Testing the attendance gap in a conservative church

Sociology of Religion, Summer, 1999 by Penny Long Marler, C. Kirk Hadaway

The GSS wording experiment (Smith 1998), Presser and Stinson's study (1998), and our own experiments in Britain and Australia (Hadaway and Marler 1998) indicate that it is possible to obtain reduced and presumably more accurate self-reports of religious behavior. Such question-wording experiments should not be seen as substitutes for experiments that provide external validation for surveys, however. In the same way that studies of voting behavior compared self-reports to voting records, research is needed that compares self-reported religious behavior to observations of that behavior. The studies by Hadaway, Marler, and Chaves (1993) and Chaves and Cavendish (1994) were efforts to provide external validation of poll data in several regions by aggregating counts of attenders from many churches. In the present study we examine the possibility of attendance misreporting by individuals in a single church.

A CHURCH ATTENDANCE TEST

One of the methodological criticisms directed at the Hadaway, Marler, and Chaves (1993) study was that attendance data from local churches in Ashtabula County, Ohio were collected at various times during the year using a variety of methods (church records of average attendance, single Sunday counts in the Fall, estimates by local church ministers, attendance counts by the authors, and in some cases estimates based on cars in church parking lots). According to Caplow (1998), "nearly all the numbers were estimates, not actual counts." Caplow (1998) and Hout and Greeley (1998) do not consider such estimates to be reliable and suggest that inaccuracies may have produced artificially low attendance counts. Caplow (1998) also indicated that the survey data used by Hadaway, Marler, and Chaves (1993) did not provide a "direct connection between the time periods for which respondents reported attendance and the month or year in which the attendance was counted." Against these criticisms it can be argued that: (1) church officials have no reason to underestimate attendance in their own congregations; and (2) polling is not necessary during the week attendance was counted so long as polling and counting are conducted during months in which attendance is not usually high or low.(2) Nevertheless, in response to criticisms about the time frame disconnection and the multi-faceted method of collecting attendance data, we conducted a test in which we counted actual attendance on a specific Sunday, and during the following week asked a sample of members if they attended church "in the last seven days." This church will not be identified. It is a very large, white, middle-class congregation, with a membership over 2,000. The congregation is affiliated with a well-established, evangelical denomination and is located in an older metropolitan suburb in the deep South.(3) It is not one of the new, rapidly growing "mega-churches," nor a Willow Creek-inspired "seeker sensitive"congregation, although it has made an effort to incorporate some contemporary elements (such as praise choruses) into a rather traditional worship service. We will call the congregation "Broadview Evangelical Church."