Religious pluralism and US church membership: a reassessment

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

Catholics

A fourth objection is more complicated and requires more extensive treatment. This objection concerns whether one can observe the true effects of religious pluralism without statistically controlling for the proportion of the population that is Catholic (hereafter, Catholic population share). The issue is important since nearly all of the published North American research findings supporting a positive relationship between pluralism and religious involvement depend on including statistical controls for Catholic population share (e.g., Finke and Stark 1988, 1989; Finke 1992).

The current debate over pluralism began when Finke and Stark (1988) showed, using the the 1906 Census of Religious Bodies, that among the 150 largest US cities, religious pluralism is positively related to percent of church adherence rates after controlling for Catholic population share (see also similar regressions in Finke and Stark 1989 and Finke 1992). However, Breault (1989a, 1989b), Land et al. (1991) and Blau et al. (1992) charge that Finke and Stark get positive regression results only because of severe multicollinearity problems created by adding statistical controls for Catholic population share (and/or Mormon population share).

Breault (1989b) notes that in the 1906 large city data used by Finke and Stark (1988) the actual bivariate correlation between pluralism and adherence rates is strongly negative (r = -.40). Using weighted least squares and ridge regression methods Breault (1989a) confirms that adding controls for Catholic population share creates multicollinearity problems sufficient to explain a reversal of pluralism's apparent effects from a negative correlation with adherence rate to a positive OLS regression beta.

The issue here is not whether multicollinearity, exists in the Finke and Stark (1988 and 1989) regressions. Finke and Stark (1989) acknowledge that Catholic population share has a high negative correlation with religious pluralism. But they argue that this is unavoidable since Catholic population share is an important part of the complete causal process and must be included in regressions to fully understand the true effects of pluralism. Why? Turn of the century Catholics, they maintain, had higher than average rates of religious mobilization for several reasons not reflected in the pluralism index. First, they were internally more diverse than other faiths so that pluralism not reflected in the pluralism index served to heighten Catholic, and thus overall, church involvement. Second, Catholics were a culturally oppressed, if not numeric, minority, a condition that increased Catholic religious involvement. Finally, in the early 20th century cities, where Catholics were concentrated, they were seen as a threat to Protestants who responded with vigorous Sunday School campaigns to thwart Catholic inroads. For all of these reasons, none of which are reflected in the pluralism index, adherence rates tended to be higher in cities with a large Catholic population share. Without statistical controls for Catholic population share, they argue, it is impossible to identify the independent effects of religious pluralism.