Abstract
Survey quality is an important goal for National Statistics Offices(NSOs) and non-response negatively affects this quality. Accordingly, it is important to understand the determinants and patterns of non-response, with the aim of providing recommendations as to how to reduce or mitigate non-response. This study looks to better categorise sampled firms based on their previous response behaviour, using trajectory modelling. Trajectory modelling is a tool for distinguishing patterns of behaviour over time, with dual trajectory modelling allowing us to analyse how outcomes in one time period are related to distinct outcomes in time two [6]. Using dual trajectory modelling to examine longitudinal response patterns is a new application of this model.
Statistics New Zealand's Retail Trade Survey is used as a test case and the model is fitted across the response history 2003 to 2007.We were able to identify three response groups, the good-responders, the group whose likelihood of response improves over time and the non-respondents. These three groups then form the basis for four groups we identified in time two. This then informs some tentative recommendations as how to best target sample members, based on previous response behaviour.
Get full access to this article
View all access options for this article.