Content uploaded by Tobias Ebert
Author content
All content in this area was uploaded by Tobias Ebert on Apr 20, 2020
Content may be subject to copyright.
Religious People only Live Longer in Religious Cultural Contexts:
A Gravestone Analysis
Tobias Ebert
University of Mannheim
Jochen E. Gebauer
University of Mannheim & University of
Copenhagen
Jildou R. Talman
Leiden University
P. Jason Rentfrow
University of Cambridge
Tobias Ebert, Mannheim Centre for European Social Research, University of Mannheim,
Germany;
Jochen Gebauer Mannheim Centre for European Social Research, University of Mannheim,
Germany and Department of Psychology, University of Copenhagen, Denmark;
Jildou R. Talman, Department of Psychology, Leiden University, The Netherlands;
P. Jason Rentfrow, Department of Psychology, University of Cambridge, UK
Corresponding author: Tobias Ebert, Mannheim Centre for European Social Research,
University of Mannheim, A5, 6, D-68159 Mannheim, Germany (email: tobias.ebert@mzes.uni-
mannheim.de).
© 2020, American Psychological Association. This paper is not the copy of record and
may not exactly replicate the final, authoritative version of the article. Please do not
copy or cite without authors' permission. The final article will be available, upon
publication, via its DOI: 10.1037/pspa0000187
Abstract
Religious people live longer than non-religious people according to a staple of social science
research. Yet, are those longevity benefits an inherent feature of religiosity? To find out, we
coded gravestone inscriptions and imagery in order to assess the religiosity and longevity of
6,400 deceased people from religious and non-religious U.S. counties. We show that in
religious cultural contexts, religious people lived 2.2 years longer than did non-religious
people. In non-religious cultural contexts, however, religiosity conferred no such longevity
benefits. Evidently, a longer life is not an inherent feature of religiosity. Instead, religious
people only live longer in religious cultural contexts where religiosity is valued. Our study
answers a fundamental question on the nature of religiosity and showcases the scientific
potential of gravestone analyses.
Keywords: Religiosity, Gravestones, Longevity, Cross-Cultural Differences
GRAVESTONE RELIGIOSITY AND LONGEVITY 2
There is widespread consensus that religious people live longer than non-religious
people (Chida, Steptoe, & Powell, 2009; McCullough, Hoyt, Larson, Koenig, & Thoresen,
2000; Shor & Roelfs, 2013). The mainstream view in the social sciences is that those
longevity benefits are inherent to religiosity (e.g., via religious practices that promote health).
Consequently, this mainstream view is a cultural-universal view stipulating that all religious
people enjoy the same longevity benefits, irrespective of their cultural context. Alternatively,
religiosity’s longevity benefits may not be inherent to religiosity, but driven by a contextual
feature shared by most published research (Sedikides & Gebauer, 2010). In religious cultural
contexts, religious people receive much social valuation for being religious (Gebauer,
Sedikides, & Neberich, 2012; Gebauer et al., 2017). Thus, the longevity benefits of religiosity
might result from that greater social valuation, rather than from religiosity itself. If this
alternative view was correct, the longevity benefits of religiosity should be culture-specific,
i.e., strong in religious cultural contexts and absent in non-religious contexts.
There is a huge amount of evidence for religiosity’s longevity benefits. However, little
is known about those benefits’ cross-cultural variability. The only two studies that
investigated cross-cultural differences arrived at different conclusions.
First, Stavrova (2015) used large-scale representative survey data and found strong
cultural variation in the religiosity-longevity link across 59 countries and 9 U.S. census
regions. However, past research found that self-reported religiosity data can be heavily biased.
Hadaway, Marler, and Chaves (1993), for example, compared self-reported church attendance
with actual headcounts in the U.S. The researchers estimated that U.S. Americans
overestimate their number of church visits by 100%. Issues with data confidentiality and
anonymity pose an additional limitation to survey data. Specifically, researchers typically
have no access to fine-grained information about the participants’ places of residence (i.e.,
counties or census regions). Religiosity is a highly localized phenomenon (Warf & Winsberg,
2008), rendering such fine-grained geographic information particularly relevant (cf.
GRAVESTONE RELIGIOSITY AND LONGEVITY 3
Oberwittler & Wikström, 2009). Second, Wallace, Anthony, End, and Way (2019) examined
religious content in 1,042 obituaries across 42 U.S. cities. They found no evidence for an
interaction effect between individual-level and city-level religiosity on longevity. However,
this study was restricted to a selective subpopulation (urban people from select cities, who had
obituaries in newspapers). Thus, it is unknown whether the findings generalize to the
population at large.
Taken together, unequivocal evidence for cross-cultural variation in the religiosity-
longevity link would have far-reaching scientific and societal implications. For example, such
evidence would call for a revision of the idea that religiosity’s longevity benefits are inherent
and suggest that turning to religiosity is not beneficial everywhere. Unfortunately, the existent
evidence is not unequivocal as it (a) used self-reported assessments of religiosity with
restricted validity, and/or (b) (if at all) compared large-scale areas that might be too broad to
represent truly relevant contexts, and/or (c) relied on samples that excluded large societal
strata.
The present investigation uses a unique and entirely novel empirical approach to
examine whether religiosity’s longevity benefits are restricted to religious cultural contexts.
Specifically, we use cemeteries as our data source and measure people’s longevity and
religiosity by the inscriptions and imagery on their gravestones.
Geographers, historians, and cultural scientists (Hijiya, 1983; Saller & Shaw, 1984;
Zelinsky, 2007) have long recognized that gravestones preserve important information about
the people buried beneath them. Gravestones often carry religious imagery, which makes
them a particularly useful indicator of deceased people’s religiosity. For one thing, the
available space on a gravestone is restricted, and, hence, only the most important or
memorable characteristics of a person are usually included. For another thing, “[a decedent,
spouse, or close family member] is more likely to devote greater time and effort in deciding
what, if anything, to put on a permanent memorial – usually a decidedly costly commodity
GRAVESTONE RELIGIOSITY AND LONGEVITY 4
and once-in-a-lifetime transaction” (Zelinsky, 2007, p. 447). Consequently, gravestone
imagery might provide particularly valid information about deceased people. That way,
studying gravestones could be less prone to the perennial problem of response bias in survey
research. Figure 1 provides illustrative examples from our data contrasting prototypical
religious from non-religious gravestones.
Beyond the “set-in-stone” aspect of our methodology, gravestone analyses provide
two additional advantages. First, cemeteries are one of the few institutions in today’s world
that bring together otherwise heavily segregated groups. For example, in the U.S., black and
white people share the same cemeteries since 1969 (Rogers, 2005). Consequently, studying
gravestones comes close to a probabilistic sampling approach. Thus, gravestone analyses are
relatively immune to otherwise frequent sampling biases, such as under-sampling rural,
working-class, and ethnically diverse subpopulations (Gurven, 2018). Second, cemeteries are
everywhere and we know their exact locations. Hence, sampling from cemeteries allows
examining cultural variation at those fine-grained geographical levels that are particularly
relevant (Kashima et al., 2004; Warf & Winsberg, 2008), but often unavailable to researchers.
GRAVESTONE RELIGIOSITY AND LONGEVITY 5
Figure 1. Prototypical religious gravestones (top row) and gravestones with non-
religious imagery (bottom left) or no imagery (bottom right).
Data and Methods
Sampling and Coding Approach
We analyzed the association between religiosity and longevity based on large-scale
gravestone data. Our study focuses on Christians (vs. non-believers) in the U.S.—the
denomination and country that is typically studied in religiosity research. We set up an a
priori sampling strategy (see Sampling and Coding Description in Online Supplement), to
select which areas and graves to include in our analyses. In short, we collected data from a
sample of 6,400 deceased people from 64 U.S. counties. To do so, we relied on publicly
available information from the U.S. religion census and the internet archive findagrave.com.
First, we used the U.S. religion census to calculate the share of Christians (i.e.,
cumulated share of Catholics and Protestants) within counties. Based on this county
religiosity measure, we stratified the sampling of the 64 counties so that 50% represent
religious cultural contexts (i.e., high share of Christians) and 50% non-religious cultural
GRAVESTONE RELIGIOSITY AND LONGEVITY 6
contexts (i.e., low share of Christians). In each cultural context (i.e., religious and non-
religious) 50% of the counties were urban (i.e., inside metropolitan areas) and 50% were rural
(i.e., outside metropolitan areas). Figure 2 depicts the geographical distribution of the sampled
counties.
Figure 2. Geographical distribution of sampled counties.
Second, we used the gravestone photographs provided on findagrave.com to assess
the religiosity and longevity of 100 deceased people in each sampled cemetery (50% male and
50% female, all passed away since January 1st 2000). On average, findagrave.com provided a
gravestone photograph for 81% of all deceased people across all cemeteries. To measure
religiosity we coded the imagery displayed on each gravestone following Zelinsky’s
gravestone classification scheme (Zelinsky, 2007). Tables S1-S2 provide the coding manual,
coding examples, and tests of coding reliability. The descriptive statistics (Table S3)
showcase the gravestone approach’s merits. First, gravestones indeed frequently carry
religious imagery (44% of all gravestones). Second, life expectancy in our sample (78.79
years) closely matched official statistics (78.54 years in 2017). Third, our sample of counties
GRAVESTONE RELIGIOSITY AND LONGEVITY 7
also covered remote populations that conventional studies often neglect, such as counties with
less than one inhabitant per square mile.
Gravestone Religiosity Index
We distinguish between five elements of religious imagery that occurred frequently
(Figure 3): the cross (appearing on 29% of all gravestones), book symbols representing the
bible or the book of life (10%), hands folded in prayer (7%), passages of scripture (6%), and
angels (2%). Consistent with the expectation that religiosity is reflected on deceased people’s
gravestones, we observed significantly more gravestones with religious imagery in religious
counties (M = .48) than in non-religious counties (M = .37), t(6,398) = -9.42, p < .001. Three
elements discriminated particularly well between religious and non-religious counties: books,
t(6,398) = -9.12, p < .001, praying hands, t(6,398) = -6.77, p < .001, and angels, t(6,398) = -
3.48, p < .001. The presence of Book-Praying-Angel imagery performed equally well in urban
counties (Mreligious = 0.21 / Mnon-religious = 0.10; t(3,198) = -8.03, p < .001) and rural counties
(Mreligious = 0.23 / Mnon-religious = 0.12; t(3,198) = -8.22, p <.001). By contrast, the presence of
Cross-Verbal imagery only discriminated in urban settings (Mreligious = 0.32 / Mnon-religious =
0.41; t(3,198) = -5.13, p < .001), but not in rural settings (Mreligious = 0.29 / Mnon-religious = 0.30;
t(3,198) = -0.931, p = .35).
GRAVESTONE RELIGIOSITY AND LONGEVITY 8
Figure 3. Prevalence of religious imagery across religious and non-religious counties.
To further ensure that Book-Praying-Angel imagery indeed reflects deceased
people’s religiosity (i.e., to avoid the ecological fallacy; Robinson, 1950) we cross-validated
them with additional individual-level information (see Gravestone Religiosity Measure in
Online Supplement). Specifically, we researched short biographies of deceased people that are
occasionally provided on findagrave.com (448 out of 6,400 observations). We then used
natural language processing based on the well-established LIWC dictionary (Pennebaker,
Boyd, Jordan, & Blackburn, 2015) to analyze the relative frequency of religious words within
these biographies. In line with our aggregate-level findings, deceased people with Book-
Praying-Angel imagery on their gravestone featured a 48% higher share of religious words
within their biographies than people without such imagery (Mbpa= .044 / Mnon-bpa = .029;
t(446) = -4.56, p < .001). By contrast, for the poorly discriminating imagery elements (i.e.,
cross and verbal scriptures) only a non-significant difference of 15% in the share of religious
words emerged (Mcv = .035 / Mnon-cv = .030; t(446) = -1.82, p = .07).
GRAVESTONE RELIGIOSITY AND LONGEVITY 9
Taken together, evidence from the aggregate level and the individual level
converges, suggesting that a deceased person’s religiosity can be validly assessed by the
presence of books, praying hands, and angels on that person’s gravestone. We, thus, devised a
gravestone religiosity index (see Gravestone Religiosity Measure in Online Supplement) that
sums up the three well discriminating elements and relativizes that sum by the total number of
imagery elements on the gravestone (results were conceptually identical with other
operationalizations; see robustness checks later).
Analysis Method
Previous research found that religiosity is not related to unusually early mortality (e.g.,
due to traffic accidents). For instance, McCullough, Friedman, Enders, and Martin (2009)
found that religious and non-religious people start to differ in their mortality from age 70-75
onwards, with the greatest difference at age 90-100. Consequently, our main-text analyses
include people who were between 70 to 99 years old when they died. In our robustness
checks, however, we used other age cut-offs and found our results robust to those different
cut-offs.
Our main-text analyses relied on data from 4,946 deceased people. To account for the
nested data structure (deceased people nested in counties), we used linear mixed-effects
modeling (Snijders & Bosker, 1999). Longevity served as the criterion in all mixed-effects
models. In the basic model, gravestone religiosity and the two stratification criteria (gender
and urbanity) served as level-1 predictors, census information on the share of religious
adherents per county (county religiosity) served as a level-2 predictor, and the cross-level
interaction between gravestone religiosity × county religiosity (both z-standardized) was
GRAVESTONE RELIGIOSITY AND LONGEVITY 10
modeled as an additional predictor. We specified random intercepts and random slopes of
gravestone religiosity.
Results
Our basic model revealed no main effect of gravestone religiosity on longevity, B = -
0.001, 95% CI [-0.22, 0.21], p = .99, but a significant cross-level interaction between
gravestone religiosity × county religiosity, B = 0.30, 95% CI [0.10, 0.50], p = .004. In the
most religious county (Figure 4), the most religious people (M = 86.82, SD = 0.75) lived 2.24
years longer than the least religious people (M = 84.58, SD = 0.32), B = 0.53, 95% CI [0.18,
0.89], p = .003. This longevity benefit, however, vanished with decreasing county-level
religiosity. In the least religious counties, the most religious people (M = 82.91, SD = 0.85)
tended to live 1.72 years shorter than non-religious people (M = 84.63, SD = 0.26), B = -0.41,
95% CI [-0.81, -0.01], p = .05. Taken together, there is great cross-cultural variation in the
link between religiosity and longevity. In religious cultural contexts, the most religious people
lived an average of 3.91 years longer than in non-religious cultural contexts. This is a
longevity difference of considerable magnitude that exceeds gender differences (2.44 years)
in our models and is comparable to the impact of severe obesity on longevity (Fontaine,
Redden, Wang, Westfall, & Allison, 2003).
GRAVESTONE RELIGIOSITY AND LONGEVITY 11
Figure 4. Interaction between gravestone religiosity and county-level religiosity
on longevity.
We conducted 14 robustness checks (see Robustness Checks in Online Supplement) to
scrutinize the robustness of our basic finding (M1). First, we tested against individual-level
confounds (R1: total number of imagery elements, presence of non-religious imagery, marital
status, number of people in grave, type of gravestone, floral or wheat ornamentation) and
county-level confounds (R2: racial composition, population density, various health
determinants). Second, we controlled for influences of superordinate cultural levels by
including state-fixed effects (R3). Third, we added three-way interaction terms consisting of
Gravestone Religiosity × County Religiosity × Stratification criteria to test for moderating
effects of the two stratification criteria gender (R4) and urbanity (R5). Fourth, we lowered
(R6: 63.5-99 years) and raised (R7: 75-99 years) the age-cut offs and also included people
who reached an age of 100+ (R8). Fifth, we operationalized gravestone religiosity not as the
share of Book-Praying-Angel imagery, but in form of a dummy (R9), a sum (R10), and a ratio
(R11). Sixth, we included only those 1,065 people for which occasionally available
GRAVESTONE RELIGIOSITY AND LONGEVITY 12
information ensured that their county of death and burial were identical (R12). Sixth, we
controlled for cohort and period effects by accounting for year of birth (R13) and year of
death (R14). Finally, we corrected the p-values of our models for multiple testing (Table S5).
For all these robustness checks, the results were conceptually identical with the main-text
results. Taken together, our results strongly suggest that the longevity benefits of religiosity
are not cultural universal. Instead, those longevity benefits were restricted to religious cultural
contexts and vanished altogether in non-religious cultural contexts.
Discussion
A large body of social scientific research suggests that religious people live longer
lives (Chida et al., 2009; McCullough et al., 2000; Shor & Roelfs, 2013). However, it has
been unclear whether these longevity benefits are inherent to religiosity (and thus culturally
universal) or dependent on the religiosity of the adherent’s cultural context. We found that the
longevity benefits of religiosity are not evident everywhere (i.e., are not cultural universal),
but are limited to religious cultural contexts. Our study is the first psychological study to ever
rest on gravestones and it strongly suggests that gravestone imagery can provide valid
information about deceased people’s religiosity. By studying gravestone imagery, we were
able to examine a valid religiosity marker across representative samples and fine-grained local
cultures. We, thus, understand our study as a conservative, comprehensive, and much-needed
extension of the influential research on the religiosity-longevity relationship.
Although our gravestone approach has key advantages over previous approaches, our
approach also has its limitations. First, gravestone analyses are necessarily cross-sectional
(people only die once). While we controlled for a wide variety of potential confounds in our
14 robustness checks, we cannot fully rule out reversed causality (i.e., older people in some
areas being particularly likely to put religious imagery on their gravestones). However, this
interpretation would be at odds with findings showing that religiosity is fairly stable across
the lifespan, especially in the later stages of life (McCullough et al., 2009). Second,
GRAVESTONE RELIGIOSITY AND LONGEVITY 13
gravestone data circumvent several biases inherent in survey data, but gravestone data likely
have their unique biases. For example, gravestone imagery may sometimes be chosen by the
deceased people themselves (akin to self-reports) and sometimes by people close to the
deceased (akin to peer-reports). Likewise, little is known about how the financial cost of
imagery, the availability of craftsmanship, and local imagery norms affect gravestone data.
Importantly, the biases inherent in gravestone data are probably very different from the biases
inherent in survey data. Thus, the validity of any given hypothesis is particularly well
supported if both types of data reveal evidence for that hypothesis. The present research in
tandem with previous survey research (Gebauer et al., 2012, 2017) renders the cultural
specificity of religiosity’s benefits such a well-supported hypothesis.
Limitations notwithstanding, our findings have at least four significant implications.
First, our findings suggest that the longevity benefits of religiosity only emerge in religious
cultural contexts. Our study, thus, challenges a longstanding staple of research and it does so
using a rigorous methodological approach (i.e., a set-in-stone measure of religiosity,
representative sampling, and fine-grained local cultures). The findings also shed light on the
mechanisms driving religious longevity benefits. Evidently, religiosity’s longevity benefits
are not due to some inherent feature of religiosity. Instead, they may be due to the more
general phenomenon that people enjoy health benefits if they receive social valuation from
their ambient cultural context (Gebauer et al., 2012, 2017). Second, from a broader data
science perspective, we provide a large-scale, empirically driven validity check of gravestone
information. Gravestones provide insights on important life domains beyond religiosity (e.g.,
hobbies/interests, political views or patriotism, see Table S1). Our study, thus, highlights that
gravestones are a widely overlooked, but valuable scientific data source. Third, from a
societal perspective, becoming religious has been thought of as a powerful health
“intervention” (Lucchetti, Lucchetti, & Koenig, 2011). However, our findings suggest that
such religiosity-based health interventions might be effective only in religious cultural
GRAVESTONE RELIGIOSITY AND LONGEVITY 14
contexts. Finally, our study holds important implications for a Western world that is becoming
increasingly secularized (Joshanloo & Gebauer, 2019). Religiosity does not delay death
everywhere and rising secularization will probably not curb collective life expectancy in the
future.
GRAVESTONE RELIGIOSITY AND LONGEVITY 15
References
Association of Statisticians of American Religious Bodies (2010). Religious congregations
and membership study, 2000 and 2010. Retrieved from
http://www.usreligioncensus.org/.
Bates, D., Kliegl, R., Vasishth, S., & Baayen, H. (2015). Parsimonious mixed models. arXiv
preprint arXiv:1506.04967.
Centers for Disease Control and Prevention (2019). Social Determinants of Health: Know
What Affects Health. Retrieved from:
https://www.cdc.gov/socialdeterminants/index.htm.
Chida, Y., Steptoe, A., & Powell, L. H. (2009). Religiosity/spirituality and mortality: A
systematic quantitative review. Psychotherapy and Psychosomatics, 78, 81-90.
County Health Rankings & Roadmaps (2019). County Health Rankings & Roadmaps -
Building a Culture of Health, County by County. Retrieved from:
http://www.countyhealthrankings.org
Find A Grave (2019). World’s largest gravesite collection. Over 170 million memorials
created by the community since 1995. Retrieved from https://www.findagrave.com/.
Fontaine, K. R., Redden, D. T., Wang, C., Westfall, A. O., & Allison, D. B. (2003). Years of
life lost due to obesity. JAMA - Journal of the American Medical Association, 289,
187-193.
Gebauer, J. E., Sedikides, C., & Neberich, W. (2012). Religiosity, social self-esteem, and
psychological adjustment: On the cross-cultural specificity of the psychological
benefits of religiosity. Psychological Science, 23, 158-160.
Gebauer, J. E., Sedikides, C., Schönbrodt, F., Bleidorn, W., Rentfrow, J., Potter, J., &
Gosling, S. (2017). The religiosity as social value hypothesis: A multi-method
replication and extension across 65 countries and three levels of spatial aggregation.
Journal of Personality and Social Psychology, 113, e18-e39.
GRAVESTONE RELIGIOSITY AND LONGEVITY 16
Gurven, M. D. (2018). Broadening horizons: Sample diversity and socioecological theory are
essential to the future of psychological science. Proceedings of the National
Academy of Sciences, 115, 11420-11427.
Gwet, K. (2014). Handbook of Inter-Rater Reliability: The Definitive Guide to Measuring the
Extent of Agreement Among Raters. Gaithersburg, MD: Advcanced Analytics, LLC.
Hadaway, C. K., Marler, P. L., & Chaves, M. (1993). What the polls don’t show: A closer
look at US church attendance. American Sociological Review, 58, 741-752.
Hijiya, J. A. (1983). American gravestones and attitudes toward death: A brief history.
Proceedings of the American Philosophical Society, 127, 339-363.
Joshanloo, M., & Gebauer, J. E. (2019). Religiosity’s nomological network and temporal
change: Introducing an extensive country-level religiosity index based on Gallup
World Poll data. European Psychologist. Advance online publication. doi:
10.1027/1016-9040/a000382
Kashima, Y., Kokubo, T., Kashima, E. S., Boxall, D., Yamaguchi, S., & Macrae, K. (2004).
Culture and self: Are there within-culture differences in self between metropolitan
areas and regional cities? Personality and Social Psychology Bulletin, 30, 816-823.
Lucchetti, G., Lucchetti, A. L., & Koenig, H. G. (2011). Impact of spirituality/religiosity on
mortality: comparison with other health interventions. Explore, 7, 234-238.
Maas, C. J. M., & Hox, J. J. (2005). Sufficient sample sizes for multilevel modeling.
Methodology, 1, 85-91.
McCullough, M. E., Friedman, H. S., Enders, C. K., & Martin, L. R. (2009). Does devoutness
delay death? Psychological investment in religion and its association with longevity
in the Terman sample. Journal of Personality and Social Psychology, 97, 866-882.
McCullough, M. E., Hoyt, W. T., Larson, D. B., Koenig, H. G., & Thoresen, C. (2000).
Religious involvement and mortality: A meta-analytic review. Health Psychology,
19, 211-222.
GRAVESTONE RELIGIOSITY AND LONGEVITY 17
National Center for Health Statistics (2019). NCHS Urban-Rural Classification Scheme for
Counties. Retrieved from https://www.cdc.gov/nchs/data_access/urban_rural.htm.
Oberwittler, D., & Wikström, P. O. H. (2009). Why small is better: Advancing the study of
the role of behavioral contexts in crime causation. In D. Weisburd, W. Bernasco, &
G. J. N. Bruinsma (Eds.), Putting crime in its place (pp. 35-59). New York: Springer.
Pennebaker, J.W., Boyd, R. L., Jordan, K. & Blackburn, K. (2015). The development and
psychometric properties of LIWC2015. Austin, TX: University of Texas at Austin.
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals. American
Sociological Review, 15, 351-357.
Rogers, K. (2005). Integrating the city of the dead: The integration of cemeteries and the
evolution of property law, 1900-1969. Alabama Law Review, 56, 1153-1187.
Saller, R. P., & Shaw, B. D. (1984). Tombstones and Roman family relations in the
Principate: Civilians, soldiers and slaves. The Journal of Roman Studies, 74, 124-
156.
Sedikides, C., & Gebauer, J. E. (2010). Religiosity as self-enhancement: A meta-analysis of
the relation between socially desirable responding and religiosity. Personality and
Social Psychology Review, 14, 17-36.
Shor, E., & Roelfs, D. J. (2013). The longevity effects of religious and nonreligious
participation: A meta-analysis and meta-regression. Journal for the Scientific Study
of Religion, 52, 120-145.
Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and
advanced multilevel modeling. Thousand Oaks, CA: Sage
Stavrova, O. (2015). Religion, self-rated health, and mortality: Whether religiosity delays
death depends on the cultural context. Social Psychological and Personality Science,
6, 911-922.
GRAVESTONE RELIGIOSITY AND LONGEVITY 18
United States Census Bureau (2019), American Community Survey, Retrieved from
https://www.census.gov/programs-surveys/acs/.
Wallace, L. E., Anthony, R., End, C. M., & Way, B. M. (2019). Does religion stave off the
grave? Religious affiliation in one’s obituary and longevity. Social Psychological
and Personality Science, 10, 662-670.
Warf, B., & Winsberg, M. (2008). The geography of religious diversity in the United States.
The Professional Geographer, 60, 413-424.
Welbers, K., Van Atteveldt, W., & Benoit, K. (2017). Text analysis in R. Communication
Methods and Measures, 11, 245-265.
Yekutieli, D. & Benjamini, Y. (2001). The control of the false discovery rate in multiple
testing under dependency. Annals of Statistics, 29, 1165-1118.
Zelinsky, W. (2007). The Gravestone Index: Tracking personal religiosity across nations,
regions, and periods. Geographical Review, 97, 441-466.
Supplementary Materials for
Religious People only Live Longer in Religious Cultural Contexts:
A Gravestone Analysis
2
Sampling and Coding Description
Sampling Approach
To select appropriate areas and gravestones for this study, we used a two-step sampling
approach. Step 1 aimed at sampling a set of U.S. areas representing (a) religious and non-
religious cultural contexts from (b) a diverse background of rural and urban living environments.
Geographical differences in religiosity follow rather fine-grained distributional patterns (Warf &
Winsberg, 2008). We, therefore, chose counties as our level of analysis as they are the smallest
spatial level for which comprehensive religiosity information is available. To distinguish
between religious and non-religious counties, we calculated the proportion of Christians in each
of the 3,143 U.S counties (i.e., number of people with Catholic, Evangelical, or Mainline
Protestant
1
denomination relative to the county’s population according to census data, averaged
across years 2000 and 2010; Association of Statisticians of American Religious Bodies, 2010).
To distinguish between urban and rural counties, we identified counties that are located outside
metropolitan areas (i.e., rural, N = 1,335) or inside metropolitan areas (i.e., urban, N = 1,808)
(based on the NCHS Urban-Rural Classification Scheme for Counties; National Center for
Health Statistics, 2019). To sample religious and non-religious counties from rural living
environments, we used the list of 1,335 counties outside metropolitan areas and identified those
133 counties belonging to the 10% most religious and those 133 counties belonging to the 10%
least religious counties. From each of these two subsets (top and bottom 10%), we randomly
chose 16 counties. To sample religious and non-religious counties from urban living
environments, we used the list of 1,808 counties inside metropolitan areas and identified those
1
There are different ways to group the Protestant population. We differentiate between Evangelical and Mainline
Protestants. That differentiation is commonly used (e.g., by the U.S. religion census) and largely exhaustive of the
Protestant population.
3
360 counties belonging to the 20% most religious counties and those 360 counties belonging to
the 20% least religious counties. In the urban setting, expanding to the top and bottom 20%
(instead of 10%) was necessary, as otherwise only small metropolitan areas would have been
sampled, while excluding all larger metropolitan areas. From each subset (i.e., top and bottom
20%), we again randomly picked 16 counties. As a result, the counties within each of our two
conditions (i.e., religious / non-religious) were stratified by urbanity, but otherwise random.
Figure 1 in the main text shows that our sampling approach led to a fair geographical spread
across the U.S. Moreover, the reported descriptive statistics in Table S3 show that our sampled
counties indeed cover very diverse living environments (e.g., ranging from remote areas with
less than one inhabitant per square mile to urban cores with more than 1,200 inhabitants per
square mile).
Step 2 sampled people buried in the 64 counties selected in step 1 (32 religious counties
and 32 non-religious counties). To do so, we relied on the world’s largest collection of gravesite
information, which holds more than 180 million entries of deceased people: findagrave.com
(Find A Grave, 2019). For each of the 64 sampled counties, we identified the largest cemetery
(most memorials) on findagrave.com. From each of these largest cemeteries, we randomly
sampled 50 men and 50 women who passed away since January 1st 2000 and featured a
gravestone photo (which was the case for 81% of all deceased people in the sampled cemeteries).
Our sampling approach eventually resulted in a data set comprising 6,400 deceased people from
64 different counties. With 64 level two units (i.e., counties) each containing 100 observations
(i.e., 50 men and 50 women) our sample is sufficiently large and well-powered for mixed-effects
modeling (Maas & Hox, 2005).
4
Coding Approach
We used the gravestone photographs provided on findagrave.com to extract (a) deceased
people’s age at death and (b) their religiosity. To identify religious imagery elements on
gravestones, we relied on Zelinsky’s (2007) scheme that classifies frequently occurring elements
of gravestone imagery as stemming from a secular or religious background. To code our data, we
transferred Zelinsky’s scheme into the coding manual provided in Table S1.
To assess the reliability of our coding approach, a second independent coder used the
same manual (Table S1) to code a randomly drawn subset of 200 gravestones. This subset was –
unbeknownst to the coder – stratified in a way that it contained 100 graves with religious
imagery and 100 graves without religious imagery. We then evaluated intercoder reliability
(Cohen’s Kappa for categorical variables, the intraclass correlation coefficient [ICC] for metric
variables; Gwet, 2014). Table S2 shows that all types of gravestone elements featured near-
perfect intercoder reliability.
Gravestone Religiosity Measure
We chose the five most frequently occurring religious imagery elements (i.e., cross,
book, praying hands, passages of religious scripture, and angels; Table S3) and examined their
ability to discriminate between religious and non-religious counties. As shown in Figure 3,
symbols that discriminated very well between religious and non-religious counties were the book
(102% higher likelihood of occurrence in religious than non-religious counties), praying Hands
(87% higher likelihood in religious counties), and the angel (102% higher likelihood). The
presence of any Book-Praying-Angel imagery discriminated very well between religious and
non-religious counties in urban settings (Mreligious = 0.21 / Mnon-religious = 0.10; t(3,198) = -8.03, p
< .001) and in rural settings (Mreligious = 0.23 / Mnon-religious = 0.12; t(3,198) = -8.22, p < .001). By
contrast, the cross (only 18% higher likelihood of occurrence in religious than non-religious
5
counties) and passages of scripture (13% higher likelihood) discriminated rather poorly. The
presence of Cross-Verbal imagery only discriminated between religious and non-religious
counties in urban settings (Mreligious = 0.32 / Mnon-religious = 0.41; t(3,198) = -5.13, p < .001), but
not in rural settings (Mreligious = 0.29 / Mnon-religious = 0.30; t(3,198) = -0.93, p = .35). In fact, the
cross is the most common image in our data. It seems plausible that the cross has become a
standard feature of gravestone design in the U.S. For example, military badges on veterans’
gravestones almost by default carry crosses. In our data, 78% of the gravestones that contained a
military marker, also featured a cross. Taken together, deceased people who were only loosely
affiliated with their religion might often feature standard Christian symbols (such as the cross) on
their gravestones. In contrast, more specific religious imagery (like bibles, praying hands, or
angels) might only be picked by (or for) truly devout people.
In the previous paragraph, we sought to validate our gravestone religiosity measure by
aggregating it to the county level and testing its association with county religiosity. However,
findings from the aggregate level not necessarily need to generalize to the individual level—a
problem known as ecological fallacy (Robinson, 1950). To cross-validate Book-Praying-Angel
imagery as a marker of deceased people’s religiosity, we sought to perform an additional
individual-level validity test. Specifically, we made use of the fact that findagrave.com
sometimes provides additional information about deceased people in form of short texts written
by descendants. If our gravestone religiosity measure indeed reflects deceased people’s
religiosity, we would expect that for people with Book-Praying-Angel imagery on their
gravestones, these texts contain more religious content.
If a text was given at all, in most cases it was a very short obituary simply stating
information regarding the date, time and location of the funeral ceremony as well as the names of
6
the closest descendants. In some cases, however, these texts were no simple obituaries, but rather
short biographies—that is, texts that are longer and contain more information about the deceased
person. To focus on these informative biographies, we only kept texts that were at least 300
words long (N = 448). Before analyzing these texts, we followed standard approaches in text
analysis (Welbers, Van Atteveldt, & Benoit, 2017) and removed any signs that were not letters,
removed all filler words that have no substantial meaning (so-called stop words) and reduced
inflected words to their word stem (so-called stemming). To identify words with religious
meaning within these cleaned texts, we used the religiosity dimension of the well-established
LIWC dictionary (Pennebaker, Boyd, Jordan, & Blackburn, 2015). Given our focus on
Christianity, from the list of 174 religious words within the 2015 LIWC dictionary, we erased 46
words that clearly stemmed from a non-Christian background (e.g., allah, buddha, imam, kosher,
krishna, mosque, rabbi, sikh). We then counted how often the remaining 128 religious words
occur within each of the 448 cleaned texts. Finally, we divided the number of religious words by
the total number of words in the cleaned text. On average, 3.19% of all words within the cleaned
texts were religious. Importantly, we found that in biographies of people with Book-Praying-
Angel imagery on their gravestone religious words occurred much more frequently (48 % more
often; Mbpa= .044 / Mnon-bpa = .029; t(446) = -4.56, p < .001) than for people without such
imagery. By contrast, for the poorly discriminating imagery elements (i.e., the cross and verbal
scriptures) this difference was much smaller and not significant (15% more often; Mcv = .035 /
Mnon-cv = .030; t(446) = -1.82, p = .07).
Given the converging evidence from the aggregate and individual level, we used the
Book-Praying-Angel imagery to measure deceased people’s religiosity. To transfer Book-
7
Praying-Angel imagery into a specific gravestone religiosity measure, we devised four different
indicators of gravestone religiosity:
(1) dummy variable indicating whether any of the three imagery elements are present,
(2) sum variable indicating the number of the three imagery elements,
(3) share variable indicating the share of the three elements relative to the total number of
imagery elements,
(4) ratio variable indicating the ratio between the sum of the three elements and the sum
of non-religious imagery elements.
To select one of these four indicators, we tried to find the indicator that showed the
highest external validity. To do so, we aggregated the individual scores for each indicator to the
county-level and then correlated these aggregated scores with the share of religious adherents in
the county (N = 64). We found the strongest correlation for the share indicator (β = .56) followed
by the dummy indicator (β = .54), the sum indicator (β = .51), and the ratio indicator (β = .50).
Therefore, we used the share indicator (i.e., the share of Book-Praying-Angel imagery on the
total number of imagery elements) to measure gravestone religiosity in our main analyses.
Nevertheless, we also tested the remaining three indicators and found similar results (see
robustness checks later).
8
Table S1
Coding Manual and Coding Examples from our Data
Variable
Description
Operationalization
Coding Examples
Basic gravestone information
Type of gravestone
The type of gravestone does not refer to
imagery on the gravestone, but
describes the type of stone itself, i.e., is
it an upright stone, a stele, or a flat
plaque in the ground.
1 = upright, rectangular stone
2 = stele, an upright narrow gravestone,
usually with a triangular top (not very
common)
3 = plaque, stone placed flat on the
ground
4 = miscellaneous
Wheat ornamentation
Bunch of wheat (most common) or
similar agricultural produce (e.g.,
corn).
0 = not on gravestone
1 = engraved on gravestone
Floral ornamentation
Floral ornamentation, such as
decoration around a frame. Sometimes
a single flower is presented as a symbol
instead of aesthetic decoration, in
which case code as [non-religious
symbol].
0 = not on gravestone
1 = engraved on gravestone
Month/Year of marriage
Always referred to explicitly (e.g.,
“Married [date]”), often including
interlocking rings (i.e., Non-religious
symbol) or church bells (i.e., Other
religious).
Month in digits 1-12; year in four-digit
format (e.g., 1962)
Non-religious imagery
Mention of family role
Only when inscribed on stone (not in
findagrave biography) and explicitly
mentioned.
0 = not on gravestone
1 = engraved on gravestone
Military reference
Textual references as well as emblems
(e.g., Airforce, Navy, Army; very
uncommon) evidence of having served
in a military sphere.
0 = not on gravestone
1 = engraved on gravestone
9
Patriotic Reference
American flag or bald eagle.
0 = not on gravestone
1 = engraved on gravestone
Mention occupation
Anything non-military, textual as well
as symbolic.
0 = not on gravestone
1 = engraved on gravestone
Picture of deceased
Can be photograph as well as
engraving, so long as it is a
recognisable portrait.
0 = not on gravestone
1 = engraved on gravestone
No example due to confidentiality reasons.
Verbal message of love
Anything referring to love or memory.
Also includes “Together forever” on
shared gravestones. “Beloved” as well
as “loving [family member]”.
0 = not on gravestone
1 = engraved on gravestone
Verbal non-religious
message
Can be about favourite pastimes or
character descriptions of the deceased.
Also sometimes quotes from the
deceased, from films or
music―importantly, the media these
citations come from should not be
explicitly religious, in which case the
quote would go under “Verbal religious
message”.
0 = not on gravestone
1 = engraved on gravestone
Non-religious symbols
Often symbols referring to favourite
pastimes, pets, or scenery. Also
includes masonic symbolism and non-
Christian religious symbolism.
0 = not on gravestone
1 = engraved on gravestone
Religious imagery
10
Book
Book symbols can represent either the
Bible or the Book of Life (i.e., the book
in which God records the names of the
persons destined for heaven). Books
can be used either to frame names or as
separate symbolism.
0 = not on gravestone
1 = engraved on gravestone
Religious passage of
scripture
Mentions of God, Jesus, and/or
Heaven, references to psalms, and/or
citations from explicitly religious
media. “Until we meet again” was also
included in this category.
0 = not on gravestone
1 = engraved on gravestone
Fish (Ichtus)
Make sure to distinguish between fish
depicting the hobby of fishing (i.e.,
“non-religious symbol”) and Ichtus, the
religious symbol. Oftentimes the
difference is detailing: The latter is far
more simplified.
0 = not on gravestone
1 = engraved on gravestone
Dove
Make sure to distinguish between
doves specifically and other birds (i.e.,
“non-religious symbol” or bald eagles
under “American flag”). Oftentimes the
difference is an olive branch.
0 = not on gravestone
1 = engraved on gravestone
Hands in prayer
Whenever a religious character is
portrayed with their hands in prayer,
code “1” for “Hands in prayer” as well
as for “Other religious” and write down
who the religious character is (e.g.,
Jesus, Mother Mary, shepherd).
0 = not on gravestone
1 = engraved on gravestone
Cross
Includes Orthodox, Celtic, Methodist,
or classic crosses. Also includes
gravestones shaped like a cross,
although those are also coded “4”
under “Type of gravestone’.
0 = not on gravestone
1 = engraved on gravestone
Lamb
Uncommon, but comes in many
different forms: more simplistic or
detailed. Make sure to separate sheep
from lambs.
0 = not on gravestone
1 = engraved on gravestone
11
Angel
Or cherubs. Winged, (often) robed
icons, often with halos. Note that
angels are frequently depicted with
their hands in prayer, which is coded
for separately as 1 under Hands in
prayer.
0 = not on gravestone
1 = engraved on gravestone
Other religious
Among others, Gothic windows,
Death’s Head, fingers [pointing
toward/from Heaven], symbols of
divinity, statuaries, [Heaven’s] gates,
hands reaching down, effigies, and
symbols of churches.
0 = not on gravestone
1 = engraved on gravestone
12
Table S2
Tests for Intercoder Reliability
Variable
Measure
Intercoder Reliability
Gravestone Religiosity
Any Religious Imagery (Dummy)
Cohen’s Kappa
.89, p < .001
Sum of All Religious Imagery (Sum)
ICC (1,1)
.85, p < .001
Cross-Scripture (Dummy)
Cohen’s Kappa
.88, p < .001
Book-Praying-Angel (Dummy)
Cohen’s Kappa
.89, p < .001
Book -Praying-Angel (Sum)
ICC (1,1)
.91, p < .001
Book -Praying-Angel (Share)
ICC (1,1)
.86, p < .001
Book -Praying-Angel (Ratio)
ICC (1,1)
.89, p < .001
Further Gravestone Information
Reached Age
ICC (1,1)
.996, p < .001
Total Number Imagery (Sum)
ICC (1,1)
.83, p < .001
No Imagery at All (Dummy)
Cohen’s Kappa
.78, p < .001
Sum of Non-Religious Imagery (Sum)
ICC (1,1)
.77, p < .001
Type of Gravestone (Dummy)
Cohen’s Kappa
.89, p < .001
Number Persons in Grave (Sum)
ICC (1,1)
.96, p < .001
Floral Ornamentation (Dummy)
Cohen’s Kappa
.90, p < .001
Wheat Ornamentation (Dummy)
Cohen’s Kappa
.76, p < .001
13
Table S3
Descriptive Statistics
Variable
N
M
SD
Min
Max
Demographic Characteristics
Reached Age
6,400
78.77
15.24
.17
115.17
Female
6,400
.50
.50
0
1
Religious Imagery
Any Religious Imagery
6,400
.44
.50
0
1
Cross Symbol
6,400
.29
.45
0
1
Book Symbol
6,400
.10
.30
0
1
Praying Hands
6,400
.07
.26
0
1
Verbal Scripture
6,400
.07
.25
0
1
Angels
6,400
.02
.13
0
1
Fish
6,400
.001
.03
0
1
Dove
6,400
.01
.11
0
1
Lamb
6,400
.002
.04
0
1
Other Religious
6,400
.03
.16
0
1
Other grave information
Type = Upright Stone
6,400
.44
.50
0
1
Number Persons in Grave
6,400
1.66
.73
1
11
Ornamentation Present
6,400
.43
.50
0
1
Total Number Imagery
6,400
1.46
1.21
0
8
No Imagery at All
6,400
.25
.43
0
1
Any Non-Religious Imagery
6,400
.60
.49
0
1
Marriage Information
6,400
.15
.35
0
1
County Characteristics
Religious Condition
64
0.50
0.50
0.00
1.00
Urban Condition
64
0.50
0.50
0.00
1.00
Population Density
64
56.86
172.35
0.25
1285.51
% Religious Adherents
64
546.09
309.02
124.54
1099.39
% White
64
86.49
12.36
52.77
98.08
% Uninsured
64
11.48
4.58
4.80
26.10
% With Diploma
64
16.76
8.04
4.60
42.50
% In Poverty
64
10.40
4.37
3.90
21.60
% Vacant Housing
64
21.61
13.80
4.30
70.60
% Obese
64
29.16
4.74
13.97
37.27
% Smoking
64
17.57
3.60
10.53
25.40
% Excessive Drinkers
64
17.20
2.75
10.67
22.90
14
Robustness Checks
We conducted 14 robustness checks (Table S4) to scrutinize the robustness of our basic
findings (M1). First, we accounted for a wide variety of further gravestone information (R1),
including (a) total number of imagery elements, (b) absence of any imagery, (c) presence of non-
religious imagery, (d) marital status, (e) number of deceased people in grave, and (f) floral or
wheat ornamentation. Additionally, we controlled for various county-level characteristics (R2),
including (g) racial composition and population density (five-year estimates from the 2015
American Community Survey; United States Census Bureau, 2019), (h) estimates for the social
determinants of health (i.e., 2015 five-year estimates in rates of health insurance, educational
attainment, poverty prevalence, and vacant housing; Centers for Disease Control and Prevention,
2019), and (i) estimates for the most relevant behavioral health risk indicators (i.e., 2017 three-
year estimates in rates of obesity, excessive drinking, and tobacco use; County Health Rankings
& Roadmaps, 2019).
Second, some of our sampled counties were located in a certain proximity to each other.
To rule out that our results are an artefact of a superordinate level, we included state-level fixed
effects (R3).
Third, we tested whether gender (R4) and urbanity (R5) (i.e., our two stratification
criteria) moderated the interaction between gravestone religiosity and county religiosity. To this
end, we added three-way interaction terms consisting of Gravestone Religiosity × County
Religiosity × Stratification criteria to our models.
Fourth, we altered our age-related inclusion criterion. Specifically, we lowered it
(including all people down to minus one standard deviation below the mean age; R6), raised it
(75-99 years; R7), and included all people who reached an age of 70+ (including people who
reached an age of 100+; R8).
15
Fifth, we tested alternative empirical specifications of gravestone religiosity. To this end,
we measured gravestone religiosity not as the share of Book-Praying-Angel imagery, but also in
the dummy (R9), sum (R10), and ratio (R11) operationalizations.
Sixth, for some deceased people, their burial county might not be the cultural context that
was relevant throughout their lifetime. Therefore, we tried to rule out that our findings are driven
by wrongfully assigning deceased people to cultural contexts that were not relevant to them. To
do so, we rerun our analyses using only those 1,060 deceased people for whom additional
information on findagrave.com allowed us to assure that their places of death and burial were
identical (R12).
Seventh, deceased people who reached a higher age are also more likely (a) to belong to
an earlier birth cohort and/or (b) to have died in a later period. To rule out that our findings are
due to cohort and period effects, we controlled for the year of birth (R13) and the year of death
(R14).
2
Taken together, for all 14 robustness checks
3
, the main-text results remained conceptually
unchanged. In other words, our results held against a wide variety of alternative explanations,
different empirical specifications, and generalized across gender and living environments.
Finally, given that we ran 15 models (our main-text model and the 14 robustness checks
described in the previous section), we adjusted for the increased chance of Type I errors in
2
Note that R12 and R13 are particularly conservative robustness tests. Specifically, reducing our sample to only
1,065 observations in R12 greatly decreased the statistical power of our analyses. Likewise, accounting for year of
birth in R13 restricted the explainable variance in longevity to only 18 years (i.e., in what year between 2000 and
2018 a person died).
3
Note that all multi-level models reached convergence. However, models R3, R4, R8, R12, and R13 were singular
(i.e., variances of one or more linear combinations of effects were close to zero) and standard errors for the random-
effect parameters could not be calculated. Therefore, for each of these five models, we additionally tested a more
parsimonious alternative model (Bates, Kliegl, Vasishth, & Baayen, 2015) without random slopes. Those alternative
models lead to identical results.
16
multiple testing, using Simes Step-Up False-Discovery Rate (Yekutieli & Benjamini, 2001).
Table S5 shows the p-values for the original Gravestone Religiosity × County Religiosity cross-
level interaction and also the new, adjusted p-values (typically called q-values). Corroborating
the robustness of our findings, all q-values fell below the 95% significance threshold.
17
Table S4.
Multi-Level Regression Results for Gravestone Religiosity and Longevity alongside 14 Robustness Checks.
M1
R1
R2
R3
R4
R5
R6
R7
R8
R9
R10
R11
R12
R13
R14
DV = Reached Age
Base
Model
Grave
Controls
County
Controls
State
Fixed
Effects
Gender
Moderat.
Urbanity
Moderat.
Age>=
-1SD
Age >=75
Age all
>=70
Gravest.
Religiosity
= Dummy
Gravest.
Religiosity
= Sum
Gravest.
Religiosity
= Ratio
Burial =
Death
Place
Cohort
Effects
Period
Effects
Base Variables
Gravestone Religiosity
-0.00
0.01
0.02
0.03
0.23
0.08
-0.04
-0.02
0.12
0.07
0.02
0.05
-0.03
-0.13
0.05
(0.11)
(0.13)
(0.13)
(0.13)
(0.19)
(0.16)
(0.15)
(0.12)
(0.13)
(0.37)
(0.14)
(0.13)
(0.28)
(0.07)
(0.13)
County Religiosity
0.10
0.18
0.11
1.11**
0.10
0.32
0.02
0.20
0.10
-0.01
0.10
0.10
0.54
0.12
0.09
(0.15)
(0.15)
(0.19)
(0.39)
(0.22)
(0.21)
(0.21)
(0.15)
(0.20)
(0.19)
(0.18)
(0.19)
(0.33)
(0.11)
(0.19)
Gravestone X
0.30**
0.31**
0.32**
0.30**
0.42**
0.35**
0.28*
0.24**
0.29**
0.69*
0.28*
0.29**
0.70**
0.13*
0.30**
County Religiosity
(0.10)
(0.10)
(0.10)
(0.10)
(0.16)
(0.12)
(0.12)
(0.09)
(0.11)
(0.31)
(0.11)
(0.11)
(0.23)
(0.06)
(0.10)
Female
2.44***
2.06***
2.05***
2.06***
2.08***
2.05***
2.51***
1.76***
2.34***
2.06***
2.06***
2.06***
2.97***
0.86***
1.99***
(0.20)
(0.21)
(0.21)
(0.21)
(0.21)
(0.21)
(0.23)
(0.19)
(0.21)
(0.21)
(0.21)
(0.21)
(0.44)
(0.12)
(0.20)
Urban
0.11
0.09
0.03
0.33
0.03
0.14
-0.27
0.18
-0.12
0.07
0.08
0.08
0.10
0.26
-0.03
(0.30)
(0.29)
(0.36)
(0.35)
(0.36)
(0.35)
(0.41)
(0.29)
(0.38)
(0.36)
(0.35)
(0.36)
(0.59)
(0.21)
(0.37)
Grave Controls
Sum of All Imagery
-0.52***
-0.51**
-0.48**
-0.51**
-0.50**
-0.59***
-0.44**
-0.64***
-0.54**
-0.53**
-0.53**
-0.20
0.17
-0.58***
(0.16)
(0.16)
(0.16)
(0.16)
(0.16)
(0.17)
(0.14)
(0.16)
(0.17)
(0.17)
(0.16)
(0.33)
(0.09)
(0.16)
No Imagery
0.14
0.17
0.22
0.15
0.17
0.47
0.17
0.39
0.10
0.09
0.13
1.42
0.04
0.18
(0.40)
(0.40)
(0.40)
(0.40)
(0.40)
(0.44)
(0.36)
(0.41)
(0.37)
(0.37)
(0.38)
(0.83)
(0.23)
(0.39)
Non-Relig. Imagery
-0.51
-0.50
-0.49
-0.50
-0.49
-0.54
-0.42
-0.28
-0.55
-0.55
-0.53
0.21
0.08
-0.56
(0.35)
(0.35)
(0.35)
(0.35)
(0.35)
(0.39)
(0.32)
(0.36)
(0.33)
(0.34)
(0.33)
(0.73)
(0.20)
(0.35)
Marriage Mentioned
-0.86**
-0.85**
-0.94**
-0.85**
-0.88**
-0.77*
-0.83**
-0.97**
-0.86**
-0.86**
-0.85**
-0.06
0.30
-0.98**
(0.31)
(0.32)
(0.32)
(0.32)
(0.32)
(0.35)
(0.29)
(0.33)
(0.32)
(0.32)
(0.32)
(0.65)
(0.18)
(0.32)
Persons in Grave
0.51***
0.50***
0.49***
0.49***
0.51***
0.82***
0.30**
0.50***
0.49***
0.49***
0.49***
0.18
0.13*
0.50***
(0.11)
(0.11)
(0.11)
(0.11)
(0.11)
(0.12)
(0.10)
(0.12)
(0.11)
(0.11)
(0.11)
(0.24)
(0.06)
(0.11)
Type of Gravestone
-0.17
-0.16
0.07
-0.15
-0.22
-0.24
-0.03
-0.10
-0.12
-0.12
-0.13
0.09
0.17
-0.20
(0.24)
(0.25)
(0.25)
(0.25)
(0.25)
(0.28)
(0.22)
(0.26)
(0.25)
(0.25)
(0.25)
(0.53)
(0.14)
(0.25)
Ornamentation
0.70**
0.72**
0.76***
0.70**
0.75***
0.76**
0.47*
0.86***
0.72**
0.72**
0.72**
0.79
-0.07
0.79***
(0.22)
(0.22)
(0.22)
(0.22)
(0.22)
(0.25)
(0.20)
(0.23)
(0.22)
(0.22)
(0.22)
(0.47)
(0.13)
(0.22)
County Controls
Share Whites
-0.11
-0.09
-0.11
-0.08
-0.12
-0.06
-0.12
-0.14
-0.16
-0.14
0.13
-0.03
-0.11
(0.21)
(0.26)
(0.21)
(0.20)
(0.24)
(0.17)
(0.22)
(0.21)
(0.20)
(0.21)
(0.35)
(0.12)
(0.22)
Population Density
-0.28
0.25
-0.28
-0.27
-0.41*
-0.21
-0.33
-0.28
-0.28
-0.27
-0.41
0.11
-0.33
(0.17)
(0.23)
(0.17)
(0.17)
(0.19)
(0.13)
(0.18)
(0.17)
(0.16)
(0.16)
(0.39)
(0.10)
(0.18)
Uninsured
-0.19
-0.53
-0.20
-0.24
-0.19
-0.20
-0.15
-0.23
-0.28
-0.27
-0.15
-0.06
-0.20
(0.22)
(0.39)
(0.22)
(0.21)
(0.24)
(0.17)
(0.23)
(0.22)
(0.21)
(0.21)
(0.37)
(0.12)
(0.22)
18
Diploma
0.04
0.01
0.04
0.01
-0.01
0.06
0.07
0.04
0.04
0.04
-0.13
0.04
0.04
(0.17)
(0.17)
(0.17)
(0.16)
(0.19)
(0.13)
(0.18)
(0.17)
(0.16)
(0.16)
(0.32)
(0.10)
(0.17)
Poverty
-0.12
-0.34
-0.13
-0.07
-0.33
0.01
-0.20
-0.14
-0.14
-0.14
0.22
-0.07
-0.12
(0.27)
(0.28)
(0.27)
(0.26)
(0.31)
(0.22)
(0.29)
(0.27)
(0.27)
(0.27)
(0.49)
(0.16)
(0.28)
Vacant Housing
-0.01
-0.02
-0.02
0.05
-0.18
-0.04
-0.08
-0.01
-0.02
-0.02
0.10
0.04
-0.02
(0.19)
(0.20)
(0.19)
(0.19)
(0.21)
(0.15)
(0.20)
(0.19)
(0.19)
(0.19)
(0.32)
(0.11)
(0.20)
Obesity
0.05
-0.13
0.06
0.11
-0.07
0.08
-0.04
0.06
0.08
0.08
0.37
0.02
0.03
(0.25)
(0.32)
(0.25)
(0.24)
(0.28)
(0.20)
(0.26)
(0.25)
(0.24)
(0.24)
(0.44)
(0.14)
(0.26)
Smoking
-0.29
0.45
-0.29
-0.23
-0.20
-0.25
-0.22
-0.27
-0.25
-0.27
-0.75
-0.07
-0.29
(0.25)
(0.29)
(0.25)
(0.25)
(0.29)
(0.21)
(0.27)
(0.26)
(0.25)
(0.25)
(0.52)
(0.15)
(0.26)
Excessive Drinking
-0.06
-0.02
-0.07
-0.06
-0.20
-0.00
0.06
-0.10
-0.08
-0.10
0.07
-0.27*
-0.02
(0.20)
(0.29)
(0.20)
(0.19)
(0.23)
(0.16)
(0.21)
(0.20)
(0.20)
(0.20)
(0.35)
(0.12)
(0.21)
State Fixed Effects
YES
Three-Way Interactions
Gravestone X County
-0.17
X Female
(0.21)
Gravestone X County
-0.23
X Urban
(0.24)
Cohort, Period Effects
Year of Birth
-0.73***
(0.01)
Year of Death
0.17***
(0.02)
Constant
83.24***
83.61***
83.61***
84.84***
83.61***
83.53***
82.18***
85.11***
83.66***
83.61***
83.62***
83.60***
82.32***
1482.24***
-257.62***
(0.24)
(0.39)
(0.39)
(1.04)
(0.39)
(0.39)
(0.44)
(0.34)
(0.41)
(0.40)
(0.38)
(0.38)
(0.82)
(13.78)
(43.86)
Intercept Variance
0.80
0.68
0.50
0.00
0.50
0.42
0.66
0.03
0.61
0.67
0.48
0.48
0.09
0.17
0.60
Slope Variance
0.00
0.00
0.01
0.00
0.00
0.00
0.08
0.00
0.01
0.46
0.04
0.03
0.04
0.00
0.01
Within Between Variance
50.29
49.30
49.30
49.09
49.23
49.29
65.80
36.45
54.16
49.25
49.30
49.29
47.00
15.99
48.65
Observations
4,946
4,946
4,946
4,946
4,946
4,946
5,339
4,440
5,071
4,946
4,946
4,946
1,065
4,946
4,946
Standard errors in parentheses
* p < .05, ** p < .01, *** p < .001
19
Table S5
FDR-Corrected q-Values to Account for Multiple
Testing
Model
p-value
FDR-corrected
q-value
M1
.004
.008
R1
.002
.008
R2
.002
.008
R3
.003
.008
R4
.009
.012
R5
.003
.008
R6
.022
.025
R7
.009
.012
R8
.006
.011
R9
.024
.026
R10
.013
.016
R11
.007
.011
R12
.002
.008
R13
.031
.031
R14
.003
.008
20
Other Supplementary Materials
Gravestone data set (gravestone_data.csv).
Gravestone analysis script (gravestone_script.do)
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Personality and Social Psychology
This content is subject to copyright. Terms and conditions apply.
