Religious pluralism and US church membership: a reassessment

Sociology of Religion, Summer, 1999 by Daniel V.A. Olson

OBJECTIONS

The results shown in Figure 1 are consistent with Breault (1989a) who reports a negative correlation of -.22 using the 1980 Glenmary data. The results in Table 1 are roughly consistent with the county level results of Land et al. (1991) and Blau et al. (1992) for US counties from 1850 to 1930. However, Finke and Stark (1989) and Finke et al. (1996) have raised four serious objections to these previous analyses.

First, Finke and Stark (1989) and Finke, Guest, and Stark (1996, n 4: 206) have questioned the reliability of Breault's work. Using the same 1980 Glenmary data and variables as Breault (1989a) they report a positive correlation (r = .21) between religious pluralism and church adherence rates. Olson (1998) shows that Finke and Stark's (1989) results are in error due to a mistake in the SAS program used to calculate the pluralism indexes. Also using the 1980 Glenmary data, Olson obtains results identical to Breault (r = -.22).

Second, Finke and Stark (1989) and Finke (1992) claim that correlations with pluralism based on the Glenmary data are biased in unknown ways since the 1980 study includes only 111 denominations, entirely missing several large Black Baptist denominations, several Jewish groups, and hundreds of very small denominations. In contrast, I find the correlations with pluralism in the 1990 Glenmary data to be quite robust with respect to excluding small or large denominations. The 1990 data includes 133 denominations (including estimates for Black Baptist groups and Jews). Using all 133 denominations, the correlation between pluralism and adherence rates is -.35. Using only the 79 denominations found in both the 1980 and 1990 Glenmary studies, thus excluding Black Baptists and many Jews, the correlation is -.36. Using only the 15 denominations in the 1990 Glenmary data that each report more than a million adherents, and thus excluding many small denominations, the correlation is -.33. Combined with the fact, noted above, that the 1990 Glenmary data already include the vast majority of US church adherents, these analyses suggest that the negative correlation between pluralism and participation is quite robust and not subject to unpredictable variations, much less a reversal of sign, based on the exclusion of many small, or a few very large denominations.

Third, Finke and Stark (1989) and Finke et al. (1996) charge that counties are too large an area to test their theories. Thus, Finke et al. (1996) show that among towns in New York State in 1855 and 1865, religious pluralism has a positive relationship with attendance rates, but when data from these towns are aggregated to the county level, the relationship becomes statistically insignificant. They argue this is because 19th century religious markets were no larger than the distance people could travel by foot or horse to attend church. Though apparently true in 19th century New York, automobiles may well have extended the relevant effects of pluralism to larger areas such as counties. Thus, Olson and Hadaway (1998), using the 1991 Canadian census, find a strong negative relationship between religious pluralism and religious affiliation rates among both counties and cities. Taken together, the Glenmary data yield county-level results that are consistent with results from contemporary Canadian cities and counties and with historical results for US counties from 1850 to 1930 (Land, Deane, and Blau 1991; Blau et al. 1992; Blau et al. 1993).