An empirical analysis of the association between neighborhood income and unit non-Response in the survey of health, ageing, and retirement in europe

This paper estimates associations between individual and neighborhood characteristics and unit non-response in a survey of the population aged 50 and over in the Netherlands in 2004. The statistical model includes interviewer fixed effects to control for the non-random distribution of addresses over interviewers. The empirical analysis shows that, relative to individuals living in apartments, there is a lower unit non-response among individuals living in houses and a higher unit non-response among individuals living in old age institutions. Unit non-response is positively associated with the size of a city. No age and gender effects are found. Unit non-response is about 25 percent lower among individuals in the top than among individuals in the bottom of the distribution of neighborhood average income. This latter result implies that the response sample is biased toward individuals living in the more wealthy neighborhoods.