Lecture 8: Social Preferences I

EC4394: Behavioral Economics

National University of Singapore

Johannes Haushofer

Spring 2023
Recap and plan for today
• Last few weeks: Time and risk preferences, reference-dependent utility
• Risk preferences: people are risk-averse in the gains domain, risk-seeking in the loss
domain
• Reference-dependent utility: evaluation of prospects happens relative to a reference
point (i.e. in terms of gains and losses), not relative to final outcomes
• Loss aversion and the endowment effect: losses weigh more heavily than gains of the
same absolute magnitude
• Probability weighting: people underweight large probabilities (except for certainty); and
overweight small probabilities

• Today: Social Preferences

Pro-social behavior
• What we just saw are examples of pro-social behavior
• An important branch of behavioral economics has asked whether these
behaviors can be studied experimentally.
• In the early days of behavioral economics, researchers began to ask in earnest
whether such behavior could be studied experimentally.
• Three workhorse games:
• Dictator game
• Ultimatum game
• Trust game
The Dictator Game
• Two players, 1 and 2
• Player 1 receives an endowment (e.g. SGD 10).
Player 2 has 0.
• Player 1 can send any share of their
endowment to Player 2, including 0% or
100%.
• Player 2 receives , Player 1 keeps .
• The game ends
• Often seen as a way to measure “pure altruism”
The Ultimatum Game
• A modified dictator game: Again Player 1 receives an endowment and decides
how much of it to send to Player 2.
• But in contrast to the Dictator Game Player, Player 2 is not a passive recipient:
they can decide whether to accept or reject the split proposed by Player 1.
• If they accept, the split proposed by Player 1 gets implemented:
• Player 2 receives
• Player 1 receives
• If they reject, both get nothing.
• Typically seen as a way to measure adherence to social norms (Player 1), and
enforcement of social norms (Player 2).
The Trust Game
• Two players, 1 and 2
• Player 1 receives an endowment, e.g. SGD 10.
1
Player 2 has 0.
• Player 1 can send any share of their endowment
to Player 2.
• The money travels to Player 2 via the
experimenter, who triples it.
• So Player 2 receives 3 times what Player 1 sent, .
Player 1 now has .
• Player 2 can now decide if they want to send
some share of their new endowment back to
2
Player 1.
• Player 1 now has .
• Player 2 now has .
• Typically seen as a way to measure Trust (Player
1) and Trustworthiness (Player 2)
What does standard theory predict about behavior in these
games?
• What are the predictions of economic theory about
behavior in these games?
• Utility is typically modeled with only the decision-
maker’s own payoff in the utility function:

• I.e., people are selfish: they only care about

themselves and their own payoffs.
What does standard theory predict about behavior in these
games?
• Dominant strategy in the Dictator Game: Player 1:
give nothing
• Dominant strategy in the Ultimatum Game:
• Player 1: Share nothing
• Player 2: Accept any amount > 0
• Dominant strategy in the Trust Game: Reverse
induction 
• Player 2: Return nothing
• Player 1: Send nothing
How do people actually behave in these games?
• Dictator Game:
• People give 28.4% of their endowment on average (Engel, 2011)
• Giving is similar for hypothetical and real stakes, and for large and small stakes

• Ultimatum Game:
• Modal offer is 50–50 split
• Average offer is 40.4% of the pie (Oosterbeek et al., 2004)
• Offers less than 20–30% are typically rejected
• Average rejection rate is 16.2%

• Trust Game:
• Player 1 sends 50% of their endowment on average
• Player 2 typically returns 37% of the amount they have available on average (Johnson &
Mislin, 2011)
Cross-country variation
Cross-country variation
Cross-country variation
Trust-
worthiness

Trust
(Trustworthiness)
Pro-social behavior is universal
• Many people around the world exhibit pro-social behavior
• They give non-zero amounts in the Dictator Game!
• They make non-zero offers in the Ultimatum Game, and reject “unfair”
offers!
• They send money in the Trust Game, and the receiver sends money back!
• This is so weird! According to our standard model, it shouldn’t
happen!
• Remember from above: Utility is typically modeled with only the
decision-maker’s own payoff in the utility function: . People are
selfish.
• So our standard economic model may need adjusting! It doesn’t
Modeling pro-social behavior
Modeling pro-social behavior
• Fehr-Schmidt utility function with social preferences
(two-player case):

• measures aversion to disadvantageous inequality: player ’s

utility is reduced if player ’s payoff is higher than her own.
• How much is it reduced? By times the difference.
• measures aversion to advantageous inequality: player ’s utility
is reduced if player ’s payoff is lower than her own.
• How much is it reduced? By times the difference.
• Assumptions: ; (being “behind” the other person is worse than
being ahead)
Modeling pro-social behavior
Marginal utilities
• Disadvantageous inequality (the other person is ahead)
• You get 1 additional dollar:
• Utility goes up by more than 1: you get an extra dollar, and you’re now less behind.
• The other person gets 1 additional dollar:
• Utility goes down by because you’re now further behind.
• Advantageous inequality (you are ahead)
• You get 1 additional dollar:
• Utility goes up by less than 1: you get an extra dollar, but you’re now more ahead, which you dislike.
• The other person gets 1 additional dollar:
• Utility increases by because you’re now less ahead of the other person.
• Notice that when and you’re ahead, your utility goes up by 0.5 regardless of whether you or the other person gets
a dollar.
• Under what circumstances would you allocate an extra dollar to the other person rather than yourself when you’re
ahead?
• The marginal utility of giving the dollar to the other person must be bigger than the marginal utility of giving it to yourself: .
• Under what circumstances would you allocate an extra dollar to the other person rather than yourself when you’re
behind? Doesn’t happen because the marginal utility of giving it to yourself, , is always larger than that of giving it
away, .
The Fehr-Schmidt model and the Dictator Game

• How does the Fehr-Schmidt model explain behavior in the

dictator game?
• Remember what we said on the last slide:
• When you’re behind, you always allocate extra dollars to yourself
• When you’re ahead (as you are, at the beginning of the dictator
game), you allocate money to the other person if . (If that’s not the
case, you keep everything to yourself.)
• How much to you allocate to the other person?
• Maximization problem:
The Fehr-Schmidt model and the Dictator Game

• Advantageous inequality (i.e. or ):

• Corner solution at
 If : Utility is maximized at , i.e. when Player 1 gives half their endowment to
Player 2.
 If : Corner solution at ; no sharing.
The Fehr-Schmidt model and the Ultimatum Game
• Start with Player 2: Would they ever reject a positive offer? Remember this would never
happen with “standard” preferences, because even a small amount is better than nothing.
• Example with Fehr-Schmidt preferences: Denote as and the disadvantageous and
advantageous inequality aversion parameters of Player 2, respectively.
• Utility for Player 2:

• Where does disadvantageous inequality lead utility to go below zero?

• Utility drops below 0 when Player 1 offers a share smaller than . In this case, Player 2
suffers so much from having a lower payoff than Player 1 because of disadvantageous
inequality aversion that it’s better for Player 2 to earn 0 than to earn a positive amount.
The Fehr-Schmidt model and the Ultimatum Game
• How will Player 1 behave? Suppose they know perfectly what Player 2 will do.
• If Player 1 offers : Player 1 receives , Player 2 receives
• If Player 1 offers less, both receive 0
• If : Player 1 is ahead at the beginning of the game. We said that they want to
allocate money to Player 2 iff . How much do they propose to Player 2 in this case?

• Advantageous inequality () (remember Player 1 is ahead):

• Again a corner solution at  Player 1 shares equally with Player 2 if .

• If : How much does Player 1 propose to Player 2?
• In this case, they want to keep as much as possible for themselves. So they propose the
minimum that Player 2 will accept, i.e. .
The Fehr-Schmidt model and the Trust Game
• Start at the end: How much will Player 2 send back?
• Notice the second stage of the trust game is a modified dictator game, with initial
endowments:
• Player 2:
• Player 1: .
• If Player 2 is ahead, i.e. , and : Player 2 wants will transfer enough to create equality
between the two players.
• The total pie is now ). So Player 2 sends back just enough that Player 1 has half of that, i.e.
.
• If Player 2 is ahead and : Player 2 sends nothing back. Player 1 then has .
• If Player 2 is behind: they send nothing back. Player 1 has .
• How much does Player 1 send?
• If : Player 1 will end up with . This means it’s in their interest to send everything.
• If : Player 2 sends nothing back, so Player 1 sends nothing in the first place.
Why do people behave pro-socially?
1. Pure altruism: The Fehr-Schmidt model explores one possible reason why people
behave pro-socially: they have internalized other people’s preferences. This is
“pure altruism”. But there are other possible motives:
2. (Anticipated) Reciprocity: I do good because it may come back to me in the
future.  Helping could be purely selfish.
Evidence: The average Player 1 offer in the trust game is about 50%, i.e. much higher than
in the dictator game (30%).
3. Social norms/social signaling: I want to be perceived as a good person by others.
4. Self-signaling: I feel bad about myself if I’m not nice.

Not entirely mutually exclusive (e.g. 4 could be modeled as 1, 3 could be driven by 2).

Which of these motives actually drive behavior?

Pure altruism?
• Think back of e.g. the Ukraine examples from the beginning: isn’t that
evidence of pure altruism?
• Not necessarily: people are observed while helping, so they could be doing it
for social image concerns
• This is why the dictator game was an important advance:
• Purely anonymous: You don’t know who the other player is  No scope for social
image concerns
• One-shot game: The game is only played once; you can’t hope that your generosity will
come back to you in the future
• Giving in the dictator game is quite generous! Remember, about 30% of the
pie.
• So is this evidence of pure altruism?
Re-interpreting dictator game giving
• The standard dictator game is played with positive outcomes only; Player 1
cannot “take” from Player 1
• When that is allowed, Player 1 does take a little bit on average!
Re-interpreting dictator game giving
Design:
• Baseline treatment:
• Player 1 has $10 can can allocate anywhere from $0 to $5 to Player 2
• Player 2 has $5 endowment, and gets whatever Player 1 gives them
• “Take 1” Treatment: Player 1 can allocate anywhere from –$1 to +$5
• “Take 5” Treatment: Player 1 can allocate anywhere from –$5 to +$5
Re-interpreting dictator game giving
• In the fully symmetric game, Player 1 takes on average $2.5 from Player 2. (What does
Player 1 end up with? How about Player 2? Can this be explained by Fehr-Schmidt
Preferences?)
• But: some people (10% in the symmetric game) do make positive offers.
• So when the game is fully symmetric, the evidence for pure altruism is a lot weaker.
Face-saving concerns?
Face-saving concerns?
Face-saving concerns?
• People are less likely to answer the door to a donation opportunity when they know it is coming: they want to
“avoid the ask” to not lose face
• Could be accommodated in the utility function by including a “social pressure” term which the individual can
avoid if they aren’t home
• This has been called “avoiding the ask”

• But it turns out people don’t just want to

save face in front of others — they also
want to do it in front of themselves!
Saving face *in
front of
yourself*?
Saving face *in front of yourself*?
Saving face *in front of yourself*?
• In the baseline dictator game, only 26% of participants make the “unfair” (keep $6, give other $1) choice.
• Hidden information game: Dictator has a choice whether to decide under ignorance, i.e. not knowing the
payoff to the recipient; or whether to decide with full information. Importantly, the recipient doesn’t know
whether the dictator chose to reveal. So the only motive for not revealing is “feeling bad about yourself”.
• Only between 50 and 63% choose to find out what the other person received.
• Between 63% and 81% of dictators make the unfair choice.

Interpretation: Self-image is an important motivator in dictator game giving; when people are given a chance
to preserve their self-image, they act much more selfishly.
Summary
• Basic insight from today’s lecture: people’s decisions are powerfully socially
motivated
• The simple self-interested model doesn’t fully capture people’s decisions in
social situations

 Next week: Examples of interventions with social preferences