Unveiling participant level determinants of unit non-response in business tendency surveys

Business Tendency Surveys (BTS) continue to be an important source of timely information on business cycles in many countries. We address quality of economic survey data by uncovering the relation between unit non-response and participant characteristics on company respectively respondent level. We use a unique, matched dataset that merges rich business tendency survey panel data with data from an exclusively conducted meta survey. Our meta information enhances the set of firm characteristics by information such as valuation of business tendency surveys or perceived response burden. We use different count data models to explain non-response count. Our models include weighted count data regressions as well as a two part hurdle model. We find that response burden, a company's survey track record, timeliness and participation mode are the strongest and most robust predictors of unit non-response. We also find a weaker negative effect of the business situation on unit response. Remarkably we do not find a significant influence of neither company size nor valuation of BTS on the propensity to respond to periodical qualitative BTS.