BEHAVIORAL ECONOMICS
REVIEW
Ernesto Reuben
System 1 (Intuition)
 Fast
 Parallel
 Automatic
 Effortless
 Associative
 Slow-learning
 Emotional
System 2 (Reasoning)
 Slow
 Serial
 Controlled
 Effortful
 Rule-governed (deductive)
 Flexible
 Neutral
TWO COGNITIVE SYSTEMS
2
Accessibility
 System 1 judgements are spontaneous
but the accessibility of
specific properties/interpretations
depends on context and
saliency
Reference-dependence
 System 1 judgements depend on the
contrast with their context
IMPORTANT CONCEPTS TO UNDERSTAND SYSTEM 1
3
HOW DO WE FORM BELIEFS?
4
New information
Cultural transmission
Own experience
Bears are
dangerous!
Prior belief
Prior belief
Stereotypes
 Accessibility
 Implicit and explicit
Updating process
Implicit
associations
Affective response ‒
Posterior belief
Explicit belief updating
‒ Bayes’ rule
‒ Conservatism
‒ Base-rate neglect
 Stereotypes are associations between
individuals/objects of a group and a
set of characteristics.
 Implicit associations are activated
automatically and might influence
people without their awareness
 Measures of implicit associations like the IAT are not very precise but they are
better than explicit measures for sensitive topics (e.g., to capture racism)
 Implicit associations are more likely to measure cultural rather than individual
attitudes
STEREOTYPES
5
 Stereotypes can become self-fulfilling through various channels
 Costly performance-enhancing investments
 When anticipating discrimination due to a
negative stereotype about the performance
of one’s group reduces individual incentives
to invest in improving performance,
confirming the stereotype
SELF-FULLING STEREOTYPES
6
BELIEF UPDATING
7
posterior belief = new evidence × prior belief
BELIEF UPDATING
8
Bayes’ rule
P(A|B) =
P(B|A)
P(B|A)P(A) + P(B|¬A) P(¬A)
P(A)
Base-rate neglect / representativeness
 Too much weight on new information
 New information is consistent with
important values/beliefs
 New information is salient and/or strong
(even if inaccurate)
 New information produces affect
Conservativism
 Too little weight on new information
 New information is inconsistent with
important values/beliefs
 New information is not salient and/or
weak (even if accurate)
 New information lacks affect
BELIEF UPDATING
9
posterior belief =
new evidence× prior belief
posterior belief =
new evidence × prior belief
Overestimation
Overprecision
OVERCONFIDENCE
10
Overplacement
Actual
distribution
Expected
distribution
Actual
performance in
the population
Your
performance
Expected
performance in
the population
Your
performance
Your
expected
performance
COMPETITIVENESS
11
[adjective] a desire and/or ability to perform in competitions
Time difference when competing  faster
–0.050
+0.058
+0.201
+0.135
vs.
vs.
vs.
vs.
vs.
vs.
+0.016
+0.016
40%
66%
60%
34%
0%
10%
20%
30%
40%
50%
60%
70%
Men Women
Percent
Piece-rate Tournament
DECISIONS UNDER RISK AND UNCERTAINTY
12
Expected utility
theory
 Decisions are made
based on utility
levels and objective
probabilities
 Incompatible with
observed behavior
 e.g., small-stake
risk aversion
U
$
U($)
The degree of risk
aversion depends on
the concavity of the
utility function
DECISIONS UNDER RISK AND UNCERTAINTY
13
Prospect theory
 Descriptive model
of risky choice
characterized by a
value function
and a probability
weighing function
V
$
The value function
Risk aversion for gains
Risk loving for losses
Steeper slope for
losses than for gains
Explains:
 Loss aversion
 Endowment effect
DECISIONS UNDER RISK AND UNCERTAINTY
14
Prospect theory
 Descriptive model
of risky choice
characterized by a
value function
and a probability
weighing function
1
1
0
½
½
Actual probability: p
Weighted
probability:
w(p)
The probability
weighing function
 Underestimation of
high probabilities
 Overestimation of
low probabilities
Reference-dependent risk attitudes
 Reference point for gains/loses corresponds to an individual’s expectations
DECISIONS UNDER RISK AND UNCERTAINTY
15
 Mimics prospect theory for surprising events
 A silver medal when expecting a gold feels like a loss and a bronze medal when expecting no medal feels like a gain
 But makes different predictions for expected events
 Getting a gold medal when expecting a gold is neither a gain nor a loss
t
D
1
0
Exponential discounting
 The discount rate is approximately
1 – δ for all time periods
 Dynamically consistent
 Preferences do not change over time
unless new information arrives
INTERTEMPORAL CHOICE
16
𝑈𝑈0 = 𝑢𝑢0 + 𝛿𝛿𝑢𝑢1 + 𝛿𝛿2𝑢𝑢2 + ⋯ + 𝛿𝛿𝑇𝑇𝑢𝑢𝑇𝑇
D(t)=δt
Present-biased preferences: quasi-hyperbolic discounting
 Weighs more the present if β < 1  impatient for choices that involve the
present but patient for choices
that involve only the future
 The discount rate is approximately
1 – βδ if t = 0 and 1 – δ if t > 0
 Implies
 Non-stationarity
 Time inconsistency
 Demand for commitment
t
D
0
1
INTERTEMPORAL CHOICE
17
𝑈𝑈0 = 𝑢𝑢0 + 𝛽𝛽𝛽𝛽𝑢𝑢1 + 𝛽𝛽𝛿𝛿2𝑢𝑢2 + ⋯ + 𝛽𝛽𝛿𝛿𝑇𝑇𝑢𝑢𝑇𝑇
β = ½
β = 1
β = ¾
 If benefits are in the future but costs are in the present, present-biased
individuals want to postpone paying the costs
Naïve individuals
 Are present-biased β < 1 and incorrectly believe that
they will not be present-biased in the future βe = 1
 Tend to always procrastinate
 Sophisticated individuals
 Are present-biased β < 1 and correctly anticipate that they will be present-
biased in the future βe = β
 Might not procrastinate as much depending on the distribution of costs and
benefits over time
PROCRASTINATION AND PRESENT-BIASED PREFERENCES
18
Reduce immediate costs
 Buy a gym membership (reduce daily cost of going)
 Transportation subsidies
Increase immediate rewards
 Exercise with a friend
 Pay people to exercise
Commitment  reduce your future choices
 Remove ice-cream from your house
 Impose yourself binding deadlines
WHAT CAN YOU DO TO SELF-REGULATE?
Altruism
∂Ui/∂π¬i > 0
Derive utility from the payoff
of others
Envy / Spite
∂Ui/∂π¬i < 0
Derive utility from a higher
payoff compared to others
SOCIAL PREFERENCES
20
Outcome-based models: Ui = Ui(πi, π¬i)
Inequity aversion
∂Ui/∂π¬i > 0 if πi > π¬i
∂Ui/∂π¬i < 0 if πi < π¬i
Dislike differences between
my income and the income
of others
Quasi-maximin preferences
∂Ui/∂π¬i > 0
∂Ui/∂π¬i >> 0 if π¬i = min{π1,…,πN}
People care about the sum of all
payoffs (efficiency) and the payoff
of the poorest
Moral judgements are influenced
by System 1
 They are often immediate and
susceptible to emotions
Five foundations of moral values
 Care/Harm
 Fairness/Cheating
 Loyalty/Betrayal
 Authority/Subversion
 Purity/Degradation
MORAL VALUES
21
Social norms
 A preference to comply with the expectations of
others and to punish intentional deviations from
the norm
 Conditional on expecting that
 Sufficiently many others comply with the social
norm  empirical expectations
 Sufficiently many others think you should comply
with the norm  normative expectations
 Social norms can be misperceived
 pluralistic ignorance
SOCIAL NORMS
22
Nudging [to push someone gently]
 Policies that help you improve your behavior without
(greatly) limiting your choices  libertarian paternalism
 Good tool to address problems of internalities (conflict
between systems 1 and 2)
 Decisions with delayed consequences, that are complex or
overwhelming, and for which there is little scope to learn
 Common tools for nudging include:
 Changing the choice and salience of alternatives, deciding
the timing of choices and deadlines, decision aids, default
options, and using information about social norms
NUDGES
23
Automatic enrollment
 Optimal when:
 Preferences are similar  a default is
attractive to many
 Little procrastination  it is not too costly to
change for those with different preferences
Active choice
 Optimal when:
 Preferences are different  a default is
attractive to few
 Lots of procrastination  it is costly to change for
those with different preferences
GOOD NUDGING
24
Problem
Identify specific cause
Welfare analysis
Theoretical benchmarks
 Use a behavioral theory and a well-defined
welfare criterion to evaluate the optimal policy
 What if the model is wrong? How do we know what
is considered welfare-relevant?
Measure social welfare
 Use direct or indirect measures of social welfare
(e.g., subjective well-being or willingness to pay)
 Issues of reliability and comparability of subjective
measures. Incentivized measures assume individuals
are sophisticated enough.
APPROACHES TO SOCIAL WELFARE
25
TWO COGNITIVE SYSTEMS
26
I value for feedback, please fill out the
course evaluations
YOUR EVALUATIONS
27

Behavioral Economics Slides explanation.pdf

  • 1.
  • 2.
    System 1 (Intuition) Fast  Parallel  Automatic  Effortless  Associative  Slow-learning  Emotional System 2 (Reasoning)  Slow  Serial  Controlled  Effortful  Rule-governed (deductive)  Flexible  Neutral TWO COGNITIVE SYSTEMS 2
  • 3.
    Accessibility  System 1judgements are spontaneous but the accessibility of specific properties/interpretations depends on context and saliency Reference-dependence  System 1 judgements depend on the contrast with their context IMPORTANT CONCEPTS TO UNDERSTAND SYSTEM 1 3
  • 4.
    HOW DO WEFORM BELIEFS? 4 New information Cultural transmission Own experience Bears are dangerous! Prior belief Prior belief Stereotypes  Accessibility  Implicit and explicit Updating process Implicit associations Affective response ‒ Posterior belief Explicit belief updating ‒ Bayes’ rule ‒ Conservatism ‒ Base-rate neglect
  • 5.
     Stereotypes areassociations between individuals/objects of a group and a set of characteristics.  Implicit associations are activated automatically and might influence people without their awareness  Measures of implicit associations like the IAT are not very precise but they are better than explicit measures for sensitive topics (e.g., to capture racism)  Implicit associations are more likely to measure cultural rather than individual attitudes STEREOTYPES 5
  • 6.
     Stereotypes canbecome self-fulfilling through various channels  Costly performance-enhancing investments  When anticipating discrimination due to a negative stereotype about the performance of one’s group reduces individual incentives to invest in improving performance, confirming the stereotype SELF-FULLING STEREOTYPES 6
  • 7.
    BELIEF UPDATING 7 posterior belief= new evidence × prior belief
  • 8.
    BELIEF UPDATING 8 Bayes’ rule P(A|B)= P(B|A) P(B|A)P(A) + P(B|¬A) P(¬A) P(A)
  • 9.
    Base-rate neglect /representativeness  Too much weight on new information  New information is consistent with important values/beliefs  New information is salient and/or strong (even if inaccurate)  New information produces affect Conservativism  Too little weight on new information  New information is inconsistent with important values/beliefs  New information is not salient and/or weak (even if accurate)  New information lacks affect BELIEF UPDATING 9 posterior belief = new evidence× prior belief posterior belief = new evidence × prior belief
  • 10.
  • 11.
    COMPETITIVENESS 11 [adjective] a desireand/or ability to perform in competitions Time difference when competing  faster –0.050 +0.058 +0.201 +0.135 vs. vs. vs. vs. vs. vs. +0.016 +0.016 40% 66% 60% 34% 0% 10% 20% 30% 40% 50% 60% 70% Men Women Percent Piece-rate Tournament
  • 12.
    DECISIONS UNDER RISKAND UNCERTAINTY 12 Expected utility theory  Decisions are made based on utility levels and objective probabilities  Incompatible with observed behavior  e.g., small-stake risk aversion U $ U($) The degree of risk aversion depends on the concavity of the utility function
  • 13.
    DECISIONS UNDER RISKAND UNCERTAINTY 13 Prospect theory  Descriptive model of risky choice characterized by a value function and a probability weighing function V $ The value function Risk aversion for gains Risk loving for losses Steeper slope for losses than for gains Explains:  Loss aversion  Endowment effect
  • 14.
    DECISIONS UNDER RISKAND UNCERTAINTY 14 Prospect theory  Descriptive model of risky choice characterized by a value function and a probability weighing function 1 1 0 ½ ½ Actual probability: p Weighted probability: w(p) The probability weighing function  Underestimation of high probabilities  Overestimation of low probabilities
  • 15.
    Reference-dependent risk attitudes Reference point for gains/loses corresponds to an individual’s expectations DECISIONS UNDER RISK AND UNCERTAINTY 15  Mimics prospect theory for surprising events  A silver medal when expecting a gold feels like a loss and a bronze medal when expecting no medal feels like a gain  But makes different predictions for expected events  Getting a gold medal when expecting a gold is neither a gain nor a loss
  • 16.
    t D 1 0 Exponential discounting  Thediscount rate is approximately 1 – δ for all time periods  Dynamically consistent  Preferences do not change over time unless new information arrives INTERTEMPORAL CHOICE 16 𝑈𝑈0 = 𝑢𝑢0 + 𝛿𝛿𝑢𝑢1 + 𝛿𝛿2𝑢𝑢2 + ⋯ + 𝛿𝛿𝑇𝑇𝑢𝑢𝑇𝑇 D(t)=δt
  • 17.
    Present-biased preferences: quasi-hyperbolicdiscounting  Weighs more the present if β < 1  impatient for choices that involve the present but patient for choices that involve only the future  The discount rate is approximately 1 – βδ if t = 0 and 1 – δ if t > 0  Implies  Non-stationarity  Time inconsistency  Demand for commitment t D 0 1 INTERTEMPORAL CHOICE 17 𝑈𝑈0 = 𝑢𝑢0 + 𝛽𝛽𝛽𝛽𝑢𝑢1 + 𝛽𝛽𝛿𝛿2𝑢𝑢2 + ⋯ + 𝛽𝛽𝛿𝛿𝑇𝑇𝑢𝑢𝑇𝑇 β = ½ β = 1 β = ¾
  • 18.
     If benefitsare in the future but costs are in the present, present-biased individuals want to postpone paying the costs Naïve individuals  Are present-biased β < 1 and incorrectly believe that they will not be present-biased in the future βe = 1  Tend to always procrastinate  Sophisticated individuals  Are present-biased β < 1 and correctly anticipate that they will be present- biased in the future βe = β  Might not procrastinate as much depending on the distribution of costs and benefits over time PROCRASTINATION AND PRESENT-BIASED PREFERENCES 18
  • 19.
    Reduce immediate costs Buy a gym membership (reduce daily cost of going)  Transportation subsidies Increase immediate rewards  Exercise with a friend  Pay people to exercise Commitment  reduce your future choices  Remove ice-cream from your house  Impose yourself binding deadlines WHAT CAN YOU DO TO SELF-REGULATE?
  • 20.
    Altruism ∂Ui/∂π¬i > 0 Deriveutility from the payoff of others Envy / Spite ∂Ui/∂π¬i < 0 Derive utility from a higher payoff compared to others SOCIAL PREFERENCES 20 Outcome-based models: Ui = Ui(πi, π¬i) Inequity aversion ∂Ui/∂π¬i > 0 if πi > π¬i ∂Ui/∂π¬i < 0 if πi < π¬i Dislike differences between my income and the income of others Quasi-maximin preferences ∂Ui/∂π¬i > 0 ∂Ui/∂π¬i >> 0 if π¬i = min{π1,…,πN} People care about the sum of all payoffs (efficiency) and the payoff of the poorest
  • 21.
    Moral judgements areinfluenced by System 1  They are often immediate and susceptible to emotions Five foundations of moral values  Care/Harm  Fairness/Cheating  Loyalty/Betrayal  Authority/Subversion  Purity/Degradation MORAL VALUES 21
  • 22.
    Social norms  Apreference to comply with the expectations of others and to punish intentional deviations from the norm  Conditional on expecting that  Sufficiently many others comply with the social norm  empirical expectations  Sufficiently many others think you should comply with the norm  normative expectations  Social norms can be misperceived  pluralistic ignorance SOCIAL NORMS 22
  • 23.
    Nudging [to pushsomeone gently]  Policies that help you improve your behavior without (greatly) limiting your choices  libertarian paternalism  Good tool to address problems of internalities (conflict between systems 1 and 2)  Decisions with delayed consequences, that are complex or overwhelming, and for which there is little scope to learn  Common tools for nudging include:  Changing the choice and salience of alternatives, deciding the timing of choices and deadlines, decision aids, default options, and using information about social norms NUDGES 23
  • 24.
    Automatic enrollment  Optimalwhen:  Preferences are similar  a default is attractive to many  Little procrastination  it is not too costly to change for those with different preferences Active choice  Optimal when:  Preferences are different  a default is attractive to few  Lots of procrastination  it is costly to change for those with different preferences GOOD NUDGING 24 Problem Identify specific cause Welfare analysis
  • 25.
    Theoretical benchmarks  Usea behavioral theory and a well-defined welfare criterion to evaluate the optimal policy  What if the model is wrong? How do we know what is considered welfare-relevant? Measure social welfare  Use direct or indirect measures of social welfare (e.g., subjective well-being or willingness to pay)  Issues of reliability and comparability of subjective measures. Incentivized measures assume individuals are sophisticated enough. APPROACHES TO SOCIAL WELFARE 25
  • 26.
  • 27.
    I value forfeedback, please fill out the course evaluations YOUR EVALUATIONS 27