Church Attendance in the United States 89
church attendance is not tenable. It is, after all, themorehighly educated who are most
likely to overreport other behaviors, such as voting. Moreover, the operationalization
of “skeptical professional” used by the critics includes athletes, artists, television an-
nouncers, and university professors, among others. This eclectic group holds no com-
mon disposition or training that would lead them to be less likely to exaggerate their
church attendance. More generally, it is not credible to rely on comparisons of self-
reports among subgroups of survey respondents rather than on comparisons between
self-reports and an external criterion such as head counts.
A fourth criticism of the Hadaway et al. claim that weekly church attendance is sub-
stantially lower than 40 percent was that their results were based on aggregate rather
than individual-level data (Hout and Greeley 1998). Hadaway et al., after all, based
their conclusions on comparisons between survey data and head-count data that did
not permit any direct examination about which specific individuals might be overre-
porting their own attendance. It would be more persuasive if one could compare the
actual church attendance of the exact same individuals who claimed in a survey to
have attended. In a different study, Marler and Hadaway did just this (1999). After con-
ducting telephone interviews of adults belonging to a single large evangelical church,
asking them if they had attended church services during the previous week, Marler and
Hadaway matched each individual’s response to attendance sheets from the previous
week kept by the church. The result: Only 115 of the 181 people who claimed to have
attended church actually had attended. Although approximately 60 percent of these
people said that they had attended, only 38 percent actually had attended.
Evidence from other studies consistently supports the conclusion that religious ser-
vice attendance is substantially overreported in conventional surveys. Marcum (1999)
compares attendance reports based on head counts within Presbyterian congregations
to self-reports obtained through conventional survey designs. He finds that the self-
reports produce attendance levels almost double what they actually are: seventeen peo-
ple report attending for every ten that actually are there. Hadaway and Marler (1997b)
find substantial overreporting when they compare surveys to actual counts of attending
Catholics in a Canadian county. As far as we know, no researcher who has compared
self-reported to actual attendance has found something other than that the latter is
much smaller than the former.
Other researchers have investigated this issue by using innovative survey techniques
designed to minimize overreporting. Presser and Stinson (1998) examine data from
studies in which people are asked to complete diaries concerning their daily activi-
ties. Although this method still relies on respondents’ self-reports, it is likely to reduce
overreporting for two reasons. First, the respondent is not engaged in face-to-face inter-
action with an interviewer, which ought to reduce respondents’ propensity to engage
in impression management. Second, the respondent is not made aware of the fact that
religious participation is of particular interest to the researchers, making the issue of
religion much less salient to informants and reducing the pressure to conform to per-
ceived social norms regarding religious participation. Using this arguably more valid
method of measuring church attendance, Presser and Stinson find that claimed rates of
church attendance are approximately one-third lower than with the traditional survey
approach. Similar results were found in a study of British respondents in which tra-
ditional surveys predicted a church attendance rate of about 21 percent and the time
diary approach yielded a lower estimate of 14 percent (Hadaway and Marler 1997a).