Valuing public goods: Discrete versus continuous contingent-valuation responses

Independent applications of open-ended and dichotomous-choice formats are compared using tests of means, estimating joint likelihood functions and nonparametric tests of distributions. The null hypothesis of no difference in the open-ended and dichotomous-choice estimates of central tendency cannot be rejected for two out of three data sets, while estimated standard deviations are significantly different for all three data sets. In addition, actual dichotomous-choice means and standard deviations exceed those from comparable synthetic dichotomous-choice data sets, suggesting either open-ended questions underestimate values or dichotomous-choice bid structures may lead to systematic overestimates.