Keywords: survey mode, religion, 3MC

Pew Research Center has been exploring whether and how to migrate our CATI-based international surveys to online panels while preserving our trends. To that end, we imbedded experiments in three probability-based online panels in France, Germany and the UK to explore the most suitable way to measure questions about religion without distorting comparisons with previous results from phone surveys.

First, we looked at religious identification. In our long-standing CATI trends, we ask people “What is your current religion, if any” and interviewers are trained to use a pre-coded list. In our CAWI experiment, we randomly assigned people to one of two different lists: a short form with only six of the most common responses and a long form with 12 responses. We discuss the differences between these two forms, as well as how both compare to the nationally representative CATI surveys we fielded at the same time.

Second, we looked at questions about religious attendance and prayer. Our CATI-based trends for both of these questions are about frequency: how often people attend religious services outside of weddings and funerals, as well as how often people pray, aside from religious services. On these CATI surveys, we order the response categories from most frequent to least. But, as part of the online panels we randomly assigned people both to this traditional ordering and reverse-coded them. We explore the difference between these two options in CAWI and discuss which one best mirrors the concurrent CATI surveys.

We explore how our results for these experiments are similar and different across the three countries and conclude with an assessment of how we plan to migrate our trends, going forward.