Effects of attrition in the Norwegian Survey on statistics on income and living conditions (EU-Silc)
In this paper we investigate attrition by using the Norwegian Survey on statistics on income and living conditions, EU-Silc, as a case. We define attrition as unit nonresponse of eligible cases that occurs after the first wave of the panel survey. There are two issues we investigate:
- What characteristics are associated with attrition?
- What it the effect of attrition?
The reason we are concerned about attrition is that loss of respondents over time will reduce the sample size, and may lead to less precise estimates. Also, attrition may be of a selective nature and thereby increase the bias of certain estimates.
By using relevant theory, we discuss causes for attrition, and try to identify relevant socio-demographic variables that are related to attrition. We are concerned if nonresponse in later waves differs in significant ways from initial nonresponse, and therefore introduces an unknown bias in survey estimates. We theoretically argue that the socio-demographic variables household income, education, household, age, employment and material status are important when it comes to attrition.
In the paper we demonstrate that household income, education and age influence the response rate in the Norwegian EU-silc. When it comes to attrition, we also find that these variables are important; we have also added two survey variables we think are important for attrition. These are employment status and health. We find that attrition follows many of the same patterns as nonresponse, and that employment status and health affect the likelihood to continue in the panel. Still, the most important characteristics for attrition are low education and a low household income. Since these variables are highly correlated we use a statistical model to adjust for other background variables so we can find the propensity to participate in previous waves. The model indicates that there is a much lower propensity to continue participating if you are in older age groups and have low education.