S S ocial T ocial Tra ransm nsmission Bias in ission Bias in - - PowerPoint PPT Presentation

s s ocial t ocial tra ransm nsmission bias in ission bias
SMART_READER_LITE
LIVE PREVIEW

S S ocial T ocial Tra ransm nsmission Bias in ission Bias in - - PowerPoint PPT Presentation

S S ocial T ocial Tra ransm nsmission Bias in ission Bias in Economics and Fina Eco nomics and Finance nce David Hirshleifer Slides and transcript will be available at: https:/ /ssrn.com/ abstract=3513201 Presidential Address American


slide-1
SLIDE 1

S

  • cial T

S

  • cial Tra

ransm nsmission Bias in ission Bias in Eco Economics and Fina nomics and Finance nce

David Hirshleifer

Slides and transcript will be available at:

https:/ /ssrn.com/ abstract=3513201

Presidential Address American Finance Association January 4, 2020

slide-2
SLIDE 2

Social economics and finance

  • M issing chapter in our understanding of finance:
  • The social processes that shape economic thinking, behavior
  • Social economics and finance:
  • The study of how social interaction affects economic outcomes
  • Recognizes that people observe each other, “talk” to each other
  • A key intellectual building block:

Social transmission bias

slide-3
SLIDE 3

Some recent intellectual revolutions

  • Information economics
  • Recognized that some people know things that others do not
  • Behavioral economics, finance
  • Recognized that people make systematic mistakes
slide-4
SLIDE 4

Do we already know?

  • Scholars “knew” these facts before each revolution
  • But considered informally, sporadically
  • Not systematically, explicitly, routinely incorporated in models,

tests

  • Same now for social economics and finance
slide-5
SLIDE 5

Behavioral finance: Path from assumptions to conclusions often very direct

  • Beliefs
  • Investors trade too aggressively?

Overconfident

  • Expectations rise after price run-ups?

Overextrapolate

slide-6
SLIDE 6

Preferences

  • Investors:
  • Buy lottery stocks?
  • Sell winners more than losers?
  • Save too little?

Taste for: skewness, realizing gains not losses, immediate

consumption

  • Y

es, but…

slide-7
SLIDE 7

Attr ttraction action to a behavior A pre prefer erence ence for it

  • Moths attracted to flame
  • M oths not flame-loving
  • Navigations systems designed by natural selection to work with

distant light sources

  • Nearby light sources fool navigation systems
slide-8
SLIDE 8
slide-9
SLIDE 9

Social emergence

  • Purely individual-level navigation errors (moths)
  • One kind of indirect effect
  • Another: social emergence
  • Aggregate outcomes not just sum of individual propensities
slide-10
SLIDE 10

Example of a socially emergent effect

  • Death spirals
  • Rotative instinct?
  • A heuristic or bias for circular motion?
  • Vs. instincts for random search, following others
  • Akin to information cascades
  • Banerjee (1992), Bikhchandani, Hirshleifer & Welch (1992)
  • Aggregate outcome looks nothing like individual propensities
slide-11
SLIDE 11
slide-12
SLIDE 12
slide-13
SLIDE 13
slide-14
SLIDE 14

Implication of emergence

  • Unwarranted:
  • Observed behavior Direct psychological bias “for” that behavior
  • In Finance field, emergent social effects usually neglected
  • Transmission bias missing from standard toolkit
slide-15
SLIDE 15

Goal of social economics and finance

  • Social economics & finance
  • Build on standard ingredients
  • Preferences, optimization, psychological bias,

equilibrium

  • As in behavioral economics:
  • Well-motivated assumptions
  • Psychological evidence
  • Evolutionary plausibility
  • Capture systematically, tractably:
  • Socially emergent, as well as direct, effects
slide-16
SLIDE 16

Plan for rest of this talk

  • Some milestones of the social economics and finance

revolution

  • What is social transmission bias?
  • Five fables of social transmission bias, economics and

finance

  • Does transmission bias offer novel messages, wide-ranging

applications?

  • Emergent themes and closing words
slide-17
SLIDE 17

S

  • me miles

S

  • me mileston

tones of the social es of the social economics & fi economics & finance rev nance revolution

  • lution
slide-18
SLIDE 18

S

  • me milestone

S

  • me milestones of

s of the s the social e

  • cial economi

conomics & cs & finance revolution finance revolution

  • M odels of
  • Biased social influence in networks
  • Ellison & Fudenberg (1993),

DeM arzo, Vayanos & Zwiebel (2001, 2003), Eyster & Rabin (2010), Golub & Jackson (2010), Bohren (2016)

  • Surveys:
  • Jackson (2008), Golub & Sadler

(2016)

  • Cultural transmission of ethnic,

religious, & cooperative traits

  • Bisin & Verdier (2000), Tabellini

(2008)

  • Payoff interactions/ games
  • Empirical literatures on:
  • Narratives, folk models & finance
  • Shiller, Konya & Tsutsui (1996), Shiller (2000, 2017, 2019)
  • Culture, ideology, & economic outcomes
  • E.g., Grinblatt & Keloharju (2001), Barro & M cCleary (2003),

Guiso, Sapienza & Zingales (2003, 2004), Graham et al. (2019)

  • Contagion of economic/ financial behaviors
  • Individuals
  • E.g., Glaeser, Sacerdote & Scheinkman (1996), Kelly &

Ograda (2000), Brock & Durlauf (2001), Duflo & Saez (2002, 2003), Hong, Kubik, & Stein (2004), Bailey et al. (2018)

  • Firms
  • E.g., Bizjak, Lemmon & Whitby (2009), Chiu, Teoh &

Tian (2013), Fracassi (2017)

slide-19
SLIDE 19

What What is s is social trans

  • cial transmission

mission bias? bias?

slide-20
SLIDE 20

Social transmission

  • Signals, ideas pass from person to person
  • Social transmission bias:
  • Signals, ideas, systematically modified in transfer from a sender, or
  • bservation target, to a receiver, or observer
  • Derives from both sender, receiver incentives, psychological biases
  • Underexplored building block
slide-21
SLIDE 21

Social tr Social transm ansmission ission bias as signal dist bias as signal distortion

  • rtion
  • 1. Signal distortion
  • Shifts in sign, intensity of what is transmitted
  • Example:
  • Owner of a stock “talks up” the firm
  • Listener fails to discount
slide-22
SLIDE 22

S

  • cial tr

S

  • cial transm

ansmission ission bias as sel bias as selectio ection bias n bias

  • 2. Selection bias
  • Bias in whether something is transmitted
  • Example: Self-enhancing transmission bias
  • Investors discuss their trades with high returns
  • Silent about their low returns
  • Escobar & Pedraza (2019)
  • Listeners fail to adjust
slide-23
SLIDE 23

Five f Five fables of social transm ables of social transmission ission bias i bias in economics and f n economics and finance inance

slide-24
SLIDE 24

Fable 1: Bandwidth constraints and simplistic thinking

slide-25
SLIDE 25

Bandwidth constraints and simplistic thinking

Hirshleifer & Tamuz (in progress)

  • Suppose loss of nuance as ideas

communicated

  • TV “Sound bites”
  • Bandwidth constraints
  • Twitter character limits
  • Time, cognitive constraints
slide-26
SLIDE 26

Failure to adjust

  • Suppose receivers do not adjust for loss of nuance
  • Consistent with standard limited attention effects
slide-27
SLIDE 27

Outcome

Then:

  • Infer senders have simple or extreme

belief

  • Adopt actually-simplistic beliefs
  • Sequential
  • Iterated loss of nuance
  • Society Extreme simplistic thinking
  • Worse than judgements made in

isolation

slide-28
SLIDE 28

Fable 2: Self-enhancing transmission bias

slide-29
SLIDE 29

Self-enhancing transmission bias

Han, Hirshleifer & Walden (2019a)

  • 2 Strategies
  • A – Active
  • Higher variance, or higher skewness
  • P – Passive
  • Investors of type A or P randomly selected to meet
  • Sender may report profit to Receiver
  • High more than low returns
slide-30
SLIDE 30

Receivers

  • Standard behavioral biases
  • Don’t adjust for selection bias
  • Think past performance predicts future performance
slide-31
SLIDE 31

Result

  • Upward selection bias in return reports
  • Stronger effect for high-variance strategy
  • High-variance, underperforming A can spread through

population Nondiversification, price anomalies…

  • Empirical support for this mechanism
  • Escobar & Pedraza (2019)
slide-32
SLIDE 32

A variation

  • High salience of extremes:
  • Positive skewnessstrategies spread
slide-33
SLIDE 33

Lessons

Attraction to variance, skewness:

  • Investors don’t like variance, skewness
  • Don’t have belief
  • “High variance, skewness Good opportunity”
  • M ay be unaware of variance, skewness
  • Attraction socially emergent

Distinctive empirical implications:

  • Personality traits (e.g., self-enhancing transmission bias)
  • Social network position
  • Overall network connectivity
slide-34
SLIDE 34

Fable 3: Visibility bias and

  • verconsumption
slide-35
SLIDE 35

Visibility bias and overconsumption

Han, Hirshleifer & Walden (2019b)

  • Visibility bias
  • Engaging in a consumption activity often more visible than refraining
slide-36
SLIDE 36

Basic idea and assumptions

  • Observers don’t adjust for this
  • (a standard behavioral bias)

Infer others consuming heavily

  • x = people’s need-for-saving
  • E.g., probability of a personal wealth disaster
  • Same, for all
  • Uncertainty about x
  • Diverse private information
slide-37
SLIDE 37

Outcome

Visibility bias, naiveté

People mistakenly “ learn” from “ high” consumption that x low Undersaving

  • Self-reinforcing effect
slide-38
SLIDE 38

Nonobvious consequences

  • Y
  • ung overconsume more than old
  • Wealth dispersion (information asymmetry) weakens effects
  • Opposite of wealth signaling models (Veblen effects)
slide-39
SLIDE 39

Moral of the story

  • No direct bias for overconsumption
  • Vs. behavioral finance
  • Present-biased preferences
  • (hyperbolic discounting)
slide-40
SLIDE 40

Why should we care?

  • Different empirical, policy implications
  • E.g.:
  • Disclosure helps!
  • Target interventions by position in social network

For good policy, empirical testing vital

  • Empirical support for both mechanisms
  • See, e.g., D‘Acunto, Rossi & Weber (2019)
slide-41
SLIDE 41

Fable 4: (Main model) Biased information percolation, action booms, and price bubbles

slide-42
SLIDE 42

Biased information percolation, action booms, and price bubbles

  • “Beyond all reason'' flavor of booms, bubbles
  • Religious awakenings, Bitcoin…
slide-43
SLIDE 43

Assumptions

  • Continuum of agents
  • Each takes an action with intensity in continuous time
  • E.g.
  • Engage in political action
  • Adopt an innovation
  • Buy a stock
  • Random payoff X per unit of activity
  • Payoff realized at terminal date T
slide-44
SLIDE 44

Signals

  • Each agent endowed with one private signal about

fundamental X

  • Public signals arrive at discrete dates
  • Quarterly earnings surprises, or other discrete events
  • Distributions jointly normal
slide-45
SLIDE 45

Meetings and signal sharing

  • Information percolation
  • Duffie, Malamud & Manso (2009)
  • M eet randomly in pairs in continuous time
  • Share accumulated signals
  • Or, bloggers randomly seeing others’ postings
  • Per capita number of private signals grows exponentially:
  • See Andrei & Cujean (2017)
slide-46
SLIDE 46

Timeline

slide-47
SLIDE 47

Transmission bias

Two key transmission biases I’ll focus on:

  • b-bias
  • In each meeting, a bias of b added to average signal
slide-48
SLIDE 48

b-bias

b > 0:

  • Empirical literature, senders tend to transmit positively
  • Berger & Milkman (2012), Berger (2014)
  • Helps senders be perceived as useful by receivers
  • E.g., if short-selling of products or stocks is rare
  • Senders like to be viewed as positive people

b < 0:

  • Negativity bias in observer attention, perceptions
  • Rozin & Royzman (2001), Baumeister et al. (2001)
  • Evolutionary underpinnings
slide-49
SLIDE 49

b-bias

  • b added to average signal each meeting
  • b-bias recursively amplified
  • Even if , big effects
  • But if , people may mistakenly use a “

” heuristic.

slide-50
SLIDE 50

The decision problem

  • M ean-variance preferences
  • M yopic decision-making at each date s
slide-51
SLIDE 51

Two model versions

Action Booms M odel

  • Aggregate action endogenous
  • E.g., political activism

Price Bubbles M odel – M y focus tonight

  • Action:
  • Buying a single risky security
  • Aggregate buying = Exogenous aggregate supply shock
  • Price endogenous
  • Learning from price
slide-52
SLIDE 52

Expected price path

  • M ultiplicative growth in:
  • Per capita signal count
  • Exponential
  • Per-signal bias

Convex growth

  • For a while
slide-53
SLIDE 53

Correction

  • Public news arrival of increasing precision

Bubble eventually corrects

  • Hump-shaped expected price path
  • Or U-shaped when b < 0
slide-54
SLIDE 54

Expected price path in Price Bubbles Model: S moothed S moothed

slide-55
SLIDE 55

Evidence and comparison

  • Expected price starts growing slowly, accelerates
  • Initial convexity
  • Consistent with evidence of Greenwood, Shleifer & Y
  • u (2019)
  • Price bubble
  • Without usual ingredients
  • Overconfidence, overextrapolation
slide-56
SLIDE 56

Predictability of expectations and returns

  • Smoothed expected price path
  • Similar to behavioral models that generate momentum, reversal
  • After price run-ups, beliefs more optimistic
  • Supporting evidence
  • Greenwood & Shleifer (2014)
  • Emergent effect
  • Accumulation of b-bias
  • High returns do not cause overoptimism
  • No extrapolation
  • Disagreement, trading rises, falls with bubble
slide-57
SLIDE 57

Event-based return predictability

  • Corrective public information (earnings) arrives at discrete

dates

  • News-date returns predictable
  • Post-news-event return predictability
  • Other return predictability patterns
slide-58
SLIDE 58

Actual Expected Price Path—Stegosaurus

slide-59
SLIDE 59

Oscillation and unifying anomalies

  • Oscillations in expected price path
  • Short-term reversal too—unifying anomalies?
  • If high frequency public news arrival
slide-60
SLIDE 60

Peak inst eak instability ability

  • Tug-of-war
  • Biased percolation
  • Public news arrival
  • Before the peak, percolation wins
  • So oscillation intensifies near the

peak

  • Peak instability
  • Empirically testable
  • Damped oscillations as bubble

subsides

  • “Dead cat bounce”

Illustration: Robert Neubecker

slide-61
SLIDE 61

Fable 5: Biased transmission of folk models

slide-62
SLIDE 62

Biased transmission of folk models (work-in-progress)

Folk model:

  • An understanding of how the world

works.

  • E.g.,
  • Belief in Heaven and Hell
  • Payback criterion for capital budgeting
slide-63
SLIDE 63

Folk models spread from person to person

  • Drive human behavior, markets?
  • Shiller (1990, 2019), Denzau & North (1994), Hoff & Stiglitz (2016)
  • Often shaped by vivid narratives
  • Shiller (2017, 2019)
  • Or, mundane beliefs about correlations, cause & effect
slide-64
SLIDE 64

Example of competing competing folk models

  • Return continuation
  • “ The trend is your friend”
  • Return reversal
  • “Buy on the dips''
  • Behavioral finance
  • Taken as given
  • Social finance
  • Spread contagiously
slide-65
SLIDE 65

Compartmental models from epidemiology

  • Infection spreads through population via random

contacts

  • Proposed that SIR M odel helps explain:
  • Spread of folk models, behavior
  • Epidemic rise-and-fall time path
  • Shive (2010), Burnside, Eichenbaum & Rebelo (2016), Shiller

(2017, 2019), Chinco (2019)

slide-66
SLIDE 66

SIR Model epidemic curve

3 types in population:

slide-67
SLIDE 67

SIR model assumptions

  • At start, almost all
  • In a meeting, probability that becomes
  • M easures transmission bias
  • Random recovery
  • randomly becomes
  • Self-reinforcing effects
slide-68
SLIDE 68

Ada Adapting pting compartmental approach

  • T
  • transmission of folk models
  • Effects on economic behaviors
  • “ Infected”
  • Folk model Overoptimism
  • Overestimate expected terminal dividend
slide-69
SLIDE 69

SIRSmodel + Buzz Buzz + Asset m Asset mark arket et

  • S: spontaneously
  • Standard SIRS model
  • Introduce an asset market
  • Somewhat like Fable 4
slide-70
SLIDE 70

Buzz

  • Growing excitement makes folk model more

contagious

  • Transmission bias toward increasing in buzz,
  • E.g., Bitcoin “hot ” M eetings more persuasive
  • Negative buzz
  • M eetings trigger recovery from the folk model
  • Vs. disease models
  • Meetings never cure infection
slide-71
SLIDE 71

Effect of buzz

  • Exaggerates
  • Boom on way up
  • Collapse on way down
  • Intensifies self-reinforcing feature of bubbles
slide-72
SLIDE 72

Price Path: Modified SIRS model with Buzz

slide-73
SLIDE 73

Damped oscillation

  • Bubble collapse
  • Correction can overshoot

Damped oscillation in fraction, stock price

  • Suggests rich return autocorrelation patterns
slide-74
SLIDE 74

Emerg Emergen ent themes t themes

slide-75
SLIDE 75

Emergent theme (1): Compounding

  • 1. Social transmission bias compounds recursively.
  • Small bias can have large effects.
slide-76
SLIDE 76

Emergent theme (2): Idiosyncrasy

  • 2. Social transmission bias helps explain the

idiosyncrasy of aggregate outcomes.

  • Error-prone, unpredictable
  • Sensitivity to small biases, shocks
slide-77
SLIDE 77

Emergent theme (3): Dynamics

  • 3. Social transmission bias offers an endogenous social

explanation for action booms, price bubbles, and swings in investor sentiment.

  • Without “standard” ingredients for bubbles
slide-78
SLIDE 78

Emergent theme (4): Emergence

  • 4. Socially emergent behavior often look completely

different from individual propensities.

  • Self-enhancing transmission and attraction to volatility
  • Visibility bias and overconsumption
slide-79
SLIDE 79

Emergent theme (5): Mimicry

  • 5. Social emergence can easily create the illusion of a direct

individual propensity “for” a behavior when no such propensity exists.

  • Apparent lottery preferences
  • Self-enhancing transmission model
  • Apparent present-oriented consumption preference
  • Visibility bias model
  • Apparent extrapolative beliefs
  • Biased percolation model
  • We often overstate inferences from empirical tests.
slide-80
SLIDE 80

Emergent theme (6): Proxies

  • 6. The social transmission bias approach suggests new

test variables.

Sources of transmission bias:

  • Psychological traits
  • Environmental cues
  • Content of folk models
  • Textual characteristics

General social:

  • Sociability
  • Communication technologies,

media

  • Individual social network

position

  • Overall social network

connectivity

slide-81
SLIDE 81

In closing

slide-82
SLIDE 82
slide-83
SLIDE 83

S

  • ci

S

  • cial economics and f

al economics and finance inance as a resea as a research opp ch opportu

  • rtunity

nity

  • We academics sometimes caught in closed loops
  • At worst, ritualistic cycles
  • At best, blinders
  • John Keats felt:

...like some watcher of the skies When a new planet swims into his ken…

  • Outline of a new field of inquiry now discernable:
  • Social economics and finance
  • Networks, folk models, social transmission bias