Download Constructed" Preferences

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

False consensus effect wikipedia , lookup

Transcript
“Constructed” preferences


Study effects relative to “complete, transitive” u(x)
“Constructed” means expression of preference is like problemsolving:








Will you vote for John Kerry?
Answered by rapid intuition (tall, good hair) and deliberate logic
(positions on issues)
Context-dependence (comparative)
Description-dependent “framing
(descriptions guide attention)
Reference-dependence (changes, not levels; anchoring)
Some values “protected”/sacred (health, environment)
Is too much choice bad?
Open questions:




Are effects smaller with familiar choices?
Experts?
Markets?
New predictions (e.g. “big tip” labor supply experiment)
1/n heuristic & partition dependence in the lab (cf.
“corporate socialism”
Partition-dependence in prediction
markets for economic statistics
An Example: Price of Digital Options
Auction on Retail Trade Release for April 2005; Held May 12, 2005
.1
.094 .092
.087
.08
.087 .085
.074
.065
.06
.052
.04
.035
.063 .063
.052
.051
.034
.025
.02
0
.015
.017
.011
-0 <
.2 -0
0 .2
0
t
-0 o .1 0.1
0
t 0
0. o 0
00 .0
t 0
0. o 0
10 .1
t 0
0. o 0
20 .2
t 0
0. o 0
30 .3
t 0
0. o 0
40 .4
t 0
0. o 0
50 .5
t 0
0. o 0
60 .6
t 0
0. o 0
70 .7
t 0
0. o 0
80 .8
t 0
0. o 0
90 .9
t 0
1. o 1
00 .0
t 0
1. o 1
10 .1
t 0
1. o 1
20 .2
t 0
1. o 1
30 .3
to 0
1.
40
>1
.4
0

What if prices
(forecasts p(x)) are a
mixture of “best
forecast” f(x) and 1/n
probability?
Can back out “truth”
from t(x)=(1-π)f(x)+
π(1/n) and make a
better prediction of the
actual result
Price / Probability

Context-dependence (comparative)

Objects judged relative to others in a
choice set



Asymmetric dominance
Compromise effects
Economic question: What is seller’s
optimal choice set given contextdependent preferences?
Description-dependent “framing” (descriptions
guide attention)


Analogy to figure-ground in perception
Actual study with n=792 docs (Harvard Med, Brigham &Women’s,
Hebrew U; McNeil et al JAMA ’80s)



Surgery
Radiation







Surgery
Radiation
treatment
10%
0%
treatment
90%
100%
1 yr
32%
23%
1 yr
68%
77%
5 yrs
choice
5 yrs
choice
66%
78%
34%
22%
53%
47%
82%
18%
both frames
60%
40%
Asian disease problem
Pro-choice vs pro-life
Economic idea: Competitive framing; which frame “wins”?
Politics: “spin” (Lakoff)


e.g. aren’t we better off w/ Hussein gone?
Liberation vs. occupation
Reference-dependence

Sensations depend on reference points r

E.g. put two hands in separate hot and cold water, then
in one large warm bath


Loss-aversion ≡ v’(x)|+ < v’(x) |

Hot hand feels colder and the cold hand feels hotter
a “kink” at 0; “first-order risk-aversion” aka focussing
illusion?
Requires theory of “mental accounting”



What gains/losses are grouped together?
When are mental accounts closed/opened?
Conjecture: time and space matter, and cognitive
boundaries

Example: Last-race-of-the-day effect (bets switch to longshots
to “break even”, McGlothlin 1956)
Reference-dependence modeling

Where does r come from?




Experiments: Usually status quo or pre-experiment
condition
Koszegi-Rabin ’03: Reference point is based on
(lagged) expectations
Solves problem of why r is not chosen to be superlow
Concept of “personal equilibrium” in which decision
fulfills expectations (multiple equilibria, endowment
effect, Giffen good effects…)
Prospect theory value function:
Note kink at zero and diminishing marginal sensitivity
(concave for x>0, convex for x<0)
Endowment effects (KKT JPE ’90)
KKT “mugs” experiment (JPE ‘90)
Plott-Zeiler review
Data from young (PCC) and old (80 yr olds) using
PZ instructions (Kovalchik et al JEBO in press 04)
“Status quo bias” and defaults in organ
donation (Johnson-Goldstein Sci 03)
Loss-aversion in savings decisions (note few points
with actual utility <0) from Chua & Camerer 03
Actual Utility Vs Optimal Utility
Actual Utility Gains/Losses
50
40
30
20
10
-50
-30
0g
-10
-10
10
-20
-30
-40
-50
Optimal Utility Gains/Losses
Data Points
Jack Knife Regression
30
50
Disposition effects in housing (Genesove and
Mayer, 2001)



Why is housing important?
It's big: Residential real estate $ value is close to stock market
value.
It’s likely that limited rationality persists




Advice market may not correct errors


most people buy houses rarely (don't learn from experience).
Very emotional ("I fell in love with that house").
Like the "big, rare" decisions -- mating, fertility, education, jobhunting…
buyer and seller agents typically paid a fixed % of $ price (Steve Levitt
study shows agents sell their own houses more slowly and get more $).
Claim:

People hate selling their houses at a "loss" from nominal [not inflationadjusted!]
original purchase price.
Boston condo slump in nominal prices
G-M econometric model
Model: Listing price L_ist depends on “hedonic terms” and m*Loss_ist
(m=0 is no disposition effect)
…but *measured* LOSS_ist excludes unobserved quality v_i
…so the error term η_it contains true error and unobserved quality v_i
…causes upward bias in measurement of m
Intuitively: If a house has a great unobserved quality v_i, the purchase
price P^0_is will be too high relative to the regression. The model will
think that somebody who refused to cut their price is being loss-averse
whereas they are really just pricing to capture the unobserved
component of value.
Results: m is significant, smaller for investors (not
owner-occupants; less “attachment”?)
Cab driver “income targeting” (QJE 97)
Cab driver instrumental variables
(IV) showing experience effect
Anchored valuation: Valuations for listening to
poetry framed as labor (top) or leisure (bottom)
(Ariely, Loewenstein, Prelec QJE 03 and working
paperhttp://sds.hss.cmu.edu/faculty/Loewenstein/downloads/Sawyersubmitted.pdf
“Arbitrary” valuations


Stock prices?
Wages (what are different jobs really worth?)





Depends on value to firm (hard to measure)
& “compensating differentials/disutility (hard to
measure)
Exotic new products
Housing (SF Pittsburgh tend to buy “too much
house”; Simonsohn and Loewenstein 03)
Exec comp'n (govt e.g. $150k for senator, vs
CEO's, $38.5 million Britney Spears)
What econ. would happen if valuations are arbitrary?









Perfect competition price=marginal cost…anchoring influences
quantity, not price; expect large Q variations for similar products
Attempts to influence the anchor (QVC home shopping, etc., "for
you just $59.95”).
Advertising!!!
If social comparison/imitation is an anchor, expect geographical,
temporal, social clustering (see this in law & medical practice)
E.g., CEO pay linked to pay of Directors on Board's comp'n
committee. Geographical differences in housing prices,
London,Tokyo, NYC, SF.
Interindustry wage differentials for the same work (Stanford
contracts out janitorial service so it doesn't have to pay as much; cf.
airline security personnel??)
Sports salaries: $100k/yr Miami Dolphins 1972 vs $10million/yr
modern football
Huge rise in CEO comp'n from 1990 (42 times worker wage) to 2000
(531 times); big differentials between US and Europe
Consumers who are most anchorable or influenceable will be most
faddish -- children and toys!!? (McDonald's happy meal etc)
Is too much choice bad?

Jams study (Iyengar-Lepper):



Assignment study:





Short list
Long list
40% stopped, 30% purchased
60% stopped, 3% purchased
74% did the extra credit assignment
60% did the extra credit assignment
Participation in 401(k) goes down 2% for every 10 extra funds
Shoe salesman: Never show more than 3 pairs of shoes…
Medical


6 jams
24 jams
65% of nonpatients said they would want to be in charge of medical
treatment…but only12% of ex-cancer patients said they would
Camerer conjecture: The curse of the composite


Paraphrased personals ad: “I want a man with the good looks of Brad
Pitt, the compassion of Denzel Washington…”
Is there “too much” mate choice in big cities?
Choice-aversion

How to model “too much choice”?




Anticipated regret from making a mistake
“grass is greener”/buyer’s remorse
Direct disutility for too-large choice set (e.g. too complex)
Policy question:




Markets are good at expanding choice…what is a good
institution for limiting choice?
Example: Bottled water in supermarkets
Limit “useless” substitution? What is the right amount?
Pro-govt example: Swedish privatized social security



Offered hundreds of funds
Default fund is low-fee global index (not too popular)
Most popular fund is local tech, down 80% 1st yr
Experimental markets & prob judgment
1. Abstract stimuli vs natural events??
pro: can precisely control information of individuals
can conpute a Bayesian prediction
con: maybe be fundamentally different mechanisms than for concrete events...
2. Do markets eliminate biases?
Yes: specialization














Market is a dollar-weighted average opinion
Uninformed traders follow informed ones
Bankruptcy
No: Short-selling constraints
Confidence (and trade size) uncorrelated with information
Camerer (1987): Experience reduces pricing biases but *increases* allocation
biases
Contingent claims markets:
Markets enforce correct prices..BUT probability judgment influences
allocations and volume of trade (example: Iowa political markets)
Illusions of transparency

“Curse of knowledge”
Difficult to recover coarse partition from fine-grained one
Piaget example: New PhD’s teaching
EA Poe, “telltale heart”
Computer manuals
“ The tapper” study (tapping out songs with a pencil)
Hindsight bias
Recollection of P_t(X) at t+1 biased by whether X occurred
“I should have known!”
“You should have known” (“ignored warning signs”)
--> juries in legal cases (securities cases)
implications for principal-agent relations?

Spotlight effect (Tom Gilovich et al)















Eating/movies alone
Wearing a Barry Manilow t-shirt
 psychology: Shows how much we think others are attending when they’re not