Keywords: Multi-Mode Surveys, Multi-Mode Design, ISSP, Item Nonresponse

The General Social Survey, conducted since 1972, is the USA’s only nationally representative study of non-institutionalized adults that has systematically studied major social topics, and has conducted systematically thematic modules for the International Social Survey Programme. In 2021, due to the COVID-19 pandemic, the GSS shifted from a face-to-face interviewer-administered survey to a web-based mode. In 2022, the GSS formalized an experiment to compare two different data collection sequences that reverse the order of mode administration; namely, (i) face to face first with a web-based follow-up for nonrespondents and (ii) web-based data collection first, with an in-person nonresponse follow-up.
The shift to multi-mode data collection presents new challenges for understanding not only the demographic background of GSS respondents, but also the ways in which their attitudes, behaviors, and beliefs appear susceptible to the mode of survey administration. In the 2022 GSS, after ten weeks of data collection, nonresponse follow-up began. For nonresponse follow-up, sampled households were reassigned to the opposite mode, presenting each respondent with the opportunity to complete the survey in a different mode. Initial research conducted by the GSS teams with select opinion variables, suggests that the two experimental data collection sequences appear to be comparable; however, analyzed on ISSP variables is yet to be conducted. This paper aims to analyze variables from the two ISSP modules included in the 2022 GSS (i.e., Family and Changing Gender Roles V and Health and Healthcare II) that were collected with the two experimental data collection sequences. This research will help to understand the effect of mode and data collection sequence. The goal is twofold: (i) determine if the two experimental sequences yield comparable estimates for ISSP variables and (ii) identify which ISSP variables appear to be mode-sensitive. We will discuss implications from findings and whether post-data collection strategies such as post-stratification can help address potential differential patterns across.