运用Spike模型分析CVM中零响应对价值评估的影响——以南京市居民对长江流域生态补偿的支付意愿为例

条件价值法调查中会有相当比例的受访者不愿意支付费用,对这部分群体如果处理不当会影响到价值评估的可靠性和有效性。以南京市居民对长江流域生态补偿的支付意愿为例,采用双边界两分式问卷调查获取1 403份有效样本。分析了忽略零响应问题的传统模型所造成的支付意愿(Willingness to Pay,WTP)估算偏差,并采用Spike模型对WTP估算进行了校正。结果显示:若直接删除零响应样本,中位数WTP在359~396元/(户·a)之间;而将零响应的实际支付用接近零的正数进行替代,得到的中位数WTP在154~208元/(户·a)之间。根据Spike模型,将真实零响应样本纳入分析后,得到的WTP估算结果在259~288元/(户·a) 之间。表明直接删除零响应会明显高估WTP,而用零或接近零的正数替代零响应又会低估WTP。未来5 a内长江南京段的生态补偿总价值至少为24.4×10⁸元。将决策过程划分为"加入决策"和"支付决策"两个阶段后,论文还发现,"加入决策"和"支付决策"是两种不同的机制,二者在影响因素上存在差异。此外,不同收入家庭的WTP存在明显异质性。

Using Spike Model to Evaluate the Effect of Zero Responseon Welfare Measurement—Evidence from the Willingness to Pay for the Ecological Compensation of the Yangtze River Basin in Nanjing Section

In contingent valuation surveys, data is often complicated by zero willingness to pay (WTP) responses in the sample. Without properly dealing with, zero responses would deeply reduce the reliability and validity of welfare valuation. Based on a case study on the WTP for ecological compensation of the Yangtze River Basin in Nanjing city, 1403 individual questionnaires in closed-ended double bounded dichotomous choice format were collected. Given conventional methods dealing with zero responses will result in estimation bias in terms of WTP, special attention was paid to the issue of zero response in the survey, and a Spike model that allows for a non-zero probability of zero response was applied. Because different types of zero responses have different economic implications, without distinguishing between true zero response and protest responses will result in non-neglecting measure bias. The conventional methods to deal with zero response have been known as dropping it from the analysis, or substituting it with either zero or a near zero positive number. When dropping zero responses, it yielded a WTP between 359 and 396 yuan per household per year. While replacing zero responses with a near zero positive number, it yielded a WTP between 154 and 208 yuan per household per year. However, dropping zero responses will run the risk of sample selection bias, while replacing it with near zero number seems somewhat too subjective. Thus, a Spike model that allows for a non-zero probability of zero WTP in referendum-style was brought into practice. It includes many of the often used models as special cases and allows various analyses that are not possible in standard applications. Under the Spike model, the zero response samples were further divided into protest response and true zero response according to their behind motivations. Then the value of WTP was estimated between 259 and 288 yuan per household per year, which is significantly smaller than the dropping method but larger than the replacing method. The total compensation value paid by Nanjing households was at least 24.4×10⁸ yuan in the following five years. It was also found that there were obvious differences in the determinations of decision-making between the willing to join step and willing to pay steps. Environmental services are luxury, and only when people's income reaches a certain level will they buy them. Thus at the practical level, considering the heterogeneous acceptability and burden of expenditure, special attention needs to be paid to the low-income families.