Uncovering the black box of (non)response in online time-use surveys

Time-use studies are a highly valued research tool to better understand behaviour and the context of conduct. Mainly due to the high participation costs of respondents, the response on these studies is usually relatively low. This raises questions on the selectivity of the response and, correspondingly, the representativeness of the data. This paper attempts to answer these questions. Specifically, we use the Belgian Time Use Survey-data (BTUS) from 2013 for this task. These data have the advantage that they can be combined with the Belgian Labour Force Survey 2013 (LFS). This means that we have an unprecedented potential of information of non-participants, allowing us to take a very broad range of background variables into account. The paper begins with a bivariate description of the (non)response. In a second phase, we hypothesise that 1) “busy” (in an objective and subjective sense of the concept) people are less inclined to participate in a time-consuming time-use study; and 2) that people who are socially well integrated tend to participate more often. To test these hypothesis we use multivariate modelling in order to control for other potentially influential background variables. The paper concludes with some recommendations on how to deal with the nonresponse.