Publication type
Conference Paper
Series
BHPS-2007 Conference: the 2007 British Household Panel Survey Research Conference, 5 July -7 July 2007, Colchester, UK
Author
Publication date
June 1, 2007
Abstract:
Attrition is a form of unit non-response in panel studies, culminating in the unintended and permanent loss of target sample members. Attrition is problematic for two reasons. First, the loss of sample members reduces the overall sample size resulting in reduced precision in estimates obtained from panel models. Second, attrition may not be random and as a result substantive findings derived from panel analysis may suffer bias. Analytically, panel attrition is considered an absorbing state, distinct from interim unit non-response, although distinguishing between them is not an entirely straightforward task. While some panel studies treat initial nonresponders as attritors by design, many other studies allow for interim unit non-response and attempt to interview non-respondents at future waves. Typically, methodologists studying attrition treat the first instance of nonresponse as panel attrition whether the survey’s design does the same or not. Using response history data from the British Household Panel Survey, I depart from this usual approach to defining attrition and study panel non-response as a repeat process. I combine a discrete time hazard model of non-response with a parallel hazard model of study compliance subsequent to wave non-response. This approach addresses the panel participation process in total and allows me to unpack the causes of different response trajectories over the life of a panel study. How researchers deal with unit non-response depends on their choice of estimators, whether balanced or unbalanced panel data are acceptable, etc… To facilitate such substantive analysis of the British Household Panel Study, I document the correlates of panel non-response for use in otherwise substantive models. Moreover, by examining the response process in total, this research will inform the decisions survey designers and research organisations make about maximising unit response over the life of a panel.
Subject
Link
- http://www.iser.essex.ac.uk/bhps/2007/programme/data/abstracts/Uhrig.pdf
Related Publications
-
The Right End of Attrition: Repeated non-response in the British Household Panel Study
S.C. Noah Uhrig,Conference Paper - 20070518
#509290