Accounting for Response Biases in Latent-Class Models for Choices and Attitudes

Abstract

We propose a latent-class model (LCM) to analyze contingent-valuation surveys incorporating attitudinal questions capturing protest reasons to identify classes of respondents with similar preferences and attitudes. In contrast to a standard LCM, our model ensures that classes are not contaminated by different types of response biases. Using data regarding the preservation of a world-heritage recreation site, low- and high-protester classes are identified. The difference in estimated willingness to pay (WTP) in these classes is €26, reflecting protest issues. If response biases were ignored, different classes would be identified and the corresponding difference in WTPs would be inflated by 130%. (JEL C35, Q51)