分離效應存在嗎？資產可比較性對風險決策之影響

Is There an Isolation Effect? Effects of Asset Comparability in Decision-Making under Risk

在賭局選擇中，依據Kahneman與Tversky（1979, 2000）之看法：1.決策者將兩組賭局中共同訊息（如：個人身上「資產」等）予以忽略，這種分離訊息之現象稱之分離效應（isolation effects, IE），2.賭客面對邏輯相同之賭局，將因賭局採「贏或輸」之描述而有不同的風險偏好。因此，影響決策的因素是來自賭客對賭局「贏或輸」之分類而非其身上之資產。道意謂著：風險下偏好行為中，賭客著重決策當下財富「贏或輸」之變化而非預估賭局結束後個人身上最後之資產（final asset）。本文認為IE實驗中「資產」缺乏比較對象以致無法將其歸為某一類型之「財富」，這是導致「資產」分離之主因，研究中採分離與合併兩種作業檢定此看法。實驗一設計分離作業以單一方式呈現「資產」（如分別呈現「1000元」或「2000元」），這時單一「資產」缺乏「可比較」對象而無法歸類；設計合併作業以成對方式呈現「資產」（如同時呈現「1000元」與「2000元」），此時「資產」雖與分離作業相同，但成對的「資產」彼此可相互比較並歸為「少」與「多」之類別。實驗二如同實驗一設計分離與合併兩種作業，但「資產」改採已歸類之詞彙，如：「富翁」與「三級貧民」。實驗一分離作業實驗結果複製IE，但實驗一合併作業及實驗二實驗結果：1.歸類為「多」或「富」類別者風險偏好傾向追逐風險或中立，2.歸類為「少」或「貧」者傾向嫌惡風險或中立。顯示經由資產比較後之財富類別會影響風險偏好，亦即個人「資產」是介入於決策之中，IE所指「資產」分離並未被支持。

Kahneman and Tversky (1979, 2000) argued that what influences a decision most are not the final asset but the change of wealth. According to their interpretation, in most cases it is the status quo, rather than the final asset, that serves as the reference point. The isolation effect (IE) occurs as a consequence of information about the asset being isolated from the current situations. Kahneman and Tversky further highlights the importance of the classification of current decision outcomes; when the outcomes are classified as "positive (for example, gains)", decision makers tend to be risk-averse. They become risk-seeking when decision outcomes are classified as "negative (for example, losses)". This article, however, contends that the cause of this differentiated attitude is actually the comparability of the asset. Since isolation arises because subjects oftentimes are not able to classify asset information in choice situations. Distinction between the relative asset and the absolute asset is then made; asset in the presence of compared stimulus is relative, rather than absolute. Traditional IE experiments present no compared stimulus rendering the asset absolute, as a result the asset can have no impacts on the choice, hence the isolation effects. In the case of compared stimulus which makes asset relative, information about the asset can be classified into meaningful categories such as "less" or "more", which then proceeds to impact the choice. Two experiments reported in this article have been designed to examine asset comparability. Experiment one uses absolute asset as the control condition (separation tasks), and in the experimental condition comparative information is provided (joint tasks). The separation task presents information about one asset. However in the joint task, information about several assets is presented simultaneously (for example, both NT $1,000 and NT $2,000 are introduced). Results show that although the given asset for both tasks is the same, only in the joint task condition could one classify the given asset as "less" or "more". In experiment two, the asset is given in classified terms such as "poor" or "rich". Subjects are predicted to take into consideration such classification and demonstrate risk-averse attitude when the asset is classified as "less" or "poor", while they become risk-seeking when the asset is classified as "more" or "rich". The isolation effect is observed in the separation task condition in experiment one, which replicates the Kahneman and Tversky finding. However isolation effect was not found in the joint task conditions for both Experiment one and Experiment two. Instead, when the asset has been classified as "more" or "rich", subjects showed risk-seeking or neutral attitude, whereas risk-averse or neutral attitude was observed when the asset was classified as "less" or "poor". This result is inconsistent with the findings of traditional IE studies, yet in general supports the conception of asset comparability, Findings of this article suggest that Isolation Effect is an exception rather than a norm in decision situations, that is, the isolation effect stands fast only in cases of absolute asset where compared stimulus is not available.