Calculating Risks? The Spatial and Political Dimensions of Hazardous Waste Policy.

Porter, Richard C.

By James T. Hamilton and W. Kip Viscusi.

Cambridge, MA: MIT Press, 1999. Pp. xi, 326. $37.50.

The Environmental Protection Agency (EPA) perceives its mandate as the protection of every American from every risk that is due to past ill-advised disposal of toxic wastes. Trying to achieve that impossible task, and on a limited budget, leads to compromises that are far from cost-effective. As a result of these compromises, "95% of the expenditures at Superfund sites are devoted to eliminating only 0.5% of the cancer risk" (p. 242). That is the bottom-line message of this important book.

The major EPA error is in its calculations of cancer risk and exposure to that risk. Hence the title. (The subtitle is misleading: There is little about space or politics in the book, and it has nothing to say about policy concerning the current generation and treatment of hazardous waste.) Hamilton and Viscusi look carefully at a sample of 150 Superfund sites and find that the EPA regularly overstates the cancer risks and then remediates the sites without proper concern for the number of people affected by those risks.

Chapters 2 through 5 contain the core of their charge of cost-ineffectiveness. In Chapter 2, they review the EPA approach to risk. The EPA lists all the possible "pathways" (e.g., residential ingestion by drinking polluted groundwater accounts for one-fourth of the pathways) by which people may be affected, now or in the future. The EPA then somehow adds the various pathway cancer risks. The authors do not tell us how, but they do state that in one case the EPA estimate of the total probability that a person contracts cancer from contact with a site is 5.1 (p. 32). These pathways involve future hypothetical populations as well as current residents, and the EPA assumes that people would flock to the area even if it were not remediated. Only 18 of the authors' sample of 150 sites have any current population at the site (p. 50). The authors point out that future risks should at least be discounted, if not prevented by zoning restrictions or toxic containment. They do not consider a third possibility: Postponing remediation until an actual population appears and becomes at risk.

The authors are not building just another dump-on-Superfund case. They note that on many Superfund pathways, the individual lifetime cancer probability is greater than 0.01, and sometimes the probability is extremely high. Under EPA estimates, albeit conservative, 652 cancers could be expected to occur if one California site had not been remediated.

In Chapter 3, the authors look more carefully at the EPA's risk estimates. Not only does the EPA simply add various pathway risks, but it overstates them by making a "reasonable maximum" estimate of exposure duration, then multiplying that by a reasonable maximum estimate of ingestion rate, then multiplying that by a reasonable maximum estimate of contaminant concentration, and then multiplying that by a reasonable maximum estimate of the toxicity of the contaminant (p. 64ff). Recalculating this risk using mean risks, the authors find that pathway risks drop on average by more than an order of magnitude--from about 1*[10.sup.-2] to about 5*[10.sup.-4] (p. 74). The authors note that conservative risk estimates not only overstate the benefits of remediation--critical because the EPA always remediates sites where the individual lifetime risk exceeds [10.sup.-4]--but also may warp the prioritization of risks and hence the order of their remediation.

In Chapter 4, the size of the affected population at Superfund sites is examined. Although the EPA does consider population in deciding whether a site is seriously hazardous (i.e., is placed on the National Priorities List (NPL)), it does not explicitly consider population when remediation decisions are made. The authors remedy this failure by matching site location data to census data (and assuming that the population continues to grow at its 1980-1990 rate). They then reestimate the number of expected cancer deaths--698 on EPA cancer probabilities, as few as 204 on the authors' revised (mean, not reasonable maximum) probabilities. The dearth of cancers suggests to the authors the "potential importance of analyzing [ldots] the risks to cleanup workers and surrounding residents of remediating these sites" (p. 105).

In Chapter 5, the costs of remediation and the benefits of reducing expected cancer cases all come together. The authors estimate the average cost of remediation (excluding the costs of legal and emergency actions) per site in their sample to be $18 million. Cryptically, these costs "reflect a 7% discount rate [ldots] [and] are expressed in 1993 dollars" (p. 288n). That dollar costs of different years need to be adjusted by a price index to become comparable is obvious, but the use of a discount rate is not. Only if the authors were comparing the cost-effectiveness of doing one thing now or something else later would discount rates come into play (and the appropriate discount rate would be a real rate, which for most of us is something much less than 7%). Their estimated remediation costs vary a lot, from an average of less than $4 million per site in the lowest quintile of remediations to more than $40 million per site in the highest quintile.

They then go on to estimating the cost per expected life saved. Using the EPA's conservative risk assessments and not discounting lives, the median cost in their sample is $388 million per life saved--a cost about 50 to 100 times the value of life estimated in the now numerous studies of other markets in which people are compensated for accepting risk of death (many of which have been done by Viscusi). They make the point in several ways, but the basic way is the most telling: By doing the most cost-effective cleanups first, the EPA could "eliminate 99.5% of the expected cancer cases from hazardous wastes with only 5% of the expenditures" (p. 125).

This might have been a good place to stop. The point to here is well and persuasively made, but they go on. Chapter 6 asks why the EPA's cleanup decisions are so cost-ineffective, but Chapters 2-5 have already answered the question. It adds little in this chapter to notice that the true benefits (i.e., the benefits estimated by the authors) do not seem to affect the cleanup decisions and that other (largely political) variables do.

And there is little new or surprising in Chapter 7, dealing with environmental justice. "The minority population percentage is higher than the national percentage of minorities at about a quarter of the Superfund sites, but these are the most populous sites" (p. 159). They go on to add many figures to what we already know. There are two kinds of Superfund sites, those beside the old downtown factories, where the poor and minorities now live, and those way out of town where it seemed safe to dump years ago, where the wealthy and White have now sprawled out to cohabit. It remains hard for rational people to write a scenario that leads to the conclusion that years ago, firms sought out areas to dump in where in the future minorities would live.

Chapter 8 is an ambitious effort to uncover how much people are willing to pay to avoid the risk of living near a Superfund site by examining what happens to house prices there. On the surface, the authors seem to have it all right. After adjusting for structural and neighborhood characteristics, there are lots of sensible, negative Superfund effects on house prices. For example, house prices go down with the cancer riskiness of the site and go up with distance from the site. But it all seems too neat. For example, they use the EPA's Hazard Ranking Score (HRS) as an independent variable. It is highly significant (t of 7) and says that if the HRS goes from 28.5 (the lowest score for an NPL site) to 100 (the highest possible score), ceteris paribus, the house price goes down by about $850. But the authors have mentioned the HRS system only a few times in the first 200 pages of the book, calling it a "rough index system" (p. 8), and they have used the HRS only once before in a regression, where it proved quite insignificant. Indeed, they have pooh-poohed the EPA's measures of risk throughout. Finally, it only takes one year, says the regression that finds a $850 loss of price, for the house price to return to its previous level, even though everyone knows that the EPA is never going to remediate the risks there within a year of a site's placement on the NPL.

The implied willingness to pay (WTP) for risk-free housing is also suspect. The authors find that, ceteris paribus, if an average house (with a value of $74,176) with the average NPL individual lifetime cancer risk (1.81*[10.sup.-6]) is sold before the EPA releases its Remedial Investigation (RI, which is the document in which the EPA first releases its estimates of the cancer dangers at an NPL site), the house price will be lower by $237, but if it is sold after the EPA releases its RI, the price will be lower by only $18 (p. 201, Equation 3, using the parameter estimates of -1779 and +1644). The authors' explanation is that people overestimate the risk before the EPA's RI appears, but once the EPA tells them what the risk really is, people accept the EPA number. Then the authors calculate that the $18, for an average house with average NPL risk and 2.5 people living in it, represents a WTP of about $4 million per expected cancer death avoided (very close to other estimates of the value of life). Does any one read the RIs, though? (I have, and they are always dust-covered in their local repositories.) And does anyone really believe them (especially if they have read anything by Hamilton or Viscusi)?

What's more, they get the WTP of $4 million only through arithmetic error. The authors divide the change in the house price, a present value figure, by the undiscounted lifetime probability of a cancer, something they themselves recognized earlier as a mistake. Doing the estimation correctly means dividing $237 by the present value of expected cancer deaths (if the site is not remediated). This denominator is 2.5 people per household times 2.26*[10.sup.-8] times 100 (using a real discount rate of 1%). (If there is a lifetime cancer probability of 1.81*[10.sup.-6], then the annual cancer probability is about 1/80th of that, or 2.26*[10.sup.-8]. And the present value of a unit-stream console at a discount rate of 1% is 100. One should use roughly 25 instead of the text's 100 if one believes that the hazard will disappear by itself in 30 years. The WTP estimate in the next sentence is then much higher.) That WTP is $43 million, which is very close to the authors' estimate of $51 million, but this figure is base d on an accurate estimate of the true risk, not the excessive initially alarmist" estimate that the authors suggest (p. 203).

My guess is that the issuance of the RI does not change anyone's idea of the real risk, but rather it heralds that a remediation is not far away and that after the typical (excessive) remediation, the site will be restored to a (near) zero-risk state. If that is correct, then the $18 house price fall after the RI measures the WTP when there is, say, only eight more years of cancer risk. Then the WTP equals $18 divided by

(2.5)(2.26 X [10.sup.-8])[1 - [(1.01).sup.-8]/0.01],

which is an estimate of WTP equal to $43 million--almost exactly the same WTP estimate as before. The $43 million is, of course, the result of assuming eight years to remediation, but in any case the implied WTP is high--way too high to add anything with confidence to our value of life estimates.

After these digressions, Chapter 9 returns to the stuff of the first few chapters. How much less would we spend on Superfund remediation if we did it cost-effectively? Lots. Only 3% of the total Superfund expenditures was spent at sites with a cost per expected life saved of less than $5 million per life. Indeed, only 29% was spent at sites with a cost per expected life saved of less than $100 million per life (p. 228). A serendipitous side effect of cost-effective cleanups would be that more minority-dominanted sites would be cleaned up relative to White-dominated sites since actual current (usually minority) populations would then count more relative to hypothetical future (usually White) populations. Cost-effectiveness yields environmental equity--at least more so than reliance on political pressures.

Does all this mean that I feel fifty-fifty about this book?' No way. If there were a Pulitzer prize for important economic policy analysis, this book would be on the short list. It goes beyond the anecdotal to demonstrate how and why we are wasting so much money on Superfund cleanups, and it makes credible estimates of the amount of money we are wasting. As a member of the applied microeconomic policy choir, I really enjoyed this book, and I recommend it highly to other choristers (and their students). (Chapters 1-5 add 128 pages to a coursepack and can be profitably read by senior economics majors.) My biggest hope is that Hamilton and Viscusi are not just preaching to the choir.