A Pricing Mechanism for Externalities that Incorporates Future Revisions of Estimates of the Costs of Today s Externalities

Transcription

1 A Pricing Mechanism for Externalities that Incorporates Future Revisions of Estimates of the Costs of Today s Externalities T. Nicolaus Tideman Department of Economics, Virginia Polytechnic Institute and State University Blacksburg, VA ntideman@vt.edu Florenz Plassmann * Department of Economics, State University of New York at Binghamton Binghamton, NY fplass@binghamton.edu This version: June 4, 2008 Abstract: The efficiency of mechanisms to control socially costly activities is limited by uncertainty about their social costs. Thus existing regulatory control mechanisms require negotiated compromise about either the prices of activities or the levels to be tolerated. As an alternative to conventional mechanisms, we offer a mechanism in which today s price of an activity is a market-based estimate of future median informed beliefs about the social cost of the activity. This increases the precision and accuracy of estimates of the right price and makes it likely that agents will base their decisions on better estimates of the harm they cause. Journal of Economic Literature Classification Codes: D84, H23, Q51 Keywords: Prediction market, Environmental regulation, Climate change * Corresponding author. Phone: (607) , Fax: (607)

2 1. INTRODUCTION Efficiency requires that every person bear the full marginal cost and receive the full marginal benefit of his actions. When a person bears less than the full marginal cost of his actions, a tax equal to the difference between the social and the private marginal cost of his activity provides this person with an incentive to reduce his activity to the efficient amount. Two common objections to such Pigouvian taxes are first, that it is often difficult to establish the social cost of negative externalities with sufficient accuracy, and second, that such taxes may come to be influenced by goals other than the desire to improve economic efficiency. In this paper we describe a pricing mechanism that reduces the severity of both objections. While it is relatively straightforward to assess the social costs of activities that affect marketable goods directly for example, the social cost of a factory whose smoke emissions harm the business of a nearby laundry it is much more difficult to determine these social costs if there is uncertainty about the extent of the externality and the amount of harm that the externality causes. For example, while there seems to be a consensus that today s emissions of CO 2 have negative externalities, there is much less agreement about the costs of these emissions, partly because of uncertainty about the impact of emissions of CO 2 on climate and partly because of uncertainty about the economic costs of climate change. What is the appropriate Pigouvian tax on CO 2 emissions under these circumstances? We address questions of this sort on the basis of the following five insights. First, if the social cost of an activity is uncertain, then one cannot reasonably require that the activity be undertaken at the level at which its actual marginal social cost equals its marginal benefit. Rather one should discourage activities whose benefits fall short of the best estimate of their 1

3 harm, including the expected harm caused by the fact that the harm is uncertain. Second, while it is straightforward in principle to adjust Pigouvian taxes to account for the cost of uncertainty, such adjustments require that regulators be able to estimate the degree of social risk aversion with sufficient precision and accuracy. Because such estimates are notoriously difficult to obtain, it would be advantageous to assign the social cost of uncertainty to those who cause the uncertainty, in ways that do not require regulators to estimate this cost. Third, while someone needs to bear the cost of uncertainty, not everyone has the same ability to bear that cost. Thus if a regulatory policy assigns the cost of uncertainty to those whose actions cause the negative externality, then it should also permit them to pay others who are better bearers of uncertainty to bear the cost of the uncertainty. Fourth, the degree of uncertainty regarding the social cost of many activities decreases over time as society learns more about the physical, social, and economic effects of these activities. For example, while the social costs of using the insecticide DDT were not well known in the 1940s and 1950s, our understanding of these costs increased as we learned about the persistence of DDT and its danger to human and animal health. By the 1970s, the social cost of any usage of DDT during the 1940s and 1950s could be estimated with much less uncertainty than before. Similarly, recent epidemiological research has much improved our understanding of the social cost of second-hand tobacco smoke, and we can now estimate the social cost of second-hand smoking in the 1970s with much more precision than we could do then. Whenever such increases in understanding reduce the uncertainty regarding the social cost of today s activities, future estimates of the social costs of today s activities will be more accurate than today s estimates. Thus if it were possible to levy today s Pigouvian taxes on 2

4 future rather than today s estimates of the social cost of today s externalities, then we could most likely allocate resources more efficiently today. Finally, and for our purpose most importantly, experience has shown that the best way to predict a statistic is to create a market in which those with superior insight into the process that generates the statistic can make money by betting on the value that the statistic will take. For example, the futures market for wheat provides a better estimate of the future price of wheat than the average of the predictions of those who buy and sell wheat. Similarly, a market in future externality payments can be expected to provide a better estimate of the future perceived costs of externalities than the average of the opinions of those who generate, experience or regulate externalities. We show in this paper that for harmful activities whose quantities can be measured with sufficient accuracy at a non-prohibitive cost, it is possible to charge for their negative externalities according to market-based estimates of future expert opinion of their costs. This is accomplished with a combination of government bonds that those who cause the negative externality in question are required to purchase and whose redemption value is a function of the estimate, at some specified date in the future, of the cost of today s externality, and a prediction market in which these bonds can be traded. The mandatory purchase of such government bonds is equivalent to the mandatory payment of a Pigouvian tax that reflects the best available estimate of the social cost of the externality. The bond s redemption value is lower the higher the current informed estimate of the future estimate of the social harm of today s activity. A positive price of the bond in the prediction market provides a guess of what the estimate of the social cost of today s activity will be at the pre-specified future date. By 3

5 selling the bonds in the prediction market, those who are required to purchase the bonds from the government can transfer the cost of uncertainty regarding the bond s redemption value to those who can bear this uncertainty better. In addition to the direct efficiency gain that comes from using a prediction market to estimate the externality s social harm, the combination of government bonds and the prediction market makes it more difficult than it is with conventional methods to base regulation on goals other than the desire to improve economic efficiency. Because the total value of the obligations to pay for today s externalities will be determined at a time that is many years in the future when the political circumstances may have changed, it is less attractive to use such regulation for short-term political gain. The prediction market makes willful underassessment of the costs of today s externalities visible because the bonds will be valued at zero immediately. By providing an opportunity to sell the bonds, the prediction market also permits those who generate negative externalities to escape willful overestimation of social costs. The remainder of this paper is organized as follows: We describe the details of our pricing mechanism in Section 2 and compare our mechanism with conventional emission taxes and tradable allowances in Section 3. Because our mechanism is particularly well-suited to accommodate the current uncertainty about the social cost of emissions of CO 2, we discuss its application to the regulation of CO 2 emissions in Section 4. We show that for the regulation of CO 2 emissions, our pricing mechanism is neither more difficult nor more costly to implement than conventional corrective taxes and tradable allowances. We conclude in Section 5. 4

6 2. A PRICING MECHANISM FOR NEGATIVE EXTERNALITIES WITH UNCERTAIN SOCIAL COSTS Consider an activity with uncertain social cost, whose quantity can be measured with sufficient accuracy at a non-prohibitive cost for example, the number of cigarettes sold or the amount of particulate matter or CO 2 emitted by a factory. For simplicity, we refer to the agent who undertakes the activity as the emitter and to the quantity of the activity as the level of emissions. Our pricing mechanism consists of two parts. Part one is a requirement that every emitter at time t buy, for each unit emitted, a special zero-coupon government bond with a maturity period of T years, whose face value equals the estimated upper limit of the present value evaluated as of time t + T, of the total harm of one unit of emissions at time t. 1 Denote this estimate by H U t (t), where the subscript indicates the year when the estimate is made and the date in parentheses is the year in which the emission occurred. When the bond matures in year t + T, a prescribed process (see below) will be used to identify the year t + T estimate of the present value, discounted to year t, of the harm of one unit of emissions in year t, (t) H E t+ T. Because the mechanism s main purpose is to take advantage of future reductions in the uncertainty regarding the social cost of today s emissions, T must be sufficiently long say 30 years for such reductions in uncertainty to occur with some reasonable likelihood. The bond s redemption value in year t + T, R t + T, is U E T [( H ( t) H ( t) ) (1 r ), 0] R + = max + +, (1) t T t t T t 1 We address the possibility that this upper limit is infinite in Section 3. 5

7 where r t is the interest rate for conventional T-year zero-coupon government bonds that prevailed in year t. If the redemption value is positive, then an emitter of one unit who buys a bond in year t and holds it for T years will have paid the year t + T estimate of the costs that he caused T years earlier by emitting one unit. If the year t + T estimate of harm exceeds the year U E t estimate of the upper limit (that is, if H ( t) < H ( t) ), then the bond s redemption value is t t+ T zero. It seems appropriate for society to bear the burden of any harm greater than the greatest that could be imagined in year t. The bond s face value must be set at the upper limit of imaginable harm rather than at today s estimate of the most likely cost of a unit of emissions, to ensure that bond holders pay for the harm from emissions if this harm turns out to exceed the harm that was considered most likely in year t. If R + > 0, then emitters receive a refund for the amount they overpaid in year t T t, plus accrued interest. However, the requirement that emitters buy bonds of an amount that almost certainly (by today s expectations) exceeds the future estimate of today s harm leads to three concerns. First, such a requirement may reduce emissions below their optimal level if liquidity constraints force emitters to reduce emissions solely because they cannot afford the required bonds. Second, even emitters who do not face liquidity constraints may prefer not to invest in T-year bonds. Third, uncertainty about the size of the refund makes it more difficult to assess the liabilities of emitters and thereby to determine the market value of emitting companies. We address these concerns by introducing a prediction market in emission bonds as the second part of our pricing mechanism. This prediction market is simply a secondary market for the bonds that ensures that at any point in time, t + τ, the bond s price at that time reflects 6

8 the present value of the bond s expected redemption value at t + T. Let x denote a possible value of the year t + T estimate of the damage of one unit emitted in year t, and let f(x) be the market consensus in year t + τ of the probability density function of these possible values. The bond s price in year t + τ, P t+ τ, is then (1 + r ) P t+ τ = (2) T T U t max[ H t ( t) x,0] f ( x) dx 0 ) ( ) ( τ rt + τ where r t+ τ denotes the interest rate for a conventional zero-coupon (T τ)-year bond in year t + τ. Equation (2) makes no allowance for the possibility that uncertainty in the bond s expected redemption value is costly. This is appropriate if the uncertainty in the redemption value is uncorrelated with the overall uncertainty in financial markets, in which case the cost of the uncertainty can be diversified away. If there is a positive correlation, then the uncertainty in the bond s redemption value will be costly, and markets will subtract the present value of this cost from the price in equation (2). Thus the prediction market ensures that emitters bear the social cost of the uncertainty that their actions have caused, and no separate estimate of the cost of uncertainty is needed. As long as P t +τ > 0, the market predicts the year t + T estimate of the harm of one unit of emissions in year t. The market fails to predict this future estimate if P t +τ = 0. But because emitters will not be held responsible for harm above the upper limit of today s estimates, the bond s price nevertheless permits straightforward assessment of the liabilities of emitters. The prediction market also provides emitters with liquidity, because emitters who do not wish to hold bonds can sell their bonds and thereby recover the present value of the current estimate of their overpayment. 7

9 For the prediction market to operate as intended, the rules according to which the bond s redemption value will be determined in year t + T must be specified in year t. Determining the bond s redemption value requires the joint effort of scientists who are trained to assess the physical consequences of the emissions and economists who are trained to assess the monetary value of these physical consequences. The prediction market will be able to predict the harm from emissions in year t only if market participants have access to the information on which this assessment will be based. Thus a principal requirement for the assessment procedure is that it use fairly easily accessible information and that it be open to public scrutiny. For some types of emissions, a suitable framework for estimating the social cost of these emissions may already be in place (we discuss an example in Section 4). For other types of emissions, the following protocol can serve as a starting point: form a pool of all persons with a rank of associate professor or higher in PhD-granting departments that research the emissions in question. Have them elect (using a pre-specified election method, for example, the single transferable vote) a body of, say, 15 experts to estimate the harm per unit of emissions in year t. Specify that the 15 elected experts will meet (physically or virtually) to discuss the matter. Require that each expert offer a personal estimate after the meeting, and specify that the median of these will be the official estimate upon which the bond redemption values will be determined by a group of economists who are chosen by the same selection method and employ the same method to reach a decision. This protocol can also be used to determine the year t estimate of the upper limit of the harm per unit of emission. 8

10 3. COMPARISON OF OUR PRICING MECHANISM WITH CONVENTIONAL CORRECTIVE TAXES AND TRADABLE ALLOWANCES First, our pricing mechanism is likely to be more efficient than conventional emission taxes and tradable allowances. Conventional emission tax mechanisms require consensus or compromise regarding the appropriate tax rate, and tradable allowances require consensus or compromise regarding the appropriate level of emissions. In contrast, our mechanism employs a market to estimate what the consensus or compromise will be at a time when it is reasonable to expect much less disagreement. The long experience with prediction markets suggest that such markets tend to provide more accurate information about uncertain events than alternative prediction mechanisms. 2 Because the estimate obtained in the prediction market directly affects the liability of bond holders for today s emissions, our our pricing mechanism is likely to achieve a more efficient reduction in emissions than conventional emission tax mechanisms and emission trading markets. Second, emitters ought to bear any cost related to the fact that the social cost of their emissions is uncertain. Although regulators could adjust corrective taxes to account for the cost of uncertainty, they would need an adequate estimate of the degree of social risk aversion to be able to do so. In contrast, our mechanism ensures that emitters bear the cost of the uncertainty that they cause, because the cost of this uncertainty is reflected in the bond s price in the prediction market. This further increases the efficiency of our pricing mechanism in comparison to conventional regulatory tools. 2 See, for example, Rhode and Strumpf (2004) and Wolfers and Zitzewitz (2004). 9

11 Third, conventional corrective taxes and tradable allowance schemes can accommodate improvements in our understanding of the social cost of certain activities through adjustments in either the number of allowances or the tax rate. These adjustments provide those affected by the regulation with an incentive to change their behavior from the point in time at which the adjustments become effective onwards, and they affect earlier behavior only to the extent to which these adjustments have been expected. Our mechanism incorporates improvements in understanding of the social cost through the prediction market that adjusts the expected redemption value of newly issued as well as existing bonds. With respect to newly issued bonds, our pricing mechanism thus affects the behavior of emitters in the same ways as existing regulatory mechanisms do, although the prediction market adjusts the bonds expected redemption values, and thus the cost of emissions, automatically. Furthermore, unlike existing regulatory mechanisms, our pricing mechanism also adjusts the expected redemption values of existing bonds and thereby incorporates the improvements in our understanding about social costs into the eventual liability for past emissions. Thus our pricing mechanism differs significantly from bankable allowances while the value of unused banked allowances changes with adjustments in the emissions cap, these unused allowances only affect the cost of current and future but not past emissions, and they do not ensure that emitters bear the cost of the uncertainty that they cause. Similarly, our prediction market differs significantly from existing futures markets for emission allowances (for example, the European Climate Exchange). Such futures markets permit emitters to hedge their risk with respect to increases in the price of future emissions. In contrast, our prediction 10

12 market permits emitters to hedge their risk with respect to their eventual liability for current and past emissions. Fourth, if there is a consensus that there is a noticeable positive probability that the activity causing the negative externality will lead to a catastrophic event whose discounted costs are infinite, then no finite face value for the bond will be adequate and the activity should be prohibited. Thus with respect to such catastrophic events that are expected to occur with positive probability, our pricing mechanism yields the same result as an infinite corrective tax or an emission allowance of zero. Fifth, our mechanism makes it more difficult to pursue regulations that further purposes other than efficient emissions. A major problem of all regulatory mechanisms that do not employ prediction markets is that it is too easy for regulations to come to be based on political goals rather than on the best estimates of what control parameters are optimal. The fact that the redemption value of the bonds in our mechanism will be determined in year t + T makes it harder to implement environmental regulation in year t that deviates from the best understanding of the harm from emissions in year t. If H U t (t) is set too low, then the bond s price will quickly fall to zero, thereby making the official underassessment of the harm visible. This will make it more difficult to defend underassessing the harm again in the following year when setting H U t ( 1). Conversely, if (t) is set higher than what the market expects to be + 1 t + H U t necessary, then the bond s price in the prediction market will be correspondingly higher, permitting emitters to escape the overly strict regulation by selling their bonds in the market. Although the price of tradable allowances also varies with the number of allowances, markets 11

13 for emission allowances lack a mechanism that indicates when the number of allowances is set too low, and therefore they do not provide a guard against overly strict regulation. There is no guarantee that year (t + T) s estimate of the social cost of emissions in year t will not be affected by undue political considerations. However, the political considerations in year t + T are likely to be different from those that may affect conventional regulatory mechanisms in year t. While corrective taxes and tradable allowances have a direct impact on today s emissions, the estimate that determines the redemption value of our bonds refers to emissions that have already occurred T years earlier; it therefore affects only the ultimate liability for these earlier emissions but not the emission themselves. Because those who were responsible for the emissions at time t may not be among the bondholders at time t + T, it is not meaningful to either punish or reward bondholders for past emissions. Thus the considerations that might affect the estimate of the bond s redemption value in year t + T differ from those that might affect the choice of the tax rate or the allowances in year t. Nevertheless, to the extent that the estimates of the social cost of emissions in earlier years may serve as signals to current emitters about their future liabilities, political considerations similar to those that affect corrective taxes and tradable allowances could still play a role. As long as the assessment of the harm requires human decisions, it is impossible to ensure that the bond s redemption value is determined without any inappropriate political considerations. Thus our mechanism it is likely to reduce though not eliminate the propensity to consider goals other than achieving efficiency in the regulation of externalities. Finally, while it is straightforward in principle to require emitters to pay for their initial allowances, emission trading mechanisms often provide emitters with free allowances. Pricing 12

14 mechanisms make it more difficult to obscure the fact that every unit of emissions can be regarded as the marginal unit that ought to bear the cost of the harm that it causes. However, it may be inappropriate for emitters to bear the entire burden of purchasing our mandatory bonds, because, to the extent that capital investment make it unreasonably costly to cut emissions in the short run, having emitters bear the entire burden assumes that emitters should have anticipated the regulatory policy. Pezzey (2003) has argued that proposed price control mechanisms should exempt some level of emissions because it is politically infeasible to require emitters to pay taxes (which are equivalent to our mandatory bonds) for all of their emissions. For our pricing mechanism, this can be accomplished by giving each emitter bonds for some specified level of emissions for a specified length of time. Because emitters would receive the bonds regardless of their levels of emission, such compensation would be lumpsum compensation and would therefore preserve incentives for emissions to be efficient. 4. APPLICATION OF THE PRICING MECHANISM TO CO 2 EMISSIONS During the past five years, several tradable allowance schemes for CO 2 emissions have been established, and several more are under consideration. 3 An important task for all such initiatives is to determine the optimal number of allowances, which depends on the expected harm from today s CO 2 emissions. However, current estimates of the magnitude of the harm 3 Trading at the Chicago Climate Exchange began in 2003 and the European Union Emissions Trading Scheme (EUETS), which is currently the world s only mandatory carbon trading program, started in Nine Northeastern and Mid-Atlantic states in the United States have signed the Regional Greenhouse Gas Initiative (RGGI), which is expected to become effective in Similar tradable allowance schemes are currently under consideration in California, New Zealand, and Australia. 13

15 from CO 2 emissions vary widely. For example, the 2007 Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) states that 4 Peer-reviewed estimates of the social cost of carbon (net economic costs of damages from climate change aggregated across the globe and discounted to the present) for 2005 have an average value of US$12 per tonne of CO 2, but the range from 100 estimates is large (-$3 to $95/tCO 2 ) It is very likely that globally aggregated figures underestimate the damage costs because they cannot include many non-quantifiable impacts. While the range of -$3 to $95/tCO 2 is very wide, it is about 20 percent smaller than the $5 to $125/tCO 2 range of estimates mentioned in the IPCC s 1995 Second Assessment Report. 5 The fact that the range of estimates has decreased noticeably during the relatively short span of only 12 years suggests that future scientific research may reduce the uncertainty about the social cost of CO 2 emissions even further, making the best estimates of the harm from today s CO 2 emissions much more precise and accurate 30 years from now than they are today. Existing mechanisms to control CO 2 emissions favor tradable allowances that emitters receive free of charge, although some allowance schemes in planning involve auctions. 6 Some others have argued that controlling greenhouse gases through either prices or a combination of price and quantity controls would yield greater social benefits (see, for example, Pezzey, 1992, 4 IPCC (2007, p.69). 5 IPCC (1996, p.51). Regarding the range of estimates, the report emphasizes that [t]his range of estimates does not represent the full range of uncertainty. The estimates are also based on models that remain simplistic and are limited representations of the actual climate processes in being and are based on earlier IPCC scientific reports. The wide range of damage estimates reflects variations in model scenarios, discount rates and other assumptions. It must be emphasized that the social cost estimates have a wide range of uncertainty because of limited knowledge of impacts, uncertain future technological and socio-economic developments, and the possibility of catastrophic events or surprises. (Ibid.) 6 For example, the EUETS and the Chicago Climate Exchange currently provide allowances free of charge to their member states. Germany has began auctioning off 10 percent of its CO 2 allowances in January 2008, and plans to auction off all of its CO 2 allowances beginning in The European Commission has recently proposed that beginning in 2013, 60 percent of the allowances under the EUETS be auctioned off. Similarly, states that have signed the RGGI plan to auction of 100 percent of the emission allowances for CO 2. 14

16 and Pizer, 2002). The various emissions markets that have already been established or are under consideration suggest that it may be politically acceptable to implement regulatory policy regarding CO 2 emissions. Because our pricing mechanism incorporates both a reasonable effort to extract the best possible estimate of the cost of emissions from information available today and the cost of the fact that the cost of emissions is uncertain, it would be sensible to use our pricing mechanism instead of conventional regulatory tools. It is unlikely that our understanding of the harm of CO 2 emissions will improve sufficiently over the next 30 years to permit the development of an uncontested formula that translates emissions into monetary harm. Thus it is necessary to have specific rules about how to combine different assessments. One possible set of rules is those adopted by the IPCC. The IPCC does not undertake its own research, but bases its conclusions regarding climate change on panel evaluation of peer-reviewed, published analyses (see IPCC, 2004). Such a rule would increase the probability that the bond s redemption price would be determined by scientific consensus. The same set of rules could also be used to estimate the upper limit of harm per ton of current CO 2 emissions. 7 Because those who assess the harm from CO 2 emissions may be tempted to base their assessment on undue political considerations, it is important to ensure that the assessment criteria are as objective as possible. The participants in the prediction market will also need access to the information on which this assessment will be based. The procedure adopted by 7 While it would be straightforward to devise more precise rules to determine the bond s redemption value, this would reduce the possibility of incorporating improvements in the process of assessing the harm of CO 2 emissions. On the other hand, the more scope there is for revision of the rules, the greater is the risk of insider trading on the basis of prospective changes in the rules. Fairness to traders requires that whatever process may be used to revise the rules in the future be delimited in the initial contracts. 15

17 the IPCC fits these requirements reasonably well those in charge of determining the harm do not make any assessments on their own but use only freely accessible results of peer-reviewed and published research. A multi-stage approval process that is open to all emitters serves as control mechanism. The fact that the IPCC is able to regularly issue reports that are not summarily rejected by the public suggests that this combination of open sources and multistage control is operational. Because any public process for determining cost-based prices that requires human decisions may be influenced by political considerations, the shortcomings of the procedure used to derive consensus about the harm of emissions in earlier years should not count against our pricing mechanism. The parameters of existing pollution taxes and tradable allowance programs are generally determined by small groups of experts and do not receive much public scrutiny. The recent debate over the Stern Review (Stern, 2006) illustrates the type of problems that arise if a small group of economists that involves independent scientists only as consultants tries to quantify the monetary costs and benefits of emissions and emission reductions. 8 This suggests that the open source framework adopted by the IPCC has a higher likelihood of leading to acceptable results than the traditional methods for setting regulatory goals. For our pricing mechanism to be efficient, the total cost of administering the bonds, implementing the prediction market, and measuring CO 2 emissions must not exceed the expected value of the harm from CO 2 emissions. The IPCC already compiles and weighs most of the information necessary to determine the bonds initial prices and their redemption values, 8 See, for example, Tol and Yohe (2006) and Nordhaus (2007). 16

18 so the marginal cost of setting the bonds prices would be fairly low. Because the sole purpose of the bonds is to generate a certain rate of return for overpaid funds, it is possible to use zerocoupon versions of existing 30-Year US Treasury bonds and add a contract clause that specifies the rules that determine the share of the bond s value that the bond holder will receive. Experience with the Iowa Electronic Markets ( and other prediction markets suggests that such markets can be implemented at low cost. Finally, experience with tradable allowance schemes indicates that it is possible to obtain measurements of CO 2 emissions from large emitters at fairly low cost. Thus our pricing mechanism is likely to be no more expensive and no more difficult to implement than existing tradable allowances. Assuming that the administration of the bonds is about as expensive as the collection of corrective taxes, our mechanism would also be no more expensive than conventional pricing mechanisms. 5. CONCLUSION Our main motivation for developing the pricing mechanism described here arises from the expectation that future scientific research is likely to reduce the uncertainty about the physical, social, and economic consequences of many of today s activities that we believe cause negative externalities. In cases in which this expectation turns out to be overly optimistic and the future assessment is as uncertain as today s assessment, our mechanism will have mainly introduced an additional layer of uncertainty about the price of these activities. If those who are required to purchase bonds immediately sell them in the secondary market, then their payments today for their emissions will not differ from those that they would have to pay under 17

19 conventional emission taxes with rates set to include the cost of the fact that the social costs of the negative externalities are uncertain. Thus the damage that our mechanism could cause if it is adopted in place of conventional regulatory mechanisms seems small. However, the fact that our understanding of the consequences of several harmful activities has increased considerably over time suggests that our mechanism is likely to be superior to conventional mechanisms. For example, global awareness of the likelihood of human induced climate change has increased noticeably over the past 10 years, and the confidence intervals of the IPCC s predictions of the effect of climate change have decreased considerably. Thus it seems appropriate to expect that our understanding of the effect of CO 2 emissions on climate change will continue to improve. Our mechanism assures that someone will voluntarily pay the future estimate of the cost of all current CO 2 emissions, up to today s maximum estimate of the harm. Furthermore, it assures that the price that today s emitters take into account when deciding how much to emit is the best available estimate of the combination of the expected cost of the emission and the cost of the uncertainty about that cost. For our pricing mechanism to be attractive, the bonds future redemption price must be determined primarily by scientific rather than political considerations. Open-source procedures like the one adopted by the IPCC provide some control against political pressure, but it is impossible to ensure that political influence will be completely eliminated. As long as the participants in the prediction market believe that the bond s redemption price will be more closely based on scientific considerations than a price today would be, our mechanism provides better protection than conventional regulatory taxes and allowances against undue political influence today. There can be no guarantee that the total payment for today s emissions will be 18

20 based exclusively on the best future estimate of the emissions harm. But if citizens find it impossible to put minimal trust in the regulators who value and administer the bonds, then it is unlikely that they will find any regulatory policy acceptable. 19

21 REFERENCES: IPCC IPCC Second Assessment: Climate Change Contribution of Working Groups I, II and III to the Second Assessment Report of the Intergovernmental Panel on Climate Change. IPCC, Geneva, Switzerland. IPCC The Preparation of IPCC Reports. IPCC Secretariat World Meteorological Organization Building. IPCC, Geneva, Switzerland. IPCC Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.). IPCC, Geneva, Switzerland. Nordhaus, William D A Review of the Stern Review on the Economics of Climate, Journal of Economic Literature 45:3, Pezzey, John C. V The Symmetry between Controlling Pollution by Price and Controlling It by Quantity. The Canadian Journal of Economics 25: 4, Pezzey, John C. V Emission Taxes and Tradeable Permits: A Comparison of Views on Long-Run Efficiency. Environmental and Resource Economics 26: Pizer, William A Combining Price and Quantity Controls to Mitigate Global Climate Change Journal of Public Economics 85:3, Rhode, Paul and Kolman Strumpf Historical Presidential Betting Markets. Journal of Economic Perspectives 18:2, Stern, Nicholas The Economics of Climate Change. The Stern Review. Cambridge University Press. Tol, Gary W. and Richard S.J. Yohe A Review of the Stern Review. World Economics 7:4, Wolfers, Justin and Eric Zitzewitz Prediction Markets. Journal of Economic Perspectives 18:2,