Publication type
Understanding Society Working Paper Series
Series Number
2025-13
Series
Understanding Society Working Paper Series
Authors
Publication date
June 19, 2025
Abstract:
Nonresponse is a critical challenge in surveys aiming to be representative, necessitating effective correction strategies. To tackle nonresponse, one must identify predictors available for both respondents and nonrespondents. Address-based samples often provide limited information from the sampling frame; however, full address details can be linked to external data sources that may include predictors of nonresponse. While the quantity of external data and potential predictors is vast, linking such data is a labour-intensive and costly task.In this paper, we utilize linked datasets and systematically evaluate a large volume of potential predictors to identify the most significant ones. We also assess the stability of these predictors across different modes of data collection. Using data from the wave 14 boost sample of UKHLS, we analyse nonresponse predictors for face-to-face first with web follow up and web-first followed by face-to-face. Our findings will guide researchers working with address-based samples by identifying the most influential nonresponse predictors and offering practical insights for survey design and implementation.
Subjects
Paper download#588664