HOW CONSISTENT ARE ALTERNATIVE SHORT-TERM

CLIMATE POLICIES WITH LONG-TERM GOALS?

by

Valentina Bosetti
(Fondazione Eni Enrico Mattei, valentina.bosetti@feem.it)

Marzio Galeotti
(Università di Milano and Fondazione Eni Enrico Mattei)

Alessandro LANZA
(Eni and Fondazione Eni Enrico Mattei)

March 2004.

Abstract. Choosing long-term goals is a key issue in the climate policy agenda. Targets should be
easily measurable and feasible, but also effective in damage control. Once goals are set globally, given
the uncertainty affecting long-term strategies and region-specific preferences for different policy
instruments, policies will be better represented by a diversified portfolio to be revised over time, rather
than “once and forever” decisions. It therefore becomes crucial to understand to what extent different
strategies (or policy portfolios) are consistent with long-term targets, that is, when they imply emission
paths which do not irreversibly diverge from globally set goals.
The present paper aims to investigate emission paths implied by plausible policy scenarios against
those derived by imposing alternative long-term targets, comparing, for example, differences in peak
periods. Plausible policy scenarios are for instance Kyoto-type targets with or without participation by
the U.S. and/or by developing countries. Long-term targets considered are stabilisation of CO2
concentrations and limits to the increase in atmospheric temperature relative to pre-industrial levels.
In order to account for the uncertainty surrounding the climate cycle, for each long-term goal multiple
paths of emission - the most probable, the optimistic and the pessimistic ones - are considered in the
comparison exercise. Comparative analysis is performed using a newly developed version of the
FEEM-RICE model, a regional economy-climate model of optimal economic growth which is based
on Nordhaus and Boyer’s RICE model crucially extended in order to account for induced technical
change. In particular, both carbon and energy intensity are affected by a new endogenous variable –
Technical Progress – which captures both the role of Learning by Researching and of Learning by
Doing. These are in turn determined by the optimal levels of Research and Development and of
Emission Abatement.

Keywords: Climate Policy, Long-term climate targets.

JEL: H0, H2, H3

This paper is part of the research work being carried out by the Climate Change Modelling and Policy
Research Programme at Fondazione Eni Enrico Mattei.

1
1. Introduction

This paper uses the FEEM-RICE model in order to investigate the magnitude of
anthropogenic emissions in time, when different climate policies, such as Kyoto or other
plausible policy scenarios, are adopted. Such emission paths are then compared with
emissions deriving from scenarios where global targets, as a constraint on global temperature
or on atmospheric carbon concentrations, have been imposed.
The main motivation for this exercise derives from the consideration that uncertainty
concerning the climate change cycle makes it very hard for the scientific community to agree
upon an “acceptable” level of greenhouse gasses concentrations, see Schneider (2001). In
addition to this, climate sensitivity, defined as the global mean climatological temperature
change resulting from a doubling of atmospheric CO2 content, is known perhaps only to a
factor of three or less, for a detailed discussion see Caldeira, et al. (2003). For these reasons,
the main policy objective should be to keep human activities on a reversible path of emissions
which leaves open future options to stabilize concentration to some “secure” level.
“It is obvious that no ‘once forever’ solution exists....the most promising approach to
climate policy is sequential decision making...Short term strategies are then crafted so that
both GHG emissions and the underlying socio-economic processes (resource use,
technologies) evolve in a direction which makes future course corrections in any direction the
least expensive.” Toth and Mwandosya (2001)
Finally, in order to understand, at least partially, the effects played by uncertainty on
key factors, a set of globally-constrained scenarios has been simulated for different hypothesis
on the climate sensitivity parameter and on the carbon cycle transition coefficients.
This paper aims at getting some insights on how short term policies, which strongly
connected with the actual socio-politic scenario, are related with long-term goals, which are
generally proposed by the scientific community, with a particular attention to the main
uncertainty factors.
The rest of the paper is organized as follows. Section 2 summarizes and compare
advantages and disadvantages of alternative targets. Section 3 briefly presents the new
extension of the FEEM-RICE model adopted in the present analysis and describes in depth
how endogenous technical change is modeled. Section 4 is a commented presentation of
results.

2
2. Comparison of Alternative Targets

The climate change cycle (see Figure 1) consists of earlier stages, namely human and
natural activities producing greenhouse gasses emissions, and final stages, namely the damage
feedback effect, both on human activity and on the ecosystems. It is possible to consider
constraints and targets on each of these different phases of the cycle (dotted arrows in the
scheme of Figure 1). Different targets present advantages and shortcomings, see also Pershing
and Tudela (2003) for a comprehensive presentation.

GHG Average
Concentrations Atmospheric
Temperature

Emissions of Impacts on
GHGs Human,
Ecosystems

Human Activities:
Energy Production
and Consumption.
Industrial Processes,
Mitigation Land Use. Adaptation

Figure 1. The climate change cycle.

In general, focusing on earlier stages (as production or emissions), means having

clearer information on what the required effort should be, but it may not produce effectively
the desired effects, mainly because of the loosely relationship between actions and climatic
impacts. The reverse is true for targets imposed on later stages.
The IPCC conventionally has concentrate its attention on the earlier phases of the
cycle, namely imposing constraints on atmospheric emissions of tons of greenhouse gasses or
on emission intensity (emissions per unit of output), see IPCC (2001). Targets on emissions
are relatively easy to identify, to implement and to measure. However, given a certain level of
emissions through time, resulting concentrations of gasses in different layers of the

3
atmosphere are extremely uncertain and, consequently, it is hard to forecast how severe the
final impacts on the climate system and on human activities are going to be. Indeed, each
phase of the chain is highly characterized by uncertainty, thus making exact previsions
approximated and scarcely reliable. Anyhow, targets on actual and future emissions allow to
understand better who and when should undertake abatement efforts, thus providing clearer
grounds for the international equity debate and negotiations through the parts.
This becomes clearly less immediate when considering targets on atmospheric
concentrations of greenhouse gasses or on the rate and magnitude of the change of
atmospheric temperature, see also Manne and Richels (2003). Given the global nature of these
targets, it becomes necessary to think of ways of distributing accordingly emission rights
among different countries, following some rationale (as, for example, GDP per capita
considerations). Both in the case of atmospheric concentration and in the one of temperature
there are not complication in the measurement phase. However, as we mentioned,
complications arise when actually trying to commeasure efforts to the defined targets.
Finally, for what concerns targets on final stages of the climate cycle, the major
benefit would be that of having direct control on the entity of damage, that is exactly what a
policy finally would aim to control. The target could take the form of a limit on sea level rise,
loss of ecosystems or of economic activities or some other identifiable indicators. However,
the link between the object under control and the consequent abatement effort grows even
thinner, leaving this category of targets, although important within a more general debate on
climate change, aside from actual implementation phases.
We concentrate our attention on investigating and comparing the first three categories
of targets. Indeed we will consider policy scenarios, which commonly constrain either total or
relative emissions, and long-term stabilization scenarios, which constrain aggregate
atmospheric concentrations or atmospheric temperatures.

3. The FEEM-RICE Model

The FEEM-RICE model which we present hereafter is an extended version of the so-
called RICE-99 model by Nordhaus and Boyer (2000). RICE-99 is a Ramsey-Koopmans
single sector optimal growth model suitably extended to incorporate the interactions between
economic activities and climate. There is one such model for each of the eight macro regions
into which the world is divided: USA, Other High Income countries (OHI), OECD Europe

4
(Europe), Russia and Eastern European countries (REE), Middle Income countries (MI),
Lower Middle Income countries (LMI), China (CHN), and Low Income countries (LI).1

3.1 The Model General Structure

Within each region a central planner chooses the optimal paths of two controls, fixed
investment and carbon energy input, so as to maximize welfare, defined as the present value
of per capita consumption. The value added created via production (net of climate change)
according to a constant returns technology is used for investment and consumption, after
subtraction of energy spending. The technology is Cobb-Douglas and combines the inputs
from capital, labor and carbon energy together with the level of technology. Population (taken
to be equal to full employment) and technology levels grow over time in an exogenous
fashion, whereas capital accumulation is governed by the optimal rate of investment.
Compared to the previous RICE-96 model of Nordhaus and Yang (1996), this
specification contains a more detailed regional disaggregation of the world. However, the
main novelty of the new model is that a major previous limitation is overcome by the
introduction of an energy input. Because carbon dioxide is the only greenhouse gas
considered, the input is directly measured in carbon units. The carbon energy can be thought
of as the energy services derived from fossil-fuel consumption (e.g. derived from coal,
petroleum, and natural gas). An implication of the introduction of this crucial input is that its
market must be specified. While demand for carbon energy stems naturally from the first
principles of the entrepreneur (or social planner)’s problem, a supply curve for this input is
introduced somewhat ad hoc, and it allows for limited (albeit huge) long-run supplies at rising
costs. Because of the optimal-growth framework, carbon-energy is efficiently allocated across
time, which implies that low-cost carbon resources have scarcity prices (Hotelling rents) and
that carbon-energy prices rise over time.2 The carbon-energy input is modeled as being the
source of GHGs emissions in the production process, and the cumulated emissions (i.e.
concentrations) cause an increase in the worldwide temperature. To close the circle, global

1
The countries belonging to each one of the macro-regions above indicated are listed in Nordhaus and Boyer’s
book. We refer to it as RICE-99 because it has been made available by the authors through the Internet in 1999.
2
Thus the new version of RICE incorporates a treatment of energy supply, which is seen as an exhaustible
resource. Another addition is a revised and extended climate module which now includes a three-reservoir model
calibrated to existing carbon-cycle models. The equation of the original model retained in FEEM-RICE are
reported in the appendix.

5
temperature (relative to pre-industrial levels) is responsible for the wedge between gross
output and net of climate change effects.
In FEEM-RICE each country plays a non-cooperative Nash game in a dynamic setting
which yields an Open Loop Nash equilibrium. This is a situation where in each region the
planner maximizes its utility subject to the individual resource and capital constraints and the
climate module for a given emission production of all the other players.3 Kyoto-type of
international environmental agreements can be easily accommodated by adding a constraint
stating that regional emissions cannot exceed a given upper limit. It is also possible to
simulate the model in the presence of international emission trading: in this case the standard
identity between sources and uses of resources specifies that output is used for consumption
and investment, to which proceeds or sales from net imports of permits should be added.
In the case of permit trading a region’s emissions may exceed the limit set in Kyoto if
permits are bought, and vice versa in the case of sales of permits. Under the possibility of
emission trading the sequence whereby an equilibrium à la Nash is reached must be revised as
follows. Each region maximizes its utility subject to the individual resource and capital
constraints, now including the Kyoto constraint, and the climate module for a given optimal
set of strategies of all the other players and a given price of permits (in the first round this is
set at an arbitrary level). When all regions have made their optimal choices, the overall net
demand for permits is computed at the given price. If the sum of net demands in each period
is approximately zero, a Nash equilibrium is obtained; otherwise the price is revised in
proportion to the market unbalance and each region’s decision process starts again. Notice
that, in equilibrium, the marginal cost of domestically abating one more unit of GHGs
emissions is equal to the price of emissions rights, as in a perfectly competitive market;
however, in converging toward the equilibrium, each region is able to influence the price
formation. This might induce some regions to behave strategically on the emissions market.
Finally, note that the price of a unit of tradable emission permits is expressed in terms of the
numéraire output price and that in this case there is an additional policy variable, i.e. net
demands for permits.

3
As there is no international trade in the model, regions are interdependent just through climate variables. The
absence of trade of goods among regions is n important limitation of all regional versions of RICE. We plan to
address this issue in the near future.

6
3.2 The Treatment of Technical Change
The original RICE-99 model specifies the following production function (n indexes
regions, t time periods):

(1) Q(n, t ) = A(n, t )[ K F (n, t )1−γ −α n CE (n, t ) α n L(n, t ) γ ] − p nE CE (n, t )

where Q is output (gross of climate change effects), A the exogenously given level of
technology and KF, CE and L are the inputs from physical capital, carbon energy and labor,
respectively, and pE is fossil fuel price. Carbon emissions are proportional to carbon energy,
that is:

(2) E(n,t) = ς(n,t)CE(n,t)

where E is industrial CO2 emissions, while ς is an idiosyncratic carbon intensity ratio which
also exogenously declines over time.4 In this way, Nordhaus and Boyer (2000) make the
assumption of a gradual, costless improvement of the green technology gained by the agents
as time passes. For the reasons expounded in the previous section we consider this treatment
of TC inadequate for a model designed to study issues related to climate change.

In previous work we use the less satisfactory version of RICE and endogenized the
process of TC. We formulated two alternative versions of ETC-RICE. In the “Learning-by-
Researching” model an endogenously generated stock of knowledge affected both factor
productivity and the emission-output ratio (there was no energy input and emissions were
linked directly to unabated output). Knowledge was the result of intertemporal optimal
accumulation of R&D, where R&D is a choice variable describing a new investment
opportunity in addition to consumption and physical investment. This version was followed
by a “Learning-by-Doing” model, where knowledge, conceived as a stock of experience, was
approximated by installed capacity, in turn represented by physical capital accumulating
through periodic investment. Again, this stock of experience, affected both factor productivity
and the emission-output ratio. This LbD approach entailed one less choice variable with

4
Throughout the paper we will indifferently refer to ‘environmental’ technology or ‘green’ technology when
mentioning the time-varying coefficient ς.

7
respect to the R&D approach, but no further claim on resources created was made, in addition
to consumption and physical investment.

The main shortcomings of these earlier formulations mainly derived from the absence,
in the core model, of an explicit energy module. The absence of an energy production factor
made it impossible to capture the effects of TC on the energy intensity of production.
Moreover, the “Learning-by-Researching” and the “Learning-by-Doing” features of TC were
modeled separately, while it would appear appropriate to include both sources of TC in the
same model. Finally, approximating the stock of experience with physical capital was not
very accurate, but the presence of the abatement rate as a model control variable made it
difficult, if not impossible, to account for cumulated abatement efforts as the force driving the
learning process.5

In FEEM-RICE we are able to focus on two distinct sources of potential TC: the
energy intensity of production and the carbon intensity of energy use. These two aspects
allow us to address energy-saving as well as energy-switching issues. In this junction it is
perhaps useful to refer to a time-honored concept in environmental economics, the so-called
Kaya identity. In its generalized form, it can be represented as follows. Let i = 1,.., I be the
various GHGs emissions, E, j = 1,.., J be the various energy sources, S, k = 1,.., K be the
sectors in the economy, Y, and n = 1,.., N be the countries in the world. Then, the world
emissions of GHGs, Et, can be decomposed as follows:

 Eijkn  S jkn  Ykn  Yn 

(3) Et = ∑∑∑∑     L

i  S jkn
 Y
n k j  kn  Yn  Ln 

where L is the total population. Hence, world emissions are a product of two ‘forces’:
techno-economic forces, given by carbon intensity (E/S) and energy intensity (S/Y), and socio-
economic forces, given by output composition (Yk/Y) and output levels, as well as
demographic aspects (Y/L) and L. As mentioned above the RICE-99 model on which FEEM-
RICE is based, has a single economic sector, a single energy source, namely carbon energy,
CE, and endogenous emission are limited to industrial CO2. Thus, the relationship stated in
(3) can be rewritten for our specific case as:

5
Recall that cumulated abatement was the variable used by Goulder and Mathai (2000) in the LbD version of
their cost minimization model. The absence of the energy input, and therefore of an explicit price, made it also
impossible to carry out any policy exercise in energy or carbon taxation.

8
  E  CE n  Yn 
(3’) Et = ∑  n    L
n  CE n  Yn  Ln 

In addition to socio-economic forces, which are commonly modeled in endogenous

growth models, this extended version of the model allows to account for both the two techno-
economic forces, energy and carbon intensity6.

The main novelty of our new formulation hinges on a new variable, which we call
(with poor inventive activity) Technical Progress, which accounts both Learning-by-
Researching and Learning-by-Doing at the same time. As before, we assume that innovation
is brought about by R&D spending which contributes to the accumulation of the stock of
existing knowledge.7 In addition to this Learning-by-Researching effect, the model accounts
also for the effect of Learning-by-Doing, now modeled in terms of cumulated abatement
efforts. Thus, Technical Progress TP is defined as follows:

(4) TP (n, t ) = f [ ABAT (n, t ), K R (n, t )]

The variable TP is conceived to affect both energy intensity (i.e., the quantity of carbon
energy required to produce one unit of output) and carbon intensity (i.e., the level of
carbonization of primarily used fuels). In terms of Kaya identity we therefore endogenize the
“techno-economic forces” of emissions and are able to disentangle the two sources. More
specifically TP is formulated as a convex combination of the stocks of knowledge and
abatement:

(5) TP(n, t) = [ ABATS (n, t)c * KR (n, t) d ]

where K R (n, t ) is the stock of knowledge and ABATS represents the stock of cumulated
abatement, in turn defined as:

6
As in most of models present in literature, population is exogenously determined. An important future
development would be that of endogenizing demographic changes, including migration flows across regions.
7
It has to be pointed out that analysing R&D expenditures is complicated because i) R&D is not always
amenable to measurement and ii) there is a great deal of uncertainty in the ability of R&D to generate
technological change. These words of caution should be therefore borne in mind by the reader when going
through the paper.

9
(6) ABATS (n, t + 1) = (1 − δ A ) * ABATS (n, t ) + ABATF (n, t ) a

δ A being the depreciation rate of cumulated experience and ABATF the abatement flow. The
stock of knowledge, K R (n, t ) instead accumulates in the usual fashion:

(7) K R (n, t + 1) = R & D (n, t ) + (1 − δ R ) K R (n, t )

δ R being the depreciation rate of knowledge.

How does the Technical progress affect the rest of the economy? As seen in equation
(1), the factors of production are labour, physical capital and effective energy. Let us first
consider the effect of technical progress on factor productivity (the energy intensity effect). In
this case the production function is modified so that (1) is replaced by the following
specification:

(1’) Q(n, t ) = A(n, t )[ K F (n, t )1−α n (TP )−γ CE (n, t ) α n (TP ) L(n, t ) γ ] − pe (n, t ) * CE (n, t )

where:
β0n
(8) α n = g[TP (n, t )] =
2 − exp − β1n *TP ( n,t )

and β 0 n and β1n are region specific parameters. Thus, an increase in the endogenously

determined Technical Progress variable reduces – ceteris paribus – the output elasticity of the
energy input.
Let us now turn to the effect of technical progress on the carbon intensity of energy
consumption. As shown in (2) effective energy results from for both fossil fuels input use and
(exogenous) TC in the energy sector. We postulate in this case that TP serves the purpose of
reducing, ceteris paribus, the level of carbon emissions. More precisely:

 1 
(2’) E (n, t ) = h[CE (n, t ), TP (n, t )] = ς(n, t )  −ψ n *TP ( n ,t ) 
CE (n, t )
 2 − exp 

10
Here an increase in TP reduces progressively the amount of emissions generated by a unit of
fossil fuel consumed.
We finally recognize that R&D spending absorbs some resources, that is:

(9) Y (n, t ) = C (n, t ) + I (n, t ) + R & D(n, t ) + p (t ) NIP(n, t )

where Y is output net of climate change effects, C is consumption, I gross fixed capital
formation, R&D research and development expenditures, pP is the equilibrium price on the
emissions rights, and NIP is the net quantity of permits demanded on the relative markets
(when positive; otherwise, it just indicates the supplied quantity on the same market). In
summary, our formulation introduces R&D as a further strategic variable of the model that
contributes to output productivity. Knowledge is a substitute for experience, but both
quantities are typically positive and therefore affect carbon and energy intensities. It may be
noted that there is only one type of R&D efforts that both help saving energy and switching
the energy needs away from fossil fuels. Although in principle it could be argued that
innovation activity resulting in technologies using less energy is different from the innovation
activity resulting in the development in clean energy technologies, in practice accounting for
this fact in highly aggregate models like FEEM-RICE is probably not worth the
complications. Finally, “red” TC – that is purely productivity-enhancing TC (captured by the
A index in the production function) – has been kept exogenous, albeit time varying. This in
keeping with Clarke and Weyant (2000), though we believe in principle optimal “red” R&D
can be introduced in the model in a relatively easy way, and this is something we plan to do in
future work.

4. Simulations and Results Discussion

FEEM-RICE has been used to simulate both policy and long-term stabilization
scenarios, accounting for uncertainty on different key parameters.
The policy scenarios which have been simulated using the FEEM-RICE model are
four and are defined by different assumptions on the involvement of different areas of the
world and different time frames. In particular, Scenario 1 and 2 (see Table 1) represent usual
assumptions regarding the nearest future, while in Scenario 3 and 4 some kind of

11
engagements for subsequent commitment periods are considered. In the “Business as Usual”
case (Scenario 1) emissions are unconstrained. In the “Kyoto Forever without US” (Scenario
2), Annex B countries, except the US, have to comply with the Kyoto target for the first and
subsequent commitment periods. They are also allowed to trade emission permits on an
international market, while the US undertake abatement efforts according to an intensity
target. For the rest of the world no constraints on emissions are considered. As far as Annex B
countries without US are concerned, “2020 Global Target” (Scenario 3) replicates Scenario 2.
US hold the same reduction in terms of intensity target for 2010, minus 10% with respect to
Scenario 1 for the second and subsequent commitment periods. Developing countries adopt
the same unconstrained Scenario 1 hypothesis for 2010 and 2020, while 10% reduction vis-à-
vis Business as Usual scenario is considered for 2020 onwards. Scenario 4 differs from
Scenario 1 in the third commitment period when 2000 emission levels have to be replicated.
Long-term emission stabilization scenarios target are considered in Table 2. A first set
of constraints is on level of aggregate atmospheric concentration, from 500 to 750 ppm. A
second set is related to the increase in atmospheric temperature above the pre-industrial levels
(the range of maximum increases allowed considered goes from 2.3 to 3 degrees C). In the set
of long-term scenarios, stabilization level at 550 ppm, for concentration, and 2.5 degree C, for
temperature, have been considered as an appropriate target by the IPCC (2001).
Finally, each of the long-term stabilization scenarios have been simulated for different
values of the GHG – temperature sensitivity coefficient and of the GHG atmospheric
retention rate (upper box to upper box transition parameter), see Joos et al.(1999), Joos et al.
(2001) and Nordhaus and Popp (1997) for a more extensive treatment of carbon cycle
representation and parameter uncertainty. The first parameter specifies the increase in
temperature for the doubling of CO2 concentration. The value of this parameter is not known
with certainty, as we have mentioned beforehand, and may affect crucially forecasting
exercise results. Scenarios considering a cap on temperature increases have thus been
simulated letting the sensitivity parameter take the values 1.5, 2.5 and 4.5 degree C/per CO2
doubling, as these are commonly considered the most optimistic, probable and pessimistic
potential realization of the parameter. The second parameter considered is the transition
parameter representing the retention rate of emissions by the biosphere, in the three box
climate module of the original RICE-99 model, Nordhaus and Boyer (2000). Again, scenarios

12
considering cap on atmospheric concentration have been simulated for different values of this
parameter.
Table 3 describes different emission peak periods in all the considered scenarios. As
expected, peak periods occur earlier for the reported long-term emission stabilization
scenarios, for most of the areas. Moreover, a constraint on temperature increase above pre-
industrial levels of 2.5 degree C seems to imply an earlier emission peak period than those
implied by a constraint on carbon concentration at 550 ppm.
Figure 1 plots the emission paths for different Annex B countries when no constraints
on emission are imposed (Scenario 1), together with maximum emission levels for the same
areas, compatible with a global cap on concentration of 550 ppm. Figure 2 reports the same
plot for the rest of the world. Different assumptions for constrained and unconstrained
scenarios imply emission paths at world level that are reported in Figure 3. The most plausible
long-term stabilization target, namely the one derived imposing a cap of 550 ppm on
atmospheric carbon concentration, starts to diverge radically from Scenario 1 from 2015. This
seems to suggest that policy actions should be very tight from this date even if the difference
in terms of global emissions in 2035 is less than 1 gton of carbon. Differences among long-
term and policy scenarios are larger if temperature stabilization scenarios are considered. In
order to reach at maximum an increase of 2.5 degree C above pre-industrial levels, policy
actions should be implemented even sooner. Differences in terms of global emissions are
evident as early as 2005.

Finally, Figure 6 and 7 plot emission paths for the four policy scenarios, together with
those predicted by simulating just one ‘sensible’ long-term scenario with a global target on
atmospheric concentrations of CO2 (550 ppm) and on temperature increase (2.5 degree C),
respectively, but for different values of the above mentioned key uncertain parameters.
Indeed, in Figure 6 a ‘pessimistic’, an ‘optimistic’ and a ‘most probable’ emission paths are
indicated given the same constraint on global atmospheric concentrations of 550 ppm. The
above discussed consideration that Scenario 1 would start to diverge radically from the 550
ppm scenario from year 2015 on, is here reinforced by the fact that this appears to be true also
considering an ‘optimistic’ value of the carbon transition parameter. A parallel results can be
derived from the final set of simulations, shown in Figure 7, where again the emission path
derived from a simulation with a cap on the maximum increase of atmospheric temperature,

13
when considering an ‘optimistic’ value of the climate sensitivity parameter, starts as early as
at the second commitment period to diverge dramatically from the business as usual scenario.

14
References

Caldeira, K. Jain, A. T., Hoffert, M.I. (2003), “Climate Sensitivity Uncertainty and the Need
for Energy Without CO2 Emission”, Science, March, 299.
Gritsevskyi, A., Nakicenovic, N. (2000), “Modeling uncertainty of induced technological
change”, Energy Policy, 28 907-921.
Goulder, L.H. and K. Mathai (2000), “Optimal CO2 Abatement in the Presence of Induced
Technological Change”, Journal of Environmental Economics and Management, 39, 1-
38.
Joos, F. et al. (1999), “Correcting the carbon cycle representation: how important is it for the
economics of climate change?”, Environmental Modeling and Assessment, 4, 133-140.
Joos, F. et al. (2001), “Global warming feedbacks on terrestrial carbon up take under the
Intergovernmental Panel on Climate change (IPCC) emission scenarios”, Global
Biogeochemical Cycles, 15-4, 891-907.
IPCC (2001), Climate Change 2001: The Scientific Basis, Cubasch et al., Contribution of
Working Group I the Third Assessment Report of the Intergovernmental Panel on
Climate Change. Houghton et al. (eds.), Cambridge University Press, Cambridge, UK.
Manne, A. S. and Richels, R.G. (2003), “Stabilizing Long-Term Temperature”, Working
Paper.
Nordhaus, W.D. and Z. Yang (1996), “A Regional Dynamic General-Equilibrium Model of
Alternative Climate-Change Strategies”, American Economic Review, 4, 741-765.
Nordhaus, W. D. and Popp, D. (1997), “What is the value of Scientific Knowledge? An
application to global warming using the PRICE Model”, The Energy Journal, 18-1, 1-
44.
Nordhaus, W. D. and Boyer ,J. (2000) Warming the World: Economic Models of Global
Warming, The MIT Press.
Pershing, J. and Tudela, F. (2003 ), “A long-term target: framing the climate effort”, Working
Draft, July Pew Center.
Schneider, S. H. (2001), “What is ‘Dangerous’ Climate Change?”’, Nature 411, 17.
Toth, F. and M. Mwandosya. (2001), “Decision-Making Frameworks,” in Climate Change
2001: Mitigation. Contribution of Working Group III to the Third Assessment Report of
the Intergovernmental Panel on Climate Change, B. Metz et al., eds. Cambridge:
Cambridge University Press.
Webster, M. D. (2002), “The Curious Role of ‘Learning’ in Climate Policy: Should We Wait
for More Data?”, Energy Journal 23 (2), 97-119.

15
Appendix: Model Equations

In this appendix we reproduce the remaining equations that make up the whole model. These
equations are reported here for the sake of completeness and are the same as the ones found in the
original RICE-99 model.

In each region, n, there is a social planner who maximizes the following utility function (n indexes the
world’s regions, t are 10-years time spans):

(A1) Wn = ∑ U [C n (t ), Ln (t )]R (t ) = ∑ Ln (t ){log[c n (t )]}R (t )

t t

where the pure time preference discount factor is given by:

t
R(t ) = ∏ [1 + ρ (v)]
−10
(A2)
v =0
and the pure rate of time preference ρ(v) is assumed to decline over time.
The maximization problem is subject to:
(A3) { }
Qn (t ) = Ω n (t ) An (t ) K n F (t )1−γ −α Ln (t ) γ CE n (t ) α − p nE (t )CE n (t )

C n (t )
(A4) c n (t ) =
Ln (t )

(A5) K n F (t + 1) = (1 − δ K ) K n F (t ) + I n (t + 1)

(A6) Qn (t ) = C n (t ) + I n (t )

(A7) E n (t ) = ς n (t )CE n (t )

(A8) p nE (t ) = q(t ) + markup nE

(A9) [ ]
M AT (t + 1) = ∑ E n (t ) + LU j (t ) + φ11 M AT (t ) + φ 21 M UP (t )
n

(A10) M UP (t + 1) = φ 22 M UP (t ) + φ12 M AT (t ) + φ32 M LO (t )

(A11) M LO (t + 1) = φ33 M LO (t ) + φ 23 M UP (t )

(A12) { [ PI
F (t ) = η log M AT (t ) / M AT ]
− log(2) + O(t ) }
(A13) T (t + 1) = T (t ) + σ 1 {F (t + 1) − λT (t ) − σ 2 [T (t ) − TLO (t )]}

1
(A14) Ω n (t ) =
(
1 + θ1,nT (t ) + θ 2,nT (t ) 2 )

List of variables:
W = welfare
U = instantaneous utility

16
C = consumption
c = per-capita consumption
L = population
R = discount factor
Q = production
Ω = damage
A = productivity or technology index
KF = capital stock
CE = carbon energy
pE = cost of carbon energy
I = fixed investment
E = carbon emissions
MAT = atmospheric CO2 concentrations
LU = land-use carbon emissions
MUP = upper oceans/biosphere CO2 concentrations
MLO = lower oceans CO2 concentrations
F = radiative forcing
T = temperature level
q = costs of extraction of industrial emissions

List of parameters:
α, γ = parameters of production function
δK = rate of depreciation of capital stock
ζ = exogenous technical change effect of energy on CO2 emissions (carbon intensity)
φ11, φ12, φ21, φ22, φ23, φ32, φ33 = parameters of the carbon transition matrix
η = increase in radiative forcing due to doubling of CO2 concentrations from pre-industrial levels
σ1, σ2 = temperature dynamics parameters
λ = climate sensitivity parameter
markupE = regional energy services markup
θ1, θ2 = parameters of the damage function
PI
M AT = pre-industrial atmospheric CO2 concentrations
O = increase in radiative forcing over pre-industrial levels due to exogenous anthropogenic causes
ρ = discount rate
TLO = lower ocean temperature

17
Tables and Figure

Table 1: A summary of the four short term policy scenarios

2010 2020 from 2020 onwards

Scenario 1 “Business-as-usual”

Annex B-US business-as-usual business-as-usual business-as-usual

US business-as-usual business-as-usual business-as-usual

Developing countries business-as-usual business-as-usual business-as-usual

Scenario 2 “Kyoto Forever without US”

Annex B-US Kyoto target: -5.2% 2010 level 2010 level

US -18% intensity target business-as-usual business-as-usual
Developing countries business-as-usual business-as-usual business-as-usual
Scenario 3 “2020 Global Target”

Annex B-US Kyoto target: -5.2% 2010 level 2010 level

US -18% intensity target -10% 2020 level

Developing countries business-as-usual business-as-usual -10%

Scenario 4 “2020 Global Cap on Emissions”

Annex B-US business-as-usual business-as-usual

US business-as-usual business-as-usual 2000 emissions

Developing countries business-as-usual business-as-usual

Table 2: A summary of the long-term target scenarios

Constraint Level
500
550

Aggregate Atmospheric Concentrations of Carbon in ppm 600

650
750

3°

Atmospheric Temperature (deg C above pre-industrial) 2,5°

2,3°

18
 Table 3. Emissions Peak Periods for Different Scenarios.
USA OHI EUROPE EE MI LMI CHINA LI World

Scenario 1 Above 2015 2015 Above Above Above Above Above Above
2105 2105 2105 2105 2105 2105 2105
Scenario 2 Above 2005 2005 2085 Above Above Above Above Above
2105 2105 2105 2105 2105 2105
Scenario 3 Above 2005 2005 2085 2035 2045 2075 2065 2105
2105
Scenario 4 2025 2015 2015 2075 2025 2025 2025 2025 2025
550 2025 2005 2015 2005 2085 2065 2065 2075 2065
2.5° 2015 2005 2005 2005 2065 2045 2025 2015 2015

19
Figure 2. Comparison of Emissions Paths in a BaU Scenario and a Scenario with a Cap
on Concentration of 550 ppm, Annex 1 Countries.
2

1,8

1,6

1,4

1,2
Gton C

0,8

0,6

0,4

0,2

0
1995 2005 2015 2025 2035 2045 2055 2065 2075 2085 2095 2105

USA-550 ppm OHI-550 ppm EUROPE-550 ppm EE-550 ppm USA-BaU OHI-BaU EUROPE-BaU EE-BaU

Figure 3. Comparison of Emissions Paths in a BaU Scenario and a Scenario with a Cap
on Concentration of 550 ppm, Rest of the World.
4,5

3,5

2,5
Gton C

1,5

0,5

0
1995 2005 2015 2025 2035 2045 2055 2065 2075 2085 2095 2105

MI-550 ppm LMI-550 ppm CHINA-550 ppm LI-550 ppm MI-BaU LMI-BaU CHINA-BaU LI-BaU

20
Figure 4. World Emission Paths. Comparison between Different Policy and Stabilization
of CO2 Concentration Scenarios.
16,5

14,5

12,5

10,5
Gton C

8,5

6,5

4,5

2,5
1995 2005 2015 2025 2035 2045 2055 2065 2075 2085 2095 2105

750 ppm 650 ppm 600 ppm

550 ppm 500 ppm BaU
Kyoto-US Intensity Target 2020 Global Target CAP on 2030 Emissions

9,5

8,5

8
Gton C

7,5

6,5

5,5
1995 2005 2015 2025 2035

750 ppm 650 ppm 600 ppm

550 ppm 500 ppm BaU
Kyoto-US Intensity Target 2020 Global Target CAP on 2030 Emissions

21
Figure 5. World Emission Paths. Comparison between Different Policy and Stabilization
of Temperature Scenarios.

12,5

10,5
Gton C

8,5

6,5

4,5

2,5
1995 2005 2015 2025 2035 2045 2055 2065 2075 2085 2095 2105

3° 2,5° 2,4°
CAP on 2030 emissions BaU Kyoto-US Intensity Target
2020 Global Target

9,5

8,5

8
Gton C

7,5

6,5

5,5
1995 2005 2015 2025 2035

3° 2,5° 2,4°
CAP on 2030 emissions BaU Kyoto-US Intensity Target
2020 Global Target

22
Figure 6. World Emission Paths. Uncertainty on carbon transition parameter.
16

14

12

10
Gton C

0
1995 2005 2015 2025 2035

550 'Pessimistic' carbon transition parameter value CAP 550 ppm 'Most probable' carbon transition parameter value
550 'Optimistic' carbon transition parameter value CAP on 2030 emissions
BaU Kyoto-US Intensity Target
2020 Global Target

10

8
Gton C

4
1995 2005 2015 2025 2035

550 'Pessimistic' carbon transition parameter value CAP 550 ppm 'Most probable' carbon transition parameter value
550 'Optimistic' carbon transition parameter value CAP on 2030 emissions
BaU Kyoto-US Intensity Target
2020 Global Target

23
Figure 7. World Emission Paths. Uncertainty on climate sensitivity parameter

12,5

10,5
Gton C

8,5

6,5

4,5

2,5
1995 2005 2015 2025 2035 2045 2055 2065 2075 2085 2095 2105

2.5°C 'Optimistic' climate sensitivity parameter 2.5°C 'Most Probable' climate sensitivity parameter
2.5°C 'Pessimistic' climate sensitivity parameter CAP on 2030 emissions
BaU Kyoto-US Intensity Target
2020 Global Target

10

8
Gton C

4
1995 2005 2015 2025 2035

2.5°C 'Optimistic' climate sensitivity parameter 2.5°C 'Most Probable' climate sensitivity parameter
2.5°C 'Pessimistic' climate sensitivity parameter CAP on 2030 emissions
BaU Kyoto-US Intensity Target
2020 Global Target

24