We compare different contingent valuation question formats with each other
and with observed behaviour for a non-monetary estimation task, the expecte
d number of kilometers travelled by automobile. Open-ended questions, open-
ended follow-up questions, dichotomous choice (DC) questions, and double-bo
und DC questions are included. The single and double-bound DC questions res
ult in an estimated mean about twice as high as the actual value and the op
en-ended mean. The DC question overestimation seems to be due to an anchori
ng effect leading to "yea-saying" behaviour. Our results about the differen
ce between DC questions and open-ended questions is consistent with the pat
tern observed in contingent valuations studies of the willingness to pay. O
ur results indicates that DC questions seem to be associated with a general
overestimation problem that is present even for simple non-monetary estima
tion tasks.