《Using paradata to identify》.pdf

Using paradata to identify response styles and respondent profiles for adaptive survey design Barry Schouten - Stats Netherlands Utrecht University RSS meeting – Paradata: From Survey Research to Practice Outline – Motivation – Response styles, paradata and profiles – Case studies LFS 2008 and CVS 2011 – Case study LFS – LISS panel 2013-2014 – Discussion 2 Motivation – multi‐mode surveys Research and case studies mostly motivated from need to: – Reduce mode effects; – Include survey mode as design feature in adaptive survey design; – Find heuristics to perform costly transition periods and parallel runs; Options to reduce mode effects: 1. Prevent through questionnaire design; 2. Avoid through data collection design (possibly adaptive); 3. Adjust through the estimation design; 4. Stabilize through estimation design; Questionnaire profiles, respondent profiles and response styles may help: – To set a benchmark design; – To reduce dimensions for multi-purpose, diverse surveys; – To filter for pre-testing and for transition periods; – To adapt to respondent; 3 Adaptive survey design Adaptive survey designs and responsive survey designs are motivated by two observations: • Persons have different preferences for communication and interview, i.e. respond differently to difference data collection strategies; • Different data collection strategies are associated with different costs per person; Scope/implementation/approach dependent on setting: • Length of data collection period and number of instances for intervention • Refusal conversion • Strength of prior knowledge (frame/registry data, paradata previous waves) • Focus on learning during data collection or from wave to wave • Focus on structural and inc