Non-response is a matter of great concern to national statistical offices and a key issue for any survey because it can introduce bias to survey estimation. However, in this presentation, we focus on a business's decision to participate in a survey as an example of organisational behaviour and draw on basic organisational theory to explain why businesses may not respond to surveys (eg Tomaskovic-Devey et al 1994). The data are drawn from the Statistics New Zealand Respondent Management System, which links the response history of individual businesses in all Statistics NZ surveys with information from the Statistics NZ Business Frame. We apply cross-sectional and longitudinal logistic regression random effects models to investigate the ways that business characteristics, business dynamics, previous business response behaviour, and Statistics NZ practices affect the likelihood of a business responding. The Annual Enterprise Survey is used as a test case and the model is fitted across the response history from 2003 to 2007. The analysis presented here is the first phase of a larger scheduled project to assess the impact of non-response bias in Statistics NZ business surveys.