Behavior Towards Endogenous Risk in the Laboratory Glenn

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 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(x. Ls )+(1 -ps)*U(x. Hs)

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 3 D 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, 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