涵蓋抗議性、無法確定/不知道答覆樣本之條件評估法一般化模型建構

Abstract

在使用條件評估法進行自然資源價值評估時，常常會發生受訪者未填答、不知道到如何填答的現象，這類樣本可在詢價過程中進一步的詢問受訪者如此填答的原因，再根據受訪者的回答將其歸類為抗議性樣本或無法確定/不知道答覆樣本。因此，本研究將所有詢價方式依據其詢價過程及結果，分為呈現WTP金額詢價模式、純選擇詢價模式及選擇加開放詢價模式等三類，以建構一個能適用於所有詢價方式，並將抗議性、無法確定/不知道答覆樣本涵蓋於其中的一般化模型。而本研究所建構的模型，所使用的估計方式，皆是一般常用的迴歸方法，故此一般化模型除了能夠完整的處理抗議性及無法確定/不知道答覆樣本外，在計算上也能夠則相對的容易。依此，本研究利用此一能夠處理多種詢價模式、並涵蓋抗議性或無法確定/不知道答覆樣本之一般化模型，以墾丁國家公園及彰化海岸未定濕地兩套樣本為驗證。由於兩套樣本皆採用二元選擇加開放詢價法，恰好能取其結果作為三種詢價模式實證檢驗。此外，同時也利用傳統將抗議性及無法確定/不知道答覆樣本刪除，或是將抗議性樣本視為零的處理方式進行估計，並與本研究之結果做比較。 實證結果顯示，發現選擇加開放詢價模式中的選擇過程與最後估計結果是有關聯的，故以本研究所建構的一般化模式將選擇過程納入考慮是適當的。實證上也證明本研究所建構之模型能夠以較有效、且簡單的估計方式處理涵蓋抗議性或無法確定/不知道答覆樣本之WTP估計，且能適用於過去條件評估法所用的所有詢價方式，同時也能夠修正過去將樣本刪除時造成願付價值的高估，或是將抗議性樣本視為零的低估現象。 在抗議性或無法確定/不知道答覆樣本佔整體樣本比例在一般化模型與傳統不恰當估計及處理方式的比較，以平均每戶的願付價值而言，傳統將樣本刪除高估墾丁國家公園的平均願付價值，在不同模式間高估的幅度則介於2.57%到3.38%；彰化海岸未定濕地的部分，高估的幅度則介於1.27%到3.35%；傳統將抗議性及無法確定/不知道答覆樣本刪除低估墾丁國家公園的平均願付價值，在不同模式間高估的幅度則介於65.77%到141.82%；彰化海岸未定濕地的部分，高估的幅度則介於51.67%到60.78%。由於受到樣本特性等因素影響，以平均每戶願付價值探討一般化模型修正傳統處理的幅度上，有部分的結果並不如所預期當處理的比例越高則幅度越高，但大部分的結果能夠看出此趨勢。然進一步估算兩項自然資源的總價值時，在墾丁國家公園的部分，一般化修正傳統不適當處理低估的幅度則更明顯，約介於30.57%到141.87%之間；彰化海岸未定濕地的部分則低估介於26.41%到60.85%。以自然資源的總價值比較時，可發現同時納入抗議性及無法確定/不知道答覆樣本時，利用本研究一般化模型修正傳統不適當處理之效果將會越明顯。

Conducting contingent valuation method to estimate value of nature resources or environment, there are various reasons for respondents to give protest replies or reply don’t know/uncertainty. Past studies to deal with or estimate sample with these two groups of responses are either to delete the protest and don’t know/uncertainty samples or to consider them as zero values. However, deleting protest sample, especially, might overestimate the final mean willingness to pay. On the other hand, while accounting these samples as zero responses might underestimate mean willingness to pay. There are other methods to deal with protest and/or don’t know/uncertainty responses to solve the potential bias, all these methods, however, are in certain way too complicate to get appropriate estimation results, not to mention the computation of the mean willingness to pay. Among all the elicitation methods employed in contingent valuation method, three types of responses are classified. The first type of elicitation approach is the mode of presenting willingness-to-pay value, open-ended, bidding game, et al. are belonged to this mode. The second type is pure binary choose mode. Single-bounded discrete choice and double-bounded discrete choice are members of this mode. The third type is mode of binary choose with open question follow-up. This study then structures a general model to systematically estimate all three types of elicitation modes mentioned above. Furthermore, this general model treats protest and/or don’t know/uncertainty responses sample in an appropriate way. Most importantly, the general model constructed in this study is easy to estimates and computation of mean willingness to pay will then not confront any difficulty accordingly. Such general model is testified for two sets of data. They are a sample of benefit evaluation of Kenting National Park and a sample of benefit evaluation of Changhua Coast Wetland. The results from the general model constructed in this study demonstrate that the standard errors are lower than those accomplished by traditional models. This indicates that our general model is more efficient. As a result, the mean willingness to pay computed from traditional models is overestimate while protest responses eliminated and underestimate while protest responses are treated as zero values. The aggregate total benefits of the related natural resources are all underestimated by the traditional estimation methods. The underestimated total values are ranged between 30.57% and 141.87% for sample of Kenting National Park and between 26.41% and 60.85% for sample of Changhua Coast Wetland. It is obvious that underestimation of the total values are expected while aggregated benefits of the natural resources are the concern.