Use of nonresponse weighting and calibration for bias reduction in quota sampling

초록

One limitation of quota sampling is its inability to address unit nonresponse, where a nonresponding unit is simply replaced by another willing participant until the quota criteria are met. In such cases, using the sample mean from quota samples can lead to biased estimates due to unit nonresponse. In this study, we explored the use of nonresponse weighting and calibration to reduce bias arising from unit nonresponse in quota sampling through a simulation approach. The simulation population was constructed to reflect aggregated results from public opinion polls conducted over a specified period. Under our simulation settings, incorporating additional auxiliary information, such as final level of education and political orientation, for both respondents and nonrespondents proved effective in reducing bias. In particular, the combined use of calibration and nonresponse adjustment achieved the best performance in reducing both bias and mean squared error. We also proposed a bootstrap method for variance estimation under these adjustment strategies, which demonstrated satisfactory performance within the scope of our simulation. © 2025 The Korean Statistical Society, and Korean International Statistical Society. All rights reserved.

키워드