DOCUMENT liESUME

ED 128 745 CG 010 858

AUTHOR 0/Malley, Patrick M.; Bachman, Jerald G.

TITLE Self-Esteem and Educational Attainment: A
Longitudinal Analysis.
PUB DATE 22 Apr 76
NOTE 58p.; Paper presented at the annual meeting of the
American Educational Research Association (San
Francisco, California, April 19-23, 1976)

EDRS PRICE MF-$0.83 HC-$3.50 Plus Postage.

DESCRIPTORS Academic Achievement; *Achievement; Correlation;
Longitudinal Studies; *Males; National Surveys;
Occupations; Research Projects; *Self Esteem; *Self
Evaluation; *Success Factors; *Young Adults

ABSTRACT
This paper explores the relationship between
self-esteem, measured five times during a span of eight years, and
the educational and occupational attainments of a nationwide sample
of over 1600 young men. An overall increase in self-esteem was
observed between the start of 10th grade (1966) and a point five
years after high school (1974) . Surprisingly, 10th grade self-esteem
correlated better with eventual educational attainment (as of 1974)
than did 1974 self-esteem. Longitudinal multivariate analysis led to
the conclusion that factors associated with educational success
become less central to self-evaluations during high school and the
following years. (Author)

***********************************************************************
Documents acquired by ERIC include many informal unpublished
* materials not available from other sources. ERIC makes every effort *
* to obtain the best copy available. Nevertheless, items of marginal *
* reproducibility are often encountered and this affects the quality *
*51_ and hardcopy reproductions ERIC makes available *
* of the micrc
* via the ERIC ument Reproduction Service (EDRS). EDRS is not
* responsible the quality of the original document. Reproductions *
* supplied by EDRS are the best that can be made from the original.
***********************************************************************
LCN

(NI

Self-Esteem and Educational AttainmewL: A Longitudinal Analysis

1=21

iti Patrick M. O'Malley and Jerald G. Bachman

Institute for Social Research, The University of Michigan

U.S. OEPARTMENT OF HEALTH,

EDUCATION & WELFARE
NATIONAL INSTITUTE OF
EDUCATION
THIS DOCUMENT HAS BEEN REPRO
DUCED EXACTLY AS RECEIVED FROM
THE PERSON OR ORGANIZATION ORIGIN
ATING it POINTS OF VIEW OR OPINIONS
STATED DO NOT NECESSARILY REPRE
SENT OFFICIAL NATIONAL INSTITUTE OF
EOUCATION POSITION OR POLICY

Paper presented at American Educational Research

Association meeting, April 22, 1976, San Francisco, California

2
 2

Self-Esteem and Educational Attainment: A Longitudinal Analysis

Most people, including social Scientists, believe that social

personality
environments and experiences can have a lasting impact on

characteristics. But it is also commonly assumed that variations in

personality characteristics lead different people to be exposed to

different -sorts of environments and experiences.

Thus the causal

y
linkage between environments and personality is seen as a two-

street--environments and events shape people, but people also play

experiences.
an important part in selecting and shaping their own

This paper presents findings from a nationwide longitudinal

between
study designed to deal with some of these interrelationships

persons and environments. The study has followed a sample of

adulthood, focusing
young men through late adolescence and early

especially on the impact of major educational and occupational

experiences. Our conceptual approach has acknowledged that, in

differences
the real world, there are important prior individual
never-
which predispose toward different environmental experiences;
still further
theless, we also assume that these experiences lead to

individual differences. Thus we assume that differences correlated

greater
with exposure to various environments and experiences are
(See Bachman, Kahn, Mednick,
after the exposure has actually occurred.
conceptual frame-
Davidson, & Johnston, 1967, for an overview of the
Stated in such a general form, the
work and purposes of the study.)
But the purpose of our research
above proposition may seem indisputable.

3
 Self-Esteem

peen to explore the P roposition in a number of more concrete forms.

ha5
Ole Present analysis, we have selected a dimension of personality
For

1-1 mlnaY illustrate the two-way street of causation particularly

whie
t'le dimension of self-esteem or self-evaluation.

5e1f-esteem has long been the subject of theoretical speculation,

ere recently has been the top ic of several major empirical studies
and
petatnith, 1967; R.os enherg, 1965; Rosenberg & Simmons, 1971).
(coo
who have revieWed the literature in this area
ell (1971) and others
Gers
em has been used in various ways
fInt ed that the terM self-este
haVe
di.
ffer era. authors, our own app roach has been heavily influenced
bY
Like these
tle work of Rosenberg and also that of Coopersmith.
bY
self-esteem to refer to an individual's
Pios, (a) wn use the terra
ant'
vevaluation or judgme nt of his own worth, (b) we treat it as a
sel
and
1) 01 di niension rather than as a number of more specific ones,
glo
uring characteristic rather than
ve view it as a rela tively end
(c)

eOling which shifts abruPtly from one situation to another.

tive on self-esteem has been provided by
Pal additional persP ec

who argued that "The various dimensions used

Froil
c a nd Kahn (1962),
himself and others are not all equally
tile Person for perceiving
bY
They may vary in centrality, defined as
impo("nt to the person.
letermine the person's self-esteem" (p. 19).
the
degree to wh ich the- '

differentially
Mtion that differ cn t aspects of the person may be
Th15
r.fel to self-esteem is also fou nd in the work of Douvan and Gold
cell"
 Self-Esteem

(1967), who noted that one of the more important components of

self for most adolescent boys (in contrast to girls) is a sense of

competence and achievement expressed in competitive fields such

as athletics, work and intellectual activity. The concept of

centrality, as used by these authors, is not incompatible with a

treatment of self-esteem as a single global dimension; it simply sug-

gests that there can be variations and changes in the relative weights--i.e.,

the centrality--of the several components which jointly constitute

global self-esteem.

Our focus in this paper is upon the links between self-esteem and

success--specifically, educational and occupational attainments of

young men in their early twenties. A number of authors have presented

or summarized evidence that self-esteem is linked to educational and

occupational attainment (Coopersmith, 1967; Gergen, 19711 Rosenberg,

1965; Rosenberg & Simmons, 1971; Wylie, 1961), and our own earlier

work provides further support (Bachman, 1970, Appendix D). Particularly

relevant is Purkey's summary of the research relating the self-concept

to school achievement. He concludes that "...there is no question that

there is a persistent relationship between the self and academic

achievement..." And, although the data do nor provide clear-cut

evidence about causal direction, he interprets the findings as showing

"...that there is a continuous interaction between the self and academic

achievement, and that each directly influences the other" (Purkey, 1970, p.23).

5
 Self-Esteem

In the area of occupational success, Luck and Heiss (1972) report

that self-esteem is positively linked to a number of job dimensions,

including income, prestige, upward mobility, and personal satisfaction

with occupational achievement; and Kaufman (1973) reports that loss

of a job resulted in diminished self-esteem among a group of profes-

sional engineers and scientists.

ln sum, there is reason to believe that self-esteem is linked to

educational and occupational attainment, and that this linkage probably

involves a number of different and complexly interrelated patterns

of causation. While it is a rather simple matter to point out the

complexities, the problem of disentangling such reciprocal causation

is extremely difficult and vexing. In most cases, the empirical

evidence is limited to a static relationship at a single point in

time--e.g., survey respondents with higher levels of educational

attainment also have higher mean scores on a measure of self-esteem

(Weidman, Phelan, & Sullivan, 1972). While such findings are important

in demonstrating that a relationship does exist, they leave us largely

in the dark about causal dynamics.

The present study is not limited to a single point in time.

Our data were obtained by surveying a sample of young men (and

measuring their self-esteem) five different times, beginning at the

start of tenth grade and ending eight years later. Thus our measures

of self-esteem span the high school years as well as the five years

6
 Self-Esteem

after high school--a period in which nearly all of our respondents

completed formal education and entered the labor market. Our

measures of educational and occupational attainment are based on

data collected at the end of the study. These data on self-esteem

and attainment, plus a number of additional measures of background,

ability, and other charactsristics, permit us to go several steps

beyond the usual cross-sectional analyses and interpretations.

Before turning to the analysis of these data, let us be more

specific about the relationships we expect to find and the bases for

these expectations. We begin by distinguishing three logically

distinct patterns of causation, all of which may play some part in

the relationship between self-esteem (A) and attainment (B):

A causes B: Self-esteem contributes directly to attainment.

Individuals with positive self-concepts are likely to be ambitious,

i.e., they will set relatively high levels of aspiration. Moreover,

their positive views of their own abilities and competence will help

them to withstand the occasional setbacks and reversals along the

xoad to educational and occupational attainment.

B causes A: Attainment contributes, both directly and indirectly,

to heightened self-esteem. Educational and occupational attainments

represent important sources of direct feedback about the self, and

this may be particularly true for adolescents and young adults.

There are indirect effects as well; those individuals with the most

7
 Self-Esteem

education and the highest status jobs are most likely to be exposed

to "ego-boosting" experiences in their day-to-day activities.

C causes both A and B: Some Of the underlying determinants of

self-esteem are also important determinants of attainment. Academic

ability, past educational accomplishments, and'family socioeconomic

level are all likely to contribute to a young person's self-esteem.

But these factors of background and ability also directly influence

educational and occupational attainment. Thus, an additional reason

for expecting self-esteem to be correlated with attainment is not that

one causes the other, but rather that there is a substantial overlap

in the factors which determine them (the "C" variables).

The three patterns of causation outlined above do not represent

testable hypotheses; rather, they form the basis for deriving a number

of such hypotheses spelled out below. The hypotheses are all stated in

terms of attainment in general; the specific applications to be trusted

here involve educational attainment (amount of schooling completed) and

several aspects of occupational attainment.

Hypothesis 1. Early self-esteem (measured during high school)

is positively correlated with later attainment (based on A causes B

and also on C causes both A and B).

Hypothesis la. Early self-esteem is positively correlated with

later attainment, after statistically controlling background and

ability (based only on A causes B).

8
 Self-Esteem

Hypothesis 2. Later self-esteem (measured five y..tars after high

school) is positively correlated with attainment (based primarily on B

causes A but perhaps also reflecting A causes B and C causes both A and B).

Hypothesis 2a. Later self-esteem is positively correlated with

attainment, after statistically controlling background and ability and

also earlier self-esteem (based only on B causes A).

Hypothesis 3. Attainment (five years after high school) is more

strongly correlated with later self-esteem than with earlier self-esteem.

This hypothesis is based on B causes A, plus our assumption that the

reciprocal pattern of causation between A and B will involve an upward

(or downward) spiraling effect, similar to the notion. that "the rich

get richer and the poor get poorer." Given that one never does a

completely adequate job of statistically controlling other possible

causes (C variables), we consider this hypothesis especially important

as providing clear evidence of the impact of attainment.

Hypothesis 3a. Attainment is more strongly correlated with later

self-esteem than with earlier self-esteem after statistically controlling

background and ability (excludes C causes A and B).

HyRothesis 4. Background and ability are correlated with self-

esteem and with attainment (based on C causes both A and B).

The hypotheses outlined above summarize the kinds of bivariate

relationships (Hypotheses 1, 2, 3, and 4) and multivariate relation-

ships (Hypotheses la, 2a, and 3a) which would be expected, based on

41Ib
9
 Self-Esteem

our assumptions about.the several patterns of causation between

self-esteem and attainment. However, these hypotheses make no

allowance for shifts along the dimension of centrality. To deal

with that omission we offer the following much more open-ended hypothesis.

Shifting Centrality Hypothesis. A more-or-less gradual shift

in the strength of relationship between self-esteem and any factor

of background, ability, or attainment may indicate an increase or

decrease in that factor's centrality for self-esteem. For example,

we might expect that things having to do with occupational attainment

would become increasingly important to young men in general as they

leave school and enter the job market. On the other hand, things having

to do with educational success might become less important, at least

among those who do not continue their education beyond high school.

Method

Sample

The data for this report came from the Youth in Transition

project, a nationwide longitudinal study of young men. Details of

the design can be found in several earlier reports (Bachman et al,,

1967; Bachman, 1970; Bachman, Green, & Wirtanen, 1971). The sample

is a multi-stage probability sample, clustered by school, of all tenth-

grade boys in public high school in 1966 in the 48 contiguous states.

Data collections took place in Fall of 1966 (tenth grade), Spring of

1968 (eleventh grade), Spring of 1969 (twelfth grade), Spring of

 Self-Esteem

10

1970, and Spring of 1974. The first four data collections consisted

of interviews and questionnaires administered by professional inter-

viewers on the staff of the University of Michigan's Survey Research

Center; the fifth was a self-completed mail questionnaire.

Of the original selected sample of 2,277 boys located in 87

schools, data were collected from 2,213 (97.2%) in 1966. In the

fifth wave (1974), sample attrition reduced the number of completed

questionnaires to 1,628 (73.5% of the original respondents). For

purposes of this paper we will limit the sample to only those 1,628

men whose participation in the study extended through 1974.

The sample attrition has, of course, reduced the generaliza-

bility of the results. While the 585 who responded in 1966 but not

in 1974 do differ from the 1,628 retained respondents along several

dimensions, we believe the sample remains reasonably representative of

the original population, particularly with regard to relationships

among variables. (See Bachman et al., 1971, pp. 18-19, and Johnston,

1973, pp. 231-239, for further discussion on this point.) Along

the dimension of primary interest for this report, self-esteem, the

1966 scores of the 585 panel dropouts averaged insignificantly higher

(by less than 10% of a standard deviation) than the corresponding

scores for the 1,628 retained respondents.

Measures

The ten item measure of self-esteem is close to that used by

Rosenberg (1965). The first six items in Table 1 were adapted directly

11
 Self-Esteem

11

from his scale; the other four items, similar in content, were

developed by Cobb, Brooks, Kasl, and Connelly (1966). Respondents

were asked to indicate on a five-point scale how often each item

was true for them. The five response categories--almost always,

often, sometimes, seldom, and neVer--were coded from 1 to 5, with

higher.values assigned to the higher self-esteem responses. The

scale is nearly balanced with six positive items and four negative

items. The self-esteem index is an unweighted mean of the ten items,

with up to two missing values allowed. Table 1 includes the means,

standard deviations, and item index correlations (uncorrected for

part-whole inflation) for 1966 (Time 1) and 1974 (Time 5).

While there is some variability, the item-index correlations are

fairly similar across items; generally, the correlations for 1974

are slightly stronger than those for 1966. Factor analyses performed

on both the 1966 and 1974 items revealed a strong first factor in

each case, explaining 64% of the 1966 common variance and 69% of

the 1974 common variancc, The item loadings ranged from .38 to .69.

Insert Table 1 About Here

Coefficient alphas are .75, .76, .79, .80, and .81 for 1966,

1968, 1969, 1970 and 1974, respectively. A single test-retest

reliability was estimated using a path analysis approach as developed

12
 Self-Esteem

12

by Heise (1969). This estimate, which assumes that the reliability

is the same at each time, is .75. Test-retest reliabilities were

also computed for each pair of measurements, using formulas suggested

by Heise (1969). (If three measurements are made at Times A, B, and

C, the reliability is computed as r AB r BC /r AC. ) With five measure-

ment points, there are ten possible combinations of three measure-

ments. Each of the ten reliability coefficients was computed; the

range is .69 to .74, with a mean of .71. All these estimates of

reliability, both internal consistency and test-retest, seem

acceptably high. Stability coefficients--to be discussed in the

results section--were also computed. (A stability coefficient is

the correlation between true scores at one time with true scores

at another time.)

While the reliability and stability of a measure can be esti-

mated in several w.ays, validity is another matter. The kind of

validity appropriate for a variable like self-esteem is construct

validity (Wylie, 1974,.p. 38). Self-esteem we hypothesize, should

relate to other variables in certain directions, and this provides

a possible source of evidence on construct validity. Self-esteem

(1966) correlates in the expected directions with measures of:

intellectual ability (.21), somatic symptoms (-.34), negative

affective states (-.52), happiness (.54), rebellious behavior in

school (-.33), and needs for self-development (.44) and social

13
 Self-Esteem

13

approval (.29). See Bachman (1970, especially pp. 242-243) for

more information on these correlations and the measures.

Other measures used in this report will be defined briefly;

more complete information can be found in Bachman et al. (1967)

and Bachman (1970).

Educational attainment as used in this report is a six-

,
category variable based on data collected in 1974: (1) high

school dropouts, i.e., those without a high school diploma,

(2) high school graduates with no further education, (3) those

with some college e-Terience but no degree, (4) those with an

associate degree, (5) those with a bachelor's degree, and (6) those

with a bachelor's degree plus some graduate work.

Socioeconomic level, which refers to the respondent's family

background, is an equally weighted mean of the following six items:

status of father's occupation (Duncan, 1961), father's education,

mother's education, a checklist of possessions in the home, number

of books in the home, and the ratio of rooms per person in the home.

Ability is a mean of three equally weighted measures of

intellectual ability--Quick Test (Ammons & Ammons, 1962), Gates

Test of Reading Comprehension (Gates, 1958), and the General Aptitude

Test Battery Part J, Vocabulat,.

College plans is a dichotomy, coded 1 if the respondent indicated

intentions to attend college, 0 otherwise.

Self-concept of school ability is an index of three items

which asked the respondent to rate himself, compared to his peers,

14
 Selthsceem
14

on school ability, intelligence, and reading ability.

Grades were ascertained by self-report; respondents were

asked to report an overall average grade for the previous year.

The 1966, 1968, and 1969 measures refer to ninth, tenth, and

twelfth grade grades, respectively. (Actual grades in Grade 12 obtained

from school personnel correlated .70 with self-reported grades for

a sub-sample of 766 for whom the actual grades were later ot...;..-ined.)

Rebellious behavior in school is a mean of thirteen items

dealing with disruptive behavior in school, rule-breaking, and

poor school work.

Delinquent behavior in school is similar to rebellious

behavior but the seven behaviors asked about are more serious.

Status of aspired occupation is the Duncan (1961) status

he might
score assigned to the work that the respondent thinks

do for a living in the long run.

Status of attained occupation is the Duncan status score assigned

to the respondent's actual occupation during the fifth data collection

(1974) or, for those not employed at that time, their most recent

occupation.

Statistical Significance

The multi-stage sampling design of the present study produces

larger sampling errors than would a simple random sampling design

(Kish, 1957), making the use of the usual tables of significance

values inappropriate. The design effect can be estimated, however,

and used to adjust downward the actual number of cases. The adjusted

frequency is then used in the usual tables. The design effect of the

statistics in this report is estimated to be about 2.25 (Kish & Frankel,

15
 Self-Esteem

15

1970), so in all assessments of statistical significance, the frequenciesi

used are equal to the actual n's divided by 2.25.

Results

Our presentation of findings proceeds through several stages. We

begin with an overview of changes in self-esteem scores across time.

Next, we consider bivariate relationships, focusing primarily on the

links between attainment measures and the five different self-esteem

measurements. Finally, we turn to multivariate analyses. As we move

through these stages of analysis we will note implications for specific

hypotheses; however, we will not undertake any summary evaluation of

hypotheses and causal linkages until we reach the discussion section.

Overall Changes in Self-Esteem

Self-esteem scores were fairly high for our respondents at the

start of tenth grade, and gradually rose to a level one standard

deviation higher during the nearly eight-year span of the longitudinal

study. The mean scores included in Table 2 show a modest increase in

self-esteem throughout high school (1966 to 1969), rather little

change during the following year (1969 to 1970), and a more substantial

increase during the next four years (1970 to 1974). The pattern of

change is consistent across all ten of the items in the self-esteem

scale; Table 1 shows for each item an increase in mean score and a

decrease in standard deviation from 1966 to 1974. The decrease in

standard deviation may reflect a "ceiling effect" at the item level,

1(i
 Self-Esteem

16

since the mean score for each item moved closer to the upper limit of

self-esteem. At the index level, however, there is less evidence of

a "ceiling effect," since the standard deviation dropped only slightly

from .52 to .48.

Insert Table 2 About Here

The pattern of correlations among self-esteem measures, presented

in Table 2, suggests a process of gradual evolution and shift throughout

the period we studied. As the table indicates, the longer the time

interval between any two self-esteem measurements, the lower the

correlation. The table also includes stability coefficients, which

are estimates of the "true " cross-time correlations after correcting

for unreliability in the measures (Heise, 1969). Again, the longer

the time interval, the lower the stability.

Links to Educational Attainment and Related Factors

Educational Attainment. Self-esteem scores for six categories of

educational attainment, ranging from high school dropouts to those in

graduate school are presented in Figure 1. Consistent with Hypotheses

1 and 2, we find a positive correlation between self-esteem and level

of educational attainment--the higher the level of education a respondent

eventually attained, the higher was his self-esteem throughout the course

of the study. Indeed, the degree of consistency in the relationship is

17
 Self-Esteem

17

the
start ef tenth grade to a point nearly eight
tro 1Y. stl-jking; From

yeaft; later , the ordering of mean self-esteem scores (from lowest to

hoest) is jdonticni uo the ordering of educational attainment

vho became high school dropouts always lowest
cae5(pris, with those

jsieraY?.,- self-esteem and those

who went on to graduate training always
in

st (differences among the six group means are significant at

_001 for all five times). moreover, as the eta values and product-
/

erlt correlations au the bottom of Figure I indicate, the relationship

noIT'

hcoieen ;,;,21j_est,,,em and educational attainment (as we scaled it) is

a/510-4t Perfectly linear.

The for sel f-esteem is basically upward, with

tre nd over time

largely parallel pattern of increases in mean

.r) subgroup showing
eac.

cs. One exception, while not extremely large, is noteworthy;

elf-esteem ef high school graduates who-did not continue their

the
whereas
(Group 2) rose during the year following high school,
edti

core5 for dropouts (Group 1) declined slightly during the same period.
the
5elf-esteem for the dropout group was mostly of
4q relat ive
Tb
4ced duration; fo.: years later ( in 1974) the difference between
liw
the same level as occurred during
Groop8 1 and 2 had shru nk to nearly

first three data co llections.

the
The most surprisi ng finding shown in Figure 1 is the fact that

the beginning of tenth grade correlates

si-
educational attainment than does self-esteem
e str ongly n eventual
 Self-Esteem

18

measured after the educational attainment levels had been reached.

This tendency for the self-esteem trend lines to converge during

high school and the years that followed is evident in the figure

itself (especially if we disregard the relatively small Group 1), and

it is also reflected in the drop in correlation values (from r = .26

in 1966 to r = .15 in 1974, p < .01). We will discuss this finding

at some length later; for the present, it is sufficient to note

that a steady decrease in self-esteem differences among educational

attainment subgroups is quite the opposite of what we projected in

Hypothesis 3, assuming a pattern of reciprocal causation between

early self-esteem, educational attainment, and later self-esteem.

Other Factors Linked to Self-Esteem. Given the fairly strong link

between early ;1966) self-esteem scores and later educational attain-

ment, it is important to examine the several possible causes for

such a relationship. One possibility is that early self-esteem is

indeed a determinant of later educational success (A causes B:

Hypotheses 1 and la). Another possibility is that factors of background,

ability, past school performance, and aspirations, all of which may

contribute to educational attainment, are also among the causes of

self-esteem (C causes both A and B: Hypothesis 4). We measured a

number of such dimensions at several points throughout the study; their

correlations with self-esteem (at all five points in time), as well

as their correlations with educational attainment, are presented in

Table 3. (Also included in Table 3 for comparison purposes are three

19
 Self-Esteem-

19

measures of attainment as of 1974: level of education completed, employ-

ment versus unemployment, and status of occupation.)

Insert Table 3 About Here

We can begin our examination of Table 3 by noting the correlations

hi the fiolumn which is second from right; they indicate that all of the

ur inluded in the tAblo are linked to educational attainment. Many

ot: the relationships, e.g., those involving academic ability (test scores)

and classroom grades, are quite substantial.

The more important finding displayed in Table 3 is that each of the

dimensions included in Table 3 is correlated with self-esteem. Of

particular interest is the fact that in every instance the early measnres

of self-esteem, usually the 1966 scores, show the strongest correlations,

whereas the weakest correlations involve self-esteem as measured in

1974. It is worth noting that the self-concept of school ability

measured in 1968 shows the highest correlation with self-esteem measured

at the same point in time, and this is also true for the 1968 measure

of rebellious behavior in school (but not for the 1968 measure of

delinquent behavior in school). We were not surprised to find some

heightening of correlations among variables measured in the same

data collection; on the contrary, the surprising fact is that this

"time matching" phenomenon does not appear more often in Table 3. Other

dimensions measured at multiple points in time, such as college plans

20
 Self-Esteem

20

(Times 1 through 3) and occupational aspirations (Times 1 through 5),

show little or no "time matching" effect; instead, the overwhelming

tendency is that plans or aspirations, no matver when they were

measured, show the strongest association with tenth grade (1966)

self-esteem and progressively weaker relationships with later

2
self-esteem.

Two implications drawn from these findings are particularly

relevant for our exploration of the linkage between success and

self-esteem. First, consistent with Hypothesis 4, it seems clear that

those factors which help to determine later educational attainment

are also determinants of self-esteem. Second, it appears that these

educationally-relevant factors make a significant contribution to self-

esteem during the early high school years, but this contribution becomes

steadily less important as young men progress through high school and

continue into other educational and/or occupational -nvironments.

These implications are entirely consistent with our earlier findings about

self-esteem and educational attainment (Figure 1). Taken as a whole, the

pattern of results suggests that things having to do with educational success--

academic skills, past classroom performance, future aspirations, and

the like--undergo some reduction in "centrality" or "salience" for

the self-esteem of young men during the late high school years and

the period that follows.

21
 Self Esteem

21

We noted earlier, when we introduced the shifting centrality

hypothesis, that a reduction in the centrality of academically relevant

factors might take place among the young men in our study who did not

continue their education beyond high school, whereas among those who

did go on to college the centrality of such factors might remain high.

The correlations shown in Table 3 were examined separately for four

categories of respondents: (a) those who initially (at the start

of tenth grade) planned to go to college and later did, (b) those

who planned to go but did not, (c) those who did not initially plan

on college but later did attend, and (d) those who neither planned

to attend nor did. A detailed review of the findings of this analysis

is beyond the scope of the present paper; however, we can report that

there was no indicatton that the centrality of academic factors

dropped more sharply among those who did not go on to college

(categories b and d):'

Impact of Occupational Experiences

Occupational success is a more complicated concept than educational

success. In this paper we deal with two aspects of occupational

experience: employment versus unemployment, and occupational status

3
(Duncan scale).

Unemployment. Table 4 contrasts self-esteem scores for those

who were unemployed at the 1974 survey and those who were employed

full-time or part-time in the civilian work force (full-time students

 Self-Esteem

?')

The bottom
and those in military service are omitted from the table).

line of the table indicates only small and non-significant differences

in self-esteem between the two groups until 1974, at which point the

difference reaches about one-third of a standard deviation. Although

this difference is not so large as to suggest a devastating impact

failure
of unemployment upon self-esteem, the finding does indicate that

to have a job has some negative effect on self-esteem.

Insert Table 4 About Here

The negative impact of unemployment may be linked to educational

level. Those who have followed the socially-approved path of high school

graduation and perhaps also college have "done their part," so to

their own
speak, and thus may be less inclined to view unemployment as

fault. High school dropouts, on the other hand, have followed a path

proclaim that dropping out

which is disapproved; public announcements

"doubles the chances of being unemployed," and the unemployed dropout

We
may thus be more likely to blame himself for his predicament.

examined the impact of unemployment (as of 1974) separately for three

levels of educational attainment: dropouts without high school diplomas,

entered
high school graduates with no further education, and those who

college (including those who completed associate and bachelor degrees).

Among the college entrants, the mean gain in self-esteem scores from

23
 Self-Esteem

'73

1966 to 1974 is .44 for the employed (N=659) and .38 for the unemployed

(N=49); this difference is small and of little consequence. Among the

high school graduates who did not enter college, the mean gain in self-

esteem is .58 for the employed (N=-.387) and .50 for the unemployed (N=39)--

again, a rather small difference. Among high school dropouts, on the

other hand, the difference is more substantial; the mean gain in self-

esteem is .62 among the employed (N=83) but only .31 among the unemployed

(N=18). While this difference falls short of statistical significance

given the small number of unemployed dropouts, it is certainly in a

direction consistent with the explanation outlined above.

In sum, we find for the sample as a whole, that the dimension of

employment-unemployment seems to have an impact on self-esteem measured

at the same point in time (1974) but little relationship to earlier

measures of self-esteem. This pattern of findings is consistent with

Hypotheses 2 and 3 (B causes A). On the other hand, Hypothesis 1

which posits an impact of self-esteem upon attainment (A causes 8), is not

supported when we treat employment (versus unemployment) as a measure

of attainment. Our analysis of educational subgroups produced some

differences which, although not statistically significant, suggest that the

negative impact of unemployment on self-esteem may be more pronounced

high school.
among high school dropouts than among those who have completed

Occupational status. Given our initial theorizing about the positive

impact of success on self-esteem, our expectation was that thbse young

24
 Self-Esteem

24

men who attained higher status jobs would show above average gains in

self-esteem; thus we expected that occupational status in 1974 would

be positively correlated with self-esteem in 1974, and that the

relationship with earlier measures of self-esteem would be weaker (Hypothesis

3). The correlations presented in Table 3 show that occupational

status in 1974 has basically the same strength of association with self-

esteem no matter when self-esteem was measured. (The slightly higher

correlation with 1970 self-esteem is not significantly different from

the other correlations and may represent nothing more than chance

variation.) One of the problems with these zero-order correlations

is that a number of the factors which may be among the causes of

educational attainment and self-esteem are also related to occupational

status. Thus we may be able to get a clearer picture of what occupational

success adds to self-esteem if we introduce statistical controls

for these other prior influences.

Multivariate Analyses

Controlling background and ability. Of all the dimensions shown in

Table 3, the first few--family socioeconomic level, academic ability, and

ninth grade classroom grades--seem particularly appropriate to treat as

factors which may play a causal role in determining both educational

attainment and self-esteem. The other dimensions in Table 3, such as

self-concepts of ability, patterns of delinquent or rebellious behavior

in school, educational and occupational aspirations, may also play a

25
 Self-Esteem
95

and also
causal role; however, their position in a causal sequence,

their theoretical relationship to a global measure of self-esteem,

are a good deal more debatable. Accordingly, we will take the

somewhat conservative approach of controlling only grades, ability

what may be
and familiy socioeconomic level in our efforts to learn

added to self-esteem by educational and occupational success.

The necessary data are provided in Table 5. The first row of

predicting self-
the table presents multiple correlation coefficients,

esteem at each point in time from our measures of background, ability, and

grades. Just as we found when we looked at these measures separately in

from these
Table 3, we see that self-esteem was most predictable
the eight
factors in 1966, and became steadily less predictable in

years that followed. But the question of greater interest to us

and our
here is what happens to the relationships between self-esteem
when we statistically
measures of educational and occupational attainment

control background, ability and grades.

Insert Table 5 About Here

between
The second row in Table 5 presents the partial correlations
background, ability
educational attainment and self-esteem, controlling

and grades. The partial correlations are much smaller than the zero-

order relationships shown in Table 3. Moreover, there is much less

and
evidence of a shift in correlation between educational attainment

26
 Self-Esteem

26

self-esteem; educational attainment has partial correlations of .10 with

1966 self-esteem and .07 with 1974 self-esteem. This finding provides

little support for either Hypothesis la (A causes B) or Hypothesis 2a

(B causes A). On the other hand, the multiple correlation data in

the first row of the table fit Hypothesis 4 (C causes both A and B);

additionally, the downward trend in multiple correlations with self-

esteem from 1966 to 1974 is fully consistent with the shifting centrality

explanation that those factors of background and ability which lead

eventually to educational success are more central to self-esteem

during the early high school years than during the period which

follows.

The third row in Table 5 shows the partial correlations between

1974 occupational status and all five self-esteem measures, with

background, ability and grades controlled. Again, the partial

correlations are lower than the zero-order relationships (in Table 3),

but the reduction is not nearly so dramatic as in the case of educational

attainment. Most important, the partial correlations show a small but

theoretically important shift: job status in 1974 is more closely linked

to self-esteem in 1970 and 1974, than to self esteem in 1966, 1968

and 1969. This finding is consistent with Hypotheses 2a and 3a (B causes A).

The fourth row in Table 5 presents partial correlations between

employment-unemployment in 1974 and the five self-esteem measures, again

controlling background, aLilitN and grades. In this case, the controls

27
 Self-Esteem

make very little difference; the partial correlations are quite

similar to the zero-order relationships shown in Table 3. Employment

(versus unemployment) in 1974 shows a small positive correlation with

1974 self-esteem but no correlation with earlier measures of self-

esteem. This pattern of findings, like the findings for occupational

status, is consistent with Hypotheses 2a and 3a and indicates a small

positive impact of occupational attainment upon self-esteem.

Controlling background, ability, and initial self-esteem. Our

analysis approach thus far has given basically equal treatment to

the five self-e:;teem measurements taken throughout the course of the study.

Now we will limit our focus to self-esteem at the end of the study (1974)

and attempt to isolate any unique contribution that can be attributed

to educational and/or occupational attainment. The technique again is

partial correlation, predicting 1974 self-esteem from our three measures

of attainment or success: (a) educational attainment, (b) status of

attained occupation, and (c) employment (versus unemployment). Table

6 shows what happens to the relationship between each of these variables

and self-esteem as we introduce an increasing number of statistical

controls: first we control the background, ability, and school perfor-

mance measures available in 1966; then we also control 1966 self-esteem

scores; and finally we extend the controls to include grades and self-

esteem measured in 1968 and 1969. (This controlling for earlier self-

esteem provides the most thorough test of hypothesis 2a.)

Insert Table 6 About Here

28
 Self-Esteem

28

Looking first at the column of relationships between educational

attainment and self-esteem we see that the original (zero-order)

correlation is cut almost in half simply by controlling initial

background, ability, and school performance measures. When we

control also for initial self-esteem scores, the partial correlation

becomes still smaller; and when we extend the controls to include self-

esteem and performance measures throughout the high school years

(1966 through 1969), the partial correlation between educational attain-

ment and 1974 self-esteem is a trivial .042. (If we also control 1974

occupational status, the partial correlation between educational attainment

and self-esteem shrinks still further to .020.) In short, once we

control background and ability differences, plus grades and self-esteem

scores throughout high school, we find virtually no unique contribution

of educational attainment to self-esteem five years after high school.

This clearly fails to support Hypothesis 2a for the dimension of

educational attainment.

The findings for occupational factors are somewhat different.

Looking next at the relationship between status of attained occupation

and self-esteem, we find that the initial correlation is reduced

appreciably when background, ability, and grades measures are controlled;

however, the further controls for self-esteem and additional grades measures

throughout high school produce virtually no further change in the

partial correlation between job status and self-esteem. (Indeed,

even when we also control educational attainment, the partial correlation

between job status and self-esteem shrinks less than one point to .085.)

29
 Self-Esteem

Neither the partial correlation, nor the unadjusted (zero-order)

correlation between occupational status and self-esteem, is very

strong; thus it would be inaccurate to claim a large impact for

job status. Nevertheless, it appears that occupational status

five years after high school does make a significant, albeit modest,

unique contribution to self-esteem. Thus, the findings along this

dimension are consistent with Hypothesis 2a.

Turning finally to the impact of unemployment, we find that

the initial (zero-order) correlation between the employment measure

and 1974 self-esteem is basically unchanged by partialling out back-

ground, ability, grades, and earlier self-esteem measures. Time 5

unemployment is virtually uncorrelated with earlier self-esteem and

with the other factors that are correlated with self-esteem; thus

the unemployment effect on self-esteem seems entirely unique, and

represents the clearest instance in the present data of a contempo-

raneous environmental factor--the "unemployment environment"--showing

an impact on selt-esteem (consistent with Hypothesis 2a). It should

be added that the correlation of about .10 between employment and

self-esteem may tend to understate the impact of being out of a job.

Only a relative]y small proportion of respondents (8.4%) were

unemployed in 1974, but their self-esteem scores were appreciably

lower than scores for the employed (see Table 4).

30
 Self-Esteem

30

Discussion

We stated at the outset of this paper that self-esteem is likely

to be linked to educational and occupational attainment--probably

through a number of complexly interrelated patterns of causation.

That general expectation has certainly been confirmed by our analyses,

although some of our specific hypotheses about causal connections

have not. Near the end of this section we draw some conclusions

about causation. Before that, we discuss our findings on changes in

self-esteem, and links to both educational and occupational outcomes.

Change and Stability in Self-Esteem

Our findings suggest that self-esteem is not a characteristic of

personality that is firmly fixed by the time a young man enters high

school. On the contrary, we found a gradual but quite substantial

rise in average self-esteem throughout high school and particularly

during the five years following high school. Our analysis was not

designed to uncover the probable cause of this rise. However, it

does seem reasonable to rule out the notion that leaving high school

produces a sudden and marked shift in self-evaluation; mean self-esteem

scores did not show a greater than usual increase during the first year

following high school, and stability estimates are virtually identical

for the one-year periods that preceded and followed graduation.

Perhaps the gradual rise in self-esteem scores among young men

during this period simply reflects their increasing maturity and the

resulting increase in status, opportunities,and privileges.

31
 Self-Esteem

31

Although we found a substantial amount of change in self-esteem

over time, we also found a good deal of stability in scores from orle

year to another. We estimate that, among young men in their late

teens and early twenties, self-esteem (after adjustments for measure-

ment unreliability) has a stability of nearly .9 for one-year intervals.

Over longer periods the stability is proportionately lower, so that

for the total eight-year span of the study we estimate the stability of

self-esteem to be slightly over .4. These findings concerning stability,

coupled with the fact that the overall rise in scores was quite gradual,

provide some support for the view of self-esteem as a relatively enduring

characteristic rather than something which shifts abruptly from one

situation to another. Change certainly does occur during and following

late adolescence, but the change seems to be gradual and developmental

rather than revolutionary.

The Changing Link with Education

We found self-esteem linked with educational attainment; however,

the linkage is complex and shows evidence of change over time. Those

young men in our sample who eventually completed college and entered

graduate training showed the highest mean self-esteem; those who never

attained the high school diploma had the lowest mean self-esteem, and

the groups between these two extremes had mean self-esteem scores

which neatly matched their level of educational attainment. The finding

that self-esteem is positively related to educational attainment is

 Self-Esteem

32

not surprising. But,the relationship is not strongest and clearest

at the end of the study, when individuals had sorted themselves into

the several levels of educational attainment; instead, the differences

are strongest among self-esteem scores obtained at the beginning

of the study, apparently "anticipating" educational attainment.

And that is surprising.

One plausible explanation for this tendency for self-esteem

scores during high school to anticipate later educational attainment

is that self-esteem is among the causes of such attainment--the A causes B

interpretation which is reflected most clearly in Hypothesis la.

Our multivariate analyses provide only very limited support for this

interpretation; the partial correlations between high school self-

esteem and later educational attainment, with background, ability,

and earlier school performance controlled, range from .09 to .11.

While such relationships remain statistically significant, they are much

lower than the corresponding zero-order relationships; moreover, we

suspect that if our set of control variables were more extensive

or more perfectly measured, the partial correlations between

high school self-esteem and later educational attainment might be

reduced nearly to zero.

We should also note that the multivariate analyses provide no

support at all for the notion that educational attainment contributes

to self-esteem--the B causes A interpretation. The partial correlations

33
 Self-Esteem

33

between self-esteem and educational attainment fail to confirm either

Hypothesis 2a (see Table 6) or Hypothesis 3a (see Table 5).

The analyses indicate rather clearly that the primary basis for

the correlation between high school self-esteem scores and later

educational attainment is the fact that those aspects of family

background, academic ability, and past school performance which

predict later educational success also play an important part in the

self-esteem of young men in tenth grade. This is the C causes both

A and B interpretation reflected in Hypothesis 4. In addition,

it appears that factors of background, ability, and past school

performance become less and less important for self-esteem as a

young man continues through high school and beyond (Table 5, top

row). This pattern of declining importance for self-esteem appears

with great consistency across quite a number of dimensions--family

socioeconomic level, test scores, grade-point averages, rebellious and

delinquent behaviors in school, and educational and occupational

aspirations. (The very uniformity of this pattern, shown in Table 3,

might suggest that the later self-esteem scores are simply less

"predictable," no matter what variables we attempt to correlate with

them; however, our other findings showing the impact of unemployment

and occupational status on later self-esteem rule out such an explanation.)

These declining correlations with self-esteem can be interpreted

within the Shifting Centrality Hypothesis; those attritubes of self-

identity which have to do with conventional educational success have less

34
 Self-Esteem

34

centrality, and thus less impact on self-esteem, as young men move

through the final years of high school and go on to other experiences.

This decreasing centrality of the factors linked to academic success

is not limited to those who end their education with high school; it

holds true also for those who enter college. In this connection it

is worth noting a finding from one of our earlier analyses; the value

placed on striving for academic achievement--studying hard to get

good grades and academic honors--showed a decline throughout the

l-igh school years, and that decline was steepest among those who

later went on to college (Bachman, Green, & Wirtanen, 1971, pp. 118-120).

We conclude that, at least among young men who completed high school in

the late 1960's, educational success became a less vital part of the

self; and this was particularly true among those who were most

successful, the ones who went on to college. We discuss later the

questim: of whether these conclusions about the shifting centrality

of educational factors can be generalized to other time periods, and

to young women as well as young men.

impact of Occupational Attainment

One dimension of occupational success is status or prestige.

Our findings, like the findings of many other researchers, show a

substantial correlation between educational attainment and job status.

Additional analyses (not reported here) have shown that the same

dimensions of family background, ability, And aspirations which predict

35
 Self-Esteem

35

educational attainment also predict job status. Thus we would

expect to find at least some association between occupational status

self-esteem--especially during the
and self-esteem simply because

early high school years--is linked to family background, ability,

and aspirations. The more important question is whether the job

status that a young man attains in his early twenties makes any

additional contribution to his self-esteem.

indeed
Our multivariate analyses indicate that job status does
self-esteem. The
make a unique, though very modest, contribution to
background and
data in Tables 5 and 6 show that after controls for
self-esteem,
ability, and even after controls for earlier measures of
contribution to
the status of a young man's job makes some marginal

his present level of self-esteem.

occupational attainment,
Another and perhaps more basic dimension of
simply having a
particularly during a time of high unemployment, is
provided the
job. We stated earlier that employment versus unemployment
environmental factor--
clearest instance in our data of a contemporaneous

environment"--having an impact on self-esteem. Those

the "unemployment
the time of the last data
young men in the study who were unemployed at
Moreover, the
collection showed lower than average self-esteem scores.
self-esteem may
findings suggest thar the impact of unemployment upon
high school diploma.
be felt most heavily by those who fail to attain a
for
Perhaps these young men are most inclined to blame themselves
is something which
their unemployment, since their lack of a diploma

36
 Self-Esteem

36

many sources constantly remind them that they could--and should--take

steps to correct. (For a discussion of the "anti-dropout" campaign

in the media, see Bachman et al., 1971, and Bachman, 1972.)

In sum, we conclude that occupational attainment, reflected in

simply having a job, and also in the status of that job, has an

impact on the self-esteem of young men in their early twenties.

These findings support the B causes A interpretation, reflected in

Hypotheses 2a and 3a.

Conclusions About Causation

We have already drawn several conclusions about the causal

connections between attainment and self-esteem. Let us summarize

them more systematically here, noting the extent to which each of

our hypotheses and the underlying assumptions about causation has been

confirmed or disconfirmed.

First of all, we find very little evidence to support the view

that self-esteem during high school makes a unique causal contribution

to later educational and occupational attainment, (A causes B), at

least as we have measured these dimensions. Positive zero-order

correlations appear between early self-esteem and later measures of

both educational and occupational attainment, consistent with Hypothesis

1; however, the more stringent partial correlational analyses show

little or no relationship, thus providing little support for Hypothesis la.

37
 Self-Esteem

37

Second, we do find evidence that factors of background, ability,

and prior experience influence both self-esteem and later attainments--

specifically level of education attained and status of attained

occupation. Thus there is some support for Hypothesis 4 (C causes

both A and B), particularly for the dimension of educational

attainment.

Third, we note that the strength of connection between other

variables ("C" variables) and self-esteem shows some shift over

time, consistent with the Shifting Centrality H'rpothesis.

Fourth, the evidence sugges s that educational attainment, at

least in terms of the amount of schooling completed, makes no

independent contribution to self-esteem five years after high school.

Although a zero-order correlation exists between educational attainment

and self-esteem, consistent with Hypothesis 2, the relationship does

not survive statistical controls for background, ability, past

school performance, and earlier self-esteem; thus the findings fail

to support Hypothesis 2a. Moreover, the pattern of correlations with

self-esteem across time is opposite to that predicted by Hypothesis 3

(for zero-order correlations) or Hypothesis 3a (for partial correlations).

Finally, we find that occupational attainment makes a modest

but apparently unique, contribution to self-esteem (B causes A).

For both dimensions, employment (versus unemployment) and

status of attained occupation, we find a relationship with 1974

38
 Self-EsterT1

38

self-esteem, even after controlling for background, ability, past

school performance;and earlier self-esteem. This finding is consistent

with Hypothesis 2a. Furthermore, when we examined the relationships

with self-esteem at all five points in time, controlling background

and ability and past school performance,we found the attainment

measures more strongly related to later self-esteem than to earlier

self-esteem. This pattern of findings is predicted by Hypothesis 3a.

In sum, only a portion of the hypothesized relationships

between attainment and self-esteem were found to exist in the present

study. The evidence of any causal impact of early self-esteem on

later educational attainment is weak at best, and there is no

evidence of an impact on occupational attainment. On the other hand,

occupational success shows a positive impact, albeit quite small,

on self-esteem, whereas educational attainment shows no such

independent effect. It appears that much of the association between

attainment and self-esteem is best explained as reflecting a common

set of prior causes--background, ability, and earlier scholastic

success.

Some Limitations to These Conclusions

The ability to generalize from any single study, no matter how

extensive, is always limited. Some limitations to the present work

have already been mentioned; others are obvious. Most notably, our

sample was limited to young men, and thus we cannot be sure that the

39
 Self-Esteem

39

same relationships between self-esteem and educational and occupational

attainment would be found for young women. Douvan and Gold (1967),

in their review of the literature on adolescent self-esteem, concluded

that the self-esteem of boys and girls depends to some extent on

different components, and that boys are more likely than girls to

establish a sense of esteem ". . . by asserting competence to achieve

in any one of a number of competitive fields (athletics, a career-line,

intellectual activity, leadership in school affairs, responsibility in

a job)" (p. 250). Our own guess--and at present it can be little

more than that--is that some such differences between boys and girls

did exist at the start of our longitudinal study in 1966, that they

probably still exist to at least some degree, but that the differences

are likely to diminish as views about women's and men's roles continue

to change.

Another limitation to our sample and the present analysis is that

we have not looked separately at racial subgroups, particularly blacks.

Rosenberg and Simmons (1971) have recently discussed racial differences

in self-esteem at some length, and our own earlier analyses included

a brief examination of self-esteem scores for three different sets

of black respondents (distinguished by geographical region and school

segregation/integration). Our sample was not designed to support

accurate generalizations about blacks, so our early analyses

were tentative and carefully qualified. Perhaps our most important

40
 Self-Esteem

40

conclusion concerning racial differences in self-esteem and many

other dimensions treated in the study was that it seemed unwise

to treat blacks as a single analysis group (Bachman, 1970, pp. 197-201).

The present paper has made no racial distinctions, partly because prelim-

inary analyses indicated that controlling race would have made

virtually no difference for the sample as a whole, and partly because

we continue to feel that our sample is not adequate for separate

racial analyses--particularly analyses as complex as those presented

here.

We noted that views about sex roles seem to be undergoing

considerable change, and thus we would be very cautious about

using data collected in the nineteen-sixties or earlier as a basis

for generalizing about sex differences in the seventies or eighties.

But this is simply one example of a very broad problem. Cronbach

(1975) argued convincingly that social science data-especially

correlational data gathered in real-life situations--are subject

to many interactions and extraneous factors, at least some of which

are likely to change over time. His summary appraisal bears

quoting here:

Generalizations decay. At one time a conclusion describes the

existing situation well, at a later time it accounts for rather

little variance, and ultimately it is valid only as history.

The half-life of an empirical proposition may be great or small.

41
 Self-Esteem

41

The more open a system, the shorter the half-life of relations

within it are likely to be. (Cronbach, 1975, pp. 122-123)

Those aspects of the social system we have been dealing with

in this paper--the links between attainment and self-esteem in

young men--may have been especially open to change during recent

years. During the period in which the young men in our sample

were making the transition from adolescence (age 15) to young

adulthood (age 23), they and the rest of society were also making the

transition from the mid-sixties (1966) to the mid-seventies (1974).

This was a turbulent interval invo1'fin substantial changes in the

job market and increasingly critical discussion about the value

of education. Thus the relationships we have been studying and trying

to isolate may have been changing all the while--perhaps at a rate

rapid enough to lead us to faulty conclusions.

The problems posed by social change--secular trends--are particu-

larly troublesome for the longitudinal study which follows a single

cohort for some period of time. As several authors have pointed

out (Buss, 1973; Schaie, 1965), the data from such a study do not

permit us to distinguish with certainty between genuine developmental

or maturational trends and those changes which affect society as a

whole.

Given the limitation outlined above, there are at leas two quite

different ways of interpreting our finding that educational success and

42
 Self-Esteem

49

its correlates became much less central to the self-esteem of the

young men in our sample as they moved through high school and beyond.

The first interpretation, and the one which we have emphasized, is

that this shift in centrality is a fairly typical part of the

developmental sequence followed by young people in this society.

During the late high school years and the period which follows,

the young person in the process of becoming an adult increasingly

anticipates and experiences situations in which self-evaluation

depends on factors quite different from academic success, and this

results in a reduced emphasis on the academic side of things. An

alternative interpretation of our findings is that they reflect a

particular secular trend or cultural change during the late sixties

and early seventies--a general decline in the importance or value that

society places upon education and educational success. Trust in

government declined dramatically during this period, and it may be

that faith in education as the pathway to success has also suffered

a setback. The developmental and secular trend interpretations are

not, of course, mutually exclusive; both may have played a part in

shaping our findings. A clear estimate of the relative importance of

each requires additional data from more than one cohort.

43
 Self-Esteem

43

References

Ammons, R. B., & Ammons, C. H. The Quick Test (QT): Provisional manual.

Psychological Reports, 1962, 11, 111-161, (Monograph Supplement I-V11.)

Bachman, J. G. Youth in Transition, volume II: The impact of family

background and intelligence on tenth-grade boys. Ann Arbor:

The Institute for Social Research, 1970.

Bachman, J. G. Anti-dropout campaign and other misanthropies. Social

Science and Modern Society (Trans/action), 1972, 9(5), 4-5.

Bachman, J. G., Green, S., & Wirtanen, I. D. Youth in Transition,

volume III: Dropping out--problem or symptom? Ann Arbor: The

Institute for Social Research, 1971.

Bachman, J. G., Kahn, R. L., Mednick, M. T., Davidson, T. N., & Johnston,

L. D. Youth in Transition, volume I: Blueprint for a longitudinal

study of adolescent boys. Ann Arbor: The Institute for Social Research, 1967.

Buss, A. R. An extension of developmental models that separate ontogentic

changes and cohort differences. Psychological Bulletin, 1973, 80,

466-479.

Cobb, S., Brooks, G. H., Kasl, S. V., & Connelly, W. E. The health of

people changing jobs: a description of a longitudinal study.

American Journal of Public Health, 1966, 56, 1476-1481.

Coopersmith, S. The antecedents of self-esteem. San Francisco; W. H.

Freeman and Company, 1967.

44
 Self-Esteem

44

Cronbach, L. J. Beyond the two disciplines of scientific psychology.

American Psychologist, 1975, 30, 116-127.

Douvan, E., & Gold, M. G. Modal patterns in American adolescence.

In L. W. Hoffman & M. L. Hoffman (Eds.), Review of child development

research. New York: Russell Sage Foundation, 1966.

Duncan, 0. D. A socioeconomic index for all occupations, and properties

and characteristics of the socioeconomic index. In A. J. Reiss, Jr.

(Ed.), Occupations and social status. New York: The Free Press, 1961.

French, J. R. P., Jr., & Kahn, R. L. A programmatic approach to studying

the industrial environment and mental health. Journal of Social

Issues, 1962, 18(3), 1-47.

Gates, A. I. Gates Reading Surve -Form 1. New York: Teachers College,

Columbil University, 1958.

Gergen, K. J. The concept of self. New York: Holt, Rinehart, & Winston,

1971.

Heise, D. R. Separating reliability and stability in test-retest corre-

lation. American Sociological Review, 1969, 34, 93-101.

Johnston, L. D. Drugs and American youth. Ann Arbor: The Institute for

Social Research, 1973.

Johnston, J., & Bachman, J. G. The transition from high school to work:

the work attitudes and early occupational experiences of young men.

Ann Arbor: Institute for Social Research, November, 1973.

(NTIS NO. 225938/AS)

40
 Self-Esteem

45

Kaufman, H. G. Relations between unemployment-reemployment experience

Pi.oceedings of the 81st Annual
and self-esteem among professionals.
8, 601-602.
Convention of the American Psychological Association, 1973,

Confidence intervals for clustered samples. American Sociolog-

Kish, L.

ical Review, 1957, 22, 154-165.

Kish, L. & Frankel, M. R. Balanced repeated replication for standard

errors. Journal of the American Statistical Association, 1970,

65, 1071-1094.

Luck, P. W., & Heiss, J. Social determinants of self-esteem in adult

males. Sociology and Social Research, 1972, 57, 69-84.

Self concept and school achievement. New Jersey:

Purkoy, W. W.

Prentice-Hall, 1970.

Society and the adolescent self-image. Princeton:

Rosenberg, M.

Princeton University Press, 1965.

Black and white self-esteem: the

Rosenberg, M., & Simmons, R. G.

urban school child. Washington: Rose Monograph Series, American

Sociological Association, 1971.

Schale, K. W. A general model for the study of developmental problems,

Psychological Bulletin, 1965, 64, 92-107.

Sewell, W. H., & Hauser, R. M. Causes and consequences of higher

Models of the status attainment process. American

education:

Journal of Agricultural Economics, 1972, 54, 851-861.

46
 Self-Esteem

46

Weidman, J. C., Phelan, W. T., & Sullivan, M. A. The influence of

educational attainment on self-evaluations of competence.

Sociology of Education, 1972, 45, 303-312.

Wylie, R. C. The self-concept: a critical survey of pertinent research

literature. Lincoln, Nebraska: University of Nebraska Press, 1961.

Wylie, R. C. The self-concept. (Rev. ed., Vol. 1). Lincoln, Nebraska:

University of Nebraska Press, 1974.

 Self-Esteem

47

Footnotes

1
The eta statistic is not restricted to linear relationships,

whereas the product-moment correlation is; therefore, when the

two statistics are nearly identical, as is true for the values

presented in Figure 1, it indicates that the relationship is linear.

2 Several of the measures included in Table 3 represent aspects

of the self-concept. The clearest example is self-concept of

school ability; but it could be argued that college and occupational

aspirations, and even self-reports of grades and misbehaviors in

school, are also parts of an individual's self-concept. As such,

these dimensions may be viewed as different components or facets

of global self esteem, rather than as separate and logically prior

causes. (This might help to account for some of the "time-matching"

noted above; another possible explanation is correlated measurement

error.) In any event, the general pattern of gradually lower correla-

tions with later measures of self-esteem holds for these measures as

well as the other ones in Table 3.

3
We also examined hourly wage rates, but found them essentially

uncorrelated with self-esteem measures at any point in time. This

finding is consistent with other analyses of wage rate data in this

study (currently unpublished) which show very few relationships

48
 Self-Esteem

48

between wage rates and other variables. We suspect that this general

lack of relationship is an age-specific phenomenon, resulting from

the fact that many of our respondents had only recently entered the

work force, and their wage rates had not begun to reach the levels

that would be expected for more experienced workers. A similar finding

and interpretation has recently been presented by Sewell and Hauser (1972).

This work was supported in part by a National Institute of Educa-

tion Grant #NE-G-00-3-0198. The authors thank T. Davidson, E. Dixon,

J. French, M. Gold, J. Johnston, R. Kulka, and M. Rosenberg for their

helpful comments on an earlier draft.

49
 Table 1

Self-esteem Items and Item-Index Correlations

1966 1974
c c
SD r SD r

I feel that I am a person of worth, at least

a
on an equal plane with others. 3.89 .91 .57 4.52 .68 .62

a
I feel that I have a number of good qualities. 3.72 .82 .56 4.36 .69 .66

a
I am able to do things as well as most other people. 3.73 .79 .57 4.31 .71 .61

b
I feel I do not have much to be proud of. 3.87 1.15 .54 4.03 1.01 .55

I take a positive attitude toward myself.a 3.64 .89 .59 4.12 .85 .69

b
Sometimes I think I am no good at all. 3.47 1.06 .58 4.06 .85 .67

I am a useful guy to have around.a 3.70 .77 .48 3.99 .75 .63
cn
cz, b
I feel that I can't do anything right. 3.83 1.02 .56 4.11 .82 .60

When I do a job, I do it well.a 3.67 .81 .51 4.35 .63 .51

b
I feel that my life is not very useful. 3.87 1.08 .62 4.39 .82 .58

Self-esteem index 3.74 .52 --- 4.22 .48 ---

Note. N = 1,628 with up to 2.5% missing data.

a
Response of "alMost always true" coded 5 (high self-esteem).
b
Response of "never true" coded 5 (high self-esteem).
c
Item-Index correlation.
 50

Table 2

Cross-time Self-esteem Correlations and Stability Coefficientsa

Self-esteem

1966 1968 1969 1970 1974

Self-esteem in 1966 1.00 .73 .64 .57 .40

Self-esteem in 1968 .53 1.00 .89 .79 .55

Self-esteem in 1969 .48 .64 1.00 .89 .62

Self-esteem in 1970 .42 .56 .65 1.00 .69

Self-esteem in 1974 .30 .40 .44 .49 1.00

M
_ 3.74 3.83 3.88 3.90 4.22

SD .52 .49 .50 .49 .48

N 1622 1501 1492 1408 1594

a
Product-moment correlations are below the diagonal; stability coeffi-

cients are above. The stability coefficients were computed using a path

analysis approach as developed by Heise (1969).

51
 Table 3

Self-esteem Correla. . . with Measures of Atta4mment and Related Factors

Educational Status of

Self-esteem Attainment Occupation

1q66 1968 1969 1970 1974 1974 1974

Socioeconomic level 1966 .16 .11 .12 .12 .07 .44 .20

Ability 1966 .21 .18 .17 .20 .13 .50 .30

Grades 1966 .26 .21 .17 .16 .12 .51 .32

Grades 1968 .25 .25 .20 .16 .14 49 .29

Grades 1969 .21 .25 .21 .14 .12 .50 .30

Self-concept of school ability '66 .36 .28 .27 .24 .17 .43 .24
CR
N.)
Self-concept of school ability '68 .29 .32 .29 .25 .18 .44 .21

Rebellious behavior in school '66 -.36 -.26 -.18 -.20 -.15 -.28 -.18

Rebellious behavior in school '68 -.21 -.27 -.19 -.17 -.10 -.26 -.15

Delinquent behavior in school '66 -.19 -.10 -.04 -.06 .00 -.32 -.14

Delinquent behavior in school '68 -.18 -.14 -.11 -.07 -.06 -.38 -.17

Delinquent behavior in school '69 -.15 -.12 -.10 -.08 -.03 -.36 -.19
 Table 3

(continued)

Educational Status of

Self-esteem Attainment Occupation

1966 1968 1969 1970 1974 1974 1974

College plans 1966 .22 .13 .12 .10 .10 .41 .23

College plans 1968 .22 .19 .16 .14 .11 .41 .20

College plans 1969 .25 .22 .20 .16 .16 .61 .30

Status of aspired occupation '66 .19 .13 .09 .11 .11 .43 .26

Status of aspired occupation '68 .20 .15 .13 .12 .09 .50 .28

Status of aspired occupation '69 .21 .18 .16 .14 .08 .52 .30

Status of aspired occupation '70 .21 .17 .15 .14 .10 .56 .31

Status of aspired occupation '74 .15 .15 .12 .12 .11 .55 .45

Educational attainment 1974 .26 .21 .20 .17 .15 1.00 .42

Status of attained occupation '74 .15 .13 .14 .20 .16 .42 1.00

Employment vs unemployment '74 .03 .00 .05 -.01 .10 .06 .12

Note: Based on an N of 1600 and a design effect of 2.25, an r of .08 is significant at

.05, and an r of .10 is significant at p < .01 .
 53

Table 4

Cross-time Self-esteem Scores for Employed and Unemployed

Self-esteem

1974 Employment Status 1966 1968 1969 1970 1974

Employed (N = 1,205) M 3.72 3.82 3.88 3.89 4.23

SD .51 .47 .49 .48 .47

Unemployed (N = 111) M 3.66 3.81 3.79 3.91 4.05

SD .56 .55 .52 .53 .58

*
'Difference M
.....
.06 .01 .09 -.02 .18

t = 3.78, IL <.05, two-tailed, incorporating design effect

54
 54

TABLE 5

Multiple and Partial Correlations with Cross-Time Self-Esteem

Self-Esteem

1966 1968 1969 1970 1974

Multiple Ra .30** .28** .24** .22** .16**

b 11** .09** .10** .06 .08*

Partial r (with educational attainment)

Partial rc (with occupational status) .05 .04 .07 .13** .11**

d .05 -.02 .10*

Partial r (with employment) .03 .00

aMultiple correlation of self-esteem with ability, socioeconomic level, and

ninth grade grades.

b Partial correlation between self-esteem and educational attainment, holding

constant ability, socioeconomic level, and ninth grade grades.

c
Partial correlation between self-esteem and status of attained occupation,

holding constant ability, socioeconomic level, and ninth grade grades.

d
Partial correlation between self-esteem and employment, holding constant

ability, socioeconomic level, and ninth grade grades.

< .05, incorporating design effect.

**11..< .01, incorporating design effect.

 55

Table 6

Partial Correlations with 1974 Self-esteem

Educational Status of Employed vs

Attainment Occupation Unemployed

1974 1974 1974

** ** *
Zero-order r with 1974 self-esteem .15 .16 .10
* *
Partial ra with 1974 self-esteem .08 .11** .10
* *
b
Partial r with 1974 self-esteem .06 .10 .09

Partial rc with 1974 self-esteem .04 .09 .10

a ji_ty. socioeconomic level, and

Partial correlation, holding cons 'int a

grades (1966).

b Partial correlation, holdimg consnt ability, socioeconomic level, grades

(1966), and self-esteem (1966).

cPartial correlation, holding covstant ability, socioeconomic level, grades

(1966, 1968, 1969), and self-esteem (1966, 1968, 1969).

s.05, two-tailed, incorporating design effect.

**
2.<.01, two-tailed, incorporating design effect.

56
 Self-Esteem

56

Figure Captions

Figure 1. Cross-time self-esteem scores for six categories of

educational attainment.

57
 +1.4

+1.2
5
+1.0 SD 4
3
+0.8 2

+0.6

+0.4
2 +0.2

tn1966Mean

-J
-0.2
-0.4
I High School Dropout
-0.6 2 High School Graduate
3 Some Col lege
-0.8 4 Associate Degree
5 Bachelor Degree
6 Post-Groduate Educotion
-1.050
I

1966 '68 '69'70 1974

TIME
eta .27 .2/ .20 ./8 ./6
r .26 .2/ .20 ./7 .15
N = /602 /487 /476 /393 /594

Figure 1. Cross-time self-esteem scores for six categories

of educational attainment

58