JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
Back to the library
經濟學 1979

展望理論:風險下決策的分析

丹尼爾·康納曼 與 阿莫斯·特沃斯基

一筆損失之痛,約是等量獲得之樂的兩倍——我們按「變化」、而非最終財富來選擇。

Choose your version
In depth · the introduction

給你一筆穩拿的 $3,000,或是 80% 的機會拿到 $4,000,多數人會抓住那筆穩拿的——可一旦同樣的數字變成損失,他們又掉頭去賭。兩位心理學家,把這個怪癖變成了一套理論。

核心想法

古典經濟學假定,人們按最終財富來權衡選擇,挑那個平均回報最高的。丹尼爾·康納曼與阿莫斯·特沃斯基卻表明,頭腦並非如此運作。我們評判一個結果,不是拿它去比自己的銀行存款總額,而是把它看作相對於此刻所站之處——一個參照點——的「獲得」或「損失」。而且我們對二者並不一視同仁:一筆損失之痛,大約是同等獲得之樂的兩倍。

正是這一不對稱——損失趨避——加上「我們對一筆變化的第一塊錢,感受比第一千塊更強烈」這一事實,以可預測的方式扭彎了我們的選擇。我們會求穩,以鎖定一筆獲得;卻會冒險,以逃開一筆確定的損失。我們還會對極小的機率反應過度——這正是為什麼同一個人,能在同一週裡既買彩券、又買保險。

它是如何誕生的

康納曼與特沃斯基是以色列的心理學家,1969 年起在耶路撒冷的希伯來大學開始合作,成了科學史上最著名的搭檔之一——兩人如此對等,以至於有時靠擲硬幣來決定誰的名字排在前面。整個 1970 年代,他們做著一些看似簡單的問卷:你更想要這個賭局,還是那個?而答案,一次次地擊穿了當時主流的理性選擇理論。

1979 年,他們在《計量經濟學》上擺出了這一規律——特意發表在一份經濟學期刊上,以觸達他們意圖改變的那個領域。特沃斯基於 1996 年去世;康納曼獲得了 2002 年的諾貝爾經濟學紀念獎,而他此後一生都說,這個獎屬於他們兩人。

它為何重要

它撬開了「完全理性的經濟人」這一假設,開創了行為經濟學。一旦你接受人們會以系統、可預測的方式誤判風險,你就能解釋舊模型解釋不了的行為——市場為何恐慌,我們為何死抱虧損的投資,一個選擇的措辭為何會改變我們的取捨——並據此圍繞「人們實際如何決策」,設計出更好的預設選項、警示與政策。

一個可以想像的畫面

想像一支衡量金錢的溫度計,它的零點不設在絕對零度,而設在室溫——也就是你此刻所在之處。一切都被讀作比「現在」更「暖」(獲得)或更「冷」(損失),而從不是一個絕對的數字。而且刻度盤上「冷」的那半邊被拉長了:掉十度的難受,遠勝過升十度的舒服。挪動你的參照點——比方說,習慣了更高的薪水——整條刻度便隨你滑動,於是昨日的奢侈,成了今天的尋常。

一條 S 形的「價值」曲線穿過一個中心點:獲得在右側緩緩上升,損失在左側陡然下落。滑桿設定賭注多大、獲勝機會多少;一個「獲得/損失」開關翻轉整個情境。工具會告訴你:展望理論預測你會選穩拿還是去賭——而當你切到損失時,預測就翻了過來。

它的位置

兩個世紀以來,經濟學運轉在期望效用之上——這一來自丹尼爾·伯努利、後由約翰·馮·諾伊曼發揚的想法認為,人們最大化的是最終財富的平均效用;本館的納許(1950)與馬可維茲(1952)正建立在它之上。展望理論是經驗上的反駁:一份對「真實選擇如何偏離那個理想」的細緻測繪,秉持著赫伯特·西蒙「有限理性」的精神。從這裡,一條路徑直通向理查德·塞勒的「推力」,通向今天的行為公共政策。

The original document
Original source text
D. Kahneman & A. Tversky · Econometrica 47, no. 2 (1979): 263–291
A choice between gambles
Decision making under risk can be viewed as a choice between prospects or gambles. A prospect (x₁, p₁; …; xₙ, pₙ) is a contract that yields outcome xᵢ with probability pᵢ.
The paper opens by treating expected utility theory not as a description of how people choose but as a hypothesis to be tested — and then tests it with a series of simple two-option "problems" put to respondents, whose near-unanimous answers repeatedly violate the theory's axioms.
The certainty effect
[P]eople overweight outcomes that are considered certain, relative to outcomes which are merely probable — a phenomenon which we label the certainty effect.
Asked to choose between a sure 3,000 and an 80% chance of 4,000 (which is worth 3,200 on average), most people take the sure thing — and this preference for certainty, demonstrated in variants of Maurice Allais's 1953 paradox, cannot be reconciled with the independence axiom of expected utility.
The reflection effect
When the sign of every outcome is reversed, preferences reverse with it. Offered an 80% chance of 4,000 against a certain 3,000, 80% of respondents chose the sure gain; mirror the same problem into losses — an 80% chance of losing 4,000 against a certain loss of 3,000 — and 92% now chose the gamble. People are risk-averse over gains but risk-seeking over losses.
Value is defined on changes, not on final wealth
[T]he value function is (i) defined on deviations from the reference point; (ii) generally concave for gains and commonly convex for losses; (iii) steeper for losses than for gains.
Probabilities, in turn, enter not directly but through decision weights π(p) that overweight small probabilities and are subadditive — the device that lets the same person buy both lottery tickets and insurance.
[ … ]
The paper distinguishes an early editing phase, in which prospects are simplified and coded as gains and losses, from a later evaluation phase that applies V = Σ π(pᵢ) v(xᵢ); it documents the isolation and framing effects, restricts itself to simple prospects, and sketches extensions. The full article is available at the source below.
Econometrica · March 1979