The cost of time: haphazard discounting and the undervaluation of regulatory benefits.

• Author: Rowell, Arden

When performing cost-benefit analyses, regulators typically use willingness-to-pay studies to determine how much to spend to avert risks. Because money has a time-value, when a risk is valued is inextricable from how much it is valued. Unfortunately, the studies on which regulators rely are insensitive to this fact: they elicit people's willingness to pay for risk reductions without identifying the time at which the risk reduction will occur. Relying on these time-indeterminate studies has led to a systematic skew in regulatory cost-benefit analysis, toward the undervaluation of risks to human lives. Insofar as cost-benefit analyses inform regulation, this suggests that the current system systematically under regulates against risks to health and safety.

INTRODUCTION I. TIME, MONEY, AND RISK: AN ORIENTATION A. Risk and Money B. Time and Money C. Risk and Time II. MONETIZATION AND TIME-INDETERMINACY: PROBLEMS WITH PRACTICE A. Regulatory Cost-Benefit Analysis in Practice B. One Agency's Practice: EPA 1. Cost-Benefit Analysis at EPA 2. A Review of the Studies C. Policy Implications of Time-Indeterminacy III. TOWARDS A (PARTIAL) SOLUTION A. Improving Time-Determinacy B. Choosing When to Value Preferences IV. OBJECTIONS V. IMPLICATIONS OF TIME-INDETERMINACY IN OTHER CONTEXTS CONCLUSION INTRODUCTION

To perform cost-benefit analyses, regulators often trade off immediate costs with benefits that will not accrue until some future time. How this trade-off is performed has important implications, particularly where--as with climate change, nuclear power, and the preservation of endangered species--most of the benefits of a regulatory action will be enjoyed by the future.

How do regulators value risks to the future? At first glance, much the way they value immediate risks: they rely on studies that measure people's willingness to pay for a risk reduction and assume that those preferences are constant across contexts. If study participants are, on average, willing to pay $80 to ameliorate a risk of 1-in-100,000 of dying from cancer, regulators assume that preventing a single cancer death will justify an expenditure of $8 million. (1)

But an expenditure of $8 million ... when? If the cancer death will not occur until twenty years from now, regulators assume that it will be appropriate to spend $8 million to prevent the death in twenty years. To determine how much to spend today to reduce the future (or "latent") risk, regulators "discount" the value of the risk reduction to modern-day dollars. They do this on the assumption that money has a time-value: a dollar today is worth less than a dollar twenty years from now, because money can be invested and made to grow. The effect of discounting is marked: at a 7% discount rate--a rate currently recommended by the Office of Management and Budget (2)--regulators would be willing to spend only $2 million today to prevent the future cancer death.

This approach is highly controversial. (3) For the purposes of this Article, however, I take current practice on its own terms (4) and make a basic internal point. The point is this: study participants may discount too.

This point will turn out to be very inconvenient for regulators. To see why, consider the example above. The initial study found that, on average, people were willing to spend $80 to ameliorate a cancer mortality risk of 1-in-100,000. Regulators used this figure to calculate both the amount they would spend now to prevent a single cancer death today ($8 million) and the amount they would spend now to prevent a single cancer death in twenty years ($2 million). To avert the risk of one hundred cancer deaths in twenty years, then, regulators would be willing to spend $200 million today.

But if participants in the study gave their initial valuations on the assumption that the cancer death would come (if it did come) not today, but at some time in the future--then these numbers are completely wrong. In that case, the $80 figure was the participants' current valuation of the future risk--which means that it was already discounted by study participants. (5) If study participants made the same assumptions as the regulators--if they assumed that it would be twenty years until they might die from the cancer risk they faced, and if they applied a 7% rate, (6) then the underlying valuation of the harm is not $80, it is $310--the amount you get if you invest $80 at a 7% return over twenty years. This means that, to prevent one immediate death from cancer, regulators should be spending not $8 million, but $31 million. And if $31 million is the correct underlying figure, then regulators should be spending $8 million today per cancer death avoided in twenty years--quadruple the amount they would spend under the current system.

This is a relatively conservative estimate of how much double discounting could affect analyses, because it assumes a relatively short latency period. As the latency period increases, the effect of any double discounting increases as well. If both study participants and regulators assumed a forty-year latency period and applied a 7% discount rate, regulators would spend only 1/15 of what they ought to spend, as measured against what people would actually be willing to spend for the future benefit. And at a one-hundred-year latency, such as could exist for longer term issues like global warming, regulators would be valuing future benefits at less than 1/800 of their actual value. (7)

What does this mean for regulators? It means that, to the extent that any willingness-to-pay study participants have perceived any latency in any of the risks they have valued over the past thirty-odd years--and as I discuss in more detail below, this describes many of the studies that regulators rely upon--regulatory cost-benefit analysis systematically undervalues risk reductions. And where regulators discount future risks, they are rediscounting already-discounted valuations. To the extent that cost-benefit analysis affects regulatory policy, this means that regulatory policy is systematically (albeit inadvertently) underprotective against risks to public health and safety.

So that is the basic concern of this Article: regulatory analyses systematically undervalue the benefits of regulation. Practical readers will note that there is a seemingly straightforward solution to this issue: if this problem arises from time-indeterminate studies, then it can be solved by using better studies, ones which identify the time at which risks are being valued. And indeed, this is my initial recommendation: that regulators use time-explicit monetization studies. Unfortunately, the choice of how to make monetization studies time-explicit turns out to be highly normatively charged, and requires a sophisticated and necessarily controversial account of intertemporal choice. I do not attempt to develop that account here. I merely outline two defensible approaches to valuing future risk, and argue that either would be a significant improvement over current practice. Further recommendations will be addressed in future work.

This Article is primarily concerned with drawing the outline of the problem. Accordingly, it starts in Part I with an orientation into the application of economic cost-benefit analysis to regulatory risks. Readers who are already familiar with debates about discounting and monetization may wish to skip to Part II, where I begin my analysis of how agencies value risks across time. That Part focuses particularly on the studies and methods used by the Environmental Protection Agency. It concludes that current practice undermines much of the value of monetizing risk reductions, because it inadvertently introduces a systemic downward bias into the valuation of the benefits of regulation. Part III outlines a (partial) solution for remedying this bias and flags additional concerns in selecting a methodology for measuring risks across time. Part IV addresses potential objections and Part V identifies some additional implications of time-indeterminacy in other contexts.

1. TIME, MONEY, AND RISK: AN ORIENTATION

   The analysis that I undertake in Part II is based upon a number of assumptions about time, money, and risk.

   This Part seeks to explicate some of these assumptions. It proceeds in three sections. The first looks at the relationship between risk and money, and explains why this Article assumes that monetizing risk is helpful to regulatory policy. The second addresses the relationship between time and money. It describes the time-value of money and explains why future monetary amounts must be discounted before they can be meaningfully compared to present-day monetary amounts. The third identifies how the relationship between risk and time operates with money as an intermediary. It explains why this Article assumes that future monetized risks must be discounted just like money.

   1. Risk and Money

      Economists generally assume that regulatory benefits--such as increased safety against risks to life and health--can be dealt with just like money. (8) But monetization remains controversial, particularly among those who are concerned that money cannot accurately represent the worth of nonmonetary goods. (9) The more we are concerned that money does not capture the entire "worth" of a good, the less helpful we will find economic cost-benefit analyses, as these analyses will portray less and less of the total worth of the goods being compared.

      But even if we think that money is a deeply incomplete measure of some goods, monetized valuations are still useful as informational floors, particularly where money will be spent to secure the good in question. A person considering whether or not to adopt a second child might want to know that she is likely to spend at least $10,930 per year on child-rearing costs. (10) If she chose to adopt the child under those circumstances, an observer might reasonably conclude that, at the time of her decision...