This paper concerns design of contingent valuation experiments when interest is in knowing whether respondents have positive willingness to pay and if so, if they are willing to pay a certain amount for a specified good. A trinomial spike model is used to model the response. Locally D- and c-optimal designs are derived and it is shown that any locally optimal design can be deduced from the locally optimal design for the case when one of the model parameters is standardized. It is demonstrated how information about the parameters, e.g. from pilot studies, can be used to construct minimax and maximin efficient designs, for which the best guaranteed value of the criterion function or efficiency function is sought under the assumption that the parameter values are within certain regions. The proposed methodology is illustrated in an application where the value of environmentally friendly produced clothes is evaluated.