Item Nonresponse in Web Versus Other Survey Modes: A Systematic Review and Meta-Analysis

Abstract

Nonresponse is a fundamental issue in survey research, due to the trend of declining response rates across survey modes. This issue is particularly serious for web surveys: A recent meta-analysis found that unit nonresponse in web surveys is, on average, higher by 12 percentage points than in other modes (Daikeler et al., 2020). Although the issue of unit nonresponse in web and other survey modes has been investigated in several meta-analyses, item nonresponse has not been examined meta-analytically, a gap addressed by this paper. Are web surveys at a disadvantage compared to other survey modes in terms of item nonresponse as well? To address this question, 13 eligible experimental manuscripts reporting 23 effect sizes were identified in a comprehensive literature search. Meta-analytic findings showed that there was no statistically significant difference in the average item nonresponse rate in web versus other survey modes. However, six moderator variables were found to statistically significantly affect the relation between the survey mode and the item nonresponse rate, namely, the target population, the number of contacts, the mode to which the web survey mode was compared, the survey sponsor, the age of the survey, and the baseline item nonresponse rate of the compared mode. The main practical implication is that while web and other survey modes differ in terms of unit nonresponse (on average), item nonresponse rates in web surveys are similar compared to other modes.