A NONRECURSIVE SIMULTANEOUS-EQUATION MODEL
FOR PROBLEM DRINKING
CLAIR EUGENE MARTIN
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
I have read with incredulity the proclamations of grati-
tude that preface other disserations. Now that I have com-
pleted this manuscript, I am a believer! Words do not ade-
quately describe the debt incurred to those who have made the
completion of this task possible.
It is difficult for me to imagine a more supportive and
constructive committee than the one wnich has directed my
doctoral study. I am especially indebted to my chairman,
Dr. Gerald R. Leslie, whose scholarship has served as a model
to me throughout my doctoral study and whose suggestions
and comments have contributed immensely to the substance and
style of this manuscript.
Dr. Felix Berardo and Dr. George Warheit have taught
me much of the sociology that I know. Both have offered
support and constructive suggestions in the process of com-
pleting this study. Dr. Alan Agresti has desensitized me
from my fear of statistical analysis. He has taught me that
statistics is a valuable tool that can be mastered by the
researcher rather than vice versa. I am also grateful to
Dr. Pauline Barton, whose empathy and support have assisted
me in my efforts to combine sociology and nursing in a mean-
I am indebted to Dr. George Warheit, Director of the
Florida Health Study, and Dr. Roger Bell, Director of the
Southern Family Life Studies, for their enthusiastic con-
sent to utilize the data obtained in their studies.
Mr. Charles Holzer III is not only a respected col-
league who has offered substantive suggestions during this
research, but his ability to talk to the computer has proved
to be invaluable assistance in the analysis of these data.
I am not only grateful for his assistance but envious of
Analysis of these data was facilitated by the financial
assistance of the Computer Committee of the College of Arts
and Sciences. Typing and copying the dissertation was made
easier by the financial assistance of Alpha Theta Chapter,
Sigma Theta Tau.
Mr. Dennis Carter gave valuable assistance in helping
me to understand the statistics from the field of econo-
metrics. Dr. Kenneth Hodge provided needed consultation
Special recognition is due Ms. Sue Kirkpatrick for her
careful typing of the manuscript in its present form.
Finally, my wife, Guenn, and daughter, Sonja, deserve
a special thank you. ,Their assistance and support have
been most valued of all.
TABLE OF CONTENTS
AKNOWLEDGMENTS . . .
ABSTRACT . . . .
CHAPTER I. INTRODUCTION . .
The Problem . . .
The Data . . .
The Method . . .
Value of the Research . .
CHAPTER II. LITERATURE REVIEW .
Drinking Behavior in the United States
Abuse of Alcohol as Deviant Behavior
Propositional Review of the Literature
Summary . . . .
CHAPTER III. METHODOLOGY . .
Theoretical Specification . .
Operationalization of the Concepts
The Instrument . . .
The Sample . . .
Data Collection . . .
Data Analysis Techniques . .
Assumptions . . .
Summary . . . .
CHAPTER IV. DATA ANALYSIS AND FINDINGS
Introduction . . .
Characteristics of the Sample .
Psychosocial Characteristics of the
Drinking Patterns of the Sample .
The Theoretical Model . .
Identification of the Model ..
The Reduced-Form Model . .
The Structural Model . .
The Path Model . . .
Summary . . . .
. . 47
. . ii
j I 1 I i
CHAPTER V. SUMMARY AND CONCLUSIONS .
The Problem Statement . .
The Theoretical Foundation .
The Data and Data Analysis ..
Findings . . . .
Strengths and Limitations .
Recommendations for Further Study
BIBLIOGRAPHY . . . .
APPENDIX . . . .
BIOGRAPHICAL SKETCH . . .
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
A NONRECURSIVE SIMULTANEOUS-EQUATION
MODEL FOR PROBLEM DRINKING
Clair Eugene Martin
Chairman: Gerald R. Leslie
Major Department: Sociology
The purpose of this dissertation is to describe, explain,
and predict the conditions under which adult residents of
four central Florida counties experience problem drinking
as measured by a problem-drinking index. For heuristic
purposes a multivariate causal model is constructed. This
model includes feedback loops from problem drinking to each
of the endogenous variables with which it is reciprocally
related. While problem drinking is the major dependent
variable it is considered to be both cause and effect of
perceived value-access disjunction, general psychopathology,
and anxiety. The predetermined, independent variables in
the model are age, sex, marital status, residential mobility,
total family income, occupational prestige, and education.
In order to identify the equations, race and residence are
considered to be predetermined independent variables in the
estimation equations for perceived value-access disjunction,
general psychopathology, and anxiety. These variables are
selected for inclusion in the model on the basis of social
integration theory, current research findings in the area
of problem drinking, and the available data.
This study focuses upon one aspect of an extensive
social psychiatric epidemiological survey. A comprehensive
317-item structured interview schedule was administered by
trained interviewers to respondents who were randomly
selected by use of a multistage cluster sampling technique.
Usable interviews were obtained from 3,674 persons, 86 per
cent of the sample.
Two-stage least-squares is the technique used for data
analysis. This technique is selected because several of
the variables in the proposed model are assumed to be re-
ciprocally related. In this case, multiple regression
produces a biased estimate of the parameters and is there-
fore an inappropirate data analysis technique. The two-
stage least-squares technique provides an estimate of the
parameters that is purged of this error. The estimated
standardized beta coefficients are plotted on a path dia-
The findings support the existing theoretical formula-
tions and clarify the reciprocal relationship among problem
drinking and the other endogenous variables in the model.
Fifteen per cent of the variance in the problem-drinking
score is explained by 8 of 10 variables described above.
When total family income is included in the model, occupa-
tional prestige and education are not significantly asso-
ciated with the problem-drinking score. In each case the
standardized beta coefficient of the feedback loop from
problem drinking to the variables with which it is recip-
rocally related--general psychopathology, anxiety, and
perceived value-access disjunction--is greater than the
standardized beta coefficient leading from that variable to
These findings verify the importance of personal and
social integration in influencing drinking behavior. They
also suggest that when an individual becomes a problem
drinker it is likely that personal and social integration
will be dramatically decreased.
One of the major tasks of the social researcher is to
describe the conditions under which a social phenomenon may
be explained and predicted. Researchers from many fields,
as well as journalists, philosophers, reformers, and moral-
ists, have focused their attention on the use and abuse of
alcohol. Nevertheless, while there exists a plethora of
assumptions and theories regarding this problem,2 there is
The extensive interest in alcohol use is documented
by the existence of national organizations such as Alco-
holics Anonymous, the National Council of Alcohol Abuse and
Alcoholism, The National Council on Alcoholism, and The
Alcohol and Drug Problems Association of North America.
National publications concerned solely or primarily with
alcohol include The Quarterly Journal of Studies on Alco-
hol, The British Journal of Addiction, The International
Journal of Addictions, and Alcohol and Health Notes. In
addition, numerous articles and monographs have been writ-
ten on alcohol, its use and abuse. Medline lists 4,859
articles in its file. Cahalan reports more than 70,000
professional writings on file at the Rutgers Center of
Alcohol Studies. See Don Cahalan, Problem Drinkers (San
Francisco: Jossey-Bass, Inc., 1970),p. 1.
Siegler et al. describe eight models and related
assumptions which have been applied to the study of alco-
holism in the literature. See Miriam Siegler, Humphrey
Osmond, and Stephens Newell, "Models of Alcoholism,"
Quarterly Journal of Studies on Alcohol 29 (September,
limited consensus and less empirical verification for most
The purpose of this research is to describe, explain,
and predict the conditions under which problem drinking
occurs among the adult residents of four counties in Florida.
It is assumed that an understanding of problem drinking may
be facilitated by constructing a model in which the inter-
dependent reciprocal relationships among variables are ex-
plored rather than by assuming one-way causality.
Research begins and ends with existing theory and re-
search findings. The literature review reported in Chapter
II discusses the appropriate concepts for inclusion in the
proposed model and suggests their relationship one with the
other. The review of the literature assists in distinguish-
ing from among the variables those which are assumed to be
interdependent, reciprocal variables, termed endogenous
variables, and those which are assumed to be independent,
predictive variables, termed exogenous variables. Causal
relationships for the proposed model are assumed and ini-
tially justified primarily on the basis of the literature
review. The discussion of methodology in Chapter III
Selden Bacon, "The Process of Addiction to Alcohol:
Social Aspects," Quarterly Journal of Studies on Alcohol
34 (March, 1973), 1-27.
explores the necessary criteria for causal assumptions. The
simultaneous equation model is presented in Chapter III.
The data which are utilized as a test of the model were
collected in two separate but nearly identical epidemiologi-
cal surveys of social psychiatric impairment. The data for
the first survey were collected in 1970 and 1971, in a
county of north-central Florida. The second survey was con-
ducted in 1972 and 1973, in three counties of central
Florida. Households and respondents were randomly selected,
utilizing a systematic random sample and the Kish method.4
Members of the research team administered a comprehensive
structured interview schedule to the sample of 1,645
respondents in the first survey and 2,029 respondents in
the second. The schedule included items concerning the
respondents' demographic characteristics, health, and use
of alcohol. Chapter IV includes a more detailed descrip-
tion of the sample.
The review of the literature presented in Chapter II
suggests the variables to be included in the model proposed
in this study. Chapter III presents the methods of
Leslie Kish, "A Procedure for Objective Respondent
Selection Within the Household," Journal of the American
Statistical Association 44 (September, 1949), 380-387.
operationalizing these theoretical concepts, building in-
dices, and measuring the variables, as well as describing
data collection and selection of the sample. The final
section of Chapter III describes the methods of data anal-
ysis utilized. The descriptive characteristics of the
sample are presented in Chapter IV. The descriptive statis-
tics used are percentages, median, mean, range, and
standard deviation. The statistical measure of association
utilized is a method of simultaneous equations, the two-
stage least-squares technique of estimation, which provides
a "feedback loop" from one endogenous variable to another.
Duncan introduced sociologists to the method of path analysis
as a means of constructing causal models. The use of
simultaneous equations and the two-stage least-squares
technique may be viewed as an extension of the path model.6
Value of the Research
This research represents a potential contribution to
our understanding of a social problem, to the sociology of
deviance, to medical sociology, and to methodology.
5Otis Dudley Duncan, "Path Analysis: Sociological
Examples," American Journal of Sociology 72 (July, 1966),
Otis Dudley Duncan, Archibald O. Haller, and Alejan-
dro Portes, "Peer Influences on Aspirations: A Reinter-
pretation," American Journal of Sociology 74 (September,
The abuse of alcohol is unquestionably a social problem.
A 1970 report of the United States Comptroller General dis-
cusses the social costs of problem drinking in the United
States. Approximately 4 to 8 per cent of the work force are
estimated to have problem-drinking patterns resulting in a
cost to employers of approximately 25 per cent of the em-
ployees' annual salaries.7 In fact, Von Wiegand points out
that the abuse of alcohol at all levels interferes with
productivity and, consequently, 87 companies among the 500
largest companies in the United States listed in Fortune
magazine have employee alcoholism programs.
Alcohol is estimated to contribute to or be associated
with 50 per cent of the fatal auto accidents. In 1969, this
amounted to approximately 28,000 fatalities and 300,000 to
400,000 injuries. The report estimates that 33 per cent of
all arrests in 1969 were related to alcohol misuse, which
resulted in unestimable expense for arrest, incarceration,
and trial. The report also estimates that 30 per cent of
the patients treated by state mental health programs had a
pattern of problem drinking.
Comptroller General of the United States, "Substan-
tial Cost Savings from Establishment of Alcoholism Program
for Federal Civilian Employees" (Washington, D.C.: U.S.
Government Printing Office, 1970), pp. 1-21.
Ross A. Von Wiegand, "Alcoholism in Industry (U.S.A.),"
British Journal of Addiction 67 (September, 1972), p. 183.
Comptroller General, 1970, pp. 1-21.
A Bureau of Indian Affairs publication states that 75
to 80 per cent of all suicides among American Indians are
alcohol related. While this represents an extreme cost
to the individual, the abuse of alcohol frequently results
in high personal cost, not only for the problem drinker,
but for family, friends, and the community as well.
Sociology of Deviance
Clinard defines deviant behavior as a "violation of
social norms."1 The use of alcohol is socially controlled
by norms concerning the acceptable quantity and frequency
of consumption in specific social situations. For instance,
drinking in moderation is acceptable and even desirable at
cocktail parties, business lunches, and so on. Drinking in
excess of the normative tolerance range or in proscribed
situations is considered to be deviant behavior, however.
The model presented in Chapter III has been deductively
formulated from theories and research findings on deviance.
The major theories include the disposition toward deviance
suggested in Merton's means-ends disjunction theory of
10Bureau of Indian Affairs, "Suicide, Homicide, and
Alcoholism Among American Indians: Guidelines for Help,"
(Washington, D.C.: U.S. Government Printing Office, 1973),
lMarshall B. Clinard, Sociology of Deviant Behavior,
(Third Edition; New York: Holt, Rinehart, and Winston,
Inc., 1968), p. 26.
anomie,12 and the restraint against deviance suggested in
Durkheim's treatment of suicide in which group attachment
reinforces norms which restrain deviant behavior. In
addition, the psychological variables of anxiety and psy-
chopathology have been associated with problem drinking in
the literature.14 The model proposed by this study is a
partial test of each of these competing theories of deviance.
The theoretical concepts are operationalized and variance in
problem drinking is predicted from each theory separately
and in combination.
The abuse of alcohol, in addition to being a form of
deviant behavior, represents a major health problem in our
society. The First Special Report to the U.S. Congress on
Alcohol and Health in 1971 labels alcohol as the most abused
drug in the United States.15 Problem drinking is considered
12Robert K. Merton, Social Theory and Social Structure,
(Enlarged Edition; New York: The Free Press, 1968), p. 229.
13Emile Durkheim, Suicide, translated by J. A. Spauld-
ing and George Simpson (New York: The Free Press, 1951),
14Richard Jessor, Theodore D. Graves, Robert C. Hanson,
and Shirley L. Jessor, Society, Personality, and Deviant
Behavior: A Study of a Tri-Ethnic Community (New York:
Holt, Rinehart, and Winston, Inc., 1968), pp. 24-25.
15Shirley Sirota Rosenberg, Editor, First Special
Report to the U.S. Congress on Alcohol and Health, from the
Secretary of Health, Education, and Welfare (Washington,
D.C.: U.S. Government Printing Office: December, 1971),
to be an illness and therefore within the domain of diag-
nosis and treatment of health care practitioners. The con-
cepts, methods, and theories of sociology may be applied
to furthering our understanding and prediction of what are
basically problems with which health care practitioners are
concerned. This application of sociology fits into what
Straus suggests as a major division within medical sociol-
ogy, sociology in medicine.6 If the health care practi-
tioners are concerned with diagnosing and treating problem
drinking, sociologists in medicine legitimately may attempt
to describe, explain, and predict this problem.
Given the extensive research literature on the use and
abuse of alcohol, it becomes not only legitimate but neces-
sary to question seriously the value to be derived from
additional research. The unique contribution of this re-
search is methodological. First, few studies have collected
data from the general population. Most gathered data from
persons who were caught and identified in one of the social
control nets of society--offenders of legal norms such as
persons arrested for public intoxication or driving while
intoxicated, mental health treatment center patients, or
6Robert Straus, "The Nature and Status of Medical
Sociology," American Sociological Review 22 (April, 1957),
members of organizations such as Alcoholics Anonymous. This
study utilizes data from a random sample of the general popu-
Second, few existing cross-sectional studies have
gathered data from respondents who live in rural or small
communities. Since there is a positive relationship be-
tween rates of problem drinking and city size, most studies
have focused upon residents of urban areas. This study
focuses upon the residents of counties in which the largest
cities did not exceed 70,000 in population, at the time of
Third, few studies have utilized multivariate tech-
niques of data analysis. Most research findings have been
couched in bivariate terms, which suggest single variables
found to be related to problem drinking. In only a few
cases did researchers introduce a third variable to test
whether this association was maintained. This study is
basically a test of a multivariate model which is deductively
constructed from a review of the literature.
Fourth, in models which assume one-way causality,
problem drinking is assumed to be the dependent variable.
There has been no research which examines the interdepen-
dent, reciprocal relationships which may apply between
problem drinking and other variables. This study utilizes
1Cahalan, p. 60.
a technique of data analysis which does not assume one-way
causality but which assumes that at least two of the vari-
ables in the model are reciprocally related.
Within the general theoretical context described above,
these four specific methodological characteristics of the
present study take on significance. Understanding and ex-
planation are enhanced by conceptual and methodological
advances in the study of a phenomenon. The value of this
study is largely dependent upon the impact of these four
methodological factors in verifying, clarifying, or refut-
ing the existing knowledge of problem drinking.
In addition to the above contributions to the under-
standing and explanation of problem drinking, a contribution
of this study is exploration of the utility of the two-stage
least-squares technique for examination of the reciprocal
effects of two or more variables.
Research indicates that whether a person drinks depends
largely on sociological and cultural factors rather than
psychological ones. Significant relationships can be found
between drinking and sociological variables such as sex, age,
socioeconomic status, region of residence, size of community,
and religion, while correlations between drinking and a num-
ber of psychological variables are low.1
Drinking Behavior in the United States
The consumption of alcoholic beverages is typical,
statistically normative behavior in the United States.
Cahalan et al. conducted a national survey of drinking prac-
tices of United States adults during 1964 and 1965. While
22 per cent of the respondents report that they never drink
alcoholic beverages, 68 per cent say they have drunk at
least once during the year. The remaining 10 per cent have
stopped drinking. Mulford reports that approximately 71
Don Cahalan, Ira H. Cisin, and Helen M. Crossley,
American Drinking Patterns (New Haven, Connecticut: College
and University Press, 1969), p. 200.
Ibid., pp. 184-185. This study is being used because
it is most recent and comprehensive. Its findings verify
per cent of the adults he interviewed have drunk alcohol
during the year. A 1974 Gallup Poll indicates that 68 per
cent of people aged 18 and older drink alcoholic beverages.
According to the survey, the use of alcohol is influenced by
sex (although the discrepancy is narrowing), age, income,
region of residence, occupation, education, and religion.
Nearly 20 per cent of those polled report that they sometimes
drink more than they should and 12 per cent report that
liquor has caused trouble in their families.4 A Harris
survey done for the National Institute on Alcohol Abuse and
Alcoholism indicates that one in five persons interviewed
reports that someone close to them--frequently a family
member--drinks too much.
Cahalan et al. classify 32 per cent of their respondents
as abstainers, 15 per cent as infrequent drinkers, 28 per
cent as light drinkers, 13 per cent as moderate drinkers,
and 12 per cent as heavy drinkers. Drinkers are classified
as escape drinkers if they report drinking for reasons such as
Harold A. Mulford, "Drinking and Deviant Drinking,
U.S.A., 1963," Quarterly Journal of Studies on Alcohol
25 (December, 1964), 634-648.
George Gallup, "Alcoholism Is a National Concern,"
Jacksonville Times-Union and Journal (June 9, 1974), p. A-15.
Shirley Sirota Rosenberg, editor, First Special Report
to the United States Congress on Alcohol and Health, from
the Secretary of Health, Education, and Welfare (Washington,
D.C.: U.S. Government Printing Office, December, 1971), p.
Cahalan et al., pp. 184-185.
tension or nervousness. Thirty-two per cent of men drinkers
and 26 per cent of women drinkers are escape drinkers. Of
heavy drinkers, 64 per cent of women and 46 per cent of men
are considered escape drinkers.7
There is substantial demographic variation in drinking
patterns. Men (77 per cent) are more likely to drink than
women (60 per cent) and are more likely to be classified as
heavy drinkers (21 per cent to 5 per cent respectively).
Among men, the age cohort of 30-34 has the lowest percentage
of abstainers and a high rate of heavy drinkers. The age
cohort 45-49 also demonstrates a high rate of heavy drinkers.
The lowest percentage of heavy drinkers is found in the 65+
age cohort. Among women, the highest percentage of heavy
drinkers is found in the age cohorts 21-24 and 45-49, with a
rapid decline in heavy drinking after age 50.8
Respondents of higher socioeconomic status are much
less likely to be abstainers, and are less likely to be
heavy drinkers as well. This relationship is even more pro-
nounced for women than for men. Among the upper socio-
economic status groups, there are only slight differences
between men and women in the proportion of drinkers, while
the difference is pronounced among the lower socioeconomic
Cahalan et al., pp. 168-189.
8bid., p. 22.
Ibid., p. 25.
Ibid., p. 25.
groups. However, the difference between men and women who
are heavy drinkers is not consistent among the various socio-
economic groups. Men and women who are married are less
likely to be heavy drinkers than persons who are single or
divorced. Men raised by single parents tend to develop into
heavy drinkers relatively late in life.11
Respondents from the South are more likely to be ab-
stainers, but among those who drink the proportion of heavy
drinkers approximates that in other regions of the country.
The tendency for higher proportions of heavy drinkers to be
residents of urbanized areas holds for all age and sex
Among men, race does not greatly affect either absti-
nence or heavy drinking, but black women have considerably
higher rates of both abstinence and heavy drinking than do
white women. Religion also exerts an impact upon drink-
ing patterns. Jews and Episcopalians are most likely to be
drinkers when sex, age, and socioeconomic status are con-
trolled. Catholics have the highest proportion of heavy
drinkers, followed by liberal Prostestants. Conservative
Protestants have the smallest proportion of heavy drinkers,
1Cahalan et al., p. 27.
Ibid., p. 37.
Ibid., p. 39.
13 d., p. 48.
Ibid., p. 48.
10 per cent compared with 22 per cent of those who never
The proportion of drinkers is highest among those of
Italian origin (91 per cent), followed by those of Russian,
Polish, or Baltic (86 per cent), and Canadian origin (79
per cent). U.S. whites (of unknown origin) have the highest
proportion of abstainers (54 per cent), followed by those
of Scotch-Irish origin (50 per cent). The proportion of
heavy drinkers among drinkers is highest among those of
Latin American, Caribbean origin (30 per cent), among U.S.
non-whites (23 per cent), and among those of Italian origin
(22 per cent).15
Respondents who report active participation in inter-
personal recreational activities are more likely to drink
than the less active. Men under age 45 and of higher socio-
economic status who are socially active are also more likely
to be heavy drinkers.16
Respondents who drink are more likely to have parents
who drink and approve of drinking than parents who drink
infrequently or disapprove. Men and women drink more
1Cahalan et al., pp. 56-57.
5Ibid., p. 53.
6Ibid., p. 75.
7Ibid., p. 79.
frequently in their own homes or in friends' homes than in
bars or restaurants.1
More of the heavy drinking men and moderate to heavy
drinking women started drinking before age 18. Those of
highest socioeconomic status started drinking later than
others and have continued drinking later in life.19
Seventy-one per cent of heavy drinkers, men and women,
are more likely than other drinkers to report having a
drink to relieve depression and nervousness.2
More respondents drink distilled spirits (57 per cent)
than either beer (50 per cent) or wine (39 per cent). Heavy
drinkers are more likely to drink beer (62 per cent) than
spirits (49 per cent) or wine (5 per cent). However,
approximately 80 per cent of heavy drinkers report drinking
both beer and spirits.21
Bales discusses three major factors which influence
group rates of problem drinking. The first of these is
termed the dynamic factors described as being those anxieties,
tensions, suppressed aggressions, or needs which provide the
motivation for problem drinking. These factors may be char-
acteristic of a society to a greater or lesser extent, but,
combined, they affect the likelihood that problem drinking
1Cahalan et al., p. 99.
Ibid., p. 123.
2Ibid., p. 150.
21Ibid., p. 65.
will occur. The second set of factors is labeled alterna-
tive factors, described as the culturally defined possi-
bilities of channeling or relieving needs, tensions, and
anxieties by adopting behavior other than excessive drink-
ing. The third set of factors is labeled orienting factors,
described as the normative attitudes of the group culture
toward drinking. Bales outlines four of these cultural
attitudes. These are: 1) utilitarian--drinking for per-
sonal reasons such as the reduction of anxiety, conflict,
and so on; 2) ritual--drinking restricted to religious,
dietary, or ceremonial occasions; 3) convivial--drinking to
symbolize and facilitate group interaction and solidarity;
and 4) abstinence--prohibition of the use of alcohol for
any purpose. Utilitarian attitudes are highly associated
with high rates of problem drinking. This model has been
demonstrated to have broad utility in explaining the incidence
of cross-cultural problem drinking.2
Abuse of Alcohol as Deviant Behavior
Although drinking alcoholic beverages is acceptable,
perhaps even desirable, and encouraged in many contemporary"
social situations, drinking in excess of the normative
tolerance range may be considered deviant behavior. Jessor
et al. report that the respondents of the tri-ethnic
2Robert F. Bales, "Cultural Differences in Rates of
Alcoholism," Quarterly Journal of Studies on Alcohol 6
(March, 1946), 480-499.
community they studied rate the immoderate use of alcohol
near the "very wrong" end of a scale, as contrasted to a
number of other non-normative behaviors. Blizard reports
that 34 per cent of his respondents would not like to live
next door to an alcoholic, 73 per cent report they would
not like to work with one, 33 per cent report they want no
contact with an alcoholic, while very few report they would
accept close and continuous contact with an alcoholic.
Simmons finds that 46 per cent of his respondents list abuse
of alcohol among responses to the question, "What is
The primary theoretical foundation of this research
falls within the functionalist tradition of Durkheim and
Merton. Deviant behavior violates the socially con-
structed norms considered to be binding upon persons
2Richard Jessor, Theodore D. Graves, Robert C. Hanson,
and Shirley L. Jessor, Society, Personality, and Deviant
Behavior: A Study of a Tri-Ethnic Community (New York:
Holt, Rinehart, and Winston, Inc., 1968), p. 41.
24Peter Blizard, "The Public Image and Social Rejection
of the Alcoholic in New Zealand," Quarterly Journal of
Studies on Alcohol 30 (September, 1968), 696.
J. L. Simmons, "Public Stereotypes of Deviants,"
The Substance of Sociology, edited by Ephraim H. Mizruchi
(New York: Appleton-Century-Crofts, 1967), p. 272.
2This study deals with members of the community who
have not already been caught in a social control net and
labeled as "alcoholics" or "problem drinkers." Rather, it
deals with primary deviance, or those individuals who
deviate from the norms in their everyday functioning, but
who have not been assigned the deviant role. Therefore,
this discussion does not concern itself with labeling theory
occupying various positions within the social structure,27
and is likely to mobilize social control agencies into
corrective action.28 Furthermore, differential rates of
deviant behavior are assumed to be at least partially the
result of differences in the characteristics of the social
Analysis of the impact of differences within the social
structure upon the occurrence of deviant behavior may be
facilitated by the heuristic division of the social struc-
ture into three component substructures: the opportunity
structure, the normative structure, and the social control
structure. The individual is assumed to occupy a position
within each of these substructures simultaneously. Relative
position within each of these substructures disposes or
constrains the individual with regard to deviant behavior.30
because labeling theory is most interested in the process
of secondary deviance. For further information see Edwin M.
Lemert, Social Pathology (New York: McGraw-Hill Book Co.,
1951), and Howard S. Becker, Outsiders: Studies in the
Sociology of Deviance (New York: The Free Press of Glencoe,
Robert K. Merton, "Social Problems and Sociological
Theory," Contemporary Social Problems, edited by Robert K.
Merton and Robert A. Nisbet (New York: Harcourt, Brace,
and World, 1961), pp. 723-724.
Kai T. Erickson, "Notes on the Sociology of Deviance,"
The Other Side: Perspectives on Deviance, edited by Howard S.
Becker (New York: The Free Press, 1964), pp. 10-11.
Robert K. Merton, Social Theory and Sccial Struc-
ture (Enlarged Edition; New York: The Free Press, 1968),
3The researcher is indebted to Jessor et al. for this
The Opportunity Structure
Merton constructs a middle-range theory to explain
rates of conformity and various types of deviant behavior.
This theory expands upon Durkheim's concept of anomie--a
condition in which the social structure does not provide
the individual with clear-cut guidelines for action. Behavior
differentials are produced by characteristics of the social
structure. Specifically, Merton's theory states that
there are socially patterned differences in:
1. exposure to culturally emphasized goals and norms
which define, regulate, and control the acceptable means of
reaching these goals;
2. acceptance of these goals and norms;
3. relative access or opportunity to achieve these
4. the degree of disjunction between the culturally
emphasized goals and access to legitimate means of achiev-
ing the goals;
5. the degree of anomie; and
6. the rate of aberrant or deviant behavior.32
Merton hypothesizes that deviant behavior is predisposed by
disjunction between culturally emphasized goals and aspira-
tions and a person's socially structured means of realizing
31Merton, Social Theory and Social Structure, pp. 224-229.
3Ibid., p. 229.
them. From a sociological point of view, deviant behavior
does not necessarily involve a violation of legal norms, nor
is it always dysfunctional for society.
The theory indicates a differential in the individual's
acceptance of the culturally emphasized goals or interests
and the regulatory norms which control legitimate means of
achieving those goals. Merton constructs a typology of five
modes of individual adaptation based upon this differential.
These modes of adaptation include retreatism and rebellion.
In retreatism, the culturally emphasized goals and means are
thoroughly assimilated by the individual, but legitimate
efforts do not produce success. Unsuccessful in using legit-
imate means and unwilling to use illegitimate means which
might be successful, the individual resolves the conflict
by escape, abandoning both goals and means. Much problem
drinking fits into this mode. In rebellion, the individual
rejects normative goals and means and replaces them with
substitute goals and means, which may include heavy drink-
In his description of social stratification, Weber
presents the concept of differential life chances or life
fate, in which an individual's status determines opportuni-
ties to accomplish certain life goals.34 For instance, the
3Merton, Social Theory and Social Structure, pp. 241-
Max Weber, From Max Weber: Essays in Sociology,
edited by Hans H. Gerth and C. Wright Mills (New York:
Dohrenwends' summarization of the literature on psychopath-
ology indicates that those persons of lower socioeconomic
status and from disadvantaged racial-ethnic minority groups
are less likely to be able to improve their situations and
demonstrate higher rates of psychopathology than persons of
more advantaged status.3
One's position within ". the opportunity structure
serves as an index of degree of instigation to the use of
illegitimate means in adapting to value-access disjuction."36
The Normative Structure
The normative structure constrains deviant behavior.
The individual's acceptance or internalization of the
socially constructed norms--standards and expectations of
what constitutes appropriate behavior in various social
situations--serves to dispose behavior toward conformity and
Durkheim, in his classic study of suicide, analyzes
the societal state of anomie in which regulatory norms be-
come inoperative and clear-cut guidelines fcr behavior no
longer exist. This state of deregulation is pronounced in
times of crisis and transition and following changes in
Oxford University Press, 1946), pp. 180-195.
3Bruce P. Dohrenwend and Barbara Snell Dohrenwend,
Social Status and Psychological Disorder: A Causal Inquiry
(New York: John Wiley and Sons, Inc., 1969).
36Jessor et al., p. 58.
social status, power, and wealth which accompany economic
disaster or prosperity.3
Klapp suggests that "consensus should be considered
as a dimension and measure of social integration, and in-
versely of anomie. ."38 In other words, the strength
of the normative system depends upon the existence of be-
liefs and expectations which are widely shared and strongly
held by members of the social group. An individual who is
well integrated into social groups which provide clear-cut
expectations of what is and is not appropriate will likely
conform rather than deviate from expectations.
Durkheim and Merton relate the normative structure to
the opportunity structure by pointing out that when behavior
deviates from the normative system but is successful, the
legitimacy and force of the institutional norms for self
and others are lessened or eliminated.
The Social Control Structure
The social control structure considers the individual's
differential access to illegitimate means and opportunity
to learn specific forms of deviance and the sanctions invoked
Emile Durkheim, Suicide, translated by J. A. Spaulding
and George Simpson (New York: The Free Press, 1951), pp. 246-
3Orrin Klapp, "The Concept of Consensus and Its Impor-
tance," Sociology and Social Research 41 (May-June, 1957),
3Durkheim; and Merton, Social Theory and Social Struc-
ture, pp. 224-229.
by deviant behavior. Association with persons who perform
deviant roles not only illustrates the possibility of per-
forming this behavior, but when models are "significant
others," the individual may develop a self-conception or
self-identity which incorporates the deviant model. Integra-
tion into the nuclear family, group, and organizational
network is assumed to increase the likelihood that one will
be rewarded for conforming behavior and punished for deviant
behavior. "The less effective the operation of the social
control structure, the greater the availability of oppor-
tunities to engage in deviance and, consequently, the higher
the deviance rates.40
These three overlapping substructures have been con-
structed, heuristically, in order to facilitate analysis
of the effects of the social structure upon behavior. It is
assumed that knowledge of the individual's position within
each of these substructures will provide a probabilistic
estimate of behavior.
Propositional Review of the Literature
The prevalence of problem drinking is unequally dis-
tributed among various segments and subgroups. Furthermore,
despite prohibitionist theories, drinking per se and problem
drinking do not appear to be closely associated, as some
population aggregates demonstrate high rates of drinking and
low rates of problem drinking.
40 or et al., p. 77.
Jessor et al., p. 77.
The purpose of this section is to describe the support
found in the literature for bivariate propositions suggested
by the preceding social theory of deviance. This review
of the literature will draw from studies on alcoholism
(studies of persons who have been identified or treated as
alcoholics) and on problem drinking (epidemiological studies
of people in the community who have not been institution-
alized). While this study deals with problem drinking,
studies on alcoholism are reviewed because they may suggest
significant variables to include in a model of problem
The most sophisticated of the problem drinking studies
was done by Cahalan, who reports on the findings of the
second stage of a longitudinal study of drinking patterns.41
These data were collected in 1967, in 1,359 personal inter-
views with adults representative of the total United States
Approximately 75 items are combined into 11 specific
problem areas. These are:
(1) frequent intoxication, or exceeding what was
defined as a moderate level in a combined frequency and
amount-per-occasion measure, or getting intoxicated
fairly often; (2) binge drinking--being intoxicated
for more than one day at a time; (3) symptomatic
41Don Cahalan, Problem Drinkers (San Francisco:
Jossey-Bass, Inc., 1970).
42These persons correspond to 72 per cent of the eli-
gible respondents selected to be interviewed in the first
stage of the study in 1964-1965. Ibid., p. 20.
drinking behavior (symptomatic dependence upon alcohol),
as inferred from finding it difficult to stop drinking
once started, blackouts or memory lapses after drinking,
sneaking drinks, and so on; (4) psychological dependence
upon alcohol; (5) problems with current spouse or with
relatives related to one's drinking; (6) problems with
friends or neighbors over one's work or employment oc-
curring in relation to one's drinking; (9) health (for
example, physician advised respondent to cut down on
drinking); (10) financial problems connected with one's
drinking; (11) belligerence or fighting associated with
Each respondent is assigned a "current problem score"
based upon the respondent's experience with these 11 types
of drinking-related problems during the previous three years.
Fifteen per cent of the men and 4 per cent of the women--
9 per cent of the total--are judged to have a high score on
the index while 43 per cent of the men and 21 per cent of
the women--31 per cent of the total--report some problem in
the preceding three years.4
The following propositions are written in terms of
social and psychological variables and their relationship
with problem drinking. The propositions simply state that
there is a relationship between the variables,and the re-
view of the literature which follows describes the nature
of the relationship.
There is a relationship between sex and problem drink-
SCahalan, pp. 26-27.
Ibid., pp. 26-27.
In a national survey of drinking patterns Cahalan finds
sex to be the most influential single factor in predicting
problem drinking.45 Encel et al., Room, Edwards et al.,
Jessor et al., and Mulford find support for this proposi-
tion. However, Edwards et al. find no difference in the
number of drinking problems between men and women who report
drinking the same quantity and frequency of alcohol.47
Between 5.5 and 6 times more men than women in this
country are alcoholics. In England, the ratio is approxi-
mately 1.5 males to one female, and Scandinavian countries
have a ratio of 12 to 1.4
Lawrence and Maxwell point out that most societies
are relatively intolerant toward female drunkenness while
the norms are likely to be more ambiguous for men, permit-
ting and encouraging deviance rates among men. Ullman
Cahalan, pp. 96-97.
S. Encel, K. C. Kotowicz, and H. E. Resler, "Drinking
Patterns in Sydney, Austrailia," Quarterly Journal of Stud-
ies on Alcohol Supplement #6 (May, 1972), 8; Robin Room,
"Drinking Patterns in Large U.S. Cities: A Comparison of
San Francisco and National Samples," Quarterly Journal of
Studies on Alcohol Supplement #6 (May, 1972), 47; Griffith
Edwards, Celia Hensman, and Julian Peto, "Drinking in a
London Suburb: III. Comparisons of Drinking Troubles Among
Men and Women," Quarterly Journal of Studies on Alcohol
Supplement #6 (May, 1972), 120-128; Jessor et al., pp. 181-
182; and Mulford, p. 640.
4Edwards et al., "Drinking in a London Suburb:III," p. 128.
Charles R. Snyder, "A Sociological View of the
Etiology of Alcoholism," Alcoholism, edited by David J.
Pittman (Springfield, Illinois: Charles C. Thomas, 1959),
49oseph J. Lawrence and Milton A. Maxwell, "Drinking
Joseph J. Lawrence and Milton A. Maxwell, "Drinking
finds men more likely than women to become somewhat intoxi-
cated at the time of their first drinks and suggests that
men are more likely than women to experience their first
drinks as socially dissonant acts.50 This may be partly
because males start drinking at younger ages, when it is
less acceptable for them to drink. Wechsler and Thum, in
their study of teen-age drinking and drug use,find that at
the junior-high level many more boys than girls drink heavily,
while there is no sex differential at the senior high
Riley finds men more likely than women to drink for
personal rather than social reasons. This utilitarian or
affective drinking is more likely than social drinking to
lead to drinking problems.52 Rimmer finds that women alco-
holics start drinking at later ages than men, but Ullman
and Socioeconomic Status," Society, Culture, and Drinking
Patterns, edited by David J. Pittman and Charles R. Snyder
(New York: John Wiley and Sons, Inc., 1962),p. 144.
5Albert Ullman, "First Drinking Experience as Related
to Age and Sex," Society, Culture, and Drinking Patterns,
edited by David J. Pittman and Charles R. Snyder (New York:
John Wiley and Sons, Inc., 1962), p. 263.
5Henry Wechsler and Denise Thum, "Teen-age Drinking,
Drug Use, and Social Correlates," Quarterly Journal of
Studies on Alcohol 34 (December, 1973), 1223.
52John W. Riley, Jr., Charles F. Marden, and Marcia
Lifshitz, "The Motivational Pattern of Drinking," Quarterly
Journal of Studies on Alcohol 9 (December, 1948), 353-362.
5John Rimmer, F. N. Pitts, Jr., T. Reich, and G. Wino-
ken, "Alcoholism: II. Sex, Socioeconomic Status, and Race
in Two Hospitalized Samples," Quarterly Journal of Studies
on Alcohol 32 (December, 1971), 946.
finds that men and women start drinking at approximately the
same age. Child et al. find 53 of 189 societies to show
differences in rates of drinking and problem drinking between
men and women, showing men to have the higher rates of
In 1973, Fillmore did a follow-up study on a sample of
the respondents of the Straus and Bacon study of 1953. Re-
spondents were college students at the time of the earlier
study and were of upper-middle socioeconomic status in 1973.
In both studies, percentages of drinkers are similar for men
and women, but of those who drank, 42 per cent of men com-
pared with 11 per cent of women were problem drinkers in the
early study. In the follow-up study, 17 per cent of men and
12 per cent of women are problem drinkers. While men have
changed to moderate drinking patterns over the 20 years, the
percentage of women problem drinkers remains about the
Knupfer and Room suggest that women might be more con-
cerned than men with the consequences of intoxication. For
example, sexual fidelity has been more rigidly expected of
Ullman, p. 263.
Irvin L. Child, Herbert Barry III, and Margaret K.
Bacon, "A Cross-Cultural Study of Drinking: III. Sex
Differences," Quarterly Journal of Studies on Alcohol
Supplement #3 (April, 1965), 60.
56Kaye Middleton Fillmore, "Drinking and Problem
Drinking in Early Adulthood and Middle Age," Quarterly
Journal of Studies on Alcohol 35 (September, 1974), 819-
of women than men, but women who are intoxicated are more
likely to be vulnerable to sexual exploitation. They sug-
gest that loss of control "costs" the woman more than the
man, which has the effect of reducing problem drinking among
women. This may be changing, however. Kenyon points out
that emancipationin of the female'together with the move-
ment towards equal pay and the desire of the working wife to
obtain a better standard of living for the family have all
led to a significant increase in the proportion of female
to male alcoholics."58
There is a relationship between age and problem drink-
Cahalan finds drinking problems among men to be highest
in the age cohort 21-29, dropping fairly sharply in the age
category 30-39, gradually through age 69, and sharply drop-
ping off after age 70.59 Mulford finds a sharp drop in heavy
drinking after age 60.60 Knupfer and Room support these
findings and suggest that older persons are more likely to
5Genevieve Knupfer and Robin Room, "Age, Sex, and
Social Class as Factors in Amount of Drinking in a Metro-
politan Community," Social Problems 12 (Fall, 1964), 228.
W. H. Kenyon, "About the Illness Alcoholism,"
Alcohol and Health Notes (December, 1972), 2.
Cahalan, pp. 41-45.
6Mulford, p. 640.
be concerned about their health and less likely to move in
social circles which permit or encourage problem drinking.
Encel et al. and Room find heavy drinking to decline with
age, although Edwards et al. find drinking problems to
increase with age. While earlier literature on alcoholism
viewed it as an ever-increasing, debilitating process, it
would appear from these recent studies that drinking pat-
terns shift to moderation with age.
The Fillmore study indicates that middle-class men who
are either problem drinkers or abstainers in the late teens
or early 20's shift toward moderate drinking patterns by
the late 30's and early 40's. Only about one-fourth of
abstainers in the early study remain abstinent, while approxi-
mately 90 per cent of drinkers are still drinkers.64 Fill-
more also points out that many of the younger persons who
drink moderate amounts of alcohol infrequently experience
drinking problems, while drinking problems in adults are
correlated with drinking large amounts in greater fre-
Cahalan finds women to have a later onset of problem
drinking than men, with relatively few problem drinkers
Knupfer and Room, "Factors in Amount of Drinking,"
6Encel et al., p. 11; Room, p. 40.
6Edwards et al., "Drinking in a London Surburb: II," p, 103.
64Fillmore, p. 824. 65Ibid., p. 829.
reported in the 20's, and a dramatic drop in the incidence
after age 50. Cahalan suggests that women are likely to
be introduced to heavy drinking by their men friends or
husbands, while men are more likely to be introduced to
heavy drinking by other men while they are quite young.6
Knupfer and Room and Encel et al. also find that women in
the age cohort 30-49 are more likely to have higher rates
of problem drinking.68
There is a relationship between marital status and
Mulford, Sadown et al., and Encel et al. find that
unmarried persons are more likely to have higher rates of
problem drinking than married people.69 Williams finds a
high percentage (68 per cent) of alcoholics in a treatment
center to be divorced. Edwards et al. does not find
marital status to be significant when age is controlled,
Cahalan, pp. 41-45.
Knupfer and Room, "Factors in Amount of Drinking,"
p. 228; Encel et al., p. 11.
6Mulford, p. 640; Roland Sadown, Giorgio Lolli,
and Milton Silverman, Drinking in French Culture (New
Brunswick, New Jersey: Rutgers Center of Alcohol Studies,
1965), p. 108; and Encel et al., p. 11.
70James Williams, Characteristics of an Alcoholic
Sample (Avon Park, Florida: State of Florida Alcoholic
Rehabilitation Program, 1964), p. 3.
however.71 Lindbeck finds the incidence of divorce or
separation among female alcoholics to be higher than among
nonalcoholic women. These women describe marriage as a
painful and disappointing experience.72 Married persons
are likely to have greater social control over their drink-
ing patterns than the unmarried.
Bacon, Kinsey, and others produce research to support
this, but caution should be exercised in interpreting this
finding due to the interactive effect of the variables.
They indicate, however, that those who become alcoholic
tend to be disproportionately single, separated, or di-
vorced prior to the onset of symptoms.73
There is a relationship between race and problem drink-
Maddox finds that twice as many black college students
drink heavily as did the respondents in the Bacon and Straus
study of college students.74 Viamontes and Powell studied
Edwards et al., "Drinking in a London Suburb: III,"
Vera Lindbeck, "The Woman Alcoholic: A Review of
the Literature," The International Journal of the Addictions
7 (March, 1972), 575.
Selden D. Bacon, "Inebriety, Social Integration, and
Marriage," Quarterly Journal of Studies on Alcohol 5 (June,
1944), 86-125; and Barry Kinsey, The Female Alcoholic: A
Social Psychological Study (Springfield, Illinois: Charles C.
Thomas, 1966), p. 109.
George L. Maddox and Jay R. Williams, "Drinking
demographic characteristics of black and white alcoholic
men and find that black men begin drinking at a younger age
than whites and lose control over drinking sooner. They
generally seek treatment at a younger age. A greater propor-
tion of black than white alcoholics are married.75 Cahalan
reports that black women have much higher rates of heavy
drinking, among those who drink, than do white women.76
Jessor et al. report that American Indians drink nearly
seven times as much alcohol as Anglos. They are more
likely to drink in order to solve personal problems, and have
six times as many drinking-related problems as other
There is a relationship between religion and problem
Mulford finds the heaviest drinking to occur among
Catholics, Lutherans, and Protestants of no specified de-
nomination.78 Encel et al. find Catholics to be heavier
Behavior of Negro Collegians," Quarterly Journal of Studies
on Alcohol 29 (March, 1968), 126-127.
75Jorge A. Viamontes and Barbara J. Powell, "Research
Note: Demographic Characteristics of Black and White Male
Alcoholics," The International Journal of the Addictions
9 (June, 1974), 489-494.
Cahalan, et al., p. 48.
Jessor, et al., pp. 182-183.
78Mulford, p. 640.
Mulford, p. 640.
drinkers than Anglicans or Presbyterians.79 Straus and
Baccn find least drinking problems among Jewish students
and the most among the Mormons, Protestants, and Catholics
Walters and McCord and McCord report that,if parents
are active church members, their children, when they drink,
are more likely to become alcoholic.81 Kinsey finds that
mothers of female alcoholics are religiously active and
fathers are inactive.82 According to Walters, mothers of
alcoholics are significantly more active religiously than
mothers of nonalcoholics.83
There is a relationship between residential area and
Cahalan finds higher rates of problem drinking among
those living in urbanized'areas. He suggests that younger
men and those of lower status tend to be attracted to large
7Encel et al., p. 11.
8Robert Straus and Selden D. Bacon, "The Problems of
Drinking in College," Society, Culture, and Drinking Pat-
terns, edited by David J. Pittman and Charles R. Snyder
(New York: John Wiley and Sons, Inc., 1962), p. 253.
8Orville S. Walters, "The Religious Background of 50
Alcoholics," Quarterly Journal of Studies on Alcohol 18
(September, 1957), 407; and William McCord and Joan McCord,
Origins of Alcoholism (Stanford, California: Stanford Univ-
sity Press, 1960),p. 52.
82Kinsey, p. 110.
83Walters, p. 408.
Walters, p. 408.
cities, increasing the effect of urbanization. Mulford
and Room support this finding.85 (In the United States, the
coastal cities have higher rates of problem drinking than
the interior cities. These cities are in the traditionally
"wet" areas of the country.) Wallace finds that in coun-
tries such as France and Italy where drinking has been a
long-term pattern over generations, urbanization does not
produce a higher rate of problem drinking.87
There is a relationship between income and problem
Mulford and Encel et al. find that those persons with
a low level of income are more likely to have higher rates
of problem drinking than those with a higher level of
income. Jessor et al. find lower rates of problem drink-
ing among higher income persons. Cahalan et al. find
higher rates of heavy drinking among the higher income
8Cahalan, pp. 56-57.
8Mulford, p. 640; and Room, p. 40.
8Room, p. 40.
87Jean G. Wallace, "Drinkers and Abstainers in Norway:
A National Survey," Quarterly Journal of Studies on Alcohol
Supplement #6 (May, 1972), 150.
8Mulford, p. 640; and Encel et al., p. 11.
89Jessor et al., pp. 182-183.
category and higher rates of problem drinking among the
lower income category.90
There is a relationship between socioeconomic status
and problem drinking.
Cahalan finds that men of high status at all ages have
lower rates of problem drinking than those of low status.
After age 40, the higher-status men demonstrate a dramatic
decrease in problem drinking not shared by those in lower-
status positions. It should be noted that these data are
cross-sectional and, therefore, a competing explanation
could be the drift of high-status problem drinkers to lower-
Mulford finds the heaviest drinking to be in the status
categories next to the lowest and next to the highest.92
Knupfer and Room find heavier drinking in higher socioeconomic
categories, regardless of age.93 Encel et al. find the
heaviest drinking among those of the highest and lowest
socioeconomic status. Edwards et al. report alcoholism
9Cahalan et al., pp. 28-29.
9Cahalan, pp. 45-49.
9Mulford, p. 640.
9Knupfer and Room, "Factors in Amount of Drinking,"
94Encel et al., p. 8.
Encel et al., p. 8.
to be highest among female professionals and unskilled
males. Lindbeck finds more identified alcoholics among
the upper-middle-class women than lower class but questions
whether this may simply be the result of upper-class per-
sons seeking treatment more readily.9 Jessor et al. find
the highest rates of problem drinking among lower socio-
economic status persons. They describe fewer constraints
and greater access to deviant drinking behavior as the
explanation of the greater rates of problem drinking among
the lower socioeconomic status persons.
Schuckit and Gunderson, in their study of alcoholism
and job type in the Navy, find that men who drink heavily
tend to have high-risk nontechnical jobs and come from
lower socioeconomic backgrounds.9
There is a relationship between level of education
and problem drinking.
Cahalan et al. find no consistent relationship between
education and problem drinking, but report that men
9Edwards et al., "Drinking in a London Suburb: III,"
9Lindbeck, p. 568.
97Jessor et al., pp. 182-183.
9Marc A. Schuckit and E. K. Eric Gunderson, "The
Association Between Alcoholism and Job Type in the Navy,"
Quarterly Journal of Studies on Alcohol 35 (June, 1974),
and women who have completed high school or have gone to
college and have not graduated have higher rates of heavy
drinking. Encel et al. also find high school graduate men
and women who have attended college but did not graduate to
have the highest rates of heavy drinking.1
There is a relationship between organizational integra-
tion and problem drinking.
Sadown et al. and Bahr find that persons who have
limited participation in community activities are more likely
to have higher rates of problem drinking than those with
higher levels of participation. Less involvement in the
activities of the community produces less involvement,
commitment, and response to the normative drinking patterns
of a community. Persons who are not responsive to the
community's normative drinking patterns are likely to be
excluded from the community activities. Blizard finds that
very few of his respondents would accept close and con-
tinuous contact with an alcoholic; 73 per cent would not
9Cahalan et al., p. 140.
1Encel et al., p. 11.
1Sadown et al., p. 114; and Howard Bahr, "Lifetime
Affiliation Patterns of Early and Late-Onset Heavy Drinkers
on Skid Row," Quarterly Journal of Studies on Alcohol 30
(September, 1969), 655.
like to work with one, and 34 per cent would not like to
live next door to one.
Kinsey finds that, with alcoholics, normal peer group
activities are either absent or require heavy drinking.1
Bacon finds significantly less social participation in the
surrounding culture, and social activities tend to be
solitary or with casual acquaintances.1 Singer et al.
find greater social isolation among alcoholics than non-
Kinsey reports that as loss-of-control-type drinking
occurs, there is a decrease in church participation and
social activities, accompanied by an increase in intra-
family conflict.106 Drinking groups change as efforts are
made to associate with those who approve of the current
drinking pattern, thereby reducing the individual-group
There is a relationship between value-access disjunc-
tion and problem drinking.
102Blizard, p. 696.
103Kinsey, pp. 138-189.
104Bacon, "Inebriety, Social Integration, and Marriage,"
105Estelle Singer, Howard T. Blaine, and Richard
Kasschau, "Alcoholism and Social Isolation," Journal of
Abnormal and Social Psychology 69 (December, 1964), 681-685.
106Ki p 0, 137.
Kinsey, pp. 110, 137.
Jessor et al. construct an index of value-access dis-
junction which includes measures of age, marital status,
language spoken in home, occupation, education, intergenera-
tion mobility, religion, and social participation. They
report that, as the level of value-access disjunction in-
creases, problem drinking increases among their respondents
from a tri-ethnic community.1
Cahalan et al. report that heavy drinkers place
slightly greater emphasis upon goals of family life and
friends, and on desire for emotional security and happiness,
but they voice slightly lower levels of satisfaction in meet-
ing life goals than do lighter or nondrinkers. Female
heavy drinkers are more likely to express dissatisfaction
with their (or husband's) occupation, neighborhood, and
educational attainment, than other females.1
Wechsler and Thum's study of teen-age drinkers indi-
cates that heavy drinkers are less likely to be planning
to go to college, and tend to receive lower grades, thereby
cutting down their chances to attain the common status
goals in American society.1
Cahalan et al. and Encel et al. report that persons
who have gone to college but did not graduate have higher
107Jessor et al., p. 237.
1Cahalan et al., p. 192.
109Wechsler and Thum, p. 1224.
Wechsler and Thum, p. 1224.
rates of problem drinking.10 These persons apparently did
not attain their goal of graduating from college.
There is a relationship between psychopathology and
Kinsey finds that alcoholic women generally tend to
view self as inadequate and to define self in undesirable
terms. Gynther et al. also support this proposition.
Halpern, using the Rorschach, found that alcoholic-
prone persons have not sufficiently developed the usual
mechanisms against threat, at least to such an extent that
these mechanisms are effective.13 McCord and McCord
report that boys who consciously and openly express feelings
of inferiority become alcoholic significantly less fre-
quently than those who do not. This finding may suggest
the overuse of denial as a primitive mental mechanism
among alcoholic persons. Both Mowrer and the McCords find
Cahalan et al., p. 140; and Encel et al., p. 11.
lKinsey, p. 115.
2Malcom Gynther, Charles Presher, and Robert McDonald,
"Personal and Interpersonal Factors Associated with Alco-
holism," Quarterly Journal of Studies on Alcohol 20 (June
Florence Halpern, "Studies of Compulsive Drinkers:
Psychological Test Results," Quarterly Journal of Studies
of Alcohol 6 (March, 1946), 468-479.
114McCord and McCord, p. 136.
McCord and McCord, p. 136.
that boys who become alcoholic have earlier engaged in
activities which are associated with masculinity, such as
smoking or visiting prostitutes, supporting the above
interpretation. Horton finds alcoholism rates to be
higher in societies where anxiety is high and few substi-
tute outlets are allowed for either anxiety or aggres-
Landis, using the Rorschach, finds that individuals
less able to express aggressive feelings and acts are more
likely to become alcoholic.17 McCord et al. and Palola
et al. find self-destructiveness to be related to alco-
holism.118 If self-destructive thoughts and tendencies are
interpreted as the internalization of thoughts and feelings
which are not acceptable to the individual to express
5Harriet R. Mowrer, "Alcoholism and the Family,"
Journal of Criminal Psychopathology 3 (July, 1941), 90-99;
and McCord and McCord, p. 154.
6Donald Horton, "The Functions of Alcohol in Primi-
tive Societies: A Cross-Cultural Study," Quarterly Journal
of Studies on Alcohol 4 (September, 1943), 199-320.
Carney Landis, "Theories of the Alcoholic Personal-
ity," Lecture 11, Alcohol, Science, and Society (New Haven,
Connecticut: Quarterly Journal of Studies on Alcohol, Inc.,
1945), pp. 130-133.
W8illiam McCord, Joan McCord, and Jon Gudeman, "Some
Current Theories of Alcoholism: A Longitudinal Evaluation,"
Quarterly Journal of Studies on Alcohol 20 (December, 1959),
727-749; and Ernest Palola, Theodore Dorpat, and William
Larson, "Alcoholism and Suicidal Behavior," Society, Culture,
and Drinking Patterns, edited by David J. Pittman and Charles
R. Snyder (New York: John Wiley and Sons, Inc., 1962),pp.
externally, then the McCord and McCord research also sup-
ports this proposition.
Witkin et al., in a series of perceptual experiments,
find alcoholics to be more dependent upon external stimuli
than nonalcoholics.120 McCord and McCord support this by
identifying a cluster of behaviors which indicates unmet
dependency needs and resulting dependency conflict.1
Kinsey also finds a high level of dependency among female
Wechsler and Thum find that teen-agers who drink
heavily are more likely to have used illegal drugs and to
have been involved in delinquent (antisocial) activities,
including trouble with the police, than light drinkers and
abstainers. These teen-agers also tend to rate themselves
as having "more personal problems" than their contempo-
There is a relationship between depression and
9McCord and McCord, p. 137.
1Herman Witkin, Stephen Karp, and Donald Goodenough,
"Dependence in Alcoholism," Quarterly Journal of Studies on
Alcohol 20 (September, 1959), 493-504.
121McCord and McCord, pp. 87-91.
122Kinsey, p. 183.
123Wechsler and Thum, pp. 1222-1225.
Overall, using the MMPI, finds the alcoholic profile
to be neurotic, depressive, anxious, passive, and depen-
dent. This finding, however, is based on profiles of
labeled alcoholics, and caution should be used in inter-
preting it as an independent variable.
Lindbeck says the alcoholic woman emerges from a
frustrating, depriving background which has contributed to
dependency, self-depreciation, insecurity in her sex role,
feelings of inadequacy and social ineptness, and sensitivity
to loss. This description is one of a depressed person.
The literature review serves the purpose of 1) placing
the present study in the context of relevant past research,
2) indicating the range of opinion and research findings,
3) suggesting hypotheses for the present study, and 4) speci-
fying the unique aspects of the present study.
The majority of the past research efforts have focused
upon description of alcoholics under treatment or comparison
of alcoholics with nonalcoholics. A few studies have focused
upon persons in the community who, while not publicly
labeled alcoholics, have problems as a consequence of drink-
ing. There is considerable similarity between the findings
24ohn Overall, "MMPI Personality Patterns of Alco-
holics and Narcotic Addicts," Quarterly Journal of Studies
on Alcohol 34 (March, 1973), 104-111.
125Lindbeck, p. 575.
of both types of studies, but there is also support for sig-
nificant differences. The etiology of problem drinking may
be understood as the result of the combined effects of
social structural variables and psychological factors.
The literature review further reveals that it is
neither logical nor reasonable to consider the relationship
between problem drinking and other variables as a simple
one-way causal model. Rather, some variables may be viewed
as simultaneously interdependent, exerting a reciprocal
effect one on the other. For example, the relationship
between perceived value-access disjunction or anxiety and
problem drinking may be considered as such a reciprocal
relationship in which a change in anxiety may be accom-
panied by a change in problem drinking, followed by a change
in anxiety and so on. The researcher knows of no research
in the area of problem drinking that examines the recipro-
cal relationship which applies among problem drinking and
other variables in the model.
The purpose of this study is to describe, explain,
and predict the conditions under which adult residents of
four central Florida counties experience problem drinking
as measured by a problem-drinking index. In pursuit of a
valid and reliable answer to this problem there are certain
assumptions and criteria which serve as guides to the re-
searcher. For various reasons these criteria may not always
be met.1 Violation of some of these criteria is more serious
than violation of others,and it is incumbent upon the re-
searcher to attempt to minimize error, to know the conse-
quences of nonfulfillment, and to know how to test whether
or not the criteria are satisfied and what methods are
available when the criteria are not satisfied. This chapter
describes the decisions, operations, and methods utilized in
the effort to make some valid and reliable statements about
the above problem.
Travis Hirschi and Hanan C. Selvin, Delinquency
Research: An Appraisal of Analytic Methods (New York:
The Free Press, 1967), p. 6.
J. Johnston, Econometric Methods (Second Edition;
New York: McGraw-Hill Book Co., 1972), p. 160.
Specifically, the researcher follows a clearly defined
series of activities which begins with an identification of
the relevant existing theory and research findings and ends
with a clarification, modification, support, or refutation
of these existing statements, based upon an empirical test.
The steps between these points include formulation of the
existing theoretical statements into a series of explicit
mathematical equations, operationalization of these theoret-
ical concepts so that observable measurements may be col-
lected, construction of an instrument for data collection,
selection of an appropriate sample to whom the instrument may
be administered, collection of the data, coding, scaling,
and measurement of these data, selection and implementation
of data analysis techniques, and report and interpretation
of the findings. A description of these methodological
Research and theory are inextricably interdependent.
Land suggests that sociologists who rely upon cross-sectional
data may benefit from specifying the complex theoretical
assumptions of their research in mathematical models. By
specifying, for each dependent variable in the model, the
conditions under which its value changes, results may be
3Kenneth C. Land, "Formal Theory," Sociological Meth-
odology 1971, edited by Herbert L. Costner (San Francisco:
Jossey-Bass, Inc., 1971), p. 179.
produced that correspond to those which would be observed
if the researcher were able to manipulate the conditions and
independent variables directly. The construction of such a
model is necessarily an oversimplified representation of
social reality; that is, a limited number of variables are
linked together in such a way that the inferred causal link-
ages are assumed to operate in the real world. The value
of specifying the theory in a mathematical model lies in its
ability to inform or make explicit the empirical research
assumptions and procedures. Knoke emphasizes the un-
certainty that remains in such model building, especially in
relation to causal ordering of the variables, but because of
its explicit formulation it is possible for subsequent re-
searchers to modify the model, based upon new knowledge and/
or improved measurement. The following set of mathematical
equations specifies the theory tested in this research.
Each equation identifies those variables assumed to di-
rectly affect the value of the dependent variable:
(1) Y=C1+B Y 4+B+BiX+BX3 +BIX5+B X7+B X +R
Kenneth C. Land, "Principles of Path Analysis,"
Sociological Methodology 1969, edited by Edgar P. Borgatta
(San Francisco: Jossey-Bass, Inc., 1969), pp. 3-4; and
David R. Heise, "Problems in Path Analysis and Causal In-
ference," Sociological Methodology 1969, edited by Edgar P.
Borgatta (San Francisco: Jossey-Bass, Inc., 1969), pp. 39-44.
Land, "Formal Theory," p. 179.
David Knoke, "A Causal Model for the Political Party
Preferences of American Men," American Sociological Review
37 (December, 1972), 680.
(2) Y2=c2+BY4+B2X+B22+B2X3+B2X4+B2X6+B2X +B2X+Rb
=2 2 4 21 22 4 2 6 28 2 9 b
(4) Y4=C4+B4+B42+B43+B4X1+B X2+B4X5+B4X6+B4X7+
Equation (1) states that, holding constant the values
of all other variables in the model, perceived value-access
disjunction (Y1) is a direct function of some constant (Cl)
problem drinking (Y4), age (X1), race (X3), total family
income (X5), education (X ), marital status (X ), and a
residual term (R ) that includes the effects of all unmea-
sured factors on perceived value-access disjunction.
Equation (2) states that, holding constant the values
of all other variables in the model, general psychopathology
(Y2) is a direct function of some constant (C2), problem
drinking (Y4), age (X) sex (X2), race (X3), rural resi-
dence (X ), occupational prestige (X6), marital status (X ),
residential mobility (X9), and a residual term (Rb) that
includes the effects of all unmeasured factors on general
Equation (3) states that, holding constant the values
of all other variables in the model, anxiety (Y3) is a di-
rect function of some constant (C3), problem drinking (Y4),
age (X1), sex (X2), race (X3), rural residence (X4), total
family income (X5), residential mobility (X9), and a re-
sidual term (R ) that includes the effects of all unmeasured
factors on anxiety.
Equation (4) states that, holding constant the values
of all other variables in the model, problem drinking (Y4)
is a direct function of some constant (C4), perceived value-
access disjunction (Y1), general psychopathology (Y2),
anxiety (Y3), age (X1), sex (X2), total family income (X5),
occupational prestige (X6), education (X7), marital status
(X8), attendance at a church that formally prohibits drink-
ing alcoholic beverages (X10), and a residual term (Rd) that
includes the effects of all unmeasured factors on problem
In addition, it may be noted that a reciprocal rela-
tionship applies among the four dependent endogenous vari-
ables inasmuch as problem drinking is both cause and effect
of perceived value-access disjunction, general psychopath-
ology, and anxiety. These reciprocal relationships are
assumed to be essentially simultaneous in their feedback
one to the other. Theoretically it is possible to expect
delayed feedback from problem drinking (the major depen-
dent variable) to other variables in the model such as atten-
dance at a church that formally prohibits drinking, occupa-
tional prestige, income, and marital status. This feedback
could be examined through the introduction of lagged vari-
ables into the model. The values of these lagged variables
could be obtained through the use of longitudinal data or
retrospective self-reports. These data are not available
in the present research.
Operationalization of the Concepts7
"Problem drinking is a repetitive use of beverage
alcohol causing physical, psychological, or social harm to
the drinker or to others." 8Problem drinking is measured
by combining the scores on a six-item index. The scores
range from 0 to 20. The mean index score is 1.7 and the
standard deviation is 2.5. The overall Cronbach's Alpha is
General psychopathology is defined as general psycho-
social impairment of the individual. General psychopath-
ology is measured by scoring the eight items of an index of
psychopathology presented by Warheit et al. The scores
range from 0 to 16. The mean index score is 4.8 and the
See Appendix for a presentation of the items used in
the following measures.
Thomas F. Plaut, Alcohol Problems: A Report to the
Nation by the Cooperative Commission on the Study of Alco-
holism (New York: Oxford University Press, 1967), pp. 37-38.
Harold A. Mulford, "Drinking and Deviant Drinking,
U.S.A., 1963," Quarterly Journal of Studies on Alcohol 25
(December, 1964), 643-650; and Don Cahalan, Problem
Drinkers (San Francisco: Jossey-Bass, Inc., 1970), present
measures of problem drinking.
0C. J. Cronbach, Essentials of Psychological Testing
(Second Edition; Harper and Brothers, 1960), p. 141.
George J. Warheit, Roger A. Bell, and John J. Schwab,
Planning for Change: Needs Assessment Approaches (Washing-
ton, D.C.: The National Institute of Mental Health, 1974),
standard deviation is 3.3. The overall Cronbach's Alpha
Anxiety is defined as a high degree of tension that
interferes with the productive discharge of one's normal
role obligations. Anxiety is measured by scoring the 11
items of an anxiety function index presented by Warheit
et al. The scores range from 0 to 44. The mean index
score is 2.0, and the standard deviation is 4.3. The over-
all Cronbach's Alpha is .88.
Perceived Value-Access Disjunction
Perceived value-access disjunction is defined as the
discrepancy between the individual's perception of his cur-
rent status and his idealized status. Perceived value-
access disjunction is measured by scoring the respondent's
position on the Cantril Self-Anchoring Striving Scale.13
The scores range from 0 to 10. The mean score is 6.9 and
the standard deviation is 2.2.
Religious proscription is defined as the formal posi-
tion of the respondent's church on alcohol abstinence.
1Warheit et al., Planning for Change, p. 209.
13Hadley Cantril, The Pattern of Human Concerns
(New Brunswick, New Jersey: Rutgers University Press,1965),
Those churches with a formal position of abstinence include
Baptist, Church of Christ, Church of God, Holiness, Jehovah
Witness, Methodist, Nazarene, Pentecostal, Seventh Day
Adventist, United Church, and other conservative demonima-
tions. This variable is treated as a dummy variable. Those
who attend a church that formally proscribes alcoholic con-
sumption are assigned the value of 1, all others, 0. One
thousand four hundred eighty-six of the respondents attend
a church that formally proscribes alcohol consumption while
2,188 do not.
Residential mobility is measured by the number of times
the respondent moved in the past five years. The mean score
is 1.5 and the standard deviation is 3.1.
Residence is measured by the location of the respon-
dent's primary domicile in either a rural area or town.
This dichotomy is treated as a dummy variable; rural resi-
dents are assigned the value of 1, all others, 0.
Race is measured by the respondent's stated racial
attribute, either white or nonwhite. This dichotomy is
treated as a dummy variable; nonwhites are assigned the
value of 1, all others, 0.
Education is measured by the highest level of educa-
tional achievement reported by the respondent. Educational
achievement is classified into 12 categories: no formal
education,first grade, fourth grade or less, eighth grade
or less, tenth grade or less, high school graduate, one year
of college, two years of college, three years of college,
college graduate, some postgraduate work, and a postgraduate
degree. Missing values are assigned the median year of
school completed by black and white age cohorts reported by
the Bureau of the Census in 1970. Black respondents age
20 to 34 are assigned the value of high school graduate (12),
age 35 to 54 are assigned some high school (10), age 55 to
64 are assigned eighth grade (8), while age 65 and over are
assigned the value of fourth grade (4). White respondents
age 20 to 54 are assigned the value of high school graduate
(12), age 55 to age 64 are assigned some high school (10),
while those age 65 and over are assigned an eighth grade
education (8).14 Education levels of the respondents range
from 0 to 18 years. The median year completed is 12, the
mean, 12.3, and the standard deviation is 4.7.
Total Family Income
Total family income is measured by the total household
U.S. Department of Commerce, Bureau of the Census,
The Social and Economic Status of Negroes in the United
States, 1970 (Washington, D.C.: U.S. Government Printing
Office, 1971), p. 79.
income reported by the respondent. Missing values are as-
signed the median total family income of Southern black and
white families as reported by the Bureau of the Census in
1970. Black families are assigned the value of $5,226,
while white families are assigned the value of $9,240.15
The mean total family income is $9,453, and the standard
deviation is $7,400.
Occupational prestige is defined as the status evalua-
tion given an occupation by members of society based upon
the education, income, and other attributes associated with
it.16 This variable is operationalized by assigning the
195017 status scores constructed by Nam and Powers to the
occupations reported by the respondents. These prestige
scores are computed by
(a) arraying occupations (actually, occupation-
industry combinations) according to the median educa-
tional level of males 14 years old and over in the
experienced civilian labor force, (b) arraying
U.S. Department of Commerce, Bureau of the Census,
1Charles B. Nam and Mary G. Powers, "Changes in the
Relative Status Level of Workers in the United States, 1950-
60," Social Forces 47 (December, 1968), 158.
1Verbal communication between Charles B. Nam and
Charles E. Holzer III on November 20, 1974, informs us that
the 1970 prestige scores have not yet been completed. How-
ever, a correlation coefficient of .96 is reported between
the 1950 and 1960 scores (Nam and Powers, p. 160). It is
reasonable to assume that minimal bias is introduced by
utilizing the 1950 scores.
occupations separately according to the median income
level of the same population, (c) using the number of
persons engaged in each occupation, determining the
cumulative interval of persons in each occupation for
each of the two arrays, and (d) averaging the midpoints
of the two cumulative distributions of occupants and
dividing by the total experienced civilian labor force
to get a status score for the occupation.18
Employed males and employed single females are allocated
present occupational status scores. Married female re-
spondents are assigned the status scores of their husbands'
occupations. Student respondents are assigned their fathers'
status scores and retired, disabled, or unemployed respon-
dents are assigned the status scores of their last full-time
jobs.9 Missing values are assigned a value of 33 as sug-
gested by Nam and Powers.2 The prestige scores have a
theoretical range of 0 to 99. The actual range of scores in
this study is 2 to 99. The median score is 51 and the mean
is 53.8 with a standard deviation of 27.3.
Sex is measured by the respondent's stated sexual attri-
bute. The attribute is treated as a dummy variable. Females
8Nam and Powers, p. 159.
9This allocation procedure follows that suggested by
Kenneth J. Hodge and Marjorie F. Spencer in an unpublished
paper, "Measures of Social Stratification: A Preliminary
Report for 'Evaluating Southern Mental Health Needs and
Services,'" National Institute of Mental Health #15900,
Gainesville, Florida: Department of Psychiatry, University
of Florida, pp. 15-16.
2Nam and Powers, p. 170.
are assigned the value of 1, males, 0. There are 1,588 male
and 2,086 female respondents.
Age is measured by the respondent's stated age. The
respondents' ages range from 18 to 96. The mean age is
44.4 and the standard deviation is 18.5.
Marital status is measured by the respondent's stated
present marital status. The status is dichotomized into
married and nonmarried status and treated as a dummy vari-
able; married respondents are assigned the value of 1,
others 0. There are 2,385 married and 1,289 unmarried re-
This study focuses upon one aspect of a more extensive
social psychiatric epidemiological survey.21 The items used
to measure the concepts described above are part of a 317-
item interview schedule. This interview schedule is de-
signed to elicit information concerning:
(1) demographic data and a comprehensive social his-
tory, (2) items concerning familial and other inter-
personal relations, (3) questions concerning life
2These data were collected as one aspect of the Florida
Health study "Evaluating Southern Mental Health Needs and
Services," NIMH Grant #15900-05 and of the "Southern Health
and Family Life Studies," NIMH Contract HSM 42-73-9 (OC)
and Community Mental Health Center and Winter Haven Hospital.
satisfactions, both interpersonal and other, (4) indices
concerning religion, racial distance, anomie, percep-
tions of social change and social aspirations, (5) a
medical systems review and detailed physical symptom
data, (6) a detailed inventory of mental symptomatology
and (7) a series of items concerning attitudes
toward and utilization of health services.22
A complex combination of theoretical and statistical
processes was performed in an effort to construct items and
indices that are reliable and valid. This process was ini-
tiated by pooling a large number of items from textbooks and
research literature on social psychiatric impairment. Clini-
cians rendered an expert judgment on whether these items were
significant in diagnosing and differentiating between spe-
cific types of social psychiatric impairment. These items
were then administered to a known patient population and to
a subsample of the general population.
Subsets of items were selected on the basis of face
validity from the pool as a measure of specific social
psychiatric impairment,such as psychopathology, anxiety
and depression. Factor analysis was utilized to generate
tentative subscales. Items which appeared in more than one
subscale were eliminated from one or both and questions
which appeared to be duplicates were eliminated.
2George Warheit, John J. Schwab, Charles E. Holzer III,
and Steven Nadeau, "New Data From the South on Race, Sex,
Age, and Mental Illness" (paper presented at the Annual
Meeting of the American Sociological Association, New York,
August 28-30, 1973), p. 5.
The universe consists of the residents age 18 or over
in four central Florida counties. A sample of approximately
5 per cent was desired. Although essentially identical and
comparable, the study includes two multistage cluster sub-
samples. The sampling frame was provided by the electrical
utilities serving the counties. In each, the sampling frame
consisted of a listing of households. From each frame a
sample was selected. In subsample one, 2,315 households
were selected and in subsample two, 2,400 households were
Since a random sample was desired, the Kish technique
was used to select the respondent within each household.
A face sheet was assigned to each household. It contained
the address of the residence, interviewer's name, date,
time, and type of each attempt to secure the interview, date
and time that the interview occurred, time taken to complete
the interview, a form for listing household members,
separately for males and females in descending rank by age,
and a Kish table for selection of the respondent within the
household. A number was assigned to each adult age 18 and
over. The interviewer consulted the Kish table to determine
which adult in the household to interview.23
2Lynn Robbins, "Sampling Processes and Organization
of Field Work" (unpublished paper; Gainesville, Florida:
Department of Psychiatry, University of Florida, February,
The nonresponse rate in subsample one is 16.07 per
cent and in subsample two, 11.9 per cent. A comparison of
the major demographic variables with the 1970 census was
made and confirmed that the sample was representative.24
These data were collected between July, 1970, and
December, 1973. Person-to-person interviews were conducted
by trained interviewers. The respondents were notified of
the impending interview seven to ten days before the inter-
view. During the two-hour interview the interviewer re-
corded the respondent's answers on the preceded question-
naire. These data were then placed on IBM data cards for
data analysis. When the respondent was not at home the
interviewer made as many as three call-backs at different
times in an effort to interview the respondent.
Data Analysis Techniques
The literature review reveals a set of variables which
may explain and predict problem drinking. The major task
of the research is to construct a model based upon existing
substantive theory and research findings which makes expli-
cit the linkage of these variables in a valid, reliable
Charles E. Holzer III, "Social Status and Psycho-
logical Disorder: An Examination of Two Competing Hypothe-
ses" (unpublished master's thesis; Gainesville, Florida:
University of Florida, 1973), p. 15.
causal system. Duncan introduced path analysis into the
sociological literature as a technique which holds the prom-
ise of meeting this objective.25 This technique was devel-
oped by the geneticist Sewall Wright in 1921 and is well
integrated into the field of econometrics.2
The advantages of path analysis are interpretative and
do not add to the statistics of the regression model.
These advantages are as follows: (1) The model is an ex-
plicit, consistent expression of the underlying theory.
(2) The estimate of the parameters is a direct test of the
model. (3) The model is readily decomposed into the direct
and indirect effects of the independent variables.27
The structural equations of the path model permit
estimation of the parameters. The results of this empirical
test of the model either support the formulation or lead to
a reformulation of the model. Finally, the model may
either verify, clarify, modify, or refute the substantive
theory that originally provided the model.2
Despite the advantages of using path analysis to ac-
complish the above objective there are rather stringent
Otis Dudley Duncan, "Path Analysis: Sociological
Examples," American Journal of Sociology 72 (July, 1966),
2Kenneth C. Land, "Principals of Path Analysis,"
Sociological Methodology 1969, edited by Edgar P. Borgatta
(San Francisco: Jossey-Bass, Inc., 1969), pp. 5-7.
2Duncan, "Path Analysis," pp. 1-16.
Ibid, pp. 3-4.
assumptions, the violation of which must be minimized to
reduce bias. Nygreen lists eight basic assumptions of the
path model. These are:
1. Interval scale measurement on all variables.
2. Homoscedasticity assumption.
3. Low multicollinearity assumption.
4. Effects between variables are linear and
5. One-way causation .
6. Residuals are uncorrelated with an independent
variable directly affecting the dependent
variable upon which it acts .
[7.] High degree of measurement reliability and
[8.] "Undebatable" rankings of the chosen variables
in terms of their causal priorities--that is,
a clearly defined causal scheme.29
Blau and Duncan observe that the researcher may fre-
quently be unable to know when the tolerance level has
been exceeded or what the consequences of violating the
assumptions may be.30 Nevertheless, Johnston states that
the researcher ought to know the consequences of violating
the assumptions, how to test whether the assumptions are
adequately met, and the alternative techniques available
when the assumptions are unmet.31 The following section
briefly addresses this problem for each of the above assump-
G. T. Nygreen, "Interactive Path Analysis," The
American Sociologist 6 (February, 1971), 41.
Peter M. Blau and Otis Dudley Duncan, The American
Occupational Structure (New York: John Wiley and Sons,
Inc., 1967), p. 116.
31Johnston, p. 160.
Interval Scale Measurement of All Variables32
Labovitz, Boyle, and Lyons note that the assumption of
equal intervals on some underlying scale of X is difficult
to meet in social research. The researcher may 1) assume
that the ordered categories conform in approximation to an
underlying monotonic scale; 2) decompose the ordered cate-
gories into dummy variables; or 3) create an effects-
proportional scale.33 Boyle shows that there is little dif-
ference between the regression and path coefficient when
these alternatives are applied to identical data.34 However,
the use of dummy variables may involve some loss of data,35
and the effects-proportional scale does not necessarily
apply when a third variable is introduced into the model.36
For these reasons it seems reasonable to treat ordered
categories as interval scale data by assuming that these
Nygreen, p. 41.
33Sanford Labovitz, "The Assignment of Numbers to Rank
Order Categories," American Sociological Review 35 (June,
1970), 515-524; Richard P. Boyle, "Path Analysis and Ordinal
Data," American Journal of Sociology 75 (January, 1970), 461-
480; and Morgan Lyons, "Techniques for Using Ordinal Mea-
sures in Regression and Path Analysis," Sociological Method-
ology 1971, edited by Herbert L. Costner (San Francisco:
Jossey-Bass, Inc., 1971), pp. 147-171.
Boyle, pp. 476-479.
5Labovitz, p. 523.
3Lyons, p. 168.
ordered cateogries conform in approximation to an underlying
interval scale and to treat nominal attributes as dummy
variables in the regression model.
Homoscedasticity is defined as equal variance of the
dependent variable, Y, for any given value of the independent
variable, X. Blalock states that the regression model is
robust, tolerating moderate departures from equality of
variances.37 Bohrnstedt and Carter demonstrate that sig-
nificance tests, estimates of the intercept and regression
coefficients remain unbiased regardless of the degree of
heteroscedasticity.38 When heteroscedasticity is extreme,
the use of transformations of the original variable will tend
to equalize the variances.39
Low Multicollinearity Assumption
This assumption states that the intercorrelation among
the independent variables is not high. In the case of
highly intercorrelated independent variables it is likely
that each explains the same portion of the variation in the
3Hubert M. Blalock, Jr., Social Statistics (Second
Edition; New York: McGraw-Hill Book Co., 1972), p. 325.
George Bohrnstedt and T. Michael Carter, "Robust-
ness in Regression Analysis," Sociological Methodology
1971, edited by Herbert L. Costner (San Francisco: Jossey-
Bass, Inc., 1971), pp. 123-125.
3Blalock, p. 325.
dependent variable. Second, highly intercorrelated inde-
pendent variables cause the regression and beta coefficients
to be susceptible to sampling and measurement error. Large
samples and accurate measurement are essential when the
independent variables are highly related. Blalock notes
that in creating a new variable--the cross-product interac-
tion term--multicollinearity is an obvious result since the
new variable will be highly correlated with the two from
which it was constructed. lutaka advises that the unique
variance explained by each independent variable may be
determined by controlling the entry of each variable into a
stepwise regression model.42
Effects Between Variables Are Linear and Additive
Heise shows that when the relationship between variables
is not linear the use of mathematical transformations to
create a new variable will assist the researcher to meet
this assumption. The major problem involved is the inspec-
tion of the data and selection of the appropriate trans-
formation. In the case of interaction--the lack of ad-
ditivity--a new variable may be created by the cross products
4Blalock, pp. 456-458.
4Ibid., p. 464.
4Sugiyama lutaka, personal communication, November,
of the independent variables.43 However, as discussed
above, the problem of multicollinearity becomes pronounced
when cross products are used to create new variables.
One-way Causation (Recursive Model)
When there is reciprocal interdependence among the
variables--when the variables are both cause and effect of
each other--the estimate will be biased inasmuch as the
error term and the independent variables will be correlated.
In this case the multiple regression model is inappropriate.
In their paper on "Peer Influences on Aspirations"
Duncan et al. introduce sociologists to a technique for
analyzing the reciprocal influences which apply when two
or more dependent variables in a model are simultaneously
interdependent.44 This technique is called two-stage
least-squares and is an extension of the regression model
and path analysis. It is an appropriate technique to apply
in data analysis when the reciprocal influences have be-
come relatively crystallized such as when cross-sectional
data have been obtained.
Heise, p. 65.
4Otis Dudley Duncan, Archibald O. Haller, and Alejan-
dro Portes, "Peer Influences on Aspirations: A Reinter-
pretation," American Journal of Sociology 74 (September,
1968), 119-137; and Robert Mason and Albert N. Halter, "The
Application of a System of Simultaneous Equations to an
Innovative Diffusion Model," Social Forces 47 (December,
Residuals Are Uncorrelated with Any Independent
Variables Directly Affecting the Dependent
When measurement errors are present in the
observed values, we guarantee that the disturbance terms
will be correlated with the independent variables and with
themselves. Further, this insures that our estimates of
the path coefficients themselves will be biased."45 The
problem is that we cannot be confident that the estimates
are either attenuated or increased. Bohrnstedt and Carter
present a series of equations for correcting the path
coefficients when random measurement error is present.46
Measurement Reliability and Validity
The most serious assumption concerns measurement error.
Random measurement error may be corrected but nonrandom
error introduces a serious potential bias into the findings.
As Bohrnstedt and Carter note, sociologists have been
little concerned with the problem of measurement error,
". yet it is measurement error which produces the most
serious distortions in our regression estimates."47 Heise
and Bohrnstedt state that "sociologists almost always rely
4Bohrnstedt and Carter, p. 139.
4Ibid., pp. 139-140.
47 ., pp. 142-143
Ibid., pp. 142-143.
on fallible measuring instruments in attempting to estimate
Causal Ordering of the Model
The construction of a causal model is never a completed
exercise. Seldom is the researcher able to completely and
unequivocally defend the causal inferences made in model
construction. However, the problem of causal order among
the variables must be faced. The necessary criteria for
reasonably inferring the existence of a causal relationship
are 1) concomitant variation; 2) temporal sequence; and
3) the elimination of spurious causal factors. In other
words, X may be inferred as a cause of Y when there is a
statistically significant relationship between the variables,
when X can be demonstrated to precede Y in occurrence and
when the association between X and Y persists when the
effects of variables temporally prior to both X and Y are
controlled. In this research the focus is upon multiple
factors which combine to "cause" a particular phenomenon,
problem drinking, with a particular probability.
48David R. Heise and George W. Bohrnstedt, "Validity,
Invalidity, and Reliability," Sociological Methodology
1970, edited by Edgar P. Borgatta and George W. Bohrnstedt
(San Francisco: Jossey-Bass, Inc., 1970), pp. 104.
49Claire Selltiz, Marie Johoda, Morton Deutsch, and
Stuart Cook, Research Methods in Social Relations (Revised
Edition; New York: Holt, Rinehart, and Winston, 1959),
It is obvious that the researcher who is able to manip-
ulate the independent variables in an experimental labora-
tory is in a more secure position in making causal infer-
ences than the researcher who analyzes cross-sectional survey
data. However, this research, as most social research, is
an analysis of cross-sectional survey data. Selltiz et al.
state that the inferences of causality drawn in nonexperi-
mental studies remain tenuous and the assumption of the
time order may rest primarily upon logical considerations.50
Heise states that model building without complete informa-
tion on the causal ordering to the variables may be justi-
fied as a method of summarizing present knowledge in order
to promote future research efforts.51
The research problem of this study is to describe,
explain, and predict the conditions under which adult resi-
dents of four central Florida counties experience problem
drinking. This study is one facet of a larger social psy-
chiatric epidemiological survey. Interview questions are
selected from textbooks and research literature on social
psychiatric impairment on the basis of face validity,
factor analyzed and Crombach's Alpha calculated to enhance
the validity and reliability of the measures. Respondents
0Selltiz et al., p. 423.
5Heise, p. 66.
were selected by utilizing a multistage cluster sample tech-
nique, and trained interviewers administered an extensive
interview schedule to the respondents. The data were coded
and placed on IBM tapes for data analysis. The existing
theory and research findings pertinent to this study are
specified in a series of mathematical equations which make
explicit the theoretical assumptions. The statistical
technique utilized for estimation of the parameters in this
study is two-stage least-squares, a technique which is
appropriate when two or more of the endogenous variables in
the model are interdependent. The findings of this analysis
are presented in the next chapter.
DATA ANALYSIS AND FINDINGS
This study utilizes one aspect of a larger social
psychiatric epidemiological survey. The purpose of this
study is to describe, explain, and predict the conditions
under which adult residents of four central Florida counties
experience problem drinking as measured by a problem-drink-
ing index. Particularly, the researcher wishes to analyze
the reciprocal relationship that theoretically applies be-
tween four of the dependent variables. The respondents are
a representative sample of the adult residents of these
counties. Holzer compares the major demographic variables
of one of the counties in the sample with the 1970 U.S.
Census and confirms its representativeness.1
The purpose of this chapter is to describe the charac-
teristics of the sample and the findings of the data analy-
sis. This description provides the basis for the interpre-
tations and conclusions which follow in Chapter V.
Charles E. Holzer III, "Social Status and Psychological
Disorder: An Examination of Two Competing Hypotheses"
(unpublished master's thesis, Gainesville, Florida: Univer-
sity of Florida, 1973), p. 15.
Characteristics of the Sample
Those characteristics which may have an effect upon
rates of problem drinking as suggested in the review of the
literature are described in this section.
A majority of the respondents, 56.8 per cent, are
women, while 43.2 per cent are men. The literature review
suggests that women will experience a considerably lower
rate of problem drinking than men.
The majority of the respondents, 82.3 per cent, are
white, 17 per cent are black, 0.6 per cent are Oriental,
and 0.1 per cent report a racial identity other than white,
black, or Oriental. The literature review suggests that
nonwhites will demonstrate a higher level of problem drink-
ing than whites.
The majority of the respondents, 64.9 per cent, are
married, 13.4 per cent are single, 21.3 per cent are
widowed, separated, or divorced, and 0.4 per cent are mar-
ried under common law arrangements. According to the litera-
ture review, the 34.7 per cent who are not currently married
are more likely than married persons to experience problem
Essentially all of the respondents, 3,668, reported
their ages. The median age is 43.1 years and the mean is
44.4 years, with a standard deviation of 18.5 years. The
general distribution of those who report their ages is as
follows: 5.2 per cent are younger than age 20, 13.1 per
cent are age 20 to 24, 10 per cent are age 25 to 29, 8.1
per cent are age 30 to 34, 7.6 per cent are age 35 to 39,
7.7 per cent are age 40 to 44, 7.4 per cent are age 45 to
49, 8.2 per cent are age 50 to 54, 6.1 per cent are age 55
to 59, 7.1 per cent are age 60 to 64, and 19.8 per cent are
over the age of 64. The literature review suggests a sharp
drop-off of problem drinking after the age of 60, and 26.9
per cent of the current sample fall within that group.
Nearly all the respondents, 3,657, report whether they
live in a rural area or in town. The majority, 79.7 per
cent, live in town, while 20.3 per cent report living in
rural areas. The literature review suggests that problem
drinking increases with the size of the city in which the
respondents live. However, none of the respondents live
in a large urban area, the largest city in the area studied
having a population of less than 70,000 at the time of the
A majority of the 3,674 respondents, 60.6 per cent,
presently attend church. Most of these persons, 52 per
cent, attend church at least once a week. However, most
of the respondents, 60 per cent, do not attend a church
that formally proscribes drinking alcoholic beverages.
The largest proportion of those who attend a church
that formally proscribes drinking alcoholic beverages
(15.2 per cent of the total sample), attend the Baptist
Church. The largest proportion of those who attend a church
that does not proscribe drinking (12.4 per cent of the total
sample), attend the Catholic Church; 0.7 per cent attend a
synagogue, while 3.5 per cent did not report their church
Nearly all of the respondents, 3,652, report their level
of educational achievement. The mean level of educational
achievement is 12.3 years, with a standard deviation of 4.7
years. Less than 7 per cent of the respondents have no
formal education, 17.5 per cent have an eighth grade educa-
tion or less, 60.7 per cent have at least some college credit,
31.6 per cent of the respondents have at least graduated
from college, and 12.2 per cent have some post college work
or a post graduate degree. It should be noted that the
high level of education reflects the presence of a major
state university in one of the counties. Heavy drinking
may be expected to be higher among those of higher educa-
tional levels, while problem drinking is expected to be more
prevalent among respondents with lower educational achieve-
The mean level of total family income is $9,453, with
a standard deviation of $7,400. A fairly large proportion
of the respondents, 15.9 per cent, did not provide informa-
tion on their total family income. Of those who reported
their total family income, 16 per cent have an income of
less than $3,000, 21.4 per cent have an income of between
$3,000 and $4,999, 24.1 per cent earn between $6,000 and
$9,999, 20.9 per cent earn between $10,000 and $14,999, and
17.6 per cent have a total family income over $15,000. The
research suggests that higher rates of heavy drinking may
be expected among lower income respondents.
A level of occupational prestige is also assigned each
respondent, on the basis of present job, husband's job for
married women, father's job for students, and past job if
retired, disabled, or unemployed. The mean prestige score
is 53.8, with a standard deviation of 27.3. Of the respon-
dents, 14.1 per cent rank in the range of 0 to 19, 3-0.4 per
cent in the range 20 to 29, 8.5 per cent in the range 40 to
59, 23 per cent in the range 60 to 79, and 24 per cent rank
over 80 in their relative occupational prestige. Occupa-
tional prestige can be expected to have the same general
effect on drinking patterns as the above two socioeconomic
This study examines the impact of these three separate
measures of socioeconomic status upon problem drinking. The
product moment correlation between education and total
family income is .09; between education and occupational
prestige it is .12; and between total family income and
occupational prestige it is .31. It is assumed that each of
these measures taps a different aspect of socioeconomic
status,and consequently they are treated separately rather
than in a composite socioeconomic status score.
Psychosocial Characteristics of the Sample
Three of the variables in the model, perceived value-
access disjunction, general psychopathology, and.anxiety,
are considered to be reciprocal in their relationship with
problem drinking. These four variables are the dependent,
endogenous variables to be explained.
The mean score on the measure of perceived value-access
disjunction is 6.9, with a standard deviation of 2.3. Cantril
reports a mean score on the Cantril Self-Anchoring Striving
Scale of 6.2 for American respondents in 1963, 6.3 for whites
and 5.2 for black respondents.2 The existing theory suggests
that the greater the perceived value-access disjunction, the
greater the disposition toward problem drinking and vice versa.
The mean score on the measure of general psychopathology
is 4.8, with a standard deviation of 3.4. The existing
theory suggests that the greater the level of general psy-
chopathology,the greater the disposition toward problem drink-
ing, and vice versa.
The measure of anxiety is highly skewed toward the upper
end of the index. The mean score is 2.0, with a standard
deviation of 4.4. It should be noted in subsequent analysis
that significance tests based upon this skewed measure are
subject to error. However, this error is minimized statis-
tically by the large number in the sample and,theoretically,
by the emphasis in this research on model building rather
Hadley Cantril, The Pattern of Human Concerns (New Bruns-
wick, New Jersey: Rutgers University Press, 1965), p. 43.
than hypothesis testing. A major suggestion for further
study is the construction of a less skewed measure of anxiety.
Drinking Patterns of the Sample
The measure of problem drinking is also skewed toward the
upper end of the index. The comment above concerning the
measure of anxiety applies to this measure as well. The
majority of the respondents report a low consumption of al-
cohol and no resultant problems. The mean score on the mea-
sure is 1.7 and the standard deviation is 2.5.
The drinking patterns of the respondents may be de-
scribed in the following way. A large proportion of the re-
spondents, 40.5 per cent, state that they do not drink
alcohol. This is a higher rate of abstinence than would be
expected in light of national surveys; Cahalan et al. and
the Gallup Poll report an abstinence rate of 32 per cent
of the adult population. Of the respondents, 13.6 per cent-.
drink often or every day, 23.5 per cent seldom drink, and
22.4 per cent drink sometimes.
Of those who drink, 66.7 per cent are never intoxicated,
while 1.0 per cent are drunk every week. Of those who drink,
33.3 per cent report being drunk at some time during the
previous year, and 3.9 per cent report being drunk at least
Don Cahalan, Ira H. Cisin, and Helen M. Crossley,
American Drinking Patterns (New Haven, Connecticut: College
and University Press, 1969), p. 200; and George Gallup,
"Alcoholism Is a National Concern," Jacksonville Times-
Union and Journal (June 9, 1974), p. A-15.
several times a month. The great majority of those who
drink, 97.8 per cent, never have too much to drink for
several days at a time, and 93.4 per cent of those who drink
have no problems as a result of drinking. Of those who
drink, 90.8 per cent don't think that they drink too much for
their own good. However, 10.5 per cent of those who drink
use alcohol to help face their problems and 5.1 per cent do
so to face their problems sometimes, often, or everyday.
This description suggests that the great majority of
the respondents consume alcohol in a controlled, normative
fashion. However, a significant portion of the sample drink
excessively, drink in a utilitarian manner, and get into
trouble as a result of drinking.
The Theoretical Model
The theory and research findings presented in the litera-
ture review were specified in a series of mathematical
equations in Chapter III. This set of equations follows:
(1) Y1=C1+B 4+BX -B X3+BX5+B X7+B X+R
(2) Y2=C2+B2Y4+B2X1+B2X2+B2X3+B2X4+B2X6+B2X8+B +R
(3) Y3=C3+B Y +B 3X+B X +B X +B X +B X +B3X9+R
(4) Y4=C4+B Y1+B4Y2+B4X3+B 4X+B4X2+B4X5+B4X6+B4X7
Multiple regression was considered as a possible data
analysis technique because of its relative power to make
use of all the available data in estimating the parameters.
However, the above model shows that Y4 is theoretically con-
sidered to be both cause and effect of each of the other
endogenous variables. In such cases, the use of multiple
regression is inappropriate because it produces a biased,
inconsistent estimate of the parameters inasmuch as the ex-
planatory exogenous variable will be associated with the
residual error term. Consequently, the technique of two-
stage least-squares was selected from the field of econome-
trics for use as an appropriate method for a relatively
unbiased estimation of parameters in a set of equations
where the assumed relationship is reciprocal rather than
Identification of the Model
Prior to estimation of the parameters in the model, it
is necessary to ascertain whether or not sufficient con-
straints have been placed on the set of equations so that
the parameters can be identified. An equation may be con-
sidered just identified when the number of variables ex-
cluded from any equation is equal to one less than the num-
ber of equations in the model. An equation may be con-
sidered overidentified when variables excluded are equal to
more than one less than the number of equations. Both just
4J. Johnston, Econometric Methods (Second Edition;
New York: McGraw-Hill Book Co., 1972), p. 376.
identified and overidentified equations are identifiable
and present no problem in the next step of estimation of
parameters. The first three of these equations are over-
identified by this rule while the fourth equation is just
identified. The system may be estimated by two-stage
The Reduced-Form Model
In the first stage of two-stage least-squares, the
above theoretical model is transformed into reduced-form
equations where each endogenous variable (Y ) is expressed
as a function of all the exogenous variables (X, .
X10) in the system. Then, by using these coefficients and
the observed values of the exogenous variables, estimates of
the values of the endogenous variables are obtained. The
purpose of this first stage is to attempt to "purge" each
endogenous variable of the effects of all others with which
it is reciprocally related by estimating what the respon-
dent's score on that variable would have been if the other
variables had not had an opportunity to affect it.6 The
reduced-form model follows:
Robert Mason and Albert N. Halter, "The Application of
a System of Simultaneous Equations to an Innovative Diffusion
Model," Social Forces 47 (December, 1968), 183-195.
Melvin L. Kohn and Carmi Schooler, "Occupational
Experience and Psychological Functioning: An Assessment of
Reciprocal Effects," American Sociological Review 38
(February, 1973), 111.
(5) Yl=cl+blxl+bl2+bl3+b x+b +bx6+b x+blx8+
(+bl+b +b X3+b +b +bX +b +bXg+
(7) Y3=c3+b3X +b3x2+b3x3+b3x4+b3x5+b3x6+b x+b3x8+
3 3 1 3 2 3 3 3 4 3 5 3 6 3 7 3 8
These equations can be estimated by the multiple re-
gression technique because there is only one endogenous
variable in each equation. The estimated values obtained
from the reduced-form equations provide an efficient and
relatively unbiased estimate of the endogenous variables.
The Structural Model
The original system of equations--(1), (2), (3), and
(4)--can now be estimated using multiple regression by in-
serting the values of the estimated endogenous variables,
91, 92 9 3, and y4 and using the observed values of the
exogenous variables. The structural equation model follows:
(9) 9=Cl+bl 4+bllx+bx 3+blx5+bx 7+bx8+Ra
(10) 2=c2+b2 4+b 2x+b2x2+b2x3+b2x+b2x6+b2x8+b2x9+Rb
(11) 3=c3+b3 4+b3xl+b3x2+b3x3+b3+b3x5 +b3x+R
(12) 94=c4+b4 1+b4 2+b43+b4xl+b4x2+b4x5+b4x6+bx7+
These equations may be interpreted as an explicit formu-
lation of the existing theory. Equation (9) states that per-
ceived value-access disjunction y1 may be estimated as a
function of some constant cl, the estimated value of problem
drinking when not affected by the variables with which it is
reciprocally related, 94, age xl, race x3, total family in-
come x5, education x7, marital status x8, and a residual
error term R
Equation (10) states that general psychopathology can
be estimated as a function of some constant c2, the estimated
value of problem drinking when not affected by the variable
with which it is reciprocally related, 94, age xl, sex x2,
race x3, residence x4, occupational prestige x6, marital
status x8, residential mobility x9, and a residual error
Equation (11) states that anxiety can be estimated as
a function of some constant c3, the estimated value of
problem drinking when not affected by the variables with
which it is reciprocally related, 94, age xl, sex x2, race
x3, residence x4, total family income x5, residential
mobility x9, and a residual error term Rc
Finally, equation (12) states that problem drinking can
be estimated as a function of some constant cq, the estimated
value of perceived value-access disjunction 91, general
psychopathology 92, and anxiety 93, when these are not
affected by the influence of problem drinking, and age xl,
sex x2, total family income x5, occupational prestige x6,
education x7, marital status x8, religious proscription
xl0, and a residual error term Rd.
The Path Model
In the schematic presentation of the path model, it is
customary to use one-way arrows on a straight line leading
from each exogenous variable to the endogenous variable
dependent on it. Two-headed arrows on a curved line are
drawn between variables that may be related but which are
not analyzed in the model. These are the predetermined
exogenous variables. The path or beta coefficients are
written on the lines with one-way arrows and represent the
sign and the proportion of the standard deviation of the
endogenous variable for which a one standard deviation change
in the designated exogenous variable is directly responsible
when all other variables in the model including the resi-
dual term are controlled. A residual term (R ) is directly
linked to each endogenous variable.
Otis Dudley Duncan, "Path Analysis: Sociological
Examples," American Journal of Sociology 72 (June, 1966), 3.
Kenneth C. Land, "Principles of Path Analysis,"
Sociological Methodology 1969, edited by Edgar P. Borgatta
(San Francisco: Jossey-Bass, Inc., 1969), p. 9.
it may be helpful to think of the residual as a
dummy variable having unit variance and zero mean and
as representing all unmeasured variables which cause
variation in the endogenous variable. The residual
path coefficient, then, represents the proportion of
the standard deviation of the endogenous vari-
able that is caused by all measured variables outside
the set under consideration in the path model.9
The residual path coefficient may be computed by the formula
The simple correlation coefficients may be written on
the curved two-headed line.11 However, in this presentation
the correlation coefficients are presented in the zero-
order correlation matrix in Table 1, because the large num-
ber of variables would produce an unintelligible tangle of
The path model illustrated in Figure 1 indicates that
the variables in the model explain 15 per cent of the
variance of the scores on the problem-drinking index. As
anticipated,when all other variables in the model are con-
trolled, the most important single predictor is sex, with
males scoring higher on the problem-drinking index.
The second most important predictor is the respondent's
score on the general psychopathology index, r=.20. A break-
down of the relationship shows that, while there is a
gradual increase in the problem-drinking scores with an
Land, "Principles of Path Analysis," p. 12.
10Ibid., p. 13.
1Duncan, "Path Analysis," p. 3.
xI X x X
m L r
increase in psychopathology, the most significant impact of
psychopathology is at the upper end of the index with those
who rate high on the psychopathology index ranking high on
The third most important variable is age, r=-.16.
Compatible with other recent findings, young people are
found to rank highest in problem drinking,while there is a
gradual drop-off until age 60, after which problem-
drinking scores drop dramatically.
The fourth ranking variable is attendance at a church
that proscribes drinking. People who attend churches which
formally proscribe alcohol consumption are less likely to
experience a high degree of problem drinking.
Anxiety follows as the next most important variable,
r=.17. The pattern described above for the relationship
between genera] psychopathology and problem drinking applies
as well to anxiety and problem drinking. Those respondents
who rank high on the anxiety index are likely to rank high
on the problem-drinking index.
Total family income is the next most significant vari-
able, r=.ll. There is a gradual increase in the problem-
drinking score as the total family income increases. This
finding is at variance with the majority of existing re-
search, which indicates that problem drinking increases with
lower income levels and heavy drinking increases as income
increases. This finding may be explained by the dependence
of the present measure of problem drinking upon the quantity
and frequency of alcoholic consumption.
Marital status comes next in significance in predicting
the problem-drinking score. The unmarried person is more
likely than the married respondent to rank high on the
Perceived value-access disjunction ranks next, r=.09.
Those who demonstrate high value-access disjunction are
likely to score high on problem drinking. Most of the vari-
ation in the relationship is at both ends of the index.
Occupational prestige and education are only slightly
related to problem drinking with an r=.07 and .05 respec-
tively. When the other variables in the model are con-
trolled, neither prestige nor education is significantly
associated with problem drinking.
One of the purposes of this study is to examine the
interdependent reciprocal relationships among the four
endogenous variables. In order to estimate the parameters,
it is necessary to construct an equation for each of the
The existing theory suggests that the relationship
between perceived value-access disjunction and problem
drinking is reciprocal. Analysis of the relationship re-
veals that when other variables in the model are controlled,
a one standard deviation change in the measure of perceived
value-access disjunction causes a .04 standard deviation
change in problem drinking, with a feedback of .09 standard
deviation change in perceived value-access disjunction.
The theory also suggests that the relationship be-
tween general psychopathology and problem drinking is re-
ciprocal. Analysis of this relationship reveals that a
one standard deviation change in psychopathology causes a
.16 standard deviation change in problem drinking with a
feedback of .20 standard deviation change in psychopathology.
Anxiety and problem drinking are assumed to be recipro-
cal in their relationship as well. A one standard deviation
change in the anxiety score causes a .09 standard deviation
change in problem drinking with a feedback of .17 standard
The test of the model verifies the existing theory that,
for the most part, those integrated into the social system
and socialized to norms of social control are less likely to
be problem drinkers. Females are more likely to be social-
ized to a norm of moderation in drinking, as this research
verifies. Older persons are more likely to be responsible
for families, businesses, and community affairs, which seems
to constrain them from excessive drinking. In all proba-
bility, the major reason for low to moderate drinking in the
age cohort 60-plus is health reasons. Attendance at a
church that formally proscribes drinking is a strong con-
straint on drinking, lending support to the general theoret-
ical formulation of the study.
Race and rural residence have little impact on the
rate of problem drinking, and,while the correlation between
the three socioeconomic variables is not high, whatever
variance is explained by total family income is essentially
the same as that part of the variance explained by education
and occupational prestige inasmuch as neither is signifi-
cant when introduced into the multiple regression model. The
relationship between perceived value-access disjunction and
problem drinking is low and suggests either the presence of
other significant unmeasured variables or inadequate opera-
tionalization of perceived value-access disjunction.
The model also demonstrates that personal integration
is a constraint against immoderate drinking. Those respon-
dents with high scores on the anxiety and psychopathology
measure are also likely to rank high on the problem-drinking
index. In both cases, it is interesting to note that the
feedback from problem drinking to the personal variable is
somewhat greater than the impact of that variable on problem
drinking; in other words, problem drinking has significant
implications for the social psychological state of the
individual. Certainly the social and psychodynamics of
problem drinking are complex and highly interrelated. How-
ever, this finding suggests that "drying out" the problem
drinker will have a significant impact on some of those
psychosocial factors considered to dispose one toward
In this chapter, the characteristics of the sample are
described. It is noted that for the great majority of
respondents the use of alcohol is not problematic and
appears to be within normative bounds. Six and six-tenths
per cent of those respondents who drink report some problems
as a result of their drinking and 10.5 per cent of those who
drink report that they do so to help face their problems.
The data are analyzed using two-stage least-squares,
a technique suitable for estimation of the parameters when
two or more of the endogenous variables are interrelated.
As anticipated, the explained variance is rather low, 15
per cent. However, the most influential explanatory vari-
ables in the model are sex, general psychopathology, anxiety,
age, and religious proscription. An examination of the
reciprocal relationship between problem drinking and the
personal variables in the model reveals that problem drink-
ing has significant implications for the individual's psy-
chosocial state as well as the usually assumed influence of
psychosocial dynamics on problem drinking.