Testing the attendance gap in a conservative church

Religious activity involves many observable, quantifiable behaviors. People pray, join churches, attend worship services, and give money to religious organizations. Much of this behavior is enumerated. Nearly all churches, synagogues, and other places of worship keep financial records on giving, most have membership rolls, and many count worship attendance. Given the many objective, verifiable ways of measuring religious behavior it may seem odd that so much of what we know about "religiosity" is based on self-reports from polls and social surveys. The dependence on social surveys is not unique to information about religion, of course. As Presser and Traugott (1992: 77) note: "although in principle there are many different ways to study the social world, in practice the findings of contemporary social science are based to a remarkable degree on the accounts people give of themselves."

Part of the problem, at least in measuring religious behavior, is the lack of consistent records. Even though many churches, synagogues, and other religious organizations count attendance at worship, not all do and there is no central clearinghouse that collects all the information that might be available. There are too many independent churches, barely organized denominations, and uncooperative congregations to allow anyone to compile and compute an exact (or nearly exact) count of persons who attend religious services in the United States during a typical week.

Lacking exact data, what do we know about the number of Americans who attend churches and other places of worship? Newspaper and magazine stories, periodical reports of Religion in America, sociology texts, and scholarly articles routinely use survey data to identify the level of religious participation and religious belonging in the United States (Hour and Greeley 1987, 1998; Babble 1992; Johnson 1992; Princeton Religion Research Center 1996). The most widely referenced source of data on worship attendance is responses to a question that has been asked by national polls since the 1930s: "Did you, yourself, happen to attend church or synagogue in the last seven days?" Responses to this question have been remarkably consistent since the late 1960s. Each year around 40 percent of Americans report that they attended church during the previous week (Princeton Religion Research Center 1996).

Despite the dependence on self-reported religious behavior from polls there was, until recently, very little discussion about the accuracy of such data. In general, social scientists assumed that attendance reports were somewhat inflated, but also that the degree of inflation was small (Stark 1994; Hout and Greeley 1998). Studies of voting behavior, contributions to charity, honest reporting of photocopying, and illegal drug use suggest, however, that the assumption of minimal discrepancy between self-reported religious behavior and actual behavior may be unwarranted (Bishop and Fisher 1995; Bradburn et al. 1987; Fendrich and Vaughn 1994; Goldstone and Chin 1993; Presser and Traugott 1992).

In order to examine the possibility that survey-based rates of church attendance were inflated, Hadaway, Marler, and Chaves (1993) compared attendance counts to survey self-reports among Protestants in one Ohio county, and Catholics in eighteen dioceses. Evidence of substantial overreporting was found. Among Protestants, 19.6 percent attended church worship in Ashtabula County, Ohio during a typical week in 1992 - compared to the 35.8 percent who said they attended. Among Catholics in 18 dioceses around the nation, attendance counts suggested an attendance rate of 28 percent - compared to the national rate of 51 percent reported by the Gallup Organization. In a subsequent report, Chaves and Cavendish (1994) adjusted the count-based rate of Roman Catholic attendance at mass to 26.7 percent, using a larger number of dioceses (48).

Criticism of Hadaway, Marler, and Chaves's (1993) findings centered on the large size of the "gap" between poll-based and count-based measures of attendance. Some overreporting was assumed for any socially approved behavior, including church attendance, but overreporting of over 80 percent was difficult for many social scientists (and polling organizations) to accept. The reliability of poll-based self-reports was not the only area of concern. If real, a large attendance gap suggests that many assumptions about the robustness and exceptional nature of religion in America should be modified.

One explanation for the large attendance gap concerned the wording of the Gallup question. The question does not define what is meant by "attending church" and asks people who may have missed one Sunday service out of many to symbolically count themselves among non-churchgoers by saying, "No, I did not attend church in the last seven days." One of the first tests of the traditional Gallup wording was conducted by the Gallup Organization itself in which the traditional question was substituted with a composite item asking the respondent to think about their answer and also to name the church they attended. This new question reduced the attendance rate by only one percentage point (Princeton Religion Research Center 1994). A similar result was obtained by a National Election Study experiment (Belli et al. 1995; Smith 1998). Other efforts, notably the 1996 General Social Survey and a University of Maryland study, produced sizable drops in self-reported attendance using alternative wording (Presser and Stinson 1998; Smith 1998). Apparently, questions that shift the focus from church attendance to events occurring during the week (including doctor visits, eating out, going to the movies, etc.) minimize the problems of backward and forward telescoping, social desirability, and identity consistency (that is, behavior consistent with self-perceptions as religious, church-going people). Using these questions and adjusting for "church attendance" at non-worship events reduces self-reported attendance from 40 percent to around 30 percent.(1)