Survey design, survey response behaviour and the dynamics of self-reported health and disability

Presenter: Steve Pudney, ISER, University of Essex

Author: Steve Pudney

Co-author(s): Annette Jäckle

Academic and policy research on disability is largely based on survey questions asking respondents to self-assess their health. For example, respondents are asked whether they have a long-standing illness, disability or infirmity, and if yes, whether they have difficulties with 11 specific Activities of Daily Life. However asking respondents repeatedly whether they have long-standing problems produces year-on-year transition rates into and out of disability that are implausibly high.

We analyse an experiment carried in Waves 6 and 7 of the Understanding Society Innovation Panel that was designed (1) to identify reasons for the high rates of change in long-term illness/disability status, (2) to investigate whether the use of the initial filter question has significant impact on measured disability, and (3) to test question versions that would produce more stable measures.

Initial results suggest that long-term health problems are not well defined. A majority of respondents who stop reporting a long-term problem still have the condition, however it has improved, treatment or medication is more effective, or their activities have changed and it is less of a problem. Among respondents who start reporting a problem, a majority claim that they already had the condition at the previous interview. As a consequence, if all respondents are asked about difficulties with Activities of Daily Life, disability rates are significantly higher than if routing is based on reporting a long-term health problem. Further analyses will assess the biasing effects of errors in health status on estimates of disability incidence and models of health dynamics.