Comment on Chaim Waxman's Article, 'Religious and Ethnic Patterns of American Jewish Baby Boomers'

There appear to be several errors in Dr. Chaim I. Waxman's article "Religious and Ethnic Patterns of American Jewish Baby Boomers," (JSSR, March 1994) that require correction.

Waxman's analysis focuses on 801 respondents of the 1990 National Jewish Population Survey who were born between 1946 and 1964. This subsample, he goes on to say, "when weighted for the entire population, represent 169,255 cases." The American "core" Jewish population is more than five and a half million persons according to NJPS. A population of that size cannot have a mere 3% of its members in a twenty-year age range. In fact, the current number of American Jews in that age range is about 1.7 million.

However, it is very difficult to determine the true magnitude of errors in Waxman's tables since he confuses the number of household respondents with the number of persons. Although there are, indeed, 801 sample household respondents (a randomly chosen Jew in each household) between the ages of 26-44, there are 1,195 persons within that age range in the sampled households who report themselves as Jewish when asked, "What is your current religion?"

Moreover, it is not at all clear why Waxman limits his analysis to the selected household respondents when his conceptual unit of analysis ("baby-boomers") is the individual.

Waxman goes on to say that in comparing baby boomers to his next chosen age cohort of people between 46 and 64 years, "the sample of 398 cases when weighted represent more than 83,000 Jewish middle agers." Once again, although his conceptual unit of analysis is the individual ("middle agers"), he limits himself to the household respondent. Here, too, he confuses the number of households with the number of persons. Moreover, his projected number is again off by a factor of ten.

To add further to the confusion, the "weighted numbers" reported by Waxman vary between Tables 2 and 3 for the same age cohort. In Table 3 his reported N is off by a factor of 33. Apparently he failed to use the correct module weights, which would have given him the same number of cases even though particular modules were administered to only one-third of the sample.

The errors in Waxman's article extend beyond his miscalculation of population figures. His apparent confusion of households and persons will undoubtedly result in crucial errors in his inferences as well.

The central point of his research note is that the patterns of Jewish identification and affiliation among baby-boomers reveals a pervasive decline in Jewish identity and identification between the two age cohorts. His purported measures of decline, however, are skewed by his reliance solely on household respondents. The reason for this is that even with correct weighting the randomly chosen household respondent underrepresents endogamously married Jewish baby-boomers and therefore overrepresents Jewish single baby-boomers and the intermarried. This is particularly problematic in Table 5 where he claims to be focusing on the intermarriage patterns of all spouses of Jewish baby-boomers, but in fact his table represents only household respondents.

The respondent weight factors used in NJPS were designed to represent the total adult U.S. population of persons in households containing a self-identified Jewish person. They do not compensate for differential patterns of identification of persons within the households.

In addition to the problems resulting from errors in selection of weights and units of analysis, Waxman also misinterprets some of his own findings. He writes that:

Chanukah candle-lighting is not a very reliable index of Jewish identity. There was a small number of eases (26) in the NJPS in which neither of the two heads of household, one of whom was the baby boomer respondent, were Jewish; both were, in fact, Catholic. Nevertheless, analysis revealed that in five of those cases (24%) respondents reported that they light Chanukah candles sometimes or always. Whatever the reasons, they apparently were not manifestations of the respondents' Jewish identity.

Waxman seems not to appreciate the significance of the numbers he presents. Five cases out of 801. A little arithmetic reveals that this constitutes just .6% of Waxman's baby-boom sample. Most social scientists would regard any measure that has an error of only six-tenths of a percent as a highly reliable indicator of any concept , particularly if there is a reasonably good explanation for that .6%.

Unfortunately, Waxman falls to mention that those five households where Chanukah was observed even though both adults in the home were Catholic were cases in which there was either a child or an adult parent of the respondent or the spouse of the respondent who is Jewish. That is the basis on which such a household was included in the Jewish population survey in the first place. There are households in the NJPS sample where, resulting from custodial arrangements following divorce and remarriage, there are Jewish children but no Jewish adult. This would suggest that the presence of Chanukah candle lighting is, in fact, an excellent indicator of the tenacity of Jewishness even in the face of intermarriage as well as divorce. That is exactly the opposite of the conclusion Waxman insists on drawing.

In fact, Chanukah candle lighting is probably the best indicator of the presence of someone with Jewish identity in the home. This particular variable must be treated as an indicator of some aspect of Jewishness of the household rather than as an indicator of any individual respondent's personal identity.

The errors contained in Waxman's article should serve as a cautionary note about the dangers in doing secondary analysis in general and in working with the 1990 NJPS in particular:

Know your parameters. Knowing basic population attributes such as total size can prevent the making of inferences that are counter-intuitive about the size of subpopulations.

Know how both the sample and the respondent were chosen. NJPS randomly selected one Jewish adult respondent beth in same-faith and multifaith households.

Know your basic weights per case. Without careful attention to the weight factors that were assigned by the creators of the data set for use with each particular item in the survey instrument, one can be misled into making projections to the population that are incorrect.

In fact, the 1990 NJPS contains 4 separate population (respondent) weights and 4 separate household (person) weights. Each weight is meant to be used for a distinct set of questionnaire items depending on whether those items were to be asked only with regard to respondents or to all members of the household, and whether the particular item was administered to the entire sample or only on a rotational basis to but one-third of the sample. When correctly used these weights produce a uniform set of figures for each survey item, reflecting a core Jewish population size of roughly 5.5 million persons in roughly 2.8 million households.

Know your level of analysis. Without proper attention to the level of analysis (e.g., household, family or individual) for which particular items were designed to serve as indicators, one can erroneously misapply to one level indicators designed for another, thereby drawing inferences that are either illogical or not corroborated by other indicators.