1990 sample are in the sixty-five to seventy 
four year old age category; however, the majority of the non-veterans (61.36 percent) fall 
225 
 
   
into the sixty to sixty-four year old age category. Veterans have a median age of sixty-
five and their non-veteran peers have a median age of sixty-three years.  In the 1980 
census the median age for every demographic category was 57 years.  Most of the 
veterans are from the South (57.04 percent) and are married (84.06 percent). In 1980 
55.40 percent were from the South and 87.72 percent were married. Furthermore, 
veterans? median income in 1990 ($19,770) was less than their median income in 1980 
($37,630.45) (both adjusted to year 2000 dollars).  
226 
 
   
 
Table 8.1: 1990 Census Descriptive Summary  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
227 
 
   
Additionally, most of them are high school graduates (31.27 percent) or they have 
some college (45.43 percent). In 1980 30.44 percent were high school graduates and 
29.74 percent had some college. Veterans increased their median SEI from forty-four in 
1980 to forty-seven in 1990. As in the decade between 1970 and 1980, there was little 
change in the percentage of veteran high school graduates or veteran SEI; however, there 
was a large increase in the percent that have some college.   
Their non-veteran peers are on average sixty-two years old (median) and the 
majority of them are a part of the sixty to sixty-four year old age category.  The majority 
of the non-veteran group is from the South (59.39 percent) and married (82.24 percent) 
which is close to the 59.60 percent that were from the South and the 84.05 percent that 
were married in 1980. Non-veterans have a median income of $23,724.00 versus 
$28,225.45 in 1980 (both adjusted to year 2000 dollars). This is the first analysis period 
in which non-veteran?s median income is higher than veterans. 30.31 percent of the non-
veterans are high school graduates and 34.48 percent have some college. Non-veterans 
have a median SEI of thirty-six which is four points higher than in 1980 (see table 8.1).  
As in the other periods of analysis white veterans look much like the general 
veteran population.  White veterans have a median age of sixty-five years, are 
predominantly in the sixty-five to seventy-four year old age category (53.11 percent), are 
married (86.35 percent), and reside in the South (56.52 percent).  In the 1980 census 
88.48 percent were from the South and 55.63 percent were married. Their average 
income adjusted to year 2000 dollars is $19,770.00 versus $37,630.45 in 1980. 31.27 
percent of white veterans are high school graduates and 46.27 percent have some college 
versus 33.25 percent high school graduates and 34.50 percent with some college in 1980. 
 
228 
 
   
White veterans have a forty-seven median SEI, which is the highest of any group and a 
three point increase from 1980 (see table 8.1).   
Black veterans in this sample are sixty-five years old on average (median) and 
most of them are in the sixty-five to seventy-four year old age category (53.64 percent).  
Black veterans like their non-veteran peers are predominantly married (72.91 percent) 
and from the South (67.91 percent). In 1980 74.81 percent were married and 69.38 
percent were from the South. They have an adjusted income of $15,816.00 versus 
$27,180.45 in 1980. Like veterans in general and white veterans, black veterans? median 
income is lower than their non-veteran peers, who have a $17,351.47 median income. 
This is the first period of analysis in which black non-veterans have a higher median 
income than black veterans. Black veterans average almost a tenth grade education and 
they have an eighteen mean SEI as they did in the 1980 census (see table 8.1). This is 
also the first period in which black non-veterans have an SEI that is equal to black non-
veterans.  This suggests that black veterans left the labor force earlier than their non-
veteran peers either through normal attrition or death. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
229 
 
   
Figure 8.1: 1990 Census Veterans versus Non-Veteran 
 
Background Descriptive Statistics and Discussion 
 
Age   
 On average veterans are significantly older than their nonveteran peers by less 
than a year, the mean difference being .62.  In the 1980 sample they had a mean 
difference of .95. Based on their entry age into World War II, veterans should be between 
the ages of sixty-one and seventy-four if they entered between the ages of eighteen and 
twenty-six and for the most part they are (see table 8.2).  
 
 
 
 
 
Table 8.2: Veteran Expected Ages in 1990 by Entry Age 
230 
 
   
 
 
 White respondents have an average age (mean) of 66.07 years versus 58.72 years 
in 1980 and they are significantly older than black respondents by .37 years versus .16 
years in 1980. Black veterans have a mean age of 66.51 years versus 57.93 years in 1980 
and they are significantly older than their non-veteran peers by 1.25 years. Similarly 
white World War II veterans have a mean age of 66.34 versus 57.76 years in 1980 and 
they are significantly older than their non-veteran peers by .56 years, which is an increase 
over the decade because their average age was less that of their non-veteran peers in 
1980. The median ages for all of the demographic categories are provided in table 8.1. 
Table 8.3 provides the results of the significance tests described above. 
 
 
Table 8.3: 1990 Census Age Significance Tests 
Tests of 
Significance 
NMeanSD p NMeanSD
28,910 66.35438 4.400716
***
26,869 65.73315 6.942035
Tests of 
Significance 
White
NMeanSD p NMeanSD
3,626 65.70905 5.641815
***
52,153 66.07919 5.783134
Tests of 
Significance 
NMeanSD p NMeanSD
1,318 66.50531 4.454057
***
2,308 65.25433 6.17417
Tests of 
Significance 
NMeanSD p NMeanSD
27,592 66.34717 4.398104
***
24,561 65.77814 7.008302
White Veteran White Non-Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
231 
 
   
  
The logic for using birth cohorts and the cutoff years for the establishment of 
mobilization phases remains the same as in previous chapters and figure 8.1 (above) 
illustrates the 1990 census World War II cohort identified by mobilization phase.  
  
Region  
 More than fifty-eight percent of the sample maintains their residence in the South, 
23.83 percent in the North, and 18.00 percent in the West. In the 1980s the sample 
resided 57.82, 24.78, and 17.40 percent in the South, North, and West respectively. The 
regional distribution of respondents did not change substantially.  The veteran and non?
veteran proportions for residency are very similar. More than twenty-four percent of the 
veterans reside in the North versus 23.17 percent of the non-veterans; 57.04 percent of 
the veterans and 59.39 percent of the non-veterans reside in the South; and 18.52 percent 
of the veterans versus 17.44 percent of the non-veterans live in the West. In the 1980 
census 24.46 percent of the veterans lived in the North versus 23.93 percent of the non-
veterans; 56.40 percent of the veterans and 59.60 percent of the non-veterans lived in the 
South; and 18.15 percent of the veterans versus 16.48 percent of the non-veterans lived in 
the West.  
  I conducted a significance test assuming unequal variance to determine if the 
population proportions for veterans and non-veterans were different in terms of region.  
In this test I combined all of the regions and tested whether there was a significant 
difference between veterans and non-veterans. Moreover, the proportion significance 
tests that disaggregate each demographic category by region are all significant with the 
232 
 
   
exception of the test that compares the proportion of black veterans to non-veterans in the 
North (see table 8.4).  
Table 8.4: 1990 Region Significance Tests 
  Veteran  
Tests of 
Significance  
Non-Veteran  
 N Proportion P N Proportion 
North 7,064 0.2443 *** 6,226 0.2317
South 16,491 0.5704 *** 15,958 0.5939 
West 5,355 0.1852 *** 4,685 0.1744 
            
  Black 
Tests of 
Significance  
White 
 N Proportion P N Proportion 
North 690 0.1903 *** 12,600 0.2416 
South 2,601 0.7173 *** 29,848 0.5723 
West 335 0.0924 *** 9,705 0.1861 
            
  Black Veteran  
Tests of 
Significance  
Black Non-Veteran 
 N Proportion P N Proportion 
North 270 0.2049 *** 420 0.1820
South 895 0.6791 NS 1,706 0.7392 
West 153 0.1161 *** 182 0.0789 
            
  White Veteran 
Tests of 
Significance  
White Non-Veteran 
 N Proportion P N Proportion 
North 6,794 0.2462 ** 5,806 0.2364
South 15,596 0.5652 NS 14,252 0.5803 
West 5,202 0.1885 *** 4,503 0.1833 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001
  
Marital Status 
 Married respondents represent approximately 84.06 percent of the sample versus 
86.08 of the 1980 sample. The highest percentage of married respondents was 88.06 
percent in 1960.  As in the 1970 and 1980 censuses, whites are significantly more likely 
than blacks to be married. However, blacks are significantly more likely to be divorced, 
widowed, or single than whites (see table 8.5).  
233 
 
   
 The 1990 marital data with regard to veterans is different in many respects than 
the 1980 data. In the 1980 data veterans were significantly more likely to be married or 
divorced than non-veterans and non-veterans were significantly more likely to be 
widowed than veteran. However, in 1990 veterans were significantly more likely to be 
married but non-veterans were significantly more likely to be divorced, widowed, or 
single than veterans. 
 The proportions of married, divorced, widowed, and single veterans and non-
veterans were stable over the decade. The proportion of married veterans decreased from 
87.72 percent in 1980 to 85.74 percent in 1990, while the proportions for non-veterans 
decreased from 84.05 percent to 82.24 percent over the same period.  
 As in the 1980 census black veterans are significantly more likely to be married 
than their non-veteran peers. White veterans are significantly more likely to be married 
than their non-veteran peers and significantly less like to be divorced, widowed, or single 
than white non-veterans (see table 8.5). In the 1980 census white veterans were 
significantly more likely to be married than their non-veteran peers and significantly less 
like to be widowed or single than white non-veterans. 
234 
 
   
 
Table 8.5: 1990 Marital Status Significance Tests 
  Veteran  
Tests of 
Significance 
Non-Veteran  
 N Proportion P N Proportion
Married  24,788 0.8574 *** 22,098 0.8224 
Divorced 1,805 0.0624 ** 1,849 0.0688 
Widowed 1,577 0.0545 ** 1,629 0.0606 
Single Never Married 740 0.0256  *** 1,293 0.0481 
            
  Black 
Tests of 
Significance 
White 
 N Proportion P N Proportion
Married  2,560 0.7060 *** 44,326 0.8499
Divorced 520 0.1434 *** 3,134 0.0601
Widowed 331 0.0913 *** 2,875 0.0551
Single Never Married 215 0.0593 *** 1,818 0.0349
            
  Black Veteran  
Tests of 
Significance 
Black Non-Veteran
 N Proportion P N Proportion
Married  961 0.7291 *** 1,599 0.6928 
Divorced 179 0.1358 NS 341 0.1477 
Widowed 117 0.0888 NS 214 0.0927 
Single Never Married 61 0.0463 NS 154 0.0667 
            
  White Veteran 
Tests of 
Significance 
White Non-Veteran
 N Proportion P N Proportion
Married  23,827 0.8635 * 20,499 0.8346 
Divorced 1,626 0.0589 *** 1,508 0.0614 
Widowed 1,460 0.0529 *** 1,415 0.0576 
Single Never Married 679 0.0246 * 1,139 0.0464 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, p<.001 
  
 
 
 
 
 
235 
 
   
 Education 
 As in the preceding decade there was an increase in education for every age and 
demographic category. The average person in the sample attained better than an eleventh 
grade education level which is about the same as in the 1980 census (see figure 8.2). 
Black respondents attained a ninth grade education and whites attained an eleventh grade 
education. These are close to the educational attainment levels in the 1980 census. 
Figure 8.2: 1990 Cohort Education Levels 
0
500
1000
1500
2000
2500
3000
3500
1
90
0
1
90
2
1
9
04
1
90
6
1
90
8
1
9
10
1
91
2
1
91
4
1
9
16
1
91
8
19
20
1
92
2
1
92
4
19
26
1
92
8
1
93
0
Birth Year
Numbe
r
Grade School
Ninth to Eleventh 
 High School Graduate
 Some College
 
 
 When the data are disaggregated by veteran status, race, and age categories we 
find that veterans attained almost a full grade of education more than non-veterans. This 
is consistent with the educational differences in the 1980 census. Both black and white 
veterans attained almost a full grade more than their non-veteran peers. Although these 
differences are statistically significant they are not substantively different (see table 8.6). 
Moreover, these differences apply and are still significant for every age category with the 
236 
 
   
exception of blacks versus whites in the eighty-five to ninety year old category (see table 
8.7).  The difference in the means is greater for black veterans versus black non-veterans 
(.897) than for white veterans versus non-veterans (.647). The differences were .930 and 
.697 respectively in 1980. Compared to the 1980 census both the black and white veteran 
educational premiums decreased slightly. 
Table 8.6: 1990 Education Significance Tests 
 
Tests of 
Significance 
NMeanSD p N Mean SD
Aggregate 28910 7.021515 1.868526
***
26,869 6.309167 2.289287
Tests of 
Significance 
White
NMeanSD p N Mean SD
Aggregate 3626 5.418643 2.359807
***
52153 6.765958 2.065378
Tests of 
Significance 
NMeanSD p N Mean SD
Aggregate 1318 5.989378 2.166889
***
2308 5.092721 2.403595
Tests of 
Significance 
NMeanSD p N Mean SD
Aggregate 27592 7.070818 1.838675
***
24561 6.423476 2.244668
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
237 
 
   
 
Table 8.7: 1990 Education Significance Test by Age Category 
 
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
60-64 11,978 7.039322 1.818574 *** 16,486 6.54058 2.196205
 65-74 15,361 7.034438 1.875534 *** 6,413 5.974895 2.391271
 75-84 1,501 6.76016 2.124093 *** 3,450 5.96029 2.33927
 85-90 70 6.742857 2.435758
***
520 5.409615 2.529118
Tests of 
Significance 
White
Age Group N Mean SD p NMeanSD
60-64 1,953 5.588326 2.304332 *** 11,978 6.836068 2.01514
 65-74 1,348 5.321217 2.37281 *** 15,361 6.814844 2.044424
 75-84 288 4.770833 2.516063 *** 1,501 6.291229 2.262603
 85-90 37 5.054054 2.50495
NS
70 5.60217 2.554648
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
60-64 523 6.260038 2.011764 *** 1,430 5.342657 2.355717
 65-74 707 5.845827 2.211929 *** 641 4.74259 2.411084
 75-84 82 5.45122 2.514906 ** 206 4.5 2.470805
 85-90 6 6.666667 2.250926
NS
31 4.741935 2.462624
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
60-64 11,455 7.074902 1.801328 *** 15,056 6.654357 2.146035
 65-74 14,654 7.091784 1.838522 *** 5,772 6.111746 2.349726
 75-84 1,419 6.8358 2.075232 *** 3,244 6.053021 2.29996
 85-90 64 6.75 2.468854
***
489 5.451943 2.52979
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
Figure 8.3 illustrates the educational attainment of veterans and non-veterans by 
birth year. As described above, veterans attained significantly more education than non-
veterans; however these differences for the aggregate were not substantive. Additionally, 
this figure shows that there is a veteran premium but no additional premium for serving 
during the peak mobilization period of World War II. Figures 8.4 and 8.5 display the net 
education advantage for black and white veterans.  
Figure 8.3: Veteran versus Non-Veteran Education (College and High School Graduates Only) 
238 
 
   
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
19
0
0
19
0
2
1
90
4
19
0
6
19
0
8
1
91
0
1
91
2
1
91
4
1
91
6
1
9
18
1
92
0
1
9
22
1
9
24
1
9
26
1
9
28
1
9
30
Birth Year
Pe
rece
n
t
 V
e
terans
 and 
Non-
V
e
terans
 Non-Veteran High School Graduate
Non-Veteran  Some College
Veteran  High School Graduate
 Veteran Some College
 
Figure 8.4: Black Veteran versus Black Non-Veteran Education  
     (College and High School Graduates Only) 
0
0.2
0.4
0.6
0.8
1
1.2
1
90
0
19
0
2
1904 19
0
6
1908 1
91
0
1912 1
91
4
19
1
6
1918 19
2
0
1922 1
92
4
1926 1
92
8
19
3
0
Birth Year
Per
c
ent Bla
ck 
Veter
a
n and
 Non-Vete
r
a
n
Black Non-Veteran High School Graduate
 Black Non-Veteran Some College
Black Veteran High School Graduate
 Black Veteran Some College
 
239 
 
   
             Figure 8.5: White Veteran versus White Non-Veteran Education  
    (College and High School Graduates Only) 
0
0.2
0.4
0.6
0.8
1
1.2
190
0
1
90
2
190
4
19
06
19
08
191
0
191
2
191
4
191
6
191
8
192
0
192
2
192
4
192
6
192
8
19
30
Birth Year
Per
ecent 
V
e
ter
a
n
s
 and
 N
o
n-Vet
e
r
a
n
s
 White Non-Veteran High School Graduate
 White Non-Veteran Some College
 White Veteran High School Graduate
 White Veteran Some College
 
 There is a veteran advantage in the aggregate as well as when the group is 
disaggregated by race. Veterans in the aggregate receive a .712 educational unit 
advantage for their service which is less than the .789 that they received in 1980. This 
represents a two period decline in the net veteran educational advantage. Black veterans 
enjoyed a .897 mean difference compared to their non-veteran peers and white veterans 
enjoyed a .647 educational unit advantage over their non-veteran peers. These differences 
are statistically significant but not substantively different and smaller than the differences 
in 1980 for both black and white veterans who had differences of .930 and .697 
respectively relative to their non-veteran peers. 
 
240 
 
   
Income  
 The average (median) income in the sample is $9.9565 log dollars ($21,088.00) 
which is less than the 1980 median income of $10.4178 log dollars ($33,450.45)
6
. 
Incomes for all of the demographics decreased over the decade. Figure 8.6 illustrates the 
median log income for veterans versus non-veterans by birth year.  
 As expected the younger cohorts, both veteran and non-veteran, earn more than 
older cohorts; a trend that continues from the 1980s. Moreover, we see that non-veterans 
have a significantly higher mean income for only the sixty to sixty-four year group and 
veterans have significantly higher incomes for the remaining three. However, since the 
sixty to sixty-four year old age category is the largest category, in terms of number of 
respondents, it makes the overall mean income higher for non-veterans (see figure 8.6 
and table 8.8).  Figure 8.6 also illustrates that the veteran peak mobilization phase 
advantage has essentially ended.  
 Veterans in the sample have significantly lower mean incomes of $9.65 (ln) than 
their non-veterans peers who have mean incomes of $9.76 (ln). Although these 
differences are significant they are not substantive ($847).  In 1980 veterans had log 
incomes of $10.54 (ln) versus $10.25 (ln) for their non-veterans peers. Blacks have a 
mean income of 9.53 (ln) that is significantly and substantively less than whites who 
have a mean income of 9.72 (ln). In 1980 blacks had a mean income of $10.04 (ln) that 
was significantly and substantively less than whites who had a mean income of $10.46 
(ln). Notably, blacks in the eighty-five to ninety year old category had higher incomes 
                                                 
6
 All income figures are expressed in log dollars. Log dollars were computed by taking 
the natural logarithm of income adjusted in year 2000 dollars. The inflation factor was 
computed by multiplying the 1990 dollar amount by a factor of 1.318.   
241 
 
   
than whites. This the first period of analysis in which blacks in any category had higher 
incomes than whites.  
  Additionally, there is not a significant difference in mean income between black 
veterans and black non-veterans in the 1990s, signaling an end to the black veteran 
advantage. White non-veterans have significantly higher mean incomes than white 
veterans (see Figure 8.7 and table 8.8). Figure 8.4 also shows that in several cohorts 
(primarily the early cohorts) black veterans? incomes surpassed those of white veterans 
and non-veterans alike.    
 
Figure 8.6: 1990 Veteran versus Non-Veteran Log Income 
7.5
8
8.5
9
9.5
10
10.5
11
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
Lo
g In
com
e
 in
 $2000
 D
o
llars
Non-Veteran
Veteran
 
 In the 1980s the difference in the log income means between black veterans and 
black non-veterans was .3228. In 1990 the difference shrunk to .0023.  In 1980 white 
veterans and non-veterans had a significant difference of .3519 in mean log income and 
the difference was reduced to .1307 in 1990.  
242 
 
   
 
 
Figure 8.7: 1990 Census Black and White Veteran versus Non-Veteran Log Income 
0
2
4
6
8
10
12
14
1
90
0
1
90
2
1
9
04
1
90
6
1
90
8
1
91
0
1
91
2
19
14
1
91
6
1
9
18
1
92
0
1
92
2
1
92
4
1
92
6
1
9
28
1
93
0
Birth Year
L
o
g
 Inco
me
 in 2
000 D
o
llars
Black Veteran
Black Non-Veteran
White Veteran
White Non-Veteran
 
 When one looks at income disaggregated by age category, one finds that for each 
age category, except the sixty to sixty-four age category veterans have significantly and 
substantively higher incomes than non-veterans (see table 8.8).  
 
 
 
 
 
 
 
 
243 
 
   
Table 8.8: 1990 Income Significance Test by Age Categories 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
Socio-Economic Indicator (SEI)   
 
 The mean Duncan SEI for the sample is 41.65 versus 40.82 in 1980. The median 
SEI is forty-four, the same as it was in the 1980 census. This is the first period in which 
there was no change in the overall sample SEI.  The increase in SEI for the preceding 
decade was eleven points. Figure 8.8 illustrates the veteran premium for SEI. 
Interestingly, as with income, the veteran peak mobilization advantage is not nearly as 
244 
 
   
salient as it was in earlier decades.  In the 1980 data every veteran cohort with exception 
of the 1903 and 1929 cohorts had an SEI advantage. In the current data the 1900, 1903, 
1904, 1929 and 1930 cohorts have disadvantages or parity (see figure 7.8 of chapter 7 
and figure 8.8 below).   
 
Figure 8.8: 1990 Veteran versus Non-Veteran Duncan Median SEI score by Birth Year 
0
10
20
30
40
50
60
70
80
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
D
u
n
can
 S
E
I
Veteran
Non-Veteran
 
   As in the preceding decades, one would expect that the mean SEI differences 
between veterans and non-veterans, blacks and whites, and black and white veterans and 
non-veterans would be similar to the mean differences in income. As in the 1950 through 
the 1980 censuses, the sample dramatically illustrates this. Veterans have a significantly 
and substantively higher mean SEI than non-veterans; whites have a significantly and 
substantively higher mean SEI than blacks, black veterans have a significantly and 
substantively higher mean SEI than black non-veterans; and white veterans have a 
245 
 
   
significantly and substantively higher SEI than white non-veterans. These differences 
apply in all age groups (see table 8.9). 
 
Table 8.9: 1990 Significance Tests for Duncan SEI 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
 In the 1980 census the net advantage for white veterans (5.22) was higher than the 
net advantage for black veterans (4.94).  This changed in the 1990 census with the net 
advantage for white veterans decreasing to 4.21 and the net advantage for black veterans 
decreasing to 4.73.  Figure 8.9 illustrates the differences for black and white veterans 
246 
 
   
versus their non-veteran peers. The peak mobilization SEI premium is no longer salient.  
Furthermore, this graph illustrates the fact that although the mean black veteran 
advantage is greater than the white veteran advantage there are a few cohorts with 
substantial differences that might be considered to be outliers. If they were removed from 
the data the overall white veteran advantage might in fact be greater.   
 
Figure 8.9: 1990 Black and White Veteran Versus Non-Veteran Duncan Median SEI score by Birth 
Year 
0
10
20
30
40
50
60
70
80
90
100
19
00
19
02
1
90
4
1
90
6
1
90
8
1
91
0
19
12
1
9
14
1
91
6
1
91
8
1
92
0
19
22
19
24
1
9
26
1
92
8
1
93
0
Birth Year
D
u
ncan
 SE
I
Black Nonveteran SEI
Black Veteran SEI
White Nonveteran SEI
White Veteran SEI
 
Models and Multivariate Regression Analysis 
 In this section I follow the same conventions used in previous chapters by making 
use of the same five multivariate regression models to control for factors associated with 
earnings, education, and Duncan SEI outcomes to determine the net premium or penalty 
to veterans and non-veterans. Education Models and Regression Analysis 
 
247 
 
   
Education Models and Regression Analysis 
 
Model 1  
 As the bivariate descriptions above suggest, veterans tend to achieve higher 
education levels than their non-veteran peers, whites attained higher levels of education 
than blacks, and older birth cohorts generally had less education than younger cohorts 
(see tables 8.8 and 8.9 above).   Model 1 generally confirms the descriptions above (see 
table 8.12). When education attainment is regressed on World War II veteran status, age
2
, 
and race we see that the strongest predictor is veteran status (B = .1530), followed by 
race (B = -.13821), and age
2
 (B = -.0914). This indicates that veterans earned a premium 
while older respondents and blacks paid a penalty. While the directions of the 
coefficients for these variables were the same in 1980 the strength of the predictors 
changed. The strongest predictor in 1980 was race, followed by veteran status, and then 
age
2
. The model explains 5.3 percent of the variance in education versus six percent of 
the variance in 1980 
Model 2  
 The addition of marital status and region of residence background variables in 
model 2 modestly decreases the veteran relative to non-veteran advantage. However, it is 
still large and significant (B = .1506), while the black race penalty decreases slightly 
from -.0914 to -.0872 (both standardized). In 1980 the penalty decreased from -.1534 to -
.1399. As they did in the 1980 census, both married and single respondents have 
significant educational attainment premiums net of the effects of the other variables. 
Those residing in the North paid a .1951 (unstandardized) educational unit penalty versus 
a .1995 (unstandardized) educational unit penalty in 1980 and those living in the South 
248 
 
   
paid a .3013 (unstandardized) educational unit penalty versus a .3219 (unstandardized)  
in 1980 relative to those not living in these regions. As in previous decades the penalty 
was much higher for those living in the South than those in the North.   
 
  Model 3  
 The addition of the mobilization phase variables in model 3 shows that the 
highest premiums for educational attainment come from World War II veteran status and 
marital status.  The mobilization phase additions also serve to increase the veteran 
educational premium from   .1506 to .1559 (both standardized) versus .1395 to .1539 in 
1980 and increase the educational penalty for being black from -.1276 to -.1345 versus -
.1399 to -.1457 in 1980. The other variables were for the most part unaffected.  The 
coefficient for mobilization phase 2 (peak period) is significant and negative. Those who 
were a part of the peak mobilization phase, both veteran and non-veteran, paid a .0364 
educational unit penalty. In the 1980s model the peak mobilization veterans paid an 
educational penalty that was similar in magnitude to that of 1990.  Finally, model 3 
shows that black veterans received a significant  .0871 (unstandardized) educational unit 
premium in 1990 versus a .0777 (unstandardized) educational unit premium in 1980 
representing an increase in the black veteran advantage over the decade controlling for 
World War II veteran status, age
2
 race, marital status, geographic region, and 
mobilization phase. The addition of the mobilization phase variables in model 3 also 
increases the adjusted R-square slightly from .0646 to .0650 which is consistent with the 
small change (.0752 to .0775) from model 2 to model 3 in 1980. 
249 
 
   
 
Table 8.10: 1990 Education Models and Regressions 
 
250 
 
   
  
Model 4 
 In this model the World War II veteran status premium increases from a 
standardized .1590 to .1823. In fact, World War II veteran status is the strongest predictor 
of educational attainment in this model followed by Southern residence (B = -. 1379), 
race (B = -.1316), and then age
2
 (B = -.0849).   Furthermore, there was a slight decrease 
in the premium for black veterans. As with the 1980s census, this model shows that there 
is a large educational attainment penalty for being a part of birth cohorts with large 
proportions of veterans (B= -.0677).  There was also a decrease of .4008 (unstandardized) 
educational units in the 1980 census. The premium for black veterans changed marginally 
from .0123 to .0106 (both standardized) versus .0120 to .0098 (both standardized) in 
1980.  
 
Model 5 
 As in the 1980 census, the addition of the two veteran*mobilization interaction 
terms to the regression equation increases the veteran educational premium slightly and it 
remains significant. Veteran status remains the strongest predictor of education followed 
by being from the South, race, and then age
2
. This is consistent with change in the 1980 
census. This model also shows that there is an education penalty for being a veteran 
associated with mobilization phase 2. This is the third decade in which the education 
penalty for mobilization phase 2 veterans presents itself. The demobilization phase 
veterans pay a .1927 (unstandardized) educational unit penalty versus a .1119 
(unstandardized) educational unit penalty in 1980.     
251 
 
   
 Black veterans continue to earn a premium for their service and the coefficient 
increases from .0106 to .0124 as opposed to an increase from .0098 to .0135 in 1980. 
Black veterans earned educational premiums in every education model in the 1980s. 
 
Income Models and Regression Analysis 
 
Model 1 
Table 8.11 illustrates the results of all five multivariate models for income. In 
model 1 income is regressed on World War II veteran status, age
2
, and race and we see 
that this model and the remaining four all reliably predict the dependent variables in that 
the F values are all significant (p ? .001). In model 1 all three of the independent 
variables are significant. The strongest predictor of income in this model is age
2
 (B = -
.2976), followed by race (B = -.0412), and veteran status (B = -.0344). 
For veterans the predicted income is lower than for non-veterans.  Moreover, this 
is the first census period of analysis in which veterans paid a penalty for their service in 
model 1. In 1980 the unstandardized coefficient was .2708 versus -.0961 in the 1990 
census.  As we have seen in every census period up to this point blacks pay an earnings 
penalty relative to whites, although the penalty is much less than they paid in the 1980 
census. The unstandardized coefficient in 1980 was -.3727 versus -.2335 
(unstandardized) in 1990. This represents consecutive periods in which the race gap has 
decreased. Additionally, there is a penalty associated with age
2
 (B = -0.0005). The 
penalty in 1980 for age
2
 was almost the same (-0.0005).  This model explains 9.14 
percent of the variance in earnings versus 16.67 percent in the same model in 1980. 
252 
 
   
Table 8.11: 1990 Income Models and Regressions 
 
 
 
 
 
 
 
  
 
 
253 
 
   
Model 2 
  When marital status, region, and education are added to model 2 the 
magnitude of the World War II veteran status coefficient increases from -.0344 in model 
1 to -.0621 in model 2. Although the coefficients are negative in the 1990 model and 
positive in the 1980 model the decrease between models 1 and 2 is consistent between 
the census years.  Although the direction of the race variable was unaffected, in that 
blacks still earned less, the magnitude decreased from -.0412 in model 1 to -.0118 in 
model 2. As in the 1960 through 1980 data, model 2 produces the lowest race coefficients 
of any of the five income models. The race coefficient is also very weak in this model 
compared to other variables indicating that compared to other background factors race is 
less important for predicting income than it was in past census periods.   
The age
2
 coefficient remains the strongest predictor of income in this model as it 
did in the 1980 census and model 1 of the 1990 census. Furthermore, the unstandardized 
coefficients in 1980 (B = -.0049) and 1990 (B = -.0049) are almost the same.  As in the 
1970 and 1980 censuses we see the effects of age remaining negative throughout all the 
models. Moreover, this point is all the more obvious in light of the figure 8.2 (above). 
 As in the 1980 census being married increases earnings relative to not being 
married although the premium was much higher in the 1980 census (B = .0782) than the 
1990 census (B = .0294).  Additionally there were penalties associated with being single 
and from the South and premiums associated with being from the North, a high school 
graduate and having some college. These penalties and premiums were all consistent with 
the 1980 census penalties and premiums for these variables in both magnitude and 
direction of the coefficients.   
 
254 
 
   
Model 3   
The strongest predictors of income in this model, as the 1980 census, are age
2
 (B 
= -.2402), high school graduation (B= .0692), and some college (B = .2158). Veterans 
continue to pay a penalty in this model (B = -.0240) albeit less than the penalty in model 
2 (B = -.0668). The decrease in the magnitude is consistent with the decrease in 1980 
although veterans were still earning a premium in the 1980 model (b= .1324). This is also 
due in large part to the addition of the black*veteran interaction term. Black veterans 
receive a significant .0168 premium in this model. This interaction term was not 
significant in 1980. 
Additionally, being a part of a birth cohort in the midst of the peak mobilization 
period of World War II yields an earnings penalty (b = .2666) versus an earnings 
premium in 1980 (b = .4774).  Being born to a birth cohort that was a part of the final 
phase of the mobilization provided a premium in 1980 (b = .3818) versus a smaller 
earnings premium in 1990 (b = .0896).  
Model 4  
  In the 1990 model 4, the World War II veteran variable is not significant.  
Married respondents, those from the North, high school graduates, and those with at least 
some college all earned premiums that were relatively close to those they earned in 
model 3. As in model 3 those who are single, black, or from the South continued to pay 
significant earnings penalties. The magnitudes of these variables are all very similar to 
those described in model three.  
Additionally, the black veteran advantage continues in this model. We find that 
black veterans continue to have an income premium for their service in 1990 (b = .1378) 
versus a smaller premium (b = .0396) in 1980. It is clear that the black veteran advantage 
255 
 
   
increased over the decade although in the bivariate case this is no so obvious.  
Furthermore, we find that those who were in cohorts with larger percentages of World 
War II veterans pay a substantial penalty (b = -.6604). In 1980 they earned a premium (b 
= .4222). 
Model 5  
The removal of the percent of birth year variable and the addition of the veteran 
and mobilization phase interaction terms in model 5 do not affect the magnitudes and 
variances of most of the other the variables in the model. Most remain very close to their 
values in model 4, except that veterans earned a premium in this model. This is the first 
of the five models that produced a veteran advantage (B = .0300). Interestingly, the 1980 
census data produced the opposite effect in that model 5 was the only model to produce a 
veteran penalty.   
Veterans serving during peak mobilization phases paid a penalty in 1990 (b= -
.3730). In 1980 peak mobilization phase veterans earned a premium (b = .2933). 
Demobilization phase veterans also paid a penalty although it was less than the peak 
mobilization phase veterans (B= .0263). Much like the peak mobilization phase veterans 
they too received a premium in 1980 (b = .0922). The race*veteran interaction term was 
not significant in this model. 
 
Duncan SEI Models and Regression Analysis 
  
Model 1 
 In model 1, I regressed SEI on World War II veteran status, age
2
, and race. The 
strongest predictor in this model is race (B = -.1428), followed by veteran status (B = 
256 
 
   
.0847), and age
2
 (B = -.0373). World War II veterans receive a 4.303 SEI unit premium 
(unstandardized) versus a 4.909 SEI unit (unstandardized) premium in 1980. Older 
respondents as well as blacks paid an SEI penalty in 1980 that were similar in magnitude 
and direction to the 1990 coefficients.  This model explains three percent of the variance 
in SEI versus five percent in the same model in 1980. 
257 
 
   
Table 8.12: 1990 Duncan SEI Models and Regressions 
 
 
 
 
  
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
258 
 
   
Model 2 
 The addition of marital, regional, and educational background variables to the 
regression equation decreases the veteran status coefficient from .0847 to .0186 
(standardized). This is a very large decrease; however it is consistent with the 1980 
decrease.  The coefficient for the veteran status variable is .9458 (unstandardized) versus 
1.994 (unstandardized) in 1980. This tells us that veterans earn approximately one 
additional point on the Duncan SEI scale for their service. The magnitude of the age 
squared coefficient decreases and remains negative as a result of controlling for 
background factors. The penalty for being black decreases from .1428 to .0790 (both 
standardized) in 1990 versus .1770 to .10749 (both standardized) in 1980. Married 
respondents receive an SEI premium of 2.74 (unstandardized) points versus almost 3.352 
(unstandardized) SEI points in 1980. Single respondents paid a 2.4679 (unstandardized) 
SEI unit penalty versus a 1.637 (unstandardized) SEI unit penalty in 1980.   
 Residing in the North provides a 1.4967 (unstandardized) unit increase versus a 
1.835 (unstandardized) SEI unit increase in 1980. As in the previous three decades of 
study, the largest coefficients come from the education independent variables. As 
expected, those with some college (.5776) or a high school degree (.2093) continue to 
earn SEI premiums. 
 
Model 3  
 In this model the addition of the race*veteran interaction term and the 
mobilization phase variables does not affect the adjusted R-squared of the model.  
However, there is a modest increase in the veteran coefficient from .0186 to .0253 
(standardized). The associated change in 1980 was from .0405 to .0440 (standardized).  
259 
 
   
The age
2
 variable increased marginally and remains negative and significant. There was 
little change in the magnitudes of the married or the education variables. The only 
independent variable that was significant in this model was the peak mobilization phase 
variable. Those serving during the peak mobilization period suffered a penalty (B = -
.0326). This variable was not significant in 1980. 
  
Model 4  
 As in the 1980 census removing the mobilization phase variables and adding the 
percent of veterans in a particular birth cohort variable did not affect most of the 
coefficients of the other variables or the adjusted R-squared. The veteran and married 
premiums for SEI increased slightly, while the education premiums decreased slightly. 
Furthermore, the coefficient for the percent of veterans in a birth cohort is -2.2062 
(unstandardized) versus -.4008 (unstandardized) in 1980. Essentially, this tells us that for 
every one unit increase in the percent of veterans in a birth cohort we can expect to see a 
2.2062 unit decrease in SEI score and that the veteran SEI penalty grew over the decade.   
 
Model 5  
  In model 5 the coefficient for World War II veteran status increases from .0288 to 
.0354. In 1980 the premium decreased from .0452 to .0375.  The coefficients for race, 
married, and North residence, high school graduate and some college remained virtually 
the same in magnitude, direction, and significance, as they did in 1980. Neither the 
veteran*race interaction terms nor the veteran*mobilization phase interaction terms were 
significant in this model. 
 
260 
 
   
Summary 
 In the 1990 census many of the trends from previous decades remained true, 
however, several changes also became apparent. This was the second period of analysis 
in which veterans outnumbered non-veterans.  This suggests that non-veterans left the 
labor force in higher proportions than veterans.  In general the analysis reveals that in the 
bivariate case, median incomes generally decreased for veterans and non-veterans black 
and white alike. However, the decrease was much more substantial for veterans (47.46 
percent) than for non-veterans (15.95 percent). Whereas in the 1980 census older 
veterans generally earned less than younger veterans; in 1990 the difference between 
older and younger veterans is less clear until the 1925 birth cohort. In this census the 
1925 cohort is the starting point at which those in succeeding cohorts seem to have an 
earnings advantage. This is similar to the 1980 census when the 1915 cohort was the 
cohort that began the earnings advantage. Each of these cohorts were 65 years old (64 at 
the reporting time) during the respective census.  
 The veteran earnings premium of prior decades did not hold for veterans in the 
1990 census. In every regression model, except model 5, veterans suffered an earnings 
penalty. However, veterans earned premiums in every model in terms of education and 
Duncan SEI score in this census period. Based on these findings it is evident that 
hypothesis one holds with respect to education and SEI, but not income.  
 Black veterans received significant income and education premiums in all almost 
every model in the 1990 census as indicated by the positive race*veteran interaction term 
coefficients. However, there was no significant difference in income between black 
veterans and their non-veteran peers. White veterans on the other hand paid significant 
income penalties compared to their non veteran peers. Therefore, although blacks did not 
261 
 
   
earn a significant income premium in this period, in the aggregate, they still did better 
than white veterans in terms of income.  Additionally, the education coefficients for 
the race*veteran interaction term were slightly higher in the 1990 census than in the 1980 
census. Moreover, the black veteran educational premium (.4084) was greater than the 
white veteran education premium (.3179) relative to their non-veteran peers. As in the 
1980 census none of the Duncan SEI coefficients were significant for the race*veteran 
interaction term; however, the SEI difference between black veterans and their non-
veteran peers (4.734) was significant and greater than white veterans versus their non-
veteran peer SEI difference (4.212). Based on these findings we can conclude that 
hypothesis two holds for all three dependent variables. More specifically, black veterans 
receive more of a social status attainment premium relative to black non-veterans than 
white veterans relative to white non-veterans.  
 In almost every model of the 1990 census veterans of both the peak and 
demobilization periods paid significant education, earnings, and income penalties or the 
differences were not significant. In any event we can see that the World War II peak 
mobilization period premium faded in the 1990 census. As a result of these findings we 
can conclude that hypothesis three does no hold for the 1990 census. 
 Much like the peak mobilization effect described above, the large veteran cohort 
effect also faded in this census period. The percent of veterans in cohort variable 
produced negative coefficients in all three models for all three dependent variables. This 
would lead us to conclude that hypothesis four, which states that one would expect that 
veterans born to cohorts with larger proportions of veterans should achieve more social 
status attainment than those with a lower proportion of veterans, does not hold for the 
1990 census period.  
262 
 
   
 Much like other census periods being married or from the North generally tends 
to increase earnings income and being single or from the South tends to decrease 
earnings. Furthermore those with high school and some college continued to earn 
premiums for education as did those who were married. Married respondents, those from 
the North, high school graduates, and those with some college all received SEI premiums 
as well.  
 
 
263 
 
   
Chapter 9: World War II Veterans and the 2000 Census 
 
 
 
 In this final period of analysis for the World War II veterans we find that they are 
between the ages of seventy and ninety-three. The most interesting question of this period 
could perhaps be, who are the people who continue to have earnings income from 
employment at the age of ninety-three? Richardson and Waldrop (2003) stated that 11.6 
of the surviving World War II veterans were still working in 2000. It is my hypothesis 
that those who are still in employment in this period are those who owned their own 
businesses or had professional jobs (e.g. professors, doctors, and lawyers) and those who 
could not afford to not work. Those who could not afford to not work probably had not 
earned enough income during their prime earning years to be able to subsist without 
working in their latter years.   
 Interestingly, although the years between the 1990 and 2000 censuses were not 
nearly as turbulent as the periods between the 1960 through 1990 census periods, they 
were periods of general prosperity in the United States.  The end of the 20
th
 century 
brought with it record increases in the stock market, booming real estate prices, and 
record enrollments in institutions of higher learning (Alcaly 2003).  If one believes that a 
rising tide raises all ships then it one might also believe that these benefits probably 
accrued to the World War II cohort as well as the general population.  
 
Descriptive data 
  
In this chapter I use the same analytical methods utilized in chapters four through 
eight to describe the sample, compare the dependent measures, and examine the 
264 
 
   
dependent variables while simultaneously making comparisons to the 1990 census. As in 
the previous chapters I shaped the data to age the World War II cohort by ten years 
ageing them from a range of seventy to one hundred.  After making all of the adjustments 
that were made in previous chapters I was left with a total sample of 14,273, which is 
41,506 respondents less than the total in 1990.    
The median age for the sample is seventy-four years with an age range of seventy 
to ninety-three years.  Respondents have birth years between 1907 and 1930 with the 
exceptions of 1909 and 1910. As expected the majority of the sample is white (93.46 
percent), from the South (61.63 percent), and married (78.46 percent). In the 1990 census 
93.50 percent were white, 58.17 percent were from the South, and 84.06 percent were 
married. The average income is $12,000, which represents the second decade of declining 
incomes for the overall sample. The majority of the respondents are high school 
graduates (30.26 percent) or have some college (47.49 percent) versus 30.81 percent high 
school graduates and 40.16 percent with some college in 1990.  The median SEI for the 
sample is forty?four which is the same as it was in 1980 and 1990 (see table 9.1). This 
would suggest that we are not left with men in lower status occupations 
Of the 14,273 total respondents 6,902 are veterans, who represent approximately 
48.36 percent of the sample which is a decrease from the 1990 census (51.83 percent). 
Figure 9.1 illustrates the 2000 census proportion of veterans to non-veterans. Figure 9.1 
also illustrates the fact that there are no respondents for the 1909 and 1910 birth cohorts. 
The lack of respondents for 1909 and 1910 results in a gap during these two years in 
Figure 9.1 and all subsequent figures in this chapter. Black veterans make up 
approximately 4.72 percent (326) of the veteran population versus 4.56 percent (1,318) of 
the population in 1990.  Although the increase in the percentage of black veterans is 
265 
 
   
small this represents the first time in three decade of analysis that the proportion of black 
veterans to veterans has increased.  
266 
 
   
Table 9.1: 2000 Census Descriptive Summary 
 
 
 
 
 
 
267 
 
   
 
Figure 9.1: 2000 Census Veterans versus Non-Veteran 
 
 
The majority of the veterans in the 2000 sample are in the seventy-five to eighty-
four year old age category; however, the majority of the non-veterans (68.65 percent) fall 
into the seventy to seventy-four year old age category. Veterans have a median age of 
seventy-six and their non-veteran peers have a median age of seventy-two. Most of the 
veterans are from the South (60.40 percent) and are married (77.56 percent). In the 1990 
census 57.04 percent were from the South and 84.06 percent were married. Furthermore, 
veterans have a median income of $11,300 versus $19,770 in 1990 (adjusted to year 2000 
dollars), are high school graduates (29.53 percent) or have some college (51.23 percent). 
In 1990 31.27 percent were high school graduates and 45.43 percent had some college. 
The median veteran SEI remained the same over the decade (forty-four). As in the 
decades between 1970 and 1990, there was little change in the percentage of high school 
 
268 
 
   
graduates or SEI; however, there was a large increase in the percent with some college 
indicating that either those that did not go to college left the labor force earlier or that the 
veteran population and the population in general continued to educate themselves.   
Their non-veteran peers are on average seventy-two years old (median) and the 
majority of them are a part of the seventy to seventy?four year old age category.  The 
majority of the non-veteran group is from the South (62.79 percent) and married (77.86 
percent) which is higher than the 59.39 percent that were from the South and less than the 
82.24 percent that were married in 1990. Non-veterans have a median income of $12,000 
versus $23,724 in 1990 (adjusted to year 2000 dollars). This is the second analysis period 
in which non-veteran?s median income is higher than veterans. More than thirty-one 
percent of the non-veterans are high school graduates and 46.40 percent have some 
college. In 1990 30.31 and 34.48 percent of the non-veterans respectively were high 
school graduates or had some college. Non-veterans have a median SEI of forty-four 
which is eight points higher than in 1980 (see table 9.1).  
As in the other periods of analysis white veterans look much like the general 
veteran population.  White veterans have a median age of seventy-six years, are 
predominantly in the seventy-five to eighty-four year old age category (51.23 percent), 
are married (77.86 percent), and reside in the South (59.69 percent).  In the 1990 census 
86.35 percent were from the South and 56.52 percent were married. Their average 
income is $11,300 versus $19,770 in 1990 (adjusted to year 2000 dollars). This is the 
second period in a row in which white non-veterans? median incomes ($12,000) are 
higher than white veterans. Almost thirty percent of white veterans are high school 
graduates and 52.02 percent have some college versus 31.27 percent high school 
graduates and 46.27 percent with some college in 1990. White veterans have a forty-
269 
 
   
seven median SEI, which is the highest of any group and the same as their SEI score in 
1990 (see table 9.1).   
Black veterans in this sample are seventy-five years old on average (median) and 
most of them are in the seventy-five to eighty-four year old age category (51.23 percent).  
Black veterans like their non-veteran peers are predominantly married (71.47 percent) 
and from the South (74.85 percent). In 1990 72.91 percent were married and 67.91 
percent were from the South. They have an adjusted income of $12,000 versus $15,816 in 
1990 (adjusted to year 2000 dollars). Unlike white veterans, black veterans? median 
income is higher than their non veteran peers, who have an $11,500 median income. In 
1990 black non-veterans had a higher median income than black veterans. Black veterans 
average a little more than a tenth grade education and they have an eighteen mean SEI as 
they did in the 1990 census (see table 9.1).  
 
Background Descriptive Statistics and Discussion 
Age   
 On average veterans are significantly older than their non-veteran peers by over 
two years, the mean difference being 2.19.  In the 1990 sample they had a mean 
difference of .62. Based on their entry age into World War II, veterans should be between 
the ages of seventy-one and eighty-four if they entered between the ages of eighteen and 
twenty-six and for the most part they are (see table 9.2).  
Table 9.2: Veteran Expected Ages in 1990 by Entry Age 
 
270 
 
   
 There is no a significant difference in median age between blacks and whites in 
the 2000 census. In 1990 whites were significantly older than blacks by .37 years. Black 
veterans have a mean age of 75.90 years versus 66.51 years in 1990 and they are 
significantly older than their non-veteran peers by 1.32 versus 1.25 years in 1990. 
Similarly white World War II veterans have a mean age of 76.34 years versus 66.34 years 
in 1990 and they are significantly older than their non-veteran peers by 2.26 years versus 
.56 in 1990. The median ages for all of the demographic categories are provided in table 
9.1. Table 9.3 provides the results of the significance tests described above. 
 
Table 9.3: 2000 Census Age Significance Tests 
 
  
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 The logic for using birth cohorts and the cutoff years for the establishment of 
mobilization phases remains the same as in previous chapters and figure 9.1 (above) 
illustrates the 2000 census World War II cohort indicated by mobilization phase.  
 
 
271 
 
   
Region  
 61.63 percent of the sample maintains their residence in the South, 20.63 percent 
in the North, and 17.74 percent in the West. In the 1990 census the sample resided 58.17, 
23.83, and 18.00 percent in the South, North, and West respectively. The regional 
distribution of respondents did not change substantially.  The veteran and non-veteran 
proportions for residency are very similar.  The proportions of veterans residing in the 
North was 20.78 versus 20.49 percent of the non-veterans; 60.40 percent of the veterans 
and 62.79 percent of the non-veterans reside in the South; and 18.82 percent of the 
veterans versus 16.73 percent of the non-veterans live in the West. In the 1990 census 
24.43 percent of the veterans resided in the North versus 23.17 percent of the non-
veterans; 57.04 percent of the veterans and 59.39 percent of the non-veterans resided in 
the South; and 18.52 percent of the veterans versus 17.44 percent of the non-veterans 
lived in the West. 
  I conducted a significance test assuming unequal variance to determine if the 
population proportions for veterans and non-veterans were different in terms of region.  
In this test I combined all of the regions and tested whether there was a significant 
difference between veterans and non-veterans (see table 9.4).  In past periods of analysis 
these tests provided evidence that veterans in general, as well as black and white 
veterans, were significantly more likely to live in the North and West than their non-
veteran peers and that non-veterans are significantly more likely to live in the South. The 
only change from the previous analysis is that black veterans are significantly more likely 
to live in the South and West than their non-veteran peers. 
 
 
 
 
272 
 
   
Table 9.4: 2000 Region Significance Tests 
 
   
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
Marital Status 
 Married respondents represent approximately 78.46 percent of the sample versus 
84.06 of the 1990 sample. The highest percentage of married respondents remains at 
88.06 percent in the 1960 census.  As in the 1970 through 1990 censuses, whites are 
significantly more likely than blacks to be married. However, blacks are significantly 
more likely to be divorced, widowed, or single than whites (see table 9.5).  
 The 2000 marital data with regard to veterans is different in many respects than 
the 1990 data. In the 1990 data veterans were significantly more likely to be married but 
273 
 
   
non-veterans were significantly more likely to be divorced, widowed, or single than 
veterans. However, in 2000 veterans were significantly less likely to be married or single 
but significantly more likely to be widowed.  
 The proportions of married, divorced, widowed, and single veterans and non-
veterans were not as stable in the decade between 1990 and 2000 as they were in the 
preceding decade. The proportion of married veterans decreased from 85.74 percent in 
1990 to 77.56 percent in 2000, while the proportions for non-veterans decreased from 
82.24 percent to 79.31 over the same period.  
 White veterans are significantly more likely to be divorced or widowed than their 
non-veteran peers and significantly less like to be married or single than white non-
veterans (see table 9.5). In the 1990 census white veterans were significantly more likely 
to be married, than their non-veteran peers and significantly less like to be divorced, 
widowed, or single than white non-veterans. 
274 
 
   
 
Table 9.5: 2000 Marital Status Significance Tests 
 
 
 
 
 
 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
275 
 
   
Education 
 As in the preceding decades there was an increase in education for every age and 
demographic category. The average person in the sample is a high school graduate which 
is at least a grade level increase from the previous census (see figure 9.2). Black 
respondents attained a tenth grade education and whites were generally high school 
graduates. Both of these represent increases in educational attainment levels from the 
1990 census. 
 
 
Figure 9.2: 2000 Cohort Education Levels 
 
0
200
400
600
800
1000
1200
19
07
19
08
19
11
19
12
1
9
13
19
14
19
15
1
9
16
1
917
19
18
19
19
19
20
1
9
21
19
22
19
23
19
24
1
925
19
26
19
27
19
28
19
29
1
930
Birth Year 
Nu
mb
e
r
Grade School
Ninth to Eleventh 
 High School Graduate
 Some College
 
 When the data are disaggregated by veteran status, race, and age categories we 
find that veterans significantly attained more education than non-veterans. This is 
consistent with the educational differences in the 1990 census. Black veterans attained 
one grade level more education in the 2000 census than their non-veteran peers, which is 
276 
 
   
about the same one grade increase they had in the 1990 census. White veterans had a 
significant but not substantive higher education level than white non-veterans; however, 
the increase is less than they experienced in the 1990s. All of the mean differences 
referenced above are significant at the .05 level at least (see table 9.6).   
 The difference in the means is greater for black veterans versus black non-
veterans (.4534) than for white veterans versus non-veterans (.303). The differences were 
.897 and .647 respectively in 1990. Compared to the 1990 census both black and white 
veteran educational premiums decreased slightly. 
 
Table 9.6: 2000 Education Significance Tests 
 
Tests of 
Significance 
NMeanSD p NMeanSD
Aggregate 6902 7.231962 1.797438
***
7371 6.875729 2.115978
Tests of 
Significance 
White
NMeanSD p NMeanSD
Aggregate 934 5.842612 2.386965
***
13339 7.132394 1.916267
Tests of 
Significance 
NMeanSD p NMeanSD
Aggregate 326 6.138037 2.369543
**
608 5.684211 2.38314
Tests of 
Significance 
NMeanSD p NMeanSD
Aggregate 6576 7.286192 1.746757
***
6763 6.982848 2.056939
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
  
 
 
 
277 
 
   
Figure 9.3 illustrates the educational attainment of veterans and non-veterans by 
birth year. As described above, veterans attained significantly and substantively more 
education than non-veterans. Additionally, this figure shows that the veteran educational 
premium for serving during the peak mobilization period of World War II did not return. 
Figures 9.4 and 9.5 display the net education advantage for black and white veterans.  
 
Table 9.7: 2000 Education Significance Test by Age Category 
 
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
70-74 2697 7.276604 1.757313 *** 5060 7.038933 2.016797
 75-84 3900 7.229231 1.791452 *** 1887 6.479597 2.301616
 85-94 305 6.872131 2.155175
NS
424 6.691038 2.165999
Tests of 
Significance 
White
Age Group N Mean SD p NMeanSD
70-74 528 6.083333 2.289458 *** 7229 7.197399 1.883039
 75-84 356 5.542135 2.458883 *** 5431 7.079359 1.932698
 85-94 50 5.44 2.619861
***
679 6.864507 2.093512
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
70-74 147 6.360544 2.287385 NS 381 5.976378 2.284264
 75-84 167 6.005988 2.375825 *** 189 5.132275 2.464312
 85-94 12 5.25 3.078518
NS
38 5.5 2.501351
Tests of 
Significance 
Age Group N Mean SD p NMeanSD
70-74 2550 7.329412 1.707447 *** 4679 7.125454 1.968616
 75-84 3733 7.283954 1.741232 *** 1698 6.629564 2.233826
 85-94 293 6.938567 2.089516
NS
386 6.80829 2.097502
White Non-VeteranWhite Veteran
Non-Veteran Veteran 
Black
Black Veteran Black Non-Veteran
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
   
 
 
 
 
 
278 
 
   
Figure 9.3: Veteran versus Non-Veteran Education (College and High School Graduates Only) 
0
0.2
0.4
0.6
0.8
1
1.2
19
07
19
08
19
09
19
10
19
1
1
1
9
12
19
13
19
14
19
15
19
16
19
17
19
18
19
19
1
920
1
9
21
1
9
22
19
23
19
24
19
25
19
26
19
27
19
28
1
929
1
930
Birth Year
P
e
rcen
t o
f
 B
i
rt
h Year  Non-Veteran High School Graduate
Non-Veteran Some College
 Veteran High School Graduate
 Veteran Some College
         
Figure 9.4: Black Veteran versus Black Non-Veteran Education  
                   (College and High School Graduate Only) 
 
 
 
0
0.2
0.4
0.6
0.8
1
1.2
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
Birth Year
P
e
r
cent
 of B
i
r
t
h Year
 
Black Non-Veteran High School Graduate
Black Non-Veteran Some College
 Black Veteran High School Graduate
 Black Veteran Some College
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
279 
 
   
Figure 9.5: White Veteran versus White Non-Veteran Education  
(College and High School Graduates Only) 
 
0
0.2
0.4
0.6
0.8
1
1.2
1
90
7
1
90
8
1
90
9
1
91
0
1
9
11
19
12
1
91
3
1
91
4
1
91
5
1
91
6
1
91
7
1
91
8
1
91
9
1
9
20
19
21
19
22
1
92
3
1
92
4
1
92
5
1
92
6
1
92
7
1
92
8
1
9
29
1
9
30
Birth Year
P
e
rce
n
t
 o
f
 B
i
rt
h
 Y
ear
 Non-Veteran High School Graduate
Non-Veteran Some College
 Veteran High School Graduate
 Veteran Some College
 
 There is a veteran advantage in the aggregate as well as when the group is 
disaggregated by race. Veterans in the aggregate receive a .356 educational unit 
advantage for their service which is less than the .712 that they received in 1990. This 
represents a three period decline in the net veteran educational advantage. Black veterans 
enjoyed a .454 mean difference compared to their non-veteran peers and white veterans 
enjoyed a .303 educational unit advantage over their non-veteran peers. These differences 
are significant but not substantive and smaller than the differences in 1990 for both black 
and white veterans who had differences of .897 and .647 respectively relative to their 
non-veteran peers.   
 
 
280 
 
   
 Income  
 The average (median) income in the sample is $9.3927 log dollars ($12,000.00) 
which is less than the 1990 median income of $9.9565 log dollars ($21,088.00)
7
. Incomes 
for all of the demographics in this census decreased for the second decade in a row.  The 
decrease in earnings income was somewhat expected, because the cohort is well past its 
prime earning years.  Figure 9.6 illustrates the median log income for veterans versus 
non-veterans by birth year.  
 Whereas in other periods younger cohorts, both veteran and non-veteran, earned 
more than older cohorts, in the 2000 census incomes for all birth cohorts are similar with 
no discernable trends. Moreover, we see that non-veterans have a significantly and 
substantively higher mean income only in the seventy to seventy-four year age range. 
White non-veterans also have significantly and substantively higher mean incomes in this 
age category. Furthermore, black veterans have significantly and substantively higher 
mean incomes in the eighty-five to ninety-four year old age group.  None of the other age 
categories are significant for any of the demographic categories (see figure 9.6 and table 
9.8).  Figure 9.6 also illustrates that the veteran peak mobilization phase income 
advantage that last appeared in the 1980 census does not return in this period either. 
Moreover figure 9.6 shows that veteran and race differences disappear by the 2000 
census.  
 Veterans in the sample have significantly (but not substantive) lower mean 
incomes of $9.25 (ln) than their non-veteran peers who have mean incomes of $9.33 (ln).  
                                                 
7
 All income figures are expressed in log dollars. Log dollars were computed by taking 
the natural logarithm of income adjusted in year 2000 dollars. The inflation factor was 
computed by multiplying the 1990 dollar amount by a factor of 1.318.   
281 
 
   
In 1990 veterans had log incomes of $9.65 (ln) versus $9.76 (ln) for their non-veterans 
peers. Blacks have a mean income of $9.24 (ln) that does not differ significantly from 
whites who have a mean income of $9.29 (ln). In 1990 blacks had a mean income of 
$9.53 (ln) that was significantly but not substantively less than whites who had a mean 
income of $9.72 (ln). This is the first period in which there was not a significant 
difference in black and white incomes   
 As in the 1990 census there is no significant difference in mean income between 
black veterans and black non-veterans; however, white veterans pay a penalty relative to 
white non-veterans as in the 1990s. The mean log incomes are $9.34 (ln) and $9.25 (ln) 
respectively (see figure 9.6 and table 9.8). Figure 9.7 also shows that in several cohorts 
black veterans? incomes surpassed those of white veterans and non-veterans alike.    
 
Figure 9.6: 2000 Veteran versus Non-Veteran Log Income 
0
2
4
6
8
10
12
1
9
07
1
908
19
09
19
10
19
11
19
12
19
13
1
9
14
19
15
19
16
19
17
19
18
19
19
1
9
20
1
921
19
22
19
23
19
24
19
25
19
26
1
9
27
19
28
19
29
19
30
Birth Year
Lo
g I
n
co
me 
Veteran
Non-Veteran
 
282 
 
   
In 1990 white veterans and non-veterans had a significant difference of .1307 in mean log 
income and the difference was reduced to .0890 in 2000.  
 
Figure 9.7: 2000 Census Black and White Veteran versus Non-Veteran Log Income 
0
2
4
6
8
10
12
14
1
9
07
19
08
19
09
19
10
1
9
11
19
12
19
13
19
14
1
915
19
16
19
17
19
18
1
919
19
20
19
21
19
22
19
23
1
9
24
19
25
19
26
19
27
1
9
28
19
29
19
30
Birth Year
Lo
g I
n
co
me
White Veteran
White Non-Veteran
Black  Veteran
Black Non-Veteran
 
 When one looks at income disaggregated by age category, one finds that are very 
few significant differences (see table 9.8).  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 9.8: 2000 Income Significance Test by Age Categories 
283 
 
   
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 
Socio-Economic Indicator (SEI)   
 The mean Duncan SEI for the sample is 42.43 versus 41.65 in 1990. The median 
SEI is forty-four, the same as it was in the 1990 census. This is the second consecutive 
period in which there was no change in the overall sample SEI. Figure 9.8 illustrates the 
veteran premium for SEI. The veteran peak mobilization advantage is still evident in this 
period of analysis. In the 1990 data the 1900, 1903, 1904, 1929 and 1930 veteran cohorts 
had disadvantages or parity; however, in the 2000 census at least more of the veteran 
cohorts have disadvantages or parity (see figure 8.8 of chapter 8 and figure 9.8 below).   
284 
 
   
 
Figure 9.8: 2000 Veteran versus Non-Veteran Duncan Median SEI score by Birth Year 
0
10
20
30
40
50
60
19
07
19
08
1
911
19
12
19
13
1
9
14
19
15
19
16
1
917
19
18
19
19
19
20
19
21
19
22
19
2
3
1
9
24
19
25
19
26
19
27
19
28
19
29
1
930
Birth Year
D
u
n
can SEI
Veteran
Non-Veteran
 
   As in the preceding decades, one would expect that the mean SEI differences 
between veterans and non-veterans, blacks and whites, and black and white veterans and 
non-veterans would be similar to the mean differences in income. Veterans have a 
significantly and substantively higher mean SEI than non-veterans (43.2695 versus 
41.6439); whites have a significantly and substantively higher mean SEI than blacks 
(43.28 versus 30.23), black veterans have a significantly and substantively higher mean 
SEI than black non-veterans (32.37 versus 29.08); and white veterans have a significantly 
and substantively higher SEI than white non-veterans (43.81 versus 42.77).  Additionally 
table 9.9 details the results of the significance tests for each demographic category 
disaggregated by age category. 
285 
Table 9.9: 2000 Significance Tests for Duncan SEI 
 
 
 
  At the following levels of significance (two-tailed t-test): * p<.05, ** p<.01, ***p<.001
 In the 1990 census the net advantage for white veterans was 4.21 and the net 
advantage for black veterans was 4.73. In the 2000 census the net advantage for black 
and white veterans relative to their non-veteran peers is 1.04 and 3.29 respectively. It is 
clear that the black veteran advantage in the bivariate analysis is much larger than the 
white veteran advantage. Figure 9.9 illustrates the differences for black and white 
veterans versus their non-veteran peers. The peak mobilization SEI premium, which was 
dormant in the 1990 census, is not apparent in 2000 either.  
 
 
286 
 
   
 
Figure 9.9: 2000 Black and White Veteran versus Non-Veteran Duncan Median SEI score by Birth 
Year 
0
10
20
30
40
50
60
70
80
19
07
1
90
8
1
90
9
1
91
0
1
91
1
19
12
1
91
3
1
91
4
1
91
5
1
9
16
19
17
1
91
8
1
91
9
1
92
0
1
9
21
19
22
1
92
3
1
92
4
1
92
5
1
9
26
1
92
7
1
92
8
1
92
9
1
93
0
Birth Year
D
u
ncan
 SE
I
White Veteran
White Non-Veteran
Black  Veteran
Black Non-Veteran
 
Models and Multivariate Regression Analysis 
 In this section I follow the same conventions used in previous chapters by making 
use of the same five multivariate regression models to control for factors associated with 
education, earnings, and Duncan SEI outcomes to determine the net premium or penalty 
to veterans and non-veterans.  
 
Education Models and Regression Analysis 
 
Model 1  
 As the bivariate descriptions above suggest, veterans tend to achieve higher 
education levels than their non-veteran peers, whites attained higher levels of education 
than blacks, and older birth cohorts generally had less education than younger cohorts 
287 
 
   
(see tables 9.6 and 9.7 above).   Model 1 generally confirms the descriptions above (see 
table 9.10). When education attainment is regressed on World War II veteran status, age
2
, 
and race we see that the strongest predictor of education is race (B = -.1484), followed by 
veteran status (B = .0965), and then age
2
 (B = -.0744), whereas, in 1990 the strongest 
predictor was veteran status (B =.1530), followed by race (B = -.1382), and age
2
 (B = -
.0913).  
 
Model 2  
 The strongest predictors of education in this model are race (B = -.1390), 
Southern region (B = -.1309), and veteran status (B = .0968). The veteran relative to non-
veteran advantage remains large and significant (.0968), while the black race penalty 
decreases slightly from -.6042 to -.5661 (unstandardized). In 1990 the penalty decreased 
from -.6038 to -.5575 (unstandardized). As they did in the 1990 census, both married and 
single respondents have significant educational attainment premiums relative to 
unmarried and non-single respondents respectively net of the effects of the other 
variables. Those residing in the North paid a .1590 (unstandardized) educational unit 
penalty versus a .1951 (unstandardized)  educational unit penalty in 1990 and those living 
in the South paid a .2712 (unstandardized) educational unit penalty versus a .3013 
(unstandardized) in 1990 relative to those not living in these regions. As in previous 
decades the penalty was much higher for those living in the South than those in the 
North.  The addition of marital status and region of residence background variables in 
model 2 modestly increases the proportion of variance in education that can be predicted 
from the independent variables from 3.54 percent to 4.76 percent versus 5.3 to 6.47 
percent in 1990. 
288 
 
   
Table 9.10: 2000 Education Models and Regressions 
 
 
  
289 
 
   
 Model 3  
 As in model 2 the strongest predictors of education in this model are race (B = -
.1424), Southern region (B = -.1311), and veteran status (B = .1064).The premiums for 
educational attainment in this model come from World War II veteran status and marital 
status.  The mobilization phase additions increase the veteran educational coefficient 
from .1951 to .2143 (unstandardized) versus .3247 to .3360 (unstandardized) in 1990 and 
increase the educational penalty for being black from .5661 to .5801 (unstandardized) 
versus .5575 to .5875 (unstandardized) in 1990. The other variables were for the most 
part unaffected.  Being a part of mobilization phase 2 provided an educational penalty of 
(B = .0653) educational units.  
 
Model 4 
 Although model 4 is significant overall, in that the independent variables reliably 
predict the dependent variable (p<.001), there was very little change in the veteran status 
variable from model 3 to model 4; however, the 1990 model 4 premium (b = .3929) was 
much higher than the 2000 model 4 premium (b = . 2131).   However, as with the 1990s 
census, this model does show that there is a large educational attainment penalty for 
being a part of birth cohorts with large proportions of veterans.  For every one standard 
deviation increase in the percentage of veterans in a birth cohort, one would expect a 
.1435 standard deviation decrease in educational attainment, holding all other variables 
constant. There was also a decrease of .3079 (unstandardized) educational units in the 
1990 census. All of the other variables had little change across models and census 
periods. 
 
290 
 
   
Model 5 
 In this model the strongest predictors of education are race (B = -.1429), Southern 
region (B = -.1306), and veteran status (B = .1227). As in the 1990 census, the addition of 
the two veteran*mobilization interaction terms to the regression equation increases the 
veteran educational premium from a significant .2131 to .2472 (unstandardized) versus a 
significant .3928 to .4364 (unstandardized) in 1990. All of the other variables had little 
change across models and census periods. Furthermore, neither the veteran*race nor the 
veteran*mobilization phase interaction terms is significant.   
 
Income Models and Regression Analysis 
 
Model 1 
 Table 9.11 illustrates the results of all five multivariate models for income. In 
model 1 income is regressed on World War II veteran status, age
2
, and race and we see 
that the veteran status and age
2
 variables are significant. For veterans the predicted 
income is lower than for non-veterans (B = -.0188). This is lower than the penalty that 
veterans paid in 1990 (B = -.0344).  Moreover, this is the second census period in row in 
which veterans paid a penalty for their service.  Additionally, there is a penalty 
associated with age
2
 (-.0001 unstandardized). The penalty in 1990 for age
2
 was larger (-
0.0005 unstandardized).  
291 
 
   
Table 9.11: 2000 Income Models and Regressions 
 
292 
 
   
Model 2  
  The addition of the marital, region, and education variables increased the 
magnitude of the World War II veteran status variable from -.0188 in model 1 to -.0294 
(unstandardized) in model 2. In 1990 model 2 produced similar effects -.0961 in model 1 
to -.1736661 (unstandardized) in model 2. Those from the South suffered an earnings 
penalty (B=-.0299). This penalty remains the same in magnitude and direction 
throughout all five models    
 None of the marital variables are significant in this model or any of the other 
models for income. The age
2
 coefficient remains small and negative which indicates that 
for a one standard deviation increase in age
2
 one might expect a .0293 standard deviation 
decrease in earnings.  As in the 1970 through 1990 censuses we see the effects of age 
remaining negative and significant throughout all the models. 
 As in other periods, those with some college earned a significant premium 
(b=.4244) which is less than the 1990 premium (b=.6210). High school graduates earned 
a significant premium (b=.1192) versus the 1990 premium (b=.2111).  When marital 
status, region, and education are added to model 2 it explains approximately two percent 
of the variance in earnings income. 
 
Model 3   
The addition of the mobilization variables and the interaction term served to 
increase the income penalty for World War II veteran status. The penalty for being a 
veteran increased from   -.0294 to -.0321 (standardized). There was almost no change in 
the coefficients for the education variables although they remained significant. Neither 
293 
 
   
the mobilization phase variables nor the race*veteran interaction term is significant in 
this model.  
 
Model 4  
In this model the significant variables are World War II veteran status (B = -
.0292), age
2
, (B = -.0267), Southern residence (B = -.0299), high school graduate (B = 
.0377), and some college (B = .1457).   
Veterans continued to pay a significant penalty (B = -.0292); however, it was less 
than the penalty that they paid in model three (B =-.0321). The veteran status variable 
was not significant in 1990. There was no change in the magnitudes or direction s of the 
age
2
, Southern residence, high school graduate, or some college variables between 
models 3 and 4. The magnitude of the significant 2000 coefficients is much smaller the 
1990 significant coefficients. Moreover the amount of variance explained by the models 
decreases from 14.26 percent in 1990 to 1.99 percent in 2000.    
 
Model 5  
The removal of the percent of birth year variable and the addition of the veteran 
and mobilization phase interaction terms in model 5 has no affect on the explained 
variance in income; it remains at about two percent. Additionally, four of the variables 
had significant coefficients: age
2
, Southern region, high school graduate, and some 
college. The penalty for age
2
 increased form -.0267 to -.0347. The differences in 
magnitudes and directions of the other variables changed little from model 4 to model 5 
in this census as in 1990.   
 
294 
 
   
Duncan SEI Models and Regression Analysis 
 
Model 1 
 In model 1, I regressed SEI on World War II veteran status, age
2
, and race. This 
model explains 1.63 percent of the variance in SEI versus three percent in the same 
model in 1990. World War II veterans receive a significant 1.312 (unstandardized) SEI 
unit premium versus a 4.303 (unstandardized) SEI unit premium in 1990. This is the only 
SEI dependent variable model in which veteran status is significant. Blacks paid an SEI 
penalty of 12.88 (unstandardized) units in the 2000 census versus 14.69 (unstandardized) 
units in the 1990 census. This is the highest penalty that blacks pay in any of the 2000 
census models. 
 
Model 2 
  In model 2 the magnitude of the age
2
 coefficient increases and becomes positive 
(B = .0192).  Furthermore, age
2
 remains positive throughout the rest of the models. The 
penalty for being black decreases from 12.88 to 6.42 (unstandardized) in 2000 versus 
14.69 to 8.13 (unstandardized) in 1990. Married respondents receive an SEI premium of 
1.88 (unstandardized) points versus almost 2.74 (unstandardized) SEI points in 1990. 
 Residing in the North provides a 1.45 (unstandardized) unit increase versus a 1.50 
(unstandardized) SEI unit increase in 1990. As in the previous four decades of study, the 
largest coefficients come from the education independent variables. As expected, those 
with some college (B = .5551) or a high school degree (B = .1396) continue to earn SEI 
premiums. The addition of marital, regional, and educational background variables to the 
295 
 
   
regression equation increases the proportion of the variance that can be explained from 
1.63 percent to 24.48 percent versus three percent to 28.55 percent in the 1990 model. 
296 
 
   
Table 9.12: 2000 Duncan SEI Models and Regressions 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
297 
 
   
Model 3  
 In this model the addition of the race*veteran interaction term and the 
mobilization phase variables does not affect the adjusted R-squared of the model, nor 
does it affect the magnitude or direction of the other variables in the model. The five 
variables that are significant are race, married, Northern residence, high school graduate, 
and some college.  
 
Model 4  
 As in the 1990 census removing the mobilization phase variables and adding the 
percent of veterans in a particular birth cohort variable had little affect on the coefficients 
of the control variables or the adjusted R-squared. The adjusted R-squared remained at 
24.43 percent.  
 
Model 5  
  In model 5 the adjusted R-squared once again remained the same as it was in 
model 4. The coefficients for race, married, and North residence, high school graduate 
and some college remained virtually the same in magnitude, direction. Race, married 
South, and the education variables were the significant variables in the model and as 
stated above there was virtually no change in the magnitudes or direction from 1990 to 
2000.  
 
Summary 
 In this last period of analysis we see many of the differences between veterans 
and non-veterans, as well as black and whites dissipating. The most obvious of these is in 
298 
 
   
income where there are very few discernable trends. While there are few trends, two that 
continued from the 1990 census were (1) non-veterans continued to earn significantly 
more than veterans and (2) white non-veterans continued to earn more than white 
veterans. These findings suggest that hypothesis one does not hold for the 2000 census 
for income. This could be because veterans could afford to take lower paying jobs in later 
years because they had larger supplemental incomes from previous years. In previous 
periods we found that younger cohorts earned a premium compared to older cohorts; 
however, in this census incomes for all of the cohorts are about the same. Those in 
cohorts with a large percentage of veterans do not differ significantly from those with 
lower proportions of veterans. Furthermore we find that there are no significant income 
differences between black veterans and black non-veterans.  
 We also find that although every demographic category increased its mean level 
of education and that there were significant differences between veterans and non-
veterans, both black and white, that the differences between them was smaller than the 
differences in 1990. Furthermore, the premium for black veterans relative to black non-
veterans remained greater than the premium for white veterans relative to white non-
veterans in the 2000 census. Moreover, there was no significant difference between 
veterans of large proportioned birth cohorts and veterans of smaller veteran proportioned 
birth cohorts. 
 When we look at Duncan SEI we find that veterans (both black and white) 
continue to have higher SEIs than non-veterans. The persistence of these differences is 
expected as it would seem unlikely that one would take on a new career in the latter 
stages of one?s life and be able to change the SEI status that one has achieved over a 50 
year period. Furthermore, the veteran peak mobilization advantage continues to persist in 
299 
 
   
terms of education in the bivariate analysis; however, when background factors are 
controlled for in the regression models there is no significant advantage. Similarly when 
one controls for background factors there is no black veteran education advantage in the 
2000 census.  
 
 
300 
 
   
Chapter 10: Summary and Conclusions 
  
 I began this dissertation with a discussion of Stouffer et al.?s research on World 
War II veterans? postwar expectations which were drawn from intensive interviews with 
men after their discharge from the Army.  Stouffer believed that an Army career could be 
advantageous in two ways, ?(a) in terms of G.I. Bill benefits and veteran?s preference 
especially in government employment, and (b) in terms of intrinsic values of Army 
experience in teaching something which might be useful in civilian life? (Stouffer et al. 
1949: 609).  
 Stouffer, et al. found it interesting, as did I, that servicemen tended to be more 
optimistic about their personal chances of employment than about the chances of 
employment for veterans in general.  To be more specific, seventy-nine percent thought 
?most soldiers would find it ?very hard? or ?fairly hard? to get the kind of jobs they 
wanted after the war? (Stouffer et al. 1949: 598). The author posited two reasons that 
servicemen might feel this way about their futures. The first revolved around the 
expectation of another ?Great Depression? and the second was a general feeling that 
soldiers? time in the Army was not valuable. In fact, he states rather emphatically, ?it was 
rather difficult to induce men to admit that their Army experience had been especially 
valuable, even though in retrospect as veterans they might eventually find it valuable, at 
least in some respects? (Stouffer et al. 1949: 610).   
 With this as a backdrop I set out to test if World War II veterans receive a penalty 
or a premium for their service. Furthermore, if there are penalties or premiums for service 
I wanted to find out if they extended to both black and white veterans. In order to do this 
301 
 
   
I first turned to previous research, which generally shows that there is a veteran 
advantage, particularly for World War II and minority veterans. 
 The research on veteran social status attainment is generally rooted in three 
theories: the bridging hypothesis, human capital theory (HCT), and life course theory, all 
of which are discussed in chapter 2 of this dissertation. More specifically, the research 
shows that World War II veterans earned a premium for their service in the armed forces  
(Cooney, Segal, Segal, and Falk 2003; Elder and Meguro 1987; Fredland and Little 1985; 
Martindale and Poston 1979; Sampson and Laub 1996; Teachman and Tedrow 2004; 
Villemez and Kasarda 1976).   
 The research generally points to the expansive benefits of the G.I. Bill, heath 
factors (veterans are generally healthy), and the expanding economy after World War II 
as significant variables in this analysis.  Furthermore, the research suggests that this 
premium extends to both white men and blacks of the World War II cohort.  In fact, some 
authors suggest that the premium to blacks was greater than that of other veterans during 
the World War II and Korean eras (Detray 1982; Martindale and Poston 1979; Villemez 
and Kasarda 1976).  Fredland and Little (1985) suggest that the black veteran premium 
arose because that population was afforded preferential status in government 
employment.      
 As important as the past research is to the field, almost all of it looks at a very 
small time frame in the lives of veterans. Some of these studies (e.g. Cooney et al.) draw 
on a single census period to make generalizations about the social status attainment of 
veterans. Others use longitudinal data that span much longer time frames, but still fail to 
cover veterans? lifetimes to ascertain if the veteran advantage is a function of a specific 
period of veterans? lives or a life long advantage. For example, Teachman and Tedrow 
302 
 
   
(2004) conducted one of the relatively new studies on World War II veterans. In their 
paper they state, ?We make use of data taken from the National Longitudinal Study of 
Mature Men (NLSMM) to examine the ?long-term? effects of military service during 
WWII on occupational and income attainments. Although, they do use a much wider 
period of analysis than most, they still fail to capture the two decades immediately 
following World War II and the 2000 decade in which approximately eleven percent of 
the World War II veterans were still working (see footnote 1 of Chapter 2).  
 Furthermore, the research to date generally focuses on either income, education, 
SEI or a combination of two of the above; relatively little of the research has focused on 
all three. However, as HCT would suggest, income is tied to education, and SEI is tied to 
both. Earnings are most often used to describe differences in wages and earnings income 
that account for increased life chances (economic well being and purchasing power) 
while the Duncan SEI score is indicative of occupational prestige. Therefore, it would 
seem reasonable to assume that where the data allow, the best understanding of social 
status attainment might come from an analysis of all three of these variables. In order to 
correct these problems I used consecutive decennial censuses to conduct an analysis of 
veterans versus their non-veteran peers in the aggregate and black and white veterans 
versus their non-veteran peers. Furthermore, I used education, income, and SEI as 
dependent variables.  
 
Education  
 I chose to examine the education variable first for two reasons. The first is that 
much of the past research including Stouffer et al.?s insight suggests that G.I. Bill 
educational benefits are important factors in veteran social status attainment. Second, I 
303 
 
   
believe that it is important to understand the impact of veteran status on education prior 
to using education as a predictor for veteran income and SEI attainment.  In my first 
hypothesis I suggested that controlling for background factors such as age, race, regional 
residence, and marital status, that World War II veterans would attain greater social 
status than their non-veteran peers. My findings suggest that in terms of education, World 
War II veterans do in fact receive substantial educational premiums relative to their non-
veteran peers. What is more, these premiums are maintained throughout their lives (see 
figure 10.1).  
 Furthermore, these premiums extend to both black and white veterans although 
the magnitudes of the differences are slightly higher for black veterans than white 
veterans relative to their non-veteran peers. This finding supports my second hypothesis 
which states that black veterans receive more of a social status attainment premium than 
white veterans relative to their non-veterans peers in terms of education (see figure 10.2). 
The data also show that although World War II veteran status does not serve as an 
educational equalizer for race it does serve to close the gap between blacks and whites.   
 My third hypothesis states that selectivity bias will be lowest when the largest 
proportions of veterans serve in a cohort. Furthermore, when selection bias is low, if the 
bridging hypothesis and HCT perspectives are accurate, one would expect that veterans 
born to cohorts with larger proportions of veterans should achieve more social status 
attainment than those with a lower proportion of veterans. My findings suggest that this 
premise does not hold for education. The 1950 census was the only period of six that 
produced positive education results for cohorts with large proportions of veterans. 
Additionally, I find that although peak mobilization period veterans earn premiums 
during the early part of their lives (1950 and 1960 censuses), these premiums did not 
304 
 
   
convey to the middle or end of their lives.  Thus any advantage that they earned was 
immediate but short lived. 
 In hypothesis four I suggest that one would expect that veterans born to cohorts 
with larger proportions of veterans should achieve more social status attainment than 
those with a lower proportion of veterans. This hypothesis fails for every census period in 
terms of education. Those born to cohorts with large percentages of veterans paid 
significant educational penalties in model 4 of every census year. 
 The regression models as well as the bivariate relationships show that World War 
II veteran status provided an educational premium for World War II veterans in every 
period. Furthermore, the premiums grew in every census period and did not decrease 
until the 2000 census when veteran status was still the strongest predictor of educational 
attainment. 
 Finally, with respect to education, there are also premiums associated with being 
married or single versus being divorced or widowed, although the premiums are 
generally double for being married versus being single in every census period. 
Interestingly, residing in the North and the South versus the West generally resulted in 
educational penalties; however, the penalties for living in the South were substantially 
higher than living in the North.  
 
 
 
 
 
 
 
 
 
 
 
Figure 10.1: Veteran versus Non-Veteran Education 
305 
 
   
0
0.5
1
1.5
2
2.5
1950 1960 1970 1980 1990 2000
Census Year
E
d
uca
t
io
n Cat
e
go
r
y
 
C
o
de
 
Veteran 
Non-Veteran
 
 
Figure 10.2:  Black and White Veteran versus Non-Veteran Education 
0
0.5
1
1.5
2
2.5
1950 1960 1970 1980 1990 2000
Census Year
Mean
 E
d
uca
t
io
n Cat
e
go
r
y
 
C
o
de
 
White Veterans 
White Non-Veterans
Black Non-Veterans
Black Veterans 
 
306 
 
   
Income  
 My findings suggest that the income advantage for World War II veterans began 
when they entered the civilian labor force and remained until the 1990 census at which 
point their non-veteran peers still in the labor force caught up and surpassed them in 
terms of income. Figure 10.3 shows the mean natural logarithm of earnings for World 
War II veterans from the 1950 census through the 2000 census. This figure clearly 
illustrates the veteran advantage and substantiates hypothesis number one proposed in 
Chapter 3, which states, that controlling for background factors such as age, race, 
regional residence, marital status, and education level, World War II veterans will attain 
greater social status than their non-veteran peers. Moreover, my findings suggest that 
although both black and white veterans earned the aforementioned premium, the 
magnitude of the difference between black veterans and non-veterans was greater than 
the magnitude of the difference between white veterans and non-veterans. Figure 10.4 
dramatically illustrates this as well as the fact that although veteran status did not serve as 
a racial equalizer for black veterans relative to whites it did decrease the income gap for 
almost every period of study. Overall these findings tend to substantiate hypothesis two.  
 My findings also suggest that veterans that were a part of the World War II peak 
mobilization phases earned premiums during the beginning and mid potions of their 
lives; however, this advantage faded away during the 1970 census, returned in the 1980 
census, and was completely eroded by the 1990 census.  
 In hypothesis four I suggested that one would expect that veterans born to cohorts 
with larger proportions of veterans should achieve more social status attainment than 
those with a lower proportion of veterans. My research suggests that this hypothesis 
307 
 
   
holds for the 1950 through 1980 census periods. This is important because these were the 
prime working years for the World War II cohort.  
 Additionally, with respect to income, both the bivariate and regression analyses 
show that income for World War II veterans increased dramatically between the 1950 
and 1960 census, leveled during the 1970 and 1980 censuses, and decreased during both 
the 1990 and 200 censuses.   
  Additionally, there are effects on income above and beyond the veteran effects 
that tell us a great deal about income attainment. The first is that age (being older) 
generally provided premiums in the 1950 and 1960 censuses. However, beginning with 
the 1970 census older respondents began to pay penalties. Additionally, being from the 
North generally provided income premiums while being from the South was almost 
always a disadvantage. Furthermore, the Northern advantage and Southern disadvantage 
remained stable over all six census periods.  As expected, I found that education 
contributes a great deal to income attainment. Furthermore, high school graduates and 
those with some college earned substantial premiums in every census period. These 
premiums increased from census to census with the exception of the 1970 to 1980 and 
1990 to 2000 interludes.    
 
 
 
 
 
 
 
308 
 
   
Figure 10.3: Veteran versus Non-Veteran Income 
8
8.5
9
9.5
10
10.5
11
1950 1960 1970 1980 1990 2000
Census Year
Natu
r
a
l Lo
garith
m o
f
 In
co
me 
Veteran 
Non-Veteran
Figure 10.4:  Black and White Veteran versus Non-Veteran Income 
8
8.5
9
9.5
10
10.5
11
1950 1960 1970 1980 1990 2000
Census Year
N
a
tu
r
a
l
 Lo
ga
r
i
th
m o
f
 I
n
c
o
me
 
White Veterans 
White Non-Veterans
Black Non-Veterans
Black Veterans 
 
309 
 
   
Duncan SEI 
 This dissertation has shown that veterans clearly attain more occupational prestige 
than their non-veteran peers as measured using the Duncan SEI scale. As in the above 
discussions of education and income this finding serves to substantiate my first 
hypothesis. Moreover, this research shows that not only do World War II veterans 
receive an SEI premium for their service they maintain it throughout their lives (see 
figure 10.5).  The veteran premium extends to black and white veterans relative to their 
non-veteran peers; however, the black veteran advantage is generally larger and more 
consistent over the lifetime (see figure 10.6).  As with income and education this finding 
substantiates my second hypothesis (discussed above). 
  
 
Figure 10.5: Veteran versus Non-Veteran SEI 
20
25
30
35
40
45
50
1950 1960 1970 1980 1990 2000
Census Year
D
u
n
c
a
n
 SE
I
 
Sc
o
r
e
Veteran 
Non-Veteran
 
 
 
310 
 
   
Figure 10.6: Black and White Veteran versus Non-Veteran SEI 
0
5
10
15
20
25
30
35
40
45
50
1950 1960 1970 1980 1990 2000
Census Year
Du
ncan
 SE
I Score
White Veterans 
White Non-Veterans
Black Non-Veterans
Black Veterans 
 
 Furthermore, we see that there is a World War II veteran peak mobilization phase 
SEI premium that begins with the 1950 census and carries through the 1980 census 
period. In the 1990 and 2000 censuses World War II veterans born to peak mobilization 
cohorts paid occupational prestige penalties. Those in the World War II demobilization 
phase had similar experiences, earning premiums in the 1960 through 1980 censuses and 
paying penalties in the 1950, 1990, and 2000 censuses. This finding also strengthens my 
assertion that in order to fully understand veteran social status attainment, on must look 
beyond the short term and examine the effects of service over the lifetime. 
 However, even though this is the case, hypothesis three does not hold. The only 
period in which veterans of large cohorts earned an SEI premium controlling for 
background factors are 1950 and 2000. Therefore, hypothesis number four does not hold 
and we can conclude that World War II veterans born to cohorts with larger proportions 
311 
 
   
of veterans do not achieve more occupational prestige than those born to cohorts with 
lower proportions of veterans. 
 Furthermore, the bivariate and regression analyses show that veterans earned SEI 
premiums until the 2000 census. The veteran SEI advantage increased from the 1950 
census to the 1960 census, leveled out in the 1970 and 1980 censuses; decreased in the 
1990 census; and dropped precipitously to a penalty in the 2000 census. 
   
Conclusion 
 This dissertation has provided some additional pieces to the puzzle of veteran 
social status attainment by providing a prescription for the underenumeration problems of 
the 1950 decennial census by looking at the social status attainment of World War II 
veterans from the beginning to presumably the end of their working lives. The findings, 
as other research has suggested, are that veterans generally earn a social status premium 
for their service and that the black veteran premiums are higher than the white veteran 
premiums relative to their non veteran peers. 
 Although this dissertation has shed additional light on veteran social status 
attainment it is not without its limitations. This dissertation is limited in that the data do 
not allow for the control of important factors including length of service, military 
occupational specialty, enlisted versus officer status, or duty status (active, reserve, or 
national guard). Additionally, a large portion of the population, both veteran and non-
veteran, is systematically excluded from the sample because they have no income or no 
SEI score (see chapter 3 for details). Future research should endeavor to right these 
problems with the intent of shedding more light on veteran social status attainment. 
312 
 
   
Furthermore, there are some limitations as to how much one can generalize from 
this type of analysis. My findings can only be generalized to the World War II veteran 
population and any attempt to bridge these findings to other veteran populations might be 
specious at best. Furthermore, not only are the findings limited to the World War II 
veteran population they are also limited by several other factors to include race, gender, 
and employment status (see Chapter 3 Methods).  More specifically, during the World 
War II period, for race, only White and Black were relevant categories. Moreover, for 
Blacks, the Army was still segregated and only 6.84 percent of the force was Black. For 
gender, the Decennial Census did not ask any questions pertaining to veteran status of 
women until 1980, so for the hundreds of thousands of women who served in WWII, 
there is no data for the first 30 years of their post service lives. For employment status, as 
the veterans aged, increasingly large groups left the labor force and reported no earned 
income, so that by 2000, only about 20 percent of the original are included in this 
analysis.  
The most evident reason that one should not generalize to populations other than 
the World War II veteran population is that the pattern of veteran effects is more complex 
than the literature on veteran social status attainment suggests.  Some of these 
complexities include but are not limited to how we count veterans, who counts as a 
veteran, macro-economic and historical events, and who the comparison groups are. 
While I make every attempt to explain how I treat these important issues (see Chapter 3 
Methods), others have treated them differently in some cases which could have the effect 
of clouding the interpretation of veteran social status attainment.  
313 
 
   
 
References 
 
 
Alcaly, Roger E. 2003. The New Economy. New York: Farrar, Straus, and Giroux. 
Angrist, Joshua D. 1993. "The Effect of Veteran Benefits on Education and Earnings." 
Industrial & Labor Relations Review 46:637-652. 
Bailey, Duncan and Thomas F. Cargill. 1969. "The Military Draft and Future Income." 
Western Economic Journal 7 (December):365-370. 
Becker, Gary Stanley. 1993. Human Capital : A Theoretical and Empirical Analysis, 
With Special Reference to Education. Chicago: The University of Chicago Press. 
Berger, Mary C. and Barry T. Hirsch. 1983. "The Civilian Earnings Experience of 
Vietnam-Era Veterans." Pp. 455-479 in The Journal of Human Resources, vol. 
18. 
Binkin, Martin and Mark J. Eitelberg. 1982. Blacks and the Military. Washington, D.C.: 
Brookings Institution. 
Bishop, Matthew. 2004, "Economics A-Z ",  Retrieved   
Blair, Julie. 1999. "GI Bill Paved the Way for a Nation of Higher Learners." Education 
Week 18:32. 
Blau, Peter Michael and Otis Dudley Duncan. 1967a. The American Occupational 
Structure. New York,: Wiley. 
?. 1967b. The American Occupational Structure. New York: John Wiley and Sons, Inc. 
Booth, Bradford, William Falk, David R. Segal, and Mady Wechsler Segal. 2000. "The 
Impact and Military Presence in Local Labor Markets on the Employment of 
Women." Gender and Society 14:318-332. 
Bouffard, Leana Allen. 2005. "The Military as Bridging Environment in Criminal 
Careers: Differntial Outcomes of the Military Experience." Armed Forces & 
Society 31:273-295. 
Bound, John and Sarah Turner. 2003. "Going to War and Going to College: Did World 
War II and the G.I. Bill Increase Educational Attainment for Returning 
Veterans?" Population Studies Center, University of Michigan, Ann Arbor, MI. 
Broom, Leonard J. and J.H. Smith. 1963. "Bridging Occupations." British Journal of 
Sociology 14:321-334. 
314 
 
   
Browning, Harley L., Sally C. Lopreato, and Dudley L. Poston, Jr. 1973. "Income and 
Veteran Status: Variations Among Mexican Americans, Blacks, and Anglos." 
American Sociological Review 38:74-85. 
Campbell, Donald T. and Julian C. Stanley. 1966. Experimental and Quasi-Experimental 
Designs For Research Chicago: Rand McNally. 
Charmarette, S. and George Thomas. 1982. Earnings Effects of Military Service, or Does 
It Pay to Be in the Military. Monterey, CA: United States Naval Postgraduate 
School. 
Cohen, Jere, David R. Segal, and Lloyd V. Temme. 1986. "The Educational Cost of 
Military Service the 1960s." Journal of Political & Military Sociology 14:303-
319. 
?. 1992. "The Impact of Education on Vietnam-Era Veterans' Occupational 
Attainment." Social Science Quarterly 73:397. 
Cooney, Richard T. Jr., Mady Wechsler Segal, David R. Segal, and William W. Falk. 
2003. "Racial Differences in the Impact of Military Service on the Socioeconomic 
Status of Women Veterans." Armed Forces & Society 30:53-85. 
Cooney, Richard Thomas. 1997. "The Post-Service Socioeconomic Status Attainment of 
Women Veterans of the All-Volunteer Force." M.A. Thesis, Dept. of Sociology., 
University of Maryland, College Park, MD. 
Cowper, Diane C., Charles F. Longino, Jr., Joseph D. Kubal, Larry M. Manheim, 
Stephen J. Dienstfrey, and Jill M. Palmer. 2000. "The Retirement Migration of 
U.S. Veterans, 1960, 1970, 1980, and 1990." Journal of Applied Gerontology 
19:123-137. 
Cutright, Phillips. 1974. "The Civilian Earnings of White and Black Draftees and 
Nonveterans." American Sociological Review 39:317-327. 
Davis, J. Ronnie and Neil Palomba. 1968. "On the Shifting of the Military: A Progressive 
Tax-in-Kind." Western Economic Journal 6 (March):150-153. 
Defleur, Lois B. and Rebecca Warner. 1985. "Socioeconomic and Social-Psychological 
Effects of Military Service on Women." Journal of Political and Military 
Sociology 13:195-208. 
Detray, Dennis N. 1982. "Veteran Status as a Screening Device." American Economic 
Review 72:133-142. 
Diehl, James M. and Stephen R. Ward. 1975. The War Generation : Veterans of the First 
World War. Port Washington, N.Y.: Kennikat Press. 
Elder, Glen H. 1986. "Military Times and Turning Points in Men's Lives." 
Developmental Psychology 22:233-245. 
315 
 
   
Elder, Glen, H. Jr. and Yoriko Meguro. 1987. "Wartime in Mens' Lives: A Comparative 
Study of American and Japanese Cohorts." International Journal of Behavioral 
Development 10:439-466. 
Fredland, J. Eric and Roger D. Little. 1985. "Socioeconomic Status of World War II 
Veterans By Race: An Empirical Test of The Bridging Hypothesis." Social 
Science Quarterly 66:533-551. 
Gambone, Michael D. 2005. The Greatest Generation Comes Home : The Veteran in 
American Society. College Station: Texas A & M University Press. 
Goldberg, Matthew S. and John T. Warner. 1986. "Military Experience, Civilian 
Experience, and the Earnings of Veterans." The Journal of Human Resources 
22:62-81. 
Hansen, W. Lee and Burton A. Weisbrod. 1967. "Economics of the Military Draft." 
Quarterly Journal of Economics 81(August):395-421. 
Heckman, James J. 1979. "Sample Selection Bias As a Specification Error." 
Econometrica (pre-1986) 47:153. 
Hollingshead, A.R. 1946. "Adjustment to Military Life." The American Journal of 
Sociology 51:439-447. 
Institute of Medicine. 2006, "The National Academy of Sciences-National Research 
Council (NAS-NRC) World War II Veteran Twin Registry",  Retrieved   
Lakhani, Hyder. 1998. "The Socioeconomic Benefits of Active Military Service to 
Reservists." Armed Forces & Society 24:549-562. 
Lipset, Seymour Martin and Reinhard Bendix. 1959. Social Mobility in Industrial 
Society. Berkely: University of California Press. 
Little, Roger D. and J. Eric Fredland. 1979. "Veteran Staus, Earnings, and Race: Some 
Long Term Results." Armed Forces & Society 5:244-260. 
Lopreato, Salley C. and Dudley. L. Poston, Jr. 1977. "Differences in Earnings and 
Earnings Ability between Black Veterans and Nonveterans in the United States." 
Social Science Quarterly 57:750-766. 
Macionis, John J. 1999. Sociology. Upper Saddle River, N.J.: Pearson Prentice Hall. 
Mangum, Stephen L. and David E. Ball. 1989. "The Transferability of Military-Provided 
Occuapational Training in the Post-Draft Era." Industrial and Labor Relations 
Review 42:230-245. 
Martindale, M. and D.L. Poston, Jr. 1979. "Variations in Veteran/Nonveteran Earnings 
Patterns Among World War II. Korea, and Vietnam War Cohorts." Armed Forces 
and Society 5:219-243. 
316 
 
   
Miller, James C. and Robert Tollison. 1971. "The Implicit Tax on Relevant MIlitary 
Recruits." Social Science Quarterly 51 (March):924-931. 
Monahan, Thomas P. 1962. "When Married Couples Part: Statistical Trends and 
Relationships in Divorce." American Sociological Review 27:625-633. 
Moskos, Charles C. and John Sibley Butler. 1997. All That We Can Be: Black Leadership 
and Racial Integration the Army Way. New York, NY: Basic Books. 
Moskos, Charles C. and Frank R. Wood. 1988. The Military: More Than Just a Job. 
Washington: Pergamon-Brassey International Defense Publishers, Inc. 
Oi, Walter Y. 1967. "The Economics of the Draft." American Economic Review:39-63. 
Ortmeyer, Carl E. 1962. "Marriage and Divorce Satistics Programs of the National Vital 
Statistics Division: Current Developments and Research Potentials." American 
Sociological Review 27:741-746. 
Poston, Dudley L., Mady W. Segal, and John S. Butler. 1984. "The Influence of Military 
Service on the Civilian Earnings Patterns of Female Veterans: Evidence from the 
1980 Census." in Women in the United States Armed Forces: Progress and 
Barriers in the 1980's, edited by N. H. Loring. Chicago: Inter-University Seminar 
on Armed Forces and Society. 
Prokos, Anastasia. 1996. "Military Veteran's Earnings: Income Differences of Female 
Veterans and Non-Veteran. Paper presented at the 1996 Annual Meeting of the 
American Sociological Association." 
Reiss, Albert J. 1962. Occupations and social status. New York: Free Press of Glencoe. 
Richardson, Christy and Judith Waldrop. 2003. "Census 2000, Brief Veterans: 2000." 
edited by The United States Census Bureau: 
http://www.census.gov/prod/2003pubs/c2kbr-22.pdf  last accessed 10 December 
2006. 
Rosen, Sherwin and Paul Taubman. 1982. "Changes in Life Cycle Earnings: What Do 
Social Security Data Show?" Journal of Human Resources 17:321-338. 
Ruggles, Steven, Matthew Sobek, Trenta Alexander, Catherine Fitch, Ronald Goeken, 
Ptaricia Kelly Hall, Miriam King, and Chad Ronnander. 2004. "Integrated Public 
Use Microdata Series: Version 3.0 " Minnesota Population Center, Minneapolis, 
MN http://www.ipums.org last accessed 21 February 2006. 
Sampson, Robert J. and John H. Laub. 1996. "Socioecometric Achievement in the Life 
Course of Disadvantaged Men: Military Service as a Turning Point, circa 1940-
1965." American Sociological Review 61:347-367. 
Schwartz, Saul. 1986. "The Relative Earnings of Vietnam and Korean-Era Veterans." 
Industrial and Labor Relations Review 39:564-572. 
317 
 
   
Segal, David R. 1989. Recruiting for Uncle Sam: Citizenship and Military Manpower 
Policy. Lawrence, KS: University Press of Kansas. 
Segal, David R. and Mady Wechsler Segal. 2004. America's Military Population, vol. 4. 
Washington, D.C.: Population Reference Bureau. 
Segal, Mady W. 1986. "The Military and the Family as Greedy Institutions." Armed 
Forces and Society 13:9-38. 
Sider, Hal and Cheryl Cole. 1984. "The Changing Composition of the Military and the 
Effect on Labor Force Data." Monthly Labor Review Online 107. 
Stouffer, Samuel A., Edward A. Suchman, Leland C. DeVinney, Shirley A. Star, and 
Robin M. Williams. 1949. The American Soldier: Adjustment During Army Life, 
vol. I. Princeton, NJ: Princeton University Press. 
Teachman, Jay D. and Vaughn R. A. Call. 1996. "The Effect of Military Service on 
Educational, Occupational, and Income Attainment." Social Science Research 
25:1-31. 
Teachman, Jay and Lucky M. Tedrow. 2004. "Wages, Earnings, and Occupational Status: 
Did World War II Veterans Receive a Premium?" Social Science Research 
33:581-605. 
U.S. Bureau of the Census. 1955. The 1950 Censuses  - How They Were Taken 
Washington, D.C.: U.S. Government Printing Office. 
United States Bureau of Economic Analysis. 2005. "National Income and Product 
Accounts Table." U.S. Department of Commerce Washington, D.C. 
http://www.bea.gov/bea/dn/nipaweb/TableView.asp#Mid  last accessed 5 March 
2006. 
United States Census Bureau. 1945. "Statistical Abstract of the United States 1944 - 
1945: Sixty-Sixth Number " Department of Commerce Washington, DC 
http://www2.census.gov/prod2/statcomp/documents/1944.zip, last accessed 5 
March 2006. 
?. 1949. "Statistical Abstract of the United States  1949: Seventieth Edition." United 
States Department of Commerce Washington, DC  
http://www2.census.gov/prod2/statcomp/documents/1949.zip last accesed 5 
March 2006. 
?. 1964. "Census Population 1960: Vol. I, Characteristics of the Population." United 
States Government Printing Office Washington, D.C 
http://www2.census.gov/prod2/decennial/documents/09768103v1p1_TOC.pdf 
last accessed 27 March 2006. 
?. 2005. "National Defense and Veterans Affairs " Pp. 326-338 in Statistical Abstract of 
the United States: 2004-2005. Washington, DC: Government Printing Office. 
318 
 
   
United States Department of Veterans Affairs. 2005. History of the G.I. Bill - VA 75th 
Anniversary. Washington, DC: 
http://www.75anniversary.va.gov/history/gi_bill.htm last accessed 22 February 
2006. 
Villemez, W. J. and J. D. Kasarda. 1976. "Veteran Status and Socioeconomic 
Attainment." Armed Forces & Society 2:407-420. 
Willett, Thomas D. 1968. "Another Cost of Conscription." Western Economic Journal 6 
(December):425. 
Xie, Yu. 1992. "The Socioeconomic Status of Young Male Veterans, 1964-1984." Social 
Science Quarterly (University of Texas Press) 73:379. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
319