Survey mode and nonresponse bias: A meta-analysis based on the data from the international social survey programme waves 1996–2018 and the European social survey rounds 1 to 9

The constant increase in survey nonresponse and fieldwork costs are the reality of survey research. Together with other unpredictable events occurring in the world today, this increase poses a challenge: the necessity to accelerate a switch from face-to-face data collection to different modes, that have usually been considered to result in lower response rates. However, recent research has established that the simple response rate is a feeble measure of study quality. Therefore, this article aims to analyze the effect of survey characteristics, especially the survey mode, on the nonresponse bias. The bias measure used is the internal criteria first proposed by Sodeur and first applied by Kohler. The analysis is based on the survey documentation and results from the International Social Survey Programme waves 1996-2018 and the European Social Survey rounds 1 to 9. Random-effects three-level meta-regression models, based on data from countries from each inhabited continent, were created in order to estimate the impact of the survey mode or modes, sampling design, fieldwork experience, year of data collection, and response rate on the nonresponse bias indicator. Several ways of nesting observations within clusters were also proposed. The results suggest that using mail and some types of mixed-mode surveys were connected to lower nonresponse bias than using face-to-face mode surveys.