The Effect of Burdensome Survey Questions on Data Quality in an Omnibus Survey
Submission ID: 4604

Date: Thursday, 5/13/2021, 1:00 PM to 3:00 PM EDT
Session: Concurrent Session F (Asynchronous with live Q&A)

Primary Presenter

Additional Authors or Round Table Presenters

Angelica Phillips, University of Nebraksa-Lincoln ,

, ,

, ,

, ,

, ,

, ,

, ,

, ,

Abstract

Some survey questions are more burdensome than others due to the increased cognitive burden that they place on respondents. In interviewer-administered omnibus surveys, where researchers have to make decisions about the ordering of questions, it is unclear where in the interview to place these questions as highly burdensome questions may result in lower quality responses (Dillman, Smyth, and Christian 2014; Galesic 2006). One example of these burdensome questions is social network questions, wherein respondents are asked about members of their personal network and details about their relationships with those individuals. Although previous research has explored how social network questions influence respondent burden (e.g., Manfreda et al. 2004) and how respondent burden, in turn, influences response quality to the social network questions (e.g., Galesic 2006; Galesic and Bosnjak 2009), scant attention has been given to how social network questions can influence the response quality of answers to non-social network questions in an omnibus survey. In this study, we answer the research question of how the placement of a social network question is associated with the data quality, such as item nonresponse, on unrelated questions in an omnibus survey. To answer this research question, we use data from the 2010 General Social Survey (GSS) (Smith et al. 2019; AAPOR RR5=70.3%), which included an experiment on the location of social network questions within the interview (i.e., beginning, middle, or end). We find that the location of the social network questions within the interview was not significantly related to item nonresponse rates, nor was this relationship moderated by cognitive capacity of the respondent. We conclude with implications for questionnaire design.