Demand effects in stated preference surveys

We argue that demand effects in stated preference studies are understudied. By demand effects, we mean anything in the survey that unintentionally influences respondents' beliefs about appropriate behavior, which in turn might affect their responses in the survey. We implement two methods for measuring and implicitly reducing the influence of demand effects. The first approach-random selection of good to be valued-does not have any effect on respondent behavior. The second approach-a demand script and a control question with feedback-has a sizable and statistically significant effect on respondent behavior. In particular, estimated marginal willingness to pay for improvements in water quality are substantially (around 50 percent) lower than a control treatment; we attribute this decrease to a reduced demand effect. Our results suggest that stated preference methods tend to lead to biased willingness-to-pay estimates due to demand effects, but that the bias can be reduced using simple measures.