Econ 219B

Psychology and Economics: Applications

(Lecture 11)

Stefano DellaVigna

April 12, 2017

Outline

1. Menu Effects: Confusion

2. Persuasion

3. Emotions: Mood

4. Emotions: Arousal

5. Methodology: Lab and Field Experiments

6. Happiness

7. Market Reaction to Biases: Introduction

8. Market Reaction to Biases: Pricing

1 Menu Effects: Confusion

• Previous heuristics reflect preference to avoid difficult choices or for salientoptions

• Confusion is simply an error in the implementation of the preferences• Different from most behavioral phenomena which are directional biases• How common is it?• Application 1. Shue-Luttmer (2009)— Choice of a political candidate among those in a ballot

— California voters in the 2003 recall elections

• Do people vote for the candidate they did not mean to vote for?

• Design:— Exploit closeness on ballot

— Exploit specific features of closeness

— Exploit random variation in placement of candidates on the ballot (asin Ho-Imai)

• First evidence: Can this matter?• If so, it should affect most minor party candidates

• Model:— Share 1 of voters meaning to vote for major candidate vote forneighboring candidate

— Estimate 1 by comparing voting for when close to and when farfrom

— Notice: The impact depends on vote share of

— Specification:

= 0 + 1 ∗ + +

— Rich set of fixed effects, so identify off changes in order

• Results:— 1 in 1,000 voters vote for adjacent candidate

— Difference in error rate by candidate (see below)

— Notice: Each candidate has 2.5 adjacent candidates — Total misvotingis 1 in 400 voters

• Interpretations:1. Limited Attention: Candidates near major candidate get reminded inmy memory

2. Trembling Hand: Pure error

• To distinguish, go back to structure of ballot.— Much more likely to fill-in the bubble on right side than on left side if(2)

— No difference if (1)

• Effect is mostly due to Trembling hand / Confusion• Additional results:— Spill-over of votes larger for more confusing voting methods (such aspunch-cards)

• Spill-over of votes larger for precincts with a larger share of lower-educationdemographics — more likely to make errors when faced with large numberof option

• This implies (small) aggregate effect: confusion has a different prevalenceamong the voters of different major candidates

• Rashes (JF, 2001) Similar issue of confusion for investor choice• Two companies:— Major telephone company MCI (Ticker MCIC)

— Small investment company (ticker MCI)

— Investors may confuse them

— MCIC is much bigger — this affects trading of company MCI

• Check correlation of volume (Table III)— High correlation

— What if two stocks have similar underlying fundamentals?

— No correlation of MCI with another telephone company (AT&T)

• Predict returns of smaller company with bigger company (Table IV)• Returns Regression:

= 0 + 1 + +

• Results:

— Positive correlation 1 — The swings in volume have some impact onprices.

— Difference between reaction to positive and negative news:

= 0+1+2∗1³ 0

´++

— Negative 2 Effect of arbitrage — It is much easier to buy by mistakethan to short a stock by mistake

• Size of confusion? Use relation in volume.— We would like to know the result (as in Luttmer-Shue) of

= + +

— Remember: = ( ) ()

— We know (Table I)

5595 = =( )q

() ()=

= ∗q ()q ()

— Hence, = 5595 ∗q ()

q () = 5595 ∗

10−3 = 5 ∗ 10−4— Hence, the error rate is approximately 5 ∗ 10−4 that is, 1 in 2000

• Conclusion— Deviation from standard model: confusion.

— Can have an aggregate impact, albeit a small one

— Can be moderately large for error from common choice to rare choice

— Other applications: eBay bidding on misspelled names (find cheaperitems when looking for ‘shavre’ [shaver] or ‘tyo’ [toy]

2 Persuasion

• Persuasion: Change in opinion/action beyond prediction of Bayesian model• Persuasion: Sender attempts to convince Receiver with words/images totake an action

— Rational persuasion through Bayesian updating

— Non-rational persuasion, i.e.: neglect of incentives of person presentinginformation

— Effect of persuasion directly on utility function (advertising/emotions)

• Compare to Social Pressure: Presence of Sender exerts pressure to take anaction

• DellaVigna and Gentzkow (2010): Overview on Persuasion:— Persuading consumers: Marketing

— Persuading voters: Political Communication

— Persuading donors: Fund-raising

— Persuading investors: Financial releases

• First problem: How to measure when persuasion occurs?• Treatment group T, control group C, Persuasion Rate is

= 100 ∗ − −

1

1− 0

— is the share of group receiving the message,

— is the share of group adopting the behavior of interest,

— 0 is the share that would adopt if there were no message

Paper Treatment Control Variable t Time Treatment Control Exposure PersuasionHorizon group t T group t C rate e T -e C rate f

(1) (2) (4) (7) (9) (10) (11) (12)Persuading Consumers

Simester et al. (2007) (NE) 17 clothing catalogs sent 12 catalogs Share Purchasing 1 year 36.7% 33.9% 100%* 4.2%>= 1 item 69.1% 66.8% 100%* 6.9%

Bertrand, Karlan, Mullainathan, Mailer with female photo Mailer no photo Applied for loan 1 month 9.1% 8.5% 100%* 0.7%Shafir, and Zinman (2010) (FE) Mailer with 4.5% interest rate Mailer 6.5% i.r. 9.1% 8.5% 100%* 0.7%

Persuading VotersGosnell (1926) Card reminding of registration No card Registration Few days 42.0% 33.0% 100.0% 13.4%

Gerber and Green (2000) (FE) Door-to-Door GOTV Canvassing No GOTV Turnout Few days 47.2% 44.8% 27.9% 15.6%GOTV Mailing of 1-3 Cards No GOTV 42.8% 42.2% 100%* 1.0%

Green, Gerber, Door-to-Door Canvassing No GOTV Turnout Few days 31.0% 28.6% 29.3% 11.5%and Nickerson (2003) (FE)Green and Gerber (2001) (FE) Phone Calls By Youth Vote No GOTV Turnout Few days 71.1% 66.0% 73.7% 20.4%

Phone Calls 18-30 Year-Olds No GOTV Turnout 41.6% 40.5% 41.4% 4.5%

DellaVigna and Kaplan (2007) (NE) Availab. of Fox News Via Cable No F.N. via cable Rep. Vote Share 0-4 years 56.4% 56.0% 3.7% 11.6%+

Enikolopov, Petrova, and Zhuravskaya (2010) (NE)

Availability of independent anti-Putin TV station (NTV) No NTV

Vote Share of anti-Putin parties 3 months 17.0% 10.7% 47.0% 7.7%+

Knight and Chiang (2010) (NE) Unsurprising Dem. Endors. (NYT) No endors. Support for Gore Few 75.5% 75.0% 100.0% 2.0% Surprising Dem. Endors. (Denver) No endors. weeks 55.1% 52.0% 100.0% 6.5%

Gerber, Karlan, and Bergan (2009) (FE)

Free 10-week subscription to Washington Post No Subscr.

Dem. Vote Share (stated in survey) 2 months 67.2% 56.0% 94.0% 19.5%+

Gentzkow (2006) (NE) Exposure to Television No Television Turnout 10 years 54.5% 56.5% 80.0% 4.4%

Gentzkow and Shapiro (2009) (NE) Read Local Newspaper No local paper Turnout 0-4 years 70.0% 69.0% 25.0% 12.9%

TABLE 1, PART APERSUASION RATES: SUMMARY OF STUDIES

Paper Treatment Control Variable t Time Treatment Control Exposure PersuasionHorizon group t T group t C rate e T -e C rate f

(1) (2) (4) (7) (9) (10) (11) (12)

List and Lucking-Reiley Fund-raiser mailer with low seed No mailer Share 1-3 weeks 3.7% 0% 100%* 3.7%(2002) (FE) Fund-raiser mailer with high seed No mailer Giving Money 8.2% 0% 100%* 8.2%

Landry, Lange, List, Price, Door-To-Door Fund-raising No visit Share immediate 10.8% 0% 36.3% 29.7%and Rupp (2006) (FE) Campaign for University Center Giving Money

DellaVigna, List, and Malmendier Door-To-Door Fund-raising No visit Share immediate 4.6% 0% 41.7% 11.0%(2009) (FE) Campaign for Out-of-State Charity Giving Money

Falk (2007) (FE) Fund-raiser mailer with no gift No mailer Share 1-3 weeks 12.2% 0% 100%* 12.2%Mailer with gift (4 post-cards) No mailer Giving Money 20.6% 0% 100%* 20.6%

Engelberg and Parsons (2009) (NE) Coverage of Earnings News No coverage Trading of Shares 3 days 0.023% 0.017% 60.0% 0.010%in Local Paper of Stock in News

TABLE 1, PART BPERSUASION RATES: SUMMARY OF STUDIES

Notes: Calculations of persuasion rates by the authors. The list of papers indicates whether the study is a natural experiment ("NE") or a field experiment ("FE"). Columns (9) and (10) report the value of the behavior studied (Column (4)) for theTreatment and Control group. Column (11) reports the Exposure Rate, that is, the difference between the Treatment and the Control group in the share of people exposed to the Treatment. Column (12) computes the estimated persuasion rate f a100*(tT-tC)/((eT-eC)*(1-tC)). The persuasion rate denotes the share of the audience that was not previously convinced and that is convinced by the message. The studies where the exposure rate (Column (11) is denoted by "100%*" are cases inwhich the data on the differential exposure rate between treatment and control is not available. In these case, we assume eT-eC=100%, which implies that the persuasion rate is a lower bound for the actual persuasion rate. In the studies on"Persuading Donors", even in cases in which an explicit control group with no mailer or no visit was not run, we assume that such a control would have yielded tC=0%, since these behaviors are very rare in absence of a fund-raiser. For studies

Persuading Donors

Persuading Investors

• Persuasion rate helps reconcile seemingly very different results, e.g. per-suading voters

• More in detail: DellaVigna-Kaplan (QJE, 2007), Fox News natural ex-periment

1. Fast expansion of Fox News in cable markets

— October 1996: Launch of 24-hour cable channel

— June 2000: 17 percent of US population listens regularly to Fox News(Scarborough Research, 2000)

2. Geographical differentiation in expansion

— Cable markets: Town-level variation in exposure to Fox News

— 9,256 towns with variation even within a county

3. Conservative content

— Unique right-wing TV channel (Groseclose and Milyo, 2004)

• Empirical Results

• Selection. In which towns does Fox News select? (Table 3):2000 = +

Pres1996 + 1996 + Γ20002000 +

Γ00−9000−90 + Γ2000 +

• Controls — Cable controls (Number of channels and potential subscribers)

— US House district or county fixed effects

• Conditional on , Fox News availability is orthogonal to— political variables

— demographic variables

• Baseline effect — Presidential races• Effect on Presidential Republican vote share (Table 4):

Pres2000 −

Pres1996 = +

2000 + Γ20002000 +

Γ00−9000−90 + Γ2000 +

• Results:— Significant effect of Fox News with district (Column 3) and countyfixed effects (Column 4)

— .4-.7 percentage point effect on Republican vote share in Pres. elections

— Similar effect on Senate elections — Effect is on ideology, not person-specific

— Effect on turnout

• Magnitude of effect: How do we generalize beyond Fox News?• Estimate audience of Fox News in towns that have Fox News via cable(First stage)

— Use Scarborough micro data on audience with Zip code of respondent

— Fox News exposure via cable increases regular audience by 6 to 10percentage points

— How many people did Fox News convince?

— Heuristic answer: Divide effect on voting (.4-.6 percentage point) byaudience measure (.6 to .10)

• Result: Fox News convinced 3 to 8 percent of audience (Recall measure)or 11 to 28 percent (Diary measure)

• How do we interpret the results?• Benchmark model:1. New media source with unknown bias with ∼

µ0

1

¶2. Media observes (differential) quality of Republican politician, ∼

³0 1

´, i.i.d., in periods 1 2

3. Media broadcast: = + Positive implies pro-Republicanmedia bias

4. Voting in period Voters vote Republican if b + 0, with ideological preference

• Signal extraction problem. New media (Fox News) says Republican politi-cian (George W. Bush) is great

— Is Bush great?

— Or is Fox News pro-Republican?

• A bit of both, the audience thinks. Updated media bias after periods:

=0 +

+

• Estimated quality of Republican politician:

=∗ 0 +

h −

i+

=

h −

i+

• Persuasion. Voter with persuasion (0 ≤ ≤ 1) does not take intoaccount enough media bias:

=

[ − (1− ) ]

+

• Vote share for Republican candidate. (+ b ≥ 0) = 1− (−b )• Proposition 1. Three results:1. Short-Run I: Republican media bias increases Republican vote share:

[1− (−b )] 0

2. Short-Run II: Media bias effect higher if persuasion ( 0).

3. Long-run ( →∞) Media bias effect ⇐⇒ persuasion 0

• Intuition.— Fox News enthusiastic of Bush

— Audience updates beliefs: “This Bush must be really good” (Short-Run I)

— Believe media more if credulous or persuadable (Short-Run II)

— But: Fox News enthusiastic also of Karl Rove, Rick Lazio, Bill Frist– “They cannot be all good!”

— Make inference that Fox News is biased, stop believing it

— Fox News influences only individuals subject to persuasion (Long-Run)

• What is the evidence about persuasion bias?

• Cain-Loewenstein-Moore (JLegalStudies, 2005). Psychology Experi-ment

— Pay subjects for precision of estimates of number of coins in a jar

— Have to rely on the advice of second group of subjects: advisors

— (Advisors inspect jar from close)

— Two experimental treatments:

∗ Aligned incentives. Advisors paid for closeness of subjects’ guess∗ Mis-Aligned incentives, Common knowledge. Advisors paid for howhigh the subjects’ guess is. Incentive common-knowledge

∗ (Mis-Aligned incentives, Not Common knowledge.)

• Result 1: Advisors increase estimate in Mis-Aligned incentives treatment– Even more so when common knowledge

• Result 2. Estimate of subjects is higher in Treatment with Mis-Alignedincentives

• Subjects do not take sufficiently into account incentives of informationprovider

• Effect even stronger when incentives are known — Advisors feel free(er)to increase estimate

• Applications to many settings

• Application 1: Malmendier-Shantikumar (JFE, 2007).— Field evidence that small investors suffer from similar bias

— Examine recommendations by analysts to investors

— Substantial upward distortion in recommendations (Buy=Sell, Hold=Sell,etc)

• Higher distortion for analysis working in Inv. Bank affiliated with companythey cover (through IPO/SEO)

• Question: Do investors discount this bias?— Analyze Trade Imbalance (essentially, whether trade is initiated byBuyer)

— Assume that

∗ large investors do large trades∗ small investors do small trades

— See how small and large investors respond to recommendations

• Examine separately for affiliated and unaffiliated analysts

• Results:— Small investor takes analyst recommendations literally (buy Buys, sellSells)

— Large investors discount for bias (hold Buys, sell Holds)

— Difference is particularly large for affiliated analysts

— Small investors do not respond to affiliation information

• Strong evidence of distortion induced by incentives

3 Emotions: Mood

• Emotions play a role in several of the phenomena considered so far:— Self-control problems — Temptation

— Projection bias in food consumption — Hunger

— Social preferences in giving — Empathy

— Gneezy-List (2006) transient effect of gift — Hot-Cold gift-exchange

• Psychology: Large literature on emotions (Loewenstein and Lerner, 2003)— Message 1: Emotions are very important

— Message 1: Different emotions operate very differently: anger 6= mood6=

• Consider two examples of emotions:— Mood

— Arousal

• Psychology: even minor mood manipulations have a substantial impact onbehavior and emotions

— On sunnier days, subjects tip more at restaurants (Rind, 1996)

— On sunnier days, subjects express higher levels of overall happiness(Schwarz and Clore, 1983)

• Should this impact economic decisions?

• Field: Impact of mood fluctuations on stock returns:— Daily weather and Sport matches

— No effect on fundamentals

— However: If good mood leads to more optimistic expectations — In-crease in stock prices

• Evidence:— Saunders (1993): Days with higher cloud cover in New York areassociated with lower aggregate US stock returns

— Hirshleifer and Shumway (2003) extend to 26 countries between1982 and 1997

∗ Use weather of the city where the stock market is located∗ Negative relationship between cloud cover (de-trended from seasonalaverages) and aggregate stock returns in 18 of the 26 cities

• — Magnitude:

— Days with completely covered skies have daily stock returns .11 percentlower than days with sunny skies

— Five percent of a standard deviation

— Small magnitude, but not negligible

• After controlling for cloud cover, other weather variables such as rain andsnow are unrelated to returns

• Additional evidence (Edmans-Garcia-Norli, 2007): International soccermatches (39 countries, 1973-2004)

• Results:— Compared to a day with no match, a loss lowers daily returns (signifi-cantly) by .21 percent. (Surprisingly, a win has essentially no effect)

— More important matches, such as World Cup elimination games, havelarger effects

— Effect does not appear to depend on whether the loss was expected ornot

— International matches in other sports have a consistent, though smaller,effect (24 countries)

• Interpretations:— Mood impacts risk aversion or perception of volatility

— Mood is projected to economic fundamentals

• Simonsohn (2007): Subtle role of mood— Weather on the day of campus visit to a prestigious university (CMU)

— Students visiting on days with more cloud cover are significantly morelikely to enroll

— Higher cloud cover induces the students to focus more on academicattributes versus social attributes of the school

— Support from laboratory experiment

4 Emotions: Arousal

• Separate impact of emotions: Arousal• Josephson (1987): Arousal due to violent content— Control group exposed to non-violent clip

— Treatment group exposed to violent clip

— Treatment group more likely to display more aggressive behavior, suchas aggressive play during a hockey game

— Impact not due to imitation (violent movie did not involve sport scenes)

• Consistent finding from large set of experiments (Table 11)• Dahl-DellaVigna (2009): Field evidence – Exploit timing of release ofblockbuster violent movies

• Model. Consumer chooses between strongly violent movie , mildlyviolent movie , non-violent movie , or alternative social activity

— Utility depends on quality of movies — Demand functions ()

• Heterogeneity:— High taste for violence (Young): consumers

— Low taste for violence (Old): consumers

— Aggregate demand for group : ( )

• Production function of violence (not part of utility fct.) depends on and :

ln =X=

[X

=

(

)+(1− ( )− ( )− ( ))]

• Estimate ( is total attendance to movie of type )

ln = 0 + + + +

• Estimated impact of exposure to violent movies : = ( − ) + (1− )( − )

• First point – Estimate of net effect

— Direct effect: Increase in violent movie exposure —

— Indirect effect: Decrease in Social Activity —

• Second point – Estimate on self-selected population:

— Estimate parameters for group actually attending movies

— Young over-represented: ( +)

• Comparison with Psychology experiments— Natural Experiment. Estimated impact of exposure to violent movies:

= ( − ) + (1− )( − )

— Psychology Experiments. Manipulate directly, holding constant

out of equilibrium

=

+ + (1−

+)

• Two differences:— ‘Shut down’ alternative activity, and hence does not appear

— Weights representative of (student) population, not of population thatselects into violent movies

• Movie data— Revenue data: Weekend (top 50) and Day (top 10) from The Numbers

— Violence Ratings from 0 to 10 from Kids In Mind (Appendix Table 1)

— Strong Violence Measure : Audience with violence 8-10 (Figure 1a)

— Mild Violence Measure : Audience with violence 5-7 (Figure 1b)

• Assault data— Source: National Incident-Based Reporting System (NIBRS)

— All incidents of aggravated assault, simple assault, and intimidationfrom 1995 to 2004

— Sample: Agencies with no missing data on crime for 7 days

— Sample: 1995-2004, days in weekend (Friday, Saturday, Sunday)

• Regression Specification. (Table 3)log =

+ +

+ Γ +

— Coefficient is percent increase in assault for one million peoplewatching strongly violent movies day (

) (Similarly and )

— Cluster standard errors by week

• Results.

— No effect of movie exposure in morning or afternoon (Columns 1-2)

— Negative effect in the evening (Column 3)

— Stronger negative effect the night after (Column 4)

• Additional Results:

— No Medium-Run Effects.

∗ No effect on Monday and Tuesday of weekend exposure

∗ No effect one, two, or three weeks later

— Placebo:

∗ No effect on crime the week after

∗ No effect if randomly draw year and reassign dates

— Similar result for DVD-VHS Rentals

• Summary of Findings:

1. Violent movies lower same-day violent crime in the evening (incapaci-tation)

2. Violent movies lower violent crime in the night after exposure (lessconsumption of alcohol in bars)

3. No lagged effect of exposure in weeks following movie attendance —No intertemporal substitution

4. Strongly violent movies have slightly smaller impact compared to mildlyviolent movies in the night after exposure

• Interpret Finding 4 in light of Lab-Field debate

• Finding 4. Non-monotonicity in Violent Content

— Night hours: = −00192 versus = −00205

— Odd if more violent movies attract more potential criminals

— Model above — Can estimate direct effect of violent movies if cancontrol for selection

− = −Ã +

−

− ( − )

!

— Do not observe selection of criminals , but observe selection of cor-related demographics (young males)

— IMDB ratings data – Share of young males among raters increaseswith movie violence (Figure 2) — Use as estimate of

— Compute \ − = 011 ( = 08), about one third of total effect

— Pattern consistent with arousal induced by strongly violent movies( )

• Bottom-line 1: Can reconcile with laboratory estimates

• Bottom-line 1: Can provide benchmark for size of arousal effect

• Differences from laboratory evidence (Levitt-List, 2007): Exposure to vio-lent movies is

— Less dangerous than alternative activity ( )(Natural Experiment)

— More dangerous than non-violent movies ( )(Laboratory Experiments and indirect evidence above)

• Both types of evidence are valid for different policy evaluations— Laboratory: Banning exposure to unexpected violence

— Field: Banning temporarily violent movies

• This leaves a number of open questions

• Example: Peer Effects through the media.— To what extent do we imitate role models in the media?

— Ongoing work: Movies with Car races — Dangerous driving — Caraccidents?

— Can measure exact duration of car chases and intensity

— Is imitation higher for characters of same race and gender?

5 Methodology: Lab and Field

• What do we learn about the relationship between lab experiments and fieldevidence?

• Contentious topic recently since List-Levitt (JEP, 2007)• To simplify, define field evidence as:— Natural Experiments

— Field Experiments

• Let us start from Dahl-DellaVigna example

• Difference 1. Differences in comparison group— Lab Experiment: Activity in control group exogenously assigned

— Natural Experiment: Activity in control group chosen to max utility

— Notice: Field Experiments are (usually) like lab experiments

• Implication: Parameters estimated very different• Write down model: what parameter are you estimating?

• Difference 2. Self-Selection— Lab Experiment: Subjects are group of students unaware of nature oftask — No selection

— Natural Experiment: People self-select into a setting

— Field Experiments: Can have self-selection too

• Different purposes:— Often useful to control for self-selection and impose a treatment

— However, can lose external validity — Put people in a situation theynormally would not be in

• Example: Social preferences— I give $10 if confronted with fund-raiser asking for money

— However: I do all possible to avoid this interaction

— — Without sorting: Frequent giving

— — With sorting: No giving

• Notice: One can integrate sorting into laboratory experiments

• Difference 3. Differences in context• Example 1: Dahl-DellaVigna— Laboratory experiments on movie violence: 15-min, clips (to save time)— Field: Full-length movies

• Example 2: Dictator experiment— Laboratory: Have been given $10 — Give it to anonymous subject— Field: Have earned money — Give some of it to someone

• Example 3: Prisoner Dilemma experiment— Framed as ‘Community Game’ — Low defection— Framed as ‘Wall-Street Game’ — High defection

• Tension for laboratory experiments: Resemble field at cost of losing exper-imental controls

• Difference 4. Demand effects in the laboratory— Subjects generate the effect that they think experimenter is looking for

— Social preference!

• Example: Dictator game— I was given $10 and asked how much to give – Inference: Shouldgive some away

• Field evidence does not have this feature

• However:— This is genuine phenomenon also in field (Obedience)

— Trade-off between demand effects and loss of control in the field

• Related: Anonymity— Situations are rarely double-blind even in experiments

— If subjects worry about experimenter, this affects behavior

• Again: Same issue also in the field

• Advantage of lab: Can control for this by running double-blind sessions

• Difference 5. Differences in Stakes— Laboratory: Small stakes

— Field: Large stakes

• Examples:— Dictator Games for $10 vs. $100+ of charitable giving

— Aggressive hockey play in Violence experiments vs. violent crime

• However:— Evidence not consistent that large stakes change behavior

— In field, many repeated interactions, all with small stakes

6 Happiness

• Is there a more direct way to measure utility?

• What about happiness questions?— ‘Taken all together, how would you say things are these days, wouldyou say that you are very happy, pretty happy, or not too happy?’

— or ‘How satisfied are you with your life as a whole?’

— Response on 1 to 7 of 0 to 10 scale

— Could average response measure utility?

• There are a number of issues:

1. (Noise I) Is the measure of happiness just noise?

2. (Noise II) Even if valid, there are no incentives, how affected is it byirrelevant cues?

3. (Scale) Happiness is measured on discrete intervals, with ceiling andfloor effect

4. (Content) What exactly does the measure capture? Instantaneous util-ity? Discounted utility?

• Revealed preference approach remains heavily favored by economists (my-self included)

• Still, significant progress in last 10-15 years on taking some role in eco-nomics

• Issue 1 (Noise I). To address,— Take happiness measure

— Does it responds to well-identified, important shifters which affectimportant economic outcomes?

• Oreopoulos (AER 2006). Exploit binding compulsory schooling laws tostudy returns to education

• UK: 1947 increase in minimum schooling from 14 to 15

• Northern Ireland: 1957 increase from 14 to 15

• Clear impact on earnings: compare earnings for adults aged 32-64 as a

function of year of birth

• Implied returns to compulsory education: 0.148 (0.046)

• Oreopoulos (JPubE 2007): Did this affect happiness measures?— Eurobarometer Surveys in UK and N. Ireland, 1973-1998

— Question on 1-4 scale

• Results:— One year of additional (compulsory) education increases happiness some-where between 2 and 8 percent

— In addition, large effects on health and wealth

— Reinforces puzzle: Why don’t people stay in school longer?

• Happiness response captures real information

• Happiness answer also responds to cues (Issue 2), has scale effects (Issue3), but valid enough to use in combination with other measures

• However, Issue 4: How would we use happiness measure as part of eco-nomic research?

• Research agenda by Dan Benjamin, Ori Heffetz, Miles Kimball, Alex Rees-Jones

— Study Econ101a-type simple issues with happiness measures

— Critical to know how to correctly interpret these measures

• Paper 1. Benjamin, Heffetz, Kimball, and Rees-Jones (AER 2012)— How does happiness (subjective well-being) relate to choice?

— Compare forecasted happiness with choice in several hypothetical sce-narios

— Forecasts of happiness predict choice quite well, but other factors alsoplay a role

• Paper 2. Benjamin et al. (AER forthcoming)— Medical students choosing match for residency

— Survey to elicit ranking of medical schools for residency + Ask antici-pated happiness

— How well does happiness predict choice relative to other factors?

• Some evidence that one can also elicit intertemporal happiness

• Other important work on happiness:— Luttmer (QJE 2005): Documents relative aspect of happiness: Anincrease in income of neighbors (appropriately instrumented) lower lifesatisfaction

— Stevenson and Wolfers (Brookings 2008):

∗ Debunks Easterlin paradox (income growth over time does not in-crease happiness)

∗ Clear link over time between log income and happiness— Finkelstein, Luttmer, Notowidigdo (JEEA 2014):

∗ How does marginal utility of consumption vary with health? Neededfor optimal policies

∗ Observe changes in happiness for varying health

7 Market Reaction to Biases: Introduction

• So far, we focused on consumer deviations from standard model

• Who exhibits these deviations?1. Self-control and naivete’. Consumers (health clubs, food, creditcards, smoking), Employees (retirement saving, benefit take-up), Stu-dents (homework)

2. Reference dependence. Workers (labor supply, increasing wages),(inexperienced) traders (sport cards), Investors, Consumers (insurance),House owners

3. Social preferences. Consumers (giving to charities), Employees (ef-fort, strikes)

4. Biased Beliefs. Individual investors, CEOs, Consumers (purchases,betting)

5. Inattention. Individual investors, Consumers (eBay bidding, taxation)

6. Menu Effects. Individual investors, Voter, Consumers (loans, 410(k)plans)

7. Social Pressure and Persuasion. Voters, Employees (productivity),Individual investors (and analysts)

8. Emotions. Individual investors, Consumers

• What is missing from picture?

• — Experienced agents

— Firms

— Broadly speaking, market interactions with ‘rational’ agents

• Market interactions— Everyone ‘born’ with biases

— But: Effect of biases lower if:

∗ learning with plenty of feedback∗ advice, access to consulting∗ specialization∗ Competition ‘drives out of market’ (BUT: See last lecture)

• For experienced agents these conditions are more likely to be satisfied

• Implications? Study biases in the market

• Six major instances:— Interaction between firms and consumers (contract design, price choice)

— Interaction between experienced and inexperienced investors (noise tradersand behavioral finance)

— Interaction between managers and investors (corporate finance)

— Interaction between employers and employees (labor economics)

— Interaction between politicians and voters (political economy)

— Institutional design

8 Market Reaction to Biases: Pricing

• Consider now the case in which consumers purchasing products have biases

• Firm maximize profits

• Do consumer biases affect profit-maximizing contract design?

• How is consumer welfare affected by firm response?

• DellaVigna and Malmendier (QJE 2004). Consumers with³

´preferences

8.1 Self-Control I

MARKET (I). INVESTMENT GOODS

• Monopoly• Two-part tariff: (lump-sum fee) (per-unit price)• Cost: set-up cost , per-unit cost

Consumption of investment good

Payoffs relative to best alternative activity:

• Cost at = 1 stochastic— non-monetary cost

— experience good, distribution ()

• Benefit 0 at = 2 deterministic

FIRM BEHAVIOR. Profit-maximization

max

{− + (− ) (− )}

s.t.

(−+

Z −−∞

(− − ) ()

)≥

• Notice the difference between and • Substitute for to maximize

max

(Z −−∞

(− − ) () + (− ) (− )− −

)

Solution for the per-unit price ∗:

∗ = [exponentials]

−³1−

´³− ∗

´ (− ∗)

[sophisticates]

−³− ∗

´− (− ∗)

(− ∗)[naives]

Features of the equilibrium

1. Exponential agents ( = = 1).Align incentives of consumers with cost of firm=⇒ marginal cost pricing: ∗ = .

∗ = [exponentials]

−³1−

´³− ∗

´ (− ∗)

[sophisticates]

−³− ∗

´− (− ∗)

(− ∗)[naives]

2. Hyperbolic agents. Time inconsistency=⇒ below-marginal cost pricing: ∗ .

(a) Sophisticates ( = 1): commitment.

(b) Naives ( = 1): overestimation of consumption.

MARKET (II). LEISURE GOODS

Payoffs of consumption at = 1:

• Benefit at = 1 stochastic• Cost at = 2 deterministic

=⇒ Use the previous setting: − is “current benefit”, 0 is “future cost.”

Results:

1. Exponential agents.

Marginal cost pricing: ∗ = , ∗ = (PC).

2. Hyperbolic agents tend to overconsume. =⇒Above-marginal cost pricing: ∗ . Initial bonus ∗ (PC).

EXTENSIONS

• Perfect Competition. Can write maximization problem as

max

− +Z −−∞

(− − ) ()

s.t. {− + (− ) (− )} = 0— Implies the same solution for ∗

• Heterogeneity. Simple case of heterogeneity:— Share of fully naive consumers ( = 1)

— Share 1− of exponential consumers ( = = 1)

— At = 0 these consumers pool on same contract, given no immediatepayoffs

• Maximization (with Monopoly):max

{− + [ (− ) + (1− ) (− )] (− )}

s.t. − +Z −−∞

(− − ) () ≥

• Solution:∗ =

− (− )− (− )

(− ) + (1− ) (− )

• The higher the fraction of naives the higher the underpricing of

EMPIRICAL PREDICTIONS

Two predictions for time-inconsistent consumers:

1. Investment goods (Proposition 1):

(a) Below-marginal cost pricing

(b) Initial fee (Perfect Competition)

2. Leisure goods (Corollary 1)

(a) Above-marginal cost pricing

(b) Initial bonus or low initial fee (Perfect Competition)

FIELD EVIDENCE ON CONTRACTS

• US Health club industry ($11.6bn revenue in 2000)— monthly and annual contracts

— Estimated marginal cost: $3-$6 + congestion cost

— Below-marginal cost pricing despite small transaction costs and pricediscrimination

• Vacation time-sharing industry ($7.5bn sales in 2000)— high initial fee: $11,000 (RCI)

— minimal fee per week of holiday: $140 (RCI)

• Credit card industry ($500bn outstanding debt in 1998)— Resale value of credit card debt: 20% premium (Ausubel, 1991)

— No initial fee, bonus (car / luggage insurance)

— Above-marginal-cost pricing of borrowing

• Gambling industry: Las Vegas hotels and restaurants:— Price rooms and meals below cost, at bonus

— High price on gambling

WELFARE EFFECTS

Result 1. Self-control problems + Sophistication ⇒ First best

• Consumption if ≤ − ∗

• Exponential agent:— ∗ =

— consume if ≤ − ∗ = −

• Sophisticated time-inconsistent agent:— ∗ = − (1− )

— consume if ≤ − ∗ = −

• Perfect commitment device• Market interaction maximizes joint surplus of consumer and firm

Result 2. Self-control + Partial naiveté ⇒ Real effect of time inconsistency

• ∗ = − [ (− ∗)− (− ∗)](− ∗)

• Firm sets ∗ so as to accentuate overconfidence

• Two welfare effects:— Inefficiency: Surplusnaive ≤ Surplussoph.

— Transfer (under monopoly) from consumer to firm

• Profits are increasing in naivete’ (monopoly)• Welfarenaive ≤ Welfaresoph.

• Large welfare effects of non-rational expectations

8.2 Self-Control II

• Eliaz and Spiegler (RES 2006), Contracting with Diversely Naive Agents.

• Extend DellaVigna and Malmendier (2004):— incorporate heterogeneity in naiveté

— allow more flexible functional form in time inconsistency

— different formulation of naiveté

• Setup:1. Actions:

— Action ∈ [0 1] taken at time 2— At time 1 utility function is ()

— At time 2 utility function is ()

2. Beliefs: At time 1 believe:

— Utility is () with probability

— Utility is () with probability 1−

— Heterogeneity: Distribution of types

3. Transfers:

— Consumer pays firm ()

— Restrictive assumption: no cost to firm of providing

• Therefore:— Time inconsistency ( 1) — Difference between and

— Naiveté ( ) — 0

— Partial naiveté here modelled as stochastic rather than deterministic

— Flexibility in capturing time inconsistency (self-control, reference de-pendence, emotions)

• Main result:• Proposition 1. There are two types of contracts:1. Perfect commitment device for sufficiently sophisticated agents ( )

2. Exploitative contracts for sufficiently naive agents ( )

• Commitment device contract:— Implement = max ()

— Transfer:

∗ () = max ()

∗ () =∞ for other actions

— Result here is like in DM: Implement first best

• Exploitative contract:— Agent has negative utility:

()− () 0

— Maximize overestimation of agents:

= argmax ( ()− ())

9 Next Lecture

• Market Response to Biases— Employees: Behavioral Labor

— Investors: Behavioral Finance

— Voters: Behavioral Political Economy