DOWN 11%
An American Life Is Worth $1 Million Less Than It Was 5 Years Ago, EPA
Says  /  Jul. 10, 2008

WASHINGTON (AP) – It’s not just the American dollar that’s losing
value. A government agency has decided that an American life isn’t
worth what it used to be. The “value of a statistical life” is $6.9
million, the Environmental Protection Agency reckoned in May – a drop
of nearly $1 million from just five years ago.

The Associated Press discovered the change after a review of cost-
benefit analyses over more than a dozen years. Though it may seem like
a harmless bureaucratic recalculation, the devaluation has real
consequences. When drawing up regulations, government agencies put a
value on human life and then weigh the costs versus the lifesaving
benefits of a proposed rule. The less a life is worth to the
government, the less the need for a regulation, such as tighter
restrictions on pollution.

Consider, for example, a hypothetical regulation that costs $18
billion to enforce but will prevent 2,500 deaths. At $7.8 million per
person (the old figure), the lifesaving benefits outweigh the costs.
But at $6.9 million per person, the rule costs more than the lives it
saves, so it may not be adopted. Some environmentalists accuse the
Bush administration of changing the value to avoid tougher rules – a
charge the EPA denies. “It appears that they’re cooking the books in
regards to the value of life,” said S. William Becker, executive
director of the National Association of Clean Air Agencies, which
represents state and local air pollution regulators. “Those decisions
are literally a matter of life and death.” Dan Esty, a senior EPA
policy official in the first Bush administration and now director of
the Yale Center for Environmental Law and Policy, said: “It’s hard to
imagine that it has other than a political motivation.”

Agency officials say they were just following what the science told
them. The EPA figure is not based on people’s earning capacity, or
their potential contributions to society, or how much they are loved
and needed by their friends and family – some of the factors used in
insurance claims and wrongful-death lawsuits. Instead, economists
calculate the value based on what people are willing to pay to avoid
certain risks, and on how much extra employers pay their workers to
take on additional risks. Most of the data is drawn from payroll
statistics; some comes from opinion surveys. According to the EPA,
people shouldn’t think of the number as a price tag on a life.

The EPA made the changes in two steps. First, in 2004, the agency cut
the estimated value of a life by 8 percent. Then, in a rule governing
train and boat air pollution this May, the agency took away the normal
adjustment for one year’s inflation. Between the two changes, the
value of a life fell 11 percent, based on today’s dollar. EPA
officials say the adjustment was not significant and was based on
better economic studies. The reduction reflects consumer preferences,
said Al McGartland, director of EPA’s office of policy, economics and
innovation. “It’s our best estimate of what consumers are willing to
pay to reduce similar risks to their own lives,” McGartland said.

But EPA’s cut “doesn’t make sense,” said Vanderbilt University
economist Kip Viscusi. EPA partly based its reduction on his work. “As
people become more affluent, the value of statistical lives go up as
well. It has to.” Viscusi also said no study has shown that Americans
are less willing to pay to reduce risks. At the same time that EPA was
trimming the value of life, the Department of Transportation twice
raised its life value figure. But its number is still lower than the

The environmental agency traditionally has placed the highest value of
life in government and still does, despite efforts by administrations
to bring uniformity to that figure among all agencies. Not all of EPA
uses the reduced value. The agency’s water division never adopted the
change and in 2006 used $.7 million in current dollars. From 1996 to
2003, EPA kept the value of a statistical life generally around $7.8
million to $7.96 million in current dollars, according to reports
analyzed by The AP. In 2004, for a major air pollution rule, the
agency lowered the value to $7.15 million in current dollars.

Just how the EPA came up with that figure is complicated and involves
two dueling analyses. Viscusi wrote one of those big studies, coming
up with a value of $8.8 million in current dollars. The other study
put the number between $2 million and $3.3 million. The co-author of
that study, Laura Taylor of North Carolina State University, said her
figure was lower because it emphasized differences in pay for various
risky jobs, not just risky industries as a whole. EPA took portions of
each study and essentially split the difference – a decision two of
the agency’s advisory boards faulted or questioned. “This sort of
number-crunching is basically numerology,” said Granger Morgan,
chairman of EPA’s Science Advisory Board and an engineering and public
policy professor at Carnegie Mellon University. “This is not a
scientific issue.” Other, similar calculations by the Bush
administration have proved politically explosive. In 2002, the EPA
decided the value of elderly people was 38 percent less than that of
people under 70. After the move became public, the agency reversed

A Price on Your Head
BY W. Kip Viscusi  /  01.07.08

The Environmental Protection Agency created a political firestorm back
in 2003 with an analysis that calculated that the lives of those over
age 70 were worth 37% less than the lives of younger people. Citizen
groups for the elderly were outraged at this “senior death discount”
and ultimately the EPA withdrew the report. Discussion of age
distinctions are off the table now, but the government routinely
places a dollar value on lives saved by regulation.

Although some may consider it immoral to even raise the question of
the dollar value of life, risk regulation agencies can’t avoid doing
so. We would soon exhaust all of our resources if we tried to do
everything that would make our lives safer. A zero pollution, risk-
free society is unattainable. To see why putting a price tag on
expected lives saved makes sense, it is helpful to see where these
numbers come from. The economic value of life is not the total of
one’s lifetime earnings, the taxes we contribute or any other
accounting measure that seems like economics. Rather, the value of
life reflects what people are willing to spend to reduce small risks
of death.

Consider the market for risky jobs. Suppose that on average, workers
face a fatality risk of 1/10,000 of being killed each year and that
they accept this risk in return for an extra $700 in annual wage
compensation. This means that if 10,000 workers faced a similar risk,
on average one worker would die, and so firms would pay a total of $7
million in compensation for the one expected death. The value of a
statistical life is consequently $7 million in this example, and the
number cited generally by most reliable estimates. A considerable
economics literature has documented the extra pay that workers receive
for fatality risks, the lower prices that risky products command and
the lower housing prices for houses in dangerous areas.

This number does not imply that people would accept certain death if
paid $7 million or that they could come up with $7 million to buy out
of certain death. Rather, it captures the rate at which people are
willing to spend money to reduce risk. Most government agencies, in
assessing the cost impact of regulations, use value-of-life numbers on
the order of $5 million to $7 million per expected life saved. But how
much they really spend to save lives is another matter. Sometimes it’s
driven by legislative mandates that do not require risk-cost
tradeoffs. Superfund hazardous waste cleanups, for example, prevent
cases of cancer at a cost of billions of dollars per expected case.

The U.S. Department of Transportation, on the other hand, historically
used wrongful death judgments to value life. It now places a value of
$3 million per life for efforts such as improving airline safety, a
figure that is too low and will produce too little safety regulation.
The question that the EPA was courageous enough to confront in 2003 is
whether all lives should have the same value. Should we value the
lives of the old the same as the young, the rich the same as the poor
and voluntary risk takers the same as those who choose safer

The age difference represents a good starting point for thinking about
such distinctions. The biggest gains in life expectancy generally
result from saving the lives of the young. But going back to first
principles, what matters in valuing life is people’s willingness to
pay to reduce small risks of death. Those values go up as we age,
along with overall spending. The fact that 60-year-olds drive safer
cars and lead safer lives than their children is not a coincidence.
Labor market studies show that workers at age 60 have a higher value
of statistical life than workers at age 20.

For those with very short life expectancies, the value of statistical
life does decline. How much is not known. But to effectively reduce
risks, agencies such as the EPA must grapple with the types of
unpleasant tradeoffs raised in its senior death discount analysis. If
air pollution regulations are expected to increase the life expectancy
of those with advanced respiratory disease by two months, is doing so
really as valuable as adding 70 years of life expectancy by preventing
the deaths of as many children?

Kip Viscusi
email : kip.viscusi [at] vanderbilt [dot] edu

Laura Taylor
email : laura_taylor [at] ncsu [dot] edu

Compensation For Bhopal Set  /  June 22, 1992

“India has finally fixed compensation for the victims of the Bhopal
poison-gas disaster after seven years of legal battles, officials said
today. Babulal Gaur, the Relief and Rehabilitation Minister of Madhya
Pradesh, told reporters that the compensation to the next of kin of
those killed ranged from $3,840 to $11,530. Those injured will be paid
$1,920 to $3,840, Mr. Gaur said.

More than 3,800 people were killed and tens of thousands injured in
the world’s worst industrial disaster when a pesticides plant owned by
an Indian subsidiary of an American company, the Union Carbide
Corporation, spewed deadly mythyl isocyanate gas on Dec. 3, 1984. The
legal battle, which began soon after the disaster, ended with the
Supreme Court of India ordering compensation of $470 million agreed to
by both the Indian Government and Union Carbide in February 1989. The
court upheld the settlement again in October. Mr. Gaur said the claims
for money would be settled by courts set up for the purpose. He said
17 special courts had so far completed hearing 2,000 claims related to
deaths. About 13,000 applications have been filed for compensation in
this category, he said.”

What is a Life Worth?  /  BY Ike Brannon

An unpleasant but necessary job of policymakers is to place a value on
saving a human life. Because society has limited resources that it can
spend on health and safety improvements, it should obtain the greatest
benefit for each dollar spent, and ascertaining an appropriate value
is necessary to that effort. As one would expect, the correct
numerical value to place on a life, typically called the value of a
statistical life, or vsl, is a matter of great controversy. Hundreds
of analyses using widely varying methodologies have been conducted to
determine this value. Despite their differences, most of the studies
center on one basic idea: The vsl should roughly correspond to the
value that people place on their lives in their private decisions.
Though most people may say they will spare no expense to avoid a
possibly fatal risk, their spending patterns dictate otherwise; we do
not all drive armored trucks to work, but instead drive somewhat less
safe — and considerably less expensive — cars. Our willingness to
accept some risk in exchange for a more easily affordable vehicle
suggests there is some limit to how much we will spend to protect our
lives. This article will examine how economists assign a number to the
value of a statistical life, and will consider criticisms of both
their methodologies and the very concept of a vsl.

Economists and other researchers have used a variety of analyses to
determine the value of a statistical life. Below are some of the most
common methods, along with some problems frequently ascribed to them.

Two jobs can differ in any number of ways: One can be in a nicer city,
or it can be in a more pleasant working environment, or it can have
better fringe benefits, or it can offer better opportunities for
advancement than the other job. Or, it can be safer. To estimate the
value of a statistical life, economists can exploit the difference in
pay between two jobs and determine how much of that difference stems
from the difference in the risk of injury or death. Then, the
researchers simply multiply that number by the inverse of the risk
difference and call the result the value of a statistical life.

For example, if I make $40,000 and my twin brother makes $42,000 at a
job that is identical to mine in all respects except for a 1 percent
greater chance of death, then an economist assumes that the $2,000
difference is a premium my twin brother requires to accept the riskier
job. If he requires $2,000 for a 1 percent greater risk, then I can
infer that he is placing a value on his life of $2,000 x(1 ÷0.01), or
$200,000. There are problems with this approach. University of Wyoming
professors Jason Shogren and Tommy Stamland argue that nearly all
revealed preference studies are biased upwards to some degree. They
observe that the wage at a particular job is just enough to entice the
marginal worker. The other workers require less money to accept the
risk. Thus, the “average” vsl is well below the “marginal” vsl obtained
with this method. Another problem is the need to decide the relevant
time period over which fatality rates should be measured when
assessing risk. Should we use the actual death rate for an occupation
over the previous year or the previous five years? Death rates
fluctuate quite a bit from year to year (think about the death rate
for commercial pilots in 2001 as compared to 2000), and this choice
can crucially affect the estimated vsl. Also, do we use the actual
death rates or the workers’ perceived chances of death? After all,
wage premiums are presumably based on perceived risk, not
actual risk, and the two can diverge. Another consideration
is that most occupations do not really carry a risk associated with
work. Should we include those occupations as well in our economy-wide
estimate of a risk premium? And is it all right to assume that we can
merely multiply the risk premium by the inverse of the risk assumed?
Economists who have studied this issue in depth have found that if the
risk doubles, the risk premium does not necessarily double. Alan
Krupnick of Resources for the Future shows that in most instances the
vsl imputed from comparing the difference in wages associated with a
0.1 percent to a 0.6 percent risk would be higher than the vsl imputed
from comparing the wage differences between a 1.1 percent and 1.6
percent risk. This nonlinearity in our valuation of risk reduction may
simply be the result of sorting — those people facing higher risks in
their job do not require the same amount of money to assume an
incrementally higher level of risk. The fact that we see evidence of
the same phenomenon when we calculate a vsl using the contingent
valuation approach (described below) leads some to theorize that it
may be more complicated than mere sorting. Researchers who estimate a
vsl using the revealed preference method have come up with a wide range
of values, from roughly zero (or even negative) to over $100 million.

Economists also estimate the value people place on their lives by just
asking them. Of course, this approach is a little bit more
sophisticated than that because the likely answer to the question,
“How much money would you need to allow us to kill you?” would be an
infinite amount of money. In a contingent valuation estimation of the
value of a statistical life, the economist surveys a number of people
and asks each person the amount of money that he would require to
accept a marginally higher chance of dying in the near future.
Generally, the subject answers yes or no to a series of questions; for
example, the opening question might be, “Would you accept $1,000 to
move from a one in 10,000 chance of death to a five in 10,000 chance
of death?” If the answer is yes, then the next question might be
whether the person would accept $800 to assume the higher risk, and so
on until the person says he would refuse the money for the risk. After
surveying a few hundred people in this manner, the
researcher imputes the implicit value that each subject places on the
value of a life, as is done in the revealed preference method
(multiplying the final dollar figure by the inverse of the additional
risk taken) and averages the valuations. Of course, problems exist in
this approach as well. First, many economists dislike it because of
its subjectivity. All of the questions are hypothetical, so why should
the answers given by the subjects actually reflect the tradeoffs that
they are willing to make? Indeed, a problem endemic to such studies is
the so-called “protest” vote in which someone insists that no amount
of money would entice him to accept a higher risk. If the project
consisted of 100 subjects and one person insisted his life is worth
$100 billion, should it be included in the final average? Should
researchers throw out that observation, or truncate the sample, or use
a median rather than a mean to dampen the rogue subject’s response? On
this matter, there is no consensus other than that the high value
should not remain in the estimate. Critics also question whether
people accurately perceive the actual changes in the small differences
presented to them in the surveys. A majority of people suffer from
innumeracy and have trouble distinguishing a three in 10,000 risk
from a seven in 10,000 risk. For those people (and maybe the rest
of us as well), their answers are little more than guesswork. Should
we include their answers? Would an estimate of vsl be reflective of
society if the mathematically challenged were not included?

A small literature has developed in recent years that infers our
implicit valuation of life from our product choices rather than our
labor-market choices. For example, we know that antilock brakes reduce
the incidence of crashes and death. If we can say for certain that
buying a car with that option reduces the probability of death by one
in 100,000 and the option costs $300, then we can infer that the
person is placing an implicit valuation on his life of at least $300
x100,000, or $30 million. Again, there are many criticisms of this
approach. People purchase thousands of devices that improve safety to
some degree. If the vsl estimated from, say, buying a bicycle helmet is
vastly different than the vsl derived from the decision on whether to
buy antilock brakes, then how can we interpret those numbers? Another
question is whether we separate safety characteristics from other
product attributes. A bicycle helmet that costs $80 and is slightly
safer than a $40 helmet may also be more comfortable, more stylish, or
available at a store closer to the consumer’s house. How are we to
determine the extent to which the buyer’s decision was influenced
by safety considerations? Many of the criticisms of the revealed
preference studies also can be made here. Do consumers accurately
perceive the safety improvements inherent in a purchase? Is it sensible
to compare vsls obtained from different products that have different
levels of risk reduction?

Enough studies have been done that a number of meta-analyses have been
performed on the existing studies in order to find some
“representative” value of a statistical life. Meta-analyses can vary
wildly in sophistication; the basic difference between a meta-analysis
and a simple averaging of a range of studies is that the meta-analysis
attempts to control or adjust for the exogenous factors that could
potentially affect the estimated vsl. For example, from revealed
preference studies we know that the extent of the assumed risk affects
the resultant vsl. A typical worker who assumes a one in 1,000 chance
of death on a job has a lower vslthan an identical worker with a one
in 10,000 chance of death. A sophisticated statistical meta-analysis
can take into account the relative differences in risk assumed in
different studies and “wash out” the effects of those differences on
the final vsl. Meta-analysis may seem like a good tool to establish a
consensus, but in reality it is very difficult to perform well. For
starters, a meta-analysis can only be done on similar studies that
employ the same statistical estimation technique; a
revealed preference study cannot be in a meta-analysis with a
contingent valuation study. In addition, if studies within a
particular method differ greatly in their approaches, it may not be
possible to combine all reputable studies using the same method in a
single meta-analysis.

When estimating the value of a statistical life for regulatory
purposes, economists are most comfortable with calculating a number
that is the by-product of decisions that people make every day that
manifest their willingness to pay for increased safety. Outside of the
realm of regulation, economists often place a value on a life after a tragic
death has resulted in the loss of future income to a household. For
such matters, the procedure of calculating the value of a lost life is
fairly straightforward: The economist calculates the present value of
the future stream of income that would have accrued to the decedent,
adjusted for taxes, consumption, and the cost of living for his
community. This approach may seem straightforward, but it is dependent
on a number of contestable assumptions. For example, what assumptions
should be made about lifetime income growth and retirement age for the
deceased? Is it correct to use population averages or should we
consider certain factors that might have influenced income growth and
retirement age, such as education or the age of children of the
deceased? As anyone who followed the travails of the special
administrator of the government’s official 9/11 survivors fund can
attest, this approach can invite any number of controversies and is
far from providing a value for a life that is free from criticism.

A common critique of the role of vsl in regulatory analysis is that it
fails to distinguish between the life saved of someone young as
opposed to someone close to the end of a life. For instance, many
would argue quite sensibly that a society should be willing to pay
more for a regulation that saves the lives of 10 young children than
for one that saves the lives of 10 senior citizens. There are two
variants of the vsl that make such adjustments: the quality-adjusted
life-year (qaly) and the value of a statistical life-year (vsly).
Both attempt to calculate the value of one additional year of life
saved, with the former adjusting for the quality as well as the amount
of life saved, and the latter adjusting the value of a life-year saved by
discounting life-years saved in the future, as is commonly done in
finance. Both approaches seem intuitively more appealing to many
policymakers than vsl calculations. John Graham, the administrator
of the Office of Information and Regulatory Affairs, has expressed a
preference for using such measures to complement or even replace
the vsl when performing cost-benefit analysis. Both the qaly and the
vsly are fundamentally different than the vsl. The vsl is in essence a
metric derived from decisions made by people either directly in a
survey or observed indirectly in their market choices. Its use in
cost-benefit analysis makes perfect sense. Neither the vsly nor the
qaly are calculated in that way — no one is observing the behavior
of anyone when arriving at this metric. They are applicable only in
the context of cost-effectiveness analysis, where the researcher is
merely trying to rank a number of different policies or treatments.
For instance, if researchers are trying to decide which of a number
of different medical procedures should be done, they may decide
that the hospital feels that only procedures that cost less than
$100,000 per year of life saved, or qaly saved, make sense. Thus,
if a hospital performs a bone marrow transplant that prolongs the
life of a patient by one year, and analysts estimate that the patient
is at 80 percent of his previous life quality for the remaining year,
then they would conclude that 0.8 of a life year was saved. If the
procedure costs less than $80,000, then they would conclude that
it was cost effective under the $100,000 rule (which, incidentally,
is a rule of thumb that quite a few hospitals have been known to
use). In the case of a vsly, let us assume that we have a regulation
that prolongs the life of a young person by two years, on average,
at the end of his life, as might be the case with bans on smoking
inside of restaurants. It is not appropriate to compare that
regulation to a regulation that prolongs the life of a person by one
year today. The vsly requires the regulator to discount the two years
saved 50 years down the road so as to fairly compare it to the life-
year saved today. In this case, using a seven percent discount rate
(to reflect the cost of capital) we would find that the life-years
saved 50 years down the road are only worth 0.07 of a life-year saved
now, just as two $1 bills received 50 and 51 years in the future would
only be worth seven cents to someone today.

While the estimated vsls vary wildly between studies, a broad
consensus is beginning to coalesce around a fairly narrow range of
values, thanks to a number of very influential studies. Economists
Janusz Mrozek and Laura Taylor published a meta-analysis of a large
number of revealed preference studies that was almost universally
praised by researchers in the field for its thoroughness and
inclusiveness. After controlling for all possible factors that could
bias or influence the vsl, they estimated a number between $2 million
and $3 million. More recently, Kip Viscusi of Harvard, in his own meta-
analysis, concluded that the number was closer to $7 million. Viscusi
is one of the leading authorities in the field as the editor of the
Journal of Risk and Uncertainty, the preeminent journal
in risk analysis, as well as the author of numerous books and articles
on risk analysis. Given his stature, this paper has been taken very
seriously by regulators. But Viscusi analyzes occupational deaths not
by occupation but by industry. That distinction is important; grouping
workers by industry essentially treats the risks faced by secretaries
that work for a mining company identical to the risks faced by miners,
a result that obscures the true risk premium received by workers.
There is not necessarily any systematic bias in his analysis as a
result, but the studies he considers generally have higher standard
errors. On the contingent valuation front, Alan Krupnick, Maureen
Cropper, and a number of economists affiliated with Resources for
the Future conducted a series of sophisticated surveys in the United
States, Canada, and Asia that received kudos from other researchers
for sophistication and rigor. The resulting series of papers, which have
just begun to be published, conclude with a number surprisingly
close to the Mrozek and Taylor value, with a range of between $2
million and $3 million. A paper by John Leeth of Bentley University
and John Ruser of the Department of Commerce may prove to be the
last word when it comes to revealed preference studies. They obtained
an incredibly complete and disaggregated data set on death and injury
rates broken down by occupation, as well as a complementary data set
with wage and employment data for the same disaggregated occupations.
Leeth and Ruser estimate a vsl in the range of $2.6 to $4.7 million.

Scholars have spent many years researching and arguing about the
correct approach to determining the value of a statistical life, and
the field is only now beginning to gravitate toward a fairly narrow
range of numbers. But the federal government has been doing cost-
benefit analyses of various regulations for decades and, as a
necessary component of those analyses, has assumed different values in
order to compare costs to benefits. So what values do the feds use?
The Department of Transportation uses a figure of $3 million, which it
left unchanged after a 2002 review of the literature. Transportation
officials cited the Mrozek and Taylor research as a significant
influence of its decision. The Environmental Protection Agency
currently uses a mean value of $6.3 million for its cost-benefit
analysis, with an interval between $1 million and $10 million. While
some degree of flexibility is to be applauded (as I will explain
below), in reality every regulation issued by the epathat spent less
than $8 million to save a life has been approved. The epa commissioned
a large number of studies on the matter a few years ago, in an effort
to establish a reliable, uncontroversial number to use in its
analyses. Unfortunately, that work has led to nothing of the sort.
Having different agencies use different valuations may seem illogical,
but there is a hint of logic in this. Cass Sunstein of the University
of Chicago has argued that people place different values on avoiding
different types of risks — for instance, people fear dying of aids or
in a plane wreck much more than dying in an automobile accident.
Hence, it may make some sense for different authorities to apply
different vsls to different risks.

It is not uncommon for well-meaning people to object strenuously to
placing a value on a human life, judging such a practice to be callous
and demeaning of the value of existence. Is not every life worth an
infinite value to the person living it? Lisa Heinzerling, a Georgetown
University professor and the co-author of a book critical of the use
of vsl, claims that the difficulty in estimating such an amorphous
entity as the value of a statistical life leaves policymakers in the
position of being easily manipulated by the wonks who attempt to
estimate vsls in the first place. It would be much better, she argues,
to have “informed public debate drawing on moral, philosophical, and
societal considerations beyond market-based assessments.” While more
informed debate on regulatory matters might make sense, it is also
necessary to realize that society cannot spend an infinite amount of
money to protect and extend each person’s life, and some choices have
to be made in the realm of health and safety regulation. We have to
decide to what extent we are willing to expend resources to prevent
unnecessary death rather than improve education, increase handicap
access, or ensure a cleaner environment. To resist placing a dollar
value on a statistical life is to abdicate any sense of rational
decision-making in the regulatory realm.