Response rates and nonresponse bias in comparative surveys 1
Dr Koen Beullens (KU Leuven )|
Dr Ineke Stoop (SCP)|
In 1989 the International workshop on household survey nonresponse was founded. At the start Bob Groves proposed a number of collaborative projects: multivariate analysis of correlates of nonresponse, collection of interviewer observations on correlates of nonresponse, construction of models of persuasion and nonresponse bias, sequential design for nonresponse reduction, relationships between nonresponse propensity and measurement error, assembly of arguments for and against survey participation, and a survey of interviewers.
The presentation will show where we are after 25 years. The focus will be on the topics suggested with a focus on comparative surveys.
In the Programme for the International Assessment of Adult Competency (PIAAC), a Background Questionnaire (BQ) is administered prior to an assessment of adult competency (referred to hereafter as “literacy”). Cases that could not complete the BQ due to literacy-related nonresponse (LRNR) reasons (e.g., due to language), and an interviewer was not available who spoke the same language, are referred to as “BQ LRNR” cases. The treatment of the BQ LRNR cases has an impact on results. We discuss several improvements, including fieldwork strategies to reduce the number of BQ LRNR cases, nonresponse adjustment strategies, imputation methods and reporting
The Programme for International Assessment of Adult Competencies offers continuing global assessment of adult skills in multiple cycles. The goal is to produce data that make inferences and comparisons across national populations on the basis of survey samples. Thus countries are required to follow consistent guidelines to facilitate valid comparisons of results internationally. However, survey environments are different across countries, making adherence to guidelines more challenging for some as compared to others. Thus the goal is to minimise the impact of such variations on the international results. This paper discusses topics related to response rates and nonresponse bias in PIAAC.
Existing literature on survey methodology is not rich in studies of the relationship between the type of sample-frames and the patterns of their fieldwork execution. Most studies focus on the impact of the within-household selection of respondent in address and household samples on the imperfect coverage of the individuals comprising the population. This presentation will be devoted to the problem of what effects the different sample-frames have on the response-rates and the post-survey structure of the set of respondents and non-respondents.