This study's objective is to determine the factors that cause divergences over time (differences) between stated willingness to pay (WTP) and actual WTP (purchase behaviour), and consider whether any divergence signifies the presence of bias in contingent valuation studies. Stated WTP for insecticide-treated bed-nets (ITNs) was elicited from a random sample of respondents using three question formats in Nigeria. The question formats were the bidding game (BG), binary with follow-up (BWFU) and a novel structured haggling (SH) technique. The sales of the nets and a second survey were conducted 1 month after the first survey. In the second survey, factors that might explain the divergences were built into the questionnaire and these together with socio-economic variables were examined for causes of divergences in WTP. Data were analysed using non-parametric tests, testing of means and cross-tabulations. There were divergences in WTP in all three question formats: 69.4% in the BG, 78.7% in the BWFU and 48.8% in the SH. The higher the stated WTP, the more likely the divergence between stated and actual WTP. The attitude of the community leaders to the ITNs in the BG (p<0.05), the time respondents had to think about their WTP (p<0.05) and the external information they received about the ITNs in the BWFU (p<0.05) all led to divergences in WTP. We conclude that there are genuine causes of divergences between stated and actual WTP across the three question formats, and that the lesser the criterion validity score, the more the level of divergence in WTP. Studies that compare stated and actual WTP should explicitly determine the causes of divergences in order to assess the role of bias in the divergences.