近年來在運用雙界二分選擇法(double-bounded dichotomous choice elicitation method)來估計受訪者願付價格(willingness to pay)的研究中，不再僅局限於單一非市場財貨的探討。這類型的模型中雖然納入了相關性的考量，但並未考慮財貨的願付價格間可能存在明確的大小關係。再者，針對抗議性樣本，以往的作法多半是丟棄不用，然而這顯然不是理想的作法。本文中，我們將建構一個模型來同時探討這兩項議題。此外我們也利用「竹東及朴子地區心臟血管疾病長期追蹤研究」第五循環中「肥胖之願付價格」的資料來進行實證分析。結果顯示，居住於竹東、有工作、曾以特定活動控制體重的受訪者願意支付較高的金額來參加減肥療程。 / Recent studies on estimating WTP prices in terms of the double-bounded dichotomous choice elicitation method are no longer restricted to the situations that discuss only one non-market good. Although a couple of models have been proposed to take the correlations into consideration when multiple scenarios are presented to the respondents, none of them pay attention to the possibilities that the prices themselves might be inherently ordered. This is one of the issues that need to be addressed. Another is about the protest samples. A common but apparently problematic approach is simply ignoring them completely. In this study, we propose a model that is able to take care of both issues simultaneously. In addition, the model is used to estimate WTP values for data collected in CVDFACTS about two weight loss treatments. The results indicate that respondents residing in Chu-Dong County, employed, and ever tried any weight controlled programs are willing to pay a higher price for the new treatments.

利用條件評估法 (contingent valuation method) 來評估非市場財貨之市場隱含價值時，雙界二分選擇法 (doubled-bound dichotomous choice method) 為最普遍的詢價方式。近年來，藉由此詢價方式來估計受訪者心目中的願付價值 (willingness to pay) 之研究中，更將此方法推廣至同時估計兩個以上且具有相關性的非市場財貨。只是文獻中的相關探討多半忽略其間的相關性，此外所採用的模型也有可能導致估計的願付價值會有小於零的情形產生。因此，本文引進了衍生版本的Bivariate Generalized Gamma Distribution，來解決這上述兩個問題。我們並採用「竹東及朴子地區心臟血管疾病之危險因子長期追□研究」中，第五循環的「肥胖之願付價格問卷」來作實證分析。在其餘的條件不變的情況下，分析結果顯示，居住在竹東、女性、教育程度愈高、年紀愈小、體重愈重及收入愈高的受訪者會願意支付較高的金額來接受減肥的療程；此外，認為肥胖會影響工作及社交關係的受訪者也會願意支付較高的金額。 / In a contingent valuation survey, it is quite often that subjects were asked to respond to more than one WTP (willingness-to-pay) scenarios. Under such a circumstance, responses provided by a subject are clearly correlated. Although the issue is well recognized in the past, in practice a popular strategy in analyzing this sort of data, however, simply ignore the issue and treat them as if they were totally uncorrelated. Concerning that WTP prices can take only non-negative values along with the issue of possible correlation, we propose an “extend bivariate generalized gamma distribution” that can be used to deal with data collected under a two-scenario situation. Applying it to the CVDFACTS study, where subjects were asked to evaluate a medication-only program as well as a medication-and-exercise program, we found that, other things being equal, female subjects, subjects residing in Chu-Dung County, subjects weigh more, subjects with younger age, higher income, and more years of schooling are willing to pay more. In addition, those who think obesity would affect their social activities would also have higher WTP prices.

由於以往在願付價格的研究中，並未考慮到價格再高都願意支付的受訪者，以及價格再低都不願意支付的受訪者，因此可能會造成估計結果的偏差。本文除了沿用Tsai(2005)所建議採用的三要素混合模型，將受訪者區成價格再高都會願意付的人、願意付合理價格的人以及價格再低都不願意付的人等三類受訪者，且將願意付合理價格者的願付價格透過加速失敗模型(accelerated failure time model，簡稱AFT model)來作評估外，進一步就願意付合理價格的人可能會受到起價點的影響或是有「定錨效果」(anchoring effect)產生的問題來作探討，並就CVDFACTS的高血壓之願付價格資料來作實證分析。結果發現教育程度越高、收入越高以及男性對於能降低罹患心臟血管相關疾病新藥的願付價格也越高。此外我們也發現此筆資料確實有起價點偏誤(starting point bias)的情形。 / In the past, studies on willing-to-pay (WTP) usually assume that all the respondents are willing to pay a reasonable price for a non-market good. In reality, some of the respondents might be willing to pay a very high price, while some others unwilling to pay anything. Following Tsai(2005), we adopt a three-component mixture model to take into account these there three types of respondents and modeling those who are willing to pay a reasonable price with an accelerated failure time (AFT) model. In addition, with a slight modification, we also incorporate issues regarding “starting point bias” and “anchoring effect” into the model. An empirical study based on the data collected from “CVDFACTS” shows that males, respondents with higher educational level, and those with higher family income are willing to pay higher prices for a new hypertension treatment. In addition, it is quite likely that “starting point bias” exists in the dataset.