What the polls do show: toward enhanced survey readings of religion in Canada.
Bibby, Reginald W. ; Grenville, Andrew
IN RECENT YEARS, RESEARCHERS, private pollsters, and Statistics
Canada have been reporting levels of religious identification and
participation that frequently have been at odds with one another. For
example, the increasing number of surveys by private polling companies
has produced results on religious identification and participation that
have been significantly below those reported by Statistics Canada's
General Social Surveys as well as the 2011 National Household Survey
(NHS).
By way of illustration, a 2011 Ipsos survey found that 32 percent
of Canadians had no religion, compared to 24 percent for the NHS (Dueck
2011). The pollster further maintained that 29 percent of Canadians were
Catholic, well below the 39 percent figure of the NHS. A 2013 Angus Reid
survey claimed that the no religion figure stood at 36 percent (Hiemstra
2015).
These disparities in survey samples are significant when it comes
to exploring any number of correlates of religious identification, such
as service attendance and beliefs and practices. For example, in its
2011 survey, Ipsos reported that 13 percent of Canadians were weekly
attenders, while the Angus Reid 2013 survey pegged attendance at a
similar level of 12 percent. Ipsos further claimed that 53 percent of
Canadians believed in God, yet--just a year later--Leger maintained the
level was 67 percent (Jedwab 2012). The 2011 GSS figure for weekly
attendance was 18 percent; in 2012 it was 17 percent. Belief in God
items in most other Canadian national surveys has never produced figures
below 75 percent (see, e.g., Bibby 2011:48-50).
It is extremely important to understand why these differences in
survey research findings on religion exist. Otherwise, survey research
is adding little to a clear understanding of religion and religious
trends in Canada. The primary focus of our exploration of the topic was
service attendance. But we also gathered data that allowed us to examine
the important related issue of religious identification.
There would seem to be two main reasons why the discrepancies in
attendance and identification findings are occurring. The first is
because of different attendance measures. The second is because of the
use of different samples. Of considerable importance, they are
interrelated.
POSSIBILITY #1: DIFFERENCES ARE DUE TO DIFFERENT MEASURES OF
ATTENDANCE
One explanation for the differences in attendance results is that
researchers are using different measures. Initially, that is what we
assumed was the heart of the disparity problem.
Our review of surveys dating back through Gallup's earliest
1945 survey in Canada revealed that three primary items have been
utilized to measure service attendance:
1. "Did you attend a religious service in the past seven
days?"
2. "How often do you attend religious services?"
3. "How often did you attend religious services in the last 12
months?"
A slightly edited version of this third item has taken rites of
passage ceremonies into account, a variation that is particularly
important because it has been used in widely cited Statistics
Canada's General Social Surveys in recent years. It reads:
"Other than on special occasions such as weddings, funerals or
baptisms, how often did you attend religious services or meetings in the
last 12 months?"
While the response options for Gallup's "seven-day
item" have been simply "Yes" or "No," the
response options for the other two items have included variations around
the four categories of "weekly," "monthly,"
"yearly," and "never":
1. Weekly ... several times a week, weekly, and at least once a
week.
2. Monthly ... about once a month, two-three times a month.
3. Yearly ... several times a year, about once a year, seldom,
rarely, less than once a year.
4. Never.
We therefore wanted to compare results when (a) the three different
items and (b) the different combinations of response options are offered
to survey participants. In addition to looking at weekly attendance, we
wanted to keep a close eye on monthly plus attendance--a level that
might increasingly be "the new norm," given the general
consensus that accelerated choices and demands make weekly attendance
today increasingly difficult, even for those committed to it (see, e.g.,
Bibby 2011:27; Hadaway and Marler 1998:475). In addition, there seemed
to be merit in exploring the assumed spike in seasonal attendance (e.g.,
at Christmas and Easter) in relationship to regular attendance.
We wanted to explore question and response variations within a
fairly tight time period in order to try to control for the possible
influence of events on attendance. Therefore, the questions and response
options were administered in three successive Canadian national omnibus
surveys in March of 2012. In addition, we were able to explore the
possible "seasonal effect" on attendance by looking at
reported attendance via a fourth national survey in May of 2012--a month
following Easter Sunday (April 8). The first three samples consisted of
around 1,000 people each, while the fourth sample was considerably
larger (52,160), making potential in-depth analyses possible. The
surveys were carried out by one of Canada's leading research
companies, Angus Reid Global, a division of Vision Critical, using
online, panel samples.
Three different attendance questions were asked in each of the
three March surveys--a total of nine items altogether. "Seven
day," "past month," and "last December" items
were each run twice to provide a modest reliability check on the three
items (six in all),while three additional items were used to probe (7)
general attendance, (8) attendance in the last 12 months, and (9)
attendance in the last 12 months minus special occasions.
Measures of Who Is Attending Do Not Matter Much
What is immediately apparent is that the results were highly
consistent regardless of the self-report measure used (Table 1).
* When people were asked how often they attended services, the
weekly figure came in at around 11 percent--similar to what they
reported when asked to recall how often they actually attended in the
last 12 months. Introducing the rites of passage qualifier made little
difference. It seems most "rites only" attenders do not see
such appearances as "counting." Further, in response to being
asked whether or not they attended a service in the past 7 days, some 15
percent indicated they did, a figure that is consistent with the 11
percent who attended every week being joined by others who attended
two-or-three times a month or less.
* Similar consistency is apparent for the 22 to 23 percent who
reported they attended a service in the past month, along with the 32
percent who said they had attended a service last December.
Measures of Who Is Not Attending Do Matter Much
So far, findings across measures seem consistent. About the only
variant finding was that when people were asked how often they attend
services--something of a statement of intentions (Sample A), they were
slightly less likely to indicate that they "never" attend
services (38 percent), compared to what people recalled they actually
did when they looked back at the previous 12 months (Sample B, 45
percent and Sample C, 46 percent). But note that, here, it was not the
active attenders who overstated their participation but rather the
infrequent attenders.
The reluctance of infrequent attenders to describe themselves as
"never" attending was also accidentally uncovered in our large
Sample D. One of the main reasons for repeating the "past twelve
months minus rites of passage" item in this large survey is that we
wanted to use identical response options to those used in the 2012
General Social Survey (GSS) (our Sample C response options were slightly
different). In addition, the size of Sample D and the inclusion of
variables that included religious identification made it possible to
examine some issues beyond attendance.
In comparing our Sample D with our previous three samples, we found
little difference in levels reported for "a few times a year"
or less. However, when we did not provide a response option of
"Less than once a year" between "At least once a
year" and "Not at all" (as we did in Samples A, B, and
C), we found that the "At least once a year" figure increased
from about 7 to 11 percent while the "Not at all" level jumped
from around 45 percent to almost 60 percent (Table 2).
Here again, changes in response options seemed to most noticeably
influence what people who are not actively involved in religious groups
reported--not people who are the most active. Given some individuals
attend very seldom, many seem to need an option between "once a
year" and "never." If they don't have it, they
"round off' their attendance to "never." Simply put,
"hardly ever" is not the same as "never."
POSSIBILITY #2: DIFFERENCES ARE DUE TO DIFFERENT SAMPLES
Our analysis to this point suggests that the differences in
attendance levels reported in surveys are not due to the use of
different measures. We found very similar results regardless of the
measure used--except in the case of isolating the percentage of people
who "never" attend services for the reasons indicated.
Ironically, that measurement limitation has virtually never been an
issue in discussions about attendance, since researchers typically
emphasize the percentage of people who attend, and give minimal
attention to those who "never" attend. Future work on the
"never" category needs to be cognizant of the measurement
issue that we have uncovered.
So if measurement differences do not account for the
attendance-level discrepancies, what does? We suggest that the
differences are due primarily to the use of different samples.
Statistics Canada's General Social Surveys typically include a
greater number of Roman Catholics and fewer people with No Religion than
the samples of private pollsters such as Ipsos and Angus Reid. The
predictable result is a higher level of attendance. That assertion is
readily documented by doing a simple weighting of the large Angus Reid
"Sample D" for May of 2012 for Catholic and No Religion
participants so that the sample conforms to the 2012 GSS (see Table 3).
That adjustment results in a national increase in weekly attendance
for the Angus Reid sample from 12 to 17 percent. In the case of
Catholics, attendance rises from 13 to 19 percent--an increase from 19
to 25 percent outside Quebec and 7 to 9 percent in Quebec. As would be
expected, weighting to reduce the proportion of No Religion respondents
from 42 to 22 percent to match GSS figures does not affect the
percentage of people in that category who attend services.
The problem that existed with the deletion of religion from the
long-form census in 2011 was the lack of a gold standard for religious
identification that we have been talking so much about. An initial sense
of comfort from the fact that the GSS sample characteristics are similar
to those of the 2011 NHS (see Table 3) soon passed for well-known
reasons. Because of its voluntary nature, the NHS did not provide the
elusive absolute measurement for the population, against which
researchers could adjust their samples. As a result, in recent years
pollsters typically have continued to weight samples for variables such
as region, age, and gender. But they did not weight for religion,
primarily because of the absence since the 2001 Census of that
much-needed standard.
It is highly significant that Statistics Canada (2013) has readily
acknowledged that some parts of the population could be underrepresented
in the NHS replacement for the 2011 Census. We can safely assume that
those missing included large numbers of immigrants, Aboriginals, poorer
people, and individuals who are highly geographically mobile. The latter
would undoubtedly include many younger adults who--additionally--have
been increasingly unlikely to participate in voluntary surveys of any
kind (see, e.g., Bibby 2006:66). Thankfully, religion is being restored
to the long-form census in 2021.
Likewise, positively speaking, the Angus Reid Forum that provided
the samples from which our data were sourced allowed the company to be
the most accurate of nine pollsters in predicting the May 2011 federal
election; its sister American company was the most accurate of 28 survey
firms in predicting the 2012 Presidential election (Angus Reid Public
Opinion 2010). That said, substantively speaking, the Reid panels in
both countries suffered from the exclusion of those not literate in
English (or French in Canada), and those without access to computers.
THE ONGOING CHALLENGE OF OBTAINING REPRESENTATIVE SAMPLES
To sum up, the sampling dilemma that every survey researcher faces
in Canada today is twofold. First, highly representative national
samples are extremely hard to obtain because of low participation rates,
especially for younger adults. Second, there no longer is a census-like
gold standard for determining the religious characteristics of the
population, allowing researchers to assign weights with confidence in
correcting for unrepresentative samples.
When our government agency can only get two in three people to
participate in its major national surveys (Eagle 2011), and private
companies such as Angus Reid are sampling from large (some 150,000) but
voluntary panels that can be affected by linguistic and technological
factors, one realizes how elusive representative samples have become.
It's not that the market research industry has given up. Work
by the Advertising Research Foundation (2009) and others (Grenville
2012) have attempted to measure the variability of answers by sample
source. One way that those who follow election polling have begun to
attempt to deal with sampling problems is to agglomerate numerous polls
and use predictive modeling to estimate the probability of a forecast
being correct. Statistician Nate Silver is perhaps the best known of
those using this approach, having gained notoriety for correctly
predicting the outcome of the 2012 U.S. Presidential Election in all 50
states. He has detailed his methods in his bestselling book, The Signal
and the Noise (Silver 2012). Such an approach could serve as a useful
model for future efforts to manage sample source variability in
estimating "true" levels of church attendance.
And it is not as if reputable organizations such as the Angus Reid
Institute is not cognizant of the need to do everything possible by way
of weighting samples to try to offset some of their inadequacies. For
example, in a March 2015 online national survey on religion involving
more than 3,000 Canadians who are part of the Angus Reid Forum, the
sample was carefully weighted not only for variables such as region,
age, and gender, but also for religious identification, making use of
the most recently available, 2012 GSS. The assumption was that the GSS
provided at least "a silver if not gold standard."
What that exercise readily demonstrated is that the best efforts
possible need to be made to try to match samples with population
estimates. In the case of religion, identification, attendance, beliefs,
practices, and experiential claims, all increased accordingly as the
sample was adjusted to match GSS estimates (see Table 4).
CONCLUSION
Our foray into trying to understand the diverse survey findings on
service attendance in Canada leads us to believe that the problem does
not so much lie with measurement items as it does with the difficulty of
obtaining samples that are representative of the national population.
The issue has been given considerable attention recently by prominent
Princeton sociologist Robert Wuthnow (2015), who maintains the problem
of unrepresentative samples is so serious as to call into question the
value of current survey research into religion beliefs and behavior.
There also has been considerable debate among social scientists in
the past two decades or so about the sheer accuracy of respondent
reports when it comes to service attendance--even if the samples are not
being questioned. The controversy was stimulated in large part by the
1993 publication of the seminal paper, "What the Polls Don't
Show," by Kirk Hadaway, Penny Marler, and Mark Chaves in the
American Sociological Review. Their paper focused on the conventional
and extensively used "last seven days" Gallup attendance item
that dated back to the 1930s. Using count-based attendance data, they
maintained that Gallup's self-reported figures in the 40 to 45
percent range over the years were probably about double the actual
attendance levels of around 20 percent--a discrepancy they have held to
in subsequent work (e.g., Hadaway and Marler 2005). Around the same
time, Hadaway and Marler (1997) used data gathered in Oxford Country in
southern Ontario to argue that a large attendance gap also exists north
of the border.
More recently, University of Massachusetts sociologist Philip
Brenner (2011, 2012), in an analysis extraordinary in scope, examined
the gap between reported weekly attendance and time diary data in 14
countries--including Canada. He claimed that the "raw gap"
between poll figures and GSS diary data here was about 10 percentage
points around 1980 (30 versus 20 percent) and 7 percentage points in
2005 (19 versus 12 percent). Our examination of the 2010 GSS survey led
us to conclude that self-reported attendance remained at about 18
percent, while the diary data figure came in at 14 percent--a gap of
only about 4 percentage points.
Hadaway and his associates, along with Brenner, have maintained
that the overreporting of service attendance is due to people
interpreting the question as being more about religious identity rather
than actual attendance. As Brenner (2011) put it, "The respondent
applies a pragmatic interpretation, influenced by a desire to
self-present in an identity-consistent light" (p. 21).
Significantly, Hadaway et al. (1993) noted in their original ASR
article that some "individuals in others countries [might be] less
likely to overreport their church attendance than are Americans"
(p. 749). They also predicted that "as growing numbers of
individuals in countries such as Australia shed their nominal church
identities, the inclination to report regular attendance could be
expected to decrease" (Hadaway et al. 1993:749).
We all are well aware that, in conversations and surveys, people
will frame their responses to fit what they want us to hear (social
desirability) as well as how they view themselves (identity). But if
there is little at stake, including our religious identities, there is
little reason to distort what people tell each other. We maintain that,
in Canada, unlike the situation in the United States and some other
countries where religion is pervasive, there is little to gain socially
and psychologically by overreporting religious service attendance.
Since at least the 1980s, Canada has been a highly polarized
country when it comes to religion. About 30 percent of Canadians
currently value faith while a growing core of about the same size
explicitly say that faith is not important to them--with the remaining
40 percent or so constituting something of "an ambivalent
middle" (Bibby and Reid 2016). In addition, the rules of pluralism
that have become so pervasive in Canada provide people with
unprecedented license to call things the way they are when it comes to
faith--to be religious or not be religious, with limited stigma either
way.
In view of such a cultural environment, there is minimal pressure
to exaggerate religious involvement--apart from a small number of
possible geographical and religious group exceptions. We therefore
believe that if Canadians are asked clear, noninvasive questions about
service attendance, they will come through with answers that are in
touch with their behavior. Put succinctly, as our four survey results
suggest, if we ask fair questions, we will get fair and consistent
responses.
It is important to note that further reflections on the differences
in survey findings for Statistics Canada and private pollsters need to
take into account the possible effects of data collection methods.
Historically, Statistics Canada has relied heavily on the use of
self-administered questionnaires and telephone interviews. Today,
polling companies that are carrying out national surveys are typically
and sometimes exclusively making use of online surveys with samples that
increasingly are drawn from a large number of panel participants.
One recent exploration of the influence of using the telephone
versus the Internet to collect data found that Americans exaggerate
service attendance and religion's salience over the telephone, but
there is no such inclination when religious identification and beliefs
are involved (Cox, Jones, and Navarro-Rivera 2014). If the same pattern
holds in Canada--and that remains an empirical question--it might help
to account for some of the differences between the GSS and private polls
on attendance, but not for the differences on identification. There
appears to be more involved than data collection methods.
We conclude that, currently, the primary measurement hurdle does
not so much lie with the responses as with the respondents. We simply
are not sure if we are speaking with an adequate cross-section of the
population. Despite the consternation of people such as Hadaway and
Brenner, we think that "What the Polls Do Show" depends not so
much on the people as it does on the pollster. Representative samples
have become elusive dreams. Nonetheless, if we are to gain comprehensive
insight into religious thinking and behavior, it is essential that such
samples continue to be vigorously pursued.
REGINALD W. BIBBY
University of Lethbridge
ANDREW GRENVILLE
Vision Critical
References
Advertising Research Foundation. 2009. Foundations of Quality Brief
3: Inter Study Comparability and Benchmark Analysis. New York: ARF.
Angus Reid Public Opinion. 2010. "Electoral Forecasts Since
2007." Angus Reid's Record in the U.S. and Canada. Toronto:
Vision Critical, Angus Reid. Retrieved May 23, 2014
(http://www.angusreidglobal.com/wp-content/
uploads/2010/10/Electoral-Forecasts-Since-2007.pdf).
Bibby, R.W. 2006. The Boomer Factor. Toronto: Bastian Books.
Bibby, R.W. 2011. Beyond the Gods and Back. Lethbridge: Project
Canada Books.
Bibby, R.W. and A. Reid. 2016. Canada's Catholics. Toronto:
Novalis.
Brenner, P.S. 2011. "Exceptional Behaviour or Exceptional
Identity?" Public Opinion Quarterly 75(1): 19-41.
Brenner, P.S. 2012. "Identity as a Determinant of the
Overreporting of Church Attendance in Canada." Journal for the
Scientific Study of Religion 51(2):377-85.
Cox, D., R.P. Jones and J. Navarro-Rivera. 2014. "I Know What
You Did Last Sunday: Measuring Social Desirability Bias in Self-Reported
Religious Behavior, Belief, and Identity." Washington, DC: Public
Religion Research Institute.
Dueck, L. (executive producer). 2011. Religious Identity: Ipsos
Canadian Online Omni (television broadcast) August 29-September 6, 2011,
Toronto, Context.
Eagle, D. 2011. "Changing Patterns of Attendance at Religious
Services in Canada, 1986-2008." Journal for the Scientific Study of
Religion 50(l):187-200.
Grenville, A. 2012. "Does the Sample Source Matter: A
Comparison of Some Leading Sample SOURCES." Vision Critical
University Whitepaper. Retrieved May 22, 2014
(http://vcu.visioncritical.com/wp-content/uploads/2012/08/WPR_Does-thesample-source-matter.-Comparison-of- leading-sample-sources_24.08.12.pdf).
Hadaway, C.K. and P.L. Marler. 1997. "Do Canadians Over-Report
Church Membership and Attendance? A Case Study of Religion in a Canadian
County." Presented at the annual meeting of the Association for the
Sociology of Religion, August 8-10, Toronto.
Hadaway, C.K. and P.L. Marler. 1998. "Did You Really go to
Church this Week? Behind the Poll Data." Christian Century, May 6,
pp. 472-475.
Hadaway, C.K. and P.L. Marler. 2005. "How Many Americans
Attend Worship Each Week? An Alternative Approach to Measurement."
Journal for the Scientific Study of Religion 44(3):307-22.
Hadaway, C.K., P.L. Marler and M. Chaves. 1993. "What the
Polls Don't Show: A Closer Look at U.S. Church Attendance."
American Sociological Review 58(6):741-52.
Hiemstra, R. 2015. "Church and Faith Trends."
Presentation, Heritage College and Seminary, Cambridge, ON. Retrieved
July 28, 2015 (http://www.heritagecambridge.com/wp-content/
uploads/2015/03/RickHiemstra_Main.pdf).
Jedwab, J. 2012. "In God We Canadians Trust?" Ottawa:
Association for Canadian Studies, April.
Pew Research Center. 2013. "Canada's Changing Religious
Landscape." June 27. Retrieved December 13, 2015
(http://www.pewforum.org/2013/06/27/canadas-changing-religiouslandscape).
Silver, N. 2012. The Signal and the Noise: Why So Many Predictions
Fail--But Some Don't. New York: Penguin.
Statistics Canada. 2011. "NHS Profile, 2011--About the
Data." Retrieved January 15, 2014
(http://wwwl2.statcan.gc.ca/nhs-enm/2011/dp-pd/prof/help-aide/aboutdataaproposdonnees).
Statistics Canada. 2013. "National Household Survey: User
Guide." Retrieved July 28, 2015
(http://www12.statcan.gc.ca/nhs-enm/2011/ref/nhs-enm
_guide/index-eng.cfm).
Wuthnow, R. 2015. "In Polls We Trust." First Things,
August. Retrieved July 28, 2015
(http://www.firstthings.com/article/2015/08/in-polls-we-trust). [Also
worth reading is a response by the Director and Associate Director of
Research at the Pew Research Center-- Alan Cooperman and Greg Smith,
First Things, July 29.
http://www.firstthings.com/webexclusives/2015/07/response-from-
pew-research-center-to-in-polls-we-trust-along with Wuthnow's
response: http://www.firstthings.co/web-exclusives/2015/07/robert-wuthnowreplies-to-the-pew-research-center.]
This is a revision of a paper presented at the Annual Meeting of
the Canadian Sociological Association Victoria, BC, June 6, 2013.
Reginald W. Bibby, Board of Governors Research Chair, Department of
Sociology, University of Lethbridge, Lethbridge, AB, Canada T1K 3M4.
E-mail: bibby@uleth.ca
Table 1
The Three Omnibus Surveys
Sample A: Omni, March 21, Sample B: Omni, March 23,
2012 (n = 1,009) 2012 (n = 1,005)
How often did you attend
How often do you attend religious services
religious services? in the last 12 months?
Percentage Percentage
At least once a week 11 At least once a week 12
Two-to-three times a 4 Two-to-three times a 4
month month
At least once a month 3 At least once a month 3
A few times a year 13 A few times a year 13
At least once a year 7 At least once a year 7
Less than once a year 24 Less than once a year 16
Never 38 Never 45
Did you attend a religious Did you attend a religious
service in the past 7 days? service in the past 7 days?
Percentage Percentage
Yes 15 Yes 16
No 85 No 84
Did you attend a religious Did you attend a religious
service in December? service in the past month?
Percentage Percentage
Yes 32 Yes 23
No 68 No 77
Sample C: Omni, March 30,
2012 (n = 1,060)
Other than on special
occasions such as weddings,
funerals, or baptisms, how
often did you attend
religious services or
meetings in the last 12
months?
Percentage
At least once 11
a week
Two-to-three 5
times a month
At least once 3
a month
A few times 12
a year
At least once 7
a year
Less than 16
once a
year
Never 46
Did you attend a religious
service in the past 7 days?
Percentage
Yes 22
No 78
Did you attend a religious
service in December?
Percentage
Yes 32
No 68
Table 2 Omni May 2012 and GSS 2012 (Ns = 52,160 and 19,284,
Respectively)
"Other than on special occasions such as weddings, funerals or
baptisms, how often did you attend religious services or meetings
in the past 12 months?"
Sample D Omni GSS
Percentage Percentage
At least once a week 12 19
At least once a month 5 12
A few times a year 13 11
At least once a year 11 16
Not at all 59 42
Table 3 GSS, NHS, and Private Pollster Samples and Results
NHS Ipsos Reid GSS
2011 2011 2012
Percentage Percentage Percentage Percentage
of pop of pop of pop weekly+
Catholic 39 29 39 17
Outside Quebec 21 15 22 24
Quebec 18 14 17 9
Protestant 28 30 32 26
Other faith 9 9 7 24
No religion 24 32 22 <1
Totals 100 100 100 19
Angus Reid
Sample D 2012
Unweighted for religion Weighted for religion
Percentage Percentage Percentage Percentage
of pop weekly+ of pop weekly+
Catholic 27 13 38 19
Outside Quebec 13 19 23 25
Quebec 14 7 15 9
Protestant 28 26 37 26
Other faith 3 17 3 16
No religion 42a <1 22 <1
Totals 100 12 100 17
Note: Unweighted includes no religion 20 percent, atheist 10
percent, agnostic 6 percent, and spiritual 6 percent; weighted
includes no religion 10 percent, atheist 5 percent, agnostic 4
percent, and spiritual 3 percent.
Table 4
Correlates of Religious Identification: 2015 ARI, Unweighted,
and Weighted Data
Unweighted Weighted
percentage percentage
Identification Catholic 29 38
Protestant 31 33
Other faith 4 5
No religion 36 22
Attendance Weekly 13 16
Monthly+ 20 23
Belief God 68 73
Life after death 61 66
Divinity Jesus 54 59
Practice Prayer weekly 34 38
Scripture monthly 17 19
Experience Monthly+ 29 32
God Have ever 43 47
Source: 2015 Angus Reid Institute (ARI) Religion Survey.
