Environmental and Resource Economics 21: 135–160, 2002.

135
© 2002 Kluwer Academic Publishers. Printed in the Netherlands.

Estimates of the Damage Costs of Climate Change

Part II. Dynamic Estimates

RICHARD S.J. TOL

Centre for Marine and Climate Research, Hamburg University, Germany; Institute for
Environmental Studies, Vrije Univresiteit, Amsterdam, The Netherlands; Center for Integrated Study
of the Human Dimensions of Global Change, Carnegie Mellon University, Pittsburgh, PA, USA

Accepted 21 May 2001

Abstract. Monetised estimates of the impact of climate change are derived. Impacts are expressed as
functions of climate change and ‘vulnerability’. Vulnerability is measured by a series of indicators,
such as per capita income, population above 65, and economic structure. Impacts are estimated
for nine world regions, for the period 2000–2200, for agriculture, forestry, water resources, energy
consumption, sea level rise, ecosystems, fatal vector- borne diseases, and fatal cardiovascular and
respiratory disorders. Uncertainties are large, often including sign switches. In the short term, the
estimated sensitivity of a sector to climate change is found to be the crucial parameter. In the longer
term, the change in the vulnerability of the sector is often more important for the total impact. Impacts
can be negative or positive, depending on the time, region, and sector one is looking at. Negative
impacts tend to dominate in the later years and in the poorer regions.

Key words: adaptation, climate change, impacts

JEL classification: Q00, Q25, Q40

1. Introduction
A sound knowledge of the impact of climate change is a prerequisite for a well-
informed greenhouse gas emission reduction policy. Although governments of
most OECD countries are now considering, and some already implementing such
policies, insights into what may be the consequences of an enhanced greenhouse
effect are still far from complete. Watson et al. (1996) review the literature on
climate change impacts in general, and Pearce et al. (1996) and Tol et al. (2000)
assess the literature on impacts from an economic viewpoint. These reviews reveal
that most impact studies take a static approach. That is, the effect of a single,
changed climate (usually, 2×CO2 , i.e, climate if the atmospheric concentration
of carbon dioxide would be doubled) on the current system is investigated. This is
a useful starting point, but climate will change gradually and is not likely to stop
at 2×CO2 . Moreover, populations and economics will grow, and technologies and
institutions will evolve. This paper develops a model of climate change impacts
that takes account of the dynamics of climate change and the systems affected
by it.
136 RICHARD S.J. TOL

The model develop can be used in a cost-benefit analysis of greenhouse gas

emission reduction. Cost-benefit analyses have so far failed to influence climate
policy. Emission abatement targets have been more ambitious than justifiable
by welfare optimisation (Nordhaus and Yang, 1996; Tol, 1999). However, it is
becoming increasingly clear that policy targets will be missed, and most likely by
a substantial margin. This situation reopens the discussion on the aims and targets
of greenhouse gas emission reduction policy.
This model is not the first in its kind. Most integrated assessment models of
climate change (cf. Weyant et al., 1996, for a review) include such a model. Tol
and Fankhauser (1998) review these models to conclude that dynamics are usually
straightforward and assumptions typically ad hoc. Worse, assumptions are seldom
openly discussed, so that understanding let alone progress are hampered. Earlier
papers (Tol, 1994, 1996) attempt to open the discussion. This paper is a further
step in that assumptions are better supported by literature and data. In that, the
approach of Mendelsohn et al. (1996a, b; cf. Mendelsohn, 1998) is comparable
to this work, although their models excludes changing vulnerabilities to climate
change. Sensitivity analyses in this paper are more systematic and extensive.
Obviously, this paper does not result in a climate change impact model that
is adequate. The accompanying static impact assessment (Tol, 2001) is far from
perfect, with many pieces missing and a lot of questionable assumptions. Adding
the dynamics implies adding more assumptions, many of which are debatable. Yet,
this paper contributes to clarifying the full scope of the enhanced greenhouse effect,
and to the research agenda that is needed for a sound understanding of the issue.
Section 2 briefly discusses the methods used for the dynamic impact assess-
ment, focusing on the common elements for the various impact categories. Sections
3 to 10 present methods and assumptions specific to the impacts, and discusses the
results. The impact categories considered are agriculture, forestry, water resources,
energy consumption, sea level rise, ecosystems, vector-borne diseases, and cardio-
vascular and respiratory diseases. Section 11 draws these impacts together to derive
a picture of the overall vulnerability to climate change. Section 12 concludes.

2. Some Notes on Methodology

The starting point of the analysis is the static impact assessment of Tol (2001).
That study draws on the literature on the impact of climate change and monetary
valuation. Selection criteria for a primary impact study to be taken up are that it
is global (for reasons of consistency) and that the climate scenario is taken from a
General Circulation Model (or reasons for realism). Different studies are combined,
and results are extrapolated using basic statistical methods (cf. Van den Bergh et
al., 1997). These statistical employ regularities in the data that are here used to
extrapolate the static findings over different climates and different vulnerabilities
to climate change. With regard to the latter, a crucial assumption is that current
DAMAGE COSTS OF CLIMATE CHANGE 137

variations in vulnerability related to, say, per capita income can be used to estimate
the effect of economic growth on vulnerability to climate change.
Impacts are estimated for nine regions, viz. OECD-America, OECD-Europe,
OECD-Pacific, Central and Eastern Europe and the former Soviet Union, Middle
East, Latin America, South and Southeast Asia, Centrally Planned Asia, and
Africa. These regions coincides with the regions of the FUND model (Tol, 1997,
1999).
Parameters are estimated from the underlying literature or guessed. Parameter
estimates and parameter guesses are presented as their most likely values, accom-
panied by their 67% confidence interval. If the probability distribution function of
the parameter is normal – a standard assumption – the confidence interval spans the
mode (expectation, median) plus or minus the standard deviation. Standard devi-
ations are guessed or estimated from the underlying literature. The values of the
standard deviations and the shapes of the probability distributions are not crucial
as their paper contains no uncertainty analysis; the standard deviations are used for
sensitivity analysis only.
Sensitivity analyses are performed on the most important parameters. Para-
meters are varied, one at the time, between the best guess plus and minus the
standard deviation. Because so many of the assumptions are heroic, the sensitivity
analyses are perhaps more informative than the central estimates. The sensitivity
analyses also seek to inform the questions how comprehensive and complicated
a climate change impact model needs to be, and what are the minimum set of
relations and indicator that capture most of the dynamics.

3. Agriculture
3.1. MODEL
Tol (1999) presents results for the impact of climate change on agriculture, based
on the studies of Darwin et al. (1995), Kane et al. (1992), Reilly et al. (1994),
Rosenzweig and Parry (1994), and Tsigas et al. (1996). Each of these studies
combines estimates of changes in crop yield or land productivity with a model
of national and international trade in agricultural products. Table I reproduces the
results for a 1 ◦ C increase in the global mean temperature and a rate of 0.04 ◦ C/year.
The estimate for the level of climate change is based on the underlying results
including adaptation. The estimate for the rate of change is based on the difference
of the underlying results with and without adaptation. The underlying studies do
not contain information about the speed of climate change relative to the speed of
adaptation. Common sense suggests that the impact reacts more than linear to the
rate of climate change. It is likely that farmers will need a couple of years to adjust
their practices to the changed climatic circumstances. The assumed model is:
138 RICHARD S.J. TOL

Table I. Impacts of climate change on agriculture.

Region Rate of change Level of change Optimal temperature

%GAP/0.04 ◦ C) (%GAP/1 ◦ C) (
◦ C wrt 1990)

OECD-A −0.021 (0.031) 0.398 (0.530) 2.29 (1.32)

OECD-E −0.026 (0.025) 0.838 (0.450) 0.45 (0.50)
OECD-P −0.016 (0.038) 0.321 (0.648) 2.71 (0.33)
CEE&fSU −0.028 (0.027) 1.060 (0.452) 2.96 (0.43)
ME −0.017 (0.011) 0.233 (0.193) 3.08 (0.49)
LA −0.022 (0.015) 0.221 (0.280) 2.14 (0.26)
S&SEA −0.022 (0.007) 0.253 (0.132) 2.16 (0.33)
CPA −0.023 (0.023) 1.239 (0.403) 3.41 (1.01)
AFR −0.12 (0.006) 0.189 (0.111) 3.00 (0.48)

Source: Tol (2002).

β

Tt
r
at,r =a + ρatr−1,r (1)
0.04

ar denotes the change in agricultural production due the rate of climate change; t
denotes time; r denotes region;
T denotes the change in the global mean temper-
ature; α is a parameter, denoting the benchmark change in agricultural production
(cf. Table I); β is a parameter, denoting the non-linearity of the reaction to temper-
ature; β = 2 (1.5–2.5); ρ is a parameter, denoting the speed of adaptation; ρ = 10
(5–15).
The model for the impact due to the level of climate change is less specu-
lative, as the underlying literature contains more information to this regard. The
change in agriculture product (after adaptation) was regressed on the change in
the global mean temperature and the global mean temperature squared. This was
only possible for the studies by Darwin et al. (1995), Reilly et al. (1994), and
Rosenzweig and Parry (1994), as the studies by Kane et al. (1992) and Tsigas et
al. (1996) are for one GCM only. A regression on temperature and temperature
squared implies an optimal temperature, which was estimated for each region for
the three studies. The weighted average of the three estimates is used here. The
results are given in Table I. The assumed model is:
opt
−2ABr Tr ABr
l
at,r = opt Tt + opt Tt
2
(2)
1 − 2Tr 1 − 2Tr
al denotes the change in agricultural production due to the level of climate change;
t denotes time; r denotes region; T denotes the change in global mean temperature
relative to 1990; AB is a parameter, denoting the benchmark change in agricultural
production (cf. Table I); Topt is a parameter, denoting the optimal temperature (cf.
Table I).
DAMAGE COSTS OF CLIMATE CHANGE 139

Equation (2) is problematic if the optimal temperature is less than 0.5, as the
optimum then is a minimum (rather than a maximum). Therefore, for Topt < 0.5,
(2) is replaced by a bilinear form. If T = Topt , al = 2AB . If T = 0, al = 0. If T = 1,
al = AB . This ad hoc solution is of little consequence, as optimal temperature of
0.50 C or less only occur in OECD-Europe, and only occasionally so.1
The share of agricultural production in total income falls with per capita income.
The elasticity across the nine regions is −0.31. So,


GAPt,r GAP1990,r y1990,r ε
= (3)
Yt,r Y1990,r yt,r
GAP denotes gross agricultural product; Y denotes gross domestic product; y
denotes gross domestic product per capita; t denotes times; r denotes regions; ε
is a parameter; ε = 0.31 (0.15–0.45).

4. Results
Figure 1 displays a sensitivity analysis around the impacts of climate change with
full adaptationj. In all cases, impacts are fairly limited, never exceeding a positive
or negative 0.1% of GDP. For most parameter choices, impacts are slightly positive,
or negative 0.1% of GDP. For most parameter choices, impacts are slightly positive,
but start falling in the 22nd century. In the last decade, the positions of high and
low impacts reverse, because the impact on regions with a high sensitivity and a
near optimum (e.g., OECD-America, South and Southeast Asia) starts to dominate.
If the optimum temperature to grow agriculture products is high, the pattern is
the same, but stretched over time. If the optimum temperature is low, negative
consequences of climate change emerge sooner.
Figure 2 displays the result of a sensitivity analysis around the impacts of agri-
cultural adaptation to climate change. The estimated impact is more important than
the estimated speed of adaptation. However, comparing Figures 1 and 2, the impact
with full adaptation is much more important than the impact of adaptation. This
reduces the influence of the ad hoc assumptions underlying (1).

5. Forestry
5.1. MODEL
The results for the various GCMs scenarios used by Perez-Garcia (1995) suggest
that the relationship between the climate change impact on the forestry sector and
the global mean temperature is linear.
Following Mendelsohn and Schlesinger (1997), forestry is assumed grow at the
same relative pace as does agriculture.
The model is thus:


yt,r ε β
Ft,r − ar Tt (4)
y1990,r
140 RICHARD S.J. TOL

Figure 1. A sensitivity analysis around the impact of climate change on world agriculture
with full adaptation. Displayed are the central case, and cases in which the impact with full
adaptation is set at its central estimate plus or minus its standard deviation (high and low
impact), and in which the optimal temperature is set at its central estimate plus or minus its
standard devaiation (high and low optimum).

where F denotes the change in forestry consumer and producers surplus (as a share
of total income); t denotes time; r denotes region; y denotes per capita income;
T denotes the global mean temperature; α is a parameter; see Table II; ε = 0.31
(0.11–0.51); β is a parameter; β = 1 (0.5–1.5).

5.2. RESULTS
Figure 3 displays the results. The impact of climate change on forestry is positive
but small, primarily because forestry is only a small fraction of the world economy.
At the short term, the benchmark estimate is an important of uncertainty, but at the
longer term the relative growth rate of the sector takes over.

6. Water Resources
6.1. MODEL
The impact of climate change on water resources is found to be highly uncertain,
but potentially very substantial in the static assessment of Tol (1999). Part of the
uncertainty is due to the complexities of precipitation, and part of the uncertainty
is due to the paucity of global assessments which can be used in for economic
DAMAGE COSTS OF CLIMATE CHANGE 141

Figure 2. A sensitivity analysis around the adaptation of world agriculture to climate change.
The central line depicts total climate change impacts with all parameters set to their central
value. In the inner interval, the speed of adaptation is varied between its central estimate plus
or minus its standard deviation. In the outer interval, the estimated costs of adaptation are
varied between the central estimates plus or minus its standard devaiation.

Table II. Impact of α 1 ◦ C warming on current day forestry, water, heating, and
cooling, in million US dollar.

Region Forestry Water Heating Cooling

OECD-A 218 (24) −3 (3) 22 (22) −11 (11)

OCED-E 134 (16) −2 (2) 13 (13) −20 (20)
OECD-P 93 (20) −0 (0) 7 (7) −1 (1)
CEE&fSU −136 (17) −76 (76) 46 (46) −19 (19)
ME 0 (0) −1 (1) 8 (8) −1 (1)
LA −10 (2) −1 (1) 3 (3) −2 (2)
S&SEA 140 (34) −2 (2) 4 (4) −4 (4)
CPA 0 (0) 2 (2) 17 (17) −12 (12)
AFR 0 (0) −2 (2) 0 (0) −5 (5)

Source: Tol (2002).

142 RICHARD S.J. TOL

Figure 3. A sensitivity analysis around the costs of climate change to world forestry. The
central line depicts the costs if all parameters are set to their central estimates. In the inner
interval, impact estimates are varied between the central estimates plus or minus their standard
deviations. In the outer interval, the rate at which the forestry sector grows relative to the total
income is varied between its central estimates plus or minus its standard deviation.

estimates. Tol (1999) relies on a single study, viz. Downing et al. (1995, 1996),
which is incomplete in its reporting.
The model used here is therefore ad hoc. The impact of climate change on water
resources follows:


Yr,t β γ
Wt,r − ar Tt (5)
Yr,1990
where W denotes the change in water resources, expressed in billion dollars; t
denotes time; r denotes region; Y denotes income; T denotes the global mean
temperature; α is a parameter, the benchmark estimate; see Table II; α is a para-
meter; β = 0.85 (0.70–1.00), as in Downing et al. (1995, 1996); γ is a parameter;
γ = 1 (0.5–1.5).

6.2. RESULTS
Figure 4 displays a sensitivity analysis around the results. The exact impact
depends on a number of factors. The overall pattern, however, is that the impact
is substantially negative. In the central case, it gradually tends to −1.0% of world
GDP by 2200. Varying the elasticity of water demand to income, this could
DAMAGE COSTS OF CLIMATE CHANGE 143

Figure 4. A sensitivity analysis around the impact of climate change on world water resources.
Displayed are the central case, in which all parameters are set to their central estimates, cases
with high and low growth relative to total economic growth, cases in which water resources
react more and less than linearly to climate change, and the case in which the benchmark
impact is set at its central estimate plus its standard devaiation.

range between −0.5–1.5% of GDP. The uncertainty about the linearity of water
resources’ sensitivity to climate change is even more important. In the short run,
however, the estimated sensitivity of water resources of climate change is the most
important variable.

7. Energy consumption
Tol’s (1999) static assessment of the impact of climate change on energy consump-
tion is based on Downing et al. (1995, 1996). The dynamic assessment here draws
on the same source.

7.1. MODEL FOR HEATING ENERGY

Downing et al. (1995, 1996) argue that the demand for heating is approximately
linear in temperature change. The market for space heating is almost saturated.
Downing et al. (1995, 1996) suggest an income elasticity of 0.2, based on UK
data. This leads to unrealistic results for the poorer regions (Tol, 1999). Mori and
Takahashi (1998) suggest an income elasticity of 0.8, which leads to more realistic
144 RICHARD S.J. TOL

results. Demand obviously increases with the number of people. Technological

development reduces the costs of delivering heat. The model is:


 t
β yt,r ε Pt,r
SHt,r = ar Tt AEEIs,r (6)
yt,1990 Pt,1990 s=1990

SH denotes the amount of money spent less on spacing heating; t denotes time; r
denotes region; T denotes the change in the global mean temperature relative to
1990; γ denotes per capita income; P denotes population size; α is a parameter; cf.
Table II; β is a parameter; β = 1 (0.5–1.5); ε = 0.8 (0.6–1.0); AEEI is a parameter;
it is about 1% per year in 1990, converging to 0.2% in 2200; cf. Tol (1997); its
standard deviation is set at a quarter of the mean.

7.2. MODEL FOR COOLING ENERGY

Following Downing et al. (1995, 1996), the demand for cooling is roughly linear
in temperature change. The market for cooling is substantially less saturated, but
Downing et al. (1995, 1996) do not report an estimate of the elasticity, so Mori
and Takahashi’s (1998) value of 0.8 is used. The model used here is based on
elasticities which are constant over time, for want of more detailed information
about air conditioning saturation levels. Demand for cooling increases with the
number of people. Technological development reduces the costs. The model is:


 t
β yt,r ε Pt,r
SCt,r = ar Tt AEEIs,r (7)
yt,1990 Pt,1990 s=1990

SC denotes the amount of money spent additionally on space cooling; t denotes

time; r denotes region; T denotes the change in the global mean temperature relative
to 1990; y denotes per capita income; P denotes population size; α is a parameter;
cf. Table II, β is a parameter; β = 1 (0.5–1.5); ε is a parameter; ε = 0.8 (0.6–1.0);
AEEI is a parameter (see above).

7.3. RESULTS
Figure 5 displays the results for heating. Under the best guess assumptions, the
money saved from a reduced demand for heating remains below 1.0% of GDP. In
the long run, impact starts levelling out due to improved energy efficiency. The
income elasticity of heating energy demand adds more to the uncertainty than does
the AEEI.
Figure 6 displays the results for cooling. Under the guess assumptions, the addi-
tional amount spent on cooling rises to 0.6% of GDP. The patterns and sensitivities
are the same for cooling and heating energy demand. Instead of the AEEI, the
sensitivity to β of equation (7) is displayed. The linearity of cooling demand in
the temperature is important for the intertemporal distribution of the impact, and,
DAMAGE COSTS OF CLIMATE CHANGE 145

Figure 5. A sensitivity analysis around the impact of climate change on the demand for
heating energy. The central line depicts the costs if all parameters are set to their central esti-
mates. In the inner interval, the rate of technological progress is between its central estimate
plus or minus its standard deviation. In the outer interval, the rate at which the heating enegy
demand grows relative to income is varied between its central estimates plus or minus its
standard deviation.

in the long run, a little less important than the income elasticity of cooling energy
demand.

8. Sea Level Rise

8.1. MODEL
The impact of sea level rise on coastal systems is outlined in great detail in Tol
(1999), drawing on the data of Hoozemans et al. (1993) and Bijlsma et al. (1996).
Table III shows the accumulated loss of drylands and wetlands for a one metre rise
in sea level. Land loss is assumed to be a linear function of sea level rise. The value
of dryland is assumed to be linear in income density ($/km2 ), with an average value
of $4 million per square kilometre for the OECD. Tol (1999) uses this assumption
to extrapolate across regions. It is used here to extrapolate over time. Wetland value
is assumed to be logistic in per capita income, with an average value of $5 million
per square kilometre for the OECD. These values are based on Fankhauser (1995).
If dryland gets lost, the people living there are forced to move. The number of
forced migrants follows from the amount of land lost and the average population
density in the region. Following Tol (1995), the value of this is set at three times
the regional per capita income per migrant.
146 RICHARD S.J. TOL

Figure 6. A sensitivity analysis around the impact of climate change on the demand for
cooling energy. The central line depicts the costs if all parameters are set to their central
estimates. In the inner interval (after 2040), the reaction of cooling energy demand to climate
change is varied between more and less then linear. In the outer interval, the rate at which the
cooling energy demand grows relative to income is varied between its central estimates plus
or minus its standard deviation.

Land loss and migration is reasonably straightforward. However, people will not
sit idle while their land get inundated. Instead, they will try to protect what they
think is worth protection. Table III dipslays the annual costs of fully protecting
all coasts against a one metre sea level rise in a hundred years time (based on
Hoozemans et al., 1993). If sea level would rise slower than that (as projected
by the IPCC), annual costs are assumed to be proportionally lower. The level of
protection, that is, the share of the coastline protected, is assumed to be based on a
cost-benefit analysis, following Fankhauser (1994):


1 PC + WL
L = min 0, 1 − (8)
2 DL

L is the fraction of the coastline to be protected. PC is the net present value of

the protection if the whole coast is protected. The GVA reports average costs per
year the next century – see Table III. PC is calculated assuming annual costs to
be constant. This is based on the following. Firstly, the coastal protection decision
makers anticipate a linear sea level rise. Secondly, coastal protection entails large
infrastructural works which last for decades. Thirdly, the considered costs are direct
investments only, and technologies for coastal protection are mature.
DAMAGE COSTS OF CLIMATE CHANGE 147
Table III. Impact of a one metre sea level rise.

Level Dryland Dryland Wetland Wetland Protection

prot. loss value loss value costs Emigrants
(%) (103 km2 ) (106 $/km2 ) (103 km2 ) (106 $/km2 ) (109 $) (106 )

OECD-A 0.77 4.8 (2.4) 1.3 (0.6) 12.0 (8.6) 5.4 (2.7) 83 (74) 0.13 (0.07)
OECD-E 0.86 0.7 (0.4) 13.1 (6.6) 4.0 (2.3) 4.3 (2.2) 136 (45) 0.22 (0.10)
OECD-P 0.95 0.3 (0.4) 13.7 (6.7) 1.0 (1.1) 5.9 (2.9) 63 (38) 0.04 (0.02)
CEE&fSU 0.93 1.2 (2.7) 0.9 (0.5) 0.0 (0.0) 2.9 (1.5) 53 (50) 0.03 (0.03)
ME 0.30 0.6 (1.2) 0.5 (0.3) 0.0 (0.0) 1.3 (0.7) 5 (3) 0.05 (0.08)
LA 0.86 7.8 (7.1) 0.3 (0.2) 50.2 (36.4) 0.9 (0.5) 147 (74) 0.71 (1.27)
S&SEA 0.93 9.3 (9.6) 0.5 (0.3) 54.9 (48.0) 0.3 (0.2) 305 (158) 2.30 (1.40)
CPA 0.93 8.4 (15.1) 0.3 (0.2) 15.6 (17.1) 0.2 (0.1) 171 (126) 2.39 (3.06)
AFR 0.89 15.4 (18.4) 0.4 (0.2) 30.8 (14.8) 0.4 (0.02) 92 (35) 2.74 (2.85)

Source: Tol (2002).

WL is the net present value of the wetlands lost due to full coastal protection.
Land values are assumed constant, reflecting how much current decision makers
care about the non-marketed services and goods that get lost. The amount of
wetland lost is assumed to increase linearly over time.
DL denotes the net present value of the dryland lost if no protection takes place.
Land values are assumed to rise at the same pace as the economy grows. The
amount of dryland lost is assumed to increase linearly over time.
Throughout the analysis, a pure rate of time preference, ρ, of 1% per year is
used. The actual discount rate lies thus 1% above the growth rate of the economy,
g. The net present costs of protection PC thus equal
∞
t
1 1+ρ+g
PC = P Ca = P Ca (9)
t =1
1+ρ+g ρ+g

where PCa is the average annual costs of protection.

The net present costs of wetland loss WL follow from
∞
t
1 1+ρ+g
WL = t W L0 = DL0 (10)
t =1
1+ρ +g (ρ + g)2

where WL0 denotes the value of wetland loss in the first year.
The net present costs of dryland loss DL are
∞
t
1 (1 + g)(1 + ρ + g
DL = t DL0 = DL) (11)
t =1
1+ρ+g ρ2

where DL0 is the value of dryland loss in the first year.

148 RICHARD S.J. TOL

Figure 7. A sensitivity analysis around dryland losses due to sea level rise. Displayed are,
from top to bottom, high losses without protection, central losses without protection, central
losses with high protection, central losses with central protection, central losses with low
protection, and low losses without protection.

8.2. RESULTS
Figure 7 displays a sensitivity analysis around the loss of drylands to sea level
rise. The amount of land lost is varied, and the value of drylands is varied so as
to influence the level of protection. The sensitivity to sea level rise dominates the
influence of land values. The influence of the somewhat speculative model of the
adaptive behaviour of coastal zone managers is thus mitigated.
Figure 8 displays a sensitivity analysis around forced migration and its costs.
The number of migrants first rises steeply as sea level rise accelerates, then stabil-
izers, and then gradually falls as more and more land gets protected. In the central
case, the number of forced migrants never exceeds 75,000 people per year, a
relatively small number. The costs of migration follow a similar pattern, but then
against a trend of ever increasing values attached to each migrant.
DAMAGE COSTS OF CLIMATE CHANGE 149

Figure 8. A sensitivity analysis around (the cost of) forced migration due to sea level rise.
Depicted are the number of migrants in the case that dryland loss (with protection) is set to its
central estimate, plus or minus its standard deviation. Also depicted are high, middle and low
estimates of the costs for the case in which dryland loss and the number of migrants are set to
their central estimates.

9. Ecosystems
9.1. MODEL
Tol (1999) assesses the impact of climate change on ecosystems, biodiversity,
species, landscape etcetera based on the “warm-glow” effect. Essentially, the value
people are assumed to place on such impacts are independent of any real change in
ecosystems etcetera. This values is specified as
yt,r yt,r /yb
Et,r = a Pt,r (12)
y1990,r 1 + yt,r /yb
where E denotes the value of the loss of ecosystems; t denotes times; r denotes
region; y denotes per capita income; P denotes population size; α is a parameter
such that E equals $50 per person if per capita income equals the OECD average
in 1990; yb is a parameter; yb = $20,000.

9.2. RESULTS
Figure 9 displays some results. In the central case, welfare losses increase from
some 0.25% of world GDP to about 0.5%. The estimate is very uncertain, though,
particularly with respect to the value attached to species loss. Welfare losses could
be less then 0.2% of GDP, but also exceed 1.0%.
150 RICHARD S.J. TOL

Figure 9. A sensivity analysis around the costs of loss of species, ecosystems, landscapes and
so on due to climate change. The central line depicts the case in which all parameters are set
to their central value. In the inner interval, the elasticity of valuation to per capita income is
varied between its central estimate plus or minus its standard deviation. In the outer interval,
the value people attach to ecosystem change is varied between its central value plus or minus
its standard deviation.

10. Vector Borne Diseases

10.1. MODEL
Tol (1999) presents estimates of the change in mortality due to vector-borne
diseases (viz., malaria, schistosomiasis, dengue fever) as the result of a one degree
increase in the global mean temperature. The estimates results from overlaying the
model-studies of Martens et al. (1995, 1997), Martin and Lefebvre (1995), and
Morita et al. (1994) with mortality figures of the WHO (Murray and Lopez, 1996)
for the nine regions identified in this study. Table IV summarizes the findings.
Martens et al. (1995, 1997) standardize their results to an increase in the global
mean temperature of 1.16 ◦ C. Martin and Lefebvre (1995), and Morita et al. (1994),
however, present their results (for malaria only) for various increases in the global
mean temperature (2.8 ◦ C to 5.2 ◦ C). Both studies suggest that the relationship
between global warming and malaria is linear. This relationship is assumed to apply
to schistosomiasis and dengue ever as well.
Vulnerability to vector-borne diseases strongly depends on basic health care and
the ability to purchase medicine. These factors are assumed to be linearly related to
per capita income. The data of the WHO (WP and Lopez, 1996)) suggest a linear
relationship between per capita income and mortality due to malaria, schistosomi-
DAMAGE COSTS OF CLIMATE CHANGE 151
Table IV. Additional deaths due to vector-borne diseases for a 1 ◦ C global
warming.

Region Malaria Schistosomiasis Dengue fever

OECD-A 0 (0) 0 (0) 0 (0)

OECD-E 0 (0) 0 (0) 0 (0)
OECD-P 0 (0) 0 (0) 0 (0)
CEE&fSU 0 (0) 0 (0) 0 (0)
ME 155 (112) −64 (13) 0 (0)
LA 1,101 (797) −114 (22) 0 (0)
S&SEA 8,218 (5949) −116 (3) 6,745 (1,171)
CPA 0 (0) −128 (25) 393 (68)
AFR 56,527 (40,919) −503 (99) 343 (60)

Source: Tol (2002).

asis, and dengue fever for the Middle East, Latin America, and South and Southeast
Asia. Centrally Planned Asia (too low mortality) and Africa (too high) mortality
are outliers. A regression of vector-borne mortality and per capita income suggests
that populations with an income above $3100 per head, with a standard deviation
of $260/head, are not vulnerable to vector-borne diseases. Because of the outliers,
the standard deviation is increased to $1000/head.
The model for vector-borne diseases thus becomes:


β yc − yt,r γ
mr,t,d = ar,d Tt if yt,r ≤ yc (13)
yc − y1990,r
while mt,r,d = if yt,r > yc ; m denotes mortality; t denotes time; r denotes regions;
d denotes disease; α is parameter, the benchmark impact of climate change on
vector-borne diseases; cf. Table IV; y denotes per capita income; T denotes the
change in the global mean temperature relative to 1990; yc is a parameter, denoting
the per capita income at which vector-borne mortality becomes zero; yc = $3100
(2100–4100); β and γ are parameters, denoting the non-linearity of mortality in
temperature and income, respectively; α = 1 (0.5–1.5); γ = (0.5–1.5).
Mortality is valued at 200 times the per capita income, with a standard deviation
of 100, comparable to the assumptions of Tol (1995).

10.2. RESULTS
Figure 10 displays the additional number of deaths due to malaria, schistosomiasis,
and dengue fever. Figure 11 displays the value placed at that. In the central case,
the number of additional vector-borne deaths never exceeds 10,000 people per year.
The number falls to zero by about 2080, as all regions acquire a sufficient standard
of living to afford effective health care. The timing of this obviously varies with
152 RICHARD S.J. TOL

Figure 10. A sensivity analysis around the increase in vector-borne mortality due to climate
change. The sensitivity, and the per capita income at which people have sufficient means to
effectively protect themselves are varied between their central estimates plus or minus their
standard deviations.

the assumed cut-off per capita income. The uncertainty about the sensitivity of
vector- borne morality to climate change is more important for the total number of
additional diseases. In Figure 11, the uncertainty about this sensitivity is also more
important, albeit slightly, than the uncertainty about the value of a statistical life.
In all cases, the total welfare loss is low, due in part to the relatively small number
of premature deaths, and due in part to the fact that only poor people suffer from
vector-born diseases.

11. Heat and Cold Stress

11.1. MODEL
Cardiovascular and respiratory disorders are worsened by both extreme cold and
extreme hot weather. Martens (1998) assesses the increase in mortality for 17
countries. Tol (1999) extrapolates these findings to all other countries, based on
formulae of the shape:

M = a + βTB (14)
where
M denotes the change in mortality due to a one degree global warming;
TB is the current temperature in the country; and α and β are parameters. Equation
(14) is specified for populations above and below 65 years of age for cardiovascular
DAMAGE COSTS OF CLIMATE CHANGE 153

Figure 11. A sensivity analysis around the costs of increased vector-borne mortality. The
central line depicts the case in which all parameters assume their central estimate. In the inner
interval, the value of a statistical life is varied between its central estimate plus or minus its
standard deviation. In the outer interval, the sensitivity of vector-borne mortality is varied
between its central estimate plus or minus its standard deviation.

disorders. Cardiovascular mortality is affected by both heat and cold. In the case of
heat, TB denotes the average temperature of the warmest month. In the case of cold,
TB denotes the average temperature of the coldest month. Respiratory mortality is
not age-specific.
Equation (14) is readily extrapolated. If warming exceeds one degree, the
baseline temperature TB changes. If this change is proportional to the change in
the global mean temperature, the equation becomes quadratic. Summing country-
specific quadratic functions results in quadratic functions for the regions:

M = aT + βT 2 (15)

where T denotes the change in global mean temperature; α and β are parameters,
derived, as described above, from the benchmark estimates of Table V.
One problem with (15) is that it is a non-linear exrtrapolation based on a data-
set that is limited to 17 countries and, more importantly, a single climate change
scenario. A global warming of 1 ◦ C leads to changes in cardiovascular and respira-
tory mortality in the order of magnitude of 1% of baseline mortality due to such
disorders. Per cause, the total change in mortality is restricted to a maximum of
5% of baseline mortality. (This restriction is effective.) Baseline cardiovascular
and respiratory mortality derives from the share of the population above 65 in the
154 RICHARD S.J. TOL

Table V. Additional deaths (in thousands) due to cardiovascular and

respiratory diseases for a 1 ◦ C global warming.

OECD-A −64.4 (4.4) 11.4 (5.9) 3.0 (9.7)

OECD-E −99.8 (2.6) 11.7 (4.0) −2.8 (5.7)
OECD-P −13.1 (2.2) 3.5 (2.8) 1.0 (4.8)
CEE&fSU −87.5 (5.2) 10.7 (4.4) 4.5 (11.0)
ME −8.9 (1.3) 2.5 (0.4) 9.9 (2.6)
LA −20.0 (3.5) 8.1 (1.8) 11.1 (7.0)
S&SEA −63.8 (16.9) 17.5 (2.9) 141.2 (34.1)
CPA −103.4 (21.7) 24.3 (4.6) 62.8 (44.4)
AFR −18.2 (3.0) 4.7 (0.5) 24.8 (6.0)

Source: Tol (2002).

total population. If the fraction of people over 65 increases by 1%, cardiovascular

mortality increases by 0.0259% (0.0096%). For respiratory mortality, the change is
0.0016% (0.0005%). These parameters are estimated from the variation in popula-
tion above 65 and cardiovascular and respiratory mortality over the nine regions in
1990.
Mortality as in equations (14) and (15) is expressed as a fraction of population
size. Cardiovascular mortality, however, is separately specified for younger and
older people. In 1990, the per capita income elasticity of the share of the population
over 65 is 0.25, with a standard deviation of 0.08. This is used to generate the
scenario of people over 65. Heat-related mortality is assumed to be limited to urban
populations. The scenario for urbanization is based on Tol (1996, 1997).

11.2. RESULTS
Figure 12 displays some results. In the central case, climate change reduces
mortality, peaking at half a million avoided deaths around 2050. The uncertainty is
rather large. Changes in the population over 65 and consequent changes in baseline
mortality have a minor effect. Capping mortality changes at a small fraction of
baseline mortality has a major impact. The largest uncertainty, however, is the
sensitivity of cardiovascular and respiratory mortality to climate change. In the
year 2200, climate change may help to avoid almost 2.5 million premature deaths,
but also cause an additional 1 million deaths.

12. Total Impact

Figure 13 displays the aggregated impact for the nine regions. The picture is mixed,
with positive and negative climate change impacts for different regions at different
times. Overall, the impact of climate change on the OECD is positive, although
DAMAGE COSTS OF CLIMATE CHANGE 155

Figure 12. A sensivity analysis around the change in cardiovascular and respiratory mortality
due to climate change. In both panels, the central line depicts the case with all parameters
set to their central estimate. In the upper panel, in the inner interval, the reaction of the age
composition to changes in per capita income is varied bewteen its central estimate plus or
minus its standard deviation. In the lower panel, in the inner interval, the reaction of baseline
mortality to changes in age composition is varied between its central estimate plus or minus its
standard deviation. In the outer interval, the maximum allowed change in mortality is varied
between its central estimate plus or minus its standard deviation.
156 RICHARD S.J. TOL

Figure 13. Aggregated impact of climate change for nine world regions. The impact on
CEE&fSU is displayed at the left y-axis.

starting and ending in the negative. Latin America follows a similar pattern. So do
the Middle East and Centrally Planned Asia, but there the whole curve is shifted
towards the negative. The impact of climate change on South and Southeast Asia,
Central and Eastern Europe and the former Soviet Union is, on the whole, negative.
Africa sees only negative impacts. For comparison, the positive impact on the
OECD never exceeds 4% of their GDP whereas the negative impact on CEE&fSU
exceeds 8% of its GDP.
Aggregate impacts do not follow a smooth, readily simplified curve. Separate
modelling of impacts therefore seems necessary to paint a realistic picture of
aggregate impacts.

13. Discussion and Conclusions

Building a model of the impacts of climate change that takes proper account of
dynamics of the impact’s response to climate change and the changes in society’s
vulnerability proves to be a task which is beset with uncertainties and the need
DAMAGE COSTS OF CLIMATE CHANGE 157

for heroic assumptions. Despite that, a number of insights can be derived from the
model attempted here.
Climate change impacts can be positive as well as negative, depending on
the sector, region, and time one is looking at. The impact on social welfare thus
depends on how one decides to aggregate (cf. Fankhauser et al., 1997). In the poorer
regions, and in later times, the negative impacts tend to dominate the positive
impacts. The former reconfirms that climate change and greenhouse gas abatement
policy is essentially a problem of justice (cf. Arrow et al., 1996; Banuri et al.,
1996; Lind and Schuler, 1998). The latter makes that even countries which are
likely to gain from climate change have an incentive to reduce their greenhouse
gas emissions. For, the workings of the climate system are slow, and so are the
political, economic and technological mechanisms to abate emissions.
At the short term, the uncertainty about the sensitivity of a sector to climate
change is most important. As time progresses, however, the change in vulnerability
and the importance of that sector become the more important driving factors. These
factors may be subject to a proactive adaptation policy (Fankhauser et al., 1998).
One should be careful, however, to base policy conclusions on the findings
of this paper. Total climate change impacts, derived here, are less informative
than marginal costs of greenhouse gas emissions. The complexity of the dynamics
suggests that marginal costs are not trivially estimated. Uncertainty, although esti-
mated, are not assessed. Above, parameters are varied one at a time. Uncertainties
about scenarios for population and economic growth, and about the workings of
the climate systems are ignored. These issues will be treated in future papers.
Most of all, one should be careful because so many of assumptions are not
properly founded on a good understanding of the system. A lot of research remains
to be done regarding vulnerabilities to climate change.

Acknowledgements
Hadi Dowlatabadi, Tom Downing, Stale Navrud, Xander Olsthoorn, Martin Parry,
David Pearce, Pier Vellinga, Harmen Verbruggen, Aart de Vos and an anonymous
referees provided helpful comments on (parts of) earlier versions, or were good
discussion partners on the subject matter. The EU DG Research Environment and
Climate Programme through the ClimNeg, Inasud and ExternE projects (JOS3–
CT95–0002, ENV4–CT96–0179; JOS3– CT97–0015; ENV4–CT98–0810), the
US National Science Foundation through the Center for Integrated Study of the
Human Dimensions of Global Change (SBR–9521914) and the Michael Otto
Foundation provided welcome financial support. All errors and opinions are mine.

Note
1. The optimal temperatures in Table I are relative to 1990. The model, however, uses temperatures
relative to pre-industrial times. 1990 was about 0.5 ◦ C warmer than the pre-industrial climate.
158 RICHARD S.J. TOL

References
Arrow, K. J., W. R. Cline, K.-G. Maeler, M. Munasinghe, R. Squitieri and J.E. Stiglitz (1996),
‘Intertemporal Equity, Discounting, and Economic Efficiency’, in J. P. Bruce, H. Lee and E. F.
Haites, eds., Climate Change 1995: Economic and Social Dimensions – Contributions of Working
Group III to the Second Assessment Report of the Intergovernmental Panel on Climate Change.
Cambridge: Cambridge University Press, pp. 125–144.
Banuri, T., K.-G. Maeler, M. J. Grubb, H. K. Jacobson and F. Yamin (1996), ‘Equity and Social
Considerations’, in J. P. Bruce, H. Lee and E. F. Haites, eds., Climate Change 1995: Economic
and Social Dimensions – Contribution of Working Group III to the Second Assessment Report
of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press,
pp. 79–124.
Bijlsma, L., C. N. Ehler, R. J. T. Klein, S. M. Kulshrestha, R. F. McLean, N. Mimura, R. J. Nicholls,
L. A. Nurse, H. Perez Nieto, E. Z. Stakhiv, R. K. Turner and R. A. Warrick (1996), ‘Coastal Zones
and Small Islands’, in R. T. Watson, M. C. Zinyowera and R. H. Moss, eds., Climate Change
1995: Impacts, Adaptations and Mitigation of Climate Change: Scientific-Technical Analyses –
Contribution of Working Group II to the Second Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge: Cambridge University Press, pp. 289–324.
Darwin, R. F., M. Tsigas, J. Lewandrowski and A. Raneses (1995), World Agriculture and Climate
Change – Economic Adaptations, Vol. 703. Washington, D.C.: U.S. Department of Agriculture.
Downing, T. E., R. A. Greener and N. Eyre (1995), The Economic Impacts of Climate Change:
Assessment of Fossil Fuel Cycles for ExternE Project. Oxford and Lonsdale, Environmental
Change Unit and Eyre Energy Environment.
Downing, T. E., N. Eyre, R. Greener and D. Blackwell (1996), Projected Costs of Climate Change
for Two Reference Scenarios and Fossil Fuel Cycles. Oxford: Environmental Change Unit.
Fankhauser, S. (1994), ‘Protection vs. Retreat – The Economic Costs of Sea Level Rise’, Environ-
mental and Planning A 27, 299–319.
Frankhauser, S. (1995), Valuing Climate Change – The Economics of the Greenhouse. London:
EarthScan.
Fankhauser, S., R. S. J. Tol and D. W. Pearce (1997), ‘The Aggregation of Climate Change Damages:
A Welfare Theoretic Approach’, Environmental and Resource Economics 10, 249–266.
Fankhauser, S., J. B. Smith and R. S. J. Tol (1998), Weathering Climate Change: Some Simple Rules
to Guide Adaptation Decisions, Vol. D98/01. Amsterdam: Institute for Environmental Studies,
Vrije Universiteit.
Hoozemans, F. M. J., M. Marchand and H. A. Pennekamp (1993), A Global Vulnerability Analysis:
Vulnerability Assessment for Population, Coastal Wetlands and Rice Production and a Global
Scale (2nd revised edn.). Delft: Delft Hydraulics.
Kane, S., J. M. Reilly and J. Tobey (1992), ‘An Empirical Study of the Economic Effects of Climate
Change on World Agriculture’, Climate Change 21, 17–35.
Lind, R. C. and R. S. Schuler (1998), ‘Equity and Discounting in Climate Change Decisions’, in W.
D. Nordhaus, ed., Economics and Policy Issues in Climate Change. Washington, D.C.: Resources
for the Future, pp. 59–96.
Martens, W. J. M., T. H. Jetten, J. Rotmans and L. W. Niessen (1995), ‘Climate Change and Vector-
Borne Diseases – A Global Modelling Perspective’, Global Environmental Change 5(3), 195–
209.
Martens, W. J. M., T. H. Jetten and D. A. Focks (1997), ‘Sensitivity of Malaria, Schistosomiasis and
Dengue to Global Warming’, Climate Change 35, 145–156.
Martens, W. J. M. (1998), ‘Climate Change, Thermal Stress and Mortality Changes’, Social Science
and Medicine 46(3), 331–344.
Martin, P. H. and M. G. Lefebvre (1995), ‘Malaria and Climate: Sensitivity of Malaria Potential
Transmission to Climate’, Ambio 24(4), 200– 207.
DAMAGE COSTS OF CLIMATE CHANGE 159

Mendelsohn, R.O., W. N. Morrison, M. E. Schlesinger and N. G. Andronova (1996), A Global Impact

Model for Climate Change (draft).
Mendelsohn, R. O., W. N. Morrison, M. E. Schlesinger and N. G. Andronova (1996), Country-
Specific Market Impacts of Climate Change (draft).
Mendelsohn, R. O. and M. E. Schlesinger (1997), Climate-Response Functions (draft).
Mendelsohn, R.O. and (1998), ‘Climate Change Damages’, in W. D. Nordhaus, ed., Economics and
Policy Issues in Climate Change. Washington, D.C.: Resources for the Future, pp. 219–236.
Mori, S. and M. Takahashi (1998), ‘An Integrated Assessment Model for the Evaluation of New
Energy Technologies and Food Production – An Extension of Multiregional Approach for
Resource and Industry Model’, International Journal of Global Energy Issues 11(1–4), 1–17.
Morita, T., M. Kainuma, H. Harasawa, K. Kai and Y. Matsuoka (1994), An Estimation of Climate
Change Effects on Malaria. Tsukuba: National Institute for Environmental Studies.
Murray, C. J. L. and A. D. Lopez, eds. (1996), The Global Burden of Disease – A Comprehensive
Assessment by Mortality and Disability from Diseases, Injuries, and Risk Factors in 1990 and
Projected to 2020. Cambridge: Harvard University Press.
Nordhaus, W. D. and Z. Yang (1996), ‘RICE: A Regional Dynamic General Equilibrium Model of
Optimal Climate-Change Policy’, American Economic Review 86(4), 741–765.
Pearce, D. W., W. R. Cline, A. N. Achanta, S. Fankhauser, R. K. Pachauri, R. S. J. Tol and P.
Vellinga (1996), ‘The Social Costs of Climate Change: Greenhouse Damage and the Benefits
of Control’, in J. P. Burce, H. Lee and E. F. Haites, eds., Climate Change 1995: Economic
and Social Dimesnions – Contribution of Working Group III to the Second Assessment Report
of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press,
pp. 179–224.
Perez-Garcia, J. (1995), ‘Economic Impacts of Climate Change on the Global Forest Sector: An
Integrated Ecological/Economic Assessment’, conference paper, Bergendal.
Reilly, J. M., N. Hohmann and S. Kane (1994), ‘Climate Change and Agricultural Trade: Who
Benefits, Who Loses?’, Global Environmental Change 4(1), 24–36.
Rosenzweig, C. and M. L. Parry (1994), ‘Potential Impact of Climate Change on World Food
Supply’, Nature 367, 133–138.
Tol, R. S. J. (1994), ‘The Damage Costs of Climate Change – A Note on Tangibles and Intangibles,
Applied to DICE’, Energy Policy 22(5), 436–438.
Tol, R. S. J. (1995), ‘The Damage Costs of Climate Change Toward More Comprehensive
Calculations’, Environmental and Resources Economics 5, 353–374.
Tol, R. S. J. (1996), ‘The Damage Costs of Climate Change Towards a Dynamic Representation’,
Ecological Economics 19, 67–90.
Tol, R. S. J. (1997), ‘On the Optimal Control of Carbon Dioxide Emission: An Application of
FUND’, Environmental Modeling and Assessment 2, 151–163.
Tol, R. S. J. (1999), ‘Spatial and Temporal Efficiency in Climate Change: Applications of FUND’,
Environmental and Resource Economics 14(1), 33–49.
Tol, R. S. J. (2002), ‘Estimates of the Damage Costs of Climate Change, Part 1: Benchmark
Estimates’, Environmental and Resource Economics 21, 47–73.
Tol, R. S. J. and S. Fankhauser (1998), ‘On the Representation of Impact in Integrated Assessment
Models of Climate Change’, Environmental Modeling and Assessment 3, 63–74.
Tol, R. S. J., S. Fankhauser, R. G. Richels and J. B. Smith (2000), ‘How Much Damage Will Climate
Change Do? Recent Estimates’, World Economics 1(4), 179–206.
Tsigas, M. E., G. B. Frisvold and B. and Kuhn (1996), ‘Global Climate Change in Agriculture’, in
Hertel, T. W. Ed., Global Trade Analysis: Modeling and Applications. Cambridge: Cambridge
University Press.
Van den Bergh, J. C. J. M., K. J. Button, P. Nijkamp and G. C. Pepping (1997), Meta-Analysis in
Environmental Economics. Dordrecth: Kluwer Academic Publishers.
160 RICHARD S.J. TOL

Watson, R. T., M. C. Zinyowera and R. H. Moss, eds. (1996), Climate Change 1995: Impacts,
Adaptation, and Mitigation of Climate Change – Scientific-Technical Analysis – Contribution
of Working Group II to the Second Assessment Report of the Intergovernmental Panel on Climate
Change. Cambridge: Cambridge University Press.
Weyant, J. P., O. Davidson, H. Dowlatabadi, J. A. Edmonds, M. J. Grubb, E. A. Parson, R. G.
Richels, J. Rotmans, P. R. Shukla, R. S. J. Tol, W. R. Cline and S. Fankhauser (1996), ‘Integrated
Assessment of Climate Change: An Overview and Comparison of Approaches and Results’, in J.
P. Bruce, H. Lee and E. F. Haites, eds., Climate Change 1995: Economic and Social Dimensions
– Contribution of Working Group III to the Second Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge, pp. 367–396.