Senior Fellote, institute for international Economices
Global warming from the buildup of carbon dioxide, methane, and other greenhouse gases is nearly unique as an environmental problem, because it involves global rather than local effects and is irreversible on a time scale of three centuries or more. In my recent book, I have estimated that scientists’ central value of 2½°C for expected global warming by 2050 would imply warming of 10°C by 2300, and their upper-bound value of 4½°C by 2050 would mean 18°C by 2300. Economic damages under even a moderate-central estimate would be at least 1 percent of gross world product (GWP) by 2050 and 6 percent by 2300, and upper-damage cases reach over 4 percent of GWP by the first date and 20 percent by the second. Worse “catastrophic” consequences cannot be ruled out.
Reducing carbon emissions by about one third and holding them constant indefinitely thereafter would avoid the great bulk of this warming and damage. Costs to achieve this outcome could be held low at first, through a move to more efficient energy use and through low-cost carbon savings from afforestation and reduced deforestation. However, by about 2020, the abatement costs could reach some 3 percent of GWP as industrial emissions are curbed, based on several energy-economic-carbon models. The costs would decline thereafter to perhaps 2 percent of GWP, thanks to the advent of new energy technologies.
Cost-benefit analysis provides a basis for economic evaluation of policy toward global warming. This requires two key methodological decisions: what discount rate to use to compare effects over time and how to take risk into account. The discount rate has an unusually powerful influence because of the extremely long time horizon of global warming, and because abatement costs occur early, whereas greenhouse damages avoided (the “benefits” of abatement) show up only after several decades (see chart). I argue that the appropriate overall discount rate should be about 2 percent a year in real terms.
Choosing the discount rate
The debate over what discount rate to use for public policy purposes certainly is not a new one. Even within the United States, a variety of levels can be found (the Office of Management and Budget favors an inflation-adjusted rate of 10 percent a year, whereas the US Congressional Budget Office and General Accounting Office prefer a rate close to the government’s borrowing costs—about 2.5 percent real, historically, for long-term government bonds and close to the rate I apply). For a project with a life of five to ten years, the difference is not necessarily large; but for a horizon of 200 years or more, the difference is extreme (see table on page 2).
Conceptually, there are two rates policymakers can call upon: the rate of return on private capital investments (opportunity cost of capital); and the social rate of time preference (SRTP)—or the extra value people place on consumption today rather than in the future. The former tends to be higher (say, 8 percent real) because of project risk, taxation of capital income, and capital market imperfections. The latter tends to be lower. Households are net savers, and the risk-free real rate of interest they can earn on savings is the Treasury bill rate (historically around 0.5 percent real a year).
“Conservative” economists argue that the rate of return on capital is the only one that matters, as the resources devoted to any undertaking could alternatively be placed into private projects where they would earn this rate. However, over the past two decades, mainstream cost-benefit analysis has moved to take both rates into account. In the 1960s, Arnold Harberger, Otto Eckstein, and William Baumol first proposed a weighted average of the two rates, with the weights reflecting how much of a project’s resources were drawn from displaced capital investment as opposed to reduced consumption.
Today’s state of the art discounting methodology—which I draw upon for my global warming cost-benefit analysis—was developed in the 1970s by Kenneth Arrow, David Bradford, Martin Feldstein, and Mordechai Kurz (the ABFK method). Their approach uses a “shadow price on capital” to translate all capital investment effects (such as the withdrawal of resources from alternative investments) into “consumption equivalents.” The ABFK method then discounts consumption over time (including consumption-equivalents of capital effects) at the SRTP for consumption.
Social rate of time preference. To estimate the SRTP—which I calculate to be 1.5 percent—I appeal to basic economic theory, which states that this rate equals the sum of two components:
The first is “pure” or “myopic” preference for consuming a good sooner rather than later—myopic, because it implies an inability to envision future pain imposed by robbing the future to increase today’s pleasure. Ants have a low pure time preference rate; grasshoppers, a high one. There is a strong tradition among economists to set this preference at zero, especially for comparisons between the present generation and future generations, who cannot participate in today’s decisions. “Emotional distance” does not justify pure time preference because it would invite imposing damage on others just because they are not ourselves. Nor do higher consumer borrowing rates constitute empirical evidence on pure time preference; typically, net borrowers expect their income to rise, so their discount rate is based on lower expected future marginal utility of consumption.
The second is the “utility-based” discount rate, which takes account of declining “marginal utility” as income rises. Just as the third doughnut adds less satisfaction (“utility”) than the second, the utility from an extra $1,000 is smaller for an individual at a $20,000 income than for the same individual at a $10,000 income. This component, in turn, equals the product of (1) the growth rate of per capita income and (2) the responsiveness (“elasticity”) of marginal utility with respect to consumption—how fast the consumption value of an extra dollar drops off as the individual attains higher consumption levels. For my study, I set the utility-based rate at 1.5 percent. This estimate applies an “elasticity of marginal utility” of 1.5, based on independent calculations by William Fellner and Maurice Scott. It also assumes an average per capita growth rate of 1 percent over the next three centuries. Combined with rising population, even this seemingly modest rate multiplies GWP 25-fold over the horizon, which some would consider incompatible with global resources.
Pay now, enjoy later
Costs and benefits of aggressive abatement of greenhouse warming
(percent of GDP)
Citation: 30, 1; 10.5089/9781451953077.022.A002
Source: The Economics of Global Warming, 1992.
Weights. The economic cost of carbon reduction is the sacrifice in output imposed by constraining fossil-fuel energy. This production loss should affect capital investment and consumption in proportion to their shares in the economy at large. I thus assume that the capital investment versus consumption origins of resources diverted to greenhouse abatement equal their economy-wide shares, or 20 percent and 80 percent, respectively.
Shadow price of capital. This I set at approximately 2 (one unit of capital is worth two units of consumption)—consistent with 1.5 percent for the SRTP, 8 percent for capital’s rate of return, and a 15-year capital life.
The bottom line. The overall effect is approximately comparable to discounting at 2 percent real (i.e., 1.5 percent x 0.8 for consumption share in resources displaced, plus 1.5 percent x 2 capital shadow price x 0.2 for capital share).
Where does this method lead for greenhouse policy? My central scenario shows that the discounted benefits of limiting global warming would cover only about three quarters of the discounted costs. However, if risk aversion is incorporated by adding high-damage and low-damage cases and attributing greater weight to the former, benefits comfortably cover costs (with a benefit-cost ratio of about 1.3 to 1). Aggressive abatement is worthwhile even though the future is much richer, because the potentially massive damages warrant the costs. This conclusion takes on even more strength if the future is not richer, because then the SRTP should be set at zero.
In contrast, at a discount rate of 10 percent, in the central case the benefit-cost ratio falls from 0.74 to 0.33; and for the high-damage case, from 2.99 to 1.07. Not even risk-averse policymakers would adopt abatement if they discounted the future at 10 percent (unless they added “catastrophic” outcomes).
Certain questions naturally arise in interpreting the method I suggest. First, should the same method apply to both environmental and nonenvironmental projects? Because the ABFK method is mainstream cost-benefit analysis, the answer is yes. My approach is not based on the use of a different underlying methodology for environmental projects and does not mean that these projects should automatically have a lower discount rate than other projects. What matters most is where the resources come from: capital or consumption.
However, under some circumstances, the same method will generate a significantly higher discount rate. Thus, if there is severe crowding out of private investment by additional government borrowing, then a project may primarily displace investment, meaning the effective discount rate would be substantially higher than the SRTP. Moreover, the SRTP itself may be higher if the country’s growth rate is high, or its populace is near starvation and has an unusually steep rate of drop-off in marginal utility—both conditions that may be likely in low-income countries.
Another question is: Will not low discount rates cause more environmental damage by justifying more projects? This argument is a red herring. If project analysis is done properly, it will incorporate the cost of adverse environmental side effects (“full-cost pricing”), and this will tend to rule out environmentally damaging projects.
A third question is: Does not this method imply that society undersaves and underinvests today, considering that the SRTP is well below the rate of return on capital? The answer is yes. However, where society is prepared to undertake more investment, as is arguably the case in global warming, second-best strategy recommends taking action if the cost-benefit results meet the criteria outlined here, rather than opposing action on grounds that something even better might be done with the money (but will not be). Moreover, this approach tends to increase—not reduce—total productive capacity passed on to the next generation.
Finally, does an extremely long horizon mean a lower discount rate? The answer is not necessarily, although it is likely to do so in practice. Per capita growth (and thus the SRTP) over the very long term is likely to be lower than that in the near term. Even the capital opportunity cost approach should apply a declining discount rate over a horizon of centuries, as capital accumulation relative to other production factors will reduce the rate of return on capital.
The World Bank’s approach
Former Bank Chief Economist Lawrence Summers (The Economist, May 30, 1992) argues for a real discount rate of at least 8 percent in evaluating global warming. There are three possible reasons for this recommendation. First, the mainstream discount methodology yields a rate of 8 percent. Second, the SRTP is irrelevant; what matters is the opportunity cost of capital. Third, because alternative Bank projects achieve such rates, so should environmental projects.
The first of these views would be extremely difficult to defend, because it would require an implausibly high SRTP. Consider 8 percent discounting. Over 200 years, this discount factor reaches 4.8 million to one. Surely, Summers, along with Bank economists Nancy Birdsall and Andrew Steer (in this issue), do not believe society would have been better off over time if today’s chairman of the Council of Economic Advisers could have been compelled to give up $4.8 million of consumption to make her bicentennial predecessor (Alexander Hamilton) better off by just one dollar of consumption (at constant prices). Even my SRTP of 1.5 percent means that this tradeoff is $20 to $1, already steep.
As for the second view, for narrow purposes of comparison among Bank projects, this approach may be adequate (with the qualification below). But for society’s broad policy toward the greenhouse problem, it seems seriously misleading and imposes a far more stringent test than the mainstream cost-benefit analysis outlined above (e.g., ABFK). Even within this “conservative” position, 8 percent is implausibly high for a 300-year horizon. The rate of return on capital will decline as capital increases relative to labor and natural resources. Today, the Bank invests $15 billion per year. Compounded at 8 percent, just one year’s investment of this amount would grow to 100,000 times my estimate of GWP at the end of the horizon! Birdsall and Steer make a modest bow in the direction of recognizing that an 8 percent rate is implausibly high by invoking a possible 5 percent rate to reflect lower investment return in industrial countries, at least for abatement by these countries. However, they do not accept the central point that the rate should take account of resource sourcing out of consumption as opposed to displaced private investment. They thus adhere conceptually to the “conservative” position.
If the Bank authors, however, are merely arguing the third view—that greenhouse and other environmental projects must achieve a return competitive with other Bank projects, then I agree (if the Bank really achieves 8 percent return). The Bank has a limited pool of capital. But in taking the greenhouse problem into account in all its projects, there should be a shadow price penalty on carbon emissions (e.g., from coal-fueled power plants) and a shadow price benefit added for projects that reduce these emissions (e.g., afforestation). This shadow price on emissions should be based on the society-wide cost-benefit analysis of abatement, and thus on the ABFK method with discounting values similar to those outlined above. This analysis would tend to place a price of $10-$20 per ton of carbon initially (the opportunity cost in afforestation or reduced deforestation), rising to $100-$200 per ton of carbon sometime after the turn of the century (and subject to further scientific confirmation). Once the greenhouse gas externalities are counted, it makes sense to require that the project compete with alternative uses of the Bank’s scarce funds at the going rate of return. The only exception might be for funds specifically provided by donors for environmental projects, as in the Global Environment Facility.
Instead, Summers, Birdsall, and Steer seem to argue much more broadly that society should do little to limit greenhouse gas emissions. They implicitly counsel that the public would be better off to set aside money in a “Fund for Future Greenhouse Victims.” The money would be invested at 8 percent (or, conceivably, at a still high 5 percent), and the income would compensate future generations for global warming damage at a far smaller cost than through limiting emissions.
There are several problems with this notion. First, as just discussed, real rates of return are unlikely to stay as high as 8 percent. Second, we cannot identify producer goods that yield a steady chain of producer goods that at the end of two centuries disgorge consumer goods of relevance to the population at that time (the “intertemporal transfer problem”). Third, we are not sure the damage estimates for global warming truly capture the change in the relative price between goods and environment. How many video cassette recorders will the future generations really consider an adequate compensation for 10°C or more global warming—especially if there is catastrophic risk? Fourth, it is not enough to talk about the possibility of such a Fund; tax revenue would actually have to be collected to implement it. It is unclear that the public will be willing to pay such taxes, whereas considerable public support for stemming global warming was already evident at the 1992 Rio Earth Summit.
So far, the message sent by the World Bank, at least in its World Development Report 1992, is that only minimal action on global warming is likely to be justifiable. An improved discounting methodology could change this diagnosis. Environmentalists have come a long way toward accepting economists’ arguments, for example, by beginning to embrace tradable permits for emissions. Economists should not discourage this convergence by imposing a more stringent hurdle to action on global warming than would be required by mainstream cost-benefit analysis.