Policy Challenges of the Heterogeneity of the
Value of Statistical Life
W. Kip ViscusiW. Kip Viscusi
University Distinguished ProfessorUniversity Distinguished Professor
Vanderbilt UniversityVanderbilt University
Presentation at CREATE-DHS ConferencePresentation at CREATE-DHS Conference
September 24, 2010September 24, 2010
The Average Value of Statistical Life
Median U.S. value is $7 million ($2000) or Median U.S. value is $7 million ($2000) or $8.7 million ($2009) based on meta analysis $8.7 million ($2009) based on meta analysis in Viscusi and Aldy (2003)in Viscusi and Aldy (2003)
Require $870 to face risk of 1/10,000Require $870 to face risk of 1/10,000 Foreign countries have VSL estimates in Foreign countries have VSL estimates in
expected range, e.g., India is lowerexpected range, e.g., India is lower..
Heterogeneity Based on Risk Level Workers who choose higher risk levels are Workers who choose higher risk levels are
on flatter part of market offer curve. on flatter part of market offer curve.
Heterogeneity Based on Risk Level, cont’d
Early studies found workers with fatality Early studies found workers with fatality risk of 1/10,000 (Viscusi 1978, 1979) had risk of 1/10,000 (Viscusi 1978, 1979) had VSL 5 times greater than study of workers VSL 5 times greater than study of workers facing risk of 1/1,000 facing risk of 1/1,000
(Thaler and Rosen 1975).(Thaler and Rosen 1975). Differences arise from legitimate Differences arise from legitimate
heterogeneity in VSL tradeoffs not failure heterogeneity in VSL tradeoffs not failure of economists to find “the value of life of economists to find “the value of life number.” number.” ..
Heterogeneity Based on Age
VSL will vary with age because length of VSL will vary with age because length of remaining life variesremaining life varies
Imperfect capital marketsImperfect capital markets Life-cycle effectsLife-cycle effects
..
Age and the Labor Market
Series of studies over two decadesSeries of studies over two decades Most recent use age-specific risk dataMost recent use age-specific risk data Result is inverted-U shape patternResult is inverted-U shape pattern Flatter if control for consumption over the Flatter if control for consumption over the
life cycle or cohort effects life cycle or cohort effects VSL tracks lifetime income and consumption VSL tracks lifetime income and consumption
(Kniesner, Viscusi, and Ziliak 2006)(Kniesner, Viscusi, and Ziliak 2006) ..
Cohort-Adjusted and Cross-Section Value of Statistical Life, 1993-2000
0
2
4
6
8
10
18 22 26 30 34 38 42 46 50 54 58 62
VSL (millions 2000$)
Cohort-Adjusted VSL Cross-Section VSL
What Do We Know About Age-VSL?
VSL does not peak at birthVSL does not peak at birth VSL does not plummet as we ageVSL does not plummet as we age VSL for workers around age 60 is VSL for workers around age 60 is higherhigher
than for workers age 20than for workers age 20..
The “Senior Discount” Controversy EPA used a senior discount of 37% in analysis of EPA used a senior discount of 37% in analysis of
Clear Skies initiative in 2002.Clear Skies initiative in 2002. Political firestormPolitical firestorm
Seniors on saleSeniors on sale
37% off37% off
EPA backed off approach. Proposed Senate EPA backed off approach. Proposed Senate legislation banning all demographic adjustments.legislation banning all demographic adjustments.
. .
Value per Year of Life (VSLY) Not a constant, as assumed and estimated in Not a constant, as assumed and estimated in
Moore and Viscusi (1988), which developed Moore and Viscusi (1988), which developed and estimated VSLY formula and rate of and estimated VSLY formula and rate of discount discount
Advent of better data makes possible more Advent of better data makes possible more refined risk measures. Viscusi-Aldy (2007) refined risk measures. Viscusi-Aldy (2007) and Aldy-Viscusi (2008) use industry by and Aldy-Viscusi (2008) use industry by age fatality rate.age fatality rate.
VSLY not constant and not steadily VSLY not constant and not steadily declining with age even though health may declining with age even though health may declinedecline
VSLY rises fairly steadily VSLY rises fairly steadily ..
Value of a Statistical Life-Year Based on Cohort-Adjusted and Cross-Section
Value of Statistical Life, 1993-2000
$0
$50,000
$100,000
$150,000
$200,000
$250,000
$300,000
$350,000
$400,000
$450,000
18 22 26 30 34 38 42 46 50 54 58 62
VSLY (2000$)
Cross-Section VSLYCohort-Adjusted VSLY
Segmented Labor Markets
Workers may face different labor market Workers may face different labor market offer curvesoffer curves
Settle into separate labor market equilibria Settle into separate labor market equilibria (Viscusi and Hersch 2001)(Viscusi and Hersch 2001)
Test: If workers face greater risk levels but Test: If workers face greater risk levels but receive less total wage compensation for receive less total wage compensation for risk, then cannot be on same market offer risk, then cannot be on same market offer curve.curve. ..
The Viscusi-Hersch Hedonic Labor Market Model
Examples of Separate Labor Market Offer Curves
Smokers and Nonsmokers Smokers and Nonsmokers
(Viscusi and Hersch 2001)(Viscusi and Hersch 2001) Black-white VSL differences Black-white VSL differences
(Viscusi 2003)(Viscusi 2003) Mexican immigrants versus other Mexican immigrants versus other
immigrants or native Americans immigrants or native Americans
(Hersch and Viscusi 2010)(Hersch and Viscusi 2010) ..
VSL and Immigrant Status
Fatality Risk* VSL
Estimates Based on the CPS
Native U.S. 4.35 7.95
Mexican immigrants 5.97 Not significant
Estimates Based on the NIS
All immigrants 4.50 9.35
Mexican immigrants 5.70 Not significant
Mexican immigrants who speak English
5.70 3.44
*Fatality rate by industry-immigrant status-age. Risk is annual fatality rate per 100,000 workers.
Income Elasticity Estimates
Meta analysis by Viscusi and Aldy (2003) – Meta analysis by Viscusi and Aldy (2003) – elasticity in 0.51 to 0.61 range for elasticity in 0.51 to 0.61 range for
10 different specifications.10 different specifications. Within sample quantile estimates by Within sample quantile estimates by
Kniesner, Viscusi, and Ziliak (2010) imply Kniesner, Viscusi, and Ziliak (2010) imply mean elasticity across quantiles of 1.44.mean elasticity across quantiles of 1.44.
Meta analyses may suppress some income Meta analyses may suppress some income elasticity, but clearly elasticity is positive.elasticity, but clearly elasticity is positive. . .
Should Income Levels Matter?
VSL increases with incomeVSL increases with income Provide policies poor don’t value?Provide policies poor don’t value? Airline safety – should we regulate it more Airline safety – should we regulate it more
stringently than highway safety? stringently than highway safety? Planes versus guardrailsPlanes versus guardrails
DOT adopted Viscusi-Aldy (2003) elasticity DOT adopted Viscusi-Aldy (2003) elasticity estimate of 0.55.estimate of 0.55.
Rationale is stronger if beneficiaries of safety Rationale is stronger if beneficiaries of safety regulation pay for higher costs of safety.regulation pay for higher costs of safety. ..
Income at Point of Time or Over Time
DOT’s proposed adjustment is very bold DOT’s proposed adjustment is very bold policy initiative to account for within policy initiative to account for within population differences. population differences.
Income changes over time for future Income changes over time for future generations receive greater support.generations receive greater support.
Efficient, but redistributes income from Efficient, but redistributes income from poorer current generation to richer future poorer current generation to richer future generation.generation. ..
Posner–Sunstein Proposal OverviewOverview
Compensation to include hedonic damages plus Compensation to include hedonic damages plus conventional economics damages.conventional economics damages.
Results will be wildly excessive insurance and Results will be wildly excessive insurance and excessive deterrence.excessive deterrence.
Hedonic Loss PitfallsHedonic Loss Pitfalls Note any use provides excessive insurance and Note any use provides excessive insurance and
imposes harmful excessive cost.imposes harmful excessive cost. Compulsory insurance claimants don’t value.Compulsory insurance claimants don’t value. Authors present no formal theory of deterrence of Authors present no formal theory of deterrence of
insurance.insurance. ..
Posner–Sunstein Implementation How Posner–Sunstein Calculate VSLHow Posner–Sunstein Calculate VSL
Method 1 – Use the government number.Method 1 – Use the government number. Method 2 – Determine victim’s VSL – R/q, Method 2 – Determine victim’s VSL – R/q,
where R is compensation for risk q.where R is compensation for risk q. Method 3 – Ask the jury to set a “value of life’s Method 3 – Ask the jury to set a “value of life’s
pleasures lost by the victim.” How?pleasures lost by the victim.” How? Method 4 – How much would the victim pay to Method 4 – How much would the victim pay to
avoid the risk? Hindsight?avoid the risk? Hindsight? Proposal. Judges use government values.Proposal. Judges use government values. Proposal would “have a significant impact on tort Proposal would “have a significant impact on tort
awards, especially for the elderly in non-hedonic awards, especially for the elderly in non-hedonic loss states.” loss states.” ..
Compensatory Damages Bonus
Additional Damages Beyond VSLAdditional Damages Beyond VSL Amount of money needed to compensate for Amount of money needed to compensate for
economic loss.economic loss. Double counting plus excessive deterrence and Double counting plus excessive deterrence and
insurance. insurance. ..
Terrorism Risks
How do people perceive terrorism risks?How do people perceive terrorism risks? How do these risk perceptions compare with How do these risk perceptions compare with
beliefs for other risks?beliefs for other risks? What factors influence risk beliefs, and are What factors influence risk beliefs, and are
those effects reasonable?those effects reasonable? Are people subject to irrational beliefs?Are people subject to irrational beliefs? . .
Terrorism Risk Valuations
How do people value terrorism risks?How do people value terrorism risks? Do people favor continued aid to terrorism Do people favor continued aid to terrorism
victims?victims? Are people willing to trade off civil liberties Are people willing to trade off civil liberties
to reduce terrorism risks?to reduce terrorism risks? How does reducing terrorism deaths How does reducing terrorism deaths
compare to traffic safety and natural compare to traffic safety and natural disasters?disasters? ..
Subjective Level of Risk by Type of Fatality
Auto fatality risk
Natural disaster fatality risk
Terrorism fatality risk
Subjective level of risk Percent Percent Percent
Below-average fatality risk 43.9 58.6 50.6
Average fatality risk 47.6 34.9 41.4
Above-average fatality risk 8.5 6.5 8.0
Ordered Probit Regressions on Subjective Terrorism Fatality Risk
Independent Variables Coefficient (asymptotic std. error)
Age -0.005*(0.003)
Black, non-Hispanic 0.389**(0.139)
Metropolitan residence 0.508**(0.118)
More than 6 plane trips per year 0.545+(0.296)
Ordered Probit Regressions for Government Relief for Terrorism Victims
Independent variables Relief for terrorism victims
Age -0.011**(0.003)
Female 0.198+(0.106)
Black, non-Hispanic 0.570*(0.229)
Years of education -0.041+(0.025)
Republican -0.298**(0.110)
Current smoker 0.326*(0.136)
More than 6 plane trips per year 0.356(0.327)
Above-average terrorism fatality risk -0.643**(0.211)
Below-average terrorism fatality risk -0.408**(0.115)
Percentage Change in Terrorist Risk Estimates after September 11
2002Sample(N = 94)
2003 Sample
(N = 117)
2004Sample
(N = 122)
Risk is now higher 43 54 33
Risk is same as before September 11 17 24 34
Risk is now lower 40 22 33
Determinants of the Probability of Favoring Targeting of
Passengers for Airport Screening
Independent variable Coefficient (std. error)
Waiting time (in min.) 0.0038**(0.0015)
0.0065***(0.0022)
Respondents targeted for screening 0.0190(0.0627)
0.1870(0.1140)
Nonwhite -0.2653***(0.0696)
-0.2655***(0.0697)
Waiting time × Respondent targeted for screening
-0.0052*(0.0030)
Examples of Risk Tradeoff QuestionsSample Terrorism Question:
Suppose you can vote for one of two different policies that cost the same amount but reduce different kinds of risks. Traffic safety policies reduce isolated deaths. The terrorism policy prevents deaths from a single major attack. Which of the two policies would you prefer?
Traffic Safety Terrorism Policy
Type of Deaths Prevented Isolated Accidents Major Terrorism Attack
Average Number of Deaths Prevented
150 50
Which Policy would you prefer?
Policy 1 Policy 2
Relative Risk Valuations after Accounting for Risk Beliefs
Fatality Risk Tradeoff CategoriesImplied Relative
ValuationsAverage Disaster Death Risk/ Average Traffic Death Risk
1.7888
Above-Average Terrorism Death Risk/ Average Traffic Death Risk
0.6794
Average Terrorism Death Risk/ Average Traffic Death Risk
0.9940
Below-Average Terrorism Death Risk/ Average Traffic Death Risk
1.3642
Above-Average Terrorism Death Risk/ Average Disaster Death Risk
0.3798
Average Terrorism Death Risk/ Average Disaster Death Risk
0.5557
Below-Average Terrorism Death Risk/ Average Disaster Death Risk
0.7626
Conclusion
Evidence on heterogeneity of VSL has Evidence on heterogeneity of VSL has increased.increased.
Political sensitivity of recognizing Political sensitivity of recognizing heterogeneity is often great.heterogeneity is often great.
Terrorism risks are a prominent candidate Terrorism risks are a prominent candidate for possible differential treatment because for possible differential treatment because of the bundled nature of what is lost.of the bundled nature of what is lost.
Off the shelf benefits transfer approaches Off the shelf benefits transfer approaches are inappropriate. are inappropriate.
..
Top Related