Group Title: variable effects of religiosity on deviant behavior
Title: The variable effects of religiosity on deviant behavior
CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00103066/00001
 Material Information
Title: The variable effects of religiosity on deviant behavior
Physical Description: x, 202 leaves : ill. ; 28 cm.
Language: English
Creator: Cochran, John K
 Subjects
Subjects / Keywords: Religion and sociology   ( lcsh )
Religion   ( lcsh )
Crime   ( lcsh )
Criminals   ( lcsh )
Genre: bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )
 Notes
Statement of Responsibility: by John K. Cochran.
Thesis: Thesis (Ph. D.)--University of Florida, 1987.
Bibliography: Includes bibliographical references (leaves 192-201).
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00103066
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000942125
notis - AEQ3771
oclc - 16694831

Full Text











THE VARIABLE EFFECTS OF RELIGIOSITY
ON DEVIANT BEHAVIOR


By

JOHN K. COCHRAN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY











UNIVERSITY OF FLORIDA


1987


I






























Copyright 1987

by

John K. Cochran


-C

































Dedicated to Mom and Dad, Anne, and
Kelly, and Walt


1















ACKNOWLEDGEMENTS


I wish to acknowledge the contributions made by my

committee members: Professors Ronald L. Akers (chair), E.

Wilbur Bock, Charles E. Frazier, Charles W. Thomas, and Alan

Agresti. Their efforts and patience are greatly appreciated

well beyond what can be stated here. I would also like to

express my appreciation to Marvin D. Krohn, Lonn Lanza-

Kaduce, and Marcia J. Radosevich who, under the direction of

Ronald L. Akers, were collaborators on all phases of the

research project from which my data come. In addition, I

would like to thank Dr. Ronald Feldman and his staff at The

Boys Town center for the Study of Youth Development (Boys

Town, Neb.) whose support and resources made possible the

collection of these data by Akers and his associates.

I would like to mention my appreciation for Professor

Leonard Beeghley for his friendship, sage advice, and

understanding throughout this dissertation project. I also

wish to express my gratitude to Mrs. Adele Koehler whose

vast word processing skills and knowledge of graduate school

guidelines aided me immensely. It is only by the grace of

God and Adele's talents that this project was ever

completed.








On a more personal note, I would like to especially

thank Professors Charles E. Frazier and Ronald L. Akers whom

I have had the good fortune to work with and learn from.

Their guidance, patience, professional example, and

friendship have served and will continue to serve me well.

I remain forever indebted to Ron and Chuck and pray I never

let them down.

Finally, more than anyone else, I want to thank Anne,

my wife, whose continued support, encouragement, and

understanding were the driving forces behind my efforts.














TABLE OF CONTENTS


Page

ACKNOWLEDGEMENTS..... ........................ ..... ... iv

ABSTRACT.............................................. ix

CHAPTER

ONE INTRODUCTION.......................... ....... 1

TWO HELLFIRE AND BEYOND: THEORY AND HYPOTHESES
ON THE RELATIONSHIP BETWEEN RELIGION AND
DEVIANCE..................................... 5

The Early Literature......................... 5
The Hirschi and Stark Hellfire Hypothesis.... 6
First Revision: The Middleton and Putney
Anti-Asceticism Hypothesis................. 9
Second Revision: The Jensen and Erickson
Norm Qualities Hypothesis.................. 14
Third Revision: The Stark et al. Moral
Communities Hypothesis..................... 18
Fourth Revision: The Integrated Moral
Communities-Norm Qualities Hypothesis...... 21

THREE METHODOLOGY................ ....... .......... 22

Data ..................... ................... 22
Dependent Variables.......................... 25
Religious Variables.......................... 28
Conditional Variables........................ 31
Control Variables..................... ....... 35
Method of Analysis........................... 35

FOUR HYPOTHESES AND MODELS........................ 42

The Hirschi and Stark Hellfire Hypothesis
and Model.. ................................. 42
The Middleton and Putney Anti-Asceticism
Hypothesis and Model ...................... 43
The Jensen and Erickson Norm Qualities
Hypothesis and Models...................... 44
The Stark et al. Moral Communities
Hypothesis and Models...................... 45
The Integrated Moral Communities-Norm
Qualities Hypothesis and Models............ 46







Page

FIVE TESTS OF THE HIRSCHI AND STARK HELLFIRE
HYPOTHESIS................................... 48

Religiosity and Involvement in Delinquent
Behavior.......... .......................... 49
Religiosity and Involvement in Specific
Forms of Delinquent Behavior ............... 52
Religiosity and Use of Hard Drugs............ 56
Conclusions ..... ................ ............ 59

SIX TESTS OF THE MIDDLETON AND PUTNEY ANTI-
ASCETICISM HYPOTHESIS............ ........ ... 61

Religiosity and Premarital Sex............... 61
Religiosity and Marijuana Use................ 64
Religiosity and Alcohol Use.................. 65
A Comparison of the Effects of Religiosity
on Delinquent and Anti-Ascetic Behaviors... 69
Religiosity, Personal Asceticism, and
Marijuana Use............................... 70
Religiosity, Personal Asceticism, and
Alcohol Use................................. 72
Conclusions .................................. 73

SEVEN TESTS OF THE JENSEN AND ERICKSON NORM
QUALITIES HYPOTHESIS......................... 76

Marijuana Use: Attributed Denominational
Proscriptiveness... ...................... 77
Marijuana Use: Perceived Denominational
Proscriptiveness....... .... ............. 80
Alcohol Use: Attributed Denominational
Proscriptiveness..... .................... 83
Alcohol Use: Perceived Denominational
Proscriptiveness.......................... 87
Conclusions.................................. 88

EIGHT TESTS OF THE STARK ET AL. MORAL
COMMUNITIES HYPOTHESIS....................... 92

Aggregate Religiosity and Marijuana Use...... 92
Aggregate Religiosity and Alcohol Use......... 94
Conclusions .................................... 95

NINE TESTS OF THE INTEGRATED MORAL COMMUNITIES-
NORM QUALITIES HYPOTHESIS.................... 97

Marijuana Use.................................. 97
Alcohol Use................................... 99
Conclusions ........................... .. .. 101


vii











TEN SUMMARY AND DISCUSSION.......................

Summary ......................................
The Anti-Asceticism Hypothesis Revisited:
A Brief Note......... ............ ...........
Discussion.................. ...................

APPENDICES

A TABLE OF PRIOR RESEARCH ON THE RELIGIOSITY-
DEVIANCE RELATIONSHIP........................

B TABLE OF THE VARIABLES, THEIR CODINGS, AND
DISTRIBUTIONS..... .............. ....... .... .

C RESULTS OF TESTS OF THE HIRSCHI AND STARK
HELLFIRE HYPOTHESIS..........................

D RESULTS OF TESTS OF THE MIDDLETON AND PUTNEY
ANTI-ASCETICISM HYPOTHESIS....................

E RESULTS OF TESTS OF THE JENSEN AND ERICKSON
NORM QUALITIES HYPOTHESIS.....................

F RESULTS OF TESTS OF THE STARK ET AL. MORAL
COMMUNITIES HYPOTHESIS.......................

G RESULTS OF TESTS OF THE INTEGRATED MORAL
COMMUNITIES-NORM QUALITIES HYPOTHESIS.........

H TABLE OF PREDICTED PROBABILITIES ACROSS
THEORETICAL MODELS...........................

REFERENCES....................... ................ ..... ...

BIOGRAPHICAL SKETCH....................................


viii


Page

102

105

114
118




128


133


139


153


163


174


180


191

192

202











Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


THE VARIABLE EFFECTS OF RELIGIOSITY
ON DEVIANT BEHAVIOR

by

John K. Cochran

May 1987

Chair: Dr. Ronald L. Akers
Major Department: Sociology

Since its original formulation in 1969 by Hirschi and

Stark, the Hellfire hypothesis has undergone several

significant revisions. This hypothesis asserts that the

extent of involvement in deviant behavior is inversely

related to the degree of religiosity. Early revisions of

this hypothesis have stressed the particular salience of

religiosity on behaviors in violation of ascetic norms such

as alcohol and marijuana use over those forms of deviance

that are more secularly defined such as property crimes and

personal offenses. Other revisions have called for the

inclusion of both direct and indirect religiosity effects.

The indirect effects are believed to be mediated through a

religiously based morality of personal asceticism.

Additional modifications argue for the inclusion of

interaction effects. For instance, the influence of

religiosity and personal asceticism on anti-ascetic

behaviors is said to vary across religious affiliations.

That is, these religious influences are expected to be









stronger among persons who identify with religious

organizations that proscribe sensual indulgences like

alcohol or marijuana use. Weaker effects, in turn, are

expected for members of churches or denominations that are

more tolerant of such behavior. Likewise, the effects of

religiosity and personal asceticism are expected to vary by

community type. Stronger effects are hypothesized for those

persons who reside in communities where the majority express

strong religious faith.

Each of these respecifications and modifications of the

original Hellfire hypothesis are tested with survey data or

self-reported deviance based on a sample of 3,065 male and

female adolescents in grades seven through twelve in three

midwestern states. The method of analysis employed is

logistic regression and the results are quite mixed. While

religiosity is inversely related with deviant behavior, its

influence varies widely across forms of deviance. In

addition, it appears that many of the revisions of the

original hypothesis apply mainly to the consumption of

alcohol and much less so for marijuana use. Limitations of

the study are addressed and an attempt is made to ground the

results in theory.















CHAPTER ONE

INTRODUCTION




The belief that a lack of religious commitment is a

major element in the etiology of deviant behavior is a view

held by at least a third of the general populations of both

the United States (Jensen, 1981) and Great Britain (Banks et

al., 1975). This view has also been repeatedly suggested in

print over the years (Cooley, 1927; Coogan, 1945, 1954;

Teeters and Reinmann, 1950; Hronke, 1955; Lee, 1957; Webb

and Webb, 1957; Benson, 1960; Neumeyer, 1961; Travers and

Davis, 1961; Blatt, 1967; Reckless, 1967; Hoover, 1968;

Schafer and Knudten, 1970). In addition, this belief has

long been reflected in the common practice of requiring, or

at least encouraging, regular church attendance as a part of

crime/delinquency prevention and rehabilitation programs

(Gannon, 1967; Misra, 1983).

Nevertheless, many social scientists have been

skeptical about the existence of an inverse religiosity-

deviant behavior relationship (Benson, 1960; Falk, 1961;

President's Commission, 1967; Hirschi and Stark, 1969;

Schur, 1969; Sutherland and Cressey, 1974). Indeed, many

have argued that the relationship is probably positive


-1-





-2-


(Ellis, 1910; Bonger, 1916; Lombroso, 1918; Steiner, 1924;

Reckless and Smith, 1932; von Hentig, 1948; Barnes and

Teeters, 1951; Argyle, 1959; Luden, 1964). Others skeptical

of the existence of an inverse relationship between

religiosity and deviant behavior, base their objections, in

part, on what seems to be conflicting evidence in the

literature (see Martin and Fitzpatrick, 1964; O'Dea, 1966;

Knudten and Knudten, 1971).

This controversy has been fueled in recent years by a

growing body of research suggesting that religiosity and

deviant behavior are, in fact, inversely related. Tittle

and Welch (1983) report that only ten of the sixty-five

studies they reviewed failed to report a significant

negative relationship between religiosity and some indicator

of deviant behavior. However, many contemporary social

scientists still remain skeptical and are hesitant to grant

the "religious factor" much worth. Often they base

their criticisms on the absence of any sound theoretical

explanations. Others claim that the observed relationships

are essentially coincidental or spurious.

While such criticisms can be legitimately leveled

against the early research efforts, they lose much of their

validity when applied to the more recent literature. These

works consistently utilize more sophisticated methodologies

and have produced improved theoretical specifications of an

originally simplistic model. In the following chapter, I





-3-


present a critical review of this body of literature and

highlight the key theoretical developments. I analyze these

developments as essentially progressive modifications or

respecifications of the original "Hellfire and Delinquency

hypothesis" tested by Hirschi and Stark (1969). I label the

first modification, following the lead of Middleton and

Putney (1962) and Burkett and White 1974), the "Anti-

asceticism hypothesis." The next, suggested by Jensen and

Erickson (1979), I call the "Norm Qualities hypothesis." I

follow that with the "Moral Communities hypothesis," based

on the works of Stark et al. (1980; 1982). Finally, I

suggest an integration of these modifications which I label

the "Moral Communities-Norm Qualities hypothesis." In

subsequent chapters I submit each of these hypotheses and

the models derived from them to statistical tests with

survey data on self-reported adolescent deviance (i.e.,

delinquency, substance use, and premarital sex). Thus, this

dissertation represents the first time that all of the

hypotheses which dominate the extant literature regarding

the effects of religiosity on deviant and delinquent

behavior are tested with one source of data in a single

research effort. In addition, this study addresses many of

the theoretical and methodological shortcomings of previous

research. Finally, a wider variety of deviant and

delinquent behaviors is examined in this study than in any

other work to date. As such, it is this author's opinion





-4-


that this project will make a significant contribution to

the literature and will not be just another study on the

religiosity-deviant behavior relationship.














CHAPTER TWO

HELLFIRE AND BEYOND: THEORY AND HYPOTHESES
ON THE RELATIONSHIP BETWEEN RELIGION AND DEVIANCE



The Early Literature


The early literature dealing with religiosity and

deviance (i.e., studies conducted prior to 1969) is replete

with works by reformers, priests, ministers, and others with

a vested interest in demonstrating that religion acts do

discourage deviance. In some cases they evaluated the

effectiveness of church youth centers in preventing

delinquency (Fitzpatrick, 1967) and in others they attempted

to demonstrate how church sponsored homes for delinquents

were able to instill moral values (Coogan, 1954). Many,

however, discovered a relationship between religiosity and

delinquent behavior opposite to that expected. That is,

delinquents were often found to be more religious than

nondelinquents (Schlapp and Smith, 1928; Middleton and Fay,

1941; Kvaraceus, 1944; Diaz, 1952; Falk, 1961; Gannon,

1967). With few exceptions, such studies were based on

samples of institutionalized or known delinquents (e.g.,

Wattenberg, 1950). Their results, therefore, are

questionable; for in many cases, it was in the interests of

the institutionalized child to claim a highly religious

character. In other studies, extreme groups of delinquents


-5-





-6-


and nondelinquents were selected and compared. For example,

Travers and Davis (1961) compared a sample of known

delinquents to a sample of alter boys. The use of such

extreme groups insured that these investigators would find a

negative relationship between religiosity and delinquency.

Finally, some of the studies conducted during this period

did find an appreciable inverse effect (Healy and Bronner,

1936; Glueck and Glueck, 1950; Dominic, 1954; Miller, 1965).

However, these studies are clearly dated and their findings

should be viewed with some skepticism. (For a more thorough

and critical review of these early works see Knudten and

Knudten, 1971).



The Hirschi and Stark Hellfire Hypothesis


It was not until 1969 with the publication of "Hellfire

and Delinquency" (Hirschi and Stark) that the religiosity-

deviant behavior relationship was again brought into

prominence in the sociology of deviance. Hirschi and Stark

tested the notion that "religious sanctioning systems play

an important role in ensuring and maintaining conformity to

social norms" (1969: 202). Religion was said to perform

this role (1) through its system of beliefs which

legitimizes social values, (2) through its rites and rituals

that reinforce the commitment to these values, and (3)

through its system of eternal rewards and punishment which

ensures the embodiment of these values in actual behavior

(1969: 202-203). In sum, religion both ensures conformity





-7-


through its beliefs, rites, and rituals and deters deviance

through the threat of hellfire for sinners. This "Hellfire

hypothesis" may be stated formally as

H1: The greater the degree of religiosity, the
lower the probability of delinquent behavior.

Using survey data from a sample of 4,007 students

entering public junior and senior high schools in West

Contra Costa County, California, Hirschi and Stark found

that church attendance and belief in supernatural sanctions

were not significantly related to self-reported delinquency.

They claimed that religion was not an important factor in

explaining deviant behavior. Their findings "quickly became

the accepted word on the subject, frequently cited and

widely reprinted" (Stark et al., 1982: 5). For some, these

findings resulted in the argument that the "religious factor

had finally been laid to rest" (Jensen and Erickson, 1979:

158), but for others they have served as a catalyst for the

"resurgence of study into the relationship between

religiosity and delinquency" (Anderson and Wakefield,

1983: 4).

More recent research has suggested that religiosity is

relevant to predicting deviant behavior (see Ellis, 1985 for

a lucid review of this literature or Table 1 in Appendix A

for a brief presentation). The year following Hirschi and

Stark's landmark study Rhodes and Reiss (1970) examined the

relationship between delinquency and truancy and religion

using data obtained from over 20,000 junior and senior high

school students in the Nashville and Davidson County,




-8-


Tennessee, school systems. They concluded that there is a

religious factor in delinquent and truant behavior.

Five years later, in a study of high school and college

students, Rohrbaugh and Jessor found religiosity to be

"correlated positively and significantly with other measures

of personal controls . and negatively with measures of

deviance proneness and deviant behavior" (1975: 153).

In 1977, Higgins and Albrecht used self-report data

from 1,383 Atlanta tenth graders to study the relationship

between church attendance and a variety of delinquent

behaviors. They found modest to moderately strong negative

relationships between the frequency of church attendance and

a list of seventeen delinquent behaviors. They suggested

that their findings diverge from Hirschi and Stark's because

of differences in the research settings. That is,

religiosity may be more significant in the South than in the

West.

Further support for the Hellfire hypothesis has been

found by Albrecht et al. (1977). Using data collected from

244 Mormon teenagers in three western states, these authors

found that when religious indicators were combined with

measures of peer and family relationships into a contingent

consistency model, good prediction of deviance was obtained

(R2 = .54 to .69). Jensen and Erickson (1979) and Tittle

and Welch (1983) have also found support for the Hellfire

hypothesis.





-9-


However, Elifson et al., in a study of 568 students

attending grades nine through twelve in 21 public high

schools in suburban DeKalb County (Atlanta), Georgia, found

that "within a multivariate context religion's contribution

as an independent variable was not statistically

significant" (1983: 505). Likewise, Burkett and White

(1974) in a replication of the original Hirschi and Stark

study, observed that the frequency of church attendance was

not significantly related to a six-item self-reported

delinquency index for white senior class students in three

Pacific Northwest high schools. Finally, Anderson and

Wakefield (1983) found only mixed support for the Hellfire

hypothesis with data from 170 Louisiana college students.

That is, while significant inverse relationships were

observed between religiosity and victimless crimes, no

relationship was observed for crimes against the person.

These inconsistent findings suggest that the Hellfire

hypothesis may be too broad and in need of revision.



First Revision: The Middleton and Putney
Anti-Asceticism Hypothesis


Much of the prior research has used composite indices

of delinquent behavior as the dependent variable. However,

such indices fail to tap the forms of deviance that

religion should theoretically be expected to influence in a

contemporary, secular society. Agreeing with this argument,

Middleton and Putney (1962: 142) state that





-10-


there has been a failure to distinguish two
different types of ethical standards--the ascetic
and the social. Social standards proscribe
actions which in general are harmful the social
group, and . tend to be shared as a part of a
general social ideology. The fact that religious
ideology may also proscribe these actions is
incidental. . In contrast, ascetic standards--
abstinence from sensual indulgences, gambling, and
the like--derive primarily from an ascetic
religious tradition. Within the context of
religious ideology violations of ascetic standards
may be spiritually harmful to the perpetrator.
But since these violations are usually not
directly or obviously harmful to the social group
. ascetic standards have less persuasiveness
to the secularly oriented individual. He is
therefore more likely to violate them. In short,
. difference in behavior of the religious and
the nonreligious are confined to specific areas
and are a product of difference in standards
rather than of differential upholding of
standards.

Basing their ideas on this suggestion by Middleton and

Putney, Burkett and White (1974) claimed that only when

secular values are ambiguous in their definition of an

activity as wrong can religiosity have an important

deterrent impact. They suggest that behaviors such as

premarital sex, gambling, and substance use, which are not

consistently disapproved of in the secular setting, are more

likely personal or property crimes to be affected by

religiosity. Using self-report data from 855 seniors in

three Pacific Northwest high schools, Burkett and White

found religiosity to be more predictive of adolescent

alcohol and marijuana use than of a delinquency index based

on property crimes and violent offenses. They concluded

that "blanket generalizations regarding the relationship

between religious participation and delinquency are not





-11-


warranted," and that "religious participation is . more

closely related to some kinds of delinquent behavior than to

others" (1974: 459).

McLuckie et al. (1975: 132), in a survey of drug use

based on a sample of over 30,000 students in grades seven

through twelve, found that "a full 88.9% of those who

regularly attend services, i.e., attend once a week or more,

are neither users nor quitters. Only 11.1% of regular

attenders have ever used drugs. Among those who never

attend, on the other hand, 25.8% have used drugs at one time

or another." They also found that the zero-order

correlation between attendance and drug use remained

basically unchanged at the first order when several

sociodemographic factors were controlled.

In 1975, Rohrbaugh and Jessor observed that an eight-

item composite religiosity scale correlated significantly

and positively to intolerance of premarital sex and

marijuana use in a sample of male and female college and

high school students. Religiosity was also seen to

"function as a personal control by regulating problem

behavior as theoretically expected" (1975: 146). That is,

the composite religiosity scale correlated negatively with

behavioral indicators of premarital sex and marijuana use

for both high school and college students.

Albrecht and his colleagues (Albrecht et al., 1977)

also found support for the expectations of Burkett and White

when they observed that religious variables were more





-12-


strongly related to victimless than to victim deviance.

Further, Linden and Currie (1977), Jensen and Erickson

(1979), Schlegel and Sanborn (1979), Turner and Willis

(1979), Nelson and Rooney (1982), Elifson et al. (1983),

Anderson and Wakefield (1983), Cochran and Bock (1984),

Hadaway et al. (1984), Bock et al. (1985), and Cochran et

al. (1986) have all found statistically significant,

negative relationships between religiosity and some

indicator of anti-ascetic behavior such as alcohol or drug

use. In fact, the cumulative evidence suggests that

religiosity and certain forms of anti-ascetic behavior are

so related as to constitute an empirical generalization

(Bock et al., 1985). Thus, the original Hellfire hypothesis

should be revised as follows:

H2: The greater the degree of religiosity,
the lower the probability of anti-ascetic
behavior.

Unfortunately, in nearly all of the research reviewed

to this point a crucial element which serves as a key link

between religiosity and anti-ascetic behavior has been

omitted. The omitted variable is the degree of religiously

based personal asceticism. In separate analyses of

adolescent substance use, Steven Burkett (1977; 1980)

concluded that (1) the greater the degree of involvement in

religious activities, the greater the likelihood that one

will maintain religious beliefs opposed to the use of

alcohol and marijuana, and (2) the more persons adhere to

such beliefs, the less likely they are to use these





-13-


substances. Thus, according to Burkett, "it is in the

acceptance of a belief relating to a religiously based

morality of personal asceticism, then, that the relationship

between church attendance and use is most apparent" (1977:

267).

Both Rohrbaugh and Jessor (1975) and Perkins (1985)

have also recognized the importance of including a measure

of the consequential salience of religion in one's daily

life. Rohrbaugh and Jessor (1975) asked each of their

respondents to indicate how much of an influence religion

has on the way they choose to act and spend their time each

day. Perkins (1985), however, assessed the respondent's

personal attitude toward alcohol. While the Perkins measure

is more directly relevant to a study on the effects of

religiosity on anti-ascetic deviance than the consequential

religiosity indicator used by Rohrbaugh and Jessor, neither

measure is as theoretically appropriate as that used by

Burkett (1977; 1980). Burkett asked his subjects to

indicate the degree to which they believe that alcohol or

marijuana use is a sin. In addition, neither Rohrbaugh and

Jessor (1975) nor Perkins (1985) utilized their measures as

variables linking religiosity to anti-asceticism. Only

Burkett (1977; 1980) has included an indicator of

religiously based personal asceticism in the theoretically

appropriate manner. From his work we can add two

corollaries to the anti-asceticism revision of the original

Hellfire hypothesis:





-14-


H2a: The greater the degree of religiosity,
the greater the degree of personal
asceticism.

H2b: The greater the degree of personal
asceticism, the lower the probability of
anti-ascetic behavior.



Second Revision: The Jensen and Erickson
Norm Qualities Hypothesis


With few exceptions, empirical examinations of the

religiosity-deviance relationship have been restricted to

analyses at the individual level. These individual-level

models stress that religious beliefs and activities function

to deter the individual from engaging in behaviors which

violate ascetic norms through the threat of supernatural

sanctions. Though such a model is quite reasonable,

particularly given the overwhelming consistency of

supportive research, it makes no reference to social

structure or process. A model that incorporates this

individual-level process into a more sociological or

structural framework would provide a more thorough

specification and a better theoretical explanation.

Such a model can be found in the work of Jensen and

Erickson (1979). They suggest that one of the more severe

flaws of the past research in this area is the failure to

examine the differential effects of religiosity on anti-

ascetic behaviors across religious affiliations. Few

studies have specified denominational breakdowns and nearly

all have failed to consider the possibility of an





-15-


interaction between religious affiliation and religiosity on

deviance. Such studies have ignored the possibility that

religiosity may be more relevant to understanding particular

forms of deviance such as substance use in some

denominations than in others (Jensen and Erickson, 1979:

159-160).

In a study of 3,268 Arizona high school students,

Jensen and Erickson (1979) observed that participation in a

religion (i.e., Mormon) that strongly emphasizes asceticism

and clearly proscribes drug use is particularly

consequential for such activities. That is, church

attendance is more strongly related to substance use among

Mormon youth than among Catholics or Protestants. However,

Jensen and Erickson also found that the prominence of this

interaction between religion and religiosity weakens as the

seriousness of substance increases. This is somewhat

consistent with the expectations of Burkett and White (1974)

for all denominations are probably quite similar in their

stands on serious drug use. More distinctive differences

are likely where denominational standards are more varied

(e.g., alcohol and marijuana use).

Nelson and Rooney (1982) with data from nearly 5,000

high school seniors from six states in the northeast also

found that denomination and religiosity interact such that

church attendance has a special impact on self-reported

alcohol and marijuana use within the more proscriptive

denominations. That is, members of Baptist, Methodist, and





-16-


other fundamentalist Protestant sects reported a stronger

inverse relationship between religiosity and substance use

than did Catholics or members of the prescriptive Protestant

denominations. Likewise, Nelson and Rooney witnessed a

weakening of this interaction as the seriousness of the

substance used increased.

Cochran and Bock (1984) and Bock et al. (1985) each

utilize the NORC General Social Surveys data and find that

the relationship between religiosity and alcohol use varies

across religious affiliations. In these studies,

religiosity has the greatest deterrent impact on adult

alcohol use for members of proscriptive denominations (i.e.,

those which take a strong stand against alcohol). In

addition, Schlegel and Sanborn (1979) and Hadaway et al.

(1982) have also found evidence of a significant interaction

between religious affiliation and religiosity on self-

reported substance use. However, McIntosh et al. (1981)

failed to find any significant affiliation-religiosity

interactions. In fact their findings indicate that the

effect of religiosity on adolescent drug use is quite stable

across denominations. Linden and Currie (1977) even found

evidence of a denomination-religiosity interaction contrary

to that expected by Jensen and Erickson. That is, they

observed a stronger inverse relationship between church

attendance and drug use within the prescriptive

denominations than within the proscriptive denominations.





-17-


Thus, though the evidence is somewhat mixed and far

from conclusive, the varying moral messages of religious

affiliations appear to constitute important normative

climates which impact upon the religiosity-anti-asceticism

relationship. The addition of information on religious

affiliation may permit a more precise specification of the

revised Hellfire hypothesis and may be formally stated as

follows:

H3: The greater the degree of denominational
proscriptiveness, the stronger the effects
of religiosity and personal asceticism on
anti-ascetic behavior.

Without exception, however, each of the studies that

have attempted to examine the influence of denomination on

the relationship between religiosity and anti-ascetic

behavior have categorized religious affiliations according

to the proscriptiveness attributed to them by researchers.

While such categorizations are fair and valid in only a very

general sense, they are gross in the manner in which they

treat as equal denominations with significant differences in

doctrine. Only Cochran et al. (1986) have maintained these

salient distinctions by analyzing the effect of religiosity

on alcohol use within each denomination separately.

However, even this study can be faulted for failing to

consider differences that may exist in denominations with

the same name. Perhaps a better method for distinguishing

between proscriptive and prescriptive affiliations is

available. Rather than categorizing according to

attributions of proscriptiveness or prescriptiveness,





-18-


denominations can be categorized by how they are perceived

by their members. Krohn et al. (1982) found that the

perceived normative climate of one's religious reference

group was related to both attitude towards and the use

patterns of alcohol and marijuana. Employing indicators of

both attributed and perceived denominational

proscriptiveness results in the following corollaries to the

above respecification of the revised Hellfire hypothesis:

H3a: The greater the degree of attributed
denominational proscriptivness, the stronger
the effects of religiosity and personal
asceticism on anti-ascetic behavior.

H3b: The greater the degree of perceived
denominational proscriptivenss, the stronger
the effects of religiosity and personal
asceticism on anti-ascetic behavior.



Third Revision:
The Stark et al. Moral Communities Hypothesis


A second alternative model to the traditional

individual-level model can be derived from the work of

Durkheim. Durkheim (1915) defined religion as a unified

system of beliefs and practices which serves to bind

adherents into a moral community. This definition stresses

a sociological feature of religion that is relevant to the

study of deviant behavior. That is, religion is not said to

make individuals afraid to sin, but to unite many

individuals into a moral community. It is the community of

moral adherents that controls the individual member from

violating ascetic standards.





-19-


This structural feature of religion is not an

omnipresent and constant aspect of social life; rather, it

is a variable feature. That is, societies and communities

vary in the extent to which religion binds them into moral

communities (Stark et al., 1982). In contemporary,

industrial societies like the United States, where social

norms are often highly secular, ambiguous, and unshared,

many people do not live in moral communities. People who do

not reside in moral communities, even the very religious and

devout, are more likely to violate ascetic norms. It is

only in communities where religious influences permeate the

culture and the social interactions of its members, where

religion is an integral aspect of everyday life, that

individual religiosity deters anti-ascetic behavior. Where

religion is not as pervasive, the deterrent effect of

individual religiosity is substantially reduced.

Higgins and Albrecht (1977) were the first

investigators to observe such a relationship. They

suggested that a possible reason why their findings differed

from those of Hirschi and Stark (1969) might be due to the

varying religious climates of the communities studied. That

is, the null finding of Hirschi and Stark may be due to the

fact that their sample was drawn from the nonreligious West,

whereas the Higgins and Albrecht sample was drawn from the

more religious South.

Stark et al. (1980) were able to support this argument.

They found a highly significant correlation between church





-20-


membership rates and the overall crime rates in 193 SMSAs.

In a second analysis based upon self-report data, these

authors found that there is also a very substantial negative

relationship between religiosity and delinquency in those

secondary schools in which religious students are the

majority (Stark et al., 1982). This relationship diminishes

as the proportion of religious students decreases and

vanished in the highly secularized West Coast schools.

According to Tittle and Welch (1983), because most

religions are devoted to the practice and promotion of moral

behavior, it follows that participation in religious

activities exposes one to definitions unfavorable toward

anti-ascetic behavior. The probability is greatest that

religious individuals who inhabit moral communities are more

likely to be exposed to an excess of definitions unfavorable

toward anti-ascetic behavior; hence, such individuals should

be least likely to engage in anti-ascetic behaviors.

Whether a control or a differential association

framework is preferred, the resulting empirical model is the

same. This model states that the preventive effect of

individual religiosity on ascetic norm violations is

greatest in those communities where religion is most

pervasive. This model may be stated formally as a testable

hypothesis as follows:

H4: The greater the degree of aggregate
religiosity, the stronger the effects of
religiosity and personal asceticism on
anti-ascetic behavior.





-21-


Forth Revision: The Integrated
Moral Communities-Norm Qualities Hypothesis


Though presently unexplored, the hypotheses generated

from the three revisions of the Hellfire hypothesis just

reviewed can be integrated into a theoretically more

complete model. That is, the relationship between

religiosity, personal asceticism, and anti-ascetic behavior

may be conditional upon not just the proscriptive or

prescriptive norm qualities of the various religious

affiliations, nor simply upon the degree of aggregate

religiosity; but rather, upon both denominational

proscriptiveness and aggregate religiosity. As such, the

relationship between religiosity and personal asceticism on

anti-ascetic behavior varies by both denominational norms

(either proscriptive or prescriptive), and the extent to

which the communities under examination represent moral

communities (i.e., high versus low aggregate religiosity).

The following hypothesis is generated from such an

integration:

H5: The greater the degree of denominational
proscriptiveness (attributed or perceived)
and the greater the degree of aggregate
religiosity, the stronger the effects
of religiosity and personal asceticism on
anti-ascetic behavior.













CHAPTER THREE

METHODOLOGY



Data


The data for this study come from research directed by

Ronald L. Akers and Marvin D. Krohn designed to examine

adolescent substance use and delinquent behavior and

supported by the Boys Town Center for the Study of Youth

Development (See Akers et al., 1979; Krohn and Massey, 1980;

Krohn et al., 1982). These data will be referred to here as

the "Boys Town data." These data were collected by

administering anonymous, confidential questionnaires to

3,065 male and female adolescents attending grades seven

through twelve in seven school districts in three midwestern

states. In the collection of these data, Akers and his

associates followed a two-stage sampling design. First,

schools from within each of the participating school

districts were selected; second, two to three classrooms per

grade level per school from among the required or general

enrollment classes were selected. Schools were selected so

as to be representative in terms of size and location within

their districts. In the smaller districts this meant

selecting all or almost all of the junior and senior high

schools in the district. Classrooms were also selected in


-22-






-23-


terms of school and average classroom size. Although

classrooms were sampled, each student had an approximately

equal chance of being included in the sample. This design

was structured to limit the involvement of as few schools

and school personnel as possible, to minimize intrusion into

the school routine, and to facilitate the administration of

the questionnaire to groups of respondents.

The questionnaire was administered to all students in

attendance in the selected classrooms on the day of the

survey who had obtained written parental permission.

Overall, 74% of the parental permission forms distributed

one week prior to the administration of the questionnaire

were returned. Ninety-five percent of those returning these

forms were granted permission to participate. Attrition

from this parental permission procedure combined with

absenteeism on the day of the survey reduced the proportion

of students who answered the questionnaire to 67% of the

total number enrolled in the sampled classrooms. While this

is somewhat lower than the response rate obtained in many

the national studies, it is better than others have reported

when the affirmative permission procedure is adopted (see

Radosevich et al., 1979). The design followed by Akers and

his colleagues does not technically constitute a random

sample of subjects. In fact, there was no attempt to obtain

a probability sample nor to insure that the sample was

regionally or nationally representative. Nevertheless, the

sample is sufficient to allow tests of the various






-24-


religiosity-deviant behavior hypotheses derived from the

literature review presented in Chapter Two.

Careful steps were taken to insure the rights of the

research participants. The usual university procedures were

followed regarding approval of the projects procedures for

protection of the rights of the respondents and of the

school districts. Prior to the administration of the

questionnaire, the sampled classrooms were visited. During

each visit the students were informed of the survey and were

given an envelope containing a letter explaining the purpose

and content of the study to the parents and a parental

consent form. The students were told that their

participation in the study was completely voluntary, that

their participation was not a condition for class credit or

any other school requirement, and that approval of the study

by the district and school officials in no way made their

participation mandatory. The students were assured that

their anonymity would be protected and that their responses

would be held in strictest confidence. The identification

of the participating school districts, communities, and

schools was also protected.

Finally, a small purposive subsample (n = 106) of the

respondents in five of the seven districts volunteered for

follow-up interviews two to eight weeks after the

administration of the questionnaire. These follow-up

interviews were intended to serve as reliability and partial

validity checks on the questionnaire responses. Research





-25-


has consistently shown that the self-report questionnaire

technique is reliable and valid in measuring adolescent

substance use and delinquent behavior (Whitehead and Smart,

1972; Block et al., 1974; Groves, 1974; Single et al., 1975;

Hardt and Peterson-Hardt, 1980; Zimmerman and Broder, 1980;

Hindelang et al., 1981; Akers et al., 1983). The checks on

these data by Akers and his associates (1979:640-641)

confirm this; internal consistency on interlocking items was

high (Gammas = .91 and higher). In addition, a comparison

of the responses to the substance use items on the

questionnaire with responses to the same items asked during

the follow-up interviews demonstrated a high degree of test-

retest reliability with Gammas = .89 and higher (Akers et

al., 1979:641). Given the high consistency on these

sensitive items, there is little reason to suspect lower

reliability in the responses to other items.



Dependent Variables


Indicators of Delinquent Behavior


Appropriate tests of the Hellfire hypothesis (H1)

require indicators of serious deviant behavior broad enough

to tap a wide range of norm violations. In the past,

serious deviant behavior has been operationalized by

indices of self-reported offending. Two types of

indicators of serious deviant behavior are available in

these data. The first set of measures contains indices of





-26-


delinquency. Respondents were asked to indicate the

frequency--never, once or twice, several times, or very

often--that they had engaged in behaviors on a delinquency

checklist. The items on this checklist include vandalism,

motor vehicle theft, assault, use of or threatening to use a

weapon, theft of things worth $2 to $50, and theft of things

worth more than $50. The distribution of responses on these

items indicates that fewer than 11% of the subjects are

involved in any of these delinquent behaviors more

frequently than once or twice. Because of these highly

skewed distributions, binary variables distinguishing

between those youth who admit at least some involvement (1)

and those who claim that they have never engaged in each

delinquent behavior (0) are constructed. In addition to

each of these dichotomized dependent variables, a

dichotomous delinquency index is also constructed. This

index distinguishes those youth who indicate no delinquency

involvement (0) from those admitting at least some

involvement in one or more of the items on the checklist

(1).

The other measures of serious deviant behavior

available in these Boys Town data are indicators of hard

drug usage. Respondents were asked to indicate the

frequency with which they had used stimulants, depressants,

psychedelics, and narcotics. Response categories for these

items were fixed and range along a six-point ordinal scale

from 0 = never to 5 = use nearly every day. Like the






-27-


delinquency items discussed above, the distributions on

these hard drug use items are highly skewed. Therefore,

these items are also dichotomized to distinguish nonusers

(0) from users (1). In addition, a dichotomous hard drug

use index is also constructed. This index distinguishes

abstinent youth (0) from those who report at least some use

of one or more of these substances (1).



Indicators of Anti-Ascetic Behavior


Each of the modifications of the Hellfire hypothesis

assert that the effect of religiosity and personal

asceticism on deviant behavior is strongest for violations

of ascetic standards. Indicators of such anti-ascetic

behaviors have typically included alcohol use, marijuana

use, and premarital sex. Items measuring each of these

indicators are available in these data. As with the

frequency of hard drug use items discussed above,

respondents were asked to indicate the frequency with which

they had used beer, wine, liquor, and marijuana. Though the

response categories for these items range from 0 = never to

5 = use nearly every day, all values indicating at least some

use are collapsed to form a single category. The employment

of this coding procedure permits the dichotomization of

these less serious substance use items (0 = nonuser,

1 = user). A general binary variable distinguishing

abstainers from users of alcohol is also constructed. The

final dependent variable is a dichotomous distinction





-28-


between those adolescents reporting no experience with

premarital sex (0) and those claiming at least some

involvement (1).



Religious Variables


Religiosity


Religiosity is here defined as the extent to which one

is religious, pious, or devout; the degree to which one

expresses a sincere and earnest regard for religion. Such

an orientation toward religion can serve many and diverse

roles, from providing meaning to one's life, to yielding a

sense of personal fulfillment, to securing access to social

contacts and interpersonal relationships, to offering a set

of standards against which to judge and guide one's actions

(Rohrbaugh and Jessor, 1975). The present concern is with

only one such aspect, namely, the role of religiosity as a

control against deviance. The aim is to present evidence of

the relationship of religiosity to measures of deviant

behavior.

The complexity of religiosity as a concept is reflected

in debates about its operationalization and about its single

or multi-dimensionality. Religiosity has been measured in

as many disparate ways as the frequency of attendance at

religious services to the intensity of feelings of personal

inspiration. Regarding the issue of dimensionality, early

scholars were generally in agreement that the key element in





-29-


religiosity was a personal belief in a transcendent reality

such as God (James, 1902; Durkheim, 1915). More recent

scholars, however, have urged a multi-dimensional view.

Their perspective acknowledges the centrality of religious

beliefs or ideology, but encompasses other elements as well

such as the consequential influence or salience of religion

on one's daily life, and the actual participation in

religious rituals (Glock and Stark, 1965).

While it would be preferable to construct measures of

religiosity which systematically cover each of these

dimensions, the items available in the Boys Town data are

limited to two single-item measures of religiosity.

Fortunately, the validity of such single-item measures of

religiosity has been demonstrated (Gorsuch and McFarland,

1972). No measures of religious participation nor of

religious ideology were included in the survey. The two

religiosity measures available and their corresponding

response categories and metrics are


Religiousness:
How religious a person are you?
(1) not religious at all
(2) a little religious
(3) more than a little religious
(4) quite religious

Participatory Salience:
Check the importance to you of the church group
activities in which you participate or would
like to participate:
(1) not important at all
(2) unimportant
(3) important
(4) very important







-30-


Personal Asceticism


As indicated earlier, most of the previous research on

the relationship of religiosity with anti-ascetic behavior

is limited by the presence of omitted variable bias. The

key omitted variable is the degree of religiously based

personal asceticism. This measure serves as a crucial link

between religiosity and anti-ascetic behavior; for

religiosity is not only consequential in its direct

preventive effect on ascetic deviance, but is also

influential in its effect on one's beliefs and values. That

is, the greater the degree of religiosity the greater the

likelihood that one will maintain religious beliefs opposed

to violations of ascetic norms. In turn, these religiously

based personal attitudes also serve to prevent anti-ascetic

behaviors. Thus, religiosity's effect on ascetic norm

violations is both direct and mediated through personal

ascetic beliefs and values.

Personal asceticism is operationalized in these data by

two items which asked the respondents to indicate their

attitude toward both alcohol and marijuana use. The

response categories for each of these items were collapsed

to construct dichotomous distinctions between those

holding proscriptive definitions of use (1) and those with

prescriptive definitions (0).

To insure that these indicators of personal asceticism

are a reflection of religious norm qualities, the effects of

the normative orientations of significant adults and peers






-31-


are controlled for. These control variables are measured by

four questionnaire items which asked the respondents to

report their perceptions of the norms governing alcohol and

marijuana use (either proscriptive or prescriptive) held by

those adults and peers whose opinions the respondent highly

valued. Krohn et al. (1982) have shown with these data that

respondent's attitudes toward alcohol and marijuana use are

consistent with religious norms when controlling for the

effects of peer and adult norms and therefore can constitute

religiously based beliefs and values.



Conditional Variables


Religious Norm Qualities


According to the Norm Qualities hypothesis (H3), the

effects of religiosity and personal asceticism on anti-

ascetic behavior vary with the differing normative

orientations toward alcohol and marijuana of the various

religious affiliations. That is, some denominations take a

strong stand against substance use while others tend to be

more prescriptive. These varying moral messages constitute

important normative climates which should affect the impact

of religiosity and personal asceticism on anti-ascetic

behavior. To test this hypothesis it is, of course,

necessary to operationalize these varying religious moral

messages. Two measures of denominational proscriptiveness

are available in these data. The first is the more






-32-


traditional indicator of denominational proscriptiveness

that is based on normative orientations attributed to the

religion based on knowledge of doctrine or liturgy; the

other is a measure of respondents' perceptions of the norms

of the religious group with which they identify.

Prior research has frequently grouped mainstream

American religions according to attributed moral climate

(Linden and Currie, 1977; Schlegel and Sanborn, 1979; Jensen

and Erickson, 1979; McIntosh et al., 1981; Nelson and

Rooney, 1982; Hadaway et al., 1982; Cochran and Bock, 1984;

Bock et al., 1985). These efforts have consistently

categorized Baptists, Methodists, and Fundamentalist

Protestants together because of the clear ascetic stands

against sensual indulgences that these denominations are

known to take. Other categorizations typically include the

remaining Protestant denominations (i.e., Lutherans,

Presbyterians, and Episcopalians), Catholics, Jews, and the

non-affiliated. Though these groupings are rarely combined,

they are often treated similarly because of their relatively

prescriptive normative orientations regarding alcohol use.

These prescriptive/proscriptive categorizations of

mainstream religious affiliations are consistent with their

liturgical use or non-use of alcohol during religious

ceremonies (Bock et al., 1985). That is, those affiliations

that are typically attributed as prescriptive tend also to

include alcohol (wine) in their religious practices, while







-33-


the three proscriptive denominations prohibit such

liturgical consumption.

Bainton (1945) has identified a third normative camp

that encourages even excessive and unprescribed use.

However, this permissive or "orgiastic" category has no

representatives in mainstream American religious life.

Therefore, attributed denominational proscriptiveness is

operationalized by a dichotomous variable distinguishing

between these two dominant religious normative climates.

Respondents who indicated that their religious preference

was either Baptist, Methodist, or Pentacostal are grouped

together as proscriptive denominations and are contrasted

with those whose religious preference was Catholic, Jewish,

Episcopalian, Lutheran, Presbyterian, or none (0 =

prescriptive, 1 = proscriptive). Respondents who indicated

"other" as their religious preference are coded as missing.

Perceived denominational proscriptiveness is measured

by responses to a question about what the respondents

believe their religion teaches about the use of alcohol and

marijuana. As with the other perceived norm qualities

measures, perceived denominational proscriptiveness is coded

to distinguish those who perceive their religion to be

proscriptive in its stand on alcohol and marijuana use

from those who perceive their religion to be more

prescriptive.

The effects of religiosity and personal asceticism on

anti-ascetic behaviors will be evaluated for each category






-34-


of attributed and perceived denominational proscriptivness.

The effects are expected to vary by category with the

strongest preventive effects predicted for those in the

proscriptive groupings.



Aggregate Religiosity


A test of the Moral Communities hypothesis (H4)

requires a measure which, at a minimum, distinguishes moral

communities from non-moral communities. A moral community

is said to exist where "a majority profess religious faith"

(Stark et al., 1980: 44). This definition leads to an

operationalization of the concept "moral community" by an

indicator of aggregate religiosity (see Stark et al., 1980;

1982 and Tittle and Welch, 1983). Here, aggregate

religiosity is measured by the percent per school district

expressing strong religiosity.

The seven communities (i.e., school districts) that

participated in the Boys Town project are categorized into

moral communities according to their levels of aggregate

religiosity. Those communities in which a majority (i.e.,

greater than 50%) express that they are quite religious are

grouped together under the heading "high aggregate

religiosity" which serves as an indicator of the presence of

a "moral community" in the sense used by Stark et al. Those

in which 50% or less express that they are very religious

are similarly grouped, though under the heading "low

aggregate religiosity"--an indicator of the relative






-35-


absence of a "moral community" as here defined. The effects

of religiosity and personal asceticism on anti-ascetic

behavior are expected to vary across levels of aggregate

religiosity with stronger effects predicted for those

residing in moral communities than for those isolated from

such communities.



Control Variables


Eight control variables are employed in this study:

four have already been introduced (i.e., the perceived

normative orientations of both significant adults and peers

regarding the use of alcohol and marijuana), the other four

controls are the sociodemographic variables of age, race,

gender, and SES. These variables are each coded in a

straightforward manner: age is coded continuously in years,

race and gender are each dummy coded with black and female,

respectively, as the reference categories (all nonwhite,

nonblack respondents have been coded as missing), finally

SES is a measure of father's occupation (i.e., 9 point

Census occupation category index). Table 2 in Appendix B

presents the variables to be used in this analysis, their

codings, and their distributions.



Method of Analysis


The method of analysis employed in this study is an

analytic technique suitable for binary response dependent






-36-


variables. This technique utilizes a logistic regression

procedure available in SAS (SAS, 1983). Although it is

tempting to use the more common OLS multiple regression

technique to investigate the influence of religiosity and

personal asceticism on the indicators of deviant behavior,

the presence of dichotomous dependent variables introduces

several problems into the use of the general linear model

approach commonly estimated by ordinary least squares--OLS

(Hanushek and Jackson, 1977).

In order for multiple regression to provide the best,

linear unbiased estimates (BLUE), the error terms must be

normally distributed and independent of the values of the

explanatory variables. However, in the case of dichotomous

dependent variables the error terms are, by definition, also

restricted to take on but two values. In such a situation

the error terms are clearly not normally distributed. Error

terms created by the dichotomous nature of the observations

are also likely to be correlated with the values of the

explanatory variables--clearly in violation of the above

OLS assumption. Furthermore, in these situations (where the

dependent variable, and consequently the error terms, can

take on only two values), the assumption of equal variance

of the error terms is also violated. Use of an OLS

technique, therefore, would yield inefficient estimates and

would render tests of significance meaningless.

A common solution to these problems concerning the

distribution of the error term is the use of a generalized






-37-


or weighted least squares technique. However, this

"solution" does not address the additional problem of

inappropriate functional form due to the use of a linear

solution for equations involving dichotomous dependent

variables. Fitting a linear function to such variables would

create an equation that yields probability estimates that

extend beyond the possible probability range bound between 0

and 1. These observations suggest, therefore, that a non-

linear function is more reasonable. The appropriate non-

linear function is that of the general S-curve. Such a

sigmoid function implies that a change in probability

becomes more difficult to obtain as the probability

approaches the natural limits of 0 and 1. Clearly such a

solution is theoretically more appropriate.

The use of a maximum likelihood logit technique

overcomes the difficulties presented by both the violation

of the OLS error term assumptions and the inadequate

linear functional form. Moreover, as multiple regression

minimizes the distance between the observations and the

regression line, logistic regression operates to maximize

the probability of obtaining the observed data given the set

of independent variables (see Chapter 7 of Hanushek and

Jackson, 1977 and Chapter 6 of Agresti, 1984). Although

this technique involves a different algorithm than the OLS

solution, its use and interpretation are somewhat analogous.

For instance, where in OLS regression an F statistic is

computed to test the statistical significance of the model,







-38-


a similar test is available in logistic regression through

the calculation of a chi-square statistic. Likewise, the

statistical significance of the relative effects of the

explanatory variables can also be computed. A proportional

effect for each independent variable with an interpretation

analogous to that of the unstandardized regression

coefficients in OLS can also be calculated (refer to the

partial derivative formula in Hanushek and Jackson,

1977:188-189). To do so the logistic regression

coefficients must be transformed into instantaneous rate of

change coefficients which estimate the proportional change

in the predicted probability of falling into category 1

rather than category 0 of the dependent variable for each

one-unit change in an independent variable, controlling for

the influence of all other independent variables when

evaluated as a linear effect tangent to the logistic curve

at a particular point of comparison. The proportional

effect of an independent variable, Xk, is Pk, where Pk =

bk[P(l p)], where bk is the logistic regression

coefficient, and where [p(l p)] is the variance of the

dependent variable at the instantaneous point of comparison,

p (typically the mean of the dependent variable). These

instantaneous rate of change coefficients describe the

relative effects of an independent variable as a function of

its contribution to the slope of the line tangent to the

logistic curve at this single point of comparison. Thus,

the value of the proportional change in the predicted






-39-


probability of Y = 1 will vary according to the value of p

chosen. Here, Pk will be evaluated at the mean of the

dependent variable. Choosing another value of p would yield

different estimates of the proportional effects, but the

size of the coefficients relative to one another would

remain the same. While such estimates are technically

improper, they do provide an interpretation similar to that

for OLS regression coefficients and are therefore

informative. Of course the more technically correct

interpretation of these logistic regression coefficients

(i.e., multiplicative effects on the odds) is also

provided. The former are employed solely for their clarity

of interpretation.

Finally, SAS reports an additional estimate of the

relative effects of each of the independent variables (Rk).

This statistic is a measure of association for the

relationship between each independent variable and the

dependent variable controlling for the effects of the other

independent variables in the model. As such, it is

claimed by Frank Harrell, the author of the SAS logistic

regression procedure, to be analogous to the partial r

statistic available in OLS regression.

In addition to these estimates of the relative effects

of the independent variables, the predicted probability of

falling into category 1 of the dependent variables P(Y=1)

for both those who indicate high levels of religiosity and

those with low levels of religiosity is also reported.






-40-


To compute these probabilities, one must first solve for L,

the logit coefficient. This statistic is a function of the

model intercept plus the sum of the products of each of the

logistic regression coefficients (bk) and some value of Xk

(i.e., L = a + JbkXk). The means of the control variables

and both the maximum and minimum values of the religiosity

and personal asceticism variables are used for the values

of Xk in this equation. Once L is computed, the predicted

probability of falling into category 1 of the dependent

variable (i.e., P(Y=1)) can then be easily derived by

dividing the inverse of L by one plus the inverse of L,

i.e., [eL / 1 + e )]. Employing both the minimum and maximum

values of the religiosity and personal asceticism measures

enables the comparison of the predicted probabilities for

those who are strongly religious with that for those who are

weakly religious. The reader should be cautioned, however,

that such comparisons are valid only at the mean of the

control variables.

One final comment, this SAS logistic regression

procedure employs listwise deletion in its handling of

missing data. This often resulted in a sizable reduction in

the number of observations to be analyzed. Such reductions,

of course, can have quite a substantive impact on one's

findings. It was determined that the vast majority of

deleted observations contained missing information on the

SES measure, an important contextual variable necessary as a

control in these analyses. A follow-up analysis, however,






-41-


indicated that there were no substantial differences

regarding the distributions of the dependent or religion

variables between those observations with missing data on

SES and those with complete information (tables not

reported). In addition, several models were reanalyzed with

the SES variable excluded, thereby avoiding any substantive

problems due to listwise deletion. No notable differences

were observed (tables not reported). I am fairly confident,

therefore, that these reductions in the number of

observations analyzed had little or no substantive impact on

the findings reported below.














CHAPTER FOUR

HYPOTHESES AND MODELS



The review of the literature presented in Chapter Two

traces several significant revisions and modifications of

the original Hellfire hypothesis which asserts a negative

relationship between religiosity and deviance. In this

chapter I present the hypotheses derived from that

literature review and add simple path diagram models

corresponding to each hypothesis.



The Hirschi and Stark Hellfire Hypothesis and Model



H1: The greater the degree of religiosity,
the lower the probability of delinquent behavior.


RELIGIOSITY


) DELINQUENT BEHAVIOR


Figure 4.1 The Hellfire Model.


-42-






-43-


The Middleton and Putney Anti-Asceticism Hypothesis
and Model



H2: The greater the degree of religiosity,
the lower the probability of anti-ascetic behavior.


a. The greater the degree of religiosity,
the greater the degree of personal asceticism.


b. The greater the degree of personal asceticism,
the lower the probability of anti-ascetic behavior.


RELIGIOSITY> ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


Figure 4.2 The Anti-Asceticism Model.






-44-


The Jensen and Erickson Norm Qualities Hypothesis
and Models



H3: The greater the degree of denominational
proscriptiveness, the stronger the effects of
religiosity and personal asceticism on anti-
ascetic behavior.

a. The greater the degree of attributed denominational
proscriptiveness, the stronger the effects of
religiosity and personal asceticism on anti-
ascetic behavior.

b. The greater the degree of perceived denominational
proscriptiveness, the stronger the effects of
religiosity and personal asceticism on anti-
ascetic behavior.


PRESCRIPTIVE
(Attributed or Perceived)


RELIGIOSITY )-ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


PROSCRIPTIVE
(Attributed or Perceived)


RELIGIOSITY




PERSONAL ASCETICISM


ANTI-ASCETIC BEHAVIOR


Figure 4.3 The Norm Qualities Models.






-45-


The Stark et al. Moral Communities Hypothesis
and Models



H4: The greater the degree of aggregate religiosity,
the stronger the effects of religiosity and personal
asceticism on anti-ascetic behavior.


LOW AGGREGATE RELIGIOSITY
(absence of a moral community)


RELIGIOSITY ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


HIGH AGGREGATE RELIGIOSITY
(presence of a moral community)


RELIGIOSITY vANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


Figure 4.4 The Moral Communities Models.






-46-


The Integrated Moral Communities-Norm Qualities
Hypothesis and Models



H5: The greater the degree of denominational
proscriptiveness (attributed or perceived) and
the greater the degree of aggregate religiosity,
the stronger the effects of religiosity and personal
asceticism on anti-ascetic behavior.








[A] LOW AGGREGATE RELIGIOSITY
(absence of a moral community)

PRESCRIPTIVE
(Attributed or Perceived)


RELIGIOSITY > ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM




[B] LOW AGGREGATE RELIGIOSITY
(absence of a moral community)

PROSCRIPTIVE
(Attributed or Perceived)


RELIGIOSITY >ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


Figure 4.5 The Moral Communities-Norm Qualities Models.






-47-


HIGHER AGGREGATE RELIGIOSITY

[C] HIGH AGGREGATE RELIGIOSITY
(presence of a moral community)

PRESCRIPTIVE
(Attributed or Perceived)


RELIGIOSITY >ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


HIGH AGGREGATE RELIGIOSITY
(presence of a moral community)


PROSCRIPTIVE
(Attributed or Perceived)


RELIGIOSITY ANTI-ASCETIC BEHAVIOR




PERSONAL ASCETICISM


Figure 4.5 continued.












CHAPTER FIVE

TESTS OF THE HIRSCHI AND STARK
HELLFIRE HYPOTHESIS



Table 3 in Appendix C presents the results of several

logistic regression analyses each testing the Hirschi and

Stark Hellfire hypothesis (H1). Each model reported

regresses a dichotomous indicator of serious deviant

behavior on the two measures of religiosity--religiousness

and importance of church activities--as well as the four

socio-demographic control variables (i.e., age, race,

gender, and SES). Each model contains the following

statistical information: in the top portion of the table are

(1) the logistic regression coefficients (bk) and (2) their

standard errors--se(bk), (3) the proportional changes in the

predicted probability of Y = 1 for one-unit changes in the

independent variables (Pk--also referred to as instantaneous

rate of change coefficients), (4) estimates of the relative

effects of the independent variables (Rk or partial Rs), and

(5) the intercept; in the middle of the table are (6) the -2

log likelihood for the model containing the intercept only

as well as for (7) the complete model, (8) the model chi-

square, (9) the coefficient of multiple correlation for the

model (RL see SAS, 1979:183), (10) the mean of the dependent

variable (Y), and (11) the sample size (N); and at the

bottom of the table are (12) the overall predicted


-48-






-49-


probabilities of falling into category 1 of the dependent

variable, P(Y=1), for both the strongly and the weakly

religious. Parameter estimates are arbitrarily defined to

be statistically discernible from zero when the attained

significance levels for the various statistical tests are

less than .05. While such tests of significance are

technically inappropriate due to the design of the data

collection, they will be retained for their informative

value.



Religiosity and Involvement in Delinquent Behavior


The first model reported in Table 3 (Model 1) presents

a test of the Hellfire hypothesis with the dichotomous

delinquency index. The findings indicate some support for

the Hellfire hypothesis. That is, both of the religiosity

measures are related to involvement in delinquent behavior

in the theoretically expected, negative, direction (i.e.,

bk= -.223 and -.409) and each effect parameter is

statistically discernible from zero. The logistic

regression coefficient of bk = -.223 for religiousness means

that the odds of delinquent involvement is multiplied by

exp(-.223) = .8 for every one-unit increase in

religiousness. Thus, the odds of involvement in delinquent

behavior are estimated to be .83 = .51 times as high for the

"quite religious" adolescents as for those who indicate that

they are "not religious at all." The odds of delinquent





-50-


involvement is multiplied by exp(-.409) = .66 for every one-

unit increase in participatory salience. Thus, the odds of

involvement in delinquent behavior are estimated to be .663

= .29 times as high for those adolescents for whom

participation in church activities is "very important" as

for those for whom such participation is "not important at

all." The closer the value of this ratio is to zero the

stronger the inverse effect. An odds of 1 suggests no

relationship and the closer this value is to positive

infinity, the stronger the positive relationship. The

values of the partial R coefficients indicate that the

impact of the religiousness variable is less than that of

participatory salience on self-reported involvement in

delinquent behavior (Rk= -.062 versus -.103, respectively).

The instantaneous rate of change coefficients when evaluated

at the mean of Y indicate that for each one-unit increase in

religiousness, the proportional changes in the predicted

probability of delinquency involvement decreases by four

percentage points (Pk= -.041). Thus, youth who claim to be

"quite religious" have a predicted probability of

involvement in delinquent behavior that is 12% less than

that for those who indicate that they are "not religious at

all." Likewise, the proportional change in the predicted

probability of delinquent behavior decreases by 7.5

percentage points for each one-unit increase in the

importance of church activities. For those adolescents who






-51-


feel that involvement in church activities is "very

important," the predicted proportional probability of

involvement in delinquency is almost 23 percentage points

less than that for those to whom such involvement is "not

important at all."


Perhaps the strongest evidence in support of the

Hellfire hypothesis can be found in the overall probability

of self-reported delinquency involvement, P(Y=1), predicted

for the strongly religious compared to that for the weakly

religious. Again, the reader is cautioned that these

probability estimates are valid only at the means of the

control variables. For those youth who are "quite

religious" and for whom involvement in church activities is

"very important," the predicted probability of delinquency

is only .09. However, for those who are "not religious at

all" and for whom participation in church activities is "not

important at all," the predicted probability is much

greater, P(Y=1)= .40. The difference in these predicted

probabilities is .31, with the weakly religious far more at

risk than the strongly religious. In fact the weakly

religious are about 4.4 times more likely to report some

involvement in delinquent behavior than the strongly

religious. In sum, the data lend solid support to the

Hellfire hypothesis; the likelihood of involvement in

delinquent behavior is substantially reduced for those

adolescents who are most religious.





-52-


Religiosity and Involvement in Specific Forms
of Delinquent Behavior


The next six models presented in Table 3 (Models 2

through 7) report the findings of the logistic regression

analyses testing the Hellfire hypothesis on the dichotomous

dependent variables representing specific forms of

delinquent behavior (i.e., vandalism, motor vehicle theft,

assault, weapon use, and both petty and grand theft). Each

model provides some additional, though limited, support for

the Hellfire hypothesis.

With only one exception--the effect of religiousness

on vandalism--the religiosity variables are inversely

related with each form of delinquent behavior examined.

The multiplicative effects on the odds for each one-unit

increase in religiousness vary from exp(-.193) = .82 for

petty theft to exp(-.338) = .71 for assault, excluding the

positive effect of religiousness on vandalism. Thus, the

odds of involvement in these forms of delinquency are

estimated to be between .713 = .36 and .823 = .55 times as

high for "quite religious" teens as for the "not religious

at all." Similarly, the odds of such delinquency

involvements are estimated to be between [exp(-.463)]3 = .25

and [exp(-.087)]3 = .77 times as high for those youths for

whom participation in church activities is "very important"

as for those who report that such participation is "not

important at all." Finally, the instantaneous rate of

change coefficients (Pk) reported in these models range from






-53-


a low of -.014 for the effect of religiousness on self-

reported involvement in grand theft to a high of -.116 for

participatory salience and vandalism. These values indicate

that, when evaluated at the mean of the dependent variables,

and while controlling for the effects of the other

independent variables, each one-unit change in the

religiosity indicators results in from 1.4 to 11.6

percentage point decreases in the predicted probabilities of

involvement in these delinquent behaviors. These findings

suggest that religiosity does play a role in reducing

involvement in delinquent behavior as predicted by the

Hellfire hypothesis.

Similarly, the partial R coefficients (Rk), while

statistically significant, indicate that the substantive

effects of religiosity on reducing involvement in specific

types of delinquency is weak. The strongest partial

correlation reported is that for participatory salience and

vandalism (Rk= -.132); the weakest, statistically

significant, partial effect is that of religiousness on

weapon use (Rk= -.038). Overall, participation in

delinquent activity is affected by religiosity, but this

effect varies by the type of delinquent act.

Table 4 in Appendix C presents data which indicate that

this variation in effect is systematic with the severity of

the delinquent act. The severity rankings of these six

offenses presented in Table 4 are based on the findings of

Wolfgang and Figlio (1984). According to their research






-54-


vandalism is the least serious of these six delinquent

behaviors and obtains a severity score of 1.1; petty theft

is next with a score of 2.9 followed by grand theft at 3.6,

motor vehicle theft at 4.4, assault at 7.3, and finally

weapon use--the most serious--at 9.3. While the partial

correlations for religiousness tend to remain rather steady

(i.e., partial rhos generally between -.05 and -.06), the

partial correlation coefficients for participatory salience

decrease in value as the severity of the offense increases.

This provides partial support for the arguments of Middleton

and Putney (1962) and Burkett and White (1974) who claim

that the impact of religiosity on reducing involvement in

deviant behavior is strongest for low consensus or

victimless crimes. Recall that this modification of the

Hellfire hypothesis is referred to as the Anti-asceticism

hypothesis (H2).

This general support for the Hellfire hypothesis with

variations by type lending some credence to the Anti-

asceticism hypothesis is also seen in Table 5 of Appendix C.

The overall predicted probabilities of delinquency

involvement, P(Y=1), are substantially less for the strongly

religious than for the weakly religious in each model. The

predicted probability of involvement in vandalism for the

strongly religious (i.e., those adolescents who report that

they are "quite religious" and for whom involvement in

church activities is "very important") is .31. While the

predicted probability of involvement in vandalism for the






-55-


weakly religious (i.e., those adolescents who are "not

religious at all" and for whom involvement in church

activities is "not important at all") is .63. The

difference between these predicted probabilities is .32 with

the weakly religious twice as likely to report involvement

in vandalism as the strongly religious.

Involvement in motor vehicle theft is unlikely for both

the strongly and weakly religious with predicted

probabilities of .04 and .15 respectively. However, these

values indicate that the weakly religious are about 3.7

times more likely to engage in motor vehicle theft than the

strongly religious. The weakly religious are 2.9 times more

likely to engage in assaults than are the strongly religious,

though such behavior is unlikely for both groups (i.e.,

P(Y=1) = .23 and .08, respectively). Similar probabilities

are also predicted for involvement in weapon use (P(Y=1)

= .19 and .07, respectively) and grand theft (P(Y=1)= .10

and .02, respectively). For petty theft, the probabilities

of involvement are greater for each group (P(Y=1)= .60

and .22, respectively). However, weakly religious

adolescents are still about 2.7 times more likely to steal

things of minor value that are their religiously committed

peers.

Table 5 arrays these overall predicted probabilities

according to the severity of each offense. The predicted

probabilities are greatest for the two least serious

offenses--vandalism and petty theft. For the remaining






-56-


four offenses the predicted probabilities are quite low and

tend to increase only very slightly as severity increases.

Note, also, the decreasing values of the risk factors (i.e.,

how many times more likely the weakly religious are to

report involvement in delinquency than the strongly

religious) as the severity of the offense increases from

grand theft to weapon use. These values suggest, to a

marginal degree, that the effects of religiosity may wane as

offense seriousness increases.

Thus, as with the findings presented in Table 4, these

results are both supportive of the Hellfire hypothesis and

are also somewhat consistent with the claims of Middleton

and Putney (1962) and Burkett and White (1974). I will

return to this point latter (See Chapter Six). However,

before proceeding to tests of the Anti-asceticism

hypothesis, additional models testing the Hellfire

hypothesis with indicators of hard drug use will be

examined.



Religiosity and Use of Hard Drugs


The findings for hard drug use parallel those for other

delinquent behaviors, but the magnitude of the relationships

tend to be weaker. Both of the religiosity indicators in

Model 8 in Table 3 are inversely related to the index of

overall hard drug use (bk= -.379 for religiousness and -.617

for participatory salience). These logistic regression

coefficients mean that the odds of hard drug usage are






-57-


multiplied by exp(-.379) = .68 for every one-unit increase

in religiousness and by exp(-.617) = .54 for every one-unit

increase in participatory salience. Thus, the odds of hard

drug use are estimated to be .683 = .32 times as high for

the "quite religious" as for the "not religious at all"

and .543 = .16 times as high for those adolescents for whom

participation in church activities is "very important" as

for those to whom such participation is "not important at

all." The partial R for religiousness, however, is only

-.098 and the proportional change in the predicted

probability for each one-unit change in religiousness is

-.039. Thus, students who consider themselves to be "quite

religious" are 11.7 percentage points less likely to be hard

drug users than those who report that they are "not

religious at all." For participatory salience the partial R

is -.148 and the predicted proportional change in

probability is -.064; each one-unit increase in the

importance of church activities results in a decrease of 6.4

percentage points in the probability of using hard drugs.

Adolescents for whom participation in church activities is

"very important" are 19.2% less likely to report experience

with hard drugs use than those for whom such participation

is "not important at all."

The overall predicted probability of drug use for the

strongly religious (those reporting maximum values of both

religiousness and participatory salience) is .02 when

evaluated at the means of the control varliables; for the






-58-


weakly religious (those reporting minimum values) the

overall predicted probability is much greater at .28. Thus,

even though the likelihood of use is slim, the risk for the

weakly religious is still 14 times greater than that for the

strongly religious.

The final four models reported in Table 3 (Models 9

through 12) present the results of logistic regression

analyses for each of the types of hard drugs that compose

the dichotomous hard drug use index discussed above--

stimulants, depressants and tranquilizers, psychedelics, and

narcotics and opiates. The results of these analyses mirror

those just presented above in Model 8. In all four of these

models, the two religiosity indicators are inversely related

with the dependent variables. However, the strength of

these relationships is generally weak. That is, the partial

Rs range in value from -.068 to -.146. For stimulants,

depressants, and psychedelics, the impact of participatory

salience is stronger than the impact of religiousness. For

narcotics, religiousness has the stronger effect.

In sum, these data provide additional, but limited,

support for the Hellfire hypothesis. While the independent

effects of the two religiosity indicators are quite weak,

and the variance accounted for is small, the combined impact

of these religiosity measures is more substantial. That is,

weakly religious adolescents are considerably more likely to

report use of hard drugs than are the strongly religious.

The predicted probability of hard drug use overall and for






-59-


each substance is low even for the weakly religious, but

they are far more likely to use hard drugs than the

strongly religious. In fact, the chance that an adolescent

who is strongly religious is using any of these drugs is

essentially zero.



Conclusions


In conclusion, each of the twelve models presented in

Table 3 provide some support for Hirschi and Stark's

Hellfire hypothesis. From these models, the following

conclusions may be drawn:

1) The two indicators of religiosity are nearly always
inversely related with the indicators of serious
deviant behavior. As such, they support the
Hellfire hypothesis.

2) However, the strengths of these relationships are
quite weak, thus limiting the support for the
hypothesis.

3) The relative effect of participatory salience
(i.e., the importance of involvement in church
activities) is somewhat more powerful than the
effect of religiousness in predicting involvement
in delinquency per se, vandalism, petty theft,
grand theft, the use of hard drugs in general, and
the use of stimulants, depressants, and
psychedelics.

4) On the other hand, religiousness is a stronger
predictor of involvement in motor vehicle theft,
assault, weapon use, and the use of narcotics.

5) The overall predicted probabilities of involvement
in both general and offense specific delinquency,
and the use of hard drugs is substantially less for
the strongly religious (i.e., those adolescents who
claim to be "quite religious" and for whom
participation in church activities is "very
important") than for the weakly religious (i.e.,






-60-


those who report minimum values of religiousness
and participatory salience). These findings
provide the strongest support for the Hellfire
hypothesis.

6) Values of the partial correlation coefficients for
participatory salience tend to decrease in strength
as offense severity increases. Likewise, the risk
factors also tend to decrease as offense severity
increases. Both of these trends suggest that the
effects of religiosity may wane as offense
severity, and therefore, societal consensus,
increase. Such findings are consistent with the
arguments of Middleton and Putney (1962) and
Burkett and White (1974) and lend some support to
the Anti-asceticism hypothesis, a modification of
the original Hellfire hypothesis. It to tests of
this modified hypothesis to which I now turn.













CHAPTER SIX

TESTS OF THE MIDDLETON AND PUTNEY
ANTI-ASCETICISM HYPOTHESIS



The original Hellfire hypothesis of Hirschi and Stark

has been modified by arguing that it applies to a more

restricted range of deviant behavior. That is, religiosity

is more strongly related to "immorality," "vice," and

"victimless" behaviors such as premarital sex and the use of

marijuana and alcohol, which are not as consistently

disapproved of in the secular setting, than to personal or

property crimes (Burkett and White, 1974; Albrecht et al.,

1977; Anderson and Wakefield, 1983; Elifson et al., 1983).

As noted earlier, the norms against offenses of morality

have been called ascetic norms and this modification has

been referred to as the Anti-asceticism hypothesis (H2).

The statistical information reported in Tables 6 through 9

of Appendix D is similar to that presented in the previous

chapter.



Religiosity and Premarital Sex


Model 1 of Table 6 presents the results of a logistic

regression analysis testing the Anti-asceticism hypothesis

with a dichotomous indicator of self-reported involvement in

premarital sexual intercourse. These findings provide some


-61-






-62-


support for the hypothesis in that religiosity is inversely

related to premarital sex. That is, both of the religiosity

variables are inversely related to involvement in premarital

sex (bk= -.329 for religiousness and -.169 for participatory

salience).

These coefficients suggest that the odds of involvement

in premarital sexual intercourse are multiplied by

exp(-.329) = .72 for every one-unit increase in

religiousness; therefore, the odds of premarital sex are

estimated to be .723 = .37 times as high for the "quite

religious" as for the "not religious at all." For each one-

unit increase in participatory salience, the odds of

premarital sexual involvement are multiplied by exp(-.169)

= .84. Thus, for those adolescents for whom involvement in

church activities is "very important" the odds of premarital

sex are .843 = .60 as high as they are for those youth for

whom involvement in church activities is "not important at

all."

For each one-unit increase in religiousness, the

proportional change in the predicted probability of

involvement in premarital sex decreases by 7.3 percentage

points (Pk= -.073). Thus, for adolescents who report that

they are "quite religious," the proportional probability of

involvement in premarital sexual intercourse is about 22

percentage points less than that for youth who are "not

religious at all." Likewise, the proportional change in the

predicted probability of premarital sex for each one-unit





-63-


increase in participatory salience is -.038. This figure

converts into an 11.4 percentage point difference in the

likelihood of premarital sex between those youth for whom

participation in church activities is "very important" and

those for whom such participation is "not important at

all."

The overall predicted probability of involvement in

premarital sexual intercourse for the strongly religious is

only .18, while for the weakly religious, the predicted

probability is .50. The difference between these predicted

probabilities is .32 with the weakly religious almost three

times (2.8) more likely to report involvement in premarital

sex.

Though these findings show that the effects of

religiosity on premarital sexual intercourse are both

statistically discernible from zero and are in the

hypothesized direction, the relative strength of these

effects is quite weak. Indeed, the partial correlations

(Rk) between the two religiosity measures and the

dichotomous premarital sex index are only -.098 and -.042,

respectively, for religiousness and participatory salience.

Thus, while the data suggest that religiosity does serve

to reduce the likelihood of involvement in premarital

sexual intercourse among adolescents, its effect is quite

weak.






-64-


Religiosity and Marijuana Use


Model 2 in Table 6 of Appendix D reports the findings

of the logistic regression analysis testing the Anti-

asceticism hypothesis on adolescent marijuana use. Again

the support for the hypothesis is mixed; although increases

in religiosity are significantly related to decreases in the

likelihood of use, the relative effects of the religiosity

indicators are weak.

Each of the religiosity measures is inversely related

with the self-reported marijuana use (bk= -.331 for

religiousness and -.458 for participatory salience). Thus,

the odds of marijuana use are multiplied by exp(-.331) = .72

and exp(-.458) = .63 for every one-unit increase in

religiousity and participatory salience, respectively. As

such the odds of marijuana use are .723 = .37 times as high

for "quite religious" adolescents as for those who are "not

religious at all." Likewise, for youth who report maximum

levels of participatory salience, the odds of marijuana use

are .633 = .25 times as high as for those reporting minimum

levels.

When these logistic regression coefficients are

transformed into instantaneous rate of coefficients

evaluated at the mean of the dependent variable, the

resulting proportional changes in the predicted probability

of marijuana use indicate that one-unit increases in the

religiosity variables result in 8.1 and 11.2 percentage

point decreases in the likelihood of use. That is, each





-65-


one-unit increase in religiousness leads to 8.1 and 11.2

percentage point decreases in the predicted probability of

marijuana use among the youth surveyed. Thus, adolescents

who indicate that they are "quite religious" are 24.3

percentage points less likely to use marijuana than youth

who report being "not religious at all."

One-unit increases in the importance of participation

in church activities is related to 11.2 percentage point

decreases in the predicted probability of use. For those

youth for whom such participation is "very important," the

predicted proportional probability of self-reported

marijuana use is 33.6 percentage points less than that for

whom participation in church activities is "not important at

all." The cumulative effect of these religiosity indicators

is such that the weakly religious are four times more likely

to report marijuana use than are strongly religious

adolescents. The overall predicted probabilities of use,

P(Y=1), are .69 and .17, respectively, for the weakly and

strongly religious. In sum, increases in religiosity are

significantly related to decreases in the likelihood of

marijuana use, providing modest support for the Anti-

asceticism hypothesis.



Religiosity and Alcohol Use


Logistic regression analyses testing the Anti-

asceticism hypothesis with measures of self-reported alcohol

use are presented in Models 3 through 6 of Table 6. The






-66-


first of these models examines the effects of the two

religiosity indicators on the use of alcohol in general;

Models 4, 5, and 6 test these effects on the use of beer,

wine, and liquor, respectively. The results reported in

these models yield mixed support for the Anti-asceticism

hypothesis.

The effect of the religiousness variable is

statistically discernible from zero only in Model 6, the

use of liquor. The strength of this effect, however, is

quite weak (Rk= -.049). The partial correlation coefficients

for the religiousness variable indicate that it has

virtually no effect on the general use of alcohol (Rk=

-.000), and essentially no effect on the use of either beer

or wine (Rk= -.024 and .000, respectively). Participatory

salience, on the other hand, attains a weak, inverse effect

on each of the indicators of alcohol use with the exception

of wine (Rk =-.08 to -.14). The null finding with regard

to the use of wine may be due to the fact that wine is

frequently consumed during religious ceremonies such as

communion. The liturgical use of wine may result in higher

than expected rates of use among the strongly religious.

However, the multiplicative effect on the odds of

alcohol consumption for the participatory salience measure

indicates that its influence is quite substantial. The odds

of alcohol use, per se, are estimated to be [exp(-.541)]3

= .20 times as high for youths reporting maximum

participatory salience as for youth placing minimum






-67-


importance on participation in church activities. Similar

odds are estimated for the use of beer, [exp(-.517)]3 = .21,

and the use of liquor, [exp(-.403)]3 = .30.

Likewise, the instantaneous rate of change coefficients

(Pk) for participatory salience also suggests that its
influence on alcohol use may be greater than these partial

correlation statistics indicate. For instance, each one-

unit increase in participatory salience is related to an 8.1

percentage point decrease in the predicted probability of

general alcohol consumption. This translates into a

difference of -24.3 percentage points in the probability of

use predicted for those adolescents for whom participation

in church activities is "very important" compared to those

for whom such activities are "not important at all."

Similar proportional effects are observed between

participatory salience and both the use of beer and liquor

(Pk= -.078 and -.085, respectively). The proportional
effect of participatory salience on the use of wine is half

of that of its effects on the other alcoholic beverage types

(Pk= -.047).

The most telling evidence in support of the Anti-

asceticism hypothesis is shown not in the separate,

independent effects of the two religiosity variables, but

rather in their combined or cumulative effect. In each of

the four models examined, strongly religious adolescents

have an overall predicted probability of use that is

substantially less that for the weakly religious. The






-68-


probability of alcohol use among weakly religious youth is

quite high--between .84 and .93. For the strongly

religious, these probabilities are much lower, ranging

from .51 for the use of liquor to .74 for the use of wine.

The differences in these predicted probabilities of use

between the strongly and the weakly religious range

between .10 for wine and .35 for liquor. With the

exception of wine, which may be used liturgically, these

differences indicate that the weakly religious are at

least 20% more likely to use alcohol than the strongly

religious. These comparisons are, of course, evaluated at

the means of the control variables.

In sum, the inverse effects of the religiosity

indicators on anti-ascetic behavior (i.e., premarital sexual

intercourse and the use of marijuana and alcohol) are

notable and their cumulative effect is such that the

strongly religious adolescents are substantially less likely

to engage in such behaviors than are youth who are weakly

religious. While these findings provide support for the

Anti-asceticism hypothesis, they are equally supportive of

the original Hellfire hypothesis. That is, religiosity has

been shown to be inversely related to a large variety of

deviant behaviors including violations of both secular and

ascetic norms. Only by comparing the results presented here

with those from the analyses testing the Hellfire hypothesis

can a true test of the arguments of Middleton and Putney be

made. If the claims of those espousing the Anti-asceticism






-69-


argument are true, than the effects of religiosity should be

stronger for violations of ascetic norms than the more

serious and consistently disapproved delinquent behaviors.

It is to such a comparison that I now turn.



A Comparison of the Effects of Religiosity on
Delinquent and Anti-Ascetic Behaviors


Table 7 of Appendix D presents the partial R statistics

for the effects of both religiousness and participatory

salience on delinquent and anti-ascetic deviant behaviors.

The Anti-asceticism hypothesis that the effect of

religiosity is greater for violations of ascetic norms than

for violations of more widely held norms is not

supported by these data. The relative effects of the

religiosity variables are very stable across both anti-

ascetic behaviors and other forms of adolescent deviance.

In fact, there is some indication that religiosity is a less

effective restraint on anti-ascetic behaviors than on hard

drug use and delinquency. That is, the religiousness

variable has no effect (Rk= -.000) on the use of alcohol,

and the weakest partial correlation coefficient for

participatory salience is with premarital sexual

intercourse (Rk= -.042); both are indicators of anti-ascetic

behavior.

Additional evidence counter to the Anti-asceticism

hypothesis can be found in a comparison of the risk

factors reported in Table 8 of Appendix D. Larger risk






-70-


factor values are associated with the delinquent behavior

measures than with the indicators of anti-ascetic behavior.

These values suggest that the weakly religious are only

between 1.3 and 4.0 times more likely than the strongly

religious to engage in behaviors that violate ascetic norms,

but are 4.4 to 14.0 times more likely to engage in behaviors

that are more consistently disapproved of. As such, the

greatest differences between the strongly and the weakly

religious occur in behaviors where normative ambiguity is

the least and social disapproval is the greatest. These

findings are, of course, quite contrary to what is predicted

by the Anti-asceticism hypothesis. Thus, while some

evidence from Chapter Five hinted at support for the

arguments of Middleton and Putney (1962) and Burkett and

White (1974), the results of these analyses reveal no

support for the Anti-asceticism hypothesis beyond that which

is also equally supportive of the original Hellfire

hypothesis of Hirschi and Stark. Higher religiosity acts as

an insulator against all forms of adolescent deviance; for

none is it a highly effective insulator, but it works

equally in predicting delinquent acts and hard drug use as

well as the anti-ascetic offenses of morality.



Religiosity, Personal Asceticism, and Marijuana Use


Model 1 of Table 9 in Appendix D presents the findings

of analyses testing Burkett's revision of the Anti-

asceticism hypothesis (H2a and H2b) with data on adolescent






-71-


marijuana use. According to Burkett (1977; 1980),

religiosity has both a direct inverse effect and an indirect

inverse effect on anti-ascetic behavior. The indirect

effect is said to be mediated through a religiously

based morality of personal asceticism. The results of the

analyses examined here fail to support Burkett's claims.

First, religiosity is not related to personal

asceticism. The effects of both of the religiosity measures

on personal asceticism are essentially zero (Rk= -.023 for

religiousness and .000 for participatory salience), second,

personal asceticism is not related to marijuana use (Rk =

-.020). In fact, the only notable effect observed is the

direct inverse effect of participatory salience on self-

reported marijuana use (bk= -.461). Third, estimates of the

direct effect of religiousness and participatory salience

remain essentially unchanged with the addition of these

measures to the model. That is, the logistic regression

coefficients for religiousness and participatory salience

are essentially identical in both models (i.e., bk = -.331

versus -.334 for religiousness and -.458 versus -.461 for

participatory salience) indicating no change in their direct

influence on the odds with the inclusion of the personal

asceticism measure and its controls. Likewise, the partial

R statistic (Rk) for the relative influence of participatory

salience on marijuana use is -.130 in both the model with

and the model without these extra variables. The relative

effect of religiousness increases only from -.097 to -.098.






-72-


The instantaneous rate of change coefficients for the two

religiosity variables also remain unchanged by the inclusion

of these additional measures (Pk= -.081 for religiousness

and -.112 for participatory salience in each model).

Finally, the overall predicted probability of marijuana use

among strongly religious adolescents (now defined as those

adolescents who indicate maximum values on both of the

religiosity indicators as well as on the personal asceticism

variable) decreases by only .02 (from .17 to .15). The

probability of use predicted for the weakly religious (also

redefined as those adolescents who report minimum values of

religiosity and personal asceticism) increases by only .03

(from .69 to .72). The conclusion generated from each of

these findings is that Burkett's claim for the inclusion of

a measure of religiously based personal asceticism into the

Anti-asceticism model receives no support from these data on

adolescent marijuana use.



Religiosity, Personal Asceticism, and Alcohol Use


The same conclusion regarding Burkett's revision of the

Anti-asceticism hypothesis is reached when the data on

adolescent alcohol use are examined (See Model 2 in Table

9). Neither measure of religiosity is related to personal

asceticism (Rk = -.000 for both). However, personal

asceticism does show a notable inverse effect on alcohol use

(bk= -.404). The direct effects of the religiosity






-73-


indicators, on the other hand, remain unchanged (Rk= -.000

for religiousness in both models, and -.142 versus -.143 for

participatory salience). Likewise, the instantaneous rate

of change coefficients (Pk) remain stable across both

models. Finally the predicted probability of use among the

weakly religious is .93 in both models, while the

probability of use predicted for the strongly religious

decreases from .69 to .60. This decrease of .09 in the

predicted probability among the strongly religious provides

some support for Burkett's claims; however, the cumulative

evidence presented here fails to support his arguments.



Conclusions


The tables discussed in this chapter have served as

tests of both the Anti-asceticism hypothesis and Burkett's

revision of this hypothesis (1977; 1980). The results of

these tests support the Anti-asceticism hypothesis but give

little comfort to Burkett's revision. The data fit the

original Hellfire hypothesis equally as well as the Anti-

asceticism hypothesis. Some of the more specific

conclusions drawn from these analyses are

1) With the exception of the effect of religiousness
on the general use of alcohol and the use of beer
and wine specifically, both religiosity variables
are inversely related with the indicators of anti-
ascetic behavior. These findings support the
Anti-asceticism hypothesis.

2) However, all of the religiousness effects are weak
(i.e., Rk < -.100). The strength of the relative
effects of participatory salience on premarital






-74-


sexual intercourse and on the use of wine are also
weak. More moderate effects (i.e., -.100 < Rk <
-.250) of participatory salience are observed for
the remaining measures of anti-ascetic behavior.
The weak or modest nature of these religiosity
effects serves to limit the support provided to
the Anti-asceticism hypothesis.

3) The overall predicted probability estimates,
P(Y=1), indicate sufficiently large differences
(i.e., greater than .20) between the weakly
religious and the strongly religious in terms of
involvement in premarital sexual intercourse and
the use of marijuana, alcohol in general, and
wine. Smaller differences between predicted
probabilities are observed for the use of beer
and liquor. These findings also provide limited
support for the Anti-asceticism hypothesis.

4) The above conclusions provide mixed support for
the Anti-asceticism hypothesis. They also provide
equally limited support to the original Hellfire
hypothesis. However, none of these findings
provide a true test of the Anti-asceticism
hypothesis. Only by comparing the results
reported in this chapter with those in Chapter
Five can an appropriate test of the Anti-
asceticism hypothesis be made. The strengths of
the relative effects of the religiosity
indicators are very stable over all forms of
deviant behavior analyzed. Therefore, the
effects of religiousness and participatory
salience are not greater on behaviors that violate
ascetic norms as claimed by Middleton and Putney
(1962) and Burkett and White (1974). Additional
evidence of nonsupport for the Anti-asceticism
hypothesis is observed in a comparison of the risk
factors. Larger values are associated with the
delinquent behaviors. The result of these
comparisons provides no support for the Anti-
asceticism hypothesis.

5) Finally, Burkett's suggestion that the Anti-
asceticism hypothesis be revised by the addition
of indirect religiosity effects mediated through a
religiously based morality of personal asceticism
receives no support with these data. The
religiosity measures are not related to personal
asceticism and personal asceticism fails to attain
a statistically significant, inverse effect on
anti-ascetic behavior. In addition, essentially
no changes are observed in the parameter estimates






-75-


of the religiosity measures (i.e., bk, Pk' and Rk)
and no significant improvements in the predictive
capacities of the estimated models (i.e., RL
and P(Y=1)) are evidenced.














CHAPTER SEVEN

TESTS OF THE JENSEN AND ERICKSON NORM
QUALITIES HYPOTHESIS



Each of the analyses examined so far have been

restricted to simple social-psychological models which

stress the constraining function of religiosity on deviant

behavior. Though these models are based on reasonable

hypotheses, the empirical support provided to them has been

somewhat limited. This may be due, in part, to

specification error. That is, these hypothetical models

limit analysis to an overly simplistic process, perhaps

they should incorporate more structural or group-level

processes.

For instance, Jensen and Erickson (1979) argue that one

of the major shortcomings of the extant literature is the

failure of previous researchers to examine the differential

effects of religiosity across religious affiliations.

Indeed, a majority of the prior research has failed to

consider the possibility of an interaction between religious

affiliation and religiosity on deviant behavior. Jensen and

Erickson claim that the varying moral messages of different

religions constitute important normative climates which

affect the deterrent impact of religiosity on anti-ascetic


-76-






-77-


behaviors, particularly substance use. Thus, while some

denominations take a strong proscriptive stand against use,

others are more prescriptive, and, according to Jensen and

Erickson, the religiosity effect is stronger among those

from proscriptive normative climates.

The analyses reported in this chapter test the primary

hypothesis derived from these ideas (referred to as the Norm

Qualities hypothesis H3); first with marijuana use as the

dependent variable, then with alcohol use. Logistic

regression models testing the effects of religiousness,

participatory salience, and personal asceticism are

presented for each level of both attributed and perceived

denominational proscriptiveness (prescriptive and

proscriptive). The results of these tests for marijuana use

are reported in Table 10, and for alcohol use in Table 11 of

Appendix E.



Marijuana Use: Attributed Denominational
Proscriptiveness


Models 1 and 2 of Table 10 present the findings of

analyses testing the effects of religiosity and personal

asceticism on self-reported marijuana use across levels of

attributed denominational proscriptiveness. Comparing the

results in Model 2 to those in Model 1 (i.e., comparing the

attributed proscriptive denominations model to the

attributed prescriptive denominations model) provides the






-78-


statistical evidence needed to test the Norm Qualities

hypothesis. Comparisons of the statistical data in Table

10 provide little support for the hypothesis.

For instance, the independent effects of the personal

asceticism and both of the religiosity variables are

inversely related to marijuana use for those adolescents

belonging to prescriptive denominations (Model 1); however,

only the inverse effect of participatory salience is

statically discernible from zero for those youth from

denominations which traditionally have held proscriptive

norms toward sensual indulgences (Model 2). Likewise, the

relative effects of these variables are just as strong or

stronger in the prescriptive model as they are in the

proscriptive model (i.e., in the attributed prescriptive

model versus the attributed proscriptive model, bk = -.411

vs. -.098, respectively, for religiousness; -.444 vs. -.476

for participatory salience; and -.425 vs. .210 for personal

asceticism). Thus the odds of marijuana use are estimated

to be [exp(-.411)]3 = .29 times as high for the "quite

religious" as for the "not religious at all" from attributed

prescriptive denominations (Model 1) compared to a multiplica-

tive effect on the odds of [exp(-.098)]3 = .74 for those

youths belonging to denominations traditionally attributed

as proscriptive (Model 2). For participatory salience,

these estimates are more similar (i.e., [exp(-.444)]3 = .26

in the prescriptive model versus [exp(-.476)]3 = .25 in the

proscriptive model). Finally, a very large difference





-79-


contrary to that predicted by the Norm Qualities hypothesis

is observed in a comparison of the multiplicative effects on

the odds for the personal asceticism variable in these

models. That is, while the odds of marijuana use are, as

expected, much lower for the ascetic youth than for the non-

ascetic youth in the prescriptive model (Model 1), the odds

of use for ascetic youth in the proscriptive model (Model 2)

are greater than that estimated for the non-ascetic youth

(i.e., exp(-.425) = .65 versus exp(.210) = 1.23). The

instantaneous rate of change coefficients (Pk) also suggest

that greater proportional reductions in the predicted

probabilities of use are associated with the religion

variables in the prescriptive model than in the proscriptive

model.

These findings are, of course, contrary to that

predicted by the Norm Qualities hypothesis. The cumulative

effect of these variables also provides little support for

the hypothesis. While the overall predicted probability of

marijuana use for the weakly religious is smaller for those

youth from proscriptive denominations--P(Y=l) = .54

versus .73--and is consistent with the Norm Qualities

hypothesis, the predicted probabilities of use for the

strongly religious fail to support the hypothesis. That is,

strongly religious adolescents from proscriptive

denominations are more likely to use marijuana than those

from prescriptive denominations--P(Y=l) = .21 and .12,

respectively. The differences in the predicted






-80-


probabilities of use between the weakly and strongly

religious when evaluated at the means of the control

variables are greater for those from prescriptive

denominations (.61) than for those from denominations

attributed to be proscriptive (.33). Likewise, a larger

risk factor value is associated with denominational

prescriptiveness (i.e., risk factors = 6.1 and 2.6,

respectively for Models 1 and 2).

In sum, these data on the normative orientations of the

various religious affiliations (as traditionally attributed

to them) fail to support the Norm Qualities hypothesis. In

fact, each of the comparisons made suggests that the impact

of religiosity and personal asceticism on marijuana use is

greater for those adolescents with prescriptive

denominational affiliations, rather than for those from

proscriptive denominations as the Norm Qualities hypothesis

would predict.



Marijuana Use: Perceived Denominational
Proscriptiveness


Models 3 and 4 of Table 10 present the statistical data

necessary for a second test of the Norm Qualities hypothesis

on adolescent marijuana use. The comparisons made here are

based on proscriptive/prescriptive denominational groupings

measured by the respondents' perceptions of the normative

orientation of their religious affiliations. Unlike the

data on attributed norm qualities these comparisons based on





-81-


perceived norm qualities yield some support for the Norm

Qualities hypothesis. That is, the relative effect of

personal asceticism is greater among those youth belonging

to denominations that are perceived to be proscriptive

In Model 3 of Table 10, the perceived denominational

prescriptiveness model, all three of the religion variables

are inversely related to marijuana use (bk= -.364 for

religiousness, -.457 for participatory salience, and -.164

for personal asceticism); however, personal asceticism does

not have a statistically significant effect. For those who

perceive the moral climate of their religious denomination

toward marijuana use as proscriptive (Model 4), personal

asceticism is inversely related to use (bk= -1.858), so are

the two religiosity variables (bk= -.022 and -.163 for

religiousness and participatory salience, respectively), but

the effects of the religiosity measures are not

statistically distinguishable from zero. Thus the religious

variables that affect adolescent marijuana use vary by

perceived denominational proscriptiveness.

Likewise, the effectiveness of these measures varies by

perceived denominational proscriptiveness. That is, when

cross-model comparisons are made, stronger relative effects

are observed for the two religiosity measures in Model 3

(perceived prescriptiveness) than in Model 4 (perceived

proscriptiveness), but the relative effect of personal

asceticism is stronger in the proscriptive model (Model 4).

These findings suggest that the deterrent effect of






-82-


religiosity, per se, on marijuana use is stronger among

those youth who do not perceive their religion taking a

strong proscriptive stand. However, where the normative

orientation of one's religion is perceived to be

proscriptive, personal asceticism has the strongest

deterrent impact. Thus, both the variables that affect

marijuana use and their relative effects vary across levels

of perceived denominational proscriptiveness. Therefore,

these comparisons provide mixed support for the Norm

Qualities hypothesis.

This mixed support is also evident in a comparison of

the combined or cumulative effect of these measures across

models. The smallest predicted probability of

marijuana use is .13 for the strongly religious in the

proscriptive model (Model 4) and the greatest probability of

use is .69 for the weakly religious in the prescriptive

model (Model 3). These probability rankings are consistent

with that predicted by the Norm Qualities hypothesis.

However, the predicted probability of use for the strongly

religious in the prescriptive model is essentially equal to

that in the proscriptive model (i.e., P(Y=1)= .14 and .13).

Likewise, the risk factor values for both models are nearly

identical (4.9 and 4.8 for Models 3 and 4 respectively) and

the difference in the predicted probability of use between

the weakly and strongly religious in the prescriptive model

is larger than the difference in the proscriptive model (.55

versus .49). These results are counter to that expected by





-83-


the Norm Qualities hypothesis and serve to limit the support

provided to it by these data.



Alcohol Use: Attributed Denominational
Proscriptiveness


The effects of religiosity and personal asceticism on

self-reported alcohol use across levels of attributed

denominational proscriptiveness provide yet another test of

the Norm Qualities hypothesis and are presented in Models 1

and 2 of Table 11 in Appendix E. Unlike the findings for

marijuana use the findings on alcohol use provide rather

solid support for the Norm Qualities hypothesis. While

participatory salience is the only religion variable in

either model to be significantly related to alcohol use (bk

-.045 in Model 1 and -.211 in Model 2), the relative effects

of all three religion variables are greater in the

proscriptive model (Model 2). For instance, bk= -.045 for

religiousness in the attributed prescriptive model (Model 1)

indicates that the odds of alcohol use are multiplied by

only exp(-.045) = .96 for every one-unit increase in

religiousness, while in the proscriptive model (Model 2),

the multiplicative effect on the odds of use is somewhat

stronger at exp(-.211) = .81. Thus, the odds of alcohol use

are estimated to be .963 = .87 times as high for the "quite

religious" as for the "not religious at all" in prescriptive

denominations. The odds of use estimated for "quite

religious" adolescents belonging to denominations






-84-


traditionally attributed as proscriptive are .813 = .53

times as high as the odds estimated for those adolescents

from proscriptive denominations who are "not religious at

all."

Within the prescriptive model (Model 1), a one-unit

increase in participatory salience is estimated to have an

exp(-.350) = .70 multiplicative effect on the odds of

alcohol use, while within the proscriptive model (Model 2),

its multiplicative effect is much more powerful (i.e.,

exp(-.535) = .59). These effects translate into estimated

odds of alcohol use that are .703 = .35 (in Model 1)

and .593 = .20 (in Model 2) times as high for those

adolescents for whom participation in church activities is

"very important" as for those youth who feel such

involvement is "not important at all." Thus, the inhibitory

effect of participatory salience, like that of

religiousness, is stronger in the proscriptive model (Model

2). The effect of personal asceticism on alcohol use is

also more powerful in the proscriptive model (i.e.,

exp(-.560) = .57 in Model 2 versus exp(-.380) = .68 in Model

1).

The instantaneous rate of change coefficients indicate

similar effects. For adolescents from prescriptive

denominations, each one-unit increase in religiousness is

associated with a 4.2 percentage point decrease in the

predicted probability of alcohol use. For youth from

denominations that are more prescriptive in their stands on






-85-


alcohol consumption, the proportional decrease in predicted

probability for each one-unit increase in religiousness is

only 0.5 percentage points. Thus, teens from proscriptive

denominations who are "quite religious" are 12.6 percentage

points less likely to use alcohol than are those who are

"not religious at all," while almost no change in the

predicted proportional probability of use can be observed

between those adolescents from prescriptive denominations

who are "quite religious" and those who are "not religious

at all."

Likewise, within proscriptive denominations, each one-

unit increase in participatory salience results in a 10.7

percentage point decrease in the predicted probability of

use. As such, adolescents for whom involvement in church

activities is "very important" are 32.1 percentage points

less likely to use alcohol than are those for whom such

participation is "not important at all." However, among

adolescents belonging to prescriptive denominations, a one-

unit increase in participatory salience is associated with

only a 4.2% decrease in the proportional probability of use

which translates into a 12.6 percentage point difference

between those youth at maximum and minimum levels of

participatory salience.

Finally, the instantaneous rate of change

coefficients for personal asceticism also indicate a

stronger effect in the proscriptive model than in the

prescriptive model. That is, youth from proscriptive






-86-


denominations who define alcohol use proscriptively have a

predicted probability of use that is 11.2 percentage points

less than that of adolescents also from proscriptive

denominations but who hold prescriptive beliefs. The

proportional difference in the predicted probability of use

between prescriptive and proscriptive teens belonging to

prescriptive denominations is much less (4.7 percentage

points).

However, cross-model comparisons of the partial

correlation coefficients for the relative effects of these

three variables indicate small differences (i.e., Rk=

-.000 versus -.000 for religiousness, -.082 versus -.130 for

participatory salience, and -.011 versus -.013 for personal

asceticism in Models 1 and 2, respectively). In sum,

however, nearly every comparison the independent effects of

religiosity and personal asceticism on alcohol use are only

slightly stronger in the proscriptive model (Model 2) than

in the prescriptive model (Model 1). As such, they provide

some support for the Norm Qualities hypothesis.

Additional support for the hypothesis is evident in

comparisons of the cumulative impact these religion

variables have on adolescent alcohol use. Those adolescents

most likely to use alcohol are weakly religious and belong

to prescriptive denominations (P(Y=1)= .93); those least

likely to use alcohol are strongly religious and are members

of proscriptive denominations (P(Y=1)= .38). The predicted

probability of use for the weakly religious is nearly






-87-


identical in both models (.93 versus .91), yet the

probability of use predicted for the strongly religious is

far greater for those from prescriptive denominations than

from proscriptive denominations (.74 versus .38). As the

Anti-asceticism hypothesis undergoes revision and

respecification leading to the Norm Qualities hypothesis,

the predicted probability of alcohol use for the weakly

religious remains stable at about .93. However, the

predicted probability of use for the strongly religious,

originally estimated at .69, reduces down to .60 under

Burkett's revision, and reduces further to .38 when

estimated for those adolescents belonging to denominations

that traditionally take a strong stand against alcohol use.

In sum, solid support for the Norm Qualities hypothesis

is provided by these data on adolescent alcohol consumption.

The strength of both the independent and the combined

effects of the religiosity and personal asceticism variables

on self-reported alcohol use increase as attributed

denominational proscriptiveness increases.



Alcohol Use: Perceived Denominational
Proscriptiveness


This support for the Norm Qualities hypothesis is

compromised when measures of perceived denominational

prescriptiveness and proscriptiveness are used. The

findings reported in Models 3 and 4, in fact, run directly

counter to the Norm Qualities hypothesis. That is, the






-88-


effects of religiosity and personal asceticism on adolescent

alcohol use is stronger under perceived prescriptive

denominational norms (Model 3) than under perceived

proscriptive denominational norms (Model 4). This finding

may be due, in part, to the skewed distribution of the

perceived denominational proscriptiveness variable which

resulted in an n of only 137 subjects for the proscriptive

model. Perhaps the findings would be different had a larger

sample been available for analysis.



Conclusions


The four sets of comparisons presented in this chapter

(see Table 12 of Appendix F) provide grounds for the

following conclusions regarding the Norm Qualities

hypothesis:

1) Both the independent and the combined effects of
the three religion variables on self-reported
adolescent marijuana use are as strong or stronger
for those youth belonging to religious affiliations
traditionally attributed to be prescriptive in
normative orientations. These findings do not to
support the Norm Qualities hypothesis.

2) When denominational proscriptiveness is measured by
the subject's perception of their religion's
normative stance regarding marijuana use, then
support for the Norm Qualities hypothesis is
somewhat mixed. Both the variables that affect
adolescent marijuana use and the strength of their
effects vary across levels of perceived
denominational proscriptiveness. The deterrent
impact of the two indicators of religiosity is
stronger within religious affiliations perceived to
be prescriptive, while personal asceticism deters
use among those youth belonging to perceived
proscriptive denominations.






-89-


3) Strong support for the Norm Qualities hypothesis is
provided by the data on adolescent alcohol use when
analyzed across levels of attributed denominational
proscriptiveness. Nearly every comparison of the
independent effects of religiosity and personal
asceticism on alcohol use is stronger in the
proscriptive model than in the prescriptive model.
Likewise, comparisons of the cumulative impact of
these variables also support the Norm Qualities
hypothesis. While the predicted probability of use
remains constant across models for the weakly
religious, the probability of use for the strongly
religious is substantially less for those
adolescents belonging to proscriptive denominations
than for those from prescriptive religions (i.e.,
P(Y=1)= .38 versus .74, respectively).

4) However, extremely little support for the Norm
Qualities hypothesis is evident in these data on
self-reported adolescent alcohol use when
comparisons are made across levels of perceived
denominational proscriptiveness.

Why is the support for the Norm Qualities hypothesis so

varied? Why is the hypothesis supported for adolescent

marijuana use when analyzed across levels of perceived

denominational proscriptiveness but not for attributed

denominational proscriptiveness? Why is the hypothesis

so strongly supported for adolescent alcohol use when

analyzed across levels of attributed denominational

proscritiveness but not when analyzed across levels of

perceived denominational proscriptiveness? Perhaps the

following tentative explanations will serve to answer these

questions.

First, with regard to the findings on adolescent

marijuana use, the prescriptive/proscriptive denominational

attributions are based on the traditional stands these

religious affiliations take regarding alcohol use. These






-90-


may not be wholly accurate attributions, for there really

are no long-standing official doctrines relating

specifically to marijuana use. Thus, the lack of support

for the Norm Qualities hypothesis observed for marijuana use

may be a function of inaccurate attributions of

denominational norm qualities. However, it is quite

likely that during the time of this survey (1977) many

individual churches took an open and vigorous stand against

adolescent marijuana use. Those adolescents who attended

church when sermons were delivered, and Sunday school

discussions were held, denouncing the use of marijuana were

better able to accurately perceive the normative orientation

of their church while those failing to attend were less

accurate in their perceptions. Thus, accuracy in perception

may be related to religiosity which, in turn, is inversely

related to marijuana use. It may be such a spurious

relationship that accounts for the support provided to the

Norm Qualities hypothesis when analyzed across levels of

perceived denominational proscriptiveness. Then again,

perhaps the hypothesis is supported by these data simply

because it is valid.

There is, on the other hand, a long tradition of

official church doctrine regarding the use of alcohol

(Chalfant et al., 1981; Salisbury, 1964). These

prescriptive/proscriptive doctrines permit the clean and

accurate attribution of normative orientation, and, in turn,

provides perhaps the best test of the Norm Qualities




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.8 - mvs