Transcript Slide 1

Behavior Towards Endogenous
Risk in the Laboratory
Glenn W. Harrison,
E. Elisabet Rutström
& Shabori Sen
What is Endogenous Risk?
Risk is “endogenous” when individual has
mitigation or self-protection choices which
alter the risk that they face.
– Prescribed burn to reduce risk of fire
– Construction of earthquake resistant buildings
– Installing hurricane resistant windows
Why is it relevant for
Environmental Policy?
Recognizing that much of the risk that
people face is endogenous helps
informing policy
For example in Florida having hurricane
resistant windows or having a new roof
reduces insurance premium. The state gives
loans and grants for these changes.
Is behavior different when risk is
endogenous?
• Do risk attitudes stay the same?
• Do risk perceptions stay the same?
• Is there “uncertainty aversion” as well as
risk aversion?
What do we find?
• We find no evidence of framing effect on
estimated risk attitudes.
• We find a statistically significant effect
from endogenous risk on estimated
subjective beliefs.
Literature Review
• Endogenous risk: Ehrlich and Becker (1972),
Garen (1988), Shogren and Crocker (1991)
• Testing Endogeneity in the Laboratory: Shogren
and Crocker (1991) (1994)
• Virtual Experiment with Endogenous Risk: Fiore,
Harrison, Hughes and Rutström (2009)
• Betting Mechanism & Jointly Estimating Beliefs
and Risk Attitudes: Andersen, Fountain,
Harrison and Rutström (2009)
Exogenous and Endogenous Risk
in the Expected Utility Model
Inferred WTP instrument in VR
Credit $40 & House worth $18
Summary of Outline of Experiment
1.
2.
3.
4.
Experience fire simulations
Make Bets
State WTP
Make choices for Holt-Laury standard
lottery task
Simulations
• Simulating forest fires
• Computer simulation of Ashley National
Forest in Utah using FARSITE
• Virtual house in the forest
• Two policy options:
– Prescribed burn
– No prescribed burn
• Payoff determined by whether or not
house in the forest is burned by forest fire
• We create 96 simulation scenarios
• 2 fuel loads:
– High – no expansion of prescribed burns has taken place
• The risky lottery
– Low – expansion of prescribed burns has taken place
• Safe lottery
• 2 weather conditions
– Hot – high temperatures and low humidity
– Cool – low temperatures and high humidity
• 2 wind conditions
– High speed – 5 miles per hour
– Low speed – 1 mile per hour
• 2 fuel moisture levels – low and high
• 2 durations until fire is extinguished either by rain or by
fire suppression – 1 day or 2 days
• 3 lightning locations – central plains, north-east
mountains, south-west plains
• 48 scenarios for each of the fuel loads: 2 x 2 x 2 x 3
Probabilities in fire simulations
Static Image CVM. Cabin did not burn.
Static Image CVM. Cabin did burn.
• Show video
Willingness to Pay
Credit = $40 House value = $18
The task as a lottery choice
• Each row in WTP price list presents a choice
between two lotteries:
– A safer lottery (with prescribed burn) with prizes
• High $20-WTP
• Low $20-$8-WTP=$12-WTP
– A riskier lottery (with no prescribed burn) with prizes
• High $20
• Low $20-$8=$12
• Probability of cabin burning is not given to
participant
Betting task: Risk is exogenous
• Two betting tasks:
– House burns in forest fire when fuel is high or
no prescribed burn is done
– House burns in forest fire when fire is low or
prescribed burn is done.
• For each event 9 bookies offer different
odds (multiple price list)
• Subject has $5 to place a bet on each of
the 9 bookies
Betting Task $5 to bet with
Standard Lottery Task
Inferring beliefs
• Controlling for risk attitudes of each
subject
• Joint Maximum Likelihood estimation of
EUT choice model
• CRRA utility function
U=x(1-r)/1-r
• EU for safe and risky lotteries
EUs=ps*U(xLs )+(1-ps)*U(xHs)
Estimated Subjective Probabilities
assuming Homogeneity
Estimated Subjective Probabilities
allowing for Heterogeneity
Conclusion
One cannot simply assume that risk
attitudes and beliefs elicited in an
exogenous risk setting transfers to an
endogenous risk setting.
Why not use field/survey data?
• Estimation or identification problem
• Design Problem
Estimation problem
• Endogeneity problem exists. OLS estimate
of this model will be biased.
• Lower risk is likely to result in lower
mitigation expenditure. Lower mitigation in
turn results in higher risk. This circularity
makes it difficult to identify the effect of a
change in exogenous risk on endogenous
risk and mitigation.
Design problem
The survey design problem arises since it
is difficult to elicit the exogenous or
‘unmitigated risk’ that people face. The
natural response is the endogenous or
‘effective risk’ that people face
Solution:
Laboratory Experiments
Problem:
Lack of contextual and naturalistic cues that
affect decision making in the field
Virtual Experiment
• Fiore, Harrison, Hughes and Rutström
(2009):
A VX is an experiment set in a controlled lablike environment, using typical lab or field
participants, that generates synthetic field
cues using Virtual Reality (VR) technology.
• Provides the control of the laboratory as
well as the naturalistic cues of the field
• Bring some aspects of the field into the lab
• Simulate the environmental setting, or the
stimuli in a naturalistic way
– Visual 3D simulations
Testing for framing effect
• Experience 4 computer simulations of
forest fire with and without prescribed burn
• Form their own “belief” about the
probability of forest fire burning the house
• Two separate instruments are used to
elicit their subjective belief
– WTP instrument where risk is endogenous
– Betting instrument where risk is exogenous
WTP: Risk is endogenous
Risk is endogenous, subjects have the
choice to make an upfront payment for
prescribed burn to reduce the risk of the
house burning in forest fire
Identification
• How can we identify both the perception of the
risk (the subjective probability) and the risk
attitude?
• EU(lottery) = p(burn)*U(burn) +
•
(1-p(burn))*U(not burn)
• U(x) = x(1-r)/(1-r)
• p and r both affect their choices
• We need a separate task to identify risk attitude
• Holt and Laury type lottery choice
Simulating the physics of the fire
• Scientific realism
• Farsite predicts fire spread
• GIS layers
– Vegetation – fuel loads
– Topography
– Weather conditions – fuel moisture,
temperatures, wind direction and speed
– Ignition points
Scientific realism – the use of
Farsite
What is Virtual Reality really?
• Are movies VR?
• Is a game board VR?
• Interactive
– The simulation reacts to the user’s actions
• Immersive
– Stimuli from the simulation dominate those
from outside
HMD – Head Mounted Display
Discussion of VR
• Details, realism and speed of rendering
– Distant objects less detail than proximate
objects
– Objects vs textures
– Photo-realism vs. immersion
– Control interference
– Flat screen monitors, caves, curved screens,
head-mounted displays
Why should frame matter for
perception?
Theory, context and individual
characteristics
• Economic theory specifies no role for
context or for individual characteristics
– Apart from risk attitudes, risk perception
(through probability weighting) (and loss
aversion)
• Experimenters are well aware of the
presence of auxiliary influences on
behavior not captured by theory
• Psychology provide us with some theories
Theories that emphasize the role of
context
• Embodied cognition – cognition extends outside of the
mind and includes the body and the environment
–
–
–
–
Calculators, paper and pen, memory aids
Body movement and concentration
Visual and auditory environmental cues trigger certain heuristics
Heuristics develop through the interaction of a person’s mind
and the environment
• Dual Process Theory of Mind – automatic and
deliberative cognitive responses to environmental cues
– Individual differences in attention implies differences in switches
between automatic and deliberative cognitive responses and
therefore decision errors and biases
– Individual differences in working memory capacity (multitasking
while remembering) affect differences in attention