options. In particular, programs A and B are identical, respectively, to pro-
grams C and D. They differ only in that the former are framed in terms of
number of lives saved, whereas the latter are framed in terms of lives lost.
An essential element in the rational theory of choice is the requirement,
known as description invariance, that equivalent representations of a choice
problem should yield the same preferences. That is, an individual’s preference
between options should not depend on the manner in which they are described,
provided the descriptions convey the same information. The majority prefer-
ences expressed in problems 4 and 5, however, violate the principle of descrip-
tion invariance and show that framing the same problem in terms of gains or in
terms of losses gives rise to predictably different choices.
Framing effects are pervasive and are often observed even when the same
respondents answer both versions of a problem. Furthermore, they are found
in the choices of both naive and sophisticated respondents. For example, expe-
rienced physicians made markedly different choices between two alternative
treatments for lung cancer—surgery and radiation therapy—depending on
whether the outcomes of these treatments were described in terms of mortality
rates or in terms of survival rates. Surprisingly, the physicians were just as
susceptible to the effect of framing as were graduate students or clinic patients
(McNeil, Pauker, Sox, and Tversky 1982).
The effectiveness of framing manipulations suggests that people tend to
adopt the frame presented in a problem and evaluate the outcomes in terms of
that frame. Thus, depending on whether a problem is described in terms of
gains or losses, people are likely to exhibit risk-averse or risk-seeking behav-
iors.Aninterestingclassofframingeffectsarisesintheevaluationsofeconomic
transactions that occur in times of inflation.
In one study (Shafir, Diamond, and Tversky 1994), subjects were asked to
imagine that they worked for a company that produced computers in Singa-
pore, and had to sign a contract for the local sale of new computers in that
country. The computers, currently selling for $1,000 apiece, were to be deliv-
ered and paid for a year later. By that time, due to inflation, all prices, includ-
ing production costs and computer prices, were expected to increase by about
20 percent. Subjects had to choose between contract A: selling the computers a
year later for the predetermined price of $1,200 (that is, 20 percent higher than
the current price), and contract B: selling the computers a year later for the
going price at that time. For one group of subjects the options were described
relative to the predetermined price of $1,200. In this frame, contract A appears
riskless because the computers are guaranteed to sell for $1,200 no matter what,
whereas contract B appears risky because the computers’ future price will be
less than $1,200 if inflation is low, and more than $1,200 if inflation is high. A
second group of subjects were presented with the same alternatives described
relative to the computers’ expected future price. Here, contract B appears risk-
less because the computers will be sold next year for their actual price then,
regardless of the rate of inflation. Contract A, on the other hand, appears risky:
the computers are to be sold for $1,200, which may be more than they are
worth if inflation is lower than the anticipated 20 percent, and less than they
are worth if inflation exceeds 20 percent. Because of loss aversion, the con-
tract that appeared riskless in each frame was relatively more attractive than
Decision Making 607