框架效應存在嗎？籌碼效應對風險決策之影響

Is There a Framing Effect? The Asset Effect in Decision-Making under Risk

Tversky與Kahneman（1981; Kahneman & Tversky, 2000）認為決策者面對邏輯上相同的兩組決策題目，其會因題目採「正或負」詞彙描述而產生不同的風險偏好。當問題以「贏、挽救或救活」等正面詞彙描述時，決策者風險偏好傾向嫌惡風險；但若以「輸、虧損或死亡」等負面詞彙時則傾向追逐風險，此稱為框架效應（framing effects）。本文提出籌碼效應（asset effects）指出：決策者所擁有的籌碼應比決策題目本身採「正或負」的框架更會影響偏好選擇。文中依循Tversky與Kahneman（1981）作業典範，設計兩個實驗檢定這看法，實驗一、二分別為「金融危機」與「亞洲疾病」問題。兩個實驗所提供的籌碼皆同時包含兩種數量，前者的籌碼為資金：數量分別為600萬元與600億元；後者的籌碼為人口：數量分別為600人與600萬人。結果發現不論題目為「正或負」，實驗一資金籌碼多者（600億元）較傾向追逐風險，反之，則較傾向嫌惡風險；實驗二人口籌碼多者（600萬人）也較傾向追逐風險，反之，則較傾向嫌惡風險或風險中立。此與框架效應之預測有所抵觸。籌碼效應認為：「相對於籌碼少的決策者，面對相同損失時籌碼多者其損失的比例較低，應較可承擔決策失敗時的風險。」故增加籌碼將使得人們承擔失敗時損失的容忍性提高，因而增強其對風險追逐之傾向，反之籌碼一旦減少將降低此種傾向。整體而言，實驗一、二大致支持這看法。

Tversky and Kahneman (1981; Kahneman & Tversky, 2000) postulated that when a decision task is described as "positive" or "life-saving", subjects tend to be risk-averse in their choice; however, when the task is described as "negative" or "life-threatening", subjects tend to be risk-seeking in their choice. Objectively the task remains the same in both cases, nevertheless different descriptions often result in different risk attitudes-- this is known as framing effect. This article asserts that a decision-maker's asset level affects the choice more than the framing of the task does. Two experiments are reported to support this view. In experiment one an "Asian financial crisis" task was constructed. In Experiment two, Tversky and Kahneman's Asian disease task was adopted for comparison. Both tasks involved two levels of asset (NT$ 6 million versus NT$ 60 billion in the Asian financial crisis task; 600 people versus 6 million people in the Asian disease task). The results of the Asian financial crisis task show that subjects tend to be risk-seeking when the asset level is high, but they tend to become risk-averse when the asset level is low. The Asian disease task reports similar results-- subjects tend to be risk-seeking when the asset level is high, and risk-averse or neutral when the asset level is low. These results suggest that it is the asset level, rather than the framing of the task, which eventually affects the decision when an asset effect (AE) is exposed. Furthermore, relationships among the framing effect, the isolation effect and the reflection effect (Kahneman & Tversky, 1979) are clarified. The clarification could help experimenters to understand more clearly the potential impact that AE may have on framing. AE indicates that given the same amount of loss, a decision-maker with a higher asset level can take on more risks since the loss can be tolerated better than one with a lower asset level. This suggests that threshold for risk and loss becomes higher when the asset level increases, while tolerance for risk and loss drops as the asset level drops. We can thus predict that those with high asset level tend to be risk-seeking while those with low asset level risk-averse. A further implication is evidenced by the two experiments conducted-- when the information of the asset levels is manipulated, framing will have a limited impact on the final decision.