Religious Influence and Identity Survey, 1996

Data Archive > U.S. Surveys > General Population > Local/Regional > Others > Summary


The Religious Identity and Influence Survey was fielded in 1996 under the direction of Dr. Christian Smith of the Department of Sociology at the University of North Carolina at Chapel Hill, with the assistance of David Sikkink, then PhD graduate student in Smith’s department. The survey focused on the “commitments, beliefs, concerns, and practices” of evangelicals and other church-going Protestants, as well as how they viewed the relationship between Christians and the educational, political, and other institutions within American society (Smith et al. 1998). Details on the survey research methods are published as Appendices A, B, and D in Christian Smith et al., 1998, American Evangelicalism: Embattled and Thriving, Chicago: University of Chicago Press. Users must be extremely careful to use appropriate weights (described below) in their analyses.

Data File
Cases: 2,590
Variables: 168
Weight Variable: WEIGHT1, WEIGHT2, WEIGHT3
Data Collection
Date Collected: January-March of 1996
Funded By
Pew Charitable Trusts
Collection Procedures
The data were collected using a telephone survey conducted by FGI, Inc, a national survey research firm based in Chapel Hill, N.C. The research design included at least, but sometimes more than, ten calls for each number; and at least, but sometimes more than, three callbacks to convert refusals. The response rate for the survey was 69 percent. At least five percent of respondents who had completed interviews were called back and again asked key questions to verify answer reliability (information taken from Smith et al. 1998).
Sampling Procedures
Survey Sampling, Inc provided the sample for the telephone survey. It was arranged in replicates based on the proportion of working telephone exchanges nationwide. Random-digit dialing ensured equal representation of listed, unlisted, and non-yet-listed household telephone numbers. In order to randomize responses within households, and so to ensure representativeness by age and gender, interviewers asked to speak with the person in the household who has the next birthday (information taken from Smith et al. 1998).

See Smith et al. 1998 for more details about the collection and sampling procedure
Principal Investigators
Christian Smith, Department of Sociology, University of North Carolina-Chapel Hill
Related Publications
Smith, Christian et al. 1998. American Evangelicalism: Embattled and Thriving. Chicago: University of Chicago Press.

Smith, Christian. 2000. Christian America? What Evangelicals Really Want. Berkeley: University of California Press.
Notes
Important note about skip patterns: Many of the questions were only asked of Protestants that have relatively high levels of religiosity. A smaller set of questions was asked of Protestants with high religiosity and of Catholics. Each question below includes information on which set of respondents answered the question. Here are the criteria for constructing each group:

GROUP 1:
Church-going Protestants and irregularly/non-church-going Protestants who say nonetheless that their faith is Extremely Important in their lives (total sample size = 2082). Specifically, the intersection of one response from each of the two categories listed below placed respondents in Group 1:


Respondent is Protestant (RELIGION=1; or CATHOLIC=2; or RELIG2=1-4, 6-8, 10, 12-13, 15-16, 18-20, 24-27)


AND respondent attends frequently or faith is very important (ATTEND=1-3; or ATTEND=4-9 AND FAITHIMP=1).


GROUP 2:
Catholics. Primarily respondents for which RELIGION=2. In addition, respondents for which RELIGION=5 AND CATHOLIC=1.


GROUP 3:
This group includes all other respondents; that is, respondents who did not meet the above criteria for Groups 1 and 2 because they were either not Protestant or Catholic or because they were irregular attenders who did not answer that their faith was "extremely important." Group 3 answered a sub-set of the survey denoted wherever the Group 1 or Group 2 indicators are absent.
Weights
Important note about weights:
Appropriate weighting of the data is essential for obtaining nationally representative results. If you are interested in using the data to describe the non-institutionalized American population in the lower 48 states, you must apply the appropriate weight, since Protestants with high levels of religiosity have been oversampled in this study.


The weight variables in the dataset are:
WEIGHT1: Household Size Weight.
An adult in a household with a greater number of adults is less likely to be selected in a telephone survey than an adult in a household with fewer adults. Therefore, most telephone survey data is weighted by the household size to come closer to representing a random sample of American adults. In most cases, users will want to use WEIGHT3, but using WEIGHT1 may be appropriate if users are only analyzing Protestants with high levels of religiosity (Group 1 below). That is, if you want your results to represent a random sample of Protestants with high levels of religiosity, then it is usually sufficient to use WEIGHT1. WEIGHT1 is the household size of the respondent divided by the average household size in the dataset.


WEIGHT2: Protestant Oversample Weight.
Protestants with high religiosity are over-sampled in this study. While some questions on the survey were only asked of Protestants with high religiosity, other questions were asked of all adults. For questions asked of all respondents, you will use WEIGHT2 if you are interested in obtaining results that describe a random sample of Americans and you are willing to ignore the household size weighting problem mentioned above (see WEIGHT1 description). However, in most analyses you should use WEIGHT3. WEIGHT2 only adjusts for the higher probability that Protestants with high religiosity (Group 1 above) were selected for the study. (The weight is the inverse of the probability of selection for each group. In this case, the weight value is .41 for Protestants with high religiosity and 3.35 for all other respondents.)


WEIGHT3: National Weight.
Use this weight if you are interested in the results of a survey question that was asked of all types of Americans (not only Protestants with high religiosity) and you want to obtain results that are nationally representative. WEIGHT3 adjusts both for household size and the higher probability that Protestants with high religiosity were selected for the study. WEIGHT3 is the result of multiplying WEIGHT1 and WEIGHT2.


There are a few questions in the survey that were answered by Catholics and highly religious Protestants (Groups 1 and 2 above) and no one else. None of the weights included in the dataset are ideal if one wants the frequencies for these variables to describe the total population of Catholics and highly religious Protestants. One acceptable strategy for these questions is to apply WEIGHT1 and analyze Catholics and highly religious Protestants separately. This will allow you to obtain estimates for the Catholic population, and separate estimates for the highly religious Protestant population.


Furthermore, in regression analysis of survey data, it is appropriate to ignore the weights if you include in the model independent variables that capture the sources of the bias from a random sample. In a regression analysis, a binary variable for household size and one for Protestants with high religiosity (Group 1 above) would take the place of WEIGHT3 described above.