Unit nonresponse in a dwelling panel study – the role of residents’ and housing characteristics

Assessing the reasons behind nonresponse is essential in survey research. In this paper, we investigate unit nonresponse in a dwelling panel (Cologne Dwelling Panel, five waves between 2010 and 2022), in which dwellings constitute the sampling units. One resident of each dwelling is interviewed in each panel wave. Using information from previous panel waves, we model an individual’s propensity for nonresponse by applying multinomial logistic regression, with socio-demographic information and indicators of satisfaction with their current housing situation and the neighborhood as predictors. Further, we assess whether there is systematic attrition at the dwelling level. We find that when respondents express greater satisfaction with their current housing situation, nonresponse is reduced at the individual level, while smaller apartments and a more heterogeneous and deprived neighborhood are indicative of higher nonresponse on the level of dwellings.