How Many People Attend Religious Services? Part 2

Part 1 discussed two recent polls that gave different answers for how often people say they attend religious services. Using aggregated data from Gallup polls between 2021 and 2023, Jones (2024) estimated that 30% of American adults attend religious services every week or almost every week. Benz et al. (2023) estimated from a 2022 poll conducted for the American Enterprise Institute (AEI) that 24% of American adults attend “regularly” (at least a few times per month).

These estimates, however, might be affected by nonresponse (perhaps religious service attendees are more or less likely to respond to surveys than nonattendees), how the questions are worded, or people’s tendency to give answers that cast them in a favorable light (called social desirability bias). How can we tell whether nonresponse or social desirability affect the estimates? We cannot assess this using the polls themselves, since we do not know the “true” attendance frequency for the respondents (as opposed to how often they say they attend) and we do not know how often survey nonrespondents attend services since they do not tell us. We can, however, use information from other sources to explore these issues. In this post, I’ll look at two other sources of data: religious congregations and the American Time Use Survey.

Data from Religious organizations

Many religious organizations collect data on membership and religious service attendance, and a compilation of these statistics could provide an alternative measure to the self-report statistics from surveys. Each decade, in the same year as the U.S. Census of Population, the research team for the U.S. Religion Census asks religious groups around the country for statistics by county on congregational counts and attendance. In 2020, the census identified 372 religious bodies, 356,739 congregations, and 161,371,931 adherents.

Grammich et al. (2023) note that the 161 million adherents identified in the study represent 48.6% of the U.S. population in 2020. But there is no measure of how frequently (if at all) these adherents attend services — or of how many attendees are “nonadherents.” The website acknowledges that “adherent” is not precisely defined: “Exact definitions of ‘congregations’ and ‘adherents’ vary by religious body.... Adherents generally are members, children who are not members, and others who are not members but are considered participants in the congregation” (Association of Religion Data Archives, 2020). Thus, the statistic about adherents cannot be used to estimate frequency of attendance.

The U.S. Religion Census could, however, be used to obtain a direct measurement of attendance, since the statisticians taking the Census have compiled a comprehensive list of U.S. religious organizations. A probability sample of congregations could be taken from this list, and statistics on attendance obtained from the sampled congregations could be used to estimate religious service attendance. This would provide a direct measurement of behavior that is not subject to the measurement problems in self-report surveys such as social desirability bias.

American Time Use Survey

Another way to reduce social desirability bias is to ask people how they spend their time instead of asking them how frequently they attend religious services (Presser and Stinson, 1998). If people are asked how they spend each hour in a 24-hour period, they may be less likely to exaggerate socially desirable activities such as attending services.

Fortunately, a large reputable survey is available for estimating how people spend their time. The U.S. Bureau of Labor Statistics collects the American Time Use Survey (ATUS), in which respondents (persons age 15 and over) are asked to keep a diary of the start and stop time of each of their activities on the preceding day (from 4 am to 4 am).* The survey thus gives a detailed picture of how Americans spend their time on each day of the week and each hour of the day. One of the many activities measured is “attending religious services,” so the ATUS can be used to estimate the percentage of adults age 18+ who attend religious services at a place of worship on each day Sunday through Saturday.

Since a respondent is asked only about the previous day’s activities and not about other time periods, the ATUS will not give us direct estimates of the percentage of people who attend religious services weekly, at least once a month, or yearly. But we can estimate the percentage of adults age 18+ who are attending religious services on a typical Sunday, on a typical Monday, and so on for each day of the week. Table 1 gives these estimates, along with their standard errors, from the 2022 and 2023 ATUS.**

Table 1. ATUS estimates of percentage of adults attending religious services on each day of week. “Fri + Sat + Sun” gives the sum of the Friday, Saturday, and Sunday percentages and the last row sums the percentages for each day of the week. The columns on the right contain the standard errors for the estimates.

Even though the Gallup and AEI surveys estimate frequency of attendance, and the ATUS estimates the percentage of people attending on a typical Sunday, Monday, …, Saturday, we can still compare the statistics across the surveys. We do not have a direct estimate from ATUS of the percentage of adults who attend in a typical week, but we know that

% who attend on Sunday ≤ % who attend at least once during the week ≤ % who attend Sunday + % attend Monday + … + % attend Saturday.

Thus we can estimate from the 2022 ATUS that the percentage of adults who attend in a typical week is somewhere between 14.0% and 19.6%. The lower figure, 14.0%, is too small because it does not include people who attend services on other days of that week (but not Sunday, the day asked about). The upper figure, 19.6% is too large because it includes multiple attendances in the same week by the same person (for example, a Catholic who attends mass daily or a Protestant who attends church on Sunday and a prayer meeting on Wednesday). The sum of percentages of adults attending Friday, Saturday, or Sunday is probably a pretty good estimate of the percentage attending in a typical week — it includes the primary day of religious services for Islam, Judaism, and Christianity but does not include as much of the multiple attendances as the sum for all days of the week.

The estimated percentage of adults who attend a religious service in an “average” week includes weekly attendees, about a quarter of the once-a-month attendees, and a few of the people who attend less frequently. We can get rough estimates of these quantities from the Gallup and AEI statistics. Let’s assume that for Gallup, “nearly every week” corresponds to attending three out of four weeks and that “seldom” corresponds to attending twice per year. For the AEI survey, let’s assume that “regular” attendance is every week or nearly every week and is equivalent to attending on average seven out of every eight weeks, that “occasional” attendance is about once per month, and that “infrequent” attendance is twice per year. With these assumptions, the Gallup polls give an estimate of about 31.5% of adults attending religious services in a typical week, and the AEI survey gives an estimate of about 24.4% of adults attending in a typical week.***

Figure 1 shows the estimates from the Gallup and AEI polls and the 2022 ATUS (I used the 2022 ATUS since the Gallup polls were from 2021-2023 and the AEI poll was taken in 2022). The estimates of attendance on a typical week from the Gallup and AEI polls are higher than the upper bound for that statistic, 19.6%, from the 2022 ATUS. In fact, the estimate of attendance on a typical week from the Gallup polls is nearly twice as large as the 2022 ATUS sum of attendances on Friday, Saturday, and Sunday.

Figure 1. Estimates of attendance from American Time Use, Gallup, and AEI surveys. The estimates from ATUS are lower than those from the other two surveys.

The Gallup estimate of the percentage of people who attend weekly, which is measured directly in the survey, is also larger than the three attendance estimates from ATUS 2022. Logically, the percentage of adults who attend every week should be smaller than the percentage who attend in a typical week, since the latter statistic includes the occasional attendees.

The results in Figure 1 are consistent with the conclusion of Presser and Stinson (1998, p. 144) that “[r]espondents in conventional surveys substantially overreport their religious attendance.” Presser and Stinson (1998) ascribed the overreporting to social desirability bias, and that explanation is also supported by these reports, since the AEI survey, which is administered online and therefore thought to be less susceptible to social desirability bias, has lower estimates of attendance than the Gallup polls that are administered over the telephone.

Conclusions

No single source of data gives a full picture of religious service attendance, and the comparisons in this study emphasize the importance of having multiple sources of data to check results, particularly for a topic such as religious service attendance that is susceptible to measurement error. These comparisons can only be made, however, because of the free availability of high-quality federal data collections such as the ATUS.

Footnotes and References

*The ATUS is a subsample of the Current Population Survey (CPS), which is the primary source of U.S. labor force and unemployment statistics. The ATUS has an overall response rate of 36.9% for 2023 (U.S. Bureau of Labor Statistics, 2024), but, because it is subsampled from the CPS, there is a lot of information available for adjusting for nonresponse. I obtained the data from https://www.bls.gov/tus/data.htm and analyzed the data (restricted to adults age 18 and over) in the SURVEYREG procedure of SAS® software (SAS® Institute, 2020) using the final and replicate weights provided by the Bureau of Labor Statistics to calculate percentages and their standard errors. You can find the questionnaire at https://www.bls.gov/tus/questionnaires/tuquestionnaire.pdf.

**There are several ways in which one might define “religious service attendance.” I defined ATTEND to equal one if the activity code was 140101 (“attending religious services,” which includes “attending church services/synagogue/temple/mosque” but also includes activities such as attending confession, a bar mitzvah, or a wedding) and the activity took place at a place of worship. Thus, my estimate excludes people who attend “Zoom church” but it also includes a few activities that are not religious services.

Results are slightly different if I define attendance differently. If I remove the restriction that the activity takes place at a house of worship, for example, the estimated attendance on a Sunday for ATUS 2022 increases to 17.8% and the sum of attendances for Fri-Sun of the week increases to 22.8%. On the other hand, if I define attendance to include only those people who spend at least 30 minutes at the house of worship, estimated attendance on Sunday is 12.9% and the sum of attendances for Fri-Sun is 15.1%.

***The Gallup statistic is calculated as 21 + (3/4)(9) + (1/4)(11) + (2/52)(25), and the AEI statistic is calculated similarly. Note that all of the differences between the survey estimates are larger than the margins of sampling error.

References

Association of Religion Data Archives (2020). U.S. Membership Report (2020). https://www.thearda.com/us-religion/census/congregational-membership.

Benz, J., Witt-Swanson, L., and Cox, D.A. (2023). Faith After the Pandemic: How COVID-19 Changed American Religion. Washington, DC: American Enterprise Instititute.

Grammich, C., Dollhopf, E.J., Gautier, M.L., Houseal, R., Jones, D.E., Krindatch, A., Stanley, R., and Thumma, S.(2023). 2020 U.S. Religion Census: Religious Congregations & Adherents Study. Lenexa, KS: Association of Statisticians of American Religious Bodies.

Jones, J.M. (2024). Church attendance has declined in most U.S. religious groups. https://news.gallup.com/poll/642548/church-attendance-declined-religious-groups.aspx.

Presser, S. and Stinson, L. (1998). Data collection mode and social desirability bias in self-reported religious attendance. American Sociological Review, 63(1), 137-145.

SAS Institute, Inc. (2020). SAS/STAT ® 15.2 User’s Guide. Cary, NC: SAS Institute, Inc.

U.S. Bureau of Labor Statistics (2024). American Time Use Survey User’s Guide. Washington, DC: U.S. Bureau of Labor Statistics.