As documented by empirical research, tax compliance costs are a considerable burden for private businesses. However, cost estimates may be biased due to survey non-response and questionnaire framing effects. This paper investigates the impact of both aspects on cost estimates. We do not find significant evidence for a non-response bias. By contrast, our results indicate that framing effects regarding the temporal dimension of cost measurement (temporal framing effects) might alter cost estimates by about 39 percent downwards (65 percent upwards) on average and by up to 53 percent downwards (respectively 112 percent upwards) for small businesses. We also test a number of cost drivers with a focus on e-government features. We do not find any evidence that the use of Belgian e-government applications in 2002 and 2004 significantly reduced compliance costs.