The Intergenerational Ethics of Climate Change: The Failure of Cost-benefit Analysis as a Normative Framework

Transcription

1 The Intergenerational Ethics of Climate Change: The Failure of Cost-benefit Analysis as a Normative Framework By Nathan R. Lee B.S. Materials Science & Engineering University of Pennsylvania, 2010 SUBMITTED TO THE ENGINEERING SYSTEMS DIVISION IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN TECHNOLOGY AND POLICY AT THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY SEPTEMBER MIT. All rights reserved. The author hereby grants to MIT permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part in any medium now known or hereafter created. Signature of Author: Engineering Systems Division August 20th, 2014 Certified by: Lucas Stanczyk Professor of Political Science Thesis Supervisor Accepted by: Dava Newman Professor of Aeronautics and Astronautics and Engineering Systems Division Director, Technology and Policy Program

2 Abstract Climate change generates a conflict between generations: while it is in the interest of the current generation to continue to exploit inexpensive carbon-based fuel to drive economic growth, it is in the interest of future generations that we reduce our carbon emissions by making these fuels more expensive. This raises the following question: what moral framework should we use to adjudicate between the interests of different generations? In this work, I argue that the commonly used framework of cost-benefit analysis the analytic framework for public policy that developed out of the field of welfare economics fails as a normative framework for intergenerational policy. For one, by aggregating costs and benefits across all generations, it ignores that what matters is each generation. For another, by reducing all value into a unitary objective function, it ignores important distinctions between different categories of moral claims. Third, by attempting to optimize a function across all time, it reflects a false sense of knowledge about the distant future. For all these reasons and more I conclude that this approach cannot offer a reasonable normative framework for intergenerational public policy. In its stead, I propose an intergenerational threshold principle which avoids aggregating generations together, gives space for different categories of value, and, I will argue, is more robust to the epistemic limitations of intergenerational policy analysis.

3 Acknowledgements First and foremost, I want to thank Lucas Stanczyk. A graduate student s fate rests in the hands of his adviser, and if only every graduate student could be as lucky as I have been. He both challenged me to pursue excellence all the while exhibiting infinite patience for my repeated missteps. At once the overseer of my intellectual journey and a companion in it, a cheerleader against despair and a drill sergeant against laziness, he performed the conflicting duties of a graduate adviser with remarkable poise. Most of all, Lucas showed me what it means to engage in genuine analytic philosophy, and the incredible insights that a steadfast commitment to this arduous task can yield. As an aspiring academic in the field, he has been a true role model. Second, I want to thank Amanda Graham, the MIT Energy Initiative, and the Bechtel Foundation for their support this past year. Without Amanda, none of this could have happened. Besides being instrumental in facilitating the project, she offered a much-needed voice of support for the basic idea behind all of this, namely, that philosophical analysis can be practically useful for public policy and that, indeed, it is a necessary component of developing effective responses to today s most pressing energy and environmental challenges. For this opportunity and for her unflinching support, I will forever be grateful. Finally, I wish to thank my friends Alex Breckel, Michael Craig, Zak Accuardi, Tal Levy, and Dylan Boynton. I admire them for who they are as individuals and for their striking intellects, each unique in its own way. Their continual willingness to engage in political-philosophical debate whether at a coffee shop in Cambridge, on a hike in New Hampshire, in our kitchen making breakfast, or sitting on the air-conditioning vent outside our shared office space in MITEI has provided the intellectual training ground on which I have developed as a scholar. What s more, they ve made the whole thing a lot of fun.

4 Content I. Introduction... 1 A. Overview... 1 B. The Meaning of Cost-benefit Analysis... 3 II. The Failure of Intergenerational Cost-benefit Analysis... 4 A. An Illustration of the Problem... 4 B. The Origins of Cost-benefit Analysis... 8 i. Welfare as Willingness to Pay (WTP)... 8 ii. The Modified Pareto Criterion C. The Social Discount Rate D. Approaches to Calculating SDR i. The Social Rate of Time Preference (SRTP) Approach ii. Social Opportunity Cost of Capital (SOCC) Approach iii. The Ramsey Approach E. Discounting in the Intergenerational Context i. Applying the Social Rate of Time Preference Approach ii. Applying the Social Opportunity Cost of Capital Approach iii. Applying the Ramsey Approach iv. The Tactical Justification for Intergenerational Discounting F. The Breakdown of Intergenerational Cost-benefit Analysis III. Why Does Intergenerational Cost-benefit Analysis Fail? A. An Operational Critique of Intergenerational Cost-Benefit Analysis i. A Critique of Market-based Revealed Preference ii. A Critique of Substitutability iii. An Epistemic Critique of Intergenerational Cost-benefit Analysis B. A Moral Critique of Intergenerational Cost-Benefit Analysis i. The Problem with Aggregative Maximization ii. The Problem with Welfare... 38

5 IV. What is the Alternative to Intergenerational Cost-Benefit Analysis? A. Precautionary Principle? i. Precautionary Principle as General Risk Aversion ii. Precautionary Principle as Aversion to Particular Risks iii. Precautionary Principle as an Epistemic Critique B. A Proposal: An Intergenerational Threshold Principle i. An Overview of the Principle ii. Defense of the Principle C. Reconciling the Proposal with Cost-Benefit Analysis V. Conclusion Bibliography Appendix 1: Can we Owe Future Generations Anything? Appendix 2: Why do Individuals Discount?... 56

6 I. Introduction A. Overview The planet s average surface temperature has already increased approximately 1.2 F in the past century, and the International Panel on Climate Change (IPCC) predicts that, absent new mitigation efforts, it will rise another 3 to 7 F over the next century. While the magnitude of the effects remain unclear, they will almost certainly include sea level rise, more extreme weather (stronger and more frequent hurricanes, droughts, and floods), and acidification of ocean water. Besides the direct fatalities such events will cause, these changes will also put pressure on food and water supplies, reinforce geopolitical conflicts, and accelerate the spread of disease. While some of these trends have already begun, the most severe impacts of climate change will be felt by future generations (Emmanuel, 2012). While future generations are most threatened by climate change, it is the current generation which is in the best position to prevent it. And given that preventing will require sacrifice substituting away from carbon-based fuels will entail substantial costs, despite remarkable advances in low-carbon energy sources in recent years 1 we have a misalignment of generational self-interest. As such, the question of the appropriate mitigation policy is, in a large part a moral one: how should we weigh the interests of our own generation against the interests of those that will come in the future? This question, in turn, raises deeper questions. Can we reasonably say the current generation owes it to future generations to prevent climate change? Does the current generation not have more urgent issues to deal with, like the immense number of people living around the 1 The global economy remains highly dependent on carbon-intensive energy. For example, eighty seven percent of global energy consumed in 2013 was from fossil fuel (BP, 2014). 1

7 world today in desperately poor conditions? Doesn t the fact that the fossil-fuel based expansion of electric grids in countries like China and India is helping to lift millions of individuals out of poverty trump the harm that doing so might cause future generations? On the other hand, does the mere fact that the victims of our actions do not yet exist negate our moral responsibility to them? Why should future generations suffer for something which they have no control over when it could be prevented at moderate cost by those who do? 2 The answers to these questions ultimately depend on the basic standard we use to address questions of intergenerational ethics. To navigate this complex web of competing intergenerational moral claims, we need a clear framework for the normative analysis of intergenerational issues. To justify our answers that such an analysis would produce, this framework must be built on a defensible standard for intergenerational ethics more broadly. It is the development of such a standard or principle for intergenerational ethics that is the principal motivation of this thesis. My thesis is twofold. The negative component of my thesis is that cost-benefit analysis the standard analytic framework for policy dilemmas in energy and environmental policy and its underlying moral standard of aggregate welfare maximization fail to provide a reasonable basis for resolving intergenerational policy dilemmas like the appropriate climate mitigation strategy. The positive component of my thesis is to argue that a reasonable framework for normative intergenerational analysis must be anchored by a non-maximizing, non-welfarist intergenerational threshold principle. The argument is organized as follows. In the first part, I will show how cost-benefit analysis breaks down in the intergenerational context, using climate change as the canonical case. In the second part, I will show why cost-benefit analysis breaks down, exploring the 2 The current generation could respond by saying that this unfairly hoists all the responsibility on the current generation when, in fact, the responsibility is shared by several past generations as well. 2

8 question from both an operational standpoint as well as a moral-philosophical one. In the third and final part, I consider alternative approaches to cost-benefit analysis, proposing and defending the general shape of an intergenerational threshold principle which, I argue, avoids the features that doom intergenerational cost-benefit analysis. B. The Meaning of Cost-benefit Analysis But first, it is important that I define clearly what it is I am attacking. I have set out to argue that cost-benefit analysis fails as a normative framework for intergenerational public policy. One might reasonably ask: how could anyone object to the notion that a government should consider both the costs and benefits of its options before making a decision? If this is all cost-benefit analysis meant, I would heartedly endorse it. After all, this is equivalent to saying perform a comprehensive analysis. But in political discourse, cost-benefit analysis means much more than this. It generally refers to the analytic approach that developed out of welfare economics for analyzing public policy. Underlying this approach is the normative principle of aggregate welfare maximization, as well as a particular set of analytic approaches to identifying, measuring, and aggregating welfare (we will examine all this shortly). This is all to say that the term cost-benefit analysis in this paper presupposes a particular moral-operational framework, and it is this framework I have in mind as I develop my critique. Perhaps then I ought to capitalize it Cost-Benefit Analysis every time, but I believe this would actually be misleading. Given the soaring influence the field of economics continues to 3

9 have on contemporary political discourse, the welfare-economics interpretation of these words has become all but synonymous with the words themselves. 3 II. The Failure of Intergenerational Cost-benefit Analysis A. An Illustration of the Problem Let us say that, by an international agreement, all countries have signed a binding agreement to enact a global tax on carbon for ten years. How much should the tax be? The standard cost-benefit analysis approach says the optimal carbon tax is that which maximizes the benefits net costs of a policy, normalized for time. In this case, the benefits are the monetized value of the damage that would have occurred from climate change absent the imposition of the tax. This should include the averted damage to both direct economic activity (typically approximated in terms of annual GDP) as well as some monetized equivalent for environmental goods not internalized in the market. The costs are the reduction in economic activity that results from the higher prices of goods imposed by the tax (again typically approximated by GDP). Finally, to sum all these values up, we normalize for the fact that they are occurring at different points in time. To accomplish this, the standard framework discounts future value at some time-compounding rate to convert all values into present value. In the case of climate change, most of these benefits will be heavily discounted because they will occur in the distant future; the costs, on the other hand, will occur in the present. Once all future costs and benefits 3 Speaking of the soaring influence of economics, let me be clear about the fight that I am picking, and the fight that I am not (since economics-bashing has become an academic cottage industry in itself). I am not seeking to criticize the role of economics in shaping public policy analysis, per se, but rather the particular analytic lens that is cost-benefit analysis and the particular application that is long-term public policy issues spanning many generations. 4

10 have been converted into their present values, we can finally sum up all the benefits and subtract the sum of all the costs. Doing this for an array of carbon tax values, we should then choose the value which yields the greatest normalized net benefits. But why do we normalize at all? Why not simply sum up all benefits and costs irrespective of the time in which they occur? It turns out that doing this leads to absurd conclusions in many cases, including climate change. To illustrate, let s take an ultra-simplified example. Let s suppose we have three policy choices: $10, $100, and $1000 per-ton-of-co 2 emitted. Furthermore, let us suppose that (1) the benefits and costs can be measured purely in terms of GDP, (2) we can develop good estimates of the effects this tax will have on GDP over time. In particular, relative to some counterfactual business-as-usual baseline GDP trajectory, the economic costs and benefits of each policy will be: 4 Global Annual Global Annual Global Annual Global Net Value Carbon Tax GDP Impact GDP Impact GDP Impact from Time- Policy (Years 1-10) (Years 11-50) (Years ) neutral CBA $10/ton 1% 0 +1% +140% $100/ton 5% 0 +3% +400% $1,000/ton 50% 0 +5% +250% 4 We will further assume that all numbers are adjusted for inflation and that the economy after two centuries will have the same trajectory regardless of policy choice today. 5

11 While these projects are purely illustrative, they are at least plausible. A $10/ton carbon tax is on the lower end of the range of commonly proposed carbon tax values, and it would cause very limited disruption to the global economy. The $100/ton tax is on the upper end of this range, and while it would not derail the global economy, it would entail significant adjustments costs. The $1000/ton value is intentionally selected as being implausibly severe it is an order of magnitude outside the reasonable range of carbon tax proposals. As a short cut to seeing how devastating this would be for the global economy, we might reasonably approximate this as being a de facto ban on the use of fossil fuel. Forcing a global economy which continues to rely on fossil fuels for nearly 90% of its energy needs to shift entirely to carbon-free energy sources would clearly cause massive economic damage. So, without discounting, which tax does our CBA tell us to choose? If we simply sum up the costs and benefits without any special accounting for time, the optimal choice is the $100/ton policy. However, notice that all policies are net positive even the one which would cut the economy in half for an entire decade! This conclusion should give us pause, as it seems to condone what would amount to a crushing blow to the current generation for the sake of (relatively) moderate economic gain in the future. This implication is a consequence of the fact that the benefits, albeit moderate in magnitude, persist for 150 years, whereas the costs last for ten years. As a result, the sum of the benefits swamps the sum of the costs, even if the per-year cost is greater. Indeed, if we were to remove the bound on the time horizon and consider an infinite stream of benefits which is perhaps a more realistic case for climate change anyway then any amount of sacrifice on the part of the current generation would be justified according to cost-benefit analysis so long as some annual benefit existed, no matter how small. This illustrates the overwhelming power of an 6

12 infinite series in aggregative analyses: that any amount of continued benefit in the future will justify even the most extreme suffering on the part of the current generation is often referred to as the tyranny of the future problem. 5 To address this, welfare economists introduce a correction factor into cost-benefit analysis that reduces the weight of the future relative to the present. In particular, each value is converted into its present value equivalent by being exponentially discounted the further out in time it occurs: ( ) Later we will closely examine how economists justify such a move, but, for now, I will simply illustrate what happens to the results. For example, using a common value of r = 5%, we get: Global Carbon Tax Policy Net Value (Undiscounted) Net Value (Discounted 5%) $10/ton +140% 5% $100/ton +400% 35% $1,000/ton +250% 380% 5 Viewed from a standpoint of mathematics, this result is not surprising: just like a line to a point or a set to an element, an infinite series of values will always overwhelm a finite one. 7

13 The difference is dramatic. We essentially have a complete reversal: whereas in the timeneutral case even the draconian $1000/ton policy generates net benefits, with a 5% discount rate no policy is worth endorsement, even the one which would only require a ten-year sacrifice of 1% of our GDP for a century and a half of equivalent economic gain. Unlike before where the cost-benefit analysis makes the present generation a prisoner to future generations, here the costbenefit analysis indicates that the present generation need not even lift a finger (even if a little bit of effort could go a long way). What conclusion can we draw from this analysis? The conclusions are twofold. First, given the extreme sensitivity of the analysis to the choice of the social discount rate, if we are to apply cost-benefit analysis intergenerationally, we must choose this parameter very carefully. 6 Second, and more importantly, that our basic moral intuitions about what a reasonable approach to cost-benefit analysis would be (i.e., temporal neutrality) turn out to generate completely unreasonable results suggests that there are some non-intuitive questions lurking in the background which need to be addressed before applying the approach. To get a good grasp of these questions, we must begin with the origins of cost-benefit analysis itself. B. The Origins of Cost-benefit Analysis i. Welfare as Willingness to Pay (WTP) In this paper, cost-benefit analysis is understood to be the analytic application of the principle of welfare maximization, which says that the optimal policy is that which maximizes the sum of welfare across all people. What exactly does welfare mean? In Classical 6 However, we cannot conclude that intergenerational cost-benefit analysis categorically fails to produce reasonable results after all, there do exist certain values between r=0% an r=5% that generate reasonable results. For example, at a discount rate of 2.5%, we would reject the $100/ton and $1000/ton carbon policies but accept the $10/ton policy. 8

14 Utilitarianism, welfare was understood as pleasure, for which philosophers such as Jeremy Bentham and John Stuart Mill provided objective criteria (Bentham, 1907; Mill, 1906). Such a definition enables one to compare amounts of welfare across individuals. However, in the 20 th century, critics like Lionel Robbins argued that these definitions of welfare were flawed attempts to impose objectivity on a concept which is inherently subjective. Welfare is a mental state or disposition that can be meaningfully understood only within an individual. As Robbins famously said, Every mind is inscrutable to every other mind, and therefore no common denominator of feeling is possible. This line of reasoning has strongly influenced contemporary welfare economics, which defines welfare as the satisfaction of one s own preferences (Robbins, 1932). But to make the above definition practically useful, economists must have some way to actually measure preference satisfaction. Economists do this through the notion of willingness to pay (WTP), which is premised on the idea that the relative amount of money one is willing to spend on something is a direct indication of how much one values or prefers that thing. For conventional market goods, measuring WTP is a relatively straightforward affair, since WTP can be inferred from the market price. For public goods (e.g., environmental protection) or goods with imperfect or distorted markets (e.g., healthcare), however, economists must estimate WTP indirectly, employing a variety of sophisticated techniques (indeed, this is a huge part of modernday empirical economics). While economists tend to prefer methods that gather information based on actual market behavior, one common technique for environmental goods is to actually ask people what they would be willing to pay under various conditions. This method is known as contingent valuation (Samuelson, 1992). 9

15 ii. The Modified Pareto Criterion While the subjective preference satisfaction definition does not necessarily rule out interpersonal comparisons of welfare, 7 contemporary welfare economics tends avoid this contentious issue by focusing normative analysis on the pursuit of the Pareto criterion, namely, that a policy should be undertaken if and only if it would make at least one person better without making anyone else worse off (as defined by the individuals themselves). Such a policy is known as a Pareto improvement, and the point at which no alternative policies could be enacted that would make at least one person better off without making someone else worse off is known as Pareto efficiency (Samuelson, 1992). Although the Pareto criterion is theoretically attractive, it turns out to be too strict for practical use. After all, it s hard to imagine any public policy ever being enacted that wouldn t make at least one person worse off. As such, strict compliance with the Pareto criterion would paralyze the public policy process. In recognition of this, welfare economists have developed a slightly relaxed version of the criterion known variously as the modified Pareto criterion (MPC), the potential Pareto Criterion, or the Kaldor-Hicks criterion, which says that a policy should be undertaken if it could generate a Pareto improvement (i.e., if those who are made worse off by the policy could be compensated by those who are made better off). While this is may seem like a subtle distinction, it turns out to be of monumental importance: given that welfare is defined in monetary units through WTP and assuming we take welfare expressed in monetary terms as functionally equivalent to actual money then for any policy that increases aggregate WTP, there will also be a hypothetical set of monetary transfers by which those made worse off by the policy could be fully compensated such that it satisfies the modified Pareto criterion. In other 7 See Ken Arrow and Amartya Sen s work on Social Choice Theory. 10

16 words, given the above premises, any policy which increases total monetary value irrespective of its actual distributional effects is to be endorsed (Samuelson, 1992). C. The Social Discount Rate Let us now turn to the question of how to compare costs and benefits across time and, in particular, to what extent we should discount for time. This issue turns out to be the single most contentious issue in cost-benefit analysis when applied to climate change and it is at the core of my critique of multi-generational cost-benefit analysis so I will examine this issue in great detail. Before going any further, let me define what discounting means from a high-level operational standpoint: discounting is a way of converting a future value into its present-value equivalent. In particular, ( ) where SDR is the social discount rate. In the standard application of cost-benefit analysis to public policy, all costs and benefits that occur in the future are converted to present values using some pre-specified SDR. It is the sum of these present values that we aim to maximize. So that is how to discount. But why do we discount? There are several reasons for why we might engage in something that looks like discounting. Broadly speaking, we can distinguish two classes of discounting: pure discounting, which is discounting for time itself, and instrumental discounting, which is discounting for contingent reasons that happen to coincide with time. There are six common explanations for discounting in standard cost-benefit analysis: 11

17 1. Diminishing Predictive Capability (Instrumental) We should discount future costs and benefits because our confidence in predicting these values falls exponentially with time. 2. Possibility of Human Extinction (Instrumental) Given that there is always a chance that some catastrophic event (e.g., asteroid collision, volcanic eruption, runaway climate change) will destroy humanity, we should prioritize the present over the future. 3. Future will be Wealthier (Instrumental) The global economy will continue to grow in the long-term, so people will have more money in the future. Given the diminishing marginal utility of wealth, we should discount to the extent that future people will be wealthier. 4. Proper Accounting for Economic Productivity (Instrumental) Due to the marginal productivity of capital, a unit of wealth invested today yields more than one unit in the future. As such, monetary values in the future are equivalent to the monetary values in the present that would have to be invested to generate them. We should therefore discount at the relevant market rate of return. 5. Maximize Return on Investment (Instrumental) To ensure we maximize total wealth over time, we should allocate scarce resources where they will achieve their highest return. Discounting future costs and benefits at the relevant market rate of return will ensure that only policies which generate future returns as high as the equivalent resources could in the market will pass a cost-benefit analysis test. This is also known as the opportunity costs argument. 12

18 6. Discounting is Democratic (Pure) People prefer wealth in the present relative to the future. In a representative democracy, public policy should reflect people s preferences. The discount rate reflects this preference. Each of these reasons is distinct and can have very different implications for the rest of the analysis and yet they lead to functionally equivalent behavior. Moreover, we will see in the following section that different combinations of these explanations are behind the different procedures economists use to calculate the SDR. Consequently, it is often very difficult to tease out the relevance of any particular explanation for a given cost-benefit analysis. This will turn out to be of great importance because I will argue that pure justifications for discounting are impermissible in the intergenerational context, while instrumental justifications are not. The difficulty in distinguishing between these justifications and when they apply is what makes the topic of intergenerational discounting so notoriously difficult. D. Approaches to Calculating SDR There are three general approaches to calculating the SDR: the social rate of time preference (SRTP), the social opportunity cost of capital (SOCC), and the Ramsey equation (RE). 8 Let us look at all three, assuming a standard time horizon (a single generation). We will subsequently look at how these approaches might be modified for longer time horizons that include many generations. 8 There is also a fourth approach known as the shadow price of capital. This is a hybrid of the first two approaches, but discussing its components will be sufficient for our purposes. 13

19 i. The Social Rate of Time Preference (SRTP) Approach The social rate of time preference approach takes the social discount rate to be a kind of collective preference parameter on how to trade off consumption goods across time. This approach tends to be justified by appeal to its democratic pedigree, i.e., that the government s rate of time preference should reflect the preferences of its constituents. If we make all the required assumptions to link preferences to the market preferences can be understood as preferences for consuming goods, individuals market behavior is in accordance with these preferences, etc. we can then employ the market as a direct source of information about people s preferences. For example, economists use the market for government-backed securities such as U.S. Treasury bonds to infer how people discount future consumption. If one assumes that government guarantee effectively eliminates risk, this allows economists to isolate the effect of time from the effect of risk. The implied discount rate from such markets tends to be around 3% (EPA, 2010). ii. Social Opportunity Cost of Capital (SOCC) Approach The social opportunity cost of capital approach effectively looks at the government as an investor in society and asks what the minimum rate of return is needed to justify the investment. The argument goes as follows: the resources the government spends are resources that otherwise would have been spent in the private economy, since government spending is financed by taxes on private citizens and businesses. Consequently, if we wish to maximize welfare and are prepared to make the necessary assumptions to convert welfare into monetary value, we should accept only public policies that can achieve at least as a high a return as they could have in the 14

20 private economy. To ensure this, we should discount the impacts of the policy by the relevant market rate of return. One common empirical approach to calculating the SOCC is to use the stock market as a kind of economy-wide average rate of return on capital. There is some dispute about whether to use corporate debt or corporate equity, as it depends on what kind of background uncertainty one thinks the risk-free SDR should incorporate. But, merely by way of example, the average annual real rate of return on equity is around 7% (EPA, 2010). iii. The Ramsey Approach The Ramsey Approach is based on the Ramsey equation, which was developed by Frank Ramsey as part of his optimal growth theory. The Ramsey Approach provides a means of deriving the discount rate rather than inferring it. In doing so, it allows us to disaggregate and compare the relative contributions of the different factors that drive discounting. The equation is as follows: The first term, ρ, is the component of the discount rate that is known as pure discounting, that is, discounting for time itself. 9 The second term is the component of the discount rate that is driven by diminishing returns to increasing wealth, where is the elasticity of the marginal utility of wealth 10 and the expected growth rate of the economy. The final term represents the uncertainty in our expectation of future economic growth (which offsets the second term since diminishing returns to wealth leads to a certain level of risk aversion). For standard cost-benefit 9 This is sometimes considered to be equivalent to the social rate of time preference. 10 This parameter can also be used to represent aversion to inequality. 15

21 analysis (i.e., single-generation), these parameters are typically inferred from the market. For example, if we take,,, and fairly common values seen in the literature), we get an SDR of approximately 6% (Weitzman, 2013). -- There is no consensus among economists which one of these three approaches is the correct one for standard cost-benefit analysis. In light of these competing methodologies and the range of values each of them can yield, the U.S. Office of Management and Budget takes a pluralistic approach, requiring federal agencies to perform their cost-benefit analyses using both a 3% discount rate (roughly corresponding to the SRTP approach) and the discount rate of 7% (roughly corresponding to the SOCC approach) (OMB, 2003). E. Discounting in the Intergenerational Context In the case of issues like climate change, we are evaluating costs and benefits that occur not only at different points in time, but in different generations. How should we calculate the SDR in this case? Can the discounting approaches used in standard cost-benefit analysis be similarly applied in the intergenerational case? Let s consider each approach in sequence. I will ultimately conclude that no approach to discounting can be reasonably applied intergenerationally. i. Applying the Social Rate of Time Preference Approach The social rate of time preference approach, as we discussed, is justified on democratic grounds and is typically inferred from markets for risk-free assets such as government bonds. However, it is not clear how this justification would extend intergenerationally. We might argue, 16

22 for example, that future generations have no right to representation in our public policy decisions today since they don t yet exist. By this logic, we might reasonably maintain the original justification for the SRTP and simply extend the time horizon. This would imply that policy decisions which affect future generations should weight the interests of future generations no less and more than the current generation weights their interests. But even if the justification for the SRTP holds in the intergenerational context, there would remain a difficult operational question, namely, whether we could continue to use the financial markets as a basis for inferring people s (intergenerational) time preferences. I suspect that the answer is no: the idea that we can understand an individual s preferences about how the government should allocate resources between generations based on their self-interested financial behavior seems fool-hardy. Even if we take the neoclassical economic view that (1) people s preferences are exogenously determined and (2) these preferences can accurately be revealed only in their behavior (rather than through considered statements, for example), it still does not follow that a person s market behavior would represent these preferences. Such a conclusion would require that there be a market by which individuals internalize their concern for future generations as a type of good. I do not believe such a market exists. But this turns out to be a moot point since, as I will show, the basic democratic premise of the SRTP approach that a government s time preferences should reflect those of its citizens cannot hold in the case of intergenerational issues such as climate change. For one, a state can be a self-respecting democracy but defy its constituency preferences from time to time. What if the constituency wished its government to carry out a massacre of a group of foreign innocents? Should the state satisfy such a request? Perhaps one might argue that, in an absolute democracy, the state is nothing more than an organ for collective preference satisfaction. But in a 17

23 representative, constitutional democracy, the state has grounds to defy its citizenry with just cause. In such a case, where satisfying the preferences of its citizens would commit a grave injustice in violation of its constitution, the state has an obligation to override these preferences. Unchecked carbon emissions may be such a case. But more powerfully and I believe the most compelling argument against the discounting is democratic argument is a rejection of the idea that a democracy s constituency is limited to those alive today. Instead, democracy might be considered a type of project over time. On this view, merely because future generations do not yet exist does not mean they should be precluded from representation (or, more precisely, represented only insofar as the current generation wishes them to be represented). In fact, taking this view might lead to the assertion that a government that accounts only for the interests of the current generation is, far from being democratic, patently undemocratic that discounting based solely on current time preferences, for example, would be an act of disenfranchisement of future generations. And if we take this view a view which I believe is basically correct how can we possibly rely on the market for information about future generations preferences? Even if the implied discount rate in financial markets does represent the current generation s long-term time preferences (a point I have disputed), it most certainly does not represent the time preferences of all generations. At its neoclassical best, the market reveals the preferences of its participants. But future generations are not participants. After all, how could they be if they do not yet exist? So if we take the view that all generations merit representation, we cannot employ a methodology for 18

24 calculating the discount rate which relies purely on market information. Consequently, the SRTP approach in the intergenerational context is simply untenable. 11 ii. Applying the Social Opportunity Cost of Capital Approach To review, the SOCC approach takes the discount rate to be an efficiency parameter to ensure we choose policies that will generate welfare returns over time at least as high as we can achieve through private investment. It infers the discount rate from the interest rate on private capital. Economists who defend this approach in the intergenerational context tend to employ a variant on the efficiency-equity decoupling argument, namely, that a government can and should separate the efficient generation of welfare (that is, maximizing aggregate welfare) from normative concerns about the distribution of welfare between generations. Consequently, intergenerational distributional concerns should not affect what investments the government makes, but rather how the welfare generated from these investments are (re-)distributed. The selection of public investments (of which discounting is a part) is not the appropriate venue to address intergenerational distributional concerns. If the objection to the SOCC approach to intergenerational discounting is based on the implications it has for the distribution of welfare between generations, this objection is misplaced. We only have reason to be concerned with the intergenerational distributional implications of any particular investments insofar as each contributes to the comprehensive distribution of resources between generations. Economists who think along these lines tend to suggest that, instead of focusing on discounting as the 11 We can actually draw a stronger conclusion: we cannot employ a framework which relies on knowing the preferences of people who will exist in the distant future. People s preferences are endogenous to the social, political, economic, environmental, and cultural circumstances they live in, and we cannot plausibly project how those circumstances will change into the distant future. 19

25 battleground for intergenerational distributional issues, we instead focus on the overall rate of savings between generations (Arrow, 1995). Put another way, the assertion that discounting at the market rate (as the SOCC would have us do) biases us toward investments with shorter time horizons is not so much wrong as it is irrelevant. There is no necessary correlation between the average time horizon for investment and the overall rate of savings between generations (since we can simply reinvest the returns from the initial investments). On the other hand, if we modify the discount rate based on concerns for intergenerational fairness and, as a result, invest in the longer-term, lower-return projects we will effectively lock ourselves out of the higher-return investment trajectory and ultimately hurt the very people we intended to help. 12 Consequently, we can and should decouple our pursuit of intergenerational fairness from our choice of the discount rate (Posner, 2007). While I agree that the intergenerational savings rate, relative to the discount rate, is a better parameter for managing the normative issues we are concerned with, these two parameters cannot actually be decoupled for issues such as climate change. Over long periods of time, the opportunity cost of capital (which determines the discount rate in the SOCC approach) tracks the overall rate of savings. To see this, consider what happens if we all collectively save more: the intergenerational savings rate will certainly go up, but the overall rate of return on capital will go down (since the highest-return investments will already have been taken). In turn, the opportunity cost of capital will also fall. Consequently, these two parameters cannot be isolated from each other, nor can the criteria that they represent (efficiency and fairness, respectively). In 12 However, this argument is valid only if the question at hand is how to allocate scarce resources between public and private investment. But if this allocation is already decided and the remaining question is how to allocate public funds between particular projects, this point is irrelevant. 20

26 short, the positivist defense of intergenerational discounting as a benign efficiency parameter with no distributional implications is simply untrue (Broome, 2012). 13 There is a second major problem with this argument. It depends on the existence of effective ways to transfer wealth across generations (without this, we cannot reasonably decouple the generation of welfare from the intergenerational distribution of it). While this is certainly plausible between proximate generations (e.g., through direct inheritance), it is harder to see how we could enact policies that accomplish this over many generations. After all, any transfers to distant future generations must go through the hands of many intervening generations, and we have no ability to predict what these middle men will do with the resources they receive, nor do we have any reason to believe that they will follow our wishes. This absence of a reliable intergenerational transfer mechanism further undermines the decoupling defense of the SOCC approach to intergenerational discounting. iii. Applying the Ramsey Approach I have shown, so far, that neither the SRTP nor the SOCC approach can offer a reasonable approach to discounting in the intergenerational context. Thus we are left with one last option the Ramsey Approach. This is the most promising of the three approaches, since the disaggregation of the input parameters allows for a clean distinction between instrumental and pure discounting. However, this approach ultimately turns out to be untenable as well. I will show that, without pure discounting (which I have argued is indefensible in the intergenerational 13 This relates to Nick Stern s defense of his below-market discount rate choice (if I understand Stern s argument correctly). He essentially argues that his rate is below market only relative to today s market, but that our climate policy will change the market itself. This is part of his broader critique of applying cost-benefit analysis an analytic framework meant for marginal projects and applying it to policy debates such as climate change, which are non-marginal (e.g., we cannot take macroeconomic trends to be exogenous to the policy decision).however, if I understand Stern s argument correctly, this seems to take a very optimistic view of the power of public policy to shape behavior in the market. This raises some interesting empirical questions: if we passed major climate reform, to what extent would it actually change people s consumption-to-savings ratio? 21

27 context), the discount rate values that the Ramsey Approach produces are too low and lead to the prisoner of the future problem. I have already shown that, unlike in the case of the single generation, we cannot appeal to democracy to justify pure discounting in the intergenerational context (nor, for that matter, am I aware of any plausible justification for pure discounting across generations). On the other hand, we have a strong reason not to engage in pure discounting, namely, that doing so seems to arbitrarily discriminate against future generations. Thus we should adopt a standpoint of intergenerational neutrality ( = 0). If we prohibit pure discounting, however, the low discount rates that result from the Ramsey Approach generate policies which are extremely demanding on the current generation. For example, Nicholas Stern, who famously argues against pure discounting on ethical grounds, adopts a Ramsey-based SDR of 1.4% (ρ =.1% 14, η = 1, and g = 1.3%). 15 At this discount rate, the cost-benefit analysis Stern runs says that we should dramatically change our current emissions trajectory. 16 The IPCC s baseline prediction is that annual GHG emissions will grow 1-3% per year. 17 In contrast, Stern says that the optimal policy would cause our emissions to peak in years (he published this report in 2007, so years would be ) and then decline year-on-year thereafter at 1-3%. 18 These changes would be costly for many people today: Stern first said it would cost 1% of global annual GDP, but revised estimates have suggested 2% or higher. Nonetheless, this is certainly feasible (Stern, 2006). 14 He chooses a pure rate of time preference of 0.1% to account for the possibility of human extinction. 15 This excludes the uncertainty term about economic growth. Depending on the magnitude of this value, it could reduce the SDR still further. 16 It can been shown that the low discount rate Stern used is the principal driver of the demanding policy conclusions. When increasing the discount rate, the required emissions reductions drop dramatically. 17 IPCC predicts global GHG emissions will increase 25-90% between 2000 and This is corresponds to a long-term stabilized atmospheric concentration of 550 ppm CO2e. 22

28 However, that is the burden of the climate policy alone. If we apply Stern s discount rate of 1.4% consistently across all intergenerational issues, it turns out that each generation would have to save, in total, somewhere between 40 to 97.5% of annual GDP. The upper end of this range is patently absurd, and even the lower end is extremely onerous, especially in light of the fact that each generation will tend to be poorer than the generations it is saving for (Dasgupta, 2008). This is not necessarily objectionable: such austerity might be justified if, for example, there were a high enough chance of future catastrophic events which would greatly outweigh the suffering imposed today. However, justifying these policies via a low discount rate is to arrive at what may be the right answer through the wrong means. Namely, what s driving the analysis of the Stern Review is not the potential for catastrophe but rather the sheer washing-out effect of using a low discount rate when aggregating benefits no matter how minute across an unbounded series of future generations. This is an inadequate justification for imposing austerity on the current generation (Weitzman, 2007). Thus, the Ramsey Approach leaves us between a rock and a hard place. We cannot engage in pure discounting, but avoiding it leads to absurd policy conclusions. Clearly, for an intergenerational cost-benefit analysis to yield reasonable results, we need to discount beyond simply accounting for increasing intergenerational wealth, but the Ramsey Approach does not offer us a justifiable way of doing so. Is there any other argument for discounting across generations? The answer is no, but before I can draw this conclusion I must refute one final argument that has recently been put forth to this end. 23

29 iv. The Tactical Justification for Intergenerational Discounting 19 In Climate Policy: Justifying a Positive Social Time Preference, Joseph Heath attempts to move us past this impasse by exploring a number of alternative arguments for intergenerational discounting. Some of his arguments have already been critiqued in this paper (e.g., he mounts a defense of discounting by appeal to opportunity costs à la SOCC approach), but one important argument he makes which I have not yet considered I will consider here: in what I will call his tactical justification for intergenerational discounting, he argues that even if intergenerational discounting itself is morally arbitrary, it is an efficient way of coordinating intergenerational moral efforts. Making an analogy to humanitarian aid coordination problems, he argues that, in the same way we might assign different people different geographic zones to send their donations (to prevent the classic Baby Jessica problem), we might similarly assign different generations different temporal zones to focus their moral efforts. In particular, we might tell each generation to focus their efforts on their own generation and (exponentially) less on future generations. In this way, we might use temporal zoning to justify a type of instrumental discounting of future generations, despite the zones themselves having no fundamental moral significance (Heath, 2013). There is a lot to be said for this argument. For one, I agree with Heath s basic goal, namely, to find ways to reconcile philosophical critique of discounting with economic pragmatism. 20 Moreover, it is certainly worth developing operational systems by which to 19 One might also refer to this as the distributed moral responsibility or the institutionally mediated morality justification to intergenerational discounting. 20 He suggests philosophers have put themselves in a corner by tending toward a categorical stand against intergenerational discounting, and I tend to agree. However, I would argue the way out of the corner is not to find ways to bring philosophical merit to discounting (and cost-benefit analysis more generally), but rather focus on a 24