Reported attendance and observed participation
For decades, surveys in the United States and Canada asked adults whether they had attended a religious service during the previous week. The resulting estimates were often near 40 percent in the United States, while Canadian estimates were generally lower. A 1999 international survey, for example, recorded reported weekly attendance of 43 percent in the United States and 20 percent in Canada, with 8 percent of Americans and 38 percent of Canadians saying that they never or almost never attended.
These figures describe answers to survey questions, not a direct count of people present. They remain useful evidence, but they should not be treated as a precise census of congregational attendance.
Why surveys can overstate participation
People do not always report their behaviour exactly. When a question concerns a socially valued activity, some respondents may give the answer they believe reflects well on them. This is commonly called social-desirability bias. Memory, interpretation of “regularly,” and inclusion of occasional services or special events can also affect answers.
Several historical studies compared survey responses with other forms of evidence. Researchers who examined membership totals, giving patterns and attendance records found that the number of people physically present could be considerably below the level suggested by standard polls. Their work did not establish one timeless percentage for every community; it demonstrated that the measurement method changes the result.
Evidence from attendance studies
Studies associated with Kirk Hadaway, Penny Marler and Mark Chaves compared reported participation with counts in selected Protestant and Catholic settings in the United States. Their estimates suggested that actual weekly attendance was closer to one-half of the commonly reported survey level in the locations examined. Follow-up work in Ontario found a similar gap between what respondents said and what local counts indicated.
Other researchers used time-use diaries. Participants recorded what they had done rather than answering a direct question about whether they were the kind of person who attended services. Analyses of diaries from the 1960s through the 1990s suggested a decline in actual weekend attendance, from roughly 42 percent in the mid-1960s to about 26 percent in the early 1990s. These estimates should be read as historical research findings, not as current measurements of all North Americans.
| Measure | What it captures | Main limitation |
|---|---|---|
| Direct survey question | What respondents remember and report | Memory and social-desirability effects |
| Time-use diary | Activities recorded across a defined period | Participant recording and sample design |
| Congregational count | People present at selected services | Coverage of congregations and dates |
Long-term trends and changing questions
Later surveys did not all move in the same direction. Some continued to produce figures above 40 percent, while redesigned instruments produced estimates closer to 30 percent. Differences can arise from wording, sampling, the time period covered and whether respondents are asked about a specific service or about their general habits.
The evidence supports a cautious conclusion: reported religious participation has often exceeded directly observed participation, and the gap may vary by country, tradition, age group and research method. It is safer to compare like with like than to place every percentage in one continuous series.
Context matters
Attendance can rise temporarily after a major public tragedy, during holidays or when a community faces an urgent shared concern. Historical reporting after the September 2001 attacks described a short-lived increase in some areas, followed by a return toward earlier patterns. Such episodes illustrate why a single week should not be treated as a permanent national trend.
Likewise, a congregation can grow more slowly than its surrounding population while still gaining members in absolute terms. Claims about decline should therefore specify whether they concern attendance, membership, population share or institutional vitality.
A measured conclusion
No single number answers how many North Americans regularly attend religious services. Surveys reveal how people describe their participation; counts and diaries offer different perspectives. Taken together, the historical research suggests that public estimates around 40 percent were probably too high when interpreted as the number physically present each week.
The most responsible reading is comparative and transparent: identify the question asked, the population studied, the date, and the method used. This approach respects both the complexity of religious practice and the limits of social measurement.
Selected historical sources
- Hadaway, Marler and Chaves, research on reported and observed religious attendance in the United States.
- Presser and Stinson, analyses of time-use diaries and weekend religious participation.
- National Opinion Research Center and General Social Survey materials on attendance question wording.
- Historical international survey material, including the 1999 Millennium Study.
- Gallup and National Election Studies reports cited in earlier versions of this essay.