Behavioral economics
Varian: Intermediate Microeconomics, 8e, chapter 30
Goolsbee, Levitt, Syverson: Microeconomics, chapter 17
1 / 33
The disease dilemma
You are the head of public health office. An unusual flu-like disease breaks
out. Estimates: 600 people die if you don’t take any action.
The choice is between two responses, each of which costs the same:
• Response A: You save 200 people with certainty.
• Response B: You save 600 people with 1/3 chance
and no one with a 2/3 chance.
The choice is between two responses, each of which costs the same:
• Response C: 400 people will die for certain.
• Response D: There is a 2/3 chance that 600 dies
and a 1/3 change that no one dies.
2 / 33
Dilema nemoci (pokraˇcov´an´ı)
If you have chosen responses A and C or B and D, the choice was rational.
They have the same outcomes, but different descriptions.
Tversky and Kahneman (Science, 1981) – a student sample:
• 72% choose A
• 78% choose D
This outcome should not happen in a standard economic model.
The choice should not depend on a description of a problem.
3 / 33
Behavioral economics
The assumption of rational decision-makers fails in some situations.
It is not a problem if these are random errors: standard economic model
might be still the best we have.
Many authors find systematic departures from
rationality – we need a more complex theory.
Behavioral economics branch of economics that
incorporates insights from human psychology into
models of economic behavior.
4 / 33
In this lecture we examine...
...the following systematic departures from rational behavior:
1 hyperbolic discounting and self-control problems
2 falling pray to framing
3 sunk-cost fallacy
4 overconfidence
We also discuss whether people are altruistic and
have a sense of fairness.
Finally, we have a look at how economist test
economic theory using experiments.
5 / 33
1. Hyperbolic discounting and self-control problems
You are in a Christmas market – imagine your favourite drink:
Choice A: Choose one option:
1 1 drink today
2 2 drinks tomorrow
Choice B: Choose one option:
1 1 drink in a year
2 2 drinks in a year and a day
Many people probably choose this way:
• choice A: 1 drink today
• choice B: 2 drinks in a year and a day
=⇒ The discount rate differs today and in a year from now.
6 / 33
Exponential vs. hyperbolic discounting
Standard economics:
People discount future at a constant discount factor δ < 1 =⇒
They use exponential discounting.
Utility from consumption in periods 1 through n:
u = u(c1) + δu(c2) + δ2
u(c3) + ... + δn−1
u(cn)
Behavioral economics:
People prefer immediate payoffs to later payoffs, even if the later
payoff is much greater =⇒ They use hyperbolic discounting.
A possible formula for hyperbolic discounting:
u = u(c1) + δ∗
u(c2) + δ∗
δu(c3) + ... + δ∗
δn−2
u(cn),
where the discount factor δ∗ < δ.
7 / 33
Example – purchase of a mobile phone
Two payment options:
• 10,000 CZK now
• 4 monthly payments of 3,000 CZK each (1. payment next month)
Exponential discounting: the monthly rate of 5%
Hyperbolic discounting: 1st month – 25%; every following month – 5%
Which payment option will be chosen?
Exponential discounting (NPC = net present cost):
NPC =
1
1.05
3,000+
1
1.052
3,000+
1
1.053
3,000+
1
1.054
3,000 = 10,761 CZK
NPC > 10, 000 =⇒ The consumer pays now.
Hyperbolic discounting:
NPC =
3,000
1.25
+
3,000
1.25 × 1.05
+
3,000
1.25 × 1.052
+
3,000
1.25 × 1.053
= 8,936 CZK
Consumer prefers payments (he spends money saved now on other things).
8 / 33
Time consistency of exponential discounting
January: The utility from consumption in Febuary (c2) + March (c3):
uJan
= δu(c2) + δ2
u(c3)
MRSJan
23 = −
δMU(c2)
δ2MU(c3)
= −
MU(c2)
δMU(c3)
February: The utility from consumption in February and March:
uFeb
= u(c2) + δu(c3)
MRSFeb
23 = −
MU(c2)
δMU(c3)
The same decision in Jan and Feb – the consumer is willing to exchange
the same number of c3 for one unit of c2 (the same MRS23).
The decision-making is time consistent.
9 / 33
Time consistency of hyperbolic discounting
January: The utility from consumption in Febuary (c2) + March (c3)
uJan
= δ∗
u(c2) + δ∗
δu(c3)
MRSJan
23 = −
δ∗MU(c2)
δ∗δMU(c3)
= −
MU(c2)
δMU(c3)
February: The utility from consumption in February and March:
uFeb
= u(c2) + δ∗
u(c3)
MRSFeb
23 = −
MU(c2)
δ∗MU(c3)
The consumer decides differently in Jan and Feb – willing to give up more
of c3 for 1 unit of c2 in Feb (|MRSFeb
23 | > |MRSJan
23 |, as 1
δ∗ > 1
δ ).
Difference between what he thinks he does in one month and what he
really does. The decision-making is time inconsistent.
10 / 33
How markets take advantage of our self-control problems
Firms may take advantage of our need to consume now.
Example:
mKreditka (2015 conditions):
• 1% refund from each payment till the end of 2015
• interest-free period up to 54 days
• free issue and zero fees (if transactions more than 500 CZK monthly)
• automatic 100% transfer
Why does mBank offer CC when there is no refund from debit cards?
My guess: Some people use CC for borrowing (at rates above 20%).
11 / 33
2. Falling pray to framing
Framing = choices depend on the presentation of the choice problem
=⇒ incompatible (irrational) decisions.
Example: the disease dilemma (positive vs. negative framing)
Framing biases:
1 endowment effect – possessing a good makes it more valuable =
the possessor must be paid more to give up the good than he would
have paid to buy it in the first place.
2 anchoring effect – decision is influenced by specific pieces of
information given (unrelated to the decision problem at hand).
3 mental accounting – people divide their assets into isolated groups
instead of basing their decision on their assets as a whole.
4 loss aversion – loss (relative to reference point) cannot be
compensated by a same-sized gain.
12 / 33
2.1 Endowment effect – experiment
Experiment Kahneman, Knetsch and Thaler (JPE, 1990):
A randomly selected half of the group gets a mug (value $6).
• If you have a mug – at what price are you willing to sell it?
• If you don’t have it – for what price are you willing to buy it?
What share of the mugs trades if we pair buyers with sellers?
Theory: If transaction costs are 0,
half of the mugs is sold.
But are transaction costs 0?
Control experiment – trading
money tokens: 1/2 of the tokens
traded.
Result: only roughly 15% traded
13 / 33
2.2 Anchoring effect
How many people live in Venezuela? More or less than X millions?
Tversky and Kahneman (Science, 1974) asked two questions:
1 Are more or less than Y % African countries in the UN?
(people saw a roulette wheel that chose Y as 10% or 65 %)
2 What is the percentage of African countries in the UN?
Y = anchor that influences responses – the average answer to Q2:
• 25% – people with Y = 10%.
• 45% – people with Y = 65%.
Venezuela has roughly 29 million inhabitants. X in the first question
was an anchor that probably misled you.
14 / 33
2.3 Mental accounting
You go to ˇSpal´ıˇcek to se you favourite movie for 100 CZK.
(You have 300 CZK in your pocket and 10,000 CZK in your account.)
Situation 1:
In the cinema, you find out that you have lost 100 CZK.
What do you do? Do you go on and buy the ticket anyway?
Situation 2:
You buy the ticket and go for a quick
beer. After coming back to the cinema
you found out that you lost the ticket.
What do you do? Do you buy the
ticket again?
15 / 33
Mental accounting (cont’d)
You should make the same choice in both situations (the same BL).
People more often buy the ticket after losing money. Why?
According to behavioral theory, people split money into “mental
accounts”. It is difficult to transfer money among them.
Explanation: mental account with “cinema” with 150 CZK per month
• losing the ticket =⇒ balance in the account “cinema” = 50 CZK
=⇒ you don’t go to see the movie
• loss of money is debited to a different account (e.g. savings = 9,900)
=⇒ the account
”
cinema“ = 150 CZK =⇒ go on and buy the ticket
Consequence: People may buy suboptimal bundles.
Advantage: Mental accounting saves the decision time.
16 / 33
Mental accounting (cont’d)
Other examples (see also Thaler, Marketing Science, 1985):
• Mr X finds 100 CZK on a sidewalk. → He spends it.
Other day Mr X gets a tip of 100 CZK. → He saves it.
• Mr K looks for a discounted bottled beer, but at the same time he is
willing to buy any number of beers in a pub for prides 3-5times higher.
• Mr and Mrs L saved 500,000 CZK for a summer cottage (interest rate
= 2%) and buy car for 300,000 CZK on leasing (IR = 10%).
• Mrs. N likes a Chanel dress, but does not buy it – too expensive for
20,000. When she gets the dress as a birthday present from Mr. N,
she is very happy, even though they have shared money with Mr. N.
17 / 33
2.4 Loss aversion
Example – toss of a coin:
• head – you get 140 CZK
• tail – you get 100 CZK
Who would accept this offer?
Most people do not accept the offer, even though the EV = 20 CZK.
The reason for this extreme risk aversion might be loss aversion
= loss has a larger impact on utility than a same-sized gain.
Loss or gain are relative to a reference point.
Example: the reference point is the current wealth.
18 / 33
How markets take advantage of our framing problems
Endowment effect
Money-back guarantee raises the probability that I buy a product to try it
out. Then the endowment effect increases the probability that I keep it.
Anchoring effect
• 50% of originally exaggerated price (= anchor) seems a bargain
• first show an expensive product (= anchor) and then cheaper options
• quantity limit (
”
limit of 5 units per person“) increases the demand
Mental accounting
• The question of car salespeople: “How much do you want to pay for
a car?” Once you put a number into a mental account, a skilled seller
will find a way to withdraw it for you.
• high tourist prices (turists may have an account
”
holiday spendings“)
19 / 33
3. Sunk-cost fallacy
Sunk costs = the money is already spent and it cannot be recovered.
E.g. fixed costs are sunk – cannot be recovered after a shut down.
Sunk costs should not influence your decision-making. If they do,
you have succumbed to the sunk-cost fallacy.
Examples:
A season subscription to a theater series:
Arkes and Blumer (1985) – people pay randomly one of prices: $15, 13, 8.
Whether they go to the theater is independent from the price (sunk costs).
1st half of the season: people paying $15 went 25% more than others.
Project Concorde:
In the course of the project it was clear that it should be stopped. They
finished it (20 airplanes), because there was “too much invested to quit”.
20 / 33
APPLICATION: The recession and the housing market
Genesove and Mayer (QJE, 2001) study the Boston housing market.
Prices of ownership homes went down by 40% in the recession in the
1990s, but people were not willing to reduce prices. The average offer price
was 35% above the expected value of homes.
Explanation:
Admitting that their homes lost 40% of the value would cause a large loss
of utility (loss aversion, sunk-cost fallacy). Rather not selling and waiting
for the prices to go up.
21 / 33
The effect of loss aversion and the sunk-cost fallacy
Examples:
• Housing market – rational people sell their houses for a higher price,
because people with loss aversion do not want to sell at current prices.
• A common situation in poker – a promising hand (AK); calling first
(500 CZK); a minimum chance to win after the flop (8, 8, J); I lose
money to rational players if I do not fold and continue betting,
because I don’t want to accept the loss of 500 CZK (sunk costs).
22 / 33
4. Overconfidence
Overconfidence – a belief that skill and judgment are better than they
truly are, or that better outcomes are more likely to happen than their true
probability.
Most people think that they are better than average:
• 93% American university students are better-than-average drivers
• 77% people in internet dating service are more attractive than average
• 32% programmers are better than 19 out of 20 colleagues
• 94% university professors think that their intelligence is above average
Economic context:
• financial markets: one side of the
transaction thinks they are better
informed than the other
• management: self-confident managers
make larger and riskier investments
23 / 33
How markets take advantage of our overconfidence
Example: the price of membership offered by gyms looks attractive
if people visit the gym regularly.
DellaVigna and Malmendier (AER, 2006) – customers with a monthly
membership pay 70% more than if they paid for individual visits. This adds
up to a $600 overpayment in the course of their membership.
A test question:
I go to the gym only because I bought the membership. Is it rational?
No, membership is a sunk cost – it should not influence my decision.
24 / 33
Altruism and fairness
Dictator game: player A gets 100 CZK and player B nothing
1 Player A gives player B an amount X ∈ 0, 100 CZK.
When player A maximizes his payoff, he should choose X = 0.
Experimental results: people often choose X > 0.
Ultimatum game: player A gets 100 CZK and player B nothing:
1 Player A offers player B an amount X ∈ 0, 100 CZK.
2 Player B can accept or reject the offer. Accepts: player B gets X and
player A gets 100 − X. Refuses: both players get 0 CZK.
Backward induction. Player B accepts X > 0. Player A offers X → 0.
Experimental results: Roughly a half of players B refuse X < 30.
Players A offer X = 45 on average (still refused by 16%).
25 / 33
Altruism and fairness (cont’d)
These results suggest that people are altruistic/care about fairness.
A controversial question: Do experiments capture real preferences?
• People might feel ashamed before the experimenters for behaving
selfishly.
• Bardsley (Exp Econ, 2008) shows that the result of dictator game
might be driven by the framing (game about giving).
Player A can give and take the money of player B (X < 0 possible).
Most people offering X > 0 in the standard game, now choose X < 0.
26 / 33
Real altruistic behavior
But how do we explain real behavior such as charity or volunteering?
Real behavior: one of the following options should be true:
1 People do not maximize utility.
2 Utility influenced by non-monetary factors (remorse, resentment).
3 Utility function has an altruistic element (welfare of others).
Some real behavior is certainly truly altruistic, but:
• Meer and Rosen (AEJ, 2009) find out that alumni with kids donate to
universities more probably than their colleagues without them. The
donations increase with the age of the child, and fall to the level of
alumni without children if the children applied but were rejected.
• DellaVigna, List and Malmendier (QJE, 2012) sent out volunteers to
knock on doors and solicit charitable gifts. They collected less in
homes where fliers with the date of the fund-raising were distributed.
27 / 33
Is the model based on rational decision-makers useful?
Yes, it is – three arguments:
1 Behavioral economics shows that some people, under certain
circumstances, act irrationally. The neoclassical model does a good
job describing how the economic world works most of the time – an
important advance when we describe human behavior.
2 Market punishes irrationality – adjust behavior or leave the market
Example: investors (poker players) suffer less from sunk-cost fallacy.
Experts might help to take complex decisions (e.g. financial advisers).
3 In some situations, simple adjustments to the basic economic model
often provide perfectly viable alternative explanations.
Example: smoking – two explanations:
• people are irrational (self-control problem)
• Becker and Murphy (JPE, 1988):
”
A Theory of Rational Addiction“
can explain why addiction rates respond to prices
28 / 33
Testing theories on data
Behavioral critique → more emphasis on testing economic models
How to test economic theory?
1 We build a theory: e.g. consumer theory.
2 We derive a prediction (= causal relationship): e.g. p ↓→ qd ↑.
3 Test on data: investigate whether p ↓→ qd ↑ (ceteris paribus).
Problem: A lot of effects in reality – ceteris paribus rarely holds.
A possible solution:
Econometrics = statistical and analytical techniques in economics
E.g. instruments = using supply shocks for demand estimation.
29 / 33
Experimental economics
An alternative solution: testing the predictions directly in experiments.
Laboratory experiment
• randomly selected groups of subjects → versions of the same activity
– e.g. the disease dilemma (1 version: A or B; 2. version: C or D)
• typically financial motivation: a show-up fee and a result-based reward
Advantage: economists can control all aspects of the test
Disadvantages: the results may apply in reality, because in the lab:
• a different type of scrutiny (people watched by the experimenters)
• a different situation (small rewards, strange and tedious tasks)
• the tasks are new (real decision-makers are typically experienced)
• an experimenter cannot control personal habits and norms
(findings that culture influences experimental results)
30 / 33
Experiments in the real world
Laboratory experiments may not study the “real world” – solutions:
1 field experiments
2 natural experiments
Field experiments
• experimental subjects are randomly divided into the experimental
and control groups – randomized control trial (RCT)
• naturally occurring environments – ideally people do not know that
they are participating in an experiment
Example:
The experiment on charitable giving (DellaVigna et al., AER, 2012)
Many firms conduct field experiments – e.g. marketing (A/B testing),
testing pricing strategies or recruitment procedures, ...
31 / 33
Experiments in the real world (con’t)
Natural experiments
Real data have the property that allows us to eliminate all other factors
(= we can observe the causal relationship ceteris paribus).
Example:
The US soldiers in the Vietnam era were selected by a draft lottery –
similar to RCT = random division of soldiers into the experimental
group (drafted) and the control group (not drafted).
Angrist (AER, 1990): veteran status reduces the income by 15%.
Problems of natural experiments:
• they are rare
• sometimes difficult to generalize from a specific situation
(do Angrist’s conclusions apply to other wars veterans?)
32 / 33
What should you know?
• Behavioral economics: people are predictably
irrational in some situations (overconfidence,
self-control, sunk-cost fallacy, framing)
• Rational people may take advantage of it.
• Behavioral critique =⇒ more emphasis on how
we test theories – the ideal is the RCT.
• The model with rational agents works most of the
time, but sometines fails. It is useful to know when
people typically behave irrationally.
• Hope that you liked micro and will use it to make
the right decisions in the future.
33 / 33