1. Personality Predictors of Extreme Response Style
2. Welfare Evaluations in Contingent Valuation Experiments with Discrete Responses
3. A comparison of alternative certainty calibration techniques in contingent valuation
4. Some relationships among sex, anxiety, and conservatism of judgment
5. An ordered logit might appear to be a logical specification in estimating response probabilities, since distributions for WTP and bid amount could be expected to move in one direction only. We explored this possibility and found that estimation results do not seem to be ordered purely by “Bid,” but instead it appears that NS may not fall between PY and PN. We estimated model (1) using an ordered logit with the ordering moving along the continuum from DN through DY with NS in the middle and with the order of NS and PN switched. There was almost no difference in goodness of fit between the two models, so we cannot rule out that the NS and PN categories are not different. However, compared with the multinomial logit results, the covariates influence these two categories quite differently. While the ordering appears to work well for the other categories, the results on the NS category are not consistent with a simple ordering. Moreover, exploratory research suggests that the connection of this category to the covariates is not what one would expect from a simple ordering. It is likely that this is a mixed category that needs more detailed exploration in future research. For these reasons, we proceed with the multinomial logit specification, where the structural restrictions on the data are minimal.