The Value of a Statistical Life is Not the Value of Life

This article by Seth Borenstein of the AP generated quite a bit of subsequent commentary:

An American life worth less today, by Seth Borenstein, 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 in today's dollars, the Environmental Protection Agency reckoned in May — a drop of nearly $1 million from just five years ago. ...

Some environmentalists accuse the Bush administration of changing the value to avoid tougher rules — a charge the EPA denies. ...

Agency officials say they were just following what the science told them. ... EPA officials say the adjustment was ... based on better economic studies. ...

As noted below, the rest of the AP article does a decent job of explaining the reasons for the change in the value of a statistical life, but the headline "an American Life Worth Less Today," and the opening paragraph supporting that claim is what most people heard about in the follow-up coverage.

For example, the Colbert Report weighs in here. Most of the commentary that came after the AP article ran along the lines in the Colbert Report video, i.e. that the Bush administration has devalued life in an attempt to avoid costly regulation. I'm not known as a defender of the Bush administration, but I don't think this characterization is fair. Let me turn the microphone over to a colleague.

Trudy Cameron has been researching these issues for the past six years, has served on the Science Advisory Board for the US EPA for almost a decade (until just last year) -- first on the Environmental Economics Advisory Committee, then on the Advisory Council for Clean Air Compliance Analysis (the committee which monitors the EPA's in-house benefit-cost analysis of the Clean Air Act) and on the Executive Committee. She is also the current President of the Association of Environmental and Resource Economists (AERE), the main professional organization in the US for environmental economists, with about 800 members. Thus, she can speak with some authority and, after reading Borenstein article, she decided she would like to set the record straight. Here is a shorter, less technical, newspaper version of her response from an op-ed that appeared today:

‘Value of life’ figures help government measure risk, by Trudy Anne Cameron, Commentary, The Register-Guard: On July 11, The Register-Guard ran a front-page Associated Press article the lead paragraph of which trumpeted that, “A government agency has decided that an American life isn’t worth what it used to be.” The story and its headline could easily give readers the impression that government agencies assign monetary values to human life in an arbitrary, perhaps even amoral, fashion. This is not the case. ...

And here is a longer version with a bit more detail:

On July 11, on the front page, the Register-Guard ran an AP article by Seth Borenstein entitled “In the numbers game of life, we’re cheaper than we used to be.” The reporting on this issue was better than the misleading title, but there are a few points which should be clarified.

The “value of a ‘statistical’ life” is not the same thing as the “worth of a life.”

The article’s opening paragraph trumpets that “A government agency has decided that an American life isn’t worth what it used to be.” Fortunately, these agencies do not merely pick their favorite number out of the air. Instead, the numbers are based on a wide variety of large-sample peer-reviewed statistical studies. These studies measure the tradeoffs that real people actually make, or claim that they would make, when they must decide whether to give up some money for an increased margin of safety. Each study produces a different estimate of people’s willingness to pay to reduce some specific mortality (death) risk. An average willingness to pay is desired, but each study has considered a different-sized risk reduction. It is not correct to average across “apples and oranges.” Thus it is typically assumed that willingness to pay is roughly proportional to the size of the risk reduction, at least over the policy-relevant range of small risk reductions.

Example: Study A might have considered a one-in-a-million reduction in the risk of death (sometimes called a “micromort”) and come up with an average willingness to pay of $5. Study B might have looked at a three-in-a-million risk reduction (i.e. three micromorts) and found an average willingness to pay of $21. If we convert Study B's result to the corresponding value for just one micromort, it works out to a willingness to pay of $7 per micromort. We could then say that the average across the two studies - of willingness to pay $5 in one study and $7 in the other, for the same single micromort - would be $6. If we then wanted to use this set of studies to produce a best guess about what people might be willing to pay for a two-micromort risk reduction, which wasn’t specifically covered by either study, we should use twice this value, or $12.

Unfortunately, instead of scaling all estimates of willingness to pay for risk reductions to the corresponding willingness to pay for a single micromort (that is, a risk change of 0.000001), as was done in the example, somebody decided to use an equally arbitrary benchmark risk change of 1.00 (which means going from a death probability of one to a death probability of zero). The average willingness to pay, scaled up to this huge risk reduction, is called “the value of a statistical life.” This is the number which the U.S. EPA currently reckons to be about $6.9 million. In practice, however, this huge average willingness to pay is promptly scaled right back down again to be matched to the many types of small risk reductions, for each individual, that can typically be achieved by environmental regulations. The EPA number would correspond to $6.90 for one micromort.

The concept of “the value of a statistical life” causes problems because many non-experts are tempted to think about the hugely scaled-up willingness-to-pay amounts, in the intermediate averaging step, in a completely inappropriate way. It is easy to infer that this big number, say $6.9 million, is the amount that “government” thinks we should pay to reduce the probability of some particular person’s death from 1.00 (that is, death with certainty) to 0 (perfect safety). From there, it is just another short leap to think that this somehow means “the value of a life.” This interpretation is simply not correct. None of the careful studies from which the evidence about willingness to pay is garnered have considered anything but small risk changes. And nowadays in the U.S., fortunately, most regulatory questions concern modest reductions in mortality risks that are already very small.

It is certainly easy to provoke moral outrage by suggesting that “the government is putting a dollar value on an American life.” Most of us would agree that every life is priceless. Far in advance of knowing who might possibly die, we certainly would not presume to speculate upon what value to place on any specific person’s life. But there is nothing wrong at all with thinking about people’s willingness to pay for small risk reductions. Almost all of us make regular tradeoffs, at the margin, between money and safety, for ourselves or on behalf of others. We decide whether to put off replacing our tires for another month when money is a bit tight. We decide whether to buy a safer but more expensive car or bicycle helmet, or settle for a cheaper one (either for ourselves or for a spouse or child). We decide whether to vote for or against an initiative to fund bridge repairs through higher taxes. Yes, in a small fraction of cases, somebody will die because of bald tires, or because they drive an economy car or wear a cheap bicycle helmet. And old bridges sometimes collapse. But nobody bats an eye, ahead of time, about the morality of these mundane choices, which are very similar to the choices that must be made with respect to environmental policies. It is simply too expensive to try to make everything perfectly safe.

The article also leaves readers with the impression that there may be something wrong with the EPA using a revised estimate of $6.9 million for one rule and an older estimate of $7.8 million for another rule, especially since the Department of Transportation uses yet another smaller value. While we probably all share the sentiment that our government should “value” everyone equally, it is not the “value of a human being” that is in question in this case. Instead, it is people’s own willingness to pay (that is, their own demands) for small risk reductions, and there is no reason why this willingness to pay should be the same for everyone or for all types of risks.

Mortality risk reductions are just some of the many different things on which people want or need to spend their hard-earned money. Willingness to pay for mortality risk reductions, like willingness to pay for most other goods or services, can be expected to differ with the characteristics of the consumers in question. For example, for most goods, people are willing to pay more if their incomes are higher, so we can expect willingness to pay to change over time as people become more affluent. But we can also expect willingness to pay to differ across people at the same point in time, because current incomes differ across people.

Willingness to pay may also differ substantially across age groups, especially when the health risk in question may not produce any symptoms for decades. For example, an elderly couple may not wish to spend much to reduce (even further) their risk of contracting bladder cancer from trace contaminants in their water supply. It may take thirty years for symptoms of bladder cancer to appear, if they ever do. This particular risk may be a lower priority than other worries, such as paying for prescriptions to deal with their current health-care needs. A young single mother may face the difficult tradeoff between living in a high-rent area with good air quality, and living in an area with poorer air quality but lower rents so that she can afford to put a little money away to help her children attend the local university when they grow up.

We are also willing to pay different amounts to reduce different kinds of risks. For example, people tend to view traffic-related risks as controllable through careful driving, so they are less willing to pay to reduce these risks. Thus it makes sense for the Department of Transportation to use smaller values for mortality risk reductions than those used by the EPA -- many environmental risks are considered to be less controllable. Within the EPA’s domain, it also makes sense that reductions in different types of environmental risks should be valued differently, since people are more concerned about some types of risks than others.

The article also mentions that administrations have tried “to bring uniformity to that figure among all departments.” In a misguided attempt to be equitable, there is a lot of pressure on governments to use a single one-size-fits-all value for willingness to pay to reduce all types of mortality risks in all situations. This might be a good idea if all mortality risks were the same, if everyone in the population was affected to the same extent by every risk, and if risk reductions were a “gift” to the beneficiaries. However, not all risks are perceived the same way, and not all risks threaten everyone in the population equally. Likewise, these risk reductions are not provided free of charge. The beneficiaries are often compelled to pay for these risk reductions themselves, since they are provided via regulations that increase production costs and divert society’s resources from other uses. How equitable is it to for some groups in society to be required to make the necessary sacrifices to pay for more safety than they would optimally choose for themselves?

People sometimes wonder why governments need to place a dollar value on mortality risk reductions at all. Even if a government chooses not to be explicit about the dollar value it uses for the benefits from mortality risk reductions, decisions will still be made about regulations. Some regulations will be implemented and others will not. Greater safety typically means higher costs, either in the form of higher prices for consumers, lower wages for workers, and/or decreased returns for investors. Simply refusing to put an explicit dollar value on mortality risk reductions does not mean that the comparison of the costs of a regulation to its benefits can be avoided. If a regulation is put in place, we can assume that somebody decided that the mortality risk reduction it would achieve was “worth” at least as much as the costs created by the regulation. Governments use explicit values for mortality risk reductions to enhance transparency about how decisions are made.

Most environmental advocates focus on the chance that regulations might be too lax if the value that the government assigns for risk reductions is too small. But it is also a mistake to regulate too stringently, which might happen if the number that is used to value mortality reduction benefits is too large. The agencies in question are doing their best to reflect an appropriate value for the social benefits of risk reductions so that these benefits can be stacked up against the costs of achieving these risk reductions. Industry representatives and environmental advocates certainly have a vested interest in pushing the government to use a smaller or a larger number. We should of course continue to be vigilant about the influence of lobbyists in all types of government policy-making, and we should challenge any decisions which appear to be made in response to the pressures of lobbyists rather than the best empirical evidence about the well-informed preferences of the populace whose collective best interests the government is mandated to serve.