that, as in visual perception, people have a “gestalt” notion of an ideal distribution
of outcomes in time, which includes interactions across time periods that violate
simple separability axioms.
FAIRNESS AND SOCIAL PREFERENCES
The assumption that people maximize their own wealth and other personal mate-
rial goals (hereafter, just “self-interest”) is a widely correct simplification that is
often useful in economics. However, people may sometimes choose to “spend”
their wealth to punish others who have harmed them, reward those who have
helped, or to make outcomes fairer. Just as understanding demand for goods re-
quires specific utility functions, the key to understanding this sort of social pref-
erence is a parsimonious specification of “social utility,” which can explain many
types of data with a single function.
An experimental game that has proved to be a useful workhorse for identifying
departures from self-interest is the “ultimatum” game, first studied by Güth et al.
(1982). In an ultimatum game, a proposer has an amount of money, typically
about $10, from which he must propose a division between himself and a respon-
der. (The players are anonymous and will never see each other again.) If the
responder accepts the offered split, they both get paid and the game ends. If she
rejects the offer, they get nothing and the game ends. In studies in more than 20
countries, the vast majority of proposers offer between a third and a half of the
total, and responders reject offers of less than a fifth of the total about half of the
time. A responder who rejects an offer is spending money to punish somebody
who has behaved unfairly.
A “trust” game can be used to explore the opposite pattern, “positive reciproc-
ity.” Positive reciprocity means that players are disposed to reward those who
have helped them, even at a cost to themselves. In a typical trust game, one player
has a pot of money, again typically around $10, from which he can choose to keep
some amount for himself, and to invest the remaining amount X, between $0 and
$10, and their investment is tripled. A trustee then takes the amount 3X, keeps as
much as she wants, and returns Y. In standard theory terms, the investor-trustee
contract is incomplete and the investor should fear trustee moral hazard. Self-
interested trustees will keep everything (Y
5
0) and self-interested investors who
anticipate this will invest nothing (X
5
0). In fact, in most experiments investors
invest about half and trustees pay back a little less than the investment. Y varies
positively with X, as if trustees feel an obligation to repay trust.
The first to attempt to model these sorts of patterns was Rabin (1993, and this
volume). Fixing player A’s likely choice, player B’s choice determines A’s payoff.
From A’s point of view, B’s choice can be either kind (gives A a lot) or mean (gives
A very little). This enables A to form a numerical judgment about B’s kindness,
which is either negative or positive (zero represents kindness-neutrality). Simi-
larly, A’s action is either kind or mean toward B. In Rabin’s approach, people earn
a utility from the payoff in the game and a utility from the product of their kindness
and the kindness of the other player. Multiplying the two kindness terms generates
both negative and positive reciprocity, or a desire for emotional coordination: If B
Do'stlaringiz bilan baham: