Reply to Mayer and Kosmin

Abstract

Listen

Translate article

The weighted Ns in my tables were, indeed, off by a factor of 10. The error was an oversight due simply to the fact that the data set is so large that the statistics package gives results that are reduced by a factor of 10. Other than that, the data in the tables are accurate. On the matter of why I limit my analysis to the 801 respondents who report their religion as Jewish, there is no mystery involved. Respondents were selected for the simple reason that almost all of the Jewish religio-cultural questions in NJPS were asked only of respondents. Since much of my analysis, beyond what was in the article, involves those matters, I saw no reason, and still do not, to go beyond respondents. Having clearly stated that my sample is the 801 baby boomer respondents who identified their religion as Jewish, I then looked at their spouses and gave the figures on the spouses' religious identification. Since I was not, at this point, interested in whether the spouses are baby boomers that is irrelevant for my focus of analysis there is no substance to the criticism that I underrepresented Jewish baby boomers who are married to other Jews. I gave the spouse's religion, much as I did so for my GSS subsample of Protestant and Catholic baby boomer respondents. The attempted lesson by Mayer and Kosmin in statistical reliability is completely off the mark. They simply did not read carefully. I specifically stated that I ran the NJPS data and found that there were 26 cases in which both heads of household, one of whom was a respondent, were Catholic and, yet, 5 of them, or 24%, reported that they light Chanuka candles sometimes or always. All of the 801 respondents in my subsample were Jewish; obviously, they were not among these 26. My main point was, and remains so despite Mayer and Kosmin, that the lighting of Chanuka candles in the household, per se, reveals very little about Jewish identity. Ironically, when I came to that tentative conclusion, it was in no way in criticism of NJPS; rather, I had in mind a number of observers who had touted the high rate of Chanuka candle-lighting as proof of the vibrancy of American Judaism and Jewry. Finally, with respect to the various weights, the issues are much more complex than my critics suggest. I recently participated in a lengthy meeting of NJPS monograph writers, including Mayer and Kosmin, on the subject of the various weights in NJPS, at which we concluded that all of the weights are of questionable appropriateness for some of the variables in NJPS as well as for many types of multivariate analysis. In other situations, there is no one objectively correct way to use the weights; the weighting method is subject to subjective interpretation. It would be fair to say that all of us, Mayer and Kosmin included, should be more cautious in when and how we use the weights. In the future, in fact, I intend to primarily use unweighted data for analysis, except for purely demographic variables. Be that as it may, the basic patterns which I delineated in my article remain, regardless of how the data are weighted.