Comments on the Stern Review

Remarks at the Roundtable Discussion on the Stern Review
of the Economics of Climate Change and Its Critics
American Economic Association, Annual Meeting, New Orleans

January 5, 2008

The Stern Review has moved the center of the debate on the economics of climate change in the right direction, toward placing the burden of proof more on those who argue that it is uneconomic to take major abatement measures. In crucial ways the review resembles the much simpler analysis in my 1992 book, The Economics of Global Warming. Following Frank Ramsey, I used zero as the rate of discounting for pure time preference, or discounting for simple impatience as opposed to higher living standards. The Social Rate of Time Preference is then the rate of pure time preference, zero, plus the product of the elasticity of marginal utility and the growth rate of per capita income. The Stern Review adopts the same approach. I proposed a much longer time horizon, three centuries, than was the practice at the time, when the focus was on a few decades for doubling of carbon dioxide concentrations; the Stern Review leapfrogs to a horizon of infinity. My nonlinear damage function resulted in a range of damage estimates that was similar to the 5 percent to 20 percent of GDP by 2200 estimated in the review. In my analysis, abatement reducing global emissions by about one-third and then holding them flat had social benefits exceeding costs. This amounted to a cut of about 80 percent from baseline by 2100; the Stern Review calls for similar cutbacks.

There are important differences, however. The Stern Review implicitly obtains a cost-benefit ratio for abatement of about 12:1. The midpoint of its range for damage avoided is 12.5 percent of GDP, and abatement costs only 1 percent of GDP. My cost-benefit ratio was only 1.3:1. There are three key reasons for the difference. First, the review uses an infinite horizon; I used only 300 years, because carbon concentrations start falling again after that from mixing back into the deep ocean. Second, the review uses an elasticity of marginal utility of unity, whereas I used 1.5. Third, following the social cost-benefit approach of Arrow, Bradford, Feldstein, and Kurz, I formally incorporated the gap between the return on capital and the social rate of time preference by applying a shadow price to resources withdrawn from investment. This effectively raises the estimated cost of abatement by about 13 percent. The Stern Review did not shadow-price capital, making itself unnecessarily vulnerable to the critique that the discounting method does not reflect return to capital.

Because of the combination of an infinite horizon, near-zero pure time preference, and unitary elasticity of marginal utility, 93 percent of the present value of total future welfare in the review arises after 2200. This means the review is vulnerable to the critique that the distant unknowable tail wags the dog. If instead the elasticity of marginal utility is raised to my value of 1.5, the share of total future welfare that arises after 2200 is only about 30 percent, which would strike most policymakers as much more sensible. Moreover, an elasticity of marginal utility of 1.5 is far more consistent with observed progressive tax regimes than is a value of unity, on the one extreme, or two or three at the other extreme. The Stern Review's value of unity would mean the flat percentage tax (like the biblical tithe) is fair. The value of two or three suggested by Nordhaus and Dasgupta would place the income tax on the moderately rich, such as the president of a top university, at about 90 percent, whereas a value of 1.5 translates to a rate of about 40 percent, far closer to what society has chosen.

The overall effect of raising the elasticity of marginal utility somewhat and incorporating a shadow price on capital would be to reduce the cost-benefit ratio of aggressive abatement from about 12:1 to perhaps 3 or 5:1, in the Stern Review framework. The bottom line would still be that such action is justified but not by as wide a margin.

Critics have argued that the Stern Review cannot have both zero pure time preference and unitary elasticity of marginal utility; one or the other must be higher to accord with observed capital market returns of, say, 4 percent or more. But social cost-benefit analysis applies the consumption discount rate with separate capital shadow pricing, and the market today has real return of only about 1.5 percent on long-term government bonds, which is the rate for transferring consumption over time. More fundamentally, markets simply do not exist for centuries-scale horizons. Many critics also seem to assume that all resources for abatement come out of investment, whereas the social cost-benefit approach explicitly recognizes that probably only about one-fifth of the opportunity cost is investment and four-fifths is consumption. I have yet to hear the critics address this point, perhaps because it is too easy for them to criticize the review for seeming to ignore the difference between the capital opportunity cost and the social rate of time preference altogether.

The main debate increasingly seems to be whether to "ramp up" after a slow start of abatement, or act early and aggressively as in the Stern Review. My concern is that the critics would ramp up too little and too late. For example, Bill Nordhaus' optimal policy in his 2007 runs of the DICE model would allow emissions to rise 50 percent over the next 50 years to 10 billions of metric tons of carbon (GtC) at mid-century and continue to rise to about 12 GtC annually by 2100, in contrast to Stern's cut to about 3.5 GtC annually by 2100. It would ultimately limit warming late next century to 3.5°C. Although this would be lower than the baseline 5.5°C, it would almost certainly be sufficient to cause massive sea level rise over subsequent centuries. Moreover, one standard deviation above the central estimate would result in warming of 5.5°C, which could potentially be catastrophic.

Nordhaus does include a scenario limiting warming to 2oC, and finds the extra cost a remarkably cheap 0.1 percent of present-value GDP. But he does so only by delaying the most severe action for several decades, after which his 4 percent discounting collapses the measured cost. Nordhaus' 2-degrees run would not reduce the absolute level of emissions at all until mid-century, but then would cut emissions by an average 45 percent each decade for the next four decades. But it is implausible that politicians and the public will suddenly switch to stringent self-discipline at mid-century. Precommitment to such a path is not credible.

Yet a limit of 2-2-1/2 degrees would be highly desirable. I think Marty Weitzman has been correct to emphasize the need to incorporate the fat-tail of low probability and high damage at much higher warming. For example, I have been struck by recent geological work linking the Permian-Triassic mass extinction event 250 million years ago, when 90 percent of all species on land and in the ocean died out, to the release of methane from frozen hydrates and an associated runaway greenhouse effect that boosted global temperatures by 6°C. Similarly, the last time global temperatures were even 3°C higher than today, 3 million years ago, sea level was 30 meters higher than today.

In sum, I think the Stern Review has done an important service in rebalancing the policy debate. It may overstate, but appropriate technical adjustments would still leave its bottom line intact: Aggressive action is warranted. Greater attention to catastrophic risks and formulation of the issue as one of buying insurance will, I believe, only reinforce this conclusion.