Content uploaded by Andrew L. Whitehead
Author content
All content in this area was uploaded by Andrew L. Whitehead on Feb 01, 2018
Content may be subject to copyright.
Forthcoming at The Journal of Sex Research
Unbuckling the Bible Belt:
A State-level Analysis of Religious Factors and Google Searches for Porn
Andrew L. Whitehead
Department of Sociology and Anthropology
Clemson University
Samuel L. Perry
Department of Sociology
University of Oklahoma
ABSTRACT
While the link between individual religious characteristics and pornography consumption is well
established, relatively little research has considered how the wider religious context may
influence pornography use. Exceptions in the literature to date have relied on relatively broad,
subjective measures of religious commitment, largely ignoring issues of religious belonging,
belief, or practice. This study moves the conversation forward by examining how a variety of
state-level religious factors predict Google searches for the term “porn,” net of relevant
sociodemographic and ideological controls. Our multivariate findings indicate that higher
percentages of evangelical Protestants, theists, and Biblical literalists in a state predict higher
frequencies of searching for “porn,” as do higher church attendance rates. Conversely, higher
percentages of religiously unaffiliated persons in a state predict lower frequencies of searching
for “porn.” Higher percentages of total religious adherents, Catholics, or Mainline Protestants in
a state are unrelated to searching for “porn” with controls in place. Contrary to recent research,
our analyses also show that higher percentages of political conservatives in a state predicted
lower frequencies of “porn” searches. Our findings support theories that more salient, traditional
religious influences in a state may influence residents—whether religious or not—toward more
covert sexual experiences.
Key words: pornography, religion, evangelicals, state-level data, political conservatism
1
INTRODUCTION
Religious beliefs, practices, and communities have always been important factors to
consider in studies of pornography
1
consumption (Davis & Braucht, 1976; Grubbs, Exine,
Pargament, Hook, & Carlisle, 2015; Perry, 2016a; Perry & Hayward, 2017; Wilson & Abelson,
1973). Religions typically teach that the only morally appropriate place for sexual desires and
behavior is monogamous, married, heterosexual relationships. Consequently, religious groups,
and most prominently conservative Protestants in the American context, strongly discourage any
type of pornography, viewing it as a form of fornication that rouses and facilitates sexual desires
about persons outside of marriage and encourages solo-masturbation (Driscoll, 2009; Sherkat &
Ellison, 1997). It is unsurprising, then, that studies of religion’s relationship to porn viewing
among American adults consistently find that those who report greater religious commitment
(measured in a variety of ways) or who hold theologically conservative identities and beliefs are
more likely to report either not viewing pornography at all (Doran & Price, 2014; Grubbs et al.,
2015; Maddox, Rhoades, & Markman, 2011; Nelson, Padilla-Walker, & Carroll, 2010; Patterson
& Price, 2012; Perry, 2016b; Stack, Wasserman, & Kern., 2004; Wright, 2013; Wright, Bae, &
Funk, 2013) or doing so less frequently than others (Baltazar, Helm, McBride, Hopkins, &
Stevens, 2010; Bridges & Morokoff, 2011; Carroll, Padilla-Walker, Nelson, Olson, Barry, &
Madsen, 2008; Hardy, Steelman, Coyne, & Ridge, 2013; Perry, 2016c; Poulsen, Busby, &
Galovan, 2013; Short, Kasper, & Wetterneck, 2015).
Yet while the connection between religious factors and porn consumption is well-
established at the individual level, relatively few studies have considered how the broader
1
The term “pornography” is difficult to define and is often replaced by other more descriptive terms like “sexually
explicit media/material.” However, the term pornography is still widely used, and because our study deals with the
Google search term “porn,” we use “pornography” or “porn” here. Throughout the study, pornography/porn will
refer to any sexually explicit media (videos, websites, magazines, etc.) intended to arouse the viewer.
2
religious context relates to pornography use. Among the exceptions, Edelman (2009) analyzed
state-level covariates of credit card subscriptions to a leading adult entertainment website from
2006-2008. He found that subscriptions were more prevalent in states where surveys indicate
more conservative stances on religion, sexuality, and gender roles. In their state-level analysis of
pornography and divorce across time, Daines and Shumway (2011) found that Playboy magazine
sales were strongly predictive of divorce rates, which are known to be higher in states with
higher percentages of conservative Protestants (Glass & Levchak, 2014). More recently,
MacInnis and Hodson (2015) found positive associations between aggregated state-level self-
identified religiosity and political conservativism and searching for sexual content online. The
authors aggregate individual-level responses to Gallup surveys of Americans to create state-level
measures. In order to operationalize the religious context of each state they create a state
religiosity index consisting of the percentage of respondents that (a) identify as “very religious”
and (b) consider religion an important part of their daily lives. While this religiosity index is
predictive in three of the fourteen multivariate models that control for other factors, it raises
several important questions that require additional investigation. Namely, what is the character of
that religiosity? Are particular religious groups more dominant in different areas? Which
religious beliefs dominate? Are the communities active religiously?
While self-identified religiosity and importance of religion give us some insight into
whether a community is religious, it tells us little about the nature of that religiosity. Two
communities could appear equally religious when aggregating individual self-reports of
religiosity, but at the same time exhibit divergent religious service attendance rates or different
views of how the Bible should be interpreted. This could lead those same communities to having
very different views toward various moral issues, like porn, for instance. Consistently, studies
3
find that the religious groups with which individuals affiliate, their religious beliefs, and their
religious behaviors routinely influence their views toward porn (Carroll et al., 2008; Lykke &
Cohen, 2015; Patterson & Price, 2012; Sherkat & Ellison, 1997) and various other moral issues
(Hoffman, Ellison, & Bartkowski, 2016; Perry, 2015; Whitehead & Perry, 2016). Therefore, it is
important to examine the content and character of religiosity at the state-level, rather than only
examining aggregated measures of how religious individuals perceive themselves to be.
To do so, we draw on the theoretical tradition of moral communities. The concept of
“moral communities” can be first attributed to Durkheim (1912/1995). In his various works he
showed how high degrees of consensus concerning community norms, as well as the religious
legitimation of those norms, significantly influence how those collectivities operate and the
actions members might take due to their context (Durkheim, 1897/1951). A number of
researchers draw on the concept of moral communities in their study of how human behavior is
shaped by their surrounding social environment (Baker, Smith, & Stoss, 2015; Gault-Sherman &
Draper, 2012; Hill, 2009; Lee & Barkowski, 2004; Stark, 1996; Ulmer, Bader, & Gault, 2008).
Instead of seeing religion only as an individual-level trait, this literature demonstrates that
religion is also a group property and a part of the social structure (Stark, 1996, p. 164). There is
evidence that the religious context—measured through religious affiliation, belief, or behavior—
of an area or state is significantly associated with a host of outcomes. The degree to which an
area contains a greater proportion of particular religious traditions can influence cohabitation
rates (Gault-Sherman & Draper, 2012), crime (Beyerlein & Hipp, 2005), gender attitudes (Moore
& Vanneman, 2003), same-sex policy outcomes (Scheitle & Hahn, 2011), and population health
(Blanchard et al., 2008). Strayhorn and Strayhorn (2009) find that teen birth rates are higher in
states where a larger percentage of residents report higher religiosity and more conservative
4
theological beliefs. And Baker and colleagues (2015) find that the percent of the population that
hold various religious beliefs or attend religious services frequently can predict a state’s posture
toward sex education.
MacInnis and Hodson (2015) draw on the preoccupation hypothesis to explain the
correlation between greater religiosity and political conservatism in a state and searches for sex-
related material. They theorize that religious persons and conservatives, perhaps due to
underlying personality factors like authoritarianism, may become preoccupied with the very
sexual content they oppose and thus pursue it covertly. While this explanation is certainly
plausible and may be true, it is not falsifiable using aggregate state-level data, as MacInnis and
Hodson (2015) point out. Indeed, it risks the ecological inference fallacy, in which inferences
about individuals are incorrectly drawn from observations of aggregate groups (Kingston &
Malamuth, 2011; Robinson, 1950; Schuessler, 1999). As an alternative, we propose that the
theoretical tradition of moral communities is more appropriate for understanding an analysis of
state-level data. An important strength of the moral communities framework is that it is less
susceptible to the ecological inference fallacy. The moral communities view posits that religion
itself is a group-level phenomenon, and thus, the theoretical link between group-level religious
characteristics and individual behavior does not require presumptions about the thought
processes or personalities of individual actors, as with the preoccupation hypothesis. Our claims
are not about which individual-level behaviors produce a particular group-level outcome, rather,
that religious contexts—moral communities—can explain some of the variation in aggregated
individual-level behaviors.
An additional strength of drawing on the moral communities thesis to understand the
religion-pornography connection at the aggregate level is that it does not necessarily negate
5
individual-level explanations—like the preoccupation hypothesis—utilized in prior literature
(MacInnis & Hodson, 2015). In fact, it allows that there may be a number of “different possible
relationships at the individual level [that] can generate the same observation at the aggregate
level” (Schuessler 1999, p. 10,578). Rather the moral communities thesis predicts significant
effects of population-level variation in religious beliefs, behaviors, and affiliations.
Therefore, in order to further contextualize and expand upon prior macro-level analyses
of religion and consumption of pornography, we hypothesize that the adherence rates to
particular religious traditions, the percent of the population that hold traditional religious beliefs,
and the average frequency with which people attend religious services will be significantly
associated with the consumption of porn at the state level. Broadly, we propose that the religious
nature of the social context surrounding people makes the consumption of online pornography
either more or less acceptable. These measures will serve as more precise indicators of the moral
communities in which people are embedded.
Beyond predicting that aggregate levels of religious affiliation, belief, and behavior are
significantly associated with the popularity of searching for pornography at the state level, we
propose two interpretations concerning the directionality of that association. In essence, religious
moral communities could serve to either encourage or suppress searching online for sexual
content. Each of these predictions draws upon prior research on religion and the consumption of
pornography. We freely acknowledge that we cannot establish the individual-level activity that
produces either of these interpretations. We can only demonstrate that a positive or negative
association exists. First, religion, as we operationalize it below with state-level measures, may
serve to suppress searching for sexual content online. A host of studies at the individual level
demonstrate that religious individuals—especially conservative Protestants, theists, those who
6
interpret the Bible literally, and those who attend religious services frequently—report lower
levels of porn consumption (Carroll et al., 2008; Perry, 2016b, 2016c; Stack et al., 2004; Wright,
2013; Wright et al., 2013). We would expect, then, that states with higher proportions of
religious individuals, who report lower consumption of porn, will have lower levels of porn
consumption as a whole.
Alternatively, MacInnis and Hodson (2015) found that the popularity of searching for
sexual content online was higher in states with higher proportions of people who identify as
“religious” or say religion is very important to them. Using more precise measures of religion,
we could find that the popularity of searching for sexual content online is higher in states with
greater proportions of citizens with particular religious affiliations, beliefs, and behaviors. This
could be due to two reasons. First, it may be that when asked directly religious individuals tend
to underreport their consumption of explicit material given the labeling of such content as
“sinful.” This explanation rests on the assumption that the responses of religious individuals on
anonymous surveys are strongly influenced by social desirability bias. Therefore, a strength of
examining aggregate levels of porn consumption and the religiosity of varying contexts is that
while certain religious markers may exhibit a relationship with porn consumption in one
direction at the individual level—largely due to measurement error stemming from social
desirability bias—at the group level the relationship could be completely opposite. In this sense,
the anonymized aggregate data could be picking up actual behaviors rather than self-reported
ones.
A second reason that searching for sexual content online could be more popular in more
religious states is that it may not be religious individuals accessing sexual content online more
frequently, but all citizens. The strong community-level norms and moral communities cultivated
7
within more religious populations may make online searches for pornography more prevalent
because there are few other outlets for sexual expression. Prior work on religious context and
divorce rates demonstrates how religious moral communities can influence the non-religious, as
well as the religious. Glass and Levchak (2014), for example, found that counties with higher
percentages of conservative Protestants predicted higher divorce rates regardless of one’s
personal religious identification. They reasoned that the broader religious culture contributed to
earlier ages at first marriage and lower collective educational attainment, strong predictors of
divorce. Therefore, even for non-religious people, being surrounded by a social environment that
is highly religious may make more intimate and private forms of sexuality ideal.
METHOD
In order to examine the state-level correlates of the popularity of Google searches for
pornography, we draw from a number of diverse and publically available sources.
Data and Measures
Outcome Variable
The dependent variable represents the popularity of “porn” as a Google web search term
for each of the 50 states from the period ranging from January 1, 2011 to July 31, 2016.
2
Google
Trends freely provides these data, and Google itself is the ideal search engine for this type of
analysis given that it fulfills two-thirds of all internet searches. Certainly, “porn” is not the only
Google search term that is used to access explicit material. Prior research (MacInnis & Hodson
2015) examined a number of different terms. We limit our analysis to “porn” alone for several
reasons. First, when tracking the popularity of the search terms “porn”, “lesbian porn”, “sex”,
2
We begin the search period at 1/1/2011 because on this date Google applied an improved algorithm for their
geographical assignment. We include the subsequent five and a half years in order to smooth the random variation
between years. In ancillary analyses, available upon request, we conducted analyses on each individual year
included in this range. Those findings do not differ in significant or substantive ways from those described below.
8
“sex tape”, “xxx”, “amateur porn”, and “free porn”, “porn” is far and away the most popular
term entered into Google to access explicit material. Over the entire period that we analyze,
“porn” accounts for almost twice as many searches as the second most searched term (“sex”),
and accounts for more searches than all of the terms mentioned above, combined. Second, most
likely due to its greater popularity, “porn” as a Google search term has a great deal of variation
across the 50 states while the less popular terms (“sex” and “xxx”) do not. For these other terms,
the distribution is highly skewed with one state (Oregon) serving as an extreme outlier with most
other states bunched together with much lower search scores. Subsequently, when we performed
multivariate analyses on these terms very few variables were significantly associated in any of
the models. We believe this is primarily due to a lack of variation to be explained. However,
when we examined the terms “sex tape”, “lesbian porn”, “amateur porn”, and “free porn” we
found that their distributions as well as their bivariate and multivariate results were substantively
identical to “porn” where the religion measures were consistently and significantly associated
with the dependent variable in the same direction (results available upon request). Third, we
believe an added benefit of concentrating on one particular term, and in this case by far the most
popular, allows for a more focused and parsimonious analysis.
Google Trends calculates values for the popularity of a specified search term for each
state on a scale from 0 to 100 over the specified time frame. The state assigned 100 is where the
fraction of the total searches for the specified search term is the largest. The scores for all other
states are computed in relation to the highest scoring state. A score of 50 is a state where the
fraction of the total searches for the specified search term is half as large as the state with the
highest proportion of searches. The strength of using Google trends is that it normalizes the
popularity of a search term across states. Because it is a proportion of all searches, more
9
populous states that generate higher volumes of total searches are placed on equal footing with
less populous states generating lower volumes of total searches. As prior research demonstrates,
Google Trends is an effective means through which public interest can be examined (MacInnis &
Hodson, 2015; Scheitle, 2011).
3
Independent Variables
In order to account for the relative sizes of various religious groups in each state, we
draw on publically available data collected by the Association of Statisticians of American
Religious Bodies (ASARB) made freely available at the Association of Religion Data Archives
(www.theARDA.com). Since 1990, ASARB has collected data every ten years on the religious
groups operating in the United States. In this study, we use data from the 2010 Religious
Congregations and Membership Study (RCMS). The 2010 RCMS provides information on 236
religious groups and the number of congregations and adherents within each state and county in
the United States. Using population totals, adherence rates for each religious group in each state
are estimated.
Drawing on this data we estimate the total religious adherence rates for each state as well
as three broad categorizations of Christian religious traditions. These are Evangelical Protestant,
Mainline Protestant, and Catholic. Each state-level religious adherence measure shows the
number of adherents per 1,000 population.
4
3
Unfortunately, Google Trends does not track information on whether terms were searched for using a computer,
smartphone, tablet, or some other electronic device. Knowing this information might provide some insight into the
demographics of the communities in which terms are being searched and how they consume porn.
4
In other studies examining the contextual effects of religious groups at the macro level (Gualt-Sherman & Draper,
2012; Ulmer et al., 2008) researchers utilize a Christian homogeneity/heterogeneity measure calculated from the
adherence rates of three Christian groups: Evangelical Protestants, Mainline Protestants, and Catholics. The index
varies between 0 and 1 where high scores equal more heterogamous populations in regards to these groups and low
equals more homogenous populations [H=1-([Evangelical adherence rate/total adherence rate]² + [Mainline
adherence rate/total adherence rate]² + [Catholic adherence rate/total adherence rate]²)]. In ancillary analyses,
available upon request, we calculated the H index. It was not significantly associated with the state-level popularity
of online searches for “porn” in either bivariate or multivariate analyses.
10
The final group are those who identify as no religious affiliation. This data is from the
2007 U.S. Religious Landscape Survey collected by the Pew Forum on Religion and Public Life
and is also made freely available on the ARDA. The 2007 U.S. Religious Landscape Survey is a
nationally representative sample of 35,556 adults in United States households. For each state we
create a percentage of the total respondents who identify with no particular religious affiliation.
In order to investigate the influence of religious beliefs and behaviors at the population
level, we estimate the percent of each state that believes in God, the percent that identify as a
biblical literalist, and a mean frequency of attendance measure for each state. Each of these state-
level measures are drawn from the 2007 U.S. Religious Landscape Survey discussed above.
Control Variables
We include a battery of control variables in the multivariate analyses. In individual
analyses on porn consumption, political ideology, income, education, age, and marital status are
all important variables (Perry, 2016b, 2016c; Stack et al., 2004; Wright, 2013; Wright et al.,
2013). Using the 2007 U.S. Religious Landscape Survey and American Community Studies’
estimates from 2008-2012, we include measures of political ideology, income, education, age,
and percent married of each state population. First, we control for the political context of each
state with the percent of the population who identify as politically “conservative” or “very
conservative”. Regarding income, we use median household income for each state. To measure
education we include the percent of each state population that is over age 25 and has a
Bachelor’s degree. We use the median age of each state and the percent married as the final
control variables. Each of the estimates drawn from the 2007 U.S. Religious Landscape Survey
and American Community Study are freely available from the ARDA.
5
5
In addition to the control variables that appear in the final models, we also examined a host of other state-level
measures in both bivariate and multivariate models. These alternative measures included: percent voting for Obama
11
Statistical Analysis
We first present descriptive statistics for the dependent variable, the independent
variables, and control variables. Also in Table 1, we provide the correlations between the state-
level popularity of Google web searches for “porn” and the independent variables of interest and
control variables. Next, in Figure 1 we display an array of bivariate scatter plots for those
independent variables of interest that are significantly correlated with the dependent variable. We
include linear trend lines, correlation coefficients, and levels of significance within each bivariate
scatter plot. In Tables 2 and 3, we employ Ordinary Least Squares (OLS) multivariate regression
to uncover whether the strong bivariate correlations between the various religion measures and
web searches for “porn” are robust. We estimate eight separate models for Total Religious
Adherence Rate, Evangelical Protestant Adherence Rate, Mainline Protestant Adherence Rate,
Catholic Adherence Rate, Percent Identifying as No Affiliation, Percent Theist, Percent Biblical
Literalist, and Mean Religious Service Attendance. Separate models are necessary because state-
level adherence rates across various religion measures are unavoidably collinear.
6
We also
display standardized coefficients in order to evaluate substantive significance instead of only
statistical significance. We performed a series of robustness checks to ensure all OLS regression
models were producing valid results and satisfied assumptions regarding error terms. Across all
models in Table 2 and 3, the error residuals are normally distributed, their mean is equal to zero,
and they are not correlated with any of the independent variables.
RESULTS
in 2012, percent voting for Obama in 2008, state income inequality, men’s median age at marriage, women’s median
age at marriage, percent male, percent female, percent male population under 18, percent male population 18-34,
percent male population 35-64, percent male population 65+, percent of population under 18, percent of population
18-34, percent of population 35-64, percent of population 65+, median male age, median female age, and percent
white. Bivariate results are available upon request.
6
Across all 8 models, no VIF scores exceed 3.48 with only 11 VIF scores (out of 48 total) exceeding 3.
12
In Table 1 we find that all 50 states range from a score of 60 to 100 on the popularity of
Google searches for “porn”. The mean score on this measure is 78.84. For perspective, a score of
50 represents half the proportion of searches for “porn” compared to the score of 100. The mean
total religious adherence rate in the US is almost 500 people per 1,000 population. The mean
Evangelical Protestant adherence rate is 160 people. Mainline Protestants are slightly less
populous with a mean adherence rate of 86. The mean Catholic adherence rate is 169. Finally,
across all 50 states about 13 percent of people identify as no particular religious affiliation. Over
nine in 10 Americans are theists, half interpret the Bible literally, and over one-third identify as
politically conservative. The average frequency of religious service attendance is close to once or
twice a month for many Americans.
[Table 1 about here]
Table 1 also displays the correlations between each of the independent variables of
interest and the popularity of “porn” as a Google search term. We can see that there is not a
significant correlation between the proportion of web searches for “porn” and the total religious
adherence rate for each of the 50 states. Similar to the total religious adherence rate, there
appears to be no underlying relationship between the Mainline Protestant adherence rate for a
state and the popularity of “porn” as a Google search term.
[Figure 1 about here]
The scatter plots contained in Figure 1 tell a very different story for the other independent
variables of interest. Graphing the bivariate relationship between the popularity of Google
searches for “porn” and Evangelical Protestant adherence rates for each state clearly
demonstrates a strong, positive, and statistically significant correlation (r = 0.649; p<0.001). In
states with a larger proportion of Evangelical Protestants, the popularity of searching for “porn”
13
on Google is significantly higher. In contrast to Evangelical Protestant adherence rates, higher
levels of Catholic adherence rates are significantly and negatively correlated (r = -0.354; p<0.05)
with popularity of Google searches for “porn”. In states with larger populations that do not
affiliate with a religion the proportion of that state’s total Google searches being for “porn” is
lower. This relationship is both significant and negative (r = -0.355; p<0.05).
Turning from the religious affiliation measures to the religious belief and behavior
measures, Figure 1 also contains a scatter plot of the percent of the state population that is theist
and the popularity of “porn” as a search term. There is a moderately-strong, positive, and
significant correlation (r = 0.592; p<0.001) such that in states where more people believe in God
the popularity of Google searches for “porn” is higher. Similarly, there is a strong, positive, and
significant association between the popularity of “porn” as a search term and the percent of the
population that identifies as a biblical literalist (r = 0.652; p<0.001). Likewise, the last scatter
plot in the array indicates a positive and significant association between the mean level of
religious service attendance state-wide and Google searches for “porn” (r = 0.521; p<0.001). In
states with more theists, biblical literalists, or a population who attends religious services more
frequently the popularity of searching for “porn” on Google is significantly higher.
With six clear significant associations across the religious affiliation, belief, and behavior
measures, we turn to multivariate modeling to assess whether these correlations are robust. Table
2 contains the results of the five OLS regression models for each religious tradition measure.
Models 1 and 3 estimate the associations of total religious adherence rate and Mainline
Protestant adherence rates alongside the various control variables. As with the bivariate analysis,
total religious adherence and Mainline Protestant adherence are not significantly associated with
state-level popularity of Google searches for “porn” in multivariate models. Median income,
14
percent with a Bachelor’s degree, and median age are all significantly and negatively associated
with the popularity of searching for “porn” in these models, though.
[Table 2 about here]
Model 2 examines the effect of Evangelical Protestant adherence rates. Net of the control
variables, Evangelical Protestant adherence rates are significantly and positively associated with
searching for “porn” on Google (β=0.36; p<0.05). In states with higher proportions of
Evangelicals, searches for “porn” appear to be much more prevalent. Interestingly, although the
percent of the population that identifies as politically conservative is positively correlated with
Google searches for “porn” at the bivariate level (r=.315; p<0.05), controlling for the proportion
of Evangelicals and other sociodemographic controls, the association is now negative and non-
significant. Higher median income (β=-0.38; p<0.05), median age (β=-0.33; p<0.01), percent
married (β=-0.24; p<0.05) in the state are also negatively related to Google searches for “porn”.
Comparing the standardized coefficients, the Evangelical Protestant adherence rate is the second
strongest measure in the model.
In contrast, however, Model 4 reveals that the Catholic adherence rate is not significantly
associated with the popularity of “porn” as a Google search term net of the effects of the control
variables. The significant bivariate correlation for the Catholic adherence rate measure appears to
be due to other underlying factors. A state’s median income, percent of the population with a
Bachelor’s degree, median age, and percent married are all significantly and negatively
associated with the popularity of searching for “porn” on Google. Ancillary analyses, available
upon request, demonstrate that Catholic adherence rate is significantly and strongly correlated
with both median income and percent of the state population with a Bachelor’s degree. It could
be that the bivariate Catholic effect discussed above was actually due to differences in income
15
and education within those states. In Model 5 we find that the percent of the population that does
not affiliate with a religious tradition is significantly and negatively associated (β=-0.26; p<0.05)
with the popularity of searching for the term “porn” on Google, net of all other effects. In this
model, political conservatism, income, education, age, and percent married are all significantly
and negatively associated, as well. Substantively, however, percent unaffiliated is the second-
weakest significant association in the model.
[Table 3 about here]
Table 3 contains the final three OLS regression models for the religious belief and
behavior measures. In Model 1 we find, as with the bivariate correlation, that as the percent of
the state population that is theist increases, the popularity of searching for “porn” on Google
increases (β=0.47; p<0.01). Percent conservative, median income, median age, and percent
married are all significantly and negatively associated. Substantively, percent theist has the
largest standardized coefficient in the model. In Model 2 a similar story emerges where percent
of the state that identifies as biblical literalist is significantly and positively associated with the
popularity of searching for “porn” on Google (β=0.36; p<0.05). Median income, median age, and
percent married are the other significant measures. Finally, in Model 3 we find that mean levels
of the frequency of religious behavior is significantly and positively associated with Google
searches for “porn” (β=0.37; p<0.05). Percent politically conservative, median income, median
age, and percent married are each significantly and negatively associated. However, mean
attendance has the second largest standardized coefficient, net of all other effects.
DISCUSSION AND CONCLUSION
Prior research established that aggregate levels of self-identified “religiosity” are linked
with searching for porn on the internet (MacInnis & Hodson, 2015). We build upon this work
16
and provide a necessary expansion with increased precision regarding the characteristics and
contours of aggregate religiosity and spirituality. Specifically, we examine how religious
affiliation adherence rates for multiple groups, aggregated religious beliefs, and average levels of
religious behavior are all linked to the popularity of searching for the term “porn” on Google.
States with more individuals in the Evangelical Protestant tradition, a greater percentage of
theists, a larger proportion of biblical literalists, or a higher mean level of religious service
attendance are all linked to a higher proportion of searches for “porn” across all Google search
engine queries. Conversely, in states with larger proportions of individuals who do not affiliate
with a religious tradition, searching for “porn” on Google is less popular. Standardized
regression coefficients also make clear that aggregate measures of religious belief, behavior, and
affiliation are consistently among the covariates most strongly associated with the dependent
variable. This highlights the importance of accounting for religious variation at the population-
level when examining aggregate outcomes concerning behaviors or attitudes centered on
perceived moral issues.
There are a number of possible explanations for the significant and positive association
between higher proportions of Evangelical Protestants, theists, biblical literalists, and religious
services attenders and online searches for “porn.” While we openly acknowledge that we cannot
definitively determine who is searching for this content, or under what circumstances, we can
explore the possible options in light of our theoretical framework. Drawing on the moral
communities thesis, we posit that higher adherence rates to the Evangelical Protestant tradition,
more theists, more biblical literalists, or more people attending religious services creates a
cultural context, a moral community, where more overt expressions of sexuality are generally
treated with disdain (Baker et al., 2015; Gault-Sherman & Draper, 2012; Hill, 2009; Lee &
17
Barkowski, 2004; Stark, 1996; Ulmer et al., 2008). In this tradition, human behavior is shaped by
the surrounding social context and religion is a group-level trait and an important part of the
social structure (Stark, 1996). These religious moral communities might privilege—or force—
more covert forms of sexual expression for those who belong to these groups and even among
those who do not. In the midst of a strong moral community both the religious and the non-
religious may have limited opportunities for expressing certain aspects of their sexuality.
Searching online for sexually explicit material, where access is anonymous and outside the
purview of disapproving religious communities and people, may be a useful and necessary
outlet.
A related explanation not explored in prior research could be that it is not religious adults
searching for terms like “porn” on Google, but the youth in their homes and communities.
Religiously devout and conservative parents tend to provide less sex education to their children,
and they also tend to monitor their children’s overt sexual activities more closely, either directly
themselves or indirectly through their religious community (Regnerus, 2005, 2007). It could be
that the youth in strongly religious moral communities are left with few options for sexual
education or expression and thus are more likely to search for “porn” online. Related to this
point, it is important to acknowledge that just because people are searching for “porn” does not
necessarily mean that they are using it. While it is likely that most people who are surfing for
pornographic content are planning to view it, some of this searching, especially if it is performed
by younger people, may be out of sheer curiosity.
Even for those outside of these religious moral communities it may be that online
pornography is their primary (or essentially only) outlet, though the non-religious may have the
opportunity to explore various types of sexual expression in less religious contexts. Conversely,
18
in states with greater percentages of unaffiliated individuals, searching for porn may be less
necessary given their greater opportunity for interpersonal sexual exploration because the
community-level norms and surveillance against it are not as prevalent. Furthermore, the youth
in less religious homes and communities may receive more sex education through both formal
(Baker et al., 2015) and informal channels making online searches non-essential. It is important
to note that only through examining the actual content and character of aggregated religiosity
would the above explanation be plausible. Because prior research highlights the actual beliefs
and behaviors of people in particular religious groups—such as Evangelicals providing less sex
education to their youth—we are then able to offer alternative justifications for the overall
finding that more religious areas manifest more online searches for sexually explicit material.
However, it could also be that these groups—Evangelicals, theists, biblical literalists, and
frequent church-goers—actually perform more searches for online porn because it is a secret and
private form of sexual expression that they do not feel the ability to express due to group-level
norms and restrictions on explicit sexuality. As the “breastplate of righteousness” concept and
the preoccupation hypothesis would predict (MacInnis & Hodson, 2015), these individuals
outwardly disdain pornography but in their private moments consume it at relatively higher
levels. There is no way to actually verify this possibility, unfortunately, since we do not have
data on the religious affiliation, beliefs, or behaviors of individual Google searchers. A related
possibility is that devout and conservative religious persons are more likely to binge on porn in
compulsive cycles. Religious persons experience greater shame and guilt due to their porn use
and are also more likely to consider themselves “addicted” to pornography compared to others
(Abell et al., 2006; Grubbs et al. 2016; MacInnis & Hodson 2016). It could be that the more
negative self-evaluations stemming from their porn use lead religious persons into cycles of
19
binging and abstention. But again, testing this possibility would require future studies to
incorporate individual measures.
It is important to compare these findings to those that dominate the religion-porn
connection using primarily individual level data. In that literature, religiosity is almost always
associated with lower levels of porn consumption. Evangelical Protestants report the lowest
levels of porn consumption with other measures of religiosity like biblical literalism and
religious service attendance evincing similar relationships (see Perry, 2016c; Wright, 2013;
Wright et al., 2013). Yet, these studies are overwhelmingly based upon self-reports which could
be problematic given that consumption of pornography is especially stigmatized among
conservative religious Americans. Since Google Trends is an anonymized and aggregate measure
of this stigmatized activity it could be that we are able to circumvent possible measurement error
due to social desirability bias allowing a less cultivated view of the levels to which individuals
consume porn. However, the strength of using an anonymized measure of a stigmatized activity
is also a weakness. We are not able to uncover who is doing the searching for “porn” on Google,
only that it is much more likely to occur in places with more Evangelicals, theists, biblical
literalists, or people who attend religious services often. Despite this limitation, future work
should continue to examine aggregated measures of religiosity/moral communities in relation to
morality issues, including but not limited to porn.
Additionally, it is important to point out that by comparing the present analysis to
findings at the individual level it is clear that aggregate-level relationships need not be identical
to individual-level associations (Kingston & Malamuth, 2011; Robinson, 1950; Schuessler,
1999). Again, it is entirely possible that areas with more conservative Protestants, theists,
biblical literalists, and frequent church attenders simultaneously have high rates of searches for
20
explicit content online while those same individuals—conservative Protestants, theists, biblical
literalists, and frequent church attenders—are not searching for sexually explicit content
individually. Such is the power, and importance theoretically, of the concept of moral
communities and recognizing religion as not only an individual-level trait but a vital part of the
social structure. Prior research that used individual-level theories and explanations to explain
aggregate-level relationships suffered from the ecological inference problem (Kingston &
Malamuth, 2011; Robinson, 1950; Schuessler, 1999). Above we posited several possible
explanations for the religion-porn connection at the state-level. However, the focus of this study
and the moral communities hypothesis is that contextual measures of religion should be
significantly associated with aggregate behaviors in the population. In this case, the moral
communities hypothesis and our use of aggregate measures highlights the somewhat
counterintuitive relationship where religion is associated with more Google searches for “porn.”
Our analyses also demonstrate that the percentage of each state who identify politically as
“conservative” or “very conservative” is negatively related to the popularity of “porn” as a search
term on Google. This suggests that it is not merely about any type of conservatism—be it
religious or political—but the broader religious climate in particular that is associated with
higher overall levels of searching for “porn.” The moral communities established through
particular religious affiliations, beliefs, and practices are related to these practices in a way that
political conservatism is not. Put another way, it is the broader religious climate—not political
conservatism—that is driving these practices into more private arenas and constricting more
open expressions of sexuality. This line of thinking is also supported by the fact that the initial
bivariate association between state-level political conservatism and “porn” searches is positive,
but becomes negative when controlling for religious and other sociodemographic factors.
21
Political conservatism in the aggregate may serve as a proxy for religious conservatism, but
when isolated is in fact negatively related to searching for sexual content. These findings counter
prior work in this area which finds no association between overall conservatism and searching
for “porn” (MacInnis & Hodson, 2015). In fact, MacInnis and Hodson (2015) find that political
conservatism is positively related to Google image searches for “sex”, but no other search terms.
One reason why these findings differ could be that our more precise operationalization of
religion and our broader collection of covariates produce a more specified model.
Overall, this analysis establishes that the religious context of a geographic area is
significantly associated with aggregate levels of pornography use. Specifically, we show that the
aggregated character and content of religion—religious belonging, beliefs, and practice—matter
greatly when attempting to delineate its relationship to aggregate measures of private sexual
activity. Communities may look equally religious, but divergent religious beliefs, religious
service attendance rates, or greater adherence to different religious traditions has profound
implications for the community at-large. This analysis illustrates the importance of accounting
for those differences in order to better understand the possible mechanisms underlying these
associations. As the moral communities theory literature makes clear, human behavior is shaped
by the surrounding social environment and religion is a vital part of social structure. It is a group
property, not just an individual-level trait. It is important for future research of online sexual
behavior to continue to account for religion as an important and powerful social force.
22
REFERENCES
Abell, J. W., Steenbergh, T. A., & Boivin, M. J. (2006). Cyberporn use in the context of
religiosity. Journal of Psychology and Theology, 34(2), 165-171.
Baker, J. O., Smith, K. K., Stoss, Y. A. (2015). Theism, secularism, and sexual education in the
United States. Sexuality Research and Social Policy, 12(3), 236-247.
Baltazar, A., Helm, H.W., McBride, D., Hopkins, G., & Stevens, J.V. (2010). Internet
pornography use in the context of external and internal religiosity. Journal of Psychology
and Theology, 38, 32-40.
Beyerlein, K., & Hipp, J. R. (2005). Social capital, too much of a good thing? American religious
traditions and community crime. Social Forces, 84(2), 995-1013.
Blanchard, T. C., Bartkowski, J. P.., Matthews, T. L., Kerley, K. R. (2008). Faith, morality, and
mortality: The ecological impact of religion on population health. Social Forces, 86(4),
1591-1620.
Bridges, A. J., & Morokoff, P. J. (2011). Sexual media use and relational satisfaction in
heterosexual couples. Personal Relationships, 18, 562-585.
Carroll, J.S., Padilla-Walker, L.M., Nelson, L.J., Olson, C.D., McNamara Barry, C., & Madsen,
S.D. (2008). Generation XXX: Pornography acceptance and use among emerging adults.
Journal of Adolescent Research, 23(1), 6-30.
Daines, R. M., & Shumway, T. (2011). Pornography and divorce. 7th Annual Conference on
Empirical Legal Studies Paper. Available at SSRN:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2112435.
Davis, K. & Braucht, G. (1976). Exposure to pornography, character, and sexual deviance: A
retrospective survey. Journal of Social Issues, 29, 183-196.
23
Doran, K., & Price, J. (2014). Pornography and marriage. Journal of Family and Economic
Issues, 35, 489-498.
Driscoll, M. (2009). Porn again Christian: A frank discussion on pornography and
masturbation. Seattle, WA: Mars Hill Church.
Durkheim, E. (1951). Suicide: A study in sociology. New York: Free Press. (Original work
published in 1897)
Durkheim, E. (1995). The elementary forms of the religious life. New York: The Free Press.
(Original work published in 1912)
Edelman, B. (2009). Red light states: Who buys on-line adult entertainment? Journal of
Economic Perspectives, 23, 209-220.
Gault-Sherman, M., & Draper, M. (2012). What will the neighbors think? The effect of moral
communities on cohabitation. Review of Religious Research, 54(1), 45-67.
Glass, J., & Levchak, P. (2014). Red states, blue states, and divorce: Understanding the impact of
conservative Protestantism on religion variation in divorce rates. American Journal of
Sociology, 119(4), 1002-1046.
Grubbs, J.B., Exline, J.J., Pargament, K.I., Hook, J.N., & Carlisle, R.D. (2015). Transgression as
addiction: Religiosity and moral disapproval as predictors of perceived addiction to
pornography. Archives of Sexual Behavior, 44(1), 125-136.
Hoffmann, J. P., Ellison, C. G., Bartkowski, J.P. (2016). Conservative Protestantism and
attitudes toward corporal punishment, 1986-2010. Social Science Research (in press).
Kingston, D. A., & Malamuth, N. M. (2011). Problems with aggregate data and the importance
of individual differences in the study of pornography and sexual aggression: Comment on
Diamond, Jozifkova, and Weiss. Archives of Sexual Behavior, 40(5), 1045-1048.
24
Lee, M. R., Bartkowski, J. P. (2004). Love thy neighbor? Moral communities, civic engagement,
and juvenile homicide in rural areas. Social Forces, 82(3), 1001-1035.
Lykke, L. C., & Cohen, P. N. (2015). The widening gender gap in opposition to pornography,
1975-2012. Social Currents, 2(4), 307-323.
MacInnis, C. C., & Hodson, G. (2015). Do American states with more religious or conservative
populations search more for sexual content on Google? Archives of Sexual Behavior, 44,
137-147.
MacInnis, C. C., & Hodson, G. (2016). Surfing for sexual sin: Relations between religiousness
and viewing sexual content online. Sexual Addiction & Compulsivity, 23, 196-210.
Moore, L. M., & Vanneman, R. (2003). Context matters: Effects of the proportion of
fundamentalists on gender attitudes. Social Forces, 82(1), 115-139.
Nelson, L.J., Padilla-Walker, L.M., & Carroll, J.S. (2010). “I believe it is wrong but I still do it”:
A comparison of religious young men who do versus do not use pornography.
Psychology of Religion and Spirituality, 2, 136-147.
Patterson, R., & Price, J. (2012). Pornography, religion, and the happiness gap: Does
pornography impact the actively religious differently? Journal for the Scientific Study of
Religion, 51(1), 79-89.
Perry, S. L. (2015). Bible beliefs, conservative religious identity, and same-sex marriage support:
Examining main and moderating effects. Journal for the Scientific Study of Religion,
54(4), 792-813.
Perry, S. L. (2016a). Does viewing pornography diminish religiosity over time? Evidence from
longitudinal data. The Journal of Sex Research. doi: 10.1080/00224499.2016.1146203
Perry, S. L. (2016b). Spousal religiosity, religious bonding, and pornography consumption.
25
Archives of Sexual Behavior. doi: 10.1007/s10508-016-0896-y
Perry, S. L. (2016c). From bad to worse? Pornography consumption, spousal religiosity,
gender, and marital quality. Sociological Forum, 31(2), 441-464.
Perry, S. L., & Hayward, G. M. (2017). Seeing is (not) believing: How viewing pornography
shapes the religious lives of young Americans. Social Forces. doi: 10.1093/sf/sow106
Poulsen, F.O., Busby, D.M., & Galovan, A.M. (2013). Pornography use: Who uses it and how it
is associated with couple outcomes. The Journal of Sex Research, 50(1), 72-83.
Price, J., Patterson, R., Regnerus, M., & Walley, J. (2015). How much more XXX is Generation
X consuming? Evidence of changing attitudes and behaviors related to pornography since
1973. Journal of Sex Research, doi: 10.1080/00224499.2014.1003773
Regnerus, M. D. (2005). Talking about sex: Religion and patterns of parent-child communication
about sex and contraception. The Sociological Quarterly, 46, 79-105.
Regnerus, M. D. (2007). Forbidden fruit: Sex and religion in the lives of American teenagers.
New York: Oxford University Press.
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals. American
Sociological Review, 15, 351-357.
Scheitle, C. P. (2011). Google insights for search: A note evaluating the use of search engine
data in social research. Social Science Quarterly, 91(1), 285-295.
Scheitle, C. P., & Hahn, B. B. (2011). From pews to policy: Specifying evangelical
Protestantism’s influence on states’ sexual orientation policies. Social Forces, 89(3), 913-
933.
Sherkat, D., & Ellison, C. (1997). The cognitive structure of a moral crusade: Conservative
Protestantism and opposition to pornography. Social Forces, 75, 957-82.
26
Stack, S., Wasserman, I., & Kern, R. (2004). Adult social bonds and use of internet pornography.
Social Science Quarterly, 85(1), 75-88.
Stark, R. (1996). Religion as context: Hellfire and delinquency one more time. Sociology of
Religion, 57(2), 163-173.
Strayhorn, J. M., & Strayhorn, J. C. (2009). Religiosity and teen birth rate in the United States.
Reproductive Health, 6(14), doi: 10.1186/1742-4755-6-14.
Ulmer, J. T., Bader, C., Gault, M. (2008). Do moral communities play a role in criminal
sentencing? Evidence from Pennsylvania. The Sociological Quarterly, 49(4), 737-768.
Whitehead, A. L., & Perry, S. L. (2016). Religion and support for adoption by same-sex couples:
The relative effects of religious tradition, practices, and beliefs. Journal of Family Issues,
37(6), 789-813.
Wilson, W., & Abelson, H. (1973). Experiences with and attitudes toward explicit sexual
materials. Journal of Social Issues, 29, 19-39.
Wright, P.J. (2013). U.S. males and pornography, 1973-2010: Consumption, predictors, and
correlates. Journal of Sex Research, 50, 60-71.
Wright, P.J., Bae, S., & Funk, M. (2013). United States women and pornography through four
decades: Exposure, attitudes, behaviors, and individual differences. Archives of Sexual
Behavior, 42, 1131-1144.
27
Table 1: Descriptive Statistics
Mean
SD
Min
Max
Google
Web Search
Google Web Search “Porn”
78.84
9.54
60
100
---
Total Religious Adherence
Rate
483.39
103.35
276.31
791.06
0.099
Evangelical Protestant
Adherence Rate
160.13
108.67
22.81
420.41
0.649***
Mainline Protestant
Adherence Rate
86.13
56.72
8.52
292.66
0.014
Catholic Adherence Rate
169.25
104.89
35.15
449.05
-0.354*
Percent Identifying as No
Religion
12.70
3.71
5.30
19.90
-0.355*
Percent Theist
92.33
3.27
83.70
99.10
0.592***
Percent Biblical Literalist
50.33
10.07
29.70
73.00
0.652***
Mean Attendance
3.70
0.37
2.98
4.63
0.521***
Percent Conservative
37.83
6.10
28.00
54.50
0.315*
Median Household Income
53,104
8,672
38,882
72,999
-0.593**
Percent of Population with
BA
17.65
2.72
11.01
23.45
-0.624***
Median Age of Population
37.58
2.27
29.30
42.8
-0.255†
Percent Married
50.37
2.68
45.35
56.98
-0.272†
N=50; ***p<0.001; **p<0.01; *p<0.05; †p<0.10
28
Table 2: OLS Regression Analysis of State-Level Google Searches for “Porn” and Religious Adherence Rates
Total Religious
Adherence Rate
Evangelical Prot.
Adherence Rate
Mainline Prot.
Adherence Rate
Catholic Adherence
Rate
% Identifying as
No Affiliation
b (SE)
β
b (SE)
β
b (SE)
β
b (SE)
β
b (SE)
β
Total Religious
Adherence Rate
-0.01
(0.01)
-0.14
---
---
---
---
---
---
---
---
Evangelical
Protestant
Adherence Rate
---
---
0.03*
(0.01)
0.36
---
---
---
---
---
---
Mainline Protestant
Adherence Rate
---
---
---
---
0.02
(0.02)
0.15
---
---
---
---
Catholic Adherence
Rate
---
---
---
---
---
---
-0.01
(0.01)
-0.11
---
---
% Identifying as No
Affiliation
---
---
---
---
---
---
---
---
-0.64*
(0.32)
-0.26
% Conservative
-0.03
(0.25)
0.02
-0.36
(0.21)
-0.23
-0.27
(0.22)
-0.17
-0.19
(0.21)
-0.12
-0.48
(0.26)
-0.31
Median Income
-0.001**
(0.00)
-0.49
-0.0004*
(0.00)
-0.38
-0.001**
(0.00)
-0.48
-0.001**
(0.00)
-0.48
-0.001**
(0.00)
-0.49
% with BA Degree
-0.74
(0.56)
-0.21
-0.54
(0.53)
-0.15
-0.91
(0.56)
-0.26
-0.67*
(0.60)
-0.19
-0.91
(0.54)
-0.26
Median Age
-1.69**
(0.51)
-0.40
-1.41**
(0.47)
-0.33
-1.80**
(0.52)
-0.43
-1.49**
(0.50)
-0.35
-1.70**
(0.48)
-0.40
Percent Married
-1.09**
(0.39)
-0.31
-0.86*
(0.37)
-0.24
-1.15**
(0.40)
-0.33
-1.17*
(0.44)
-0.33
-0.83*
(0.38)
-0.23
Intercept
246.10***
215.39***
256.85***
242.47***
255.33***
Adj. R²
0.527
0.583
0.534
0.522
0.557
N
50
50
50
50
50
***p<0.001; **p<0.01; *p<0.05
29
Table 3: OLS Regression Analysis of State-Level Google Searches for “Porn” and Religious Belief and Attendance Rates
Percent Theist
Percent Biblical Literalist
Mean Attendance Rate
b (SE)
β
b (SE)
β
b (SE)
β
Percent Theist
1.37**
(0.45)
0.47
---
---
---
---
Percent Biblical Literalist
---
---
0.34*
(0.14)
0.36
---
---
Mean Attendance Rate
---
---
---
---
9.46*
(4.36)
0.37
% Conservative
-0.59*
(0.24)
-0.37
-0.34
(0.21)
-0.22
-0.57*
(0.28)
-0.37
Median Income
-0.0004*
(0.00)
-0.36
-0.001**
(0.00)
-0.46
-0.001**
(0.00)
-0.46
% with BA Degree
-0.65
(0.52)
-0.18
-0.16
(0.59)
-0.05
-0.65
(0.54)
-0.18
Median Age
-1.42**
(0.46)
-0.34
-1.55**
(0.47)
-0.37
-1.45**
(0.48)
-0.34
Percent Married
-0.95*
(0.36)
-0.27
-0.88*
(0.37)
-0.25
-0.89*
(0.38)
-0.25
Intercept
108.34*
206.45***
202.90***
Adj. R²
0.600
0.577
0.563
N
50
50
50
***p<0.001; **p<0.01; *p<0.05
30
Figure 1: Scatter Plots for Popularity of "Porn" as Google Search Term and Various Religion Measures by State
