Portal de Periódicos da CAPES

Hedonic Damages: Were the Body Blows to the Golden Goose Well-Founded?

2008; Volume: 20; Issue: 2 Linguagem: Inglês

10.5085/0898-5510-20.2.137

2374-8753

Ted R. Miller,

Insurance, Mortality, Demography, Risk Management

In the late 1980s, I was convinced that the value of statistical life (VSL) would soon be the basis of courtroom compensation for quality of life loss to injury and possibly even to death. I saw that changeover in compensation rules as a great opportunity that posed grave danger (Miller 1988, 1989). It seemed to offer a way to get off the verdict roulette wheel through scheduling, but without scheduling, it risked raising payouts and reducing predictability. As it turned out, quality of life valuation through a VSL regime (hedonic damages) has not become the norm. Indeed, as Ireland (2000) documented, it suffered heavy courtroom setbacks in the 1990s. It has fared better since, judging from the cases cited by Ireland (2007).This paper looks back at the arguments that kept hedonic damages from taking off and considers their validity. It then returns to my early conceptual papers on the subject, examining how well hedonic damages are being applied and what we have learned in 20 years of regulatory analysis and courtroom application.I add this preface: I am an observer, not a practitioner. I routinely use VSL to estimate societal and individual losses due to death, injury, and illness. Based on those valuations, I do benefit-cost analyses of preventive and palliative interventions. But I still have never testified to hedonic damages or other losses in a wrongful death or personal injury case.To my eye, guided in part by Ireland (2000) and Viscusi (2000), four threads of argument have barred hedonic damages from many courtrooms. The four claims are: (1) VSL is junk science, not mainstream economics. (2) The empirical estimates seem unreliable. (3) Value of life post priori is limitless so a priori VSL is irrelevant. (4) VSL is not the right conceptual basis for compensation; using it would over-insure against risk. Sections II–V of this paper argue that each of these claims is specious. Viscusi (2000) and a wealth of well-known economists would concur with that judgment of the first two claims, yet they appear to be the telling arguments. Courts, and the precedence system they rely on, however, are conservative. They move slowly and resist innovation. So hedonic damages still may reign, especially for nonfatal compensation. To forward that possibility and increase the likelihood of a smooth transition, I may yet choose to testify.Sections VI–VIII of this paper explain why one would use VSL when valuing or compensating nonfatal injury and review the valuation procedure. Sections IX–X evaluate current hedonic valuation practice and suggest future directions in two controversial areas: (1) whether VSL should vary by income, age, and other demographics and (2) what damages are subsumed in the VSL. The two concluding sections ask whether scheduling of hedonic damages is a credible outcome.Surely yes. The conceptual foundations of the method were laid by Jacques Dreze (1962) and Nobel Laureate Thomas Schelling (1968). The literature on the value of a statistical life is quite rich. A spate of theoreticians filled in details in the 1970s and 1980s, with theoretical refinements continuing to be published (e.g., Hammitt, 2007; Johansson, 2002). More than 100 articles have empirically estimated VSL, many of them in the flagship economics journals. The prestigious Journal of Economic Literature published a review of studies in the area 15 years ago (Viscusi, 1993). Empirical articles continue to appear regularly (e.g., Ashenfelter and Greenstone, 2002; Kneisner et al., 2007) as do meta-analytic syntheses.1 The Palgrave Dictionary of Economics has covered the method since the 1980s (Schelling, 1987; Viscusi, 1998, with a 2008 entry by Viscusi in press). Thus VSL is entrenched in the mainstream. Yet, for the reasons listed above, quite a few judges have been convinced that VSL lacks the scientific credibility required to form the basis for expert testimony (Ireland, 2000). Why? Because bright, articulate forensic economists were paid to oppose them. Few were experts on value of statistical life. Nevertheless, they eloquently poked holes, often based on commonsense, incorrect assumptions, and a fast read of a few articles. Judges, trained in the value of adhering to precedent, were only too happy to disallow innovation. Section VII tries to avert similar problems with forensic economic commentary about use of quality-adjusted life year loss estimates in conjunction with VSL estimates to value non-fatal injury.In large measure because the studies vary in quality, the empirical estimates range widely (Miller, 1990; Kneisner et al., 2007). That alarms casual observers. The issue, however, is not "what is the range from an exhaustive list of 100-plus published and unpublished, sound and unsound studies conducted worldwide over almost 40 years?" Rather, it is, "have the extant meta-analyses and systematic reviews independently arrived at recommended value ranges that are reasonably consistent?" The answer is a qualified "yes." Table 1 summarizes the meta-analytic and systematic review estimates. They arrive at a $4.0–$7.8 million range ($4.4–$7.5 million if we include only the six formal meta-analyses).2 That range is looser than one ideally would like but not so wide one cannot use the values credibly. Moreover, the quintessential valuation study for the U.S. (Kneisner et al., 2007) yields a most defensible estimate of $6.1–$6.2 million, toward the center of the meta-analytic range, and a $6.0–$8.1 million credible range.3 Thus, the meta-analyses might recommend estimating quality of life losses using a $5.5 million value with sensitivity analysis at $4.5 and $6.5 million.4Essentially the claim here is that no amount of compensation after the fact can restore someone's life. Under this argument, complying with laws that entitle the estate to compensation for lost quality of life requires infinite compensation. Surely that is not realistic. Tort laws prescribing compensation of the deceased's estate for lost quality of life almost certainly are passed by legislatures that believe these values have a finite upper bound or range. The forensic economist's job is to guide value selection.Liability awards and other forced expenditures divert funds that firms and their workers would have spent on other things including health and safety measures. Research into risk-risk tradeoffs suggests that $20 to $25 million in diverted spending costs some anonymous person their life (Viscusi and Aldy, 2003). Conceptually, that value is a logical ceiling on compensation. Investing more to save a life or awarding more to compensate a death is so economically inefficient that it causes someone else to die (Lutter et al., 1999). Clearly, tort compensation should not create an implied tort. That means the upper bound on compensation for our most precious commodity should not be limitless. Indeed, it should not exceed the risk-risk tradeoff value.Posner and Sunstein (2005) argue it is. They state that the compensation provided, with some distributional differences, is commensurate with compensating individuals for the risks of death that malfeasors impose on them. Moreover, the distributional difference seems an appropriate one in that the compensation goes to the unfortunate person whose risk was realized rather than being shared with those whose risks had no notable consequences.Conversely Kip Viscusi, the acknowledged central figure in the VSL literature, stated in his 1990 article for the first NAFE symposium on hedonic damages that damage awards based on VSL can only be justified for the purposes of creating a deterrence effect. Using the estimates in compensation is not "consistent with the meaning of the estimates or the purposes of compensation in tort cases" (Viscusi, 2000).From the perspective of what testimony on damages an expert currently should offer in a courtroom, the issue, however, is not what it is economically efficient for the tort system to compensate. Rather it is, what does the law prescribe should be compensated?For nonfatal injury or illness and for fatalities in a few states, the law requires compensating lost quality of life. Unfortunately, while state laws and precedents require compensating lost quality of life; they do not prescribe how to compute that loss. So we are left grasping for guidelines.I consider VSL a sensible lower bound. It provides a basis for providing compensation that derives from economic theory and estimates the values that people would have placed on preserving their quality of life prior to the tort. Importantly, it represents the way they price mortality risk when they buy goods like cars and bicycles and when they decide how safely to live their lives.To my mind, compensating people based on the values they placed on their lives pretort is fair compensation. They might like more compensation post priori, but that extra compensation would need to be built into the price of insurance and the price of goods that pose risks to health and safety. Otherwise how could we pay it? And the market only will bear the a priori pricing, i.e., VSL.Viscusi (2000) argues from logic and simple example that even VSL is too high a bound, and that it implicitly amounts to forcing people to buy too much insurance. I do not find his argument convincing, especially when compensating morbidity losses. VSL is the amount of insurance that people routinely buy as they make health and safety choices. More may be too much, but VSL is not! The issue is, what is the value of the plaintiff's loss? If the plaintiff was willing to pay $550 for a 1/10,000 reduction in probability of dying, does that tell us anything about the value (s)he places on her/his life. Since safety is not an inferior good (as defined in micro-economics), surely it must. No, the value of not dying is not simply 10,000 times willingness to pay for a 1/10,000 probability reduction, but the value cannot be lower than that product. Rather, at some point, scarcity of safety should raise safety's unit price. Nevertheless, estimates across empirical studies do not show a strong gradient (Miller, 1990; Viscusi and Aldy, 2003; Bellavance et al., 2006), so a linear approximation may be appropriate.5Thus, VSL provides a lower bound, perhaps even a recommended value from below to go with the risk-risk analysis bound from above. The risk-risk literature and Viscusi's (2000) concerns about over-insurance and market pricing suggest that the lower bound from VSL may be the wisest valuation.Importantly, although Viscusi (2000) asserts the VSL valuation is too high, it is in line with the amounts juries actually award in nonfatal cases where the law does not place a ceiling on non-pecuniary damages (Cohen and Miller, 2003; Smith et al., in press). It is built into current insurance pricing. Thus, Viscusi's claims of over-insurance at VSL are neither derived from theory nor strongly based in an empirical analysis of current awards. They appear to reflect a gut judgment that the values would be too high. For values beyond VSL, however, I agree with them.Implicit in the discussion above is an assumption that VSL is relevant when valuing nonfatal injury and illness. Why not simply value nonfatal injury and illness directly? If people only sued over a few diagnoses, one could do that. But that is not the case.Deriving WTP for injury directly is impractical since we need values that are diagnosis-specific (i.e., for a torn Achilles tendon vs. a fractured patella). Toepan intrusion in a crash crushes and shatters feet and ankles; those injuries cannot be valued fairly as generic lower limb injuries. The International Classification of Diseases lists hundreds of diagnoses for injury alone. Classifications more sensitive to disability consequences expand rather than compress the list. Both regulatory and courtroom use require values by detailed diagnosis.Valuing hundreds of diagnoses precludes regression approaches. First, where would we get risk data for hundreds of different injuries. Even if we could, since injury events deliver force to a body region, someone with a high probability of leg laceration also has a high probability of leg fracture. Picture the multicollinearity problems trying to stuff all those diagnoses into a regression. Even a regression that includes risks of occupational death, permanently disabling injury, and injury with temporary total disability has virtually intractable multicollinearity problems. If we could solve those problems, the assumption that people are sufficiently knowledgeable of their risks to behave rationally still would break down. They may have some idea of their risk of permanent disability, but they cannot even list one hundred injury diagnoses, much less tell you their risk of suffering each one.Similarly, the sheer number of diagnoses and the public's knowledge gaps about individual diagnoses and their consequences preclude a survey approach. Picture the challenge and expense to value 500 diagnoses by survey.If direct valuation of nonfatal health events is intractable, how can we get credible values? I found a practical solution. One uses objective measures to capture the functional losses over time, generically by diagnosis or for a plaintiff. Ideally, one should measure losses of physical, psychological and social/role functions. Based on population surveys or regression methods, one converts the functional losses to losses in quality-adjusted life years (QALYs, a putative measure of utility loss, defined below). One then uses the VSL or survey methods to value a QALY. Earlier papers in this symposium described how VSL is calculated and the theory underlying it. Miller (2000b) explains how VSL can be used in computing damages for nonfatal injury and illness. Here I offer an example and a summary with a few added insights. I also respond to and correct Viscusi's discussion of QALYs in his paper from this symposium.A quality-adjusted life year (QALY) instrument is a systematic tool for replicably valuing functional capacity loss in standardized non-monetary units related to individual utility. It can be used to assess the amount of health-related quality of life that a person loses to a health problem. More specifically, a QALY is a health-outcome measure that assigns a value of 1 to a year of perfect health and 0 to death. Negative values, which represent fates worse than death from a family or victim perspective, may be allowed. QALYs essentially quantify the impacts of injury and illness on functional and role status. Lifetime QALY loss due to a health problem is determined by problem duration and severity. To compute QALY loss, one estimates the fraction of perfect health lost (the QALY loss) during each year that a victim is recovering from the problem or living with a residual disability, then sums the present value of these fractions. Hedonic damages valuation can multiply the percentage of lifetime QALYs lost to an injury or illness times VSL to compute the loss incurred.6 As explained more fully below, before performing this calculation, one must remove from the VSL the value of damages such as earnings and household production loss that will be valued separately.Consider a hypothetical example of a 45-year-old woman injured in an auto crash. This plaintiff suffered a compound fracture of the right leg, a nose fracture, and a concussion. The leg fracture left the woman permanently unable to run and created some difficulty with stair-climbing. The nose fracture healed quickly but left the victim's nose oddly bulged. The concussion created mild cognitive impairment. On the Injury Impairment Index (Miller et al., 1995) these functional losses equate to Cosmetic 4: "readily observable, not amenable to cosmetic cover-up," Mobility 1: "impaired mobility with intact functional ability," Cognitive/psychological 1: "mild inappropriate behavior, increased irritability," Pain 1: "normal function with occasional use of non-narcotic drugs." These losses correspond to annual QALY losses of .1, .13, .05, and .01 (Miller et al., 1995, p. 33). Combining these figures, total QALY loss = 1 – (1 – .1) * (1 – .13) * (1 – .05) * (1 – .01) = .26 QALYs per year. In other words, this plaintiff's quality of life has been reduced by 26%. The economic expert could simply testify to that effect or could multiply times VSL net of work loss, i.e., by the value of lifetime QALYs.For hedonic damage assessment, a key question is which standardized quality of life instrument to use to assess the plaintiff. Once the scale is selected, it is probably best for medical, psychological, and/or rehabilitation experts to do the assessment. The economist then can use a scoring scheme that converts the objective losses into QALY losses.Miller (2000b) catalogued the instruments available in 2000, but at least three others since have joined the pantheon.7 All standardized quality of life instruments are segmented into multiple dimensions of functioning. My views about the criteria to use in selecting an instrument have evolved. They include:The QALY scoring step can be done with the scoring developed for the assessment instrument. In general, however, I no longer recommend that approach. For one thing, many of these instruments were scored imperfectly, using small samples, unrepresentative samples, imperfect survey questions, or rating methods that are not state-of-the-art. For a second, multiple scorings now have been developed for some of the best instruments, either ones from different countries or based on different rating methods. In work for the National Highway Traffic Safety Administration and the Veterans Administration, we found it practical to use meta-analytic and systematic review techniques to develop best scores by scale point within each dimension of loss (e.g., mobility or cognitive) and between the worst health states in the different dimensions.The strength of QALYs lies in replicability and objectivity. In clinical trials and regulatory analyses, their most controversial aspect is the method for converting functional losses to QALY losses, which theory suggests should be probabilistic. Viscusi's paper in this symposium issue states "VSL numbers are grounded in lotteries on life and death, and any usage of these [happiness and QALY] scales should be based on a lottery approach involving personal fatality risks." He goes on to state "No matter how happiness scores and disability scores are used, they fail this test." That statement is true of happiness scales and older disability scales like the one described in Berla et al., (1990) that I severely criticized in Miller (2004). Viscusi describes the latter as "QALYs based on expert judgments of quality-adjusted life years." I do not consider them to be QALY scales at all.The scales Viscusi attacks are not the scales I recommend using! They are not the modern QALY scales endemic in the health outcomes literature. Modern QALY scales do not allow the analyst or a respondent to assign subjective values to their health state. Rather they start by assessing multi-dimensional functional status. The scorer then looks up the score associated with each recorded functional level and applies an equation to compute the multi-dimensional loss. Typically that loss is computed relative to population norms, not to perfect health. Importantly, many, many of the available scorings are probabilistic. They pass Viscusi's test. Reviewing the literature, I found that virtually all scorings of the Health Utility Index, SF-6d, and AQOL are probabilistic, as are 11 scorings of the EQ5d, 15 scorings of the SF-12, and two of the three scorings for the Functional Capacity Index. Viscusi did not look at or critique modern QALY scales and scoring.In forensic and other post-event applications, however, as I observed in Miller (2000b), probabilistic rating actually does not seem like a theoretical requirement. Simple thermometer ratings can be used to compute valid multi-attribute utility loss estimates. More importantly, I recently scored almost 40,000 healthy and disabled veterans using 23 EQ5d scorings. Mean and median QALY levels using weights derived meta-analytically from eight time tradeoff scorings was virtually identical to ones derived from 15 visual analog scorings. So even if a probabilistic method is theoretically superior, if the scoring is sound, using a simpler thermometer scale rating method to derive scores for the functional levels makes negligible difference empirically.A concern raised about the majority of the published QALY scorings is that the disabled adapt to their problems, judging them as less severe than other people.8 For this reason, QALY instrument scorings sometimes are based on interviews with the disabled population. Moreover, scorings from general population surveys incorporate the views of numerous disabled respondents. In 1992, an estimated 19.4% of noninstitutionalized civilians in the United States had a disability. Almost half of these people were considered to have a severe disability (Kraus et al., 1996). If QALY-based hedonic damages became common, it would be desirable to score the observed impairments using only responses from the disabled subset of the population or to use regression to parse out and adjust for the effects of adaptation.The overwhelming majority of the VSLs come from studies of the wages workers receive to take risky jobs or studies of traffic safety behavior. The standard way to derive a value per QALY is to subtract the present value of future earnings, fringe benefits, and household production from VSL (Miller, Calhoun, and Arthur, 1989; Miller, 2000b). One then applies a life table in conjunction with the age and sex distribution of occupational and motor vehicle injury deaths to estimate the number of years of life included in the value of statistical life. The Panel on Cost-Effectiveness in Health and Medicine (Gold et al., 1996) recommends using the same discount rate in this calculation as in computing the present value of future earnings. Dividing the present value number of life years remaining into the value of statistical life yields the cost per QALY.9Viscusi (2000) objects to this subtraction procedure. His argument against it is based on commonsense logic that does not hold up to close scrutiny. Working from economic theory, Landefeld and Seskin (1982) showed mathematically that this approach was acceptable. More importantly, I view the approach as computational adjustment, not valuation. Adding VSL and lost after-tax wages clearly double-counts. VSL measures the combined value to the individual or family of wages, fringe benefits, household production, quality of life, and a host of related attributes. VSL is not a method to estimate the value of wages. But wages are a market good; $1 of after-tax wages is worth $1, no more, no less. Having a job may be worth more, but the wages themselves cannot be worth more at equilibrium. QALY measures often subsume the ability to work including the value of wages, financial security, and the satisfaction derived from work. So if one is computing damages (or presenting costs in a regulatory analysis), to avoid double-counting, one could either simply use VSL or one could subtract lost wages from VSL and compute the impact on each of the two components separately. Because the law has strong precedent for compensating lost wages, separate computations make the most sense. If the computations are separate and ability to work is a QALY dimension, the score for that dimension also must be adjusted downward to remove the value of wages and fringe benefits. That mechanical computation is built into almost all the willingness-to-pay-based injury costs used in Federal regulatory analysis.Monetized QALYs increasingly have been prescribed for use in government regulatory analyses worldwide (McCormick and Shane, 1993; Duvall and Gribbin, 2008—United States Department of Transportation. Ball et al., 1997; Pearce, 2006—United Kingdom, Department of Health and Ageing and en-Health Council, 2003—Australia; Zhang et al., 2005—Canada). Their use in the peer-reviewed literature also has grown. A modest sampling of newer citations (and some important ones missed in Miller (2000b)) includes Abelson (2003a, 2003b), Applied Economics (2003), Becker et al. (2005), Cutler and Kadiyala (2003), Hirth et al. (2000), Jackson et al. (2004), Miller, Levy et al. (2006), Miller, Zaloshnja et al. (in press), Murphy and Topel (2006), Nordhaus (2003), Pearce and Koundouri (2004), Philipson and Jena (2006), Rascati (2006), Tolley et al. (1994), Zaloshnja et al. (2004), and Zarkin et al. (1993).With increasing use, monetized QALYs have been more carefully appraised from a theoretical and empirical viewpoint. Scrutiny has revealed some issues. On methodological grounds, questions have been raised about what QALYs represent (Kenkel, 2001; Krupnick, 2004; Sunstein, 2004). Perhaps more importantly from the viewpoint of application in hedonic damage calculations, serious concerns have arisen about whether the value of a QALY is constant across personal circumstances (Hammitt, 2007; Johnson, 2005; King Jr. et al., 2005).Nevertheless, decomposing VSL into the value of a life year is accepted practice in health economics. Cutler and Kadiyala (2003), a pre-print version of Murphy and Topel (2006), and Philipson and Jena (2006) won Research!America's prestigious Eugene Garfield Economic Impact on Medical and Health Research Award in 2003, 2005, and 2007 respectively, and Murphy and Topel (2006) also received the International Health Economics Association's 2007 Kenneth J. Arrow Award for the best paper research paper in health economics. Those papers all decomposed VSL.The state of the art finally has made demographic tailoring possible. The meta-analyses and a rising stack of articles have analyzed whether VSL varies with income. It unambiguously does. The size of the variation is unclear, with meta-analytic income elasticities across countries ranging from perhaps 0.45 to 1.0 in all but one formulation (Viscusi and Aldy, 2003; Bellavance et al., 2006). Elasticities that derive from analyses of individual data, however, are more a propos to hedonic damages. The one study I know of that derives elasticities from individual U.S. data arrives at similar estimates ranging from 0.6 to 1.0 (Viscusi and Evans, 1990). Persson et al. (1995) estimate that Swedish values vary less sharply, with income elasticity between 0.37 and 0.46. Jones-Lee, Hammerton, and Abbott (1987) estimate the income elasticity in the United Kingdom is between 0.3 and 0.6.VSL also needs to be tailored by age. Sunstein (2004) argues persuasively for doing so. Mounting evidence (Aldy and Viscusi, 2007; Krupnick, 2007; Mount et al., 2001; Posner, 1995; Smith et al., 2004) shows that simply dividing VSL by present value years remaining (the procedure used to derive cost per QALY), then multiplying by the individual's present value lifespan underestimates the values that the elderly place on their own lives unless one uses a discount rate in the 10% range. In part, this result reflects the lower appetite for risk and greater wealth of elderly than younger respondents. In part, it may result from a scarcity effect, with some elderly spending more to preserve the scarce years they have remaining. At the same time, surveys show a subset of elderly respondents are unwilling to spend for their own safety, stating their glory days are done and the money would be better spent on the next generation (Miller and Guria, 1991).VSL estimates for children have been the subject of intensive EPA-funded research recently. A conference proceedings (Abt Associates, 2002, Session 5, plus Blomquist's paper in Session 1; see also Mount et al., 2001) documents much of this work, supplemented by final reports/articles on a subsequent targeted round of grants on the issue.The literature does not suggest any strong VSL differential by gender. Personally, I think tailoring by race would be unsound from a public policy viewpoint, in part because it seems inappropriate to carry over labor market wage discrimination by race into quality of life valuation.In Miller (1989), I suggested that the value of lost quality of life (hedonic damages) should be computed by subtracting lost after-tax earnings (including the value of fringe benefits) and household production from VSL. Relying on a paper by Landefeld and Seskin (1982), I also suggested subtracting a risk aversion factor that related to the value of income security. No one ever adopted the later adjustment, including me. Personally, I moved into a more fully theory-driven valuation framework in which this term did not arise. Conversely, I should have prescribed other subtractions from a forensic perspective. VSL as derived from theory and measured empirically captures the individual's entire uncompensated loss from death, injury, or illness. I should have said to subtract after-tax earnings, household production, AND any other damages valued separately except direct costs such as property damage and medical costs. I am appalled by suits that seek hedonic damages and also seek to compensate loss of consortium, for example. In my opinion, such claims double-count the losses. Indeed, in non-fatal incidents, typical QALY instruments often explicitly account not only for pain but for reduced ability to participate in social interactions.It is critical to realize that VSL is not just an individual value. A spouse rarely takes a hazardous job without conferring with his or her partner. Family members tell each other to wear a raincoat, hand over the keys when drunk, not drive in an ice storm, go to the doctor, wear a safety belt, etc. The risk behavior decisions that yield VSL estimates generally are decisions by a family who live together in a household. Thus, VSL largely accounts for losses by the family, not just by the person killed or injured.Whether to add hedonic damages and claims for pain and suffering prior to death is more ambiguous. The acid test is: how much pain of this type would one expect in a typical motor vehicle crash or occupational death, since the empirical work on VSL is dominated by those two death scenarios. Because, for example, 43% of motor vehicle deaths are not at the scene or dead on arrival at the hospital (according to 2003–2004 Vital Statistics data), substantial pre-death pain and suffering implicitly is factored into the VSL.In Miller (1988, 1989), I suggested that scheduling damages around a VSL framework could reduce the lottery nature of the tort compensation system. Conceptually, I still believe that. Practically speaking, given a tort system that intensively debates every aspect of the value of future earnings, I may have been naive. We are unlikely to move to scheduled damages any time soon, both because courts are slow adopters of new approaches and because of the uncertainty range around VSL across the meta-analyses. Nevertheless, hedonic damage testimony provides jury guidance that should reduce the variance in awards and raise their predictability. That, in turn, will encourage settlements and reduce extreme awards.Bottom line, VSL clearly is good science. Moreover, I still believe it provides useful guidance to juries, especially in non-fatal cases where survivors seek compensation for their quality of life loss resulting from another's torts. The science increasingly allows tailoring to individual claimants, with adjustment for income level critical but age adjustment modest. Guiding awards related to pain, suffering, and lost quality of life with VSL-based hedonic damages probably would not greatly affect mean awards since willingness to award is in the same range as VSL. By providing an anchor, it might reduce extreme awards, which would encourage settlement. The survey reported in Viscusi (2000) that suggested a VSL guide raised punitive awards, however, is worrisome. Still, I think the legal process has wrongly knifed a goose that may yet lay some golden eggs of equity.