Академический Документы
Профессиональный Документы
Культура Документы
3, 1978
INTRODUCTION
The research reported in this article was funded by a grant from the National Institute of
Mental Health (MH27747), which is gratefully acknowledged. The data and tabulations
utilized in this article were made available (in part) by the Institute for Social Research
Social Science Archive. The data were originally collected by Dr. Jerald G. Bachman,
Survey Research Center, Institute for Social Research, The University of Michigan. Neither
the original collector of the data nor the archive bears any responsibility for the analyses or
interpretations presented here.
This article is a revised version of a paper presented at the annual meeting of the Eastern
Sociological Society, Philadelphia, April 2, 1978.
1Lecturer, Department of Sociology, University of Maryland~ College Park, Maryland.
Received her Ph.D. from the University of Maryland and is interested particularly in the
social influence on the self-concept.
2Professor, Department of Sociology, University of Maryland, College Park, Maryland.
Received his Ph.D. from Columbia University and is interested in the structure and develop-
ment of the self-concept.
279
0047-2891/78/0900-0279505.00/0 9 1978 Plenum Publishing Corporation
280 Rosenberg and Rosenberg
are Lazarsfeld's 16-fold table (Lazarsfeld, 1948, 1968, 1972; Rosenberg and
Thielens, 1952; Rosenberg, 1957; Lipset et al., 1954; Barton and Lazarsfeld,
1962), Coleman's continuous time Markov process model (Coleman, 1964,
1968; McDill and Coleman, 1963 ; Anderson, 1953), the cross-lagged panel cor-
relation (Campbell, 1963; Campbell and Stanley, 1963; Pelz and Andrews,
1964; Rozelle and Campbell, 1969), and path analysis (Heise, 1970).
It would lead us too far astray to consider the strengths, weaknesses, and
assumptions of each procedure, and we have opted here for the cross-lagged
panel correlation. Since the technique is not widely used, Figure 1 shows the
components of this procedure. A1B1 and A2B2 are synchronous correlations,
A1A2 and B1B2 are auto-correlations, and A~B2 and B1A2 are cross-lag-
ged correlations, l f A ~B2 > BIA2, then A is said to have a greater effect on B
than the other way around; ifBaA2 > A ~B2, then the reverse interpretation is
made.
The data for this analysis are drawn from Youth in Transition, a study direc-
ted by Jerald G. Bachman at the Institute for Social Research at the University
of Michigan (Bachman et al., 1972). Thus far, five waves over an eight-year
span have been completed. Wave I, conducted in the fall of 1966, involved a
probability sample of 2213 tenth-grade boys in 87 high schools throughout
the country; 1886 of these subjects participated in the second data collection
phase (Wave II) in the spring of 1968. The data analysis in this paper will be
confined to the first two waves.
Self-esteem consists of a 10-item score based on two previous studies -
6 items are from the Rosenberg (1965) self-esteem scale, 4 from a study of indi-
viduals changing jobs (Cobb et al., 1966). It is important to stress that these
items are general and content-free, i.e., they explicitly avoid reference to any
specific aspect of the self, such as delinquency. (Typical items: "I take a positive
attitude toward myself. .... I feel that I have a number of good qualities.")
These interviews include material on a wide variety of delinquent be-
havior. In most cases a number of items have been combined to form different
indexes of delinquency. We have examined five indexes of delinquency:
1. Delinquent behavior in school (e.g., "damaged school property on pur-
pose," "hit a teacher").
2. Frequency of delinquent behavior (e.g., how often "run away from home,"
"taken part in a fight where a bunch o f your friends are against another
bunch").
3. Seriousness of delinquent behavior (e.g., "hurt someone badly enough to
need bandages or a doctor," "used a knife or g u n . . , to get something
from a person").
4. Theft and vandalism combines several items from the above indexes (e.g.,
"taken something from a store without paying for it," "set fire to some-
one else's property on purpose").
Self-Esteem and Delinquency 283
A1 auto-correlatlon A2
~oug
l;o/ y
synchro- ~ synchro-
nou8
BILM" ~.m 2
auto-correistiou
Fig. 1. Cross-lagged panel model.
5. Total delinquency combines all the above items plus several additional
ones (e.g., "got into trouble with the police because of something you
did").
All these items are measured in terms of frequency of occurrence. Figure 2 pre-
sents the cross-lagged panel correlations between global self-esteem and each of
the five measures of delinquency. An analysis of the correlations in these dia-
grams indicates the following:
(1) There are strong auto-correlations for each of the variables between time
1 and time 2. The correlations for self-esteem over time is r = 0.536. Cor-
relations for each of the delinquency indexes over time range from r =
D1 - -.594
. '~ D 2 bI .502 "D 2 DI - .422 "v 2
I; "~ I ; =o
DI~ .442 D2 D1 I ~ ,565 ~D2
Theft and Total
Vandalism Delinquency
Fig. 2. Cross-lagged and other correlations between self-esteem and delinquency.
284 Rosenberg and Rosenberg
3Both Rozelle and Campbell (1969) and Kenny (1975) point to the importance of "sta-
tionarity" (i.e., similar synchronous correlations at both waves). Criteria are unavailable to
assess whether the decline in synchronous correlations in this case is so severe as to call
into question the cross-lagged data.
Self-Esteem and Delinquency 285
4 Partial correlation, of course, represents a statistical adjustment rather than a literal "hold-
ing constant," as in subgroup classification, but we have taken the liberty of applying
phraseology appropriate to the latter method in this discussion.
GO
Table III. Conditional Cross-Lagged Panel Correlations Between Self-Esteem and Delinquency for Total Sample and for Lower
and Higher Socioeconomic Groups a
w
ga.
Self-Esteem and Delinquency 287
The data thus suggest that self-esteem has a stronger effect on delinquency
than the other way around. But while this condition may be true in general, it
need not necessarily be true, or true in equal measure, for all population sub-
groups. Since social class and delinquency are related, it is relevant to examine
these relationships within social class groups.
T a b l e III presents the cross-lagged correlations of self-esteem and the five
delinquency measures for the entire sample and separately for lower and higher
class boys. s As a crude measure of relative effect we have subtracted the cross-
lagged correlation rDISE2from the cross-lagged correlation rSEID2. A positive
figure indicates that self-esteem has the greater effect on delinquency than the
other way around. All 15 comparisons, it can be observed, are positive, which
indicates that self-esteem is more likely to be the causal variable both for the
entire sample and for each social class group on every delinquency index. For
the total sample, it may be noted, all the correlations are significantly different
at at least the 0.10 level.
It is noteworthy, however, that the differences between these correla-
tions are consistently larger in size, and more likely to be statistically significant,
in the lower class than in the higher class (Table III). In other words, although
in each case self-esteem has a greater effect on delinquency than the converse,
this is more the case among lower than higher socioeconomic class adolescents.
It is interesting to examine the basis of this difference. Consider first
rDISE2in both social classes. With regard to three delinquency measures, the
relationship is stronger in the higher socioeconomic class and in the other two
cases it is equal. Thus, to the extent that delinquency has any effect on self-
esteem, this effect is stronger in the higher than in the lower class. In other
words, the fact of delinquency will damage a higher SES youth's self-esteem more
than a lower SES youth's.
Now consider rSEID2. In four out of five cases, Table III reveals, these
correlations are appreciably stronger in the lower than in the higher socio-
economic class, and the fifth case is about equal. Thus, the effect of self-esteem
on delinquency appears to be greater in the lower than in the higher socio-
economic class. Low self-esteem, it would appear, is more likely to drive the
lower class adolescent to delinquency than the higher class youngster.
The upshot is that, in the lower socioeconomic class, self-esteem has a
stronger effect on delinquency, while delinquency has a weaker effect on self-
esteem than in the higher class. Hence, self-esteem affects delinquency more
than the other way around to a greater extent in the lower than in the higher
socioeconomic class. 6
Why should this be so? The chief reason, we believe, is to be found in
Kaplan's theory (1975) of reference group identification. According to this
theory certain youths, because of societal rejection, develop low self-esteem.
In order to restore their damaged self-esteem, they turn to delinquent groups
which share and reinforce their animus toward the larger society and which
grant them respect for delinquent behavior - a respect which the larger society
denies.
What is of particular interest in the present regard is that this theory is
more persuasive for the lower than for the higher socioeconomic class. In a lower
class environment in which delinquency is more common, the low self-esteem
youth is more likely to fred companions who will respect and admire delinquent
behavior. He gains kudos and social acceptance for daring feats of nonconformity
and illegality. Furthermore, the severity of condemnation may be weaker and
the general level of acceptance of delinquent behavior stronger in environments
in which such behavior is more widespread. Finally, there are probably fewer al-
ternative ways for the lower class youngster to command social respect, e.g.,
acquiring a car, stereo, clothes. The reverse applies to the higher class youngster
with low self-esteem. The higher socioeconomic class- contains fewer delinquent
groups which can serve as sources of status; general social condemnation of
delinquency in this environment may be more intense (e.g., middle class girls
may be less willing to associate with the delinquent youth); and the youth may
have other ways of gaining self-esteem. These factors may help to explain our
finding that low self-esteem at a given time is more likely to be associated with
subsequent delinquency in the lower than in the higher crass.
But the same normative factors may also explain why delinquency affects
self-esteem to a greater extent in the higher than in the lower class. Whatever
factors draw the higher class youth to delinquency in the first place, the result
is more likely to be severe social condemnation in his environment. This may
take the form of horrified reactions on the part of his parents, and he may be
shunned by the predominantly "straight" boys and girls in his environment. If
this social disapproval is more likely to characterize higher than lower class en-
vironments, then it is understandable, from the perspective of Mead's theory
(1934) of reflected appraisals, that delinquency should more strongly affect
self-esteem in the higher than in the lower class.
6When these data are examined by means of partial correlation (rSE1D 2. D~ and rD~ SE 2.
SEt), the pattern of relationships is maintained for each socioeconomiclevel.
Self-Esteem and Delinquency 289
ence group theory holds that self-esteem is primarily responsible for delin-
q u e n c y - that youngsters with damaged self-esteem seek to restore their self-
respect by aligning themselves with deviant groups which accord them the ap-
proval denied by the rest of society. The theory of reflected appraisals, on the
other hand, assigns causal priority to delinquency or, to be more exact, societal
reaction to deviance. Whatever the initial cause of delinquency, society's con-
demnation of the behavior and the individual's internalization of others' at-
titudes toward him will lead to low self-esteem.
In this study of high school boys, the weight of evidence is in the direc-
tion of Kaplan's speculations. For the sample as a whole, and for the lower and
higher socioeconomic classes separately, the analysis of cross-lagged panel cor-
relations suggests that self-esteem has a stronger effect on delinquency than
delinquency has on self-esteem. And this is especially tree in the lower class
where the social support for such activity may be stronger and the social con-
demnation weaker. For the same reason, delinquency appears to damage the
self-esteem of the higher SES youngster more than of the lower SES boy.
Obviously, self-esteem is but one of many factors affecting delinquency
and is certainly not the most important. Nevertheless, if it is the case, as our
data suggest, that adolescent boys may turn to delinquent subgroups in order
to restore their shattered self-esteem, then one might consider the advisability
of providing alternative activities or reference groups (e.g., athletics, jobs, music)
which can help restore their feeling of self-respect. For one thing is certain:
Everyone -delinquent or nondelinquent, upper class or lower c l a s s - prefers
to think well of himself, to feel self-respect, to be a person of worth in his own
eyes. No one accepts low self-esteem with equanimity. If a youngster must have
recourse to deviant behavior in order to burnish his tarnished self-image, he will
do so. The society which disdains a youngster as stupid, bad, or incompetent
should not be surprised when he turns his energies into deviant channels in order
to acquire what everyone w a n t s - "the sentiment of self-regard" (McDougall,
1932).
The data presented in this paper are little more than an introduction to a
complex topic. Is delinquency carried on in a group context - where the delin-
quent act excites approval and a d m i r a t i o n - or is it conducted in i s o l a t i o n -
where the individual attempts to gain self-respect by measuring up to certain
inner standards of courage, cunning, and ruthlessness? Although there is abundant
research on delinquency, we still need to learn more about the phenomenology
of the delinquent act and its place in a motivational system.
Finally, it is plain that the relative effects of self-esteem and delinquency
may not be the same for girls as for boys. Perhaps girls gain none of the social
approval that boys get from engaging in delinquent acts, in which case delin-
quency would not be a device to restore lowered female self-esteem. At the same
time the girl's deviance might be more strongly condemned in the group, thereby
damaging her self-esteem. It is possible that the successful gang leader is admired
by some girls, but the successful prostitute is not admired by boys. In tiffs case,
290 Rosenbexg and Rosenberg
REFERENCES
Anderson, T. (1953). Markov chains and panel analysis. In Lazarsfeld, P. F., (ed.), Mathe-
matical Thinking in the Social Sciences, Free Press, Glencoe, Ill.
Bachman, J. G., Kahn, R. L., Mednick, M. T., Davidson, T. N., and Johnston, L. D. (1972).
Youth in Transition, Volume L Blueprint for a Longitudinal Study of Adolescent
Boys, Institute for Social Research, Ann Arbor.
Backman, C. W., and Secord, P. F. (1962). Liking, selective interaction, and misperception
in congruent interpersonal relations. Sociometry 25 : 325-335.
Barton, A. H., and Lazarsfeld, P. F. (1962). Methodology of quantitative social research.
In Varma, B. N. (ed.), A New Survey of the Social Sciences, Asia Publishing House,
New York.
Bohrnstedt, G. W. (1969). Observations on the measurement of change. In Borgatta, E. F.
(ed.), Sociological Methodology, Jossey-Bass, San Francisco, pp. 113-133.
Campbell, D. (1963). From description to experimentation. In Harris, C. W. (ed.), Prob-
lems in Measuring Change, University of Wisconsin Press, Madison.
Campbell, D., and Stanley, J. C. (1963). Experimental and quasi-experimental designs.
In Gage, N. L. (ed.), Handbook of Research on Teaching, Rand McNally, Chicago.
Cobb, S., Brooks, G. H., Kasl, S. V., and Connelly, W. E. (1966). The health of people
changing jobs: A description of a longitudinal study. Am. J. Pub. Hlth. 56: 1476-1481.
Coleman, J. S. (1964). Introduction to Mathematical Sociology, Free Press, New York.
Coleman, J. S. (1968). The mathematical study of change. In Blalock, H., and Blalock, A.
(eds.), Methodology in SocialResearch. McGraw-Hill, New York.
Dinitz, S., Dynes, R. R., and Clarke, A. C. (1969). Deviance: Studies in the Process of
Stigmatization and Societal Reaction, Oxford University Press, New York.
Gordon, C. (1968). Self-conceptions: Configurations of content. In Gordon, C., and Gergen,
K. (eds.), The Self in Social Interaction, Wiley, New York.
Heise, D. (1970). Causal inference from panel data. In Borgatta, E. F., and Bohrnstedt, G.
(eds.), SociologicalMethodology, Wiley, New York.
Kenny, D. A. (1975). Cross-lagged panel correlation: A test for spuriousness. Psychol. Bull.
82: 887-903.
James, W. (1950). The Principles of Psychology, Dover, New York, (originally published
in 1890 by Henry Holt and Company).
Jensen, G. F. (1972). Delinquency and adolesce~at self-conceptions: A study of the personal
relevance of infraction. Soc. Probl. 20(1): 84-103.
Jones, S. C. (1973). Self- and interpersonal evaluations: Esteem theories versus consistency
theories. Psychol. Bull 79: 185-199.
Kaplan, H. B. (1975). Self-Attitudes and Deviant Behavior, Goodyear Publishing, Pacific
Palisades, Calif.
Kaplan, H. B. (1976). Self-attitudes and deviant response. Soc. Forces 54: 788-801.
Lazarsfeld, P. F. (1948). The use of panels in social research. Proc. Am. Philosoph. Soc.
92: 405-410.
Lazarsfeld, P. F. (1968). Preface to the third edition. In Lazarsfeld, P. F., Berelson, B.,
and Gaudet, H., .The People's Choice, Columbia University Press, New York.
Lazarsfeld, P. F. (1972). Mutual relations over time of two attributes: A review and integra-
tion of various approaches. In Hamner, M. et al. (eds.), Psychopathology, Wiley,
New York, pp. 461-479.
Lecky, P. (1945). Self-Consistency: A Theory of Personality, Island Press, New York.
Lipset, S. M.,Lazarsfeld, P. F., Barton, A. H., and Linz, J. (1954). The psychology of voting:
An analysis of political behavior. In Lindzey, G. (ed.), Handbook o f SocialPsychol-
ogy, Addison-Wesley, Cambridge, Massachusetts, pp. 1150-1164.
Maslow, A. H. (1954). Motivation and Personality, Harper, New York.
Self-Esteem and Delinquency 291
McDill, E. L., and Coleman, J. S. (1963). High school social status, college plans, and aca-
demic achievement. Am. Sociol. Rev. 28: 905-918.
McDougall, W. (1932). The Energies o f Men, Methuen, London.
Mead, G. H., (1934). Mind, Self and Society, University of Chicago Press, Chicago.
Pelz, D. C., and Andrews, F. M. (1964). Detecting causal priorities in panel data. Am.
Sociot. Rev. 29: 836-848.
Reckless, W. C., and Dinitz, S. (1967). Pioneering with the self-concept as a vulnerability
factor in delinquency. J. Crim. Law, Criminol. Police Sci. 58:515-523.
Reckless, W. C., Dinitz, S., and Murray, E. (1956). Self-concept as an insulator against delin-
quency. Am. Sociol. Rev. 21: 744-746.
Reckless, W. C., Dinitz, S., and Murray, E. (1957). The "good" boy in a high delinquency
area. J. Crim. Law, Criminol. Police Sci. 58: 18-26.
Rosenberg, M. (1957). Occupations and Values, The Free Press, Glencoe, Illinois.
Rosenberg, M. (1965). Society and the Adolescent Self-linage, Princeton University Press,
Princeton, N.J.
Rosenberg, M., and Simmons, R. G. (1972)..Blacktand White Self-Esteem, American Sociol-
ogical Association, Washington, D.C.
Rosenberg, M., and Thielens, W. (1952). The panel study. In Jahoda, M.,Deutsch, M., and
Cook, S. W., (eds.), Research Methods in Social Relations, Vol. II, Dryden Press,
New York, pp. 588-609.
Rozelle, R. M., and Campbell, D. T. (1969). More plausible rival hypotheses in the cross-
lagged panel correlation technique. Psychol. Bull 71 : 74-80.
Schur, E. M. (1971). Labeling Deviant Behavior: Its Sociological Implications, Harper and
Row, New York.
Schwartz, M., and Tangri, S. (1965). A note on self-concept as an insulator against delin-
quency. Am. Sociol. Rev. 30: 922-926.
Secord, P. F., and Backman, C. W. (1961). Personality theory and the problem of stability
and change in individual behavior: An interpersonal approach. Psychol. Rev. 68:
21-32.
Tangri, S. S., and Schwartz, M. (1967). Delinquency and the self-concept variable. J. CrirrL
Law, Criminol. Police ScL 58: 182-190.
References
Bachman, J. G. (1970). Youth in Transition, Vol. 2, Institute for Social Research, Ann
Arbor.
Bachman, J. G., Green, S., and Wirtanen, I. D. (1971). Youth in Transition, V01. 3, insti-
tute for Social Research, Ann Arbor.
McCord, J. (1978). A thirty-year follow-up of treatment effects. Am. Psychologist 33(3):
284-289.
McCord, J. (1977). A life history approach to criminal behavior. Paper presented at the
meeting of the American Society of Criminology, Atlanta, November 16-20.
Florence R. Rosenberg
Morris R osenberg
In her comment, Dr. McCord points out that the correlation between self-
esteem and delinquency may be due to "some third condition." We are cer-
tainly aware of the danger that extraneous factors may underly zero-order rela-
tionships (see Rosenberg, 1968). But it is precisely this danger that makes the
cross-lagged panel correlation such a valuable technique. If, as is usually the case,
we were restricted to the examination of a static relationship between self-esteem
and delinquency, we might indeed suspect that both were outcomes of some
common cause, e.g., parental status or family background. But family back-
ground does not explain why self-esteem predicts changes in delinquency be-
tween the tenth and eleventh grades. Family background might explain an initial
relationship, but how does it explain how one variable is associated with change
in another? The problem of extraneous variables in survey research must always
be considered (Hirschi and Selvin, 1973; Rosenberg, 1968), but the problem is
substantially reduced through use of the cross-lagged panel method. McCord
has elected to attack one of the distinctive strengths of the study.
We must acknowledge bafflement at McCord's observation that "the
hypothesis that low self-esteem causes delinquency predicts that among those
low in delinquency at time 1, self-esteem scores should be negatively correlated
with delinquency." What can this statement mean? McCord seems to be suggest-
ing that we examine the relationship between self-esteem and delinquency,
controlling on delinquency. The logic of such a procedure eludes us. If, in fact,
she meant (but neglected to state) that "self-esteem scores should be negatively
294 Rosenberg and Rosenberg
REFERENCES
Hirschi, T., and Selvin, H. (1973). Principles of Survey Analysis, Free Press, New York.
Kaplan, H. B. (1975). Increase in self-rejection as an antecendent of deviant responses. J.
Youth Adoles. 4(3): 281-292.
Kaplan, H. B. (1976). Self-attitudes and deviant response. Soc. Forces 54(4): 788-801.
Rosenberg, M. (1968). The Logic of Survey Analysis, Basic Books, New York.
Rosenberg, M. (1976). Beyond self-esteem. Paper presented at the annual meeting of the
American Sociological Association, New York, August.