You are on page 1of 19

Society for the Study of Social Problems

Television Violence and Violent Crime: An Aggregate Analysis


Author(s): Steven F. Messner
Source: Social Problems, Vol. 33, No. 3 (Feb., 1986), pp. 218-235
Published by: University of California Press on behalf of the Society for the Study of Social Problems
Stable URL: http://www.jstor.org/stable/800706
Accessed: 23/09/2010 02:49

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucal.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Society for the Study of Social Problems and University of California Press are collaborating with JSTOR to
digitize, preserve and extend access to Social Problems.

http://www.jstor.org
SOCIAL PROBLEMS, Vol. 33, No. 3, FEBRUARY 1986

TELEVISION
VIOLENCE
ANDVIOLENT
CRIME:
ANAGGREGATE
ANALYSIS*
STEVENF. MESSNER
State University of New Yorkat Albany

Inthispaper,Iexaminethe relationship betweenlevelsof exposureto televisionviolence


and ratesof violentcrimeforsamplesof populationaggregates.Myprimaryhypothesis
is that populationaggregates withhigh levels of exposureto violenttelevisioncontent
willexhibithigh ratesof criminalviolence.The resultsof a series of multivariate,
cross-
sectionalanalyses failto provideany supportforthis hypothesis.Contraryto expecta-
tions,aggregatelevelsof exposureto televisionareconsistentlyrelatedto ratesof violent
crimein an inversedirection.Analysesbased on differingspecificationsand alternative
measuresof televisionexposureallyieldhighlysimilarresults.I suggest thatthe theory
of "criminal
subcultures"and the "routine approachofferpossibleexplanations
activities"
for these seeminglyanomalousfindings.

Television is one of the most important media of mass communications in the United States.
National Commissions have concluded that "almost everyone"in the United States watches televi-
sion (National Institute of Mental Health, 1982:1), and surveys suggest that the television in an
average household is turned on for about seven hours everyday (Comstock, 1980:30). Given the
pervasiveness of television in U.S. society, it is not surprising that this medium has been, and
continues to be, at the center of controversy and public debate. Perhaps the most intense and
enduring controversy concerns the consequences of viewing television violence. Critics of the
industry have charged that violence on television is a cause of many of the more important social
ills of society, while representativesof the industry and a number of social scientists have disputed
the scientific basis for such sweeping claims.'
The purpose of this paper is to address an aspect of the television controversy that has received
comparatively little attention in the scientific literature but that serves as the basis for many of
the calls for some sort of censorship of television content. This is the claim that violence on televi-
sion is an important cause not simply of "aggression"but of "criminalviolence" in society at large.
For reasons to be discussed shortly, an aggregate-level, cross-sectional analysis will be performed
to explore the relationship between television violence and violent crime. The key question guiding
the analysis can be stated quite simply. Do population aggregates with high levels of exposure to
violent television content also exhibit high rates of criminal violence?

PREVIOUSRESEARCH
There is an extensive body of literature dealing with the potential effects of television viewing
on human behavior.2 The National Institute of Mental Health's (NIMH) update of the original

* An earlierversionof this paperwaspresentedat the Seventy-ninth


AnnualMeetingof the AmericanSocio-
logicalAssociation,August, 1984.I am gratefulto MitchellChamlinandAnne Roschellefor assistancewith
data collection and processing. Discussions with MurrayStraus in the course of a related, collaborative project
have contributed greatly to the development of the index of television violence. Allen Liska and Richard Hall
provided helpful comments on a proposal submitted to the Research Foundation of the State University of
New York. A grant from the S.U.N.Y. Research Foundation is gratefully acknowledged. I would also like to
thank Richard Felson for a review of an earlier draft of this manuscript, and William Behanna of the A.C.
Nielsen Company for granting me access to the relevant Nielsen reports. Of course, I alone assume full respon-
sibility for the analyses and conclusions contained herein.

1. The view that researchon television justifies some form of censorship can be found for example in Eysenck
and Nias (1978); also see Belson (1978). Verydifferent interpretations of the scientific literatureare represented
in Howitt and Cumberbatch (1975) and in Kaplan and Singer (1976). For the views of some television producers
and actors, see "Televiolence Prepetrators: Don't Blame Us" (New YorkDaily News, 1982:114).
2. The "classic"literatureon the effects of television viewing includes the field studies in the 1950s by Himmel-
Televisionand ViolentCrime 219
1972 Report of the Surgeon General'sAdvisory Commission on Television and Behavior notes that
over 2,500 publications dealing with the general behavioral influences of television have appeared
between 1972 and 1982 alone (NIMH, 1982:iii). Much of this research has been concerned with
the possibility that the viewing of television violence might lead to aggressive or violent behavior
on the part of the observer. Although there are a few dissenting voices, the most widely accepted
interpretation of the accumulated evidence is probably that contained in the NIMH update. The
authors of this report conclude, that all things considered, "recent research confirms the earlier
findings of a causal relationship between viewing televised violence and later agressive behavior"
(NIMH, 1982:89).
Even though some consensus seems to be emerging with respect to the existence of a causal
connection between television violence and agression, the larger social significance of this relation-
ship has not been firmly established. The vast majority of previous studies has dealt with relatively
minor forms of aggression. The question thus arises as to whether or not the findings for the kinds
of aggression typically studied by social scientists can be extended to more serious forms of aggres-
sion such as criminal violence.
A primary reason for the restricted focus on mild forms of aggression in previous research is
the preference of social psychologists for "micro-level"research designs. These designs, although
strong in many respects, are not well suited for studying a dependent variable such as criminal
violence. Ethical considerations preclude the use of an experimental design in the study of violent
crime because it would clearly be unethical to introduce a stimulus, either in the laboratory or
in the field, suspected of provoking harmful behavior.3 Correlation studies, in contrast, are not
subject to the same ethical constraints, but the "micro-level"correlational design (i.e., a sampling
of individuals) encounters problems of a different sort. Because serious crimes of violence are
extremely rare,the probability of selecting an appreciable number of individuals who have engaged
in such acts is highly remote unless samples sizes are unrealistically large.4
Although micro-level designs have been predominant in the research on the effects of television,
there have been several important studies which have adopted macro-level approaches and have
been able, as a consequence, to investigate the more serious forms of aggression such as criminal
violence. Clark and Blackenburg.(1972) examined the relationship between the amount of violent
television content and national rates of violent crime in the United States between 1953 and 1969.
They found that the proportion of television programs judged to be violent did not correlate with
national rates of criminal violence either directly or with a time lag. This study raises questions

weitet al. (1958)and Schrammet al. (1961),and the laboratoryexperimentsby Bandura(1965;1973)and by


Berkowitzandcolleagues(BerkowitzandAlioto, 1973;BerkowitzandRawlings,1963).Additionalfieldstudies
of specialimportanceconductedin the 1960sand 1970sarethose by Belson(1978),Eron(1982),Eronet al.
(1983),McCarthyet al. (1975a),and Parkeet al. (1977;also see Leyenset al., 1975).Onehighlysophisticated
researcheffort with findingsopposedto muchof the literatureis the panelstudyby Milavskyet al. (1982).
Surveysof this trulyextensivebody of researchcan be foundin Andison(1977),Comstock(1975),andNIMH
(1982).
3. The field studyby Milgramand Shotland(1973)probablyrepresentsthe outerlimitsof ethicallyjustified
experimenton televisionandcriminalbehavior.MilgramandShotlandprovidedsubjectswithan opportunity
to breakintoboguscharityboxesafterhavingbeenexposedto violentor prosocialtelevisioncontentandhaving
been subjectedto a frustratingexperience.Theyreportno significantdifferencein the liklihoodof breaking
into the boxes acrosstreatmentgroups.
4. Thereis somemicro-level researchthatis at leastindirectlyrelevantto thequestionof therelationship
between
televisionviolenceand criminalformsof behavior.Menziescomparedthe televisionpreferencesof inmates
with historiesof violentbehaviorto those of inmateswith no such histories(citedin Comstock,1975:197).
Contraryto whatmightbe expectedon the basisof researchfocusedon relativelymild formsof aggression,
he foundthat violentinmatesdo not exhibita greaterpreferencefor violenceon televisionthando theirnon-
violentcounterparts.Additionalstudiesthathaveexaminedthetelevisionpreferences of delinquents
andcompared
themto thepreferences of non-delinquentshaveproducedcontradictory results- contrastMcIntyreandTeevan
(1972)and Robinsonand Bachman(1972)with Halloran(1978).
220 MESSNER
about the generalizability of findings for mild forms of aggression to more serious forms; but the
small number of observations in this time-series analysis limits its significance.
Hennigan et al. (1982) also used aggregate data to assess the impact of television on different
types of crime. They examined crime rates before and after the introduction of television for a
sample of American communities. Rather surprisingly,they observed a consistent increase in rates
of larceny-theft associated with the introduction of television but no significant change in rates
of violent crime. The authors acknowledged, however, that their study may not provide a truly
accurate assessment of the effects of television violence and violent crime. During the early years
of television, violence was depicted less frequently and less realistically than it is today.
The most ambitious effort to use aggregate data to evaluate the impact of media violence on
anti-social behavior, including violent crime, is Phillips' research on various forms of "imitative"
violence. He reported that publicity surroundings suicides is associated with subsequent increases
in suicide rates (Bollen and Phillips, 1981; Phillips, 1974, 1982) and automobile fatalities, which
might be viewed as disguised suicides (Phillips, 1979). Of greatest relevance in the present context
is Phillips' researchon the relationship between the publicity surrounding heavyweight prize fights
and the total number of U.S. homicides. He discovered that homicides increase by a significant
amount three days after championship fights, and that prize fights which are discussed on the
network evening news are followed by a significantly greater increase in homicides than are fights
which do not receive this television publicity (Phillips, 1983:563). These results seem to suggest
that publicizing violence on television produces higher levels of "real life" criminal violence.
However, it should be recognized that Phillips' analyses have been limited to relatively brief periods
surrounding the publicized violence, which leaves open the question of whether media violence
has lasting effects on overall levels of criminal violence.
In short, once the focus of attention shifts from "aggression"to "violence," and to "criminal
violence" in particular, the evidence on the effects of television viewing becomes much less conclu-
sive. The cautious assessment made by George Comstock (1980) seems highly appropriate. Despite
the vast number of studies on the effects of television, the research to date does not tell us with
any certainty "the extent or degree to which violent television programmingmay contribute to crime
or serious antisocial behavior" (1980:109).
The purpose of the present study is to introduce an original source of evidence that bears upon
the relationship between television violence and criminally violent behavior. The evidence derives
from a cross-sectionalanalysis of levels of exposureto violent television content and rates of criminal
violence for samples of population aggregates. There are two noteworthy advantages associated
with this type of research design. First, the use of population aggregates rather than individuals
facilitates the investigation of the extremely serious forms of criminal aggression that are at the
center of public concern about media violence. Secondly, the selection of a cross-section of popu-
lation aggregates at a single point in time results in a fairly large sample of observations. The avail-
ability of a large samples makes it possible to conduct multivariateanalyses and to assess the impor-
tance of indicators of television violence in comparison with other independent variables.

THEORETICAL RATIONALE
The central hypothesis for this study is that there will be a significant, positive relationship
between levels of exposure to television violence and rates of violent crime. In essence, this hypoth-
esis is based on theories and empirical findings concerning the determinants of mild forms of
aggression at the individual level which I extend to account for more serious forms of aggression
at the aggregate level.
Social psychologists have identified several processes that could produce a positive relationship
between media violence and aggressive behavior. One such process is that of "modeling."
According to the major tenets of social learning theory, individuals learn to behave aggressively
Televisionand ViolentCrime 221
by observing the behavior of others (Bandura, 1973). Thus, heavy exposure to violent television
content should provide people with violent models and, in turn, raise the probability of violent
behavior, including criminal violence.
Another psychological process which has been linked to television violence is "desensitization."
Research indicates that children who have been exposed to portrayals of violence become more
accepting of violence (see Eysenck and Nias, 1978). On the basis of such findings, it has been
suggested that television violence reduces the anxiety associated with aggression and makes people
more likely to carry out acts of aggression, possibly even very serious forms of aggression (Eysenck
and Nias, 1978:184).
"Disinhibition theory" provides an additional theoretical rationale for anticipating a relation-
ship between media violence and aggression. According to this theory, people are inhibited by
socialization experiences from expressing aggressive impulses. These inhibitors break down, it has
been argued, when individuals are exposed to a steady diet of media violence (NIMH, 1982:39).
Clearly, there are ample theoretical grounds for anticipating that individuals exposed to violence
on television will be especially prone to violent behavior. A straightforward implication of this
micro-level theorizing is that population aggregates with high levels of exposure to violent televi-
sion will be characterized by large numbers of individuals prone to violence and, accordingly, will
exhibit relatively high rates of violent crime.5

METHODS
A crucial requirement in this study is a measure of the level of exposure to television violence.
Such a measure must reflect two components: the number of persons viewing television programs,
i.e., audience size, and the content of these programs scaled along the dimension of violence. The
ideal procedure would be to estimate the viewing audience for all television programs broadcast
in a given geographical area for some specified time interval, weight these viewing estimates by
evaluations of the violent content of the respective programs, and then combine the weighted esti-
mates into a composite index. However, this approach would be highly impractical because it
would requiredetailed information on the content of all shows broadcast during the specified time
interval. Conducting the content analysis required for such a measure would be a monumental
task well beyond the scope of the present project.
Instead, I used an alternativeapproach which involved identification of the "most violent"regular
television series and measurement of the typical audience size attracted to these violent programs.
Selection of the "violent" television series was based on the content analyses performed by the
National Coalition on TelevisionViolence (NCTV, 1981).The NCTV regularlymonitors the amount
of violence portrayed in prime time network telecasts and records the number of "violent acts per
hour." A list of the shows that are the most violent according to the "violent acts" criterion is then
compiled. For the time period under investigation (the Fall season of 1980), the most violent regular,
prime-time series, as judged by NCTV, were"Enos,""The IncredibleHulk," "Hartto Hart," "Dukes

5. Althoughthe social psychologicaltheoriesreviewedaboveprovidethe foundationfor the hypothesisof


an aggregaterelationshipbetweentelevisionviolenceexposureandratesof crime,themacro-levelanalysisdoes
not constitutea formaltest of these theories.The findingof a positiveassociationbetweenaggregatelevels
of exposureto televisionviolenceand ratesof violentcrimewouldnot necessarilyimplythat the individuals
who are watching violent television are the same individuals who are committing the illegal acts. My purpose
is not to assess the validity of the aforementioned social psychological perspectives but to determine whether
macro-level relationships involving criminal violence parallel the fairly well-established micro-level relation-
ships involving "aggressive behavior." Much of the concern over the potentially harmful effects of television
violence seems to be based on an implicit assumption that the social psychological evidence can in fact be
generalized to aggregate-level explanations of high rates of crime in U.S. society. The results of the present
research will permit an assessment of this important but undocumented assumption.
222 MESSNER
of Hazzard,"and "FantasyIsland."6I selectedtheseprogramsas a "strategicsample"to be used
in the constructionof the index of exposureto televisionviolence.
Theviolentcontentratingsof theNCTVareevidentlyquitereliable.TheNCTVnewsletter reports
a score-by-scoreagreementof .81 for its monitoringteam (NCTV,1981:4).In addition,previous
researchon evaluationsof televisioncontentrevealsconsiderableconsensusamongthe U.S.public
overjustwhatconstitutesviolenttelevisionprogramming (Greenberg andGordon,1972;Milavsky,
1982:80).Therefore,there is no compelling reason to expect that the programsidentifiedas most
violentbythe NCTVmonitorswoulddiffersignificantlyfromthoseidentifiedby otherevaluators.7
Estimatesof audiencesizefor theseviolenttelevisionserieswereobtainedfromthe A.C.Nielsen
Company.8Duringfour monthsof the year-November, February,May,and July-the Nielsen
Companymeasuresviewingbehaviorat the local level.The estimatesfor Novemberwereselected
for the presentanalysisbecausethis time intervalcorrespondsmost closelywith the time period
coveredbytheNCTVcontentanalyses.Unfortunately, thetemporalmatchis not perfect.TheNielsen
ratingsperiod extend from October 30 to November 26, whereasthe NCTV content analyses
encompass the November 10 to December 27 period.This slight "misalignment" should not be
a seriousproblemin the anlaysisof regulartelevisionseriesbecausetheNCTVsamplingof program
contentoveran appreciabletimeinterval(morethan four weeks)is likelyto yield a representative
indicationof typicalprogramcontentfor suchseries.Thespecificcountsof "violentactsperhour"
are probablynot veryreliableand, for this reason,no attemptwas made to use these counts as
weightsin the index. Of course,for specialprogramming,the contentratingsapplyonly to the
specificprogramsbroadcast.Consequently, networkmoviespecialswerenot includedin the sample
of programsfor the exposureindexeventhoughtheyappearin the NCTVlist of the most violent
primetime shows.
The Nielsendata on audiencesize provideestimatesof the proportionof the total numberof
televisionhouseholdsreachedby networkstations.Almost all U.S. householdsarenow television
households.Hence,the proportionof "televisionhouseholds"reachedis virtuallyequivalentto
the proportionof all householdsin the population.
Afteridentifyingthe fivemostviolenttelevisionseriesandgatheringinformationon the propor-
tion of householdsregularlytuned in to these showsoverthe courseof a one-monthinterval,I
computedan indexof exposureto televisionviolenceby simplyaveragingthe proportionsacross
thesefiveviolentshows.Theresultingscoremeasurestheaveragerelativesizeof theviewingaudience
attractedto programswith characteristically violentcontent.High valueson this indexindicate
thatlargerelativenumbersof residentsareregularlyviewingtelevisionprogramswithviolentcontent;
low valuesimply the reverse.
The computationof the televisionexposureindexinvolvedthe use of a ratherunusualunit of
analysis.The Nielsendata on audiencesize are aggregatedon the basis of "DesignatedMarket
Areas"(DMAs).DMAsaregeographicalunitscomprisedof the countiesservedby localtelevision

6. The NCTV actually lists the 10 most violent television shows. From the original 10, "Freebieand the Bean"
had to be excluded from the analysis because no data were available for this program on audience size, and
"Charlie'sAngels" had to be excluded because it was not broadcast by many local affiliates. Three network
movie specials were also excluded for reasons discussed later in the text. These deletions account for the selec-
tion of only five programs from the original list of the 10 most violent shows.
7. The NCTV is an organization with membership drawn largely from the medical profession, from research
universities, and from civic and religious associations. A listing of the board of directors and the "endorsers"
appears in the NCTV newsletter. The NCTV is unmistakably an activist organization devoted to a reduction
in levels of television violence. Consequently, it is reasonable to expect that the NCTV monitors might be espe-
cially sensitive to the presence of "violence" in television programming. It seems unlikely, however, that the
NCTV rankings of shows in terms of violent content would diverge markedly from the rankings of monitors
without the same activist orientations.
8. I am verygratefulto WilliamR. Behanna,Directorof PressRelationsfor the A.C. NielsenCompany,for
making available the television ratings data and for explaining various aspects of the Nielsen reports.
Televisionand ViolentCrime 223
stations. DMAs are larger than the more common governmental units of "StandardMetropolitan
Statistical Areas" (SMSAs). To be able to use the Nielsen data on audience size in an SMSA anal-
ysis, it was necessary to locate each SMSA within an encompassing DMA. The DMA value on
the television index was assigned to the "residing" SMSA.
The dependentvariablesare the SMSA rates(per 100,000population) of violent crime.The offenses
include criminal homicide, forcible rape, robbery,and aggravatedassault. Crime rates are measured
for 1981 since the television index refers to a period near the very end of 1980. The source for the
crime data is the FederalBureauof Investigation's(1982) Uniform CrimeReports(UCR). The offenses
are considered singly, and in combination as a composite index of violent crime. The composite
index is simply the sum of the individual rates.
The UCR data are based on reports to the police. As such, they are subject to all of the criticisms
of official crime statistics that have been discussed at great length in the literature(see among others
Hindelang, 1974;Savitz, 1978).The problem of reportingbiases is especially worrisomein the present
analysis because a plausible case could be made that such biases would be systematically related
to the independent variable of primary interest- exposure to television violence. The reporting
behavior of residents in communities with high levels of exposure to television violence might very
well differ from that of residentsin communities with low levels of exposure.This would be expected,
for example, if "desensitization" is produced by exposure to television violence. The possibility
of such biases will be considered in the context of the actual results.
In an effort to minimizethe possibility of observing"spurious"associations, data werealso gathered
for a fairly comprehensive list of controls. The selection of the control variables was guided by
previous research indicating relationships between these variables and either television viewing or
levels of violent crime. On the basis of the research revealing that television viewing is associated
with low socio-economic status, low academic achievement, and minority status (Comstock, 1975),
measuresof the following socio-demographic characteristicsof SMSAs wereincluded:the percentage
of the population below the poverty line, the percentage with less than a high school education,
and the percentage black. A measure of the level of economic inequality, the Gini coefficient of
household income concentrations, was also included. Blau and Blau (1982) reported that the level
of inequality across SMSAs is related to rates of criminal violence. Another economic control vari-
able for the analysis was the averagemonthly payment to recipients of Aid to Families with Depen-
dent Children (AFDC). Significant relationships between AFDC payments and rates of violent
crime were reported by De Fronzo (1983).
Four population variables were also measured: total population; population per square mile;
the percentage of the population ages 18-34; and the number of males per 100 females. Total popu-
lation and population per square mile were log transformed because of highly skewed distributions
(Tufte, 1974:108-11).Severalprevious studies have found associations between population size, popu-
lation density,and ratesof violent crime (Blau and Blau, 1982;Mayhewand Levinger,1976;McCarthy
et al., 1975b). Substantial evidence indicates that young people and males constitute "high risk"
groups for criminal violence (cf. Gibbons, 1982). Finally, given the well-known thesis of a "regional
culture of violence" (Gastil, 1971;Hackney, 1969; but see Loftin and Hill, 1974), a dummy variable
to represent regional location (South = 1; non-South = 0) was also included in the list of controls.
Information for the precentage below the poverty line, percentage with less than high school
education, the Gini coefficient, and males per females was taken from the 1980 U.S. Census tapes
(U.S. Bureau of the Census, 1982b). The regional dummy variable has been scored in accordance
with Census Bureau classifications. The source for all other control variables is the State and Metro-
politan Area Data Book, 1982 (U.S. Bureau of the Census, 1982a).

RESULTS
Regression equations for the violent crime rates of SMSAs are reported in Table 1. The results
are quite surprising. For each measure of violent crime, the estimate for the level of exposure to
224 MESSNER
TABLE 1

Regression Results for ViolentCrimeRates, TVViolenceExposure,


and AdditionalCharacteristicsof SMSAs (N= 281)a
DependentVariables
Violent
Criminal Forcible Aggravated Crime
IndependentVariables Homicide Rape Robbery Assault Index
Exposureto TV Violence -.17* -.17* -.15* -.12 -.17"
Pct Black .43* .42* .39* .32* .43*
Pct 18-34 -.08 -.18* -.18* -.17 -.20*
Males/Females .14* .16* .13* .19* .19*
Pct less than h.s. edu. .21* -.37* -.06 -.05 -.09
Population(In) .29* .28* .41* .09 .30*
GiniCoefficient .41* .28* .36* .36* .42*
AFDC Payments -.18* -.01 -.03 .03 -.01
Pop per sq. mi. (In) -.09 -.15* .20* -.03 .08
Pct Poor -.12 .05 -.05 .00 -.03
South - .08 .05 - .23* .05 - .09
Adj. R2 .58 .39 .65 .27 .52
Notes:
a. Standardizedregression coefficientsare reported.
* p < .05

television violence is negative, and, with the exception of assault, all coefficients are statistically
significant. In other words, SMSAs in which large audiences are attracted to violent television
programming tend to exhibit low rates of violent crime.
There are several additional findings of interest in Table 1. The percentage black has a fairly
strong, positive effect on each measure of violent crime. This is consistent with previous research
detecting a relationship between racial composition and rates of violent crime for SMSAs (cf. Blau
and Blau, 1982). The measure of economic inequality (the Gini coefficient) also yields consistently
positive effects - SMSAs with high levels of income inequality tend to have high rates of violent
crime. Similar relationships have been observed at the societal level for the offense of homicide
(Braithwaite, 1979;Krohn, 1976; McDonald, 1976) and at the intra-national level for homicide and
assault (Blau and Blau, 1982). Sex composition and population size are both positively related to
violent crime rates - i.e., large SMSAs and SMSAs with large numbers of males relative to females
tend to exhibit high rates of violent crime (cf. Mayhew and Levinger, 1976). Finally, contrary to
a "regional culture of violence thesis," the only significant effect of region is on robbery, and the
effect indicates that robbery rates tend to be lower in Southern than in non-Southern SMSAs.
Given the highly unexpected nature of the findings for the television exposure measure in Table
1, it is important to consider the possibility that the observed patterns for this variable might reflect
certain methodological features of the analysis. One very common methodological problem that
arises in macro-levelresearchon crime it that of multicollinearity.To assess the extent of this problem
in the present study, "variance inflation factors" have been computed following the procedures
proposed by Fisher and Mason (1981).9

9. The varianceinflationfactor(cjj)for a givenindependentvariableis definedas: cjj= (1-Rj2)-' whereRj2


is thecoefficientof determination resultingfromtheregressionof thejth predictoron theremainingpredictors
(Fisherand Mason, 1981:105-6). The varianceinflationfactorrepresentsthe amountthat the varianceof the
estimatoris inflateddue to the presenceof multicollinearity.Fisherand Masonsuggestthat "a good ruleof
thumbregarding varianceinflationfactorsis thatvaluesabove4.0 (i.e.,Rj2 > .75)indicatethatthecorresponding
predictorvariableis involvedin collinearitythat may haveseriouseffects on the OLS regressioncoefficient
estimatorsand their standarderrors.Valuesbelow 4.0 are not consideredsevere"(1981:109).
Televisionand ViolentCrime 225
TABLE 2

VarianceInflationFactors
Matrixa
Variables II
Exposureto TV Violence 2.04 1.58
Pct 18-34 2.98 1.51
Males/Females 2.11 1.56
Population(In) 1.98 1.75
Pop per sq. mi. (In) 1.99 1.77
Pct Black 2.00
Pct Less than h.s. edu. 3.52
GiniCoefficient 3.10
AFDCPayments 3.56
Pct Poor 3.45
South 4.08
Race/Region/SESIndex ... 1.51
Note:
a. MatrixI includes the originallist of independentvariables. MatrixIIcontains a composite index in place of the
stronglyinter-relatedpredictors.See text for explanation.

Varianceinflation factors based on the matrix for the full list of independent variablesare reported
in Column I of Table 2. Applying Fisher and Mason's rule of thumb, the variable "South"is clearly
implicated in severe collinearity (variance inflation factor = 4.08). Several other variables also yield
inflation values that are large enough to arouse concern, i.e., values of approximately 3.5. It thus
seems advisable to pursue strategies for reducing levels of multicollinearity.
One useful technique for dealing with multicollinearity is to combine highly correlatedpredictors
into composite indexes (see Weisberg,1980:185).1oAccordingly,a principalcomponents factor analysis
was performed on the "control"variables.The resultsof the factor analysis indicate that the following
six predictors all load highly on a single dimension: percent black, percent less than high school
education, the Gini coefficient, AFDC payments, percent poor, and South. A composite index
was constructed using factor score coefficients for these six predictors.Although there is no obvious
substantiveinterpretationfor the dimension underlying the index, it can be labeled a "Race/Region/
SES index" for shorthand. The somewhat ambiguous meaning of the factor index is not especially
problematic in the present analysis because the primary purpose is not to interpret the observed
relationships for the index but simply to control for confounding effects of a range of relevant
variables in the assessment of the television violence variable.
A second set of variance inflation factors was computed based on a matrix with the composite
index included in place of the index's constituent variables. The resulting values are presented in
Column II of Table 2. The substitution of the index for the component terms greatly lessens the
problem of multicollinearity. Indeed, none of the predictors in the reduced matrix approaches a

10. Anotherprocedurefor addressingthe problemof multicollinearity, suggestedby Hanushekand Jackson


(1977:90),is to experiment andobservethe fluctuationsin thecorresponding
withslightlydifferentspecifications
importantfluctuationsareproducedby minorchangesin the model,littleconfi-
coefficients.If substantively
dencecanbe placedin the parameterestimates.Twoalternativespecificationsbasedon subsetsof the indepen-
dentvariableshavebeenexaminedin additionto the modelsdiscussedin the text. Specifically,the threepairs
of independentvariableswith bivariatecorrelationsgreaterthan .60 havebeen identified,and one variable
in eachof thesepairshas beenincludedin a modelwhilethe otherhas beenexcluded.The procedureis then
reversedfor a secondspecification.The regressionresultsfor thesealternativespecificationsare fully consis-
tentwiththosepresentedin the text.Themeasureof exposureof televisionviolenceexhibitsnegativerelation-
ships will all measuresof violent crimeacrossboth specifications.Nine of the ten negativecoefficientsfor
televisionviolence(i.e.two specificationsX five crimemeasures)are statisticallysignificantat the .05 level.
226 MESSNER
TABLE3

Regression Results for ViolentCrimeRates, TVViolenceExposure,


and Controlswith Composite Index (N= 281)a
DependentVariables
Violent
Criminal Forcible Aggravated Crime
IndependentVariables Homicide Rape Robbery Assault Index
Exposureto TV Violence -.17* -.24* -.21* -.20* -.25*
Pct 18-34 .01 .16* .03 .01 .03
Males/Females - .02 .04 - .02 .06 .03
Population(In) .37* .46* .54* .21* .45*
Pop per sq. mi. (In) -.07 -.16* .21* -.03 .08
Race/Region/SESIndex .70* .41* .29* .51* .49*
Adj. R2 .46 .26 .54 .20 .40
Notes:
a. Standardizedregression coefficientsare reported.
* p < .05

critical level. Regression results based on the reduced set of predictors should not therefore be seri-
ously affected by multicollinearity.
The effects of exposure to television violence were re-estimated using the composite factor index
in place of the full list of controls (see Table 3). The results for the television measure are highly
similar to those reported previously. Levels of exposure to television violence are inversely related
to each of the five measures of violent crime. Moreover, these negative relationships are now in
every instance statistically significant at conventional levels.
An analysis of residuals has also been performed to determine whether or not the surprising
effects of the television measure might be due to the presence of a few atypical observations. An
examination of standardized and studentized residuals indicates that for each offense, the case
with the largest residual is one with a predicted crime rate considerably lower than the actual rate.
For example, the observed murder rate for Miami of 34.5 greatly exceeds the predicted rate of 17.2.
Similarly, New York City's robbery rate of 1212.5 is substantially larger than the predicted value
of 572.7. This pattern probably reflects the slight skewness in crime rates across the SMSA sample.
To lessen the impact of the cases with relatively extreme values on the dependent variables, the
crime rates were transformed by taking natural logarithms, and the regression equations depicted
in Table 3 were then re-estimated. The substantively important conclusions remain unchanged. The
measure of exposure to television violence is negatively and significantly related to each of the
five log-transformed crime rates.
The influence of special cases on the regression results was also assessed by computing measures
of Cook's Di for each case (Cook and Weisberg, 1982; Weisberg, 1980)." The computed Cook's
Di's for the equations in Table 3 imply that there are no extremely influential cases affecting the
regression estimates. All of the observed values of Di are below .46. Nevertheless, each regression
equation was re-estimated with the single most influential case (as indicated by Di) deleted. The
resulting coefficients for the television violence measure are virtually the same as those obtained
for the entire sample.
An additional question which might be raised about these findings concerns the appropriateness
of focusing on the exposure levels for total populations. The overall index of television exposure

11. Thisstatisticindicatesthe extentto whicha particularobservationinfluencesthe observedparameteresti-


mates.A largevalueof Di impliesthatthe regressionestimatesdependheavilyon a givencaseandthatentirely
differentresultsmightbe obtainedin the absenceof that case. Weisbergsuggeststhat valuesof Di of about
1.0 or largermay be judgedas largeand hence highlyinfluential(1980:108).
Televisionand ViolentCrime 227
reflects the viewing behavior of the entire population, including segments of the population that
are not very likely to be involved in the kind of behavior under investigation. Therefore, it might
be instructive to examine levels of exposure for a subsection of the population at especially high
risk of involvement in violent crime.
The available data permit the construction of an index of exposure to television violence for
one particularly "high risk"population, namely, males aged 18-34.12 An index of violence exposure
for this group was constructed in a manner similar to that used for the overall index. The new
index represents the average relative size of the male population aged 18-34 viewing the selected
violent programs during the time period under consideration. This age-sex-specific index of televi-
sion violence exposure was entered into the regression equations in place of the overall television
index. The results (not shown) are highly similar to those observed for the overall index. The age-
sex-specific measure of exposure to television violence is negatively and significantly related to
all rates of violent crime.
Another possible explanation for these unexpected findings is that they are due to reportingbiases
associated with the official statistics on crime. As suggested earlier, residents in communities with
high levels of exposure to television violence might be less likely than those in communities with
lower levels of exposure to report criminal incidents to the police. Systematic under-reporting of
this kind could conceivably be the consequence of a "desensitization"to violence, or it could reflect
high levels of fear and interpersonal mistrust which have been linked to television viewing in past
research (Gerbner et al., 1978, 1979, 1980). The effect of such under-reporting would be to make
it appear as if there is an inverse relationship between television violence and rates of crime even
though the "true" relationship could very well be in the opposite direction.
However, evidence in these data diminishes the plausibility of this "under-reporting"hypothesis.
Criminologistsgenerallyagreethat the official statisticsfor homicide arenot susceptibleto appreciable
reporting biases - the number of officially recorded homicides is probably a fair indication of the
actual volume of these offenses (Sellin, 1951;Wolfgang, 1970). Therefore, if the under-reporting
argument were correct, the results for homicide would diverge appreciably from those for the other
offenses which are subject to these reporting biases. The findings for homicide are in fact very
consistent with those for the other types of crime. This makes it unlikely that the suprising inverse
associations for the television measure are merely "artifacts"of the well-known reporting biases
associated with official crime statistics.13
One final issue that needs to be addressed concerns the measurement of exposure to television
violence. The television violence index was based on the audience size for a small sampling of
programs- the five most violent, regulartelevision series.The viewingpatternsfor these five programs
may not fully reflectexposureto violent televisioncontent more generally.Ideally,it would be desirable
to construct indexes based on a larger sampling of programs, but this is not feasible in the present
study because of the need to rely on the secondary data source (NCTV content analyses) described
above.
It is possible nevertheless to make an indirect assessment of the plausibility of the violence index
by means of an analysis of an additional television measure that is likely to be correlated with
violence exposure. Researchat the individual level has revealeda strong, positive association between

12. Hindelang(1981)documentedthe highoffendingratesof youngmalesand showedthat the demographic


correlates of offending are very similar across official police reports and victimization surveys. Hindelang's
data also indicate that young, black, males are characterized by especially high offending rates. Unfortunately,
it is not possible to construct a television viewing measure broken down by age, sex, and race with available data.
13. There is one noteworthy difference in the pattern of results for homicide in comparison with those for
the other violent offenses. The zero order correlation between homicide and the television violence measure
is significantly positive. The surprising inverse relationship appears only after the effects of other characteris-
tics of SMSAs are controlled. This pattern suggests that there may be certain features of the interrelationships
between homicide and other structural characteristics that are unique to this particular offense.
228 MESSNER
exposure to violent television content and total exposure to television programming (Milavsky,
1982:82). In other words, persons who are heavy television viewers tend to be frequent viewers of
violent television. If this pattern is generalized from the micro-level to the macro-level, two predic-
tions can be derived: (1) a measure of total levels of exposure to television programming for popu-
lation aggregates should be strongly correlated with the measure of exposure to violent television
content; and (2) the effects of the total exposure on rates of violent crime should be similar to
those for the measure of exposure to television violence.
A measure of total television exposurewas computed by combining averagequarter-houraudience
estimates for the three networks combined for each of the three major "dayparts":Monday-Friday
10:00A.M.-5:00P.M..;Sunday-Saturday8:00-11:00P.M.;and Saturdaymorning 7:00-noon. The scores
on this measure indicate relative audience size attracted to any network shows, during an average
15 minute interval,over the course of a television programmingday. The correlationbetween exposure
to television violence and this new measure (total exposure to television programming) is strongly
positive (r = .79). In other words, population aggregates in which large audiences are attracted to
violent television content tend to be those in which large audiences are attractedto television program-
ming more generally. Regression analyses also were performed with the total television exposure
measure. With the exception of homicide, total exposure to television programming is consistently
associated with rates of crime in an inverse direction (see Table 4). The consistency in the results
across the two measures- which are expected on the basis of independent evidence to be strongly
inter-related- lends additional credibility to the television violence index.14

DISCUSSION
The results of the analysis clearly raise questions about generalizing directly from previous
research on television violence and mild forms of aggression at the micro-level to more serious
forms of aggression such as criminal violence at the level of population aggregates. Contrary to
what might be expected on the basis of previous research, there is no evidence in the aggregate
data to support the claim that high levels of exposure to television violence are related to high
rates of violent crime. In fact, the relationship that does emerge is in the opposite direction. The
data consistently indicate that high levels of exposure to violent television content are accompanied
by relatively low rates of violent crime.
Certain limitations of the present study require that these empirical findings be interpreted
cautiously. First the analysis has been restricted to the programming of the three major networks.
It is possible that the most violent programs currently being broadcast are not those carried by
network affiliates but rather those appearing on independent stations and on cable stations. If
so, exposure levels to these "very violent" programs might conceivably be related to violent crime
rates in the expected positive direction. Further research dealing with a wider range of program
outlets is clearly needed.
A related limitation of the present study is the use of a brief period for the sampling of audience
size. Viewing behavior has been estimated over the course of a four-week interval. The interval

14. Additionalanalyseswereperformedto addressthe possibilitythat the unexpectedfindingof an inverse


relationshipbetweentelevisionviolenceandratesof violentcrimeis in somewayan artifactof havingmeasured
television viewing at one level of aggregation (the DMA level) and the other variables at a slightly different
level of aggregation (SMSA level). To explore this possibility, a parallel DMA file was created. Creation of
this file involved locating all SMSAs that fall within a larger DMA and aggregating the SMSA-level measures
in accordance with DMA boundaries. The result of this procedure is a file for which all variables are measured
at the DMA level. The findings for DMAs prove to be highly similar to those for SMSAs. The various measures
of television viewing (i.e. exposure to television violence, exposure of males 18-34 to television violence, and
total television exposure) consistently exhibit negative relationships with all rates of violent crime in the regres-
sion analyses. Only the results for SMSAs are presented here because of space considerations. Means, standard
deviations, and correlations for DMAs are provided in the Appendices along with those of SMSAs.
Televisionand ViolentCrime 229
TABLE 4

Regression Results for ViolentCrimeRates, TotalTVExposure,


and Controlswith Composite Index (N= 281)a
DependentVariables
Violent
Criminal Forcible Aggravated Crime
IndependentVariables Homicide Rape Robbery Assault Index
Total TV Exposure -.03 -.12b -.11 * -.13* -.14
Pct 18-34 - .01 .14* .01 .00 .01
Males/Females .00 .06 .00 .08 .05
Population(In) .39* .47* .55* .21* .45*
Pop per sq. mi. (In) -.06 -.15* .22* -.02 .09
Race/Region/SESIndex .61* .33* .22* .46* .42"
Adj. R2 .44 .23 .52 .19 .37
Notes:
a. Standardizedregressioncoefficients are reported.
b. Estimateborderson statisticalsignificance(t = - 1.92, p = .056).
* p < .05

selected - November - is a strategic one because it occurs at the heart of the fall television season,
but viewing patterns for November might not necessarily be representativeof viewing more gener-
ally. The development of measures of aggregate viewing patterns that are based on a more compre-
hensive sampling of television programming is also an important problem for future research.
Finally, it should be stressed that the analysis presented above is cross-sectional in nature. A
particularly promising extension of the research design would be to collect data on audience size
for population aggregates at multiple points in time. Longitudinal data of this sort would permit
the use of statistical techniques such as panel regression analysis and would also allow for an
assessment of the extent to which these findings reflect any unique, historical factors operating
during the time period under investigation - factors that might affect television viewing and crime
rates simultaneously"5(see Milavsky et al., 1982). The extension of the design in this manner is
very straightforwardin principle, but the practical difficulties are considerable. The major problem
is that, for every time point, data on audience size must be combined with content analyses for
samples of programs. Given the fact that the programs being broadcast are likely to change over
time, the development of temporally comparable indexes would be a major task. The collection
of comparable longitudinal data would nevertheless permit an assessment of the historical speci-
ficity of the results and also allow for a closer examination of the direction of causal influences.
However, it is interesting to note that the finding of an inverse relationship between television
violence and violent crime is rather surprising irrespective of the nature of the causal ordering.
Despite the limitations, the consistency of the empirical findings is quite impressive. Negative
associations are observed for each of the four offenses taken singly and for the combined index
of violent crime. Similar relationships are observed for indexes based on both the total population
and a strategic subgroup (i.e. males 18-34). The results are virtually identical for a variety of
specifications and for selected subsamples. Finally, the pattern of inverse associations is replicated
with an additional measure- total television exposure - which is expected to be highly correlated
with exposure to violent television content.

15. Withrespectto the issueof historicalspecificity,it is interestingto note the virtuallyidenticalresultsare


obtainedwhena two yearaverageof the crimeratesis usedin placeof the datafor a singleyear.It thusseems
unlikelythatthe observedrelationshipscan be attributedto a single,historicalepisodewitheffectsof a short
duration.
230 MESSNER
The question that naturally arises is whether or not these anomalous findings can be placed
within a meaningful theoretical context. Why should high levels of exposure to television violence
be associated with low rates of violent crime?
"Catharsis theory" offers one possible explanation. According to catharsis theory, the viewing
of violence on film and on television allows for the vicarious release of aggressive impulses and
in so doing reduces the likelihood of overt expressions of aggression. Catharsis theory thus implies
that there will be the kind of inverse relationship between media violence and rates of violent crime
that has in fact been observed in the empirical analysis. The main problem with catharsis theory
is that there is little support for this theoretical perspective in micro-level research. An early study
by Feshbach and Singer (1971) reported findings consistent with this perspective, but most studies
have not detected any evidence of cathartic processes.
A rather different theoretical account is the subcultural argument proposed by Howitt and
colleagues (Howitt and Cumberbatch, 1965; Howitt and Dembo, 1974). Howitt et al. draw upon
a long standing tradition in criminology and argue that criminal behavior results from socializa-
tion into delinquent or criminal subcultures. Further, they suggest that this socialization occurs
primarily by means of associations with criminally inclined peers. Therefore, the effect of a given
mass medium on the probability of criminal behavior will depend on whether or not that medium
promotes associations with criminallyinclined peersat the expense of associations with more "conven-
tional others."
Howitt and Cumberbatch observe that television is not the kind of medium that encourages
extensive associations with delinquent or criminal peers. In their words, television viewing "is an
activity laigely confined to the home. The child for whom his peers are important would scarcely
be expected to watch television to excess - his taste would be to be out and about with his friends"
(Howitt and Cumberbatch,1975:111).The clearimplication of this argumentis that televisionviewing,
by keeping youths at home, decreases opportunities for socialization into criminal subcultures,
and in so doing, reducesthe likelihood of criminal behavior. Moreover,since total television viewing
and the viewing of television violence are highly intercorrelated,the same sort of relationship with
indicators of crime is expected for both of these measures of television viewing behavior.
A slightly different theoretical interpretation for the surprising inverse associations between tele-
vision exposure and rates of violent crime can be formulated by means of an application of the
"routine activities approach" developed by Cohen and Felson (1979). Cohen and Felson argue that
aggregatecrime ratesreflectthe probabilitythat motivated offenders, suitabletargets,and the absence
of capable guardians convergein space and time. This probability will depend on the "routineactivi-
ties" in which people are engaged. Cohen and Felson observed that "routine activities" centered
around the household and the family involve a relativelylow risk of criminalvictimization (1979:594).
The reason for this low victimization risk is that family and household activities are unlikely to
promote the convergencein space and time of the three requisites for a criminal incident - offender,
target, no guardian.
Television viewing is by its very nature "an activity largely confined to the home." Hence, high
levels of television viewing imply that residents are spending large amounts of time within the rela-
tively safe confines of the household. The observed inverse relationship between television viewing
and rates of crime is thereforequite understandablegiven the logic of the routine activities approach.
Crime rates are low in population aggregates characterized by high levels of exposure to television
programmingbecause activities in these areas are concentrated around the household, and because
the concentration of activities around the household reducesthe available opportunities for criminal
victimization.
The routine activities interpretation also implies that the effects of television viewing on crime
rates should not be restrictedto violent crime but should extend to property crimes as well. Indeed,
the theory would seem to suggest especially strong relationships for burglary, an offense that is
Televisionand ViolentCrime 231
TABLE 5

Regression Results for PropertyCrimeRates, TVViolenceExposure,


or TotalTVExposure,and Controls(N = 281)a
DependentVariables
IndependentVariables Burglary Larceny AutoTheft
TV Violence Exposure - .35 ..... - .31* ..... -.25*
Total TV Exposure ..... - .28* ..... - .24* ..... - .15
Pct 18-34 .09 .06 .41* .38* .07 .05
Males/Females - .05 - .02 - .26* - .24* - .05 - .03
Population(In) .38* .38* .20* .19* .39* .40*
Pop per sq. mi. (In) -.07 -.05 -.19* -.17* .28* .29*
Race/Region/SESIndex .42* .35* .11 .05 .27* .20*
Adj. R2 .25 .23 .17 .15 .45 .42
Notes:
a. Standardizedregressioncoefficients are reported.
* p < .05

unlikely to occur when people are at home serving as "capable guardians."To address these predic-
tions, I estimated regression equations for the rates of major property crimes with measures of
both television violence exposure and total television exposure. The results in Table 5 are quite
striking. Both measures of television viewing yield negative and statistically significant estimates
for all three of the major types of property crimes, with especially strong effects for burglary.
Admittedly, I have applied the subcultural theory and the routine activities account in an "ex
post facto" manner. Consequently, neither has been tested in any formal sense. Nevertheless, these
two theories provide plausible explanations for resultsthat are very surprisingyet remarkablyconsis-
tent across a variety of analyses. Irrespective of the cogency of these theoretical arguments, the
evidence presented here clearly raises serious questions about the claim that high rates of urban
crime can be attributed in any simple and direct way to heavy exposure to television violence.

REFERENCES
AndisonF. Scott
1977 "TVviolence and vieweraggression:a cumulationof study results 1956-1976."Public Opinion Quarterly
41:314-31.
Bandura, Albert
1965 "Influence of model's reinforcementcontingencies on the acquisition of imitative responses."Journal
of Personality and Social Psychology 1:589-95.
1973 Agression: A Social Learning Analysis. Englewood Cliffs, NJ: Prentice-Hall.
Belson, William A.
1978 Television Violence and the Adolescent Boy. London: Saxon House.
Berkowitz, and Joseph T. Alioto
1973 "The meaning of an observed event as a determinant of its aggressive consequences." Journal of
Personality and Social Psychology 28:206-17.
Berkowitz, Leonard and Edna Rawlings
1963 "Effects of film violence on inhibitions against subsequent aggression." Journal of Abnormal and
Social Psychology 66:405-12
Blau, JudithR. and PeterM. Blau
1982 "The cost of inequality: metropolitan structure and violent crime." American Sociological Review
47:114-29.
Bollen, Kenneth A. and David P. Phillips
1981 "Suicidalmotor vehicle fatalities in Detroit: a replication."American Journal of Sociology 87:404-12.
Braithwaite, John
1979 Inequality, Crime, and Public Policy. London: Routledge and Kegan Paul.
Clark, David G. and William Blackenburg
1972 "Trendsin violent content in selected mass media." Pp. 188-243 in George A. Comstock and Eli
Rubinstein (eds.), Television and Social Behavior, Vol. I. Washington, DC: U.S. Government Printing
Office.
232 MESSNER
Cohen, Lawrence E. and Marcus Felson
1979 "Social change and crime rate trends: a routine activity approach." American Sociological Review
44:588-608.
Comstock, George A.
1975 Television and Human Behavior. The Key Studies. Santa Monica, CA: Rand.
1980 Television in America. Beverly Hills, CA: Sage.
Cook, R. Dennis and Sanford Weisberg
1982 "Criticismand influence analysis in regression."Pp. 313-61 in Samuel Leinhardt (ed.), Sociological
Methodology 1982. San Francisco: Jossey-Bass.
De Fronzo, James
1983 "Economic assistance to impoverished Americans: relationship to incidence of crime."Criminology
21:119-36
Eron, Leonard D.
1982 "Parent-child interaction, television violence, and aggression of children." American Psychologist
37:197-211
Eron, Leonard D., L. Rowell Huesmann, Patrick Brice, Paulette Fischer, and Rebecca Mermelstein
1983 "Age trends in the development of aggression, sex typing, and related television habits." Develop-
mental Psychology. 19:71-77.
Eysenck, Hans and D.K.B. Nias
1978 Sex, Violence, and the Media. New York: Harper and Row.
Federal Bureau of Investigation
1982 Uniform Crime Reports for the United States- 1981. Washington, DC: U.S. Government Printing
Office.
Feshbach, Seymour and R. Singer
1971 Television and Aggression. San Francisco, CA: Jossey-Bass.
Fisher, Joseph C. and Robert L. Mason
1981 "The analysis of multicollinear data in criminology." Pp. 99-125 in James A. Fox (ed.), Methods
in Quantitative Criminology. NY: Academic.
Gastil, Raymond
1971 "Homicide and a regional culture of violence." American Sociological Review 36:412-27.
Gerbner, George, Larry Gross, Marilyn Jackson-Beeck, Suzanne Jeffries-Fox, and Nancy Signorielli
1978 "Cultural indicators: violence profile no. 9." Journal of Communication 28:176-207.
Gerbner, George, Larry Gross, Nancy Signorielli, Michael Morgan, and Marilyn Jackson-Beeck
1979 "The demonstration of power: violence profile no. 10." Journal of Communication 29:177-96.
Gerbner, George, Larry Gross, Michael Morgan, and Nancy Signorielli
1980 "The mainstreaming of America: violence profile no. 11."Journal of Communication 30:10-29.
Gibbons, Don
1982 Society, Crime, and Criminal Behavior. Fourth Edition. Englewood Cliffs, NJ: Prentice-Hall.
Greenberg, Bradley S. and Thomas F. Gordon
1972 "Perceptions of violence in television programs: critics and the public." Pp. 244-58 in George A.
Comstock and Eli Rubinstein (eds.), Television and Social Behavior, Vol. I. Washington, DC: U.S.
Government Printing Office.
Hackney, Sheldon
1969 "Southern violence." American Historical Review 74:906-25.
Halloran, James
1978 "Studying violence and the media: a sociological approach." Pp. 287-305 in Charles Winick (ed.),
Deviance and Mass Media. Beverly Hills, CA: Sage.
Hanushek, Eric A. and John E. Jackson
1977 Statistical Methods for Social Scientists. NY: Academic.
Hennigan, KarenM., MarilynL. Del Rosario,Linda Heath, Thomas D. Cook, J. D. Wharton,and Bobby J. Calder
1982 "Impact of television on crime in the United States: empirical findings and theoretical implications."
Journal of Personality and Social Psychology 42:461-77.
Himmelweit, Hilda T., A. N. Oppenheim, and Pamela Vance
1958 Television and the Child. London: Oxford University Press.
Hindelang, Michael
1974 "The uniform crime reports revisited." Journal of Criminal Justice 2:1-17.
1981 "Variations in sex-race-age-specific incidence rates of offending." American Sociological Review
46:461-74.
Howitt, Dennis and Guy Cumberbatch
1975 Mass Media Violence and Society. New York: Wiley.
Howitt, Dennis and Richard Dembo
1974 "A subcultural account of media effects." Human Relations 27:25-41
Kaplan, Robert M. and Robert D. Singer
1976 "Televisionviolence and vieweraggression: a reexaminationof the evidence."Journal of Social Issues
32:35-70.
Televisionand ViolentCrime 233
Krohn, Marvin
1976 "Inequality,unemployment, and crime: a cross-national analysis."Sociological Quarterly 17:303-13.
Leyens, Jacque-Philippe P., Leoncio Camino, Ross D. Parke, and Leonard Berkowitz
1975 "Effects of movie violence on aggression in a field setting as a function of group dominance and
cohesion." Journal of Personality and Social Psychology 32:346-60.
Loftin, Colin and Robert H. Hill
1974 "Regional subculture and homicide: an examination of the Gastil-Hackney thesis." American Socio-
logical Review 39:714-24.
Mayhew, Bruce H. and Roger L. Levinger
1976 "Size and the density of interaction in human aggregates."American Journal of Sociology 82:86-110.
McCarthy, Elizabeth D., Thomas S. Langner, Joanne C. Gersten, Jeanne G. Eisenberg, and Lida Orzeck
1975a "Violence and behavior disorders." Journal of Communication 25:71-85.
McCarthy, John D., Omar R. Galle, and William Zimmern
1975b "Populationdensity,social structure,and interpersonalviolence: an intermetropolitantest of competing
models." American Behavioral Scientist 18:771-89.
McDonald, Lynn
1976 The Sociology of Law and Order. Boulder, CO: Westview.
McIntyre, Jennie J. and James Teevan
1972 "Televisionviolence and deviant behavior." Pp. 383-435 in George A. Comstock and Eli Rubinstein
(eds.), Television and Social Behavior, Vol. III. Washington, DC: U.S. Government Printing Office.
Milavsky, J. Ronald, Ronald C. Kessler, Horst H. Stipp, and William S. Rubens
1982 Television and Aggression: A Panel Study. New York: Academic Press.
Milgram, Stanley and R. Lance Shotland
1973 Televison and Anti-Social Behavior: Field Experiments. NY: Academic Press.
National Coalition on Television Violence
1981 NCTV News. 2(January).
National Institute of Mental Health
1982 Televisionand Behavior:TenYearsof Scientific Progressand Implicationsfor the Eighties. Washington,
DC: U.S. Government Printing Office.
New YorkDaily News
1982 "Televiolence perpetrators: don't blame us." May 30: 114.
Parke, Ross D., Leonard Berkowitz, J. Leyens, Stephen G. West, and Richard J. Sebastian
1977 "Some effects of violent and nonviolent movies on the behavior of juvenile delinquents." In Leonard
Berkowitz (ed.), Advances in Experimental Psychology, Vol. 10, NY: Academic.
Phillips, David
1974 "The influence of suggestion on suicide: substantive and theoretical implications of the Werther
effect." American Sociological Review 39:340-54.
1979 "Suicide,motor vehicle fatalities, and the mass media: evidence towarda theory of suggestion."Amer-
ican Journal of Sociology 84:1150-74.
1982 "The effect of fictional television stories on U.S. adult fatalities: new evidence on the effect of mass
media on violence." American Journal of Sociology 87:1340-59.
1983 "The impact of Mass Media Violence on U.S. Homicides." American Sociological Review 48:560-68.
Robinson, John P. and Jerald Bachman
1972 "Television viewing habits and aggression." Pp. 372-82 in George A. Comstock and Eli Rubinstein
(eds.), Television and Social Behavior, Vol. III. Washington, DC: U.S. Government Printing Office.
Savitz, Leonard
1978 "Official police statistics and their limitations." Pp. 69-81 in Leonard Savitz and Norman Johnston
(eds.), Crime in Society. New York: Wiley.
Schramm, Wilbur, Jack Lyle, and Edwin D. Parker
1961 Television in the Lives of Our Children. Stanford, CA: Stanford University Press.
Sellin, Thorsten
1951 "The significance of records of crime." Law Quarterly Review 67:489-504.
Tufte, Edward
1974 Data Analysis for Politics and Policy. Englewood Cliffs, NJ: Prentice-Hall.
U.S. Bureau of the Census
1982a State and Metropolitan Area Data Book, 1982. Washington, DC: U.S. Government Printing Office.
1982b Census of the Population and Housing, 1980:Summary Tape File #3C. Washington DC: U.S. Bureau
of the Census.
Weisberg, Sanford
1980 Applied Linear Regression. NY: Wiley.
Wolfgang, Marvin
1970 "A sociological analysis of criminal homicide." Pp. 52-60 in Bruce J. Cohen (ed.), Crime in America:
Perspectives on Criminal and Delinquent Behavior. Itasca, IL: F. E. Peacock.
234 MESSNER

APPENDIXA
Basic Statistics
SMSAs DMAs
(N= 281) (N= 159)
Variables x S.D. x S.D.
(Y1) CriminalHomicide 8.17 5.48 8.93 5.48
(Y2) ForcibleRape 35.56 17.81 38.26 17.36
(Y3) Robbery 160.95 142.56 178.72 136.43
(Y4) AggravatedAssault 281.76 152.14 288.68 146.56
(Y5) ViolentCrimeIndex 486.44 266.69 514.58 258.89
(TV1) Exposureto TV Violence 20.92 4.39 21.47 4.49
(TV2) Males 18-34 Exposureto TV Violence 13.59 3.47 14.06 3.54
(TV3) Total TV Exposure 29.07 3.37 29.44 3.58
(X1) Pct Black 10.19 9.75 11.09 10.41
(X2) Pct 18-34 30.97 4.74 31.28 3.92
(X3) Males/Females 95.22 5.30 95.64 7.94
(X4) Pct less than h.s. edu. 30.57 7.67 30.04 6.98
(X5) Population(In) 12.61 .98 13.16 1.08
(X6) GiniCoefficient .40 .02 .40 .02
(X7) AFDCPayments 236.53 91.10 223.35 90.17
(X8) Pop per sq. mi. (In) 5.41 .95 5.32 .85
(X9) Pct Poor 9.02 3.43 9.25 3.55
(X10) South .41 .49 .45 .50
APPENDIX B

CorrelationMatrixa
(SMSAs above diagonal, DMAsbelow diagonal)
Y1 Y2 Y3 Y4 Y5 TV1 TV2 TV3 X1 X2 X3 X4
Y1 - .56 .59 .51 .66 .16 .14 .18 .60 - .08 - .07 .36
Y2 .61 - .54 .64 .73 -.07 -.07 -.05 .39 .12 .10 -.12
Y3 .60 .55 - .44 .83 -.18 -.24 -.14 .39 -.11 -.14 .01
Y4 .52 .67 .47 - .86 .05 .02 .04 .40 - .01 .04 .15
Y5 .67 .75 .84 .87 - -.07 -.12 -.05 .47 -.06 -.04 .09
TVi .15 -.06 -.19 .06 -.07 - .84 .79 .46 .04 -.05 .41 -
TV2 .06 -.14 -.31 -.06 -.20 .81 - .65 .35 .00 -.03 .37 -
TV3 .20 -.06 -.16 .07 -.05 .81 .66 - .50 .00 -.05 .36 -
X1 .63 .41 .39 .42 .48 .50 .31 .53 - .02 - .05 .35
X2 -.10 .25 -.06 .11 .04 .05 .00 -.02 .09 - .57 -.55 -
X3 -.05 .01 -.03 .05 .01 .01 .03 .05 -.03 .35 - -.27 -
X4 .42 -.13 -.04 .10 .04 .44 .38 .45 .35 -.50 .35 - -
X5 .22 .21 .64 .17 .45 -.27 -.38 -.27 .12 -.18 -.09 -.04
X6 .60 .43 .23 .45 .42 .24 .16 .16 .42 .04 -.15 .42 -
X7 - .50 - .21 .04 - .30 -.17 - .56 - .51 - .52 - .53 - .05 .03 - .52
X8 .16 .10 .54 .14 .38 - .03 -.17 - .03 .28 -.12 - .09 .04
X9 .46 .16 .05 .29 .21 .43 .28 .38 .44 - .04 .00 .68 -
X10 .53 .27 .06 .38 .28 .52 .43 .46 .62 .03 - .03 .55 -
Note:
a. See AppendixA for variablekey.

You might also like