在過去的條件評估法研究中，對於抗議性答覆或是答覆「無法確定」與「不知道」之樣本，慣例的處理方式是將這些樣本由總樣本中刪除，而其剩餘的則被視為合理的有效樣本。然刪除大量抗議性答覆或是答覆無法確定者，不但將縮小原來所規劃的樣本規模，更有可能因為蓄意選擇剩餘的有效答覆而產生抽樣偏差，故只分析所謂的合理有效觀察值，對最後環境資源價值的評估結果將亦可能產生偏誤。此外，從另一個角度來看，這一些樣本之反應，事實上可視為一種不同的意見。因此，只要系統性分析填答有抗議成分及答覆無法確定或不知道者，與有效答覆之間的差別，進而為這類觀察值之反應找到合宜的解釋，即可透過合理模型將這一群人重新納入效益評估範圍中。如此不僅可以減少因大量被刪除樣本對最後評估結果之影響，更可能降低依此所作的資源配置決策上之偏誤。 有鑑於此，本文首先將歸納文獻中之抗議性樣本與無法確定或答覆不知道樣本的理由，進而分析一套2000年利用開放雙界二元選擇條件評估法，評估墾丁國家公園資源價值之資料，抗議性答覆與答覆無法確定或不知道樣本之特質，同時，將逐一對照檢定這些樣本及傳統所認定的有效答覆在屬性之差異。最後，則經由對這些樣本特質的掌握與瞭解，及對抗議性答覆與答覆無法確定或不知道樣本選擇行為的不同詮釋和歸類下，修正有效答覆樣本的內涵與範圍，並比較不同詮釋與歸類下，有無包括抗議性與答覆無法確定或不知道樣本，對資源效益評估結果之差異。 這一套資料中有37.13％樣本屬抗議性及無法確定答覆，而檢定結果顯示無法確定答覆在某些屬性表現上，部分偏向有效答覆的回答，也有部分則偏向抗議的一方。對於無法確定答覆，乃分別假設這些答覆的開放式願意支付，為其第一次所面對受訪價格、該受訪價格的一半、及零價格，或是將其視為抗議性答覆等四種方式，重新認定這一群人為有效答覆，而不論是對於抗議性答覆或是無法確定答覆都分別以OLS、Tobit與複檻式決策過程模型進行估計。 對於樣本的不同詮釋，同時採用不同的估計模式，不同樣本之估計結果顯示，過去未將抗議性答覆與無法確定答覆納入效益評估的分析中，估計所得的墾丁國家公園資源之價值是偏高的，而高估的倍數比起加入抗議性樣本之情形，以最佳解釋能力之複檻式決策過程模型之估計結果多出66％至168％，而比起進一步將無法確定樣本，以複檻式決策過程模型之估計結果多出112％至157％。這些結果與預期中認為加入抗議性答覆與無法確定答覆，會得到較低的願付價格之平均值是一致的，而本研究更進一步提供了，不納入這些樣本，同時也未採用恰當的估計模型時，平均願付價格的高估程度。

The protest responses and responses with "uncertain" are normally excluded from the final analysis in the benefit estimation in the past research. Exclusion of these responses will not only change the planned sample size but also bias the final benefit measurement. On the other hand, the responses of these samples could be deemed as alternative opinions to the survey. As long as the characteristics of these responses are realized and the effective sample size are modified accordingly the bias of benefit measurement should be controlled substantially. The purpose of this study ins then to systematically analyze the characteristics of protest responses and responses with "uncertain" for a set of data collected through double-bounded dichotomous choice with open-ended follow-up contingent valuation method for the benefit evaluation of Kenting National Park. The characteristics will be accomplished by a comparison between these two types of responses and that is deemed as effective sample traditionally. The results are to be used for the explanation and modification of traditionally effective sample. The benefit estimation will then be conducted and compared under different modified samples. There are about 37.13% of responses that are characterized as protests and "uncertain" responses in this data set. The test results have indicated that the "uncertain" responses have certain characteristics toward effective responses and some toward protest responses. The effective samples are modified by assigning different offered price to the responses with "uncertain." All these modified samples are estimated by model of OLS, Tobit, and multiple-hurdle with decision process. The benefit estimation from the sample excluding the protest responses and responses with "uncertain" is biased upward. That is, the traditionally effective sample has over estimated the benefit by about 66%-168% as compared with the sample modified by inclusion of protest responses under multiple-hurdle with decision process model. It is also consistent with the expectation that it has over estimated the benefit by about 112%-157% as compared with the sample modified by inclusion of both the protest responses and responses with "uncertain."