Publication type
Conference Paper
Series
EPUNet Conference: The 2nd Annual Research Conference of the European Panel Users' Network, Berlin, 24-26 June 2004
Authors
Publication date
June 1, 2004
Abstract:
The attrition represents one of the most important problems of panel surveys. Erosion of a panel over time might not be a serious problem if it was evenly spread across all demographic, behavioural and economical subgroups. Unfortunately, in practice, the pattern of attrition is not like this, and particular subgroups are lost in disproportionately large numbers.
Using the UDB 1-8 we examine the extent and the nature of the sample attrition in the 11 countries which have taken part to the ECHP project since its beginning (1994).
The aim of the paper is to give an answer to the question “is the attrition pattern selective across the different countries and for different types of variables?”.
The unit of analysis considered in the paper is the individual belonging to the 1994 sample (approximately 127.000 units). This set has been split into three groups depending on the pattern during the 8 years: 1) non-attritors, individuals that have taken part into the survey each year since 1994; 2) attritors, individuals who leave the sample for ever at a certain wave (the exits due to death are excluded); 3) returnees made of people who do not answer some years and then start answering
again.
Using a non-parametric approach of event history analysis we model the probability to be still in the sample at different duration conditional on some characteristics of the sample units. The paper tries to detect which variables (if there are any) drive the attrition process in the different countries considered.
Subject
Link
- http://epunet.essex.ac.uk/papers/gallo_pap.pdf
#508183