Faculty of Sociology, Adam Mickiewicz University, Poznań, Poland.
PLoS One. 2023 Mar 16;18(3):e0283092. doi: 10.1371/journal.pone.0283092. eCollection 2023.
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.
调查不回复率和实地工作成本的持续增加是调查研究的现实。加上当今世界上发生的其他不可预测的事件，这一增加带来了一个挑战：必须加快从面对面数据收集向通常被认为会导致较低回复率的不同模式的转变。然而，最近的研究已经确定，简单的回复率是衡量研究质量的一个薄弱指标。因此，本文旨在分析调查特征，特别是调查模式，对无回应偏差的影响。使用的偏差衡量标准是 Sodeur 首次提出并由 Kohler 首次应用的内部标准。分析基于国际社会调查方案 1996-2018 年各轮次和欧洲社会调查 1 至 9 轮次的调查文件和结果。为了估计调查模式或模式、抽样设计、实地工作经验、数据收集年份和回复率对无回应偏差指标的影响，基于来自每个有人居住的大陆的国家的数据，创建了随机效应三级元回归模型。还提出了几种在群集内嵌套观察的方法。结果表明，与使用面对面模式调查相比，使用邮件和某些类型的混合模式调查与较低的无回应偏差有关。
