To exclude or not to exclude? The effect of protest responses on the economic value of an iconic urban heritage tree.

It is a common practice for many studies using the contingent valuation method (CVM) to censor protest responses, but this practice may make data structures incomplete and lead to sample selection bias at the final analysis stage for economic evaluation. This study adopted the CVM to assess the economic value of Qiedong-Wang-Gong (Bischofia javanica) (QWG), the only urban heritage tree over a thousand years old in Taiwan, and analyzed the effect of protest reponses. A total of 307 valid samples were collected and 27% of the respondents were categorized to protest responses. This study used discriminant analysis and multinomial logistic regression to distinguish the types of protest samples. The results showed that the overall classification performance of the multinomial logistic regression is better than that of the discriminant analysis. The Tobit model was applied to deal protest responses as legitimate zero responses and the result showed that the model underestimated the economic values of QWG significantly. The Heckman two-stage model was utilized to evaluate the effect of censoring protest responses. The result revealed that excluding protest responses from further valuation analysis did not lead to significant sample selection bias. In this study, many respondents failed to quantify their preferences toward protecting QWG in terms of willingness to pay. This study suggested further studies could highlight the current policy about ancient urban trees management to decrease respondents' protest beliefs in CVM studies. [ABSTRACT FROM AUTHOR]