Reconceptualizing religious change: ethno-apostasy and change in religion among American Jews

Sociology of Religion, Winter, 2006 by Benjamin T. Phillips, Shaul Kelner

Switching as a rare event

Recent papers by King and Zeng (2001a, 2001b) emphasize the importance of considering the possibility of bias in logistic regression when either ones or zeroes are scarce. A rare event is defined as a binary dependent variable with "dozens to thousands of times fewer ones ... than zeroes" (King and Zeng 2001b:138). In such cases, estimated probabilities of an event occurring will be too small, unless corrections are made. The effects of corrections are largest where numbers of observations are less than a few thousand and where the events make up less than five percent of cases. Our data appear to meet these criteria, with about 4,000 observations and with ethno-apostates making up 4.2 percent of cases. As Tomz, King and Zeng's (1999) relogit procedure for Stata (StataCorp 2004) is for binary rather than multinomial logistic, analyses are run as pairs of binary logistic regression (stayers vs. switchers; stayers vs. ethno-apostates), which provide consistent estimates of the parameters of multinomial logistic models (Begg and Gray 1984; Long 1997). Comparisons between multinomial logistic models for complex survey data and relogit results show substantive differences in the predicted direction.

Defining religious change

Analyses of religious switching and apostasy generally define the population of interest as being any individual raised in the religious tradition of interest. In the case of American Jews, however, the possibility of having been raised in an intermarried household means that a somewhat broader standard must be used. Accordingly, the population eligible for analysis includes all respondents who responded yes in any way to the question "Were you raised as a Jew?" including those who were coded by NJPS interviewers as having been raised "half' or "partially" Jewish (United Jewish Communities 2003d:10). As noted above, we have decided to treat Jews who profess adherence to a non-Jewish religion as religious switchers, regardless of whether they continue to declare adherence to Judaism as well. Therefore, respondents who, as adults, identified at the time of the survey with Judaism and another religion or exclusively with another religion are classified as switchers. Ethno-apostates are defined as individuals who at the time of the survey categorized themselves as having no religion and who said they did not consider themselves to be Jewish. Stayers are classified as respondents who at the time of the survey were exclusively Jewish by religion or did not identify with a religion but still considered themselves to be currently Jewish.

Independent variables

The following sociodemographic variables are modeled: female, a quadratic term for age, and parents' migrant status. (4) As there was no significant difference in the probability of religious change between being an immigrant or having two immigrant parents, the latter term was kept (thus including migrants). Respondents with one immigrant parent were as likely to change status as those with no migrant parents.