Вы находитесь на странице: 1из 22

sep07_Article4 8/17/07 4:31 PM Page 703

Journal of Economic Literature

Vol. XLV (September 2007), pp. 703–724

A Review of The Stern Review on the

Economics of Climate Change

The Stern Review calls for immediate decisive action to stabilize greenhouse gases
because “the benefits of strong, early action on climate change outweighs the costs.”
The economic analysis supporting this conclusion consists mostly of two basic strands.
The first strand is a formal aggregative model that relies for its conclusions primari-
ly upon imposing a very low discount rate. Concerning this discount-rate aspect, I am
skeptical of the Review’s formal analysis, but this essay points out that we are actual-
ly a lot less sure about what interest rate should be used for discounting climate
change than is commonly acknowledged. The Review’s second basic strand is a more
intuitive argument that it might be very important to avoid possibly large uncertain-
ties that are difficult to quantify. Concerning this uncertainty aspect, I argue that it
might be recast into sound analytical reasoning that might justify some of the
Review’s conclusions. The basic issue here is that spending money to slow global
warming should perhaps not be conceptualized primarily as being about consump-
tion smoothing as much as being about how much insurance to buy to offset the small
change of a ruinous catastrophe that is difficult to compensate by ordinary savings.

1. Introduction on a vivid actuality to breathe new life into

otherwise arcane matters of economic analy-
T he issue of global climate change and
what to do about it has put economics to
a severe test in which economists have been
sis. Beyond the issue of whether it is right or
wrong in its conclusions, the Stern Review on
the Economics of Climate Change is an
challenged to think afresh about how to
opportunity for economists to take stock of
model (or at least how to conceptualize) such
what we know about this subject, how we
fundamental notions as risk, uncertainty, and
know it, what we don’t know, and why we
discounting. There is nothing like being
don’t know it.
asked for a specific policy recommendation
The Stern Review is a full-fledged eco-
nomic analysis of climate change that was
∗ Weitzman: Harvard University. For helpful detailed officially commissioned by the British gov-
comments on earlier drafts of this paper, but without ernment and, for reasons both economic and
implicating them for its remaining defects, I am grateful
to Scott Barrett, Roland Benabou, Olivier Blanchard, political, is an unusual—and unusually
Richard Cooper, Stephen DeCanio, Howard important—document. Sir Nicholas Stern is
Gruenspecht, Cameron Hepburn, Chris Hope, Donald a professional economist of high standing
Ludwig, Robert Mendelsohn, Larry Samuelson, Robert
Solow, Nicholas Stern, Lawrence Summers, and Hal and a distinguished public servant. Weighing
Varian. in at close to 700 pages, the Stern Review is

sep07_Article4 8/17/07 4:31 PM Page 704

704 Journal of Economic Literature, Vol. XLV (September 2007)

comprehensive in its scope and ambitious in strong and early action far outweigh the eco-
its aims, with an attractive multicolored visu- nomic costs of not acting” (p. xv) and calls for
al design that makes topics like cost–benefit stabilizing greenhouse gas atmospheric con-
analysis of dynamic externalities look almost centrations at ≈ 550 parts per million (ppm)
glamorous. Anyone wanting to get a good of CO2-equivalent (CO2 e). (The current
feel for the basic issues of global climate level is ≈ 430 ppm CO2e, compared with
change could profitably browse through this ≈ 280 ppm CO2 e before the Industrial
report, which covers well its multiple facets Revolution.) This would make temperatures
in a reader-friendly format. The Review con- a hundred years from now be at E[ΔT] ≈ 2˚C
tains much of value and interest aside from and would (hopefully) stabilize future tem-
its cost–benefit analysis of mitigation poli- peratures permanently thereafter at
cies, although that is naturally the part which Δ ≈ 3˚C. By contrast, along the more grad-
most grabs the attention of economists. A ual majoritarian optimal trajectories CO2e
detailed Review of the Review is out of place concentrations a century from now are > 600
here—it would be too long, and besides the ppm and E[ΔT] ≈ 2.5˚C—with temperatures
Stern Review reads well and is available expected to continue rising to well above
online. Instead, I concentrate here on trying E[ΔT] ≈ 3˚C after year 2105. To accomplish
to distill the Review down to what I think is the Review’s ambitious goal, greenhouse gas
its analytical essence as a piece of applied emissions would need to be progressively cut
cost–benefit analysis, because there can be by ≈ 3 percent each year, beginning more or
difficulty seeing the forest for the trees when less immediately. Which brings us to a cen-
there are so many trees. tral question. Why is there such a big differ-
To make a long story short, the Stern ence between what Stern is recommending
Review comes down very strongly on the side and what most other serious analysts favor?
of undertaking decisive—and expensive— This paper makes five basic points about
measures starting now to reduce CO2 and the economics of climate change: (1) the dis-
other greenhouse gas emissions because count rate we choose is all important and
(and this quote captures well the tone of Stern’s results come from choosing a very low
urgency about moving quickly to avoid cata- discount rate; (2) we are a lot less sure about
strophic possibilities that is evident through- core elements of discounting for climate
out the report): “Our actions over the change than we commonly acknowledge
coming few decades could create risks of because critical puzzles, projections, and
major disruption to economic and social ambiguities are yet unresolved; (3) standard
activity, later in this century and in the next, approaches to climate change (even those that
on a scale similar to those associated with the purport to treat uncertainty) fail to account
great wars and economic depression of the fully for the implications of large conse-
first half of the 20th century” (p. xv). Such a quences with small probabilities; (4) structur-
strong call to immediate decisive action is at al parameter uncertainty that manifests itself
odds with what most other economic analy- in the thick tails of reduced-form probability
ses of climate change have concluded. The distributions—not risk—is what likely matters
majority view of most other economic ana- most; (5) gathering information about thick-
lysts finds it optimal to pursue a more grad- tailed uncertainties representing rare climate
ualist course by starting with greenhouse gas disasters (and developing a realistic emer-
emissions reductions at far lower levels than gency plan were they to materialize) should
what the Stern Review advocates for the be a priority of research. To anticipate my
near future, but which after that ramp up main finding, spending money now to slow
considerably over time. The Review analysis, global warming should not be conceptualized
on the other hand, finds that “the benefits of primarily as being about optimal consumption
sep07_Article4 8/17/07 4:31 PM Page 705

Weitzman: A Review of the Stern Review 705

smoothing so much as an issue about how time profile by imposing discounting at a

much insurance to buy to offset the small bare-minimum rate of interest.
chance of a ruinous catastrophe that is diffi- Global climate change unfolds over a time
cult to compensate by ordinary savings. While scale of centuries and, through the power of
I am (along with most other economist- compound interest, what to do now is hugely
critics) skeptical of Stern’s formal analysis, I sensitive to the discount rate that is postulat-
believe that the Review’s informal emphasis ed. In fact, it is not an exaggeration to say that
on climate-change uncertainty can be recast the biggest uncertainty of all in the econom-
into sound analytical arguments that might ics of climate change is the uncertainty about
justify some of its conclusions. which interest rate to use for discounting. In
one form or another, this little secret is
2. Interest Rates and Long-Term known to insiders in the economics of climate
Discounting change, but it needs to be more widely appre-
ciated by economists at large. The insight that
Overall, I believe it is fair to say that the the strong conclusions of the Review are driv-
Stern Review consistently leans toward (and en mainly by the low assumed discount rate
consistently phrases issues in terms of) has been picked up and commented upon
assumptions and formulations that empha- already by several insider critics.1 Here I
size optimistically low expected costs of mit- want to paraphrase this important debate for
igation and pessimistically high expected outsider economists and in the process bring
damages from greenhouse warming—rela- some new ingredients to the mix.
tive to most other studies of the economics An Integrated Assessment Model—here-
of climate change. But far more crucially, after IAM—is insider lingo for a multiple-
the key assumption that drives its strong equation computer-simulated model that
conclusions is the mundane fact that a very combines dynamic economics with geophys-
low interest rate is postulated, with which ical climate dynamics for the purposes of
distant-future benefits and costs are then analyzing the economic effects of global cli-
discounted. The upward-sloping “climate mate change.2 An IAM is essentially a model
policy ramp” of ever-tighter emissions of economic growth with a controllable
reductions in the majority of other models externality of endogenous greenhouse warm-
(but not beginning just yet, please) is a ing. The Review uses an IAM called PAGE,
familiar consumption-smoothing conse- on which some numbers have been crunched
quence of discounting: the higher the inter- and some conclusions have been based, but
est rate the stronger the desire to move the exact connection between PAGE and
toward getting more pleasure now at the Stern’s conclusions is elusive, frustrating, and
expense of postponing more pain until later. ultimately unsatisfactory for a professional
An efficient trajectory has a cost minimizing economist who honestly wants to understand
substructure similar to a Hotelling extrac- where the strong policy recommendations
tion problem: consumption flows are are coming from.3 The analytical core of the
smoothed over time by maximizing present
discounted utility subject to a stock con- 1 Variants of this argument are made in Partha
straint on accumulated CO2e, which results Dasgupta (2007), Robert O. Mendelsohn (2007), William
in an “as if” CO2e shadow tax that grows over D. Nordhaus (2007), and Richard S. J. Tol and Gary W.
Yohe (2006).
time at (approximately) the rate of interest. 2 A survey of integrated asssessment models for climate
The Stern Review simultaneously raises change control is provided in David L. Kelly and Charles
overall greenhouse gas reductions and flat- D. Kolstad (1999).
3 A nice description of how the Stern Review uses (and
tens the “climate policy ramp” in its misuses) PAGE is contained in David Maddison (2006).
Hotelling-analogous consumption smoothing PAGE itself is described in Chris Hope (2006).
sep07_Article4 8/17/07 4:31 PM Page 706

706 Journal of Economic Literature, Vol. XLV (September 2007)

Review is chapter 6 (“Economic Modelling of where r is the interest rate (more on the
Climate-Change Impacts”), which is loosely interest rate later), δ is the rate of pure time
tied to PAGE. However, the rest of the book preference, g is the per capita growth rate of
contains lots of stories and examples suggest- consumption, and η is the elasticity of mar-
ing that difficult-to-quantify uncertainty ginal utility, or, equivalently, the coefficient of
about really bad climate extremes may actu- relative risk aversion. In the shorthand nota-
ally be an important informal part of Stern’s tion of (1), the parameters δ and η capture
overall case. Economists are justifiably suspi- two critical aspects of “tastes” (or “prefer-
cious when someone refuses to aggregate ences”) while the reduced-form representa-
various probability-weighted scenarios into tion of “technology” is the growth rate of
an overall cost–benefit assessment, which at consumption g. The important distinction
least can serve as a conversation starter. (How between δ and r is that δ is a more primitive
else are we to evaluate overall policy advice, rate of pure time preference that discounts
such as what Stern recommends to us, except utility, while r is the much more familiar
in the context of some overall model where interest rate used to discount consumption,
assumptions and specifications are spelled which is derived from all of the more primi-
out clearly?) As economic analysis, the Stern tive underlying parameters of tastes and
Review dwells in a nonscientific state of technology via (1). The other taste parameter
limbo where it uses an IAM but simultane- η represents the relative curvature of the
ously refuses to commit to it or to any other utility function and is simultaneously a meas-
consistent overarching framework within ure of aversion to interpersonal inequality
which its radical recommendations might be and a measure of personal risk aversion. On
deconstructed and judged by others. Instead, the technology side, formula (1) holds
the Review dances around the significance of whether g is endogenous or exogenous. In
the aggregative analysis of chapter 6 by argu- Ramsey’s time, g was conceptualized as com-
ing that conclusions from IAMs are sugges- ing from capital accumulation, and therefore
tively useful but not crucial to the basic story in long run equilibrium with diminishing
line that anything above ultimate stabilization returns to capital g → 0 and r → δ. We now
at ≈ 550 ppm of CO2e and Δ ≈ 3˚C is self- know from modern post-Solow growth theory
evidently just too risky for the planet to bear. (but Ramsey and Fisher didn’t) that, in bal-
In trying to make some overall sense of anced growth steady-state equilibrium, g is
Stern’s mixed methodology (called “multidi- essentially the underlying growth rate of labor-
mensional” in the Review), I propose here to augmenting technological progress that,
lay out the core issues of how risk, uncertain- behind the scene, is pushing the entire econo-
ty, and discounting interact with the econom- my forward (at least in a world without a
ics of climate change in terms of the simplest greenhouse-warming externality). What I pro-
general-equilibrium model I can think of. pose to do here is use the Ramsey equation as
Then I will try to clothe some parts of Stern’s a transparency-based springboard for recast-
intuitions about climate-change uncertainty ing the economics of climate change in terms
in formal garb. of the four critical variables that appear in (1):
Irving Fisher taught us that an interest δ, η, g, and r. I will ultimately argue that, in a
rate, like any other price, is the outcome of a greenhouse gas world, g needs to be seen as a
dynamic general-equilibrium interaction of random variable whose probability distribu-
tastes with technology. The modern incarna- tion has a climate-change-thickened left tail
tion of Fisher’s idea in a deterministic setting that carries most of the weight of expected
is the famous Frank Ramsey equation marginal utility in cost–benefit analysis.
To cut sharply to the essence of the core
(1) r = δ + ηg, discounting issue behind the Review’s strong
sep07_Article4 8/17/07 4:31 PM Page 707

Weitzman: A Review of the Stern Review 707

conclusions, pretend there are just two temporarily such details as the optimal
periods—the present and the future—where consumption-smoothing profile of measures
the “future” is about one hundred years from to slow greenhouse warming (and their iner-
now. For the purposes at hand, I am about to tial consequences) in favor of immediate
conduct a gigantic macroeconomic cost– transparency by focusing on the highly aggre-
benefit exercise trading off less present con- gated macroeconomic big picture of what is
sumption from greenhouse gas abatement most essential in driving the Stern Review
for more future consumption from mitigat- results, for which purpose focusing on a cen-
ing the bad effects a century hence of global tury hence is a good enough approximation.
warming. Technically speaking, the possibili- Going right to the target here, my own
ty of extreme left-tail values of g occurring rough point-guesstimate of what most econo-
with small positive probability is outside this mists might think are decent parameter val-
marginalist framework and requires us to go ues would be something like a “trio of twos”:
back to the fundamentals of expected utility δ = 2 percent, g = 2 percent (both on an
theory that lie behind cost–benefit analysis annual basis), and η = 2 percent. For the
under uncertainty, but I will cross that bridge sake of moving along, I am not going to try to
when I come to it later and the take-away defend the “trio of twos values” with a bunch
message will turn out to be similar anyway. of citations but instead I pretend for the
Of course such an incredible oversimplifi- time being that every critic of Stern thinks
cation of the economics of climate change they are about right, so we can temporarily
ignores or distorts truly monumental chunks shelve this issue. Plugging these primitives
of reality. As just one example among many, into (1) makes “the” annual interest rate be
a very important part of the global warming r = 6 percent. Other reasonable—in my
story concerns the huge stock–flow lags and view—parameter combinations, say δ = 1
enormous built-in inertias from having such percent, g = 2 percent, η = 2.5 (or even
a long pipeline between greenhouse gas δ = 0 percent, g = 2 percent, η = 3) also
emissions and ultimate temperature give r = 6 percent.
changes. This built-in inertia causes ΔT to Concerning the rate of pure time prefer-
continue to rise to levels above E[ΔT] ≈ 3˚C ence, Stern follows a decidedly minority
after a century from now (and also causes paternalistic view (which, however, includes
the tail probabilities of very high tempera- a handful of distinguished economists) that
tures ΔT > 6˚C to be relatively much bigger for social discounting selects the lowest con-
two centuries from now than one century ceivable value δ ≈ 0 according to the a priori
from now) along any trajectory that does not philosophical principle of treating all gener-
stabilize atmospheric greenhouse gases at ations equally—irrespective of preferences
≈ 550 ppm CO2e—including the majority- for present over future utility that people
opinion gradualist climate-policy-ramp tra- seem to exhibit in their everyday savings and
jectory. The ultimate high-temperature investment behavior. In a similar spirit of
consequences of the huge inertial lag of ΔT choosing extreme taste parameters, Stern
to greenhouse gases already in the pipeline selects as its base-case coefficient of relative
animate the Stern Review passion for severe risk aversion the value η = 1 that is the low-
curtailment of greenhouse gas emissions to est lower bound of just about any economist’s
begin soon, because at current flow rates we best-guess range. Some other taste-parameter
will attain a stock of 550 ppm of CO2e with- values are considered in a halfhearted sensi-
in about a half-century and move (essential- tivity analysis postscript to the original ver-
ly irreversibly) thereafter beyond any hope sion of the Review, which is reported as if
to stabilize ultimate E[ΔT] at ≈ 3°C. indicating robustness but I would interpret
However, the point here is to put aside as more nearly the opposite because, no
sep07_Article4 8/17/07 4:31 PM Page 708

708 Journal of Economic Literature, Vol. XLV (September 2007)

matter what spin is put on it, there is no equivalent to about 1 percent of GDP
escaping the impact of higher interest rates (which seems rather on the low side by
on undoing the Review’s extreme policy con- maybe a factor of two or more, but that is not
clusions. With its preferred base-case param- so relevant here).
eter values δ = 0.1 percent p.a., g = 1.3 The question for the Stern Review analy-
percent p.a., η = 1, Stern’s discount rate sis then effectively becomes: is it worthwhile
from (1) is r = 1.4 percent. The present dis- to sacrifice costs C ≈ 1 percent of GDP now
counted value of a given global-warming loss to remove damages D ≈ 5 percent of GDP a
from a century hence at the non-Stern annu- century from now? With g and r being
al interest rate of r = 6 percent is one hun- expressed on an annual basis, the benefit-
dreth of the present discounted value of the over-cost ratio of such an investment would
same loss at Stern’s annual interest rate of be CB− = 5 exp(100(g − r)). From (1), r − g =
r = 1.4 percent. The disagreement over what (η − 1)g + δ, so that by picking the extreme
interest rate to use for discounting is equiva- values η = 1, δ = 0.1 percent, Stern guaran-
lent here in its impact to a disagreement tees that the difference r − g is always the
about the estimated damage costs of global miniscule amount δ = 0.1 percent, no matter
warming a hundred years hence of two orders what value of g is chosen, which is really
of magnitude. Bingo! stacking the deck in favor of approving such
If D is aggregate damages from global cli- kind of fractional GDP swaps across time.
mate change and Y is GDP, then values of (The Review could have made life easier
the ratio D−
Y a century from now (if nothing or here by just rounding down a mere tenth of
very little is done to halt greenhouse gas a percent by assuming δ = 0, which along
emissions) are commonly taken to be some- with η = 1 would make cost–benefit analysis
where in the range of about 0 percent to 3 really simple because a fixed fraction of
percent. The Stern Review effectively uses GDP would then always be worth the same

Y ≈ 5 percent as its base case. This high
fixed fraction of GDP at any future time.)
value is consistent with what an uncharitable With Stern’s preferred parameter values, the
critic might see as a philosophy of focusing benefit–cost ratio is − C = 4.5 (close to the

on the gloomier outcomes in a heuristic- upper bound of − C = 5 from assuming a zero


intuitive attempt to include extreme dam- rate of pure time preference)—a clear slam-
ages, because in Stern’s language “when we dunk accept. The alternative non-Stern val-
try to take due account of the upside risks ues g = 2 percent, r = 6 percent make
and uncertainties, the probability-weighted −
C = 10
clear reject. This simple kind of
costs look very large.” Actually, the Review exercise is what drives the Stern Review
goes well beyond 5 percent in its multi- results and, in a nutshell, is what accounts for
dimensional approach by making numerous the difference with the more conventional
literary and numerical allusions to the dark analyses of its critics. The no-frills stripped-
possibilities lurking in the tails of the distri- down variant of the Ramsey model I am
bution of possible outcomes (and then, as it using here is liable to a thousand and one
were, rubbing salt in the wound of numeri- legitimate questions and criticisms about its
cal calibration by noting how centrist it is oversimplifications but, at the end of the day,
actually being by not choosing much higher I believe this exercise is highlighting fairly
probability-weighted distant-future dam- what really counts in the economics of cli-
ages, which could be as big as DY− ≈ 20 mate change—the hidden discounting
percent–35 percent when one considers assumptions whose role tends to be more
catastrophes that might materialize after obscured than informed by the big IAMs.
2105). Stern also estimates the annual costs For most economists, a major problem
of its ambitious abatement strategy as being with Stern’s numbers is that people are not
sep07_Article4 8/17/07 4:31 PM Page 709

Weitzman: A Review of the Stern Review 709

observed to behave as if they are operating inferences for putting welfare weights on the
with δ ≈ 0 and η ≈ 1. To gauge the magni- utilities of one’s great-grandchildren, and
tude of the headache this presents for Stern’s there might be some sporadic support for
taste-parameter values, consider the follow- Stern’s preferred taste parameters scattered
ing thought experiment expressed in terms of throughout the literature, I ultimately find
the permanent income hypothesis in a deter- such an extreme stance on the primacy of
ministic setting. Suppose that on the margin δ ≈ 0, η ≈ 1 unconvincing when super-strong
an individual representing a long-lived policy advice is so dependent upon noncon-
dynasty faces a constant interest rate, r, and ventional assumptions that go so strongly
has a level of wealth, W, representing the against mainstream economics.
capitalized value of future earnings plus ini-
tial holdings. Then permanent income is rW 3. Puzzles and Ambiguities of Uncertain
and an optimal consumption trajectory saves Discounting
a constant amount s of permanent income.
Plugging the implied balanced growth rate The most worrisome omission from any
g = sr into (1) and rearranging gives analysis based on the Ramsey approach (1) is
uncertainty. As a first-pass informal cut at
(2) s = ⎯⎯ δ
r −⎯ uncertainty, suppose we admit that we don’t
ηr . really know for sure whether Stern or Stern’s
critics are right about the interest rate to use
With Stern’s preferred values δ ≈ 0, η ≈ 1, for discounting costs and benefits a hundred
equation (2) implies s ≈ 100 percent irre- years or so from now. An important feature
spective of r—a reductio ad absurdum. In of interest rates under uncertainty is that
the economics of uncertainty, plausible values they don’t aggregate arithmetically into a
of the coefficient of relative risk aversion η simple certainty-equivalent interest rate. A −12
are commonly taken to be somewhere chance of r = 6 percent and a −12 chance of
between 1 and 4 (I use the geometric-average r = 1.4 percent are not at all the same thing
point estimate η = 2). A reader can plug as splitting the difference by selecting the
favorite parameter values into (2) and back average r = 3.7 percent. It is not discount
out implied values of δ. For me (and I sus- rates that need to be averaged but discount
pect most economists), sensible savings rates factors. A −12 chance of a discount factor of
in this and other variants of market-behavior- e−6 a century hence and a −12 chance of a dis-
based thought experiments requires the rate count factor of e−1.4 a century hence make an
of pure time preference to be significantly expected discount factor of 0.5 e−6 + 0.5 e−1.4 a
greater than zero (or at least if δ is chosen to century hence, which, when you do the
be relatively small then η should be chosen math, is equivalent to an effective interest
to be relatively big). Stern’s worldview tends rate of r = 2 percent. According to this logic,
to blow off market-based observations and the interest rate we should be using to dis-
behavioral inferences as being (for a variety count a dollar of costs or benefits a century
of reasons including market incompleteness) from now is in between the Stern value of
largely irrelevant to long-run discounting, r = 1.4 percent and the more conventional
which should instead be based primarily value of r = 6 percent, but with the above
upon the “ethical” value δ ≈ 0 that Stern numbers it is a lot closer to the Stern value
imposes on a priori grounds. Readers will and is not anywhere near the arithmetic
have to make up their own minds about “eth- average of r = 3.7 percent. More generally
ical” values of preference parameters. While here, if there is a subjective probability pi
there may be something to Stern’s position that discount rate ri is the correct rate to use,
about the limited relevance of market-based then the effective discount rate for time t is
sep07_Article4 8/17/07 4:31 PM Page 710

710 Journal of Economic Literature, Vol. XLV (September 2007)

ln∑ pi e−ri t way to begin this process is by making the

(3) r(t) = − ⎯⎯⎯⎯t , growth rate be a random variable.
Continuing here in the spirit of being sim-
which declines monotonically over time ple, suppose that the growth rate g in any
from the expected interest rate r(0) = ∑ piri given year is i.i.d. normal with known mean
to an asymptotic limit of r() = miin{ri}. The  and known variance  2. (The fact that 
moral of this story is that the Stern value may and  2 are known will later become signifi-
end up being more right than wrong when full cant when we inquire what happens under
accounting is made for the uncertainty of the greenhouse warming when  and  2 are
discount rate itself, which arguably is the most modeled as not known.) With g ∼ N(, 2),
important uncertainty of all in the economics the Ramsey formula (1) becomes
of climate change. The very same force of
compound interest that makes costs and ben- (4) r f = δ + η − 1− η 2 2,
efits a century from now seem relatively 2
insignificant, and that additionally creates the
“majority tilt” of a pain-postponing climate where r f in equation (4) denotes the risk free
policy ramp of emissions reductions starting interest rate. The introduction of uncertainty
from a low gradual base, also forces us to rec- also allows consideration of a risky asset with
ognize the logic that over such long periods a different rate of return. Following the
we should be using interest rates at the lower asset-pricing expository literature, suppose
end of the spectrum of possible values. we model comprehensive or representative
In the certain world of the Ramsey deter- equity at a high level of abstraction as being a
ministic formula (1), there is no distinction claim on the consumption dividend pro-
among rates of return on various assets and r duced by the macroeconomy itself. Suppose
is just the economywide rate of return on this abstract macroeconomy is represented
capital or, more succinctly, the interest rate. by a dynamic stochastic general equilibrium
In nondeterministic reality, there are many in the Lucas–Mehra–Prescott fruit-tree
rates of return out there and they differ con- model. Let the random variable Re be the
siderably. The point has already been estab- gross arithmetic return on equity while
lished that it makes a tremendous difference r e = lnRe is the more familiar geometric rate
for long time periods of a century or more of return on equity. When g is i.i.d. N(, 2)
what interest rate is used for discounting. To in this fruit-tree economy, the equity risk
understand better which discount rate to use, premium over the safe rate then reduces to
we need to enrich the Ramsey model by for- the well-known expression
mally introducing uncertainty, which allows
(5) −
r e − r f =  2,
us at least to distinguish between rates of
return on capital from two fundamentally dif-
where − r e is defined by the oblique-looking
ferent sorts of investments: a risky economy-
expected-value formula − r e ≡ lnE[Re], which
wide rate of return applicable to investments e
is close enough to E[r ] to make them inter-
that have payoff characteristics parallel to the
changeable for my purposes here.4
economy itself and a risk free rate of return
Combining (5) with (4) gives the average
applicable to investments whose payoffs are
return on equity as
orthogonal to the economy as a whole. After
that, we need to decide which of these two −
rates is more appropriate for discounting (6) r e = δ + η − 1− η 2 2 + η 2.
costs and benefits of mitigating climate
change. Then we need to plug in numbers 4 The formulas (4) and (5) are explained in most
and see what happens. The simplest formal graduate-level textbooks covering asset pricing.
sep07_Article4 8/17/07 4:31 PM Page 711

Weitzman: A Review of the Stern Review 711

Extending the previous “trio of twos” param- the oversimplified two-period formulation
eter values to a not-implausible knee-jerk here, a project to mitigate the effects of
“quartet of twos” δ = 2 percent, η = 2, global warming incurs consumption costs in
E[g] = 2 percent,  [g] = 2 percent (on an the present period by curtailing CO2e emis-
annual basis, for long time series) makes sions, investing in costly new technologies,
very little difference on the risk free rate and so forth, but consumption in the future
because now r f = 5.9 percent in (4) instead of period is increased by having reduced the
the previous value for “the” interest rate of detrimental impacts at that time from
r = 6 percent in (1). The corresponding greenhouse warming. The payoff is the extra
equity premium from (5) is − r e − r f = 0.1 per- consumption available in the distant-future
cent and the average return on equity from period. Suppose that the correlation coeffi-
(6) is −
r e = 6 percent. The actual empirical cient between the increased output of the
numbers are closer to r f ≈ 1 percent, − re − project and returns to the economy as a

r ≈ 6 percent, r ≈ 7 percent. (The calibra-
f e whole is β. An investment beta is intended
tion r f ≈ 1 percent refers to short-term to represent a correlation coefficient that
treasury bills, while −r e ≈ 7 percent refers to applies to discount factors as contrasted
overall returns on comprehensive indexes of with discount rates (i.e., here β is the corre-
publicly traded shares of common stocks, but lation between the investment payoff and
I don’t think the numbers would be funda- Re, not r e). It then follows from essentially
mentally different for other empirical meas- the same considerations as went into deriv-
ures of returns from investments for the ing formula (3) that the relevant interest
economy as a whole.) So with the not- rate for discounting costs and benefits at
implausible “quartet of twos” parameter val- time t here is
ues the theory does a decent job of
ln[β exp(− −
r et) + (1 − β)exp(− r ft)]
predicting the average return on equity but (7) r(t) = − ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ t ⎯,
fails miserably on the risk free rate and the
equity premium—thereby giving rise to the which declines monotonically over time
notorious “risk free rate puzzle” and the even from r(0) = β −r e + (1 − β)r f to an asymptotic
more notorious “equity premium puzzle.” limit of r () = r f. So the question here
What does all of this have to do with the becomes: what is the right β for the kinds of
economics of climate change? Well, a lot projects that the Stern Review has in mind
actually. But before getting into the relation- for mitigating global warming?
ship between the asset-return puzzles and Overall damages from climate change are
the economics of climate change, we need to modeled in most IAMs, including the
put the puzzling numerical mismatches tem- PAGE model that crunches some numbers
porarily aside in favor of first asking a funda- for the Review, as a pure production exter-
mental prenumerical question: in principle nality equivalent to losing output via a par-
(leaving aside their correct numerical val- ticular subaggregator equation of the
ues) should we be using the risk free rate or multiplicative form
the risky economywide rate of return for dis-
counting costs and benefits of climate (8) D(t) = Y∗(t) − Y(t) = f(ΔT(t))Y∗(t),
The issue of which rate of return to where t is time, D is the total damages of
choose (as between r f and −r e) for discount- greenhouse warming, ΔT is atmospheric
ing a project comes down to the extent to temperature relative to the base period, Y∗ is
which the payoffs from the project are pro- potential GDP (or NDP, no distinction being
portional to or independent from returns to made here) in the absence of any green-
investments for the economy as a whole. In house warming, and Y is actual GDP with
sep07_Article4 8/17/07 4:31 PM Page 712

712 Journal of Economic Literature, Vol. XLV (September 2007)

greenhouse warming. The standard func- Like Stern, the essentially identical earlier
tional form actually chosen in most IAMs is formulation of Cline used parameter values
f(ΔT) = k(ΔT)γ for some coefficients γ and k, that made the Ramsey formula (5) deliver a
where typically γ = 2 (quadratic loss in tem- low interest rate—in Cline’s case the assumed
perature change). The parameter k is usual- parameter values were δ = 0 percent, η = 1.5,
ly calibrated so as to make −D
Y a century hence g = 1 percent, which combined to make the
under mild or no abatement (with interest rate be r = 1.5 percent per year. Also
ΔT ≈ 2.5˚C) be somewhere between approx- like Stern, the strong activist conclusions of
imately 0 percent and approximately 5 per- Cline’s analysis fifteen years earlier traced
cent (depending on who is doing the back to the very low discount rate being
calibrating). There is no question here about used. Furthermore, Cline and Stern are soul-
the value of beta implicit in the multiplica- mates in their cri de coeur justifying δ ≈ 0 by
tive formulation (8): it is one! Therefore, by relying mostly on a priori philosopher-king
the very logic of the IAM used by the Stern ethical judgements about the immorality of
Review itself, the interest rate for discount- treating future generations differently from
ing costs and benefits should be the returns the current generation—instead of trying to
to the economy as a whole, − r e. This still back out what possibly more representative
leaves open the question of which rate to use members of society than either Cline or
for −r e—the empirical returns on a broad Stern might be revealing from their behavior
index of publicly traded shares of stocks of is their implicit rate of pure time preference.
about 7 percent (representing economy- An enormously important part of the “disci-
wide average returns and used, e.g., by the pline” of economics is supposed to be that
Congressional Budget Office for evaluating economists understand the difference
U.S. government projects) or the value of between their own personal preferences for
6 percent predicted by formula (5) from apples over oranges and the preferences of
my non-Stern “quartet of twos” parameter others for apples over oranges. Inferring
values—but the discrepancy between 6 per- society’s revealed preference value of δ is not
cent and 7 percent is insignificant for pur- an easy task in any event (here for purposes
poses here. Whatever number is used for − r e, of long-term discounting, no less), but at
if it in any reasonable way represents the least a good-faith effort at such an inference
returns to the economy as a whole then it will might have gone some way towards convinc-
completely undo the Review conclusions ing the public that the economists doing the
about drastic consumption smoothing and studies are not drawing conclusions primari-
bring the results back to the much more mod- ly from imposing their own value judge-
erate take-it-more-slowly climate-policy-ramp ments on the rest of the world.
time profile of emissions reductions advocated In part because Cline’s results, and where
by most mainstream critics of Stern. they were coming from, were more trans-
This important dispute about what interest parent (largely from not being buried within
rate to use for discounting costs and benefits a big mysterious IAM, which was not yet
of mitigating greenhouse warming duplicates readily available around 1990), his study
the same debate about the same subject more attempted to seize the analytical high
than a decade ago between William R. Cline ground by emphasizing that an assumed
and Nordhaus, two early pioneers of model- annual interest rate of r = 1.5 percent is
ing the economic effects of climate change.5 calibration-consistent with the real return on
relatively safe U.S. Treasury bills historically
5 Cline (1992), Nordhaus (1994). See also the later
being about 1 percent or so per annum.
studies and reflections on discounting for climate change Missing from Cline’s reasoning was a serious
contained in Paul R. Portney and John P. Weyant (1999). discussion of the implications of risk and of
sep07_Article4 8/17/07 4:31 PM Page 713

Weitzman: A Review of the Stern Review 713

payoff correlations for the choice of a dis- global warming involve its “outdoor” aspects
count rate that might justify using r = r f. (broadly defined) like agriculture, coastal
Nordhaus, whose careful pragmatic model- recreational areas, and natural landscapes
ing throughout his DICE series of IAMs has (including the existence value of ecosystems,
long set a standard in this arena, argued species, and so forth). Climate-affected “out-
forcefully over a decade ago that the risk free door” activities may be differently impacted
interest rate should not be used for dis- by greenhouse warming than “indoor” eco-
counting costs and benefits of climate nomic activities constituting the bulk of the
change. In this argument, Nordhaus was fol- economy, which are largely going to be dom-
lowing Robert Lind who, in a comprehen- inated by the unknown future growth rate of
sive summary of an influential book he labor-augmenting technological progress.
edited in 1984 entitled Discounting for Time Instances of changes in “outdoor” activities
and Risk in Energy Policy, concluded that under global warming include what happens
“unless there is substantial evidence to the to tropical agriculture, losing significant parts
contrary, the returns associated with public of Bangladesh (or Florida) to rising sea lev-
projects should be assumed to be highly cor- els, the “consumption” of an altered natural
related with returns to the economy as a world that is a direct argument in the utility
whole” (p. 77). function, and so forth. These kinds of
All of this having been said, there was changes, which include the existence value of
never any deep economic rationale in the natural environments, are presumably not
first place for damages from greenhouse gas highly correlated with technological progress
warming being modeled as entering utility in computing power, furniture making, or
functions through the particular reduced- better pharmaceuticals a century from now.
form route of being a pure production exter- The relevant share of the “outdoor” subset of
nality that substitutes perfectly with output the economy in investment-beta calculations
according to the multiplicative subaggrega- might be disproportionately large because it is
tor function (8). It was more due to an his- disproportionately largely impacted by green-
torical accident of stumbling upon a simple house warming. Furthermore, it might plausi-
understandable analytical form whose bly be argued that the high income elasticity
parameters could be conveniently adjusted of environmental awareness will make for a
to match various scenarios than the result of high existence value of unaltered natural habi-
serious thought about whether damages tats when per capita incomes have increased
from global warming are better specified as ten-fold over the course of a century or more.
multiplicative or additive with GDP, or even What happens to the discount rate for cli-
entering the utility function as a direct argu- mate-change investments naturally depends
ment (rather than substituting one-for-one on the actual value of β that is assumed. If
with economic output)—all of which would β = 0 in (7), then r(t) = r f = 1 percent. If
have been seen as a secondary issue. So, with β = 1 in (7), then r(t) = −r e = 7 percent. The
the benefit of hindsight, let us now ask: Is more interesting question concerns what
there any economic rationale by which green- happens to r(t) for in-between values of β.
house-warming damages are as much uncor- Suppose for the sake of argument we split
related as they are correlated with aggregate the difference and imagine that the dispro-
economic activity? The answer, when you portionate impact of climate change on
think about it, is yes. No one has ever tried to generalized-land-usage “outdoor” activities
argue that the effects of global warming will of the economy warrants an overall correla-
be evenly spread among regions of the world tion coefficient of, say, β ≈ 0.5. With β ≈ 0.5
or sectors of the economy. The parts of an in (7), the relevant interest rate for a centu-
economy likely to be most impacted by ry from now becomes r(100) = 1.7 pecent,
sep07_Article4 8/17/07 4:31 PM Page 714

714 Journal of Economic Literature, Vol. XLV (September 2007)

which is close to Stern’s r = 1.4 percent or The relevant discounting rate r(t) from (7)
Cline’s earlier r = 1.5 percent. In this case then lies between 5.9 percent and 6 percent
investments for mitigating global climate independent of the assumed value of β.
change become attractive as an insurance When 5.9 percent ≤ r(t) ≤ 6 percent, the
policy that secures food supplies, preserves Review conclusions are again undone and
coastal areas, and maintains natural environ- the more orthodox mainstream policies of
ments in a world where future aggregate moderate greenhouse gas slowing in the
growth rates are uncertain. I am not trying to near future come back. The practical ques-
defend this particular formulation or the tion of what interest rate to use for discount-
particular value β = 0.5. Rather, the moral of ing costs and benefits of climate change thus
this story is that the nature of the impacts of becomes intertwined with the interpretation
climate change determine whether we should of the equity premium and risk free rate
end up closer to using the risk-free rate or the puzzles. It is a measure of how deep and
economywide return on capital—and there serious these puzzles are that even after
are plenty of stories suggesting that the rele- thousands of articles there is still no agreed-
vant investment beta here is significantly less upon resolution of them. If we use numbers
than one. When the overall discount factor is that resolve the puzzles in the descriptive
a combination of more primitive discount direction, then r is sensitive to β and r ≈ 1.7
factors (as is the case here when the correla- percent for β = 0.5. If we use numbers that
tion coefficient β is some midrange value resolve the asset-return puzzles in the pre-
between zero and one), the risk free interest scriptive direction, then r ≈ 6 percent inde-
rate, which is close to the Stern interest rate, pendent of β. And, to whip a horse long
then may well end up being more right than dead, it makes a huge difference to the eco-
wrong. Over a time horizon of a century or nomics of climate change whether r ≈ 1.7
so, this “midrange β effect,” which is not percent or r ≈ 6 percent.
implausible when one considers the highly- Critics of the Stern Review are fond of
uneven impacts of greenhouse warming on pointing out that δ ≈ 0, η ≈ 1 is inconsistent
the different regions and sectors of the world with observed economic behavior, especial-
economy, can be a strong factor in lowering ly savings behavior. While this is true, and
discounting rates significantly—from the it is a genuine problem for Stern, it is just
same underlying analytical source as the the tip of an iceberg that threatens all such
force of compound interest and the logic of formulations—not just Stern’s. The biggest
the climate-policy ramp. Remarkably, the and most troubling disconnect between the
big IAMs with their casually built-in specifi- prescriptive numbers that theory says we
cation of β ≈ 1 obscure rather than clarify should be using for discounting and the
the critical role in climate-change analysis of descriptive discount-rate numbers that are
assumptions about investment betas. actually out there concerns the asset-return
Next, suppose we try to repeat the above puzzles. These puzzles very strongly suggest
numerical exercise but in place of the empir- that something fundamental is amiss in the
ical values r f = 1 percent, − r e = 7 percent, we paradigm framework for pricing assets and
use the values predicted by the theoretical deriving the rates of return that we are rely-
formulas via assuming the quartet of twos ing upon to produce discount rates for eval-
parameter values, which then implies r f = 5.9 uating new investment opportunities. For
percent from (4) and − r e = 6 percent from (6). example, perhaps the taste parameters δ and
Because the equity premium predicted η that we are commonly using (here δ = 2
from (5) is a miniscule 0.1 percent, there is percent p.a. and η = 2 ) are wrong. If we
essentially no difference in this case treat (4) and (5) as two equations in two
between r f = 5.9 percent and − r e = 6 percent. unknowns (δ and η), we can then invert the
sep07_Article4 8/17/07 4:31 PM Page 715

Weitzman: A Review of the Stern Review 715

two equations to back out the hypothetical climate-change-affected events a century or

values δ̂ and η̂ that would “explain” the so from now.
stylized-fact empirical observation that r f ≈ 1
percent and − r e − r f ≈ 6 percent. When this is 4. Uncertainty Tends to Matter Much More
done (for  = 2 percent,  = 2 percent), it than Risk
produces the mega-puzzle that the estimat-
ed rate of pure time preference is δ̂ ≈ 151 If the conclusion from the last section—that
percent per year and the coefficient of rela- what to do about global warming depends
tive risk aversion is η̂ ≈ 150. One does not overwhelmingly on the imposed interest
know whether to laugh or to cry at the rate—is seen as disappointing, then a second
prospect of what the Stern Review IAM conclusion is likely to seem downright
might end up recommending as its preferred unnerving. As noted, the choice of appropri-
policy for climate change in its number- ate discount rate is itself extraordinarily sen-
crunching simulations if the parameter val- sitive to seemingly arcane modeling details
ues δ̂ ≈ 151 percent, η̂ ≈ 150 were fed into like the value of the climate-change invest-
PAGE. So much for the fantasy that values of ment beta and how the asset-return puzzles
the taste parameters δ and η should be chosen are resolved. One interpretation of the asset-
to be consistent with the revealed-preference return puzzles, which could also have some
observed stylized facts of economic behavior! relevance for the economics of climate
At the end of the day, where do these change, is the idea that investors are dispro-
dizzying and disconcerting numerical exercis- portionately afraid of rare disasters. These
es leave us with respect to the economics of rare disasters are not fully reflected in the
climate change? One inescapably strong con- available data samples that, being limited,
clusion is that the emissions reductions that are naturally deficient in coverage. Besides,
go along with optimal growth under endoge- even if we had an infinite time series of past
nous climate change are extraordinarily sensi- observations, they are of restricted relevance
tive to the interest rate that has implicitly in an evolving world whose features are
been built into whatever model is being used always changing and whose past never fully
for the analysis. The present discounted value repeats itself. With this interpretation of the
of a future cost (or benefit) is the product of puzzles, people are willing to pay high pre-
an imposed discount factor times the project- miums for relatively safe stores of value that
ed future cost (or benefit). Trying to forecast might represent “catastrophe insurance”
costs and benefits of climate change scenarios against out-of-sample or newly evolved rare
a hundred years or so from now is more the disasters.6 Such an ongoing catastrophe-
art of inspired guesstimating by analogy than insurance effect could readily explain why
a science (imagine forecasting today’s world a observed r f is so low relative to the observed
century ago). But in my opinion the unsure past average of realized r e.
prediction of future costs and benefits of cli- There is little doubt that the worst-case
mate change a century or two hence is over- scenarios of global-warming catastrophes are
shadowed by the unsure interest rate to use genuinely frightening. The Stern Review
in the discount factor, which makes the dis- goes over several of these highly unlikely,
count factor more uncertain than predicted poorly understood threshold-crossing disas-
costs (or benefits) of climate change by about ters associated with abrupt large-scale irre-
an order of magnitude. Of the two multipli- versible changes in the climate system:
cands in the product of a discount factor
6 The theme of catastrophe insurance and the underly-
times an expected cost (or benefit), empiri-
ing motivation for the treatment of structural uncertainty
cally it is the discount-factor uncertainty that as tail thickening of posterior-predictive distributions is
looms much larger in practice for analyzing developed in Martin L. Weitzman (forthcoming).
sep07_Article4 8/17/07 4:31 PM Page 716

716 Journal of Economic Literature, Vol. XLV (September 2007)

sudden collapse of the Greenland and West unavoidable that the reduced-form probabil-
Antarctica ice sheets, weakening or even ity of ΔT > 6˚C increases substantially above
reversal of thermohaline circulations that 3 percent after the next century just from
might radically affect such things as the the enormous inertial lags for what by then
Gulf Stream and European climate, run- will be in the climate-change pipeline.
away climate-sensitivity amplification of Societies and ecosystems whose average
global warming due to positive-reinforcing temperature has changed in the course of a
multiplier feedbacks (including, but not lim- century or so by ΔT > 6˚C (for U.S. readers:
ited to, loss of polar albedo, weakened car- Δ6˚C ≈ Δ11˚F) are located in the terra incog-
bon sinks, and rapid releases of methane nita of what any honest economic modeler
from the thawing of arctic permafrost). would have to admit is a planet Earth recon-
More gradual but still very serious examples figured as science fiction, since such high
of uncertain climate-change effects are: sea- temperatures have not existed for some tens
level dynamics, drowned coastlines of of millions of years.
unknown magnitude, very different and pos- The idea behind analyzing climate-
sibly extreme weather patterns including change projects by converting future costs
droughts and floods, ecosystem destruction, and benefits into present discounted values
mass species extinctions, big changes in is that society has alternative investment
worldwide precipitation patterns and distri- opportunities, whose proxy rate of return is
bution of fresh water, tropical-crop failures, the discount rate, representing alternative
large-scale migrations of human populations, capital-accumulation opportunities through-
humidity-nourished contagious diseases—and out the rest of the economy that would
the list goes on and on. compensate us for the economic losses suf-
Translated into the language of the simple fered from climate change. Human-capital
model used here, such rare disasters are far investments in education or public health
out in the right tail of very high ΔT, which have consistently been found to have high
corresponds to being far out in the left tail of rates of return, arguably far greater than 10
the consumption-growth random variable g. percent for less-developed countries and
The probability distribution of long-run ΔT regions. More mundane examples of alter-
is disturbingly spread apart, largely because natives to CO2e mitigation from middle-of-
of structural-parameter uncertainty about the-probability-distribution mild warming
the unknown “climate sensitivity” multiplier might include accumulating air conditioners
that amplifies greenhouse gas concentra- to counter high temperatures or erecting
tions into ultimate steady-state greenhouse sea walls to keep the rising ocean out of
warming. The recently released Fourth coastal cities. Such alterative investments
Assessment Report of the IPCC (2007) pre- compensate mostly for potential loss of
dicts for one hundred years from now a “indoor” consumption and they tend to be a
mean temperature change of further plane- lot less expensive than wholesale abatement
tary warming (from averaging six “equally of greenhouse gases. The real problem is in
sound” marker scenarios) of E[ΔT] = 2.8˚ C the tails and it mostly concerns “outdoor”
with a thick-tailed upper-end standard devi- consumption. If the definition of consump-
ation ≈ 1.6˚C (table SPM-3). This means the tion is broadened (as it should be) to
probability that ΔT > 4.5˚C is approximately include nonmarket enjoyment of the natural
15 percent and the probability of ΔT > 6˚C is environment—like habitats, ecosystems,
very roughly about 3 percent. IPCC does not and species—then it is difficult to imagine
extend its projections beyond 2105 on the what the compensating investments are for
basis that predictions into the twenty-second which we should now be saving more as an
century are too uncertain, but it seems alternative that might substitute for holding
sep07_Article4 8/17/07 4:31 PM Page 717

Weitzman: A Review of the Stern Review 717

down ΔT directly. With roughly 3 percent a cost–benefit calculation of what to do

IPCC-4 probability, we will “consume” a about it is the greatest exercise in Bayesian
terra incognita biosphere within a hundred decision theory that we economists have ever
years whose mass species extinctions, radical performed. Formally, of course, cost–benefit
alterations of natural environments, and analysis can deal with uncertainty—by taking
other extreme outdoor consequences of a expected values, relying on expected-utility
different planet will have been triggered by theory, accounting for risk aversion, and
a geologically-instantaneous temperature using all of the other, by now familiar, para-
change that is significantly larger than what phernalia of the modern theory of the eco-
separates us now from past ice ages. nomics of uncertainty. In principle, it does
In the rest of this paper, marginal analysis not matter whether the probabilities that
is set aside and g stands for the unknown show up in our cost–benefit calculations are
growth rate of a comprehensive future “con- objective or subjective because the mathe-
sumption” that includes the consumption of matical formulas are the same for either
natural environments, ecosystems, species, case. But in lumping together objective and
and the like. The cost of low-g disasters from subjective uncertainties and thereby obscur-
high-ΔT scenarios more properly constitutes ing their distinction—to the extent that a
uncertainty in the sense of Knight or Keynes graduate student today hardly knows, or
than risk, because the scale and probability even cares, what kinds of probabilities are
of these disasters are both unknown. Not legitimate to plug into a rational expectations
only is it very difficult to estimate tail proba- equilibrium and what kinds of probabilities
bilities of high-ΔT outcomes—due, ultimate- are illegitimate for such purposes—I think
ly, to the underlying sampling-theory that contemporary macroeconomics goes too
principle that the rarer is an event the more far and leads to a mindset that too easily
unsure is our estimate of its probability—but identifies probability (and “economic sci-
translating this into g-equivalent economic- ence”) with exercises in calibration to sample
damage units introduces enormous further frequencies from past data.
fuzziness, especially when g includes exis- I do not propose to rehash here the ages-
tence values of natural habitats. With an evo- old, never-resolved foundational controversy
lutionary stochastic process like global about whether probabilities are better con-
climate change, the world is not standing still ceptualized on the most fundamental level as
long enough for us to accumulate the rele- objective frequencies or subjective beliefs.
vant information to accurately assess tail Personally, I do not think there exists a pure
probabilities. The net result is thicker left case of either extreme pole, but rather there
tails for the distribution of g under dynami- is a continuum of situations with some being
cally evolving global climate change than we closer for practical modeling purposes to the
are accustomed to dealing with in our much objective pole and others being closer for
more familiar dynamic stochastic general practical modeling purposes to the subjec-
equilibrium macro models, which in practice tive pole. Here I just want to point out that if
are based upon the stationary thin-tailed sto- something like radioactive decay is close to
chastic processes that we use to model a being a pure case of objective frequencies,
rational expectations equilibrium whose then climate change, and especially the eco-
structure is (supposedly) fully known and nomics of climate change, is as close to being
understood. a pure case of modeling probabilities by sub-
Every cost–benefit analysis is an exercise jective judgements as we economists are
in subjective uncertainty. If, as the Stern ever likely to encounter in practice. To para-
Review puts it, “climate change is the great- phrase the language of the Stern Review yet
est externality the world has ever seen,” then again, the economics of climate change is the
sep07_Article4 8/17/07 4:31 PM Page 718

718 Journal of Economic Literature, Vol. XLV (September 2007)

greatest application of subjective uncertainty but, far more importantly in this context, it
theory the world has ever seen. thins the left tail of the distribution as well.
To the extent that it makes any sense at all The thickened tails of the reduced form
to think in terms of some approximately bell- of the distribution of g that are an inevitable
shaped meta-distribution of growth rates g consequence of taking expectations of
that is out there, the part of the probability expectations can have surprisingly strong
distribution that corresponds most closely to effects on cost–benefit calculations by low-
objective-frequency risk is in its body around ering significantly expected utility and rais-
the middle because, from previous experi- ing significantly expected marginal utility.
ence, past observations, plausible extrapola- To get a sense of just how strong the effect
tions, and maybe even the law of large can be of tails thickened by having structur-
numbers, we have at least some modicum of al parameters that we do not know but
confidence in being able to construct a rea- whose values must be inferred indirectly
sonable approximation of the central regions from limited experience—and therefore a
of the probability distribution. As we move sense of how much we could be missing in
toward probabilities in the tails of the g dis- our economic analysis by ignoring the terra
tribution, however, we are increasingly mov- incognita of the greenhouse-warming
ing into the unknown territory of subjective extremes—consider this prosaic example.
uncertainty where our probability estimates Suppose that in the good old days before we
of the probability distributions themselves understood human-induced climate change
becomes increasingly diffuse because the we were sure that g ∼ N(, 2), where we
frequencies of rare events in the tails cannot somehow knew that  = 2 percent and  = 2
be pinned down by previous experiences, percent. Normalize current marginal utility
past observations, or computer simulations. to be unity. Then from using the familiar for-
The upshot of this uncertainty about uncer- mula for the expectation of a lognormally
tainties is that the reduced-form probability distributed random variable, the expected
distribution of g (after integrating out the marginal utility of an extra sure unit of con-
probabilities of probabilities)—which is a sumption in the pre-climate-change era
reduced form for the economics of climate would have been EMU = E[exp(− ηg)] =
change in the sense that g here is the growth exp(− η + − 2 η  ). (It is precisely this kind of
1 2 2

rate of comprehensive consumption that calculation that lies behind the risk free rate
includes the natural environment—has a and equity-premium formulas (4) and (5).)
thick left tail. The exact thickness of this left Imagine next that the possibility of green-
tail of g depends not only upon how bad an house warming has now made us unsure
environmental catastrophe global warming about  and . Let us preliminarily model
might induce and with what probabilities, this greenhouse-warming-induced uncertain
but also upon how imprecise are our prob- situation, where we don’t know the true val-
ability estimates of the probabilities of ues of  and  because of limited experience
those bad catastrophes. Uneasiness about with climate change, as if we are limited
projecting uncertain uncertainties prevents because we only have data from some finite
IPCC and most economic analyses from number n of past observations (or finite sim-
taking a stand on the increasing—and ulation outcomes from the data generating
increasingly diffuse—probabilities of extreme process of some model) and we run a regres-
temperatures after the year 2105, which sion to estimate  and . For simplicity, sup-
hardly eliminates the underlying problem. pose further that the point estimates ˆ = 2
Mitigating the future consequences of percent and ˆ = 2 percent from this regres-
greenhouse warming does not just shift the sion just so happen to be the very same num-
center of the distribution of g to the right bers as the presumed-known population
sep07_Article4 8/17/07 4:31 PM Page 719

Weitzman: A Review of the Stern Review 719

parameters for the normal distribution There is a general point here and a partic-
before climate change was understood to be ular application to the economics of climate
a possibility. Then the reduced-form situa- change. The general point is that from expe-
tion is as if g is distributed as a Student-t dis- rience alone one cannot acquire sufficiently
tribution with n − 1 degrees of freedom. The accurate information about the probabilities
Student-t here has the same mean as the of tail disasters to prevent the expected mar-
normal and for large n has almost the same ginal utility of an extra sure unit of consump-
standard deviation, but if you look closely tion from becoming unbounded for any
with a magnifying glass its tails are naturally utility function having everywhere-positive
thickened due to the “true” values of the relative risk aversion, thereby effortlessly
structural parameters  and  being uncer- driving cost-benefit applications of expected
tain. This kind of structural uncertainty utility theory. The degree to which this kind
about the parameters of the probability dis- of “generalized precautionary principle” is
tribution spreads apart the reduced-form relevant in a particular application must be
(“predictive posterior” in Bayesian jargon) decided on a case-by-case basis that
distribution of g, an effect that is especially depends upon the extent to which a priori
pronounced in the thickened tails because knowledge in a particular case limits the
they are especially difficult to learn about. If extent of posterior-predictive tail thickening.
we now calculate the expected marginal util- In the particular application to the econom-
ity of an extra sure unit of consumption using ics of climate change, where there is so obvi-
this Student-t distribution (which is a natural ously limited data and limited information
manifestation of limited experience or limit- about the global catastrophic reach of cli-
ed information), then EMU = E[exp(− mate extremes for the case ΔT > 6˚C, to
ηg)] = + , which is mathematically equiva- ignore or suppress the significance of rare
lent to the fact that the moment generating tail disasters is to ignore or suppress what
function of a Student-t distribution is infi- economic theory is telling us loudly and
nite. The bombshell fact that EMU = +  (as clearly is potentially the most important part
soon as we admit that we don’t know the of the analysis. While it is always fair game to
underlying stochastic structure, and there- challenge the assumptions of a model, when
fore parameters must be estimated) changes economic theory provides a generic result
the rules of the game. Such a mechanism, (like “free trade is Pareto optimal”) the bur-
for example, explains the asset-return puz- den of proof is commonly taken as resting on
zles for reasonable values of δ and η as being whomever wants to overturn the theorem in
due to a fear of relatively rare tail disasters a particular application. The take-away mes-
that is theoretically difficult or impossible to sage here is that the burden of proof in the
eliminate when the underlying tail-structure economics of climate change is presumptive-
remains uncertain. The fact that under ly upon whomever wants to model optimal-
structural uncertainty EMU = +  represents expected-utility growth under endogenous
a mathematically generic result not limited greenhouse warming without having structur-
to isoelastic utility or the normal parent dis- al uncertainty tending to matter much more
tribution and Student-t child distribution of than risk. Such a middle-of-the-distribution
the example. I claim this general result has modeler needs to explain why the
significant economic repercussions which inescapably thickened tails of the posterior-
are not easily brushed aside, not least of all predictive distribution, for which the thick
for cost–benefit analysis of climate change left tail of g represents rare disasters under
because such an effect in principle over- uncertain structure, is not the primary focus
shadows the discounting of far-future of attention and does not play the decisive
events. role in the analysis.
sep07_Article4 8/17/07 4:31 PM Page 720

720 Journal of Economic Literature, Vol. XLV (September 2007)

5. Climate Uncertainty and the Value of the “ethical judgements on the distribution
Information of income” is Stern-speak for picking η ≈ 1.
Such “ethical judgements” could appear to
Because the Stern Review is imbued with an uncharitable critic as if designed to justi-
the laudable moral imperative of not expos- fy the activist conclusions that are consid-
ing future generations to the tribulations of ered necessary to avoid the climate-change
global warming, it does not shy away from horror scenarios. In the context of its self-
emphasizing (at least discursively in its “mul- imposed “multidimensional” methodology,
tidimensional” text) the possibilities of rare by choosing η ≈ 1 the Review appears to be
high-ΔT, low-g catastrophes from climate playing both sides of the street against the
change. Indeed, reading between the lines middle. On the one hand, it wants η to be as
of the report, one has the feeling that the high as possible to reflect its tremendous
immorality of relegating future generations humanitarian concern with distributional
to live under the shadow of the open-ended inequities across space, which would allow it
possibilities of uncertain large-scale changes to argue (informally, if passionately, in scat-
in the climate system, when for a mere tered prose and numerical examples) that the
annuity cost of a percent or two (or at most disproportionate negative impact of climate
three) of GDP each year we might have pur- change on the world’s poor (whose marginal
chased an insurance policy on their behalf utility is high because η in this story is
that avoided this scary uncertainty (or at implicitly large) calls for urgent action now
least greatly reduced it), is a major underly- to avoid future massive spatial redistribution
ing leitmotif of the Review. This feeling of of relative income from the poor to the rich.
guilt has no place to go analytically (under Simultaneously, the Stern Review wants to
the conventional analytical confines adopted further exacerbate distributional inequities
by the IAMs, including PAGE), so to speak, by redistributing income across time from
except to be subliminally channelled into the relatively poor present to the much-
choosing such low values of δ ≈ 0 and of richer future a century or two from now
η ≈ 1 (and, secondarily, such high values of (when standards of living are likely to be ten

Y ≈ 5 percent and low values of −
Y ≈ 1 per-
times higher) via—an uncharitable critic
cent) as will operate through the back door might suspect—choosing the lowest imagi-
of conventional economic analysis to weight nable value η ≈ 1 that might be used (along
present-discounted future damages high with δ ≈ 0) to reverse-engineer the really low
enough relative to present mitigation costs r needed to prop up its technical case for
to make the IAM want to reduce substan- immediate urgent action. The same contra-
tially the disastrous possibilities. The Review diction about values of η shows itself in
puts it directly: “Averaging across possibili- Stern’s heuristic justification for big values of
ties conceals risks. The risks of outcomes probability-weighted − D
Y as being due, in
much worse than expected are very real and effect, to a risk-averse high-curvature high-η
they could be catastrophic. Policy on climate utility function interacting with highly
change is in large measure about reducing uncertain damages.
these risks. They cannot be fully eliminated, I think that rather than trying to go
but they can be substantially reduced. Such though the back door with unreasonably low
a modeling framework has to take account of values of δ and η (or, secondarily, “averaging
ethical judgements on the distribution of across possibilities” by heuristically making
income and how to treat future generations.” business-as-usual − Y ≥ 5 percent instead of

The “ethical judgements” in the above some smaller more plausible point estimate),
quote about “how to treat future genera- it is much better to go directly through the
tions” is Stern-speak for picking δ ≈ 0, while front door with the legitimate concern that
sep07_Article4 8/17/07 4:31 PM Page 721

Weitzman: A Review of the Stern Review 721

there is a chance, whose subjective probabil- embrace the currently politically incorrect
ity is small but diffuse (thereby resulting in a idea, which is a third rail few policymakers
dangerously thickened left tail of compre- dare to touch and Stern doesn’t even men-
hensive consumption growth rates), that tion, that in the extremely unlikely event of a
global warming may eventually cause disas- truly extraordinary unfolding disaster it
trous temperatures and environmental might be a good emergency-backup plan to
catastrophes. If one accepts that global cli- purposely geoengineer spaceship Earth to
mate change is as likely an arena as any for a reverse previous inadvertent geoengineering
valid application of the general principle that from burning too much fossil fuel? Or, per-
thickened tails from uncertain structural haps more to the point here, can anyone
parameters must dominate expected dis- imagine how the public would not embrace
counted utility calculations, then many hard such an idea and demand that we should do
questions need to be asked. What are early all in our power to avert a climate-change-
warning signs of impending runaway envi- induced catastrophe at some hypothetical
ronmental disasters like melting ice sheets, future time when environmental disaster
thermohaline inversions, or just plain know- seems imminent? And how does the tail thick-
ing beforehand that we are on a trajectory ness of climate-change disasters compare
toward ΔT > 6˚C? How much would it cost to with the tail thickness of aerosol geoengi-
put in place the very best system of sensors neering, or the tail thickness inherent in
that money can buy for detecting early-warn- widespread nuclear power possibly going
ing signals of impending climate catastro- awry, or the tail thickness of massive sudden-
phes? How early might the warning from release mishaps from buried-CO2 sequestra-
monitoring systems be before the full effects tion because some remote “hypothetical”
are felt? What could we do as an emergency materialized?
response if we received such early-warning I trust that a few readers may be able to
signals? Would last-ditch emergency geo- think of more such questions about the real
engineering measures to ward off disaster by option value of waiting to gather information
reversing the worst consequences of global (and the empirical issue of what to do about
warming be available in time to help? (Such it), some of which might hopefully be more
emergency measures are likely to be so grounded in reality than my highly specula-
extreme as to be defensible only for an even tive examples. Whether or not my particular
more extreme environmental catastrophe in hypothetical stories are realistic, these kinds
the making—perhaps they might include of questions become relevant once the focus
painting all human-made structures on the of the economics of climate change shifts
planet reflective white and creating a from the middle range of the distribution of
“Pinatubo effect” by seeding the upper what might happen with ΔT at a IPCC-4
atmosphere with metallic dust or aerosols.7) mean of ≈ 2.8˚C a hundred years hence to
Could such last-ditch measures be made thinking more about what might be in the
reversible by building in decay mechanisms, tails with ΔT > 6˚C, which is just two IPCC-
as with sulfate aerosols, while we then used 4 standard deviations out for a century from
the new information to really undertake dra- now, meaning a probability ≈ 3 percent (and
conian measures to cut greenhouse gas emis- presumably a yet higher probability after
sions drastically? Are there other aerosol 2105). This thick tail is where most of the
precursors than sulfates with possibly better cost–benefit action may well be even if—or
environmental properties? Can the public perhaps precisely because—our estimates of
the probabilities involved are themselves
7 On the feasible use of sulfate aerosol precursors to so highly uncertain. Anything is possible in
reverse global warming, see T. M. L. Wigley (2006). the tails of a nondogmatic distribution.
sep07_Article4 8/17/07 4:31 PM Page 722

722 Journal of Economic Literature, Vol. XLV (September 2007)

(“Nondogmatic” in Bayesian parlance just early detection is impossible, or it is too

means that no event is ruled out a priori by expensive, or it comes too late (this is Stern’s
having been assigned zero probability in the line, and it might, or might not, happen to be
prior distribution.) A responsible policy true), or because nothing practical can be
approach neither dismisses the horror sto- done about reversing greenhouse warming
ries just because they are two standard devi- anyway—so we should stop stalling and start
ations away from what is likely nor gets making serious down payments on catastro-
stampeded into overemphasizing false phe insurance by cutting CO2e emissions
dichotomies as if we must make costly all-or- drastically. But these are conclusions we
nothing investment decisions right now to need to reach empirically, rather than pre-
avoid theoretically possible horrible outcomes judging them initially. Instead of declaring
in the distant future. immediate all-out war on greenhouse gas
In my opinion, public policy on greenhouse emissions as advocated by Stern, maybe we
warming needs desperately to steer a middle would do better by steadily but surely ramp-
course, which is not yet there, for dealing ing up greenhouse gas cuts over the next
with possible climate-change disasters. This decade or two while simultaneously investi-
middle course combines the gradualist gating seriously the nature of the runaway-
climate-policy ramp of ever-tighter green- climate disasters in the thick tails and what
house gas reductions that comes from main- might be done realistically about them should
stream mid-probability-distribution analysis they start to materialize. We can always come
(under reasonable parameter values) with the back in ten or twenty years time and declare
option value of waiting for better information all-out war on global-warming emissions
about the thick-tailed disasters. It takes seri- then—if we then think it is the best option
ously whether or not possibilities exist for among a better-studied reasonably consid-
finding out beforehand that we are on a run- ered portfolio of possible options.
away-climate trajectory and—without “leav- Until we start seriously posing and trying
ing it all up to geoengineering”—confronts to answer tough questions about rare global-
honestly the possible options of undertaking warming catastrophes, we will not make real
currently politically incorrect emergency progress in dealing constructively with the
measures if a worst-case nightmare trajecto- nightmare scenarios and we will continue to
ry happens to materialize. The overarching cope with them inadequately by trying to
concern of such a middle course is to be con- shoehorn disaster policy into an either-or
structive by having some semblance of a response category where it won’t fit. The
game plan for dealing realistically with what Stern Review has its heart in the right
might conceivably be coming down the road. place—it is not nice for us to play the role of
The point is to supplement mainstream eco- nature’s grim reaper by bequeathing the
nomic analysis of climate change (and main- enormously unsettling uncertainty of a very
stream ramped-up mitigation policies for small, but essentially unknown (and perhaps
dealing with it) by putting serious research unknowable), probability of a planet Earth
dollars into early detection of rare disasters that in hindsight we allowed to get wrecked
and by beginning a major public dialogue on our watch. However, Stern does not fol-
about contingency planning for worst-case low through formally on this really unsettling
scenarios perhaps akin to the way Americans part of the global warming equation (which a
(at their best) might debate the pros and generous interpretation of its not-convincing
cons of an anti-ICBM early warning system. economic analysis might say is the underly-
It may well turn out that the option value of ing motivation for its overall alarmist tone)
waiting for better information about cata- except indirectly, by choosing δ ≈ 0, η ≈ 1,
strophic tail events is negligible because −
Y ≈ 5 percent, −
Y ≈ 1 percent (which an
sep07_Article4 8/17/07 4:31 PM Page 723

Weitzman: A Review of the Stern Review 723

ungenerous interpretation might say is Review puts it, “the pricing of carbon, imple-
reverse-engineering the drastic slowing mented through tax, trading, or regulation,”
measures that the Review wants to impose is “required for an effective global response”
on greenhouse gas emissions to neutralize (p. xvii). One can only wish that U.S. political
the nightmare scenarios). I don’t mean to leaders might have the wisdom to understand
imply that there is some off-the-shelf and the courage to act upon the breathtak-
turnkey consensus model of the economics ingly simple vision that steady pressure from
of uncertain catastrophes that the Stern the predictable presence of a high carbon
Review was negligent in not using or that price reflecting social costs (whether
such a model would (or should) provide imposed directly through taxes or indirectly
ammunition for an excuse not to undertake via tradable permits) would do more to
serious action soon to slow greenhouse emis- unleash the decentralized power of capitalis-
sions. We just don’t yet know and we need tic American inventive genius on the problem
badly to find out. The overarching problem of researching, developing, and finally invest-
is that we lack a commonly accepted usable ing in economically efficient carbon-avoiding
economic framework for dealing with these alternative technologies than all of the piece-
kinds of thick-tailed extreme disasters, meal command-and-control standards and
whose probability distributions are inherent- patchwork subsidies making the rounds in
ly difficult to estimate (which is why the tails Washington these days.
must be thick in the first place). But I think As we have seen, on the economic-analysis
progress begins by recognizing that the hid- side the Stern Review predetermines the
den core meaning of Stern vs. Critics may be outcome in favor of strong immediate action
about tails vs. middle and about catastrophe to curtail greenhouse gas emissions by creat-
insurance vs. consumption smoothing. ing a very low value of r ≈ 1.4 percent via the
indirect route of picking point-estimate
6. Getting it Right for the Wrong Reasons? parameter values δ ≈ 0 and η ≈ 1 that are
more like theoretically reasoned extreme
The Stern Review is a political document— lower bounds than empirically plausible esti-
in Keynes’s phrase an essay in persuasion—at mates of representative tastes. But we have
least as much as it is an economic analysis also seen that a fair recognition of the truth
and, in fairness, it needs ultimately to be that we are genuinely uncertain about what
judged by both standards. To its great credit, interest rate should be used to discount costs
the Review supports very strongly the politi- and benefits of climate changes a century
cally unpalatable idea, which no democratic from now brings discounting rates down
politician planning to remain in office any- from conventional values r ≈ 6 − 7 percent to
where wants to hear, that (however it is pack- much lower values of perhaps r ≈ 2 − 4 per-
aged and whatever spin is put on it) cent, which would create a more intermedi-
substantial carbon taxes must be levied ate sense of urgency somewhere between
because energy users need desperately to what the Stern Review is advocating and the
start confronting the expensive reality that more modest measures to slow global warm-
burning carbon has a significant externality ing advocated by many of its critics. The
cost that ought to be taken into account by important remaining caveat is that such an
being charged full freight for doing it. (This is intermediate position is still grounded in a
the most central “inconvenient truth” of all, conventional consumption-smoothing ap-
which was conveniently ignored in Al Gore’s proach to the economic analysis of climate
award-winning film.) An entire chapter 22 in change that avoids formally confronting the
Stern, entitled “Creating a Global Price for issue of what to do about catastrophe insur-
Carbon,” is devoted to this theme. As the ance against the possibility of thick-tailed
sep07_Article4 8/17/07 4:31 PM Page 724

724 Journal of Economic Literature, Vol. XLV (September 2007)

rare disasters, which from first principles of comprehensive measure of consumption

economic–statistical reasoning presumptively that includes the natural environment. I
drive expected discounted utility outcomes. have argued that this inherently thickened
In conclusion, I think the Stern Review left tail of g is an important aspect of the
deserves credit for effectively raising the economics of climate change that every ana-
level of public discourse—by increasing gen- lyst—Stern and the critics of Stern—might
eral awareness that climate change is a seri- do well to try to address more directly.
ous issue which should be taken seriously, by History will judge whether the economic
arguing cogently for what is effectively a analysis of the Stern Review ended up being
global carbon tax as an essential component more wrong or more right and, if it was more
of any reasonable solution, by openly dis- right, whether as pure economic reasoning it
cussing adaptation to climate change (as well was right for the right reasons or it was right
as mitigation), and by popularizing for a for the wrong reasons.
wider audience than economists the idea REFERENCES
that economic analysis of costs and benefits Cline, William R. 1992. The Economics of Global
might be a useful legitimate method for Warming. Washington, D.C.: Institute for Inter-
national Economics.
evaluating policies to mitigate global warm- Dasgupta, Partha. 2007. “Commentary: The Stern
ing. I think also that Stern deserves some Review’s Economics of Climate Change.” National
measure of credit for elevating to promi- Institute Economic Review, 199: 4–7.
Hope, Chris. 2006. “The Marginal Impact of CO2 from
nence the problem of genuine uncertainty PAGE2002: An Integrated Assessment Model
concerning rare but dangerous climate dis- Incorporating the IPCC’s Five Reasons for Concern.”
asters, about which decisions must be made Integrated Assessment, 6(1): 19–56.
Intergovernmental Panel on Climate Change. 2007.
but whose scale and probability cannot be Climate Change 2007: The Physical Science Basis:
known precisely. However, in my opinion, Summary for Policy Makers. http://www.ipcc.ch.
Stern deserves a measure of discredit for Kelly, David L., and Charles D. Kolstad. 1999.
“Integrated Assessment Models for Climate Change
giving readers an authoritative-looking Control.” In The International Yearbook of
impression that seemingly objective best- Environmental and Resource Economics: 1999/2000:
available-practice professional economic A Survey of Current Issues, ed. H. Folmer and T.
Tietenberg. Cheltenham, U.K. and Northampton,
analysis robustly supports its conclusions, Mass.: Elgar, 171–97.
instead of more openly disclosing the full Lind, Robert, ed. 1982. Discounting for Time and Risk
extent to which the Review’s radical policy in Energy Policy. Washington, D.C.: Resources for
the Future.
recommendations depend upon controver- Maddison, David. 2006. “Further Comments on the
sial extreme assumptions and unconvention- Stern Review.” http://www.economics.bham.ac.uk/
al discount rates that most mainstream maddison/Stern%20Comments.pdf.
Mendelsohn, Robert O. 2007. “A Critique of the Stern
economists would consider much too low. I Report.” Regulation, 29(4): 40–47.
can’t help but feel after reading the Review Nordhaus, William D. 1994. Managing the Global
that its urgent tone of morality and alarm Commons: The Economics of Climate Change.
Cambridge, Mass. and London: MIT Press.
comes not from the chapter 6 aggregative Nordhaus, William D. 2007. “A Review of the Stern
economic modeling of climate-change Review on the Economics of Climate Change.”
impacts, which deservedly has drawn strong Journal of Economic Literature, 45(3): 686–702.
Portney, Paul R., and John P. Weyant, eds. 1999.
criticism from economists, but more from a Discounting and Intergenerational Equity. Wash-
not formally articulated fear of what might ington, D.C.: Resources for the Future.
potentially be out there with greenhouse Tol, Richard S. J., and Gary W. Yohe. 2006. “A Review
of the Stern Review.” World Economics, 7(4): 233–50.
warming in (using my own ponderous termi- Weitzman, Martin L. Forthcoming. “Subjective
nology here to make sure my expression is Expectations and Asset-Return Puzzles.” American
exact) the inherently thickened left tail of Economic Review.
Wigley, T. M. L. 2006. “A Combined Migration/
the reduced-form posterior-predictive prob- Geoengineering Approach to Climate Stabilization.”
ability distribution of the growth rate of a Science, 314(5798): 452–54.