ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Combining Data from Complex Surveys to Compensate for Limitations in Targeted Surveys of Rare Groups: A Study of the Jewish Population in the United States

Elizabeth Tighe, Leonard Saxe

Building: Law Building
Room: Breakout 7 - Law Building, Room 028
Date: 2012-07-10 01:30 PM – 03:00 PM
Last modified: 2012-01-25


One of the defining characteristics of the United States is its religious diversity and the traditions of civic involvement and service of many of the religious communities. There is, however, an additional defining characteristic of the US, the constitutional separation of church and state, which precludes the government from compelling, as part of mandatory census reports, collection of religious identification information. As a result, researchers interested in the religious composition of the US need to rely on voluntary surveys, such as the American Religious Identification Survey and surveys commissioned by specific religious denominations. Single surveys as sources of population estimation are problematic. In particular for low-incidence religious groups (e.g., Mormans, Muslims, Jews) which have less than 5% the population, typical surveys include too few respondents to be able to describe these groups reliably. Moreover, any individual survey contains systematic errors that arise from questionnaire construction, sampling, sponsorship, and “house” effects. Although targeted sampling can be done to obtain over-samples of particular low incidence groups, no reliable external source of data exists on the distribution of the groups nationally that would allow one to accurately adjust population estimates for these over-samples. In addition, obtaining over-samples of multiple low incidence groups simultaneously for purposes of comparative analysis can be prohibitively expensive and introduces potential biase (e.g., those more strongly affiliated more likely to be identified through targeting). To address these challenges in assessing small religious groups, we draw on methods of cross-survey analysis (e.g., Park, Gelman & Bafumi, 2004) to combine data across a sample of over 150 independent surveys of the US adult household population. Hierarchical Bayesian analysis is used to account for clustering of respondents within surveys and to examine potential biases associated with survey characteristics. Estimates are post-stratified across surveys on basic demographics such as age, sex, race, educational attainment and geographic dispersion. These distributions are then used to provide adjustments in a smaller, substantively specific survey of the population. The results from this analysis provide a methodological framework for studying small religious groups, but also other small groups. The approach serves as an important source of data for targeted surveys by providing a reliable resource for post-stratification adjustments based on religious identification.