We conducted a choice experiment (CE) to estimate willingness to accept (WTA) values for a planned conditional cash transfer (CCT) programme designed to increase toilet use in South Africa. The payment is made conditional on using a toilet and bringing urine to a central collection point. In a split-sample approach, a segment of respondents were given time to think (TTT) (24 hours) about their responses, while the remaining respondents had to answer immediately. We found significant differences in the choice behaviour between the subsamples. To validate the stated preferences with actual behaviour, a CCT programme was implemented afterwards. The stated WTA estimates were far below those revealed by actual behaviour for both subsamples. Contrary to our expectations, the TTT group had underestimated their actual WTA values by an even larger margin. The preferences for various attributes were nevertheless useful in informing the design of the real intervention.