Discounting and Uncertainty in Climate Change Policy Analysis

Author(s): Richard B. Howarth

Source: Land Economics, Vol. 79, No. 3 (Aug., 2003), pp. 369-381
Published by: University of Wisconsin Press
Stable URL: [Link]
Accessed: 02-01-2020 01:57 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@[Link].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
[Link]

University of Wisconsin Press is collaborating with JSTOR to digitize, preserve and extend
access to Land Economics

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 Discounting and Uncertainty in Climate Change
Policy Analysis
Richard B. Howarth

ABSTRACT. Economic studies of climate of unrestricted emissions and hence the po-
change commonly discount the future at a rate tential benefits of emissions control. Based
equal to the long-run return on corporate stocks. on a survey of the existing literature, the In-
Stock market returns, however, are dominated by
tergovernmental Panel on Climate Change
a risk premium, while climate change mitigation
(IPCC 1996) gauged that a doubling of
measures would reduce important risks to fu-
greenhouse gas concentrations relative to the
ture welfare. Drawing on the theory of investment
behavior under uncertainty, this paper argues pre-industrial norm would impose quantified
that the benefits of climate stabilization policies costs equivalent to 1.5-2.0% of world
should be discounted at a rate equal to the annualeconomic output with greater impacts in de-
return on risk-free financial assets, which attains veloping countries. The panel warned, how-
an empirical value between 0 and 2.6%. In addi-ever, that the impacts of climate change on
tion, expected benefits must be adjusted to ac- sea level, agriculture, human health, and
count for the value of risk abatement. (JEL Q21)ecosystems remain poorly understood. In-
deed, the IPCC's Third Assessment Report
I. INTRODUCTION
(2001) notes that climate change might im-
pose currently unquantified yet highly delete-
Climate change response strategies
riousin-
impacts through the disruption of the
volve flows of costs and benefits that are North Atlantic thermohaline circulation, sub-
asymmetrically distributed over time. Ac- stantial melting in the Greenland and West
cording to standard estimates, stabilizing Antarctic ice sheets, or positive feedback
greenhouse gas emissions at current levelsprocesses that led to the release of large
through the use of emissions taxes or a trada-quantities of greenhouse gases stored by ter-
ble permit scheme would impose annual restrial and marine ecosystems. Hence, cli-
costs on the order of 1% of gross economic
mate change policy is a matter of risk man-
output in industrialized nations over the nextagement, reducing emissions in the short run
10-20 years (Weyant 1999). Somewhat to protect future generations against poten-
lower costs might be achieved if emissions tially catastrophic costs. This approach is ex-
tax revenues were used to offset distortionary
plicitly embraced by the Framework Con-
taxes on capital and labor or if incentive-vention on Climate Change, which calls for
based instruments were accompanied by the "stabilization of greenhouse gas concen-
measures to accelerate the full adoption of trations in the atmosphere at a level that
cost-effective energy-efficient technologies would prevent dangerous anthropogenic in-
(Shackleton et al. 1996; Interlaboratory terference with the climate system."
Working Group 2000). Since the technolo- The value judgment behind this statement
gies and mechanisms required to achieve
has been criticized by some economists, who
emissions reductions are reasonably well
characterized, abatement costs estimates are
The author is associate professor, Environmen-
clustered in a relatively narrow range. tal Studies Program, Dartmouth College. Work on
Substantial uncertainties, however, sur- this paper was sponsored by the Office of Atmo-
round the long-run environmental impacts spheric Programs at the U.S. Environmental Protec-
tion Agency. The author thanks Kjell Arne Brekke,
Stephen DeCanio, Karen Fisher-Vanden, Reyer Ger-
Land Economics * August 2003 * 79 (3): 369-381 lagh, Michael Hoel, Snorre Kverndokk, Skip Laitner,
ISSN 0023-7939; E-ISSN 1543-8325 Richard Norgaard, Alan Sanstad, Robert Shackleton,
@ 2003 by the Board of Regents of the and the anonymous referees for comments and sugges-
University of Wisconsin System tions.

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 370 Land Economics August 2003

warn that greenhouse gas

positive when the emissions
discount rate is set equal
ment might to the
divert risk-free rate of [Link]
resources This conclusionfrom
productive uses, depends
possibly on the plausible assumption that ec
slowing
emissions abatement
growth and impairing the is an effective
welfare risk man- of
generations. This point
agement is emphasi
strategy that reduces the overall un-
Nordhaus (1994b) and
certainties Manne
faced by (1995
future generations. As
argue that discussed
climate by the IPCC's comprehensive
change policies re- sh
based on the reasoned use of cost-benefit view (1996, ch. 5), the argument that returns
to corporate stocks provide a reliable mea-
analysis, in which future costs and benefits
are discounted relative to the present based
sure of individual time preference is wide-
spread in the climate change literature. The
on the revealed time preference of individu-
present analysis suggests that this point of
als. According to these authors, the discount
rate should be set equal to the real return
view abstracts away from factors that are es-
available on corporate stocks, which aver-sential to the problem under investigation.
aged 7.9% per year between 1926 and 2000,
or 5.7% accounting for the influence of taxa- II. DISCOUNTING UNDER
tion (IPCC 1996, ch. 5; Ibbotson Associates CERTAINTY
2001).' In this view, the use of lower dis-
count rates would lead to the displacement ofFor purposes of comparison, it is useful t
commence the analysis by considering
productive private investments by public ex-
penditures paying substantially lower re- model of intertemporal choice in the contex
turns. A contrasting perspective is offeredof
byperfect foresight regarding future pri
Broome (1992) and Cline (1992), who argue and economic conditions. This framework is
that discount rates should be chosen to attach central to the analysis of Nordhaus (1994b)
equal weight to the welfare of present and fu- and Manne (1995) and is linked to theoretical
ture generations. work of Arrow and Kurz (1970). In this
The purpose of this paper is to consider model, there is just one financial asset that is
the implications of uncertainty in choosing often interpreted as the wealth that investors
the appropriate discount rate for use in aggre- hold in the form of corporate stocks.2 Under
gating the costs and benefits of emissions this interpretation, there are good reasons to
abatement. For the sake of analysis, I shall equate the discount rate with average stock
set aside the moral argument that the welfare yields. Relaxing the assumptions of this
of present and future generations demand model will shed light on the limitations of
equal weight in cost-benefit analysis, instead this approach in a world of multiple assets
adopting the view that discounting proce- and financial risk.
dures should be based narrowly on individual Consider the case of a representative
preferences. In a context of uncertainty, how- household that seeks to maximize the objec-
ever, I shall argue that this premise does not tive function:
justify equating the discount rate with the re-
turns paid by corporate stocks. Under stan-
dard investment theory, stock market yields
are dominated by a risk premium that reflects
W = u(c,)/(1 + p)/. [1]
the uncertainties associated with equity mar- In this setting, c, is the level of household
kets. As we shall see, less risky assets such consumption at a sequence of dates t = 0, 1,
as U.S. Treasury Bills and corporate bonds
pay after-tax real returns of only 0.0-2.6% 2, ... ; u(.) is a concave and increasing utility
per year. Accordingly, time preference per se
cannot account for the high yields that in- 'These figures are based on average returns to large
vestors demand on stocks. capitalization stocks in U.S. financial markets.
2 This view is embraced by Nordhaus and Manne
More particularly, the paper presents a
but is discussed less favorably by Arrow and Kurz, who
model in which climate change mitigationpoint out that different forms of capital yield different
measures yield net gains in social welfare if
rates of return. Arrow and Lind's (1970) work on this
their expected present-value net benefits are topic is discussed in the next section of this paper.

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
79(3) Howarth: Discounting and Uncertainty in Climate Change Policy Analysis 371

function; and p is the utility discount rate or

pure rate of time preference. The household dW = u(c, + dc,)/(1 + p)' - u(c,)/(1 + p)'
t=O t=0
earns the wage income w, and the rate of re-
turn r, on its capital wealth k,. Dividing its
income between consumption and net capital t=0
= (u(ct +
investment, the household's budget con-
straint takes the form:
S u'(c,)dc,/(1 + p).
t=O [4]
c, = kt+, = wt + (1 + rt)kt. [2]
Since equa
= u'(co)(HI
Although we shall not explicitly consider thechange in
form:
production side of the economy, it is natural
to suppose that wages and returns on invest-
ment reflect the marginal productivity of la-
bor and capital. Questions of general equi-
librium are addressed in standard textbooks dW = u'(co)dco I+ 1u'(co)dc,
i=1 +(1+I +
(see Blanchard and Fischer 1989). Laitner,
= u'(co)NPV [5]
DeCanio, and Peters (2001) discusses this
model and its applications in the economics where:
of climate change.
Maximizing perceived household welfare
subject to the budget constraint gives rise to
the optimality condition: NPV = dco +

u'(ct) = u'(ct+)(1 + rt+)/(l + p), [3] Equation [6] calcula

value (NPV) of the cli
gime by aggregating
that is commonly known as "Ramsey's arise at each point in t
rule" based on Frank Ramsey's (1928) semi- the discount rate is s
nal work on the theory of savings and invest- return on capital inv
ment. In this expression, the gross return to holds that the impa
capital investment is set equal to the mar- household welfare ar
ginal rate of intertemporal substitution. measure of net moneta
Given equation [3], we are ready to exam- pression, the constan
ine the discount rate that is appropriate for given by the marginal
use in climate change policy analysis. Rela- in the initial period. Si
tive to a baseline path that is characterized by positive under the assu
the consumption stream co, cl, c2, . . . , Sup- it follows that the net
pose that a climate change policy regime is may be used to evalu
implemented that yields a sequence of net the model under consideration.
benefits dco, dc , dc2, .. . at dates t = 0, 1, To briefly summarize, equating the dis-
2, ... We assume that all costs and benefits count rate with the rate of return on capi-
are measured in units of consumption-equiv- tal investment is justified in the case where:
alents, possibly accounting for the impacts of (a) the prevailing objective is to maximize
climate change on nonmarket environmental the perceived welfare of a representative
services. household; and (b) savings/investment deci-
If net benefits at each date are small rela- sions are made in the context of perfect fore-
tive to the prevailing consumption level, then sight concerning future prices and economic
the change in welfare generated by the pol- conditions. If one interprets long-run yields
icy regime may be calculated using the first- on corporate stocks as a measure of the mar-
order Taylor series approximation: ket return to capital investment, then it would

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 372 Land Economics August 2003

follow that the discount rate should be set information that is presently available. Under
equal to a value between 5.7% and 7.8% per the notation of the model, E,[W] is defined
year. This discussion leaves open the issueas ofthe expected value of W given the in-
whether discount rates should be defined in formation available at date t, in which the
terms of pre- or post-tax returns on invest- welfare achieved in each possible state of
ment. While Lind (1982) suggests that thenature is weighted by that state's statistical
use of after-tax rates is theoretically superior,probability. Since the utility function is con-
this approach is valid only if analysts care-cave, the model assumes the households are
fully account for the opportunity costs that strictly risk averse. The degree of risk aver-
arise when public policies crowd out privatesion depends on the curvature of the utility
investments.3 The necessary accounting oc- function.
curs automatically in general equilibrium The household's budget constraint takes
models that capture the impacts of taxation the form:
but can be quite complex in partial equilib-
rium analysis, calling for the use of theCt + kst+l + krt+i =
"shadow price of capital" method. See Kolb
wt + (1 + rst)kst + (1 + r,rt)k,. [8]
and Scheraga (1990) and Cline (1992) for
further review and discussion.
As before, c, and w, represent the house-
hold's consumption and wage income at date
III. INVESTMENT BEHAVIOR
t. In the present model, however, the house-
UNDER UNCERTAINTY
hold faces a choice between two types of
capital wealth. A "safe" asset kst, yields the
The assumption of perfect foresight is a
after-tax return rst that is risk-free and known
useful simplifying device in the analysis of
with certainty. A "risky" asset k,rt yields the
some economic questions. The finance litera-
after-tax return r, that is uncertain and sub-
ture, however, emphasizes the fact ject
thattorisk
random variation. In addition, we as-
and uncertainty play key roles in real-world
sume that real wages fluctuate randomly over
investment behavior. By extending a frame-
time due to technology shocks, regulatory
work developed by authors such uncertainty,
as Sam- and so forth.
uelson (1969) and Lucas (1978), we next The maximization of equation [7] sub-
explore the consequences of modem invest- ject to equation [8] is a standard problem
ment theory for the analytical framework de- in sequential decision-making or stochastic
scribed above. As we shall see, accounting dynamic programming. The solution of
for uncertainty has substantial implications this problem was described by Samuelson
for the theory of discounting. (1969); see Blanchard and Fischer (1989, ch.
Consider the case of a representative 6) for an accessible textbook treatment. At
household that seeks to maximize the objec-each date t, the household observes the real-
tive function:
ized values of the prevailing wage (w,) and
the return on risky capital (r,), while updat-
ing its expectations regarding conditions that
E0[W] = Eo u(c,)/(1 + p) , [7] might prevail at more distant future dates.
Accordingly, the household tailors its deci-
sions concerning consumption and invest-
under conditions of uncertainty. This ment
objec-to take advantage of new informa-
tive function is closely similar to the for-
tion as it becomes available. In this setting,
mulation described in equation [1]. In the
present context, however, the household isthe
3 In context of the present model, Lind's reason-
certain about its income level and market ing would apply if wages and returns to capital were
measured in after-tax terms and if the net benefit mea-
conditions in the first period of the model,
but uncertain about the conditions that will sure dc, accounted for the full impacts of climate
change policies on the future consumption stream, in-
hold at dates t = 1, 2, 3,... Hence it seeks to
cluding indirect effects mediated through changes in
maximize its expected well-being given theprivate investment.

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
79(3) Howarth: Discounting and Uncertainty in Climate Change Policy Analysis 373

rational behavior generates the first-order cov,[u'(c,+1), r,+l] [12]

conditions: E,[u'(c,+ )]

u'(c,) = E,[u'(c,+l)](1 + rs,+l)]/(1 + p) [9] is positive as well.

This analytical result is closely tied to ob-
u'(ct) = Et[u'(ct+)](1 + r,,,+)]/(1 + p). [10] served data on returns to financial assets with
varying levels of risk. Between 1926 and
These conditions have a simple and plausible 2000, for example, U.S. Treasury Bills paid
interpretation, naturally extending Ramsey's an average pre-tax return of 0.7% per year
rule to allow for uncertainty. In these equa- with an after-tax return of 0.0% (Ibbotson
tions, the marginal utility of consumption at Associates 2001).4 In the finance literature,
date t is equated with the expected contribu- Treasury Bills are often interpreted as a risk-
tion that one unit of investment yields at date free asset since their real returns are highly
t + 1, accounting for uncertainties in both stable over time (Mehra and Prescott 1985;
consumption growth and returns to the risky Kocherlakota 1996; Cochran 2001). Given
asset. their short (three-month) maturities, their
One conclusion that follows from these nominal yields adjust rapidly to changes in
conditions was described by Lucas' (1978) inflation. Similar returns are paid by safe
seminal paper on capital markets and asset private-sector assets such as money market
accounts and short-term certificates of de-
pricing under conditions of uncertainty. (See
Cochran (2001) for an overview of the subse-
posit.5
quent literature.) According to this frame-Alternatively, analysts might gauge the re-
work, the returns on safe and risky invest-
turns demanded on low-risk assets using his-
ments are linked by the following equation,
torical data on corporate bonds, which gener-
which is known as the "consumption capital
ated average pre- and post-tax yields of 2.6%
asset pricing model": and 1.5% over this same period (Ibbotson
Associates 2001). Although bonds are sub-
ject to inflation risk and market fluctuations
E, [rrt+,] = rst+, - cov.[u'(c ), r, ] [11]
E,[u'(ct+l)] that do not apply to Treasury Bills and short-
term bank deposits, the risks they entail are
In this setting, the expected return to the relatively modest. In contrast, the stock mar-
risky asset is set equal to the return on the ket-which is characterized by substantial
safe asset plus a second term that defines volatility as demonstrated by the asset bubble
the effective risk premium investors demandof the 1990s and the subsequent bear mar-
ket-paid an average pre-tax return of 7.9%
to willingly accept uncertainty. The risk pre-
mium is defined in terms of the covariance with a post-tax return of 5.7% (IPCC 1996,
between the marginal utility of consumptionch. 5; Ibbotson Associates 2001). When con-
and the return to the risky asset as perceivedsidered in terms of the theoretical model un-
given the information available at date t,der discussion, these data suggest that stock
yields are dominated by a risk premium that
cov,[u'(c,+I),
value r,+l,], divided
of the marginal utility by the expected
of [Link] investors for accepting financial un-
A key point is that risky investments yield certainty. Thus risk, not pure time prefer-
high returns under circumstances where con-
sumption is high so that the marginal utility
4 The after-tax return is based on a 3.8% nominal
of consumption is low, while paying low re-
interest rate, a 3.1% inflation rate, and a 19% marginal
turns when consumption is low so that the tax rate for median-income U.S. households. This tax
marginal utility of consumption is high. In
rate is calculated using data provided by Burman, Gale,
other words, risky assets enhance the uncer- and Weiner (1998).
tainty associated with a household's con- 5 An anonymous reviewer suggested that Treasury
Bill yields might be explained in terms of investors' de-
sumption stream so that cov,[u'(c,+1), r,+1]sire
< to maintain adequate liquidity. Instruments such as
0. And since E,[u'(c,+l)] is unambiguously
mutual funds, however, provide both high liquidity and
positive, it follows that the risk premium: the returns associated with stocks.

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 374 Land Economics August 2003

ence, is the key The point here

factor is not that
that real-world
explains
behavior equityhouseholds
markets. hold negative rates of pure time
in
What are the implications of to
preference. Instead, the conclusion these
draw is o
that Nordhaus (1994b)
vations for the economics of and Manne,
climateMendel- ch
sohn, and Richelsand
Both Nordhaus (1994b) (1995) work with models
Manne, M
delsohn, and Richels (1995)
that significantly overstateassume
the rate of pure th
utility function time
is preference
logarithmic in
by failing to account for the co
tion and that role of uncertainty in
consumption financial
will markets.
grow at
annual rate. Based on
Moreover, Ramsey's
Kocherlakota (1996) presents evi- ru
dence that the
characterizes optimal logarithmic utility function
decision-makin
employed in foresight
conditions of perfect these studies substantially un-
(see e
derestimates
[3]), these authors identifythe degree of risk
the aversion re-
pure
vealed by
time preference for useactual in
investment behavior. Since
calculating
counted net benefits
greenhouse gas
of emissions
climate
abatement yields
mit
measures. Given a distant
return future benefits
to that
corporate
reduce the poten-
tial risks of
of 6% per year, these climate change, these findings
assumptions imp
suggest that
the pure rate of time the models in question may
preference is:un-
derstate the full social benefits of emissions
control.

p- C (1 + r,)-1IV. DISCOUNTING UNDER

Ct+l
UNCERTAINTY
1
1- (1.06) - 1 0.03,
1.03 [13] The discussion thus far suggests that
is inappropriate to use models that ar
premised on the assumption of perfect fore
or 3% per sight
year. Under
to evaluate climate change polici
tainty, however, equati
under conditions of uncertainty. Nordh
pure rate of
(1994b, time prefe
ch. 8) and Kolstad (1996) partia
address this problem by extending deter
ministic models of climate-economy inte
actions to allow for uncertainty in key p
p = E rametert
ci' values. Newell
(1 and Pizer (2001)
+r,,
c,+- I show that uncertainty about asset yields can
have important consequences for discounting
1
procedures even if one interprets stock yields
- 1I (11.03
+Irt)- 1 [14] as an indicator of individual time preference.
None of these studies, however, focuses on
the difficulties of choosing discount rates
if the utilitybased on the standard version of Ramsey's
function were
rule that abstracts away from perceived in-
mic. The approximation E, vestment risks. A full-blown analysis of this
holds when the expected rate
problem would need to simultaneously ac-
growth
formly is 3% per
count
distributed year
for the risks associated
across and
with both w
capi- va
borhood of tal investment
c,. Underand climate-change response
these
pure rate of strategies,
time with thepreference
rate of pure time prefer-
ence calibrated
negative value unless in accordance withthe equation re
assets assumed a of
[9]. Analyzing a model value
this type is beyond of
year. As we the scope of the present
have seen, discussion andhow
data on Treasury would present ratherBills
formidable computa-
and
yields suggests tional complexities.
that the tr
falls below this critical value. The results outlined in the preceding sec-

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
79(3) Howarth: Discounting and Uncertainty in Climate Change Policy Analysis 375

tion, however, may be extended to determine

the discount rate that should be used to eval- - u(c,))/(1 + p)'
uate climate change policies in a partial equi-
librium setting. A related problem was ex-
plored by Arrow and Lind (1970), who = Eo u'(c,)dc,(1 +p)
showed that risk-free government projects
should be evaluated using a discount rate that
reflects the return on risk-free private-sector = E[u(c,)dc,](1 + p)'. [15]
investments. In a similar vein, Starrett (1988) t=O

(see also Sandmo 1972) sought to define pro-

cedures through which analysts could ad- In interpreting this expression, it is useful to
dress the risk characteristics of public proj- note that the covariance between the terms
ects by adjusting the discount rate that is u'(c,) and dct as evaluated at date t = 0 is
used to calculate the project's expected defined as:
present-value net benefits. Here we employ
similar methods to reach a conclusion that covo[u'(c,), dc,l ] Eo[(u'(c,) - Eo[u'(c,)l)(dc,
differs from the findings of these previous
- Eo[dc,])] = Eo[(u'(c,)dc,]
studies. In evaluating climate change poli-
cies, analysts should: (1) adjust measures of - Eo[(u'(c,)]Eo[dc,]. [16]
expected net benefits for risk using a cer-
tainty equivalents approach; and (2) set the In addition, taking the expectation of equa-
tion [9] at date t = 0 implies that:
discount rate equal to the risk-free rate of re-
turn. The reasoning that supports this conclu-
sion proceeds as follows. 1+p
As in Section 2, suppose that the economy Eo[u'(c,+l)] = Eo[(u'(c,)] [17]
1 + rs,+l
may be described by the baseline consump-
tion path co, cI, c2,..., while climate change which in turn implies that:
policies yield a sequence of net benefits dco,
dc , dc2, . .. at dates t = 0, 1, 2 ... In the t

present analysis, however, both future con-

sumption levels and the net benefits of emis- Eo[(u'(c,)] = u'(co)(1 + p)' 1 + . [18]
sions abatement are explicitly uncertain. In-
deed, it is natural to assume that the specific
actions taken to slow climate change are Substituting equations [16] and [18] into ex-
state-contingent, allowing decisionmakers to pression [15], we find that the welfare change
revise future policies to take advantage of induced by greenhouse gas emissions abate-
new information as it becomes available. Un- ment policies may be written as:
der these assumptions, the expected welfare
gain generated by emissions abatement may
E[dW] = I (Eo[u'(c,)]Eo[dc,]
be calculated using the first-order Taylor se- t=O

ries expansion:
+ covo[u'(c,), dc,])/(1 + p)'

= u'(co)dco + u'(co)

E[dW] = EoZ u(c, + dc,)/(1 + p) X7(EO[dc] + covo[u'(c,), dc,]

t=1A (Eo[u'(c,)]
- u(c,)/(1 + p
S1 + ' rs
t=O ?~ ~?
= u'(co)NPV, [19]

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 376 Land Economics August 2003

where: expected net benefits of a given policy re-

gime to account for their perceived risk char-
acteristics, not by using a discount rate that
NPV = dco + EO[dc]
departs+fromcovo[u'(c,),
the returns available on safe in-dc,]
t=1 Eo[u'(c,)] vestments.

Third, as in Section 2, the impacts of

X 1
emissions abatement on perceived household
( 1=i + rsi [20] welfare are strictly proportional to the mone-
tary measure of present-value net benefits.
Under equation [19], the change in expected
There are several points to note aboutwelfare produced by the policy is given by
these formulae. First, the NPV expression
u'(co)NPV. Since the marginal utility of con-
calculates the net present value of the bene-
sumption is positive, u'(co) may be inter-
fits provided by greenhouse gas emissionspreted as a scaling parameter that converts
abatement, evaluated in monetary units. Inmonetary measures of net benefits into units
this formula, expected net benefits are ad-of well-being.
justed to reflect the uncertainties associated
In sum, this analysis shows that account-
with emissions abatement policies. If one as-
ing the role that uncertainty plays in invest-
sumes that the benefits of emissions abate-
ment decisions justifies the use of a risk-free
ment are positively correlated with the mar-
discount rate in the analysis of climate
ginal utility of consumption-that is, thatchange response strategies, provided that an-
climate change policies yield their greatest
alysts calculate net benefits using a certainty
benefits in "bad" states of nature character-
equivalents approach. This conclusion de-
ized by low levels of welfare and hence parts
a from the widely accepted argument that
high value for u'(c,)-then the expectation returns to corporate stocks provide an appro-
Eo(dc,) understates the full benefits of the
priate measure of individual time preference.
policy so that:

Eo[dc,] < EO[dc,] + covo[u'(c,), dc,] [21] V. A HEURISTIC APPROACH TO

Eo[u'(c,)] THE ANALYSIS OF UNCERTAINTY

This inequality stems in part from the as- As we have seen, the application of
sumption that the utility function is strictly cost-benefit techniques to the analysis o
increasing so that u'(c,) > 0. If, on the other greenhouse gas emissions abatement requir
hand, the covariance between the marginal assumptions concerning the covariance b
utility of consumption and the net benefits tween the marginal utility of consumptio
of emissions abatement were negative-that and the net benefits of emissions abatement.
is, if abatement benefits were highest in Unfortunately, existing data on emissions
"good" states of nature characterized by abatement benefits are incomplete and sub-
high consumption levels and hence a low ject to significant structural uncertainties.
value for the marginal utility of consump- While analysts may construct rough calcula-
tion-then Eo[dct] would overstate the cor- tions of both the expected net benefits of
rect risk-adjusted benefits at date t, and the abatement and the costs that climate change
sign of relation [21] would be reversed. might entail in particular "worst case scenar-
Second, the present value of future net ios" (see Tol 2003), adjusting expected ben-
benefits is evaluated by setting the discount efits for risk is a speculative activity given
rate equal to the rate of return on risk-free the information that is presently available.
assets. In the NPV formula, the risk-free rate To cope with this problem, it is useful to
rs, figures prominently while the rate of re- note that relation [21] shows that the risk-
turn on risky assets (r,) plays no direct role. adjusted benefits of emissions abatement ex-
According to this model, questions of uncer-
tainty should be addressed by adjusting the ceed the expected
> 0. Under benefits ifitcovo[u'(ct),
this assumption, dc,]
follows that:

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
79(3) Howarth: Discounting and Uncertainty in Climate Change Policy Analysis 377

ENPV dco (1994a), however, finds that output losses of

25% might arise with a mean probability of
5%. Since greenhouse gas emissions abate-
+ Eo[dcl ment would yield particularly large benefits
in "bad" states of nature in which unabated
< NPV. [22] emissions would impose devastating impacts
on agriculture, coastal areas, biodiversity,
and other systems, it seems natural to assume
In this expression, ENPV
that abatement is corre-
benefits are negatively the
lated with future consumption
value of the expected benefits levels and o
abatement. Since ENPV understates the full hence positively correlated with the marginal
contribution that emissions abatement mea- utility of consumption.
sures provide in terms of enhanced social Viewed more comprehensively, however,
welfare, it follows that climate change re-future levels of income and consumption are
sponse strategies yield net improvements intied to technological and behavioral uncer-
social welfare if the present value of their
tainties that are largely independent of envi-
expected net benefits is positive when theronmental conditions. Moreover, it is reason-
discount rate is set equal to the return to risk-able to assume that the benefits of emissions
abatement would be positively correlated
free assets. Hence ENPV may be used to pro-
vide a lower-bound estimate of NPV when with prevailing incomes since higher in-
comes imply a greater willingness to pay
cov0[u'(c,), dc,] is unobservable but is known
to assume a positive value. to avoid environmental damages. In other
In practical terms, this result has ratherwords, technological uncertainties suggest
far-reaching implications for the economics that that emissions abatement benefits might
of climate change. As Howarth (1998)be positively correlated with consumption
shows, relatively modest rates of greenhouse (and hence negatively correlated with the
gas emissions abatement emerge as socially marginal utility of consumption) in the ab-
efficient when the discount rate is set equalsence of other factors. Hence the sign of
to the long-run return to corporate stocks. covo[u'(c,), dcl] depends critically on the
When the anticipated net benefits of emis- interplay between uncertainties concerning
sions abatement are discounted at an annual baseline economic growth and the impacts of
rate of 1.3% per year, however, optimal con- climate change.
trol rates rise from 48% to 89% over the While a full empirical analysis is beyond
course of the next century. In Howarth's
the scope of this paper, preliminary insights
analysis, this low discount rate was justified
into this problem are provided by the follow-
by the moral premise that social decision-ing illustrative calculations. Suppose that:
makers should attach equal weight to the
welfare of present and future generations 1. Per capita consumption would grow at
(see also Broome 1992; Cline 1992). In the an average rate of either 2.5% or 3.5%
present analysis we have found that similar over the next 70 years in the absence
discount rates may be justified based on con- of climate change. For the sake of sim-
siderations of risk, implying that high rates plicity, we shall assume that these
of emissions control might be economically growth rates are equally probable.
efficient notwithstanding issues of intergen-2. With probability of 0.95, climate
erational fairness. change would reduce potential con-
One might ask whether it is reasonable to sumption by 1.75% 70 years from the
assume that covo[u'(c,), dct] > 0. According present in the absence of corrective
to the IPCC (1996), a doubling of carbon di- policies. With a probability of 0.05, cli-
oxide equivalent in the global atmosphere matic impacts would reduce consump-
might impose anticipated costs of 1.5-2.0% tion by 25% unless steps were taken to
of economic output on the global economy. reduce greenhouse gas emissions.
An expert opinion survey by Nordhaus 3. We consider a marginal reduction in

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 378 Land Economics August 2003

short-run greenhouse
ment, we examined thegas emissio
numerical size of this
that would reduce the
effect using future
Howarth's damag
(1998) general equi-
of climate change libriumby model 1%. (Thisinterac-
of climate-economy impl
that the damage [Link]
Looking 70 years into the wouldfuture, the b
reduced to 1.7325% and 24.75% in expected benefits of a marginal reduction in
these two scenarios.) present greenhouse gas emissions are five
times greater than the indirect costs that
would arise due to reductions in the rate of
Under the assumptions of this example,
the consumption level and the net benefits of growth. When the calculations pre-
economic
emissions abatement are nearly uncorrelated.
sented in this paper are adjusted to allow for
suchthe
The correlation coefficient, which reflects costs, the results described above are
two sets of factors identified above, takes on
substantially unaltered.
a value of just 0.01. The analysis therefore supports the intu-
To calculate the covariance between the ition that emissions abatement provides an
marginal utility of consumption and net ben-
effective risk-management strategy if one as-
efits it is necessary to impose specific as-
signs a small but positive probability to the
sumptions on household preferences. Sup- hypothesis that unmitigated climate change
pose, for example, that preferences would are impose major social costs. Hence in-
characterized by the utility function u(c,) =
vestments in climate stabilization might en-
c-/(1 - a) where a is the coefficient hance social welfare even if their expected
of relative risk aversion. As Kocherlakota returns were below those available on risk-
(1996) notes, data on observed investment free financial assets. Under these conditions,
behavior suggest that a assumes a value be-
comparing the returns available on emissions
tween 8.5 and 30 for the model under con- abatement and corporate stocks is conceptu-
sideration. Under this assumption, the co-ally inappropriate. More appropriately, emis-
variance between the marginal utility of sions reductions should be considered as a
consumption and the net benefits of emis- form of insurance that safeguards the inter-
sions abatement takes on a large positive
ests of future generations.
value. More specifically, equation [20] im-
plies the expected benefits of emissions VI. CONCLUSION
abatement must be adjusted upwards by be-
tween 125% and 474% to account for the The argument that long-run returns o
welfare gains associated with reductionscorporate
in stocks offer a reliable measure
the risk of catastrophic climate change. individual time preference is a famili
Although these calculations are merely theme
il- in the economics of climate chang
lustrative, their qualitative conclusionsSince
are equity markets generate real return
robust with respect to reasonable changes that
in average 5.7-7.9% per year, this a
the parameters. The expected value of proach
net suggests that the future benefits
greenhouse gas emissions abatement shoul
benefits, for example, understates the full
be discounted at a quite substantial rate i
(risk-adjusted) value of greenhouse gas emis-
sions abatement when either: a) the riskpolicy-makers
pa- aim to maximize the welfar
of a typical household. In this perspectiv
rameter a reduced to 1.4; or b) the probabil-
the use of low discount rates could be ration-
ity of the "high damage" scenario is reduced
to just 1%. Since Tol (2003) constructs a alized
sce- only if policy-makers adopted a nor-
mative framework that, in effect, set aside
nario of catastrophic climate change in which
the principle of consumer sovereignty.
potential damages are much larger than those
considered here, the present analysis might The present analysis, in contrast, finds that
be viewed as conservative. the practice of equating the discount rate
An anonymous reviewer suggested thatwith average stock yields is generally inap-
short-run emissions abatement might imposepropriate in the analysis of climate change
response strategies. This conclusion stems
costs on future society by reducing the rate
of economic growth. To explore this argu- from the premise that greenhouse gas emis-

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
79(3) Howarth: Discounting and Uncertainty in Climate Change Policy Analysis 379

sions abatement constitutes a type of insur- be evaluated using overlapping generations

ance mechanism that reduces the risk that fu-models that explicitly address the distribu-
ture generations will bear the impacts of tion of costs and benefits between present
catastrophic climate change. When confront- and future generations. This approach is con-
ing a problem of this type, the standard sistent with Brown's (1998) notion of stew-
methods of financial economics suggest that ardship-the moral judgment that future
public policies should be evaluated using a generations are entitled to inherit an undi-
three-stage approach. First, analysts should minished natural environment and must
calculate the expected net benefits of emis- therefore be compensated for climate change
sions abatement at each point in time. Sec- damages. Howarth and Norgaard (1995)
ond, they should adjust expected net benefits show how questions of uncertainty may be
to account for the social value of risk reduc- addressed in an overlapping generation
tion. Third, they should calculate the present model of climate-economy interactions. Fu-
value of adjusted net benefits with the dis-ture research might explore the interplay be-
count rate set equal to the return available on tween risk and time preference in this impor-
risk-free financial instruments. tant class of models.
This procedure is based on the assumption In addition, financial economists have
that decision-makers aim to maximize the called into question the specification of pref-
expected welfare of a representative house- erences adopted in this paper. As Kocherla-
hold given the time and risk preferences kotare- (1996) notes, at least three types of fac-
vealed in real-world financial markets. As tors affect household decisions concerning
such, the approach is fully consistent with savings
the and investment: pure time prefer-
principle of consumer sovereignty. It is note-
ence, risk, and the diminishing marginal util-
worthy, however, that low-risk assets suchityas of consumption in a world characterized
U.S. Treasury Bills and corporate bonds pay by economic growth. In the standard model
real returns of just 0.0-2.6% per year. Hence
of investment under uncertainty that under-
this analysis suggests the use of discount lies this paper, both risk and the diminishing
marginal utility of consumption are ad-
rates that are substantially lower than long-
run returns to stocks. dressed in terms of the curvature of house-
Moreover, it is plausible to assume that holds' utility functions. In reality, however,
the standard measure of expected net benefits households appear to be both highly risk-
understates the full social value of green- averse and yet at the same time quite willing
house gas emissions abatement. This as- to exchange present for future consumption
sumption would hold true if the net benefits when they can do so without risk. According
of abatement were positively correlated with to Kocherlakota (1996), accounting for these
the marginal utility of consumption-that is, effects requires more richly specified models
if climate change policies would yield their that disentangle these conceptually distinct
greatest benefits in circumstances where cli- aspects of preferences. There are good rea-
mate change impacts would impose their sons to consider the implications of such
greatest social costs. Ultimately, of course, models for the economics of climate change.
this issue must be settled by careful empirical Finally, some readers might object that the
analysis. The heuristic calculations provided approach described in this paper implies that
in this paper, however, suggest that the social policy analysts should work with discount
value of risk reduction might be positive rates that are substantially lower than those
even if the probability of catastrophic climate employed in the private sector. In this per-
change was as low as 1%-a figure that is spective, low-return public policies might
five times below the mean value reported in crowd out private investments yielding
Nordhaus' (1994a) expert opinion survey. greater social benefits. As we have seen,
The analysis, of course, rests on assump- however, the private sector demands low re-
tions that some readers might find objection- turns on investments that are characterized
able. Gerlagh and Keyzer (2000), for exam- by low degrees of financial risk. Firms regu-
ple, argue that climate change policies should larly accept negative expected returns on in-

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
 380 Land Economics August 2003

surance policies mental

and Economics, 111-38, ed. D. Bromley.
investments in risk
agement, which are Oxford, U.K.: Blackwell [Link]
valuable
Ibbotson Associates. 2001. Stocks, Bonds, Bills
protect firms against the prospect o
and Inflation. Chicago: Ibbotson Associates.
(though improbable) potential losses.
Intergovernmental Panel on Climate Change
Framework Convention (IPCC). 1996. Climateon ChangeClimate
1995: Eco-
makes clear, the purpose of of
nomic and Social Dimensions greenho
Climate
emissions abatement is to reduce the threat
Change. New York: Cambridge University
that climate change poses to future genera- Press.
tions. In short, climate change mitigation . 2001. Climate Change 2001: Impacts,
measures have risk characteristics that are Adaptation and Vulnerability. New York:
not analogous to the uncertainties that sur- Cambridge University Press.
round corporate stocks. This disparity is cast Interlaboratory Working Group. 2000. Scenarios
into sharp relief through appeals to the stan- for a Clean Energy Future. Oak Ridge: Oak
Ridge National Laboratory.
dard theory of investment under uncertainty.
Kocherlakota, N. R. 1996. "The Equity Premium:
It's Still a Puzzle." Journal of Economic Lit-
erature 34 (Mar.): 42-71.
Kolb, J. A., and J. D. Scheraga. 1990. "Dis-
References counting the Benefits and Costs of Environ-
mental Regulations." Journal of Policy Analy-
Arrow, K. J., and M. Kurz. 1970. Public Invest- sis and Management 9 (Summer): 381-90.
ment, the Rate of Return, and Optimal FiscalKolstad, C. D. 1996. "Learning and Stock Effects
Policy. Baltimore: Johns Hopkins Press. in Environmental Regulation: The Case of
Arrow, K. J., and R. C. Lind. 1970. "Uncertainty Greenhouse Gas Emissions." Journal of Envi-
and the Evaluation of Public Investment Deci- ronmental Economics and Management 31
sions." American Economic Review 60 (Jun.): (Jul.): 1-18.
364-78. Laitner, J. A., S. DeCanio, and I. Peters. 2001.
Blanchard, O. J., and S. Fischer. 1989. Lectures"Conceptual Frameworks to Reflect Behav-
on Macroeconomics. Cambridge, Mass.: MIT ioural and Social Relationships in the Assess-
Press. ment of Climate Mitigation Options." In Soci-
Broome, J. 1992. Counting the Cost of Global ety, Behaviour, and Climate Mitigation, 1-64,
Warming. Cambridge, U.K.: White Horse eds. E. Jochem, J. Sathaye, and D. Bouille.
Press. Dordrecht: Kluwer.
Brown, P. G. 1998. "Toward an Economics of Lind, R. C. 1982. Discounting for Time and Risk
Stewardship: The Case of Climate." Ecologi- in Energy Policy. Washington, D.C.: Re-
cal Economics 26 (Jul.): 11-21. sources for the Future.
Burman, L. E., W. G. Gale, and D. Weiner. 1998. Lucas, R. E. 1978. "Asset Prices in an Exchange
"Six Tax Laws Later: How Individuals' Mar- Economy." Econometrica 46 (Nov.): 1429-
ginal Federal Income Tax Rates Changed be- 45.
tween 1980 and 1995." National Tax Journal Manne, A. S. 1995. "The Rate of Time Prefer-
51 (Sep.): 637-52. ence: Implications for the Greenhouse De-
Cline, W. R. 1992. The Economics of Global bate." Energy Policy 23 (Apr./May): 391-94.
Warming. Washington, D.C.: Institute for In- Manne, A. S., R. Mendelsohn, and R. Richels.
ternational Economics. 1995. "A Model for Evaluating Regional and
Cochran, J. H. 2001. Asset Pricing. Princeton: Global Effects of GHG Reduction Policies."
Princeton University Press. Energy Policy 23 (Jan.): 17-34.
Gerlagh, R., and M. A. Keyzer. 2001. "Sus-Mehra, R., and E. C. Prescott. 1985. "The Equity
tainability and the Intergenerational Distribu- Premium: A Puzzle." Journal of Monetary
tion of Natural Resource Entitlements." Jour- Economics 15 (2): 145-61.
nal of Public Economics 79 (Feb.): 315-41. Newell, R., and W. Pizer. 2001. "Discounting the
Howarth, R. B. 1998. "An Overlapping Genera- Benefits of Climate Change Mitigation: How
tions Model of Climate-Economy Interac- Much Do Uncertain Rates Increase Valua-
tions." Scandinavian Journal of Economics tions?" Washington, D.C.: Pew Center on
100 (3): 575-91. Global Climate Change.
Howarth, R. B., and R. B. Norgaard. 1995. "In- Nordhaus, W. D. 1994a. "Expert Opinion on Cli-
tergenerational Choices under Global Environ- matic Change." American Scientist 82 (Jan.):
mental Change." In Handbook of Environ- 45-52.

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]
79(3) Howarth: Discounting and Uncertainty in Climate Change Policy Analysis 381

ciency Value of Carbon Tax Revenues." In

-. 1994b. Managing
The Economics the Global
of Climate Commons:
Change. Cam- Reducing Global Carbon Dioxide Emissions:
bridge, Mass.: MIT Press. Costs and Policy Options, ed. D. Gaskins
Ramsey, F. 1928. "A Mathematical Theory of and J. Weyant. Stanford: Stanford University
Saving." Economic Journal 38 (Dec.): 543-59. Press.
Samuelson, P. 1969. "Lifetime Portfolio Selec- Starrett, D. A. 1988. Foundations of Public Eco-
tion by Dynamic Stochastic Programming." nomics. New York: Cambridge University
Review of Economics and Statistics 51 (Aug.): Press.
239-46. Tol, R. S. J. 2003. "Is the Uncertainty about Cli-
Sandmo, A. 1972. "Discount Rates for Public In-mate Change Too Large for Expected Cost-
vestment under Uncertainty." InternationalBenefit Analysis?" Climatic Change 56 (3):
Economic Review 13 (Jun.): 287-302. 257-63.
Shackleton, R., M. Shelby, A. Christofaro, R.
Weyant, J. P., ed. 1999. The Costs of the Kyoto
Brinner, J. Yanchar, L. Goulder, D. Jorgenson,Protocol: A Multi-Model Evaluation. Energy
P. Wilcoxen, and P. Pauly. 1996. "The Effi- Journal (special issue): 1-448.

This content downloaded from [Link] on Thu, 02 Jan 2020 01:57:44 UTC
All use subject to [Link]