Energy Economics 47 (2015) 129–141

Contents lists available at ScienceDirect

Energy Economics
journal homepage: www.elsevier.com/locate/eneco

Costs of certified emission reductions under the Clean Development

Mechanism of the Kyoto Protocol☆
Shaikh M. Rahman a,⁎, Grant A. Kirkman b,1
a
Agricultural and Applied Economics, Texas Tech University, Box 42132, Lubbock, TX 79409, USA
b
Data Analysis and Intelligence, Sustainable Development Mechanisms, United Nations Climate Change Secretariat, Martin-Luther-King Strasse 8, 53175 Bonn, Germany

a r t i c l e i n f o a b s t r a c t

Article history: This paper examines the cost structure of certified emission reductions (CERs) through various types of projects
Received 18 October 2013 under the Clean Development Mechanism (CDM) of the Kyoto Protocol. Using the CDM project data, the costs of
Received in revised form 24 June 2014 CERs and their variation across technology and over time and space are estimated by applying alternative func-
Accepted 25 October 2014
tional forms and specifications. Results show that the average cost of CERs decreases with the project scale and
Available online 3 November 2014
duration, scale and duration effects significantly vary across project types, and there is an upward trend in
JEL classification:
costs. The results also show that the distribution of the projects in the CDM portfolio or a given location does
F53 not strictly follow the relative cost structure, nor does the distribution of the CDM projects in different host coun-
Q54 tries follow the principle of comparative advantage. More than 84% of the CDM portfolio consists of various en-
Q56 ergy projects with substantially higher costs of CERs than afforestation and reforestation, industrial and landfill
Q58 gas reduction, and methane avoidance projects, which are only 12% of all projects. While per unit cost of abate-
ment plays an important role in the bottom-up and top-down models to evaluate emission reduction potential
Keywords:
and analyze policy alternatives, the findings contradict the presumption of such models that project investors
Kyoto Protocol
seek out low-cost opportunities. At the aggregate level, the cost of CER by the projects in Asia and Europe is sim-
Clean development mechanism
Certified emission reductions
ilar but higher than those hosted in Africa, Americas, and Oceania. Yet more than 83% of the projects in the CDM
Levelized cost of CER portfolio are located in Asia; more than 69% of the projects are in China and India alone. China appears to have a
Economies of scale comparative advantage (i.e., lowest opportunity cost) in energy efficiency projects, while India has a comparative
Comparative advantage advantage in hydro power projects and Brazil has a comparative advantage in wind power projects. In contrast,
energy efficiency category accounts for only 8% of the CDM projects in China, hydro power accounts for 12% of the
projects in India, and wind power accounts for 18% of the projects in Brazil. The results provide a basis for eval-
uating the incentives that the mechanism offers as a cost effective policy instrument that balances greenhouse
gas mitigation across sectors and regions, while fulfilling the objective of the convention.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction invest in sustainable development projects that reduce emissions in de-

The Clean Development Mechanism (CDM) is a provision of the
Kyoto Protocol of the United Nations Framework Convention on Climate
Change (UNFCCC) that allows European Community and other coun-
tries listed in Annex I of the Protocol to help meet their binding targets
of curbing anthropogenic greenhouse gas (GHG) emissions by reducing
emissions in developing countries.2 The CDM provides an incentive to
☆ This article is an outcome of an unfunded research project conducted by the authors.
All rights reserved to the authors.
⁎ Corresponding author. Tel.: +1 806 834 0505; +1 806 742 1099.
E-mail addresses: shaikh.m.rahman@ttu.edu (S.M. Rahman), gkirkman@unfccc.int
(G.A. Kirkman).
1 4
Tel.: +49 228 815 1363; fax: +49 228 815 1999.
2
Annex I countries are committed to meet their targets for limiting or reducing GHG
emissions primarily through domestic measures. These targets are expressed as levels of
allowed emissions, or “assigned amounts,” over the 2008–2012 commitment period. The
allowed emissions are divided into “assigned amount units” (AAUs).
veloping countries, presumably at costs lower than that of domestic
measures in Annex I countries.3 For measurable and verifiable emission
reductions that are additional to what would have occurred without the
CDM project or program of activities (PoA), the project earns certified
emission reductions (CERs), each equivalent to 1 ton of CO2 equivalent
(tCO2e hereafter) abatement. The project owners can either use the
CERs to meet their obligation or sell the CERs to an Annex I Party that
can use those to meet part of its emission reduction target under the
Kyoto Protocol.4
3
The Joint Implementation (JI) is another project-based mechanism that enables the
Annex I countries to carry out bilateral or multilateral emission reduction projects among
themselves.
The Emissions Trading (ET) is the third flexibility provision that allows Annex I coun-
tries to trade assigned amount units (AAUs) as well as credits generated by the project-
based mechanisms among themselves. As set out in Article 17 of the Kyoto Protocol,
Annex I countries with fewer emissions than permitted are also allowed to sell the excess
AAUs to the countries with more emissions than permitted.

http://dx.doi.org/10.1016/j.eneco.2014.10.020
0140-9883/© 2014 Elsevier B.V. All rights reserved.
130 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141
Both industrialized and developing countries have responded to the
incentives provided through the CDM. Starting in 2004, the number of
CDM project activities as well as the investments in such projects has in-
creased exponentially. As of 31 January 2014, 8744 projects are at some
stage of the CDM project cycle.5 If those projects operate at their full
potential, those will reduce 1.1 billion tCO2e emissions every year
thereby generating an equivalent number of CERs.6 While the rapid ex-
pansion of the CDM indicates that on the whole the mechanism aligns
the incentives of the Annex I and non-Annex I country participants,
the distributions of the projects by type (technology) and locations
are skewed. Most of the CDM project activities are renewable resource
based and are principally located in China and India. Sectors with
large mitigation potentials in some other countries, such as agriculture,
construction and transport, account for a smaller share of the CDM pro-
ject activities, and least developed countries with large sequestration
potentials have fewer forestry projects (Rahman et al., 2012). Thus,
the broad expectation that CDM project activities would be relatively
evenly distributed across all developing countries is challenged, raising
the question whether the incentive mechanism for undertaking a CDM
project activity is consistent with the cost structure.
Most numerical analyses of how the CDM affects the cost of meeting
the objectives of the Kyoto Protocol are based on specified abatement
cost curves and the assumption that capital will seek out least-cost
projects.7 The same approach also leads to the prediction of the sectors
and regions likely to benefit from project investment flows. However,
the cost of CERs for a particular project type of a given size can differ
across countries due to taxation and trade barriers and even within
countries due to factors such as emission factors, costs of connecting
to the electricity grids, and the cost of site access. This paper examines
the cost of CERs by various types and sizes of CDM projects located in
different developing countries, with the objective of assessing the
cost-effectiveness of GHG reductions through the CDM to provide poli-
cy relevant perspectives.
A majority of previous studies providing useful estimates of abate-
ment costs of various pollutants are based on secondary data or approx-
imated coefficients in the abatement functions. Rahman et al.
(forthcoming) estimated GHG emission abatement costs under the
CDM using imputed project cost data. Castro (2014) collected plant-
level cost data for 109 CDM projects and calculated abatement costs,
adjusting for baseline costs.8 However, actual abatement by the projects
in operation, incurred CER issuance costs, and the difference between
the crediting period and project life are not considered in these studies.
In this paper we use project-level production and cost data for a large
sample of CDM project activities, as reported in the project design
documents (PDDs). Apart from expected CERs, byproduct outputs,
and fixed and variable costs of the projects, the data distinguishes
among various types of projects, methodologies for calculating emission
reductions, the countries hosting the projects, and sequence of new
project investments for the period 2003–2013. For calculating the cost
of CERs for each project, we adjust the expected CERs and byproduct
outputs by actual credit issuance rate, consider effective lengths of
crediting period and project life, and take account of CER issuance
costs. As our objective is to provide accurate estimates of CER costs for
5
These projects are either registered (with or without issued CERs) or in validation. See
the UNEP Risoe CDM/JI Pipeline Analysis and Database (http://cdmpipeline.org/
publications/CDMPipeline.xlsx-January2014) and Larson et al. (2008) for a discussion of
CDM implementation rules and the CDM project cycle.
6 9
This number represents the potential of projects known to be issuing as adjusted by
their rate of issuance up to January 2013 (b= 1).
7
Metz et al. (2007) provide a careful discussion of abatement cost curves in top-down
and bottom-up models of mitigation costs and how the models are used to inform policy.
8
The baseline is generally conceived as the situation without the CDM project, which
may be new investment or status quo — continuation of the current situation without a
new investment (Castro, 2014). For many energy projects, baseline is the status quo with
expenses such as buying energy from the grid or buying coal. Avoiding or reducing such
expenses is considered as revenue for the CDM project and is included in the cost calcula-
tion (Castro, 2014).
different CDM technologies, we do not consider the baseline costs for
the projects, because baseline costs are relevant for mitigation cost
estimation.
The unique features of the CDM project activity-level data allow us
to draw distinctions among projects across types (technologies),
methodologies, locations, and time, and test two hypotheses important
for policy design: (1) whether CDM projects exhibit economies of scale
in certified emission reduction, and (2) whether the average cost of CER
by the CDM projects has decreased over time, presumably due to accu-
mulated experience. Based on the estimated costs, we further examine
the role of CER costs in explaining the observed CDM investments.9 In
particular, we examine whether the distribution of projects follows a
type-specific cost structure, such that there are more projects of the
types for which the average cost of CER is relatively lower. We also ex-
amine whether the distribution of the projects in different host coun-
tries follows the principle of comparative advantage (i.e. whether
countries hosting specific types of projects generate CERs at a lower
average cost over others).10
The remainder of the paper is organized in the following way. The
next section describes the conceptual model and empirical framework
for estimating the cost of CERs by the CDM projects. Section three
describes the CDM project data. Section four delineates the estimation
results and discusses the implications. Finally, the last section concludes
the paper.
2. Estimating the cost of certified emission reduction
In the case of CDM projects, the costs of CERs are not necessarily the
same as total project costs. Many CDM projects generate byproducts of
CERs. For example, renewable resource based projects generate elec-
tricity, energy efficiency projects saves electricity, while afforestation
and reforestation projects produce forest products. Total project costs
are larger than CER costs for the multi-output CDM projects, and the
cost of CERs may not be separable from the cost of byproducts. The
cost of CERs is equivalent to the total project cost for the projects that
generate CERs only.
Previous studies estimating pollution abatement costs can be cate-
gorized into two groups. The first group of studies considered pollution
abatement as an inseparable multi-output process, and suggested that
the cost of abatement might not be separable from the cost of produc-
tion of the primary output (see Considine and Larson, 2009, 2006;
Pizer and Kopp, 2005; Maradan and Vassiliev, 2005; Newell et al.,
2003; Newell and Stavins, 2003; Boyd et al., 1996; Nordhaus, 1994).
The second group of studies estimate the pollution abatement cost
function by separating cost of abatement from the cost of production
(see Castro, 2014; Rahman et al., forthcoming; Hamaide and Boland,
2000; Bystrom, 1998; Hartman et al., 1994).
In this paper, we estimate a separable cost function for CERs.11 Our
dataset provides total fixed and variable costs for the projects; it does
not distinguish between CER costs and total costs for the multi-output
projects. Developing a conceptual model for separable cost function,
we calculate the net cost of CERs by subtracting expected value
(i.e., revenues from the sales) of the byproducts from the fixed and var-
iable costs of the project. Due to a paucity of data on various transaction
costs associated with project development, approval, and implementa-
tion, we are unable to take account of such costs.12 Such costs, however,
See Rahman et al. (2012) for a review of ex ante predictions for the CDM.
10
For the definition and explanation of comparative advantage, please see Boudreaux
(2008).
11
Rahman et al. (forthcoming) estimate both the separable emission abatement cost as
well as project costs (when abatement cost is not separable) for the CDM using imputed
cost data. They found similar results for both models upon controlling for the byproduct
in the joint cost estimation.
12
Michaelowa and Jotzo (2005) outline various transaction costs associated with the
CDM projects. In addition, projects may incur site access costs, byproduct marketing costs
(e.g., connecting electricity to the grid), etc. We do not consider PoAs nor their costs.
 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141 131

are sunk in nature and account for a very small share of the fixed costs of For comparing the costs of CER for different technologies or systems,
the projects. Moreover, the costs of preparation of project design docu- it is convenient to calculate the average cost as CR0/∑Tt = 0Rte−rt. This is
ments (PDDs), submissions of the documents to the CDM executive the so called “levelized cost” of CER (LCOR), which is equivalent to
board and other government agencies, and processing the validation levelized cost of energy (LCOE) as referred in energy cost literature
and registration of the projects are typically included in the initial (see USEIA, 2014).
investments. In the above model, the length of the crediting period and project life
are assumed to be the same for simplicity. However, the lifetime of a
2.1. The conceptual model CDM project may be longer than the emission reduction crediting
period as it is the case for most of the electricity generating projects.
As a starting point for the derivation of our applied model, consider In that case, the LCOR can be calculated as:
the expected project value function, where the value of the investment,
Xτ −rt
XT  E

−rt
V0, is determined by the discounted value of two streams of profits from I0 þ cRe þ m −p E e
t¼0 t t t¼0 t t t
a project initiated in year 0 and expected to last T years13: LCOR ¼ Xτ −rt ; ð7Þ
Re
t¼0 t
XT R −rt
XT E −rt
V0 ¼ π e
t¼0 t
þ π e ;
t¼0 t
t ¼ 1; 2; …; n; ð1Þ
where τ is the length of emission reduction crediting period in years,
where the superscripts R and E distinguish between the expected profits and T is the project life in years. The LCOR for other byproducts can be
from CERs and electricity (as a byproduct of CERs). Eq. (1) can be ex- calculated in the same fashion.
panded to distinguish revenues from costs:
2.2. The empirical framework
XT R −rt
XT E −rt
XT R −rt
XT E −rt j
π e
t¼0 t
þ π e
t¼0 t
¼ p Re
t¼0 t t
þ p Ee
t¼0 t t
−C 0 ; ð2Þ
Assuming separable costs and fixed input prices in a given location
where, pR and pE are the prices of CERs and electricity, and Cj0 is the net and time, the basic expressions for the log–log and log–quadratic
present value of total cost of the project. functional forms of annualized cost of CERs for project j can be given
For the moment, we treat the cost of producing both CERs and elec- by14:
tricity as joint, and denote the total discounted costs as: R
! !
Cj Rj
j
XT −rt
XT j −rt
ln ¼ α þ β ln þ θq j þ e j and ð8Þ
C 0 ¼ I0 þ mt e þ cRe ; ð3Þ Tj Tj
t¼0 t¼0 t t

where I0 is the initial investment, m is the annual variable (operation

R
! ! " !#2
and maintenance) cost of the project and c is the per unit cost of issuing Cj Rj Rj
CERs that may vary with project scale and over time. When the rate of ln ¼ α þ β ln þ γ ln þ θq j þ ε j ; ð9Þ
Tj Tj Tj
expected profit clears the investment hurdle, positive investments are
observed and the associated value function can be characterized as:
where C is the net present value of the total cost of CERs (as represented
  R
XT −rt E
XT −rt j by Eq. (6)), R is the total volume of CERs, T is the duration of the project
V I0 ¼ p0 Re
t¼0 t
þ p0 t¼0 Et e −C ðR0 ; E0 Þ ð4Þ
in years, and q is a vector of control variables (e.g., project types,
location, and methodologies for calculating emission reduction).15
where the pR0 and the pE0 represent the weighted average prices for CERs Eq. (8) is nested in Eq. (9) and the two are indistinguishable when γ
and electricity at the time the investment decision is made. The aggre- is indistinguishable from zero.
gate output levels consistent with solution values can be recovered The vector of input prices, w, associated with the cost functions is not
∂I
 T −rt ∂I

via the envelope theorem, as ∂pR0
¼ ∑t¼0 Rt e ≡ R0 , and ∂pE0
¼ always observed. However, for a given time period and a given location,
T −rt the prices of the inputs are likely the same. The specific type of technol-
∑t¼0 Et e ≡ E0, where R0 and E0 are the volumes of CERs and electric-
ogy employed by each project is laid out in the Project Design Document
ity that the project is expected to produce over its lifetime, weighted by
(PDD) and therefore implicitly in the project classifications. The
the discount factor used in the evaluation of the investment function.
methodology used to calculate CERs reflects the scale of the projects.
The associated joint cost function can be expressed as Cjt(wt, Rt, Et; St),
Thus, dummy variables for different project types, methodologies,
where wτ is the vector of expected input prices and Sτ is the set of
locations, and time periods can be used as a proxy for the missing input
state variables, in addition to input prices, that condition the optimiza-
price vector. More formally, let xm be the vector of inputs associated
tion problem.
with a specific CDM methodology. The associated vector of prices can
The problem can be simplified when costs are not joint, that is, when
vary by time (t) and place (l), suggesting the notation wx,l,t. Because we
C j = C R(R0) + C E(E0). In this case, Eq. (4) can be restated as:
lack specific information about this vector, we use a triplet of dummies
  R
XT −rt
XT E −rt R E (x, l, t) to proxy the missing input prices. This solves the missing price
V I0 ¼ p0 Re
t¼0 t
þ p E e − C ðR0 Þ−C ðE0 Þ:
t¼0 t t
ð5Þ
information problem, but the associated parameters on the indicator
variables likely pick up other attributes of time and place. This is a
The optimal level of CERs can be recovered via the envelope theorem
as before. However, in this case we needn't keep track of the optimal
level of electricity generated by the project, since our objective is to es- 14
Using imputed cost data, Rahman et al. (forthcoming) estimated both the non-joint
timate the portion of the non-joint cost function attributable to CERs, C-t (separable) emission abatement cost function given by Eq. (5) as well as the joint (project)
R cost function given by Eq. (3). They found similar results for both models upon controlling
(wt, Rt; St). Thus the net present value of total cost of CERs by the
for the byproduct in the joint cost estimation. As it turns out that in the case of CDM there
project can be calculated as:
are few practical consequences from choosing one version of the applied model over the
R
XT E −rt
XT −rt
XT −rt
other, we focus on the non-joint (separable) cost function for CERs given by Eq. (5).
C 0 ¼ I0 − t¼0 pt Et e þ t¼0
mt e þ cRe :
t¼0 t t
ð6Þ 15
In the current setting, we suppress the time subscript assuming that the equilibrium
level of abatement would be the same in each year. This restriction is consistent with
the flow of CERs as described in most PDDs. However, the CDM projects may have a sub-
13
The model is similar to the one developed by Rahman et al. (forthcoming). See stantial variation in abatement over time. We expect to relax this assumption in future
Timilsina and Lefevre (1999) for a related discussion. work when data on issued CERs and actual investments are available.
132 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141
mixed blessing, since the net effect is to round up otherwise unobserved
effects; however it confounds the interpretation of the estimated fixed
effects.
Separate from differences in costs and expectations explained by
start-dates of the projects, the duration of the projects may also matter.
According to Booth (1991), projects can exhibit ‘economies of time’,
such that projects with longer duration are associated with lower per
unit costs. On the other hand, CDM projects with longer duration are
likely riskier than the projects of the same size and type with a shorter
duration. To take account of such time relationships, the net present
value of total abatement cost is annualized by dividing the duration of
the project (in years). This allows for a comparison of the results with
those of other studies.
3. Data description
Analytical data maintained by the UNFCCC secretariat comprising in-
formation on the CDM project status, expected project lifetime, installed
power generation capacity, electricity output in each year over the life of
the project, expected CERs over the crediting period, CER issuance,
methodology used for calculating emission reductions, initial capital in-
vestments, annual operation and maintenance costs, and time bound
variables (e.g. project design, start, submission, registration, monitoring
and issuance dates) provides the basis for this analysis.16 The database is
generated from the individual project design documents which consid-
ered system boundaries.
As of December 2013, 7429 projects are registered under the CDM
and 1315 projects are in the process of registration.17 These projects
are (to be) hosted in 103 developing countries in 5 UN regions (Africa,
Americas, Asia, Europe, and Oceania). Approximately 83% of the projects
are (to be) hosted in Asia, 13% in the Americas, 3% in Africa, and the rest
in Europe and Oceania. About 34% of the registered projects have also is-
sued CERs. We use 7659 of these projects for which required data for
this analysis are available. The distribution of the projects in the sample
is similar to that of the CDM portfolio.
Information about inputs used in the projects are not available, but
the type of technology or mitigation action employed to replace the
baseline (typically used) technology is implicitly laid out in the project
classification. There are 27 specific types of projects as categorized
by the United Nations Environment Programme (UNEP) Risoe Cen-
ter. We consolidate the projects into 13 major types: (1) wind
power (2) hydro power; (3)biomass energy; (4) energy efficiency;
(5) CH4 reduction/avoidance; (6) solar power; (7) landfill gas
reduction; (8) industrial gas (HFCs, PFCs & SF6 , N2 O) reduction;
(9) fossil fuel switch, (10) cement, fugitive and CO2 usage (cement
& fugitive hereafter); (11) afforestation & reforestation; (12) tidal,
geothermal, and mixed renewables (tidal & geothermal hereafter);
and (13) transportation.
The CDM Board has approved various methodologies to calculate
emission reductions by these projects. While a specific methodology
can be used to evaluate emission reduction by different types of pro-
jects, approved methodologies can be categorized into two major
groups: (1) large scale methodologies; and (2) small scale methodolo-
gies. Note that ‘scale’ is a measure of project size that has changed
over time; small scale methodologies are limited to small projects and
are much cheaper and quicker to implement.
The crediting period of the projects ranges from 5 to 30 years, while
the project life ranges from 7 to 100 years. Some projects report a
16
These data — investment, costs, CERs and byproduct generation are all sourced from the
PDDs of the CDM projects. While these are planning documents, the data contained therein
are independently checked by auditors as part of the CDM validation process and thereafter
by the CDM Executive Board as part of the registration process. In most cases it is not known
how closely the actual values match the planning data from the PDD. The exception is CERs
where rates of issuance to date are available and average over 70%.
17
For further information on the status of projects, issuance of CERs, approved method-
ologies on 31 December 2013 see the CDM Insights portal bhttp://cdm.unfccc.intN.
lifetime equal to the crediting period, while others typically have a
much longer operational lifetime (e.g., hydro power). A majority of
the projects have a crediting period of either 10 years or 7 years
that is renewable up to 21 years. Note that some projects in the sam-
ple abstained from renewing the crediting period despite having the
opportunity. We consider the up-to-date crediting period for each
project.
For the purpose of CER cost estimation, we calculate expected CERs
over the crediting period and expected byproduct (e.g., electricity) out-
put over the project life. Expected CERs are adjusted by multiplying the
median CER issuance rate (i.e., issued CERs divided by expected CERs)
for each UNEP sub-category in each UN sub-region. The average issu-
ance rate of the projects with issued CERs is 0.79, while the issuance
rate ranges from 0.02 to 1.50. For the projects without issued CERs,
the issuance rates are interpolated with the medians for respective pro-
ject category and geographical region.
Table 1 provides the summary statistics on the number and expected
annual CERs by the CDM projects in the sample categorized by major pro-
ject types and methodology classes. The project categories are listed in the
table according to the descending order of their shares in the sample and
CDM portfolio. About 69.6% of these projects are renewable resource
based power projects (e.g., wind, hydro, solar, and tidal & geothermal)
that account for 53.2% of the annual expected CERs from all projects in
the sample (assuming all of these projects issue CERs at the expected
rate for specific category). While only 2% of the projects are in industrial
gas (HFCS, PFCs & SF6, and N2O) reduction category, those account
for about 12.8% of the total annual CERs. Energy efficiency, fossil
fuel switch, methane reduction/avoidance, biomass energy, and landfill
gas categories account for approximately 9.7%, 6.2%, 4.8%, 4.3%, and 3.8%
of total annual CERs, respectively. While solar power is the fifth
largest category by project numbers, those projects are expected to
generate 1.1% of total annual CERs. Afforestation & reforestation and
transportation projects together account for less than 1.0% of the total
annual CERs.
As presented in Table 1, the mean, median, minimum, and maxi-
mum annual CERs show that project size varies widely within as well
as across types. The industrial gas reduction projects are the largest
in terms of average annual expected CERs, followed in a descending
order by fossil fuel switch, tidal & geothermal, cement & fugitive,
energy efficiency, hydro power, forestation, landfill gas, wind
power, methane reduction/avoidance, transportation, biomass
energy, and solar power. Table 1 also shows that large scale
methodologies are applied to 63.3% of the projects in the sample,
which account for 93.6% of the total expected CERs. Small scale
methodologies are applied to 36.7% of the projects that account for
only 6.4% of potential CERs.
Most of the projects in the sample generate some kind of byproduct of
CERs, thus generating additional revenue from the sale of byproducts.
About 94% of the projects in the sample generate additional electricity
output (in addition to the capacity of the baseline). Expected electricity
outputs on each year over the project life are calculated by multiplying
installed megawatt capacity by expected annual hours of operations as
obtained from the PDDs. Assuming that the outputs of CERs and
byproducts maintain the same proportion over the scale of operation,
expected output of the byproducts are also adjusted by the rate of CER
issuance.
Table 2 shows a summary statistics of the number and expected
annual electricity generation by major project type and methodology
classes. About 34.6% of the total electricity is expected to be generat-
ed by the hydro power projects, followed by energy efficiency
(27.2%), wind power (18.5%), fossil fuel switch (12.5%), biomass
energy (3.6%), tidal and geothermal (1.6%), solar power (1.1%),
methane reduction/avoidance (0.6%), landfill gas (0.4%), and cement
& fugitive (0.1%) projects. In terms of annual expected electricity
output, project size varies widely within as well as across types.
Note that, for different types of projects, relative sizes by average
 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141 133

Table 1
Number, percentage, and expected annual CERs by the CDM projects in the sample.
Source: UNFCCC Project Activity Database.

Expected annual CERs (1000)

CDM projects Per project Category

No. % Total Mean Med. Min. Max. Total

Project category
Wind power 2618 34.2 91.6 88.7 1.1 1051.1 239,754.6
Hydro power 2272 29.7 148.3 46.6 0.5 7325.1 336,840.1
Biomass energy 680 8.9 71.7 47.4 2.1 484.2 48,776.2
Energy efficiency 488 6.4 224.4 84.8 0.2 4148.7 109,504.1
CH4 reduction/avoidance 458 6.0 118.3 41.3 1.4 3016.7 54,163.1
Solar power 393 5.1 31.7 17.1 0.6 361.8 12,448.9
Landfill gas 270 3.5 157.4 98.3 6.2 1717.5 42,498.8
HFCs, PFCs & SF6, and N2O 149 1.9 969.9 228.3 7.9 10,437.3 144,507.8
Fossil fuel switch 94 1.2 746.8 590.4 2.3 4017.2 70,203.7
Cement & fugitive 88 1.1 572.5 210.3 6.6 13,710.0 50,380.6
Afforestation & reforestation 67 0.9 84.7 23.6 0.6 1930.9 5673.0
Tidal & geothermal 46 0.6 288.3 156.7 0.3 1105.7 13,260.9
Transportation 36 0.5 139.3 104.5 0.2 556.9 5016.6

Methodology
Large scale (AM/ACM) 4848 63.3 218.8 100.7 2.2 13,710.0 1,060,856.0
Small scale (AMS) 2811 36.7 25.7 21.2 0.2 1717.5 72,172.2
Total 7659 100.0 147.9 65.0 0.2 13,710.0 1,133,028.0

electricity output do not correspond to the relative sizes by expected
annual CERs (see Tables 1 and 2).
The total cost of a CDM project includes initial outlay of capital,
operational and maintenance expenditure, and CER issuance costs. For
the projects that generate CERs only, the total cost of CERs is the same
as the total project cost. For the projects that generate byproducts of
CERs, the total cost of CERs is the total project cost minus revenue
from byproduct sales. Data on initial capital investments, annual opera-
tion and maintenance costs, and expected revenues from the sales of
byproducts for all projects in the sample are obtained from the UNFCCC
Project Activity Database. CER issuance cost includes the fee payable to
UNFCCC and expenses for the monitoring report and verification. The is-
suance fee is fixed at €0.15 per CER for projects issuing up to 20,000
CERs, and €0.20 per CER for issuance volumes greater than 20,000 CERs
(Ferdinand and Melum, 2013). The other two cost components are inde-
pendent of the number of CERs issued. Project owners have to finance one
monitoring report and one verification report. The costs of monitoring
and verification depend on project type, location, and deadlines.
Ferdinand and Melum (2013) provide a measure of per unit cost of issu-
ing CERs that declines at a decreasing rate with the volume of CERs issued
per monitoring period. Following their estimate and assuming one
monitoring period per year, we assign a per unit issuance cost of €1.20
if the CER volume is less than or equal to 40,000 CERs per monitoring
period, and calculate the per unit cost of issuance for volumes
larger than 40,000 CERs using a function, c = 1.1695((R − 40,000) /
10,000 + 1)−0.393, where c is the issuance cost per unit of CER and R
is the volume of CER. We convert the issuance cost into US dollar
using annual average conversion rate for the crediting start years.

Table 2
Number, percentage, and expected annual electricity generation by the CDM projects in the sample.
Source: UNFCCC Project Activity Database.

Exp. annual electricity (GWh)

CDM projects Per project Category

No. % Total Mean Med. Min. Max. Total

Project category
Wind power 2618 100.0 90.0 82.8 0.8 1358.1 235,698.1
Hydro power 2272 100.0 194.1 55.8 0.6 20,042.8 441,025.2
Biomass energy 671 98.7 68.9 48.2 0.5 812.9 46,252.7
Energy efficiency 466 95.5 743.0 74.1 3.4 15,611.7 346,233.2
CH4 reduction/avoidance 395 86.2 18.4 6.3 0.0 900.4 7283.1
Solar power 389 99.0 34.9 18.3 0.0 497.9 13,561.5
Landfill gas 253 93.7 18.8 9.6 0.0 184.0 4750.9
HFCs, PFCs & SF6, and N2O – – – – – – –
Fossil fuel switch 90 95.7 1768.0 1387.6 4.4 7758.5 159,117.9
Cement & fugitive 5 5.7 126.8 21.8 5.5 462.5 634.2
Afforestation & reforestation – – – – – – –
Tidal & geothermal 44 95.7 465.3 304.7 0.4 1804.8 20,474.2
Transportation – – – – – – –

Methodology
Large scale (AM/ACM) 4848 63.3 218.8 100.7 2.2 13,710.0 1,060,856.0
Small scale (AMS) 2811 36.7 25.7 21.2 0.2 1717.5 72,172.2
Total 7659 100.0 147.9 65.0 0.2 13,710.0 1,133,028.0
134 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141

Table 3
Calculated levelized cost of CER (LCOR) for different types of CDM projects.

CDM project category No. of proj. Levelized cost of CER (US $/CER)

Mean Median Min. Max.

Wind power 2618 58.39 58.39 −1077.44 9502.47

Hydro power 2272 51.45 34.04 −1012.36 3945.39
Biomass energy 680 122.11 62.73 −216.14 2938.85
Energy efficiency 488 189.92 21.00 −393.84 31,261.19
CH4 avoidance/reduction 458 13.43 13.14 −485.93 283.07
Solar power 393 597.10 270.40 −245.19 8568.14
Landfill gas 270 78.76 37.49 −10.77 798.40
HFCs, PFCs & SF6, and N2O 149 9.24 3.07 −118.06 191.31
Fossil fuel switch 94 247.39 150.47 −1050.76 2236.03
Cement & fugitive 88 332.29 79.24 −88.75 4481.86
Afforestation & reforestation 67 190.27 41.19 −95.55 4362.13
Tidal & geothermal 46 186.38 52.38 −82.84 2879.74
Transportation 36 18,461.79 1671.56 −7975.88 548,635.90
All projects 7659 188.98 47.38 −7975.88 548,635.90

As a project generates streams of CERs and byproducts in each
year of its duration, the present value of CER generation costs for each
project are calculated as described in Eq. (6). The annual variable costs
and revenues are discounted using borrowing interest rates for the
year of fixed capital investment (i.e., the prior year of the crediting
start period) as derived from the project documentation. The discount
rate is the rate used in the PDDs to demonstrate additionality; it is
typically expressed as a benchmark rate or hurdle rate. The interest
rates for the projects in the sample range from 2.4 to 18.0%, with a
mean of 10.9%.
For CER cost calculation, the stream of byproduct revenues
from power projects is obtained by multiplying electricity outputs
by electricity tariffs as reported in the PDDs. For the projects with
missing electricity tariff data, the median tariff in a particular coun-
try or region is used. The observations of the non-power projects
with missing revenue data are not used in the estimation. Finally,
as implied by duality, the CER cost function includes the sum of the
discounted-weighted flows of CERs and byproduct outputs. Interest
rates as mentioned above are used to discount those streams of
outputs.
4. Results and discussions
Before estimating the cost function, we calculate the levelized
cost of CER (LCOR) for each individual CDM project as specified in
Eq. (7). Table 3 shows the summary statistics of the calculated
LCORs for different types of CDM projects. Note that the total
amounts of CERs and CER issuance costs are calculated over the emis-
sion reduction crediting period for each project, while operation and
maintenance costs and byproduct revenues are calculated over the
life of the project.
As presented in Table 3, the mean and median of calculated LCORs
for all projects in the sample are US$ 188.98 and US$ 47.38 per CER,
respectively. Calculated LCORs widely vary across different CDM pro-
ject categories, and even within each category. Also, the ranges of
LCORs for some categories overlap. The median LCOR is the highest
for the transportation category, followed in a descending order by
solar power, fossil fuel switch, cement & fugitive, biomass energy,
wind power, tidal & geothermal, forestation, landfill gas, hydro
power, energy efficiency, methane reduction/avoidance, and indus-
trial gas reduction categories. This order, however, is different from
the order based on the mean LCORs for different categories. Note
that for some projects (about 9% of the sample), calculated LCORs
are less than zero. Negative LCORs imply that those projects are
profitable even without the CDM revenue. Projects with positive
LCORs may still be profitable as long as the CER price exceeds the
average cost.
These estimates of LCORs as presented in Table 3 are much higher
than the estimates in other studies (see Castro, 2014; Wetzelaer
et al., 2007; Sims, Rogner, and Gregory, 2003; Vattenfall, 2007).
Apart from the lengths of crediting period and project life, three
other factors have also contributed to the higher estimates.18 First,
the stream of CER issuance costs is accounted for in the total cost
calculation. Second, while calculating LCORs, the annual amount of
reductions is adjusted by the annual CER issuance rates for different
project types located in different countries, which are for most pro-
jects smaller than expected CERs as listed in the PDDs. Finally, as
LCOR is defined in Eq. (7), the adjusted annual CERs are further
discounted using the time value of money (i.e., borrowing interest
rates). While Castro (2014) does the same discounting, our esti-
mates are higher because of other factors as mentioned above. Note
that the levelized cost of abatement over the life of a CDM project
may be much smaller than LCOR, because the project may reduce
emissions beyond the approved crediting period even if CERs are
not issued.
Previous studies show that the average cost of abatement by
the CDM projects also differ across location and declines with
the project scale (see Rahman et al., forthcoming). To further
evaluate the CER cost structure of the CDM, using individual
project data we estimate the log–log and log–quadratic cost
functions in Eqs. (8) and (9) for the whole sample as well as for
the sample disaggregated into project categories and major host
countries. To make comparisons easier, we report calculated elastic-
ities or, in the case of the discrete regressors, semi-elasticities
(i.e., associated discrete percentage changes in abatement costs as
the value of the indicator variable switches from 0 to 1) in the resul-
tant set of tables.
4.1. Aggregate models
We begin with an ordinary least squares estimation of the log–log
model. In particular, the logarithm of total annualized cost of CERs is
regressed on the logarithm of the volume of annual CERs, and four
sets of indicator variables for major project types, methodologies
for calculating emission reduction, project start years, and broad
18
The initial capital investments and operation and maintenance costs as reported in the
PDDs may be biased upwards, due to the justification of additionality. In that case, actual
LCORs would be smaller than those reported in Table 3. However, given available informa-
tion, it was not possible to assess whether the costs reported in the PDDs were realistic or
not.
 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141 135

800
Table 4
Aggregate model results — elasticities and discrete-change effects. 2008
700

Average Project Size (1,000 CERs/year)

Log–log Log–quadratic
600
Annual certified emission reduction (CER) 0.629*** 0.627*** 2009
Indicator variables for project categories (hydro power is the base) 500
Wind power 0.287*** 0.271***
Biomass energy 0.686*** 0.703*** 400 2010
Energy efficiency 0.175* 0.144
CH4 avoidance/reduction −1.083*** −1.097*** 300
Solar power 1.723*** 1.672***
Landfill gas −0.181** −0.175** 2011
200
HFCs, PFCs & SF6, and N2O −1.887*** −2.039***
Fossil fuel switch 2.035*** 1.901*** 100
Cement & fugitive 1.138*** 1.087*** 2012
Afforestation & reforestation −0.531*** −0.671*** 0
Tidal & geothermal 0.802*** 0.667***
Transportation 3.843*** 3.809***
Indicator variables for methodology groups (small scale is the base)
Large scale 0.643*** 0.687***
Indicator variables for project start years (2005 and earlier is the base)
2006 0.033 0.019 Fig. 1. Average size (1,000 CERs/year) of the new projects in the CDM pipeline during
2007 0.02 0.025 2008-12 Source: www.cdmpipeline.org (accessed 31 January 2014).
2008 0.232*** 0.231***
2009 0.349*** 0.348***
2010 0.446*** 0.455***
2011 0.441*** 0.461***
2012 0.544*** 0.560***
2013 0.513*** 0.524*** et al. (2013) also find similar result. At the aggregate level, the aver-
Indicator variables for regions (Asia is the base) age size of all projects in the pipeline gradually decreased over time;
Africa 0.473*** 0.445*** from 132,850 CERs/year in 2008 to 112,450 CERs/year in 2012.
Americas 0.395*** 0.371***
Europe −0.396 −0.379
However, the average sizes (CERs/year) of hydro, biomass energy,
Oceania 0.897*** 0.970*** and landfill gas projects gradually increased during 2008–12
Constant 4.187*** 4.516*** (see Fig. 1). While it reflects some learning from experience, the
Observations 6774 6774 average size of the wind power and industrial gas reduction projects
R-squared 0.614 0.617
gradually decreased during the same period (see Fig. 1). The average
RMSE 1.012 1.008
size of fossil fuel switch projects decreased during 2008–11, but sub-
Note: Log of annualized total CER cost is used as the dependent variable in both models. stantially increased in 2012. No particular trend can be identified for
Parameter estimates with standard errors are given in Annex Table 1. Asterisks ***, **,
and * indicate significance at 1%, 5%, and 10% levels. The results of separate one-tailed
any other project categories (see Fig. 1).
tests indicate that the estimated elasticities for CER volumes are less than 1.00 at 1% signif- As the coefficient estimate of the annualized CERs is positive, it im-
icance level for both models. plies that the duration of the project has an inverse relationship with
CER costs (see Eqs. (8) and (9)). A separate regression of the log of
total costs on log of total CERs, log of project duration, and the same
set of indicator variables show that the estimated coefficient of log of
geographical regions. Thirteen binary variables are used to indicate the project duration is negative and significant (results are not shown
major project types as described earlier; hydro power category is to conserve space). This result implies that, all things equal, projects
used as the base. Two binary variables are used to indicate large with longer duration (but same expected CERs over the life time)
and small scale (used as the base) methodologies. Nine indicator have lower per unit cost of CER generation, thus implying ‘economies
variables for different project start years are used; eight for of time.’
each year during 2006–13, and one for year 2005 and earlier The estimated coefficient of the intercept term appears to be pos-
(used as the base).19 Five indicator variables are used for the projects itive and highly significant. It captures the joint fixed effects of the
located in Africa, Asia, Latin America, Europe, and Oceania. As a base project category (hydro power), methodology (small scale),
majority of the projects are located in Asia, the region is used as start year (2005 or earlier), and region (Asia), and can be interpreted
the base. as the conditional expected mean of log of fixed costs of CERs. As
The calculated elasticities for the log–log model are presented in the hydro power category is used as the base, any significant coeffi-
the second column of Table 4.20 The estimated elasticity of CER vol- cient of the variable indicating another project type can be
ume is positive and significant, suggesting that the cost increases interpreted as the conditional expected mean of log of additional
with the volume of CERs (Table 4). A one-tailed test indicates that fixed costs of generating CERs through that type (with all other
the elasticity is less than one at 1.0% significance level, implying things same).
‘economies of scale.’21 Thus, all things equal, larger projects Estimated coefficients of the indicator variables for wind power,
have lower per unit cost of CERs. Fig. 1 depicts average sizes biomass energy, energy efficiency, solar power, fossil fuel switch,
(1000 CERs/year) of the flow of projects in the CDM pipeline during cement & fugitive, tidal & geothermal, and efficient transport catego-
2008–2012, the first commitment period of the Protocol. Kirkman ries are positive and significant at standard critical levels (Table 4).
The coefficients for methane reduction/avoidance, landfill gas, forest,
and industrial gas reduction projects are negative and significant at
standard critical levels (Table 4). One-tailed tests for pairwise
19
The Kyoto protocol was ratified by the end of 2005. comparison of the project-type coefficients indicate that the
20
Estimated coefficients and their standard errors under each model are reported in fixed costs for transportation projects is the highest, followed in de-
Annex Table 1.
21 scending order by fossil fuel switch, solar power, cement & fugitive,
This test and subsequent ones are one-tailed tests based on a non-linear combination
of estimated parameters and implemented using Stata's (2012) NLCOM post-estimation tidal & geothermal, biomass energy, wind power, energy efficiency,
procedure. hydro power, landfill gas, forest, methane reduction/avoidance, and
136 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141
industrial gas reduction (Stata, 2012). 22 This relative fixed cost
structure is not to the same as that of mean or median LCORs as re-
ported in Table 3.
The distribution of the CDM projects in the sample, and therefore in
the CDM portfolio, does not quite follow this relative fixed cost structure
(see Table 1) even though costs for different project types are likely to
overlap each other for a given project type by location. Consistent
with the estimated set of relative fixed costs, the CDM portfolio contains
a small proportion (3.5%) of transport, fossil fuel switch, cement &
fugitive, and tidal & geothermal projects which account for about
12.3% of annual CERs. About 14.0% of the projects in CDM pipeline are
biomass energy and solar power projects which account for another
5.4% of annual expected CERs. In contrast to the relative fixed cost
structure, the lowest cost categories such as industrial gas reduction,
methane reduction/avoidance, landfill gas, and forest projects account
for 12.3% of the CDM portfolio and 21.8% of the total annual CERs.
Despite having much higher cost of CERs, about 70.2% of the projects
in the pipeline are hydro power, wind power, and energy efficiency
projects; these projects are expected to generate 60.6% of the annual
expected CERs. The distribution of the projects is heavily skewed
towards renewable resource based categories. While considering the re-
newable resource based energy projects separately, the distribution of
the projects into hydro power, wind power, biomass energy, and solar
power categories corresponds to the relative fixed costs for those
categories.
The skewed distribution of the CDM towards energy projects implies
that rather than cost consideration there may be other factors, which in-
fluence investment decisions. For example, environmental policies of
some countries favor particular project types (e.g., renewable energy
projects in China), fiscal policies and institutional settings of a host
country dictate the choice of projects (e.g., China and Brazil), or sectors
such as HFCs, PFCs & SF6, and N2O reduction opportunities are
saturated.23 Another explanation may be that, considering the uncer-
tainties about the functioning of the carbon market and declining CER
prices, investors may find it less risky to invest in projects that generate
valuable byproducts (e.g., electricity) even though projects that gener-
ate CERs only have substantially lower cost (e.g., industrial gas reduc-
tion and forestry).
The estimated coefficient of the indicator variable for the large-scale
methodology is positive and significant at standard critical levels
(Table 4). Thus, fixed costs of generating CERs are higher for large-scale
methodologies than that for the small-scale methodologies. But, only
36.8% of the projects in the sample employed small-scale methodologies.
This may be because some technologies are simply not suited to reduc-
tions of less than 60,000 tCO2e per year (as dictated under small-scale
CDM rules).
The coefficient estimates for variables indicating project start
years 2008–13 are positive and significant at standard critical levels
(Table 4). An upward trend (not continuous though) in CER genera-
tion costs is obvious from these estimates. This is contrary to the pre-
sumption that followers are likely to improve efficiency by learning
from the experience of the leaders. It might be the case that
22
One-tailed tests for pairwise comparison of the project-type coefficients indicate
that the fixed costs for cement, fugitive and biomass energy projects are not signifi-
cantly different from each other. The same results are obtained for the pairs of renew-
able resource based and energy efficiency projects, and forest and methane avoidance
projects.
23
In the early years of CDM, HFC projects were popular due to their cost effective-
ness (e.g., 11,700 CERs are generated for destroying 1 ton of HFC-23 with relatively
small investments and well-established technologies). But, such projects have been
heavily criticized for potential perverse incentives they provide for more GHG emis-
sion (e.g., HFC-23 is a byproduct of HCFC-22). To address this issue, the CDM Execu-
tive Board has decided that only pre-existing HCFC-22 production facilities are
eligible under CDM — new facilities cannot receive CERs, and CERs would be issued
for the existing plants based on their historical HCFC-22 production levels
(Gillenwater and Seres, 2011).
competition for access to natural resources became intense over
time as more and more projects were developed. Consequently,
costs increased which outweighed the benefit of learning from expe-
rience of others. For example, increased use of the same type of bio-
mass in a given location may lead to an increase of cost. However,
such additional costs could be larger or smaller than the average in-
flation rate. They may even decline in some cases due to technologi-
cal change, as is the case with declining prices of primary technology
in both the wind and solar photovoltaic sectors over the same period.
The CDM Executive Board also adjusted the additionality tests to be
more (or less) stringent, thus excluding (including) the lowest
(highest) cost projects that were (not) viable without the CER
revenue.
Estimated coefficients of the indicator variables for Africa, America,
and Oceania are positive and significant, thus implying larger fixed
costs in those regions as compared to that of Asia. The estimated
coefficient for Europe does not appear to be significant at standard
critical levels, implying similar fixed costs as of Asia. The distribution
of the CDM projects across different regions does not follow this relative
fixed cost structure; the distribution is heavily skewed towards Asia.
Kirkman et al. (2012) claim that the distribution of CDM projects and
CERs tends to match the distribution of mitigation potential across
countries, and it is also strongly dependent on the host country's gross
savings, population and GDP. Another potential explanation for this
finding is that while selecting location for the projects investors not
only consider production costs but also transaction costs in various
stages of project development (see Michaelowa and Jotzo, 2005). For
example, access to and ease of use of natural resources and infrastruc-
ture of the host country and its policies towards development of certain
types of CDM projects are crucial for investment decisions. These
findings are also consistent with other studies which show that a poor
financial sector and/or limited supply of funds for domestic investment
limits CDM activity in many countries in Africa and the group of the
least developed countries (LDCs), as well as some in Asia with small
economies and low levels of non-Annex I emissions (see Kirkman
et al., 2012).
To allow the cost function to have a more flexible functional form, we
estimate the costs of CERs with a log–quadratic specification as in Eq. (9).
In particular, the log of total cost is regressed on the same set of explan-
atory variables as in the log–log model plus squared log of the volume of
CERs. Estimated elasticities and semi-elasticities for the log–quadratic
model are presented in the last column of Table 4. The estimated coef-
ficient for the log of CER volume and squared log of CER volume both
appear to be positive and significant in model II (see Annex Table 1).
However, the inclusion of the squared log of abatement as an additional
explanatory variable does not change the resultant elasticities or semi-
elasticities significantly (see Table 4). In contrast to the log–log model,
the coefficient of the variable indicating energy efficiency projects
appears to be positive in the log–quadratic model. However, the coeffi-
cient does not appear to be significant at standard critical levels. For
other indicator variables, each coefficient estimate has the same sign
and level of significance in both models, with the magnitudes being
slightly different. Thus, log–log and log–quadratic models with the
same specification provide similar estimates for the costs of generating
CERs.
We also estimate the log–quadratic model replacing indicator
variables for project and methodology classes with a set of variables
indicating all possible combinations of project type and methodolo-
gy class (e.g., renewable resource and large scale methodology,
renewable resource and small scale methodology). The results
(CER volume and duration elasticities and fixed cost structure)
remain the same as in the log–quadratic model, and not reported to
conserve space.
In summary, the estimates of the aggregate models show that
the average cost of CER decreases with the scale and duration of
the projects (i.e., economies of scale and time). CER costs are lower
 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141 137

Table 5
Results from the project-type models — elasticities and discrete change effects.

Hydro power Wind power Biomass energy Energy eff. Methane reduce Solar power Landfill gas Industrial gas Fossil fuel

Annual CERs 0.67*** 0.67*** 0.31*** 1.10*** 0.79*** 0.40*** 0.47*** 0.43*** 0.65***
Indicator variables for methodology groups (small scale is the base)
Large scale 0.67*** 0.50*** 1.09*** −0.30 0.28* 0.81*** 1.36*** −1.27*** 1.67***
Indicator variables for project start years (2005 and earlier is the base)
2006 0.16* 0.01 0.15 −0.40 −0.20 0.78 −0.18 0.71*** 0.44
2007 0.09 0.15 0.51*** −0.37 0.09 1.28* −0.65* 0.10 0.03
2008 0.29*** 0.20** 0.30** −0.29 0.55** 1.44*** 0.06 0.06 0.02
2009 0.44*** 0.42*** 0.54*** −0.29 0.39 0.63 0.21 1.14*** 0.27
2010 0.47*** 0.54*** 0.33** −0.40 0.41 0.94* 0.70** 1.38*** 0.02
2011 0.45*** 0.57*** 0.27* 0.39 0.46** 0.67 0.32 −0.22 1.29**
2012 0.53*** 0.63*** 0.64*** 0.83* 0.58** 0.68 0.23 0.59*** 0.91
2013 0.51** 0.74*** 0.59*** 0.20 0.79** 0.84 0.50 0.18 0.67***
Indicator variables for regions (Asia is the base)
Africa 0.83*** 1.43*** −0.12 0.91*** 0.45 −0.79 −0.30 −0.52** −2.09
Americas 0.69*** 1.14*** 0.74*** 1.91*** −1.23*** 0.13 −0.45*** −1.17*** −1.09***
Europe −0.23 0.30* −0.80*** 1.01*** −3.64*** - −0.82*** −1.15** −0.55
Europe 1.91*** – – – 0.44*** – – – –
Constant 3.95*** 4.34*** 5.55*** 3.69*** 2.93*** 6.11*** 4.40*** 5.31*** 5.31***
Observations 1930 2404 613 368 396 376 260 138 86
R-squared 0.53 0.75 0.42 0.48 0.68 0.54 0.42 0.52 0.66
RMSE 1.04 0.64 1.03 1.66 1.02 0.79 1.02 0.86 1.15

Note: Log of annualized total CER cost is used as the dependent variable. Parameter estimates with standard errors are given in Annex Table 2. Asterisks ***, **, and * indicate
significance at 1%, 5%, and 10% levels. The results of separate one-tailed tests indicate that the estimated elasticities for CER volumes are less than 1.00 at 5% significance level
for all models.

under small-scale methodologies for evaluating emission reduction. curves show that, except for energy efficiency projects, average
There is an upward trend in CER costs for new projects. Most impor- cost of CERs continuously decrease with the volume of CERs at a
tantly, CER costs vary depending on the type and location of the decreasing rate while the degree of economies of scale vary across
projects. This calls for further investigation of the project type and underlying technologies. More importantly, Fig. 2 allows us to com-
location specific cost structures, as some projects may be cheaper pare the average costs of CERs by different technologies at observed
and easier to deploy than others in the same or neighboring project scales.
countries. At the mean project scales (see Table 1), the average cost of CER is
the highest for the transportation projects, followed in a descending
order by fossil fuel switch, solar power, energy efficiency, wind power,
4.2. Project-type models
biomass energy, hydro power, methane avoidance, landfill gas,
afforestation and reforestation, and industrial gas reduction projects.
We estimate the log–log model for each project category separately. In
However, at a scale smaller than 80,000 CERs per year, emission reduc-
particular, the log of annualized cost is regressed on the log of annualized
tion by biomass energy projects is costlier than by hydro power projects.
CERs and three sets of indicator variables for methodology for calculating
Also, at a scale smaller than 50,000 CERs per year, the average cost of
CERs, project start year, and region. The elasticity and semi-elasticity
CERs by landfill gas projects is higher than that of methane reduction/
estimates for selected project categories are presented in Table 5, and
avoidance projects.
the coefficient estimates with standard errors are reported in Annex
The distribution of the projects in the CDM portfolio, however,
Table 2.24
does not quite follow this cost structure. Relatively higher costs of
From Table 5 cost appears to be elastic to the volume of CERs for
CERs for transportation, fossil fuel switch, solar power, and energy
all selected project categories but energy efficiency. The estimated
efficiency projects may be the reason for smaller shares of these
CER elasticity for energy efficiency category is 1.10, although a one-
categories in the CDM portfolio. Despite having the lowest cost
tailed test indicates that the coefficient is not different from 1.0 at
of CERs (see Fig. 2), afforestation and reforestation category
standard significance levels. For all other project categories, estimat-
accounts for less than 1% of the projects in the CDM portfolio.
ed elasticities are significantly less than one at standard critical levels
Industrial gas reduction, landfill gas, and methane avoidance pro-
(Stata, 2012). The magnitude of CER elasticities at the mean levels
jects together account for only 11.5% of the CDM portfolio although
varies across project types. Consequently, the shape of the average
the costs of CERs by these categories are much lower compared to
cost curve varies across project types over observed project scale.
renewable energy categories (see Fig. 2). About 64% of the CDM
To illustrate this point, the annualized average costs of CERs for dif-
portfolio consists of hydro and wind power projects. Thus, the
ferent types of CDM projects at different levels of CERs are calculated
types of projects that have attracted larger amounts of investments
using the coefficient estimates as reported in Table 5. Fig. 2 depicts
are not the ones associated with lower per unit cost of emission
annualized average cost curves for different types of projects hosted
reduction.
in Asia. The markers on the average cost curves indicate average
While this finding contradicts the presumption that investors
cost of abatement at the mean scales (measured by thousand
will seek out low-cost opportunities, there are potential explana-
CERs per year) for the respective project types.25 The average cost
tions (in addition to the one given earlier). First, it may be the case
that the lowest-cost opportunities identified in the analysis have
24
been fully exploited and cannot be duplicated (e.g., industrial gas
Estimation results for cement & fugitive, afforestation & reforestation, tidal & geother-
mal, and transportation categories are not reported to conserve space. These categories ac-
reduction and methane avoidance projects); as a consequence,
count for a very small share of the CDM. investors have moved to higher-cost alternatives. Second, when
25
Median scales for different project categories are lower than the means. risks associated with the aspects of investments and transaction
138 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141

1500

15
Average Cost (US$/CER)
Average Cost (US$/CER)
1000

10
500

5
0
0

0 200 400 600 0 200 400 600

Emissions Reduction (1,000 CERs per year) Emissions Reduction (1,000 CERs per year)

Transport FFS Solar CH4 LFG Ind. Gas Forest

60
Average Cost (US$/CER)
40
20

Note:
The range of vertical scale is not the same
0

0 200 400 600 for each group of average cost curves.

The marker on each line indicates
Emissions Reduction (1,000 CERs per year)
average cost at mean scale for the
EE Wind Hydro Biomass respective project category

Fig. 2. Estimated average costs of CERs under different types of CDM projects.

are substantial relative to unit costs of generating the CERs, investors
and/or buyers of CERs and byproducts hold preferences about
the underlying technologies used to generate offsets. Lastly,
investors from some countries may not invest in CDM projects
for the CER revenue gain, but for other reasons, such as compliance
or the prospect of compliance in other emergent markets
(e.g., renewable energy).
4.3. Host-country models
About 73.8% of the CDM projects in the sample are located in
three host countries: China (45.4%), India (23.9%), and Brazil
(4.6%). To examine whether the distribution of the projects in
these countries corresponds to the relative average cost of CER, we
separately estimate the log–log model for each of these countries.
The elasticity and semi-elasticity estimates are presented in Table 6,
and the coefficient estimates with standard errors are reported in
Annex Table 3.
Based on the estimates as reported in Table 6, cost appears to be
elastic to the volume of CERs for each of the selected countries imply-
ing economies of scale. 26 A somewhat upward trend in costs is
apparent for the flow of projects in China and India, which is not
surprising since these countries account for such a large share of
the total sample of projects. More importantly, the results do not
indicate that per unit costs of CERs are lower in the host countries
where a large share of CDM investments are made and, within
the host countries, investments are not concentrated in project
categories with the lower per unit costs. Despite having higher
cost of CER, China hosts a majority of CDM projects. Industrial
gas reduction projects have the lowest average cost of CER,
followed by methane reduction/avoidance and forest projects in
all three of the selected countries. However, the low-cost project
categories account for a small share of these countries' CDM
portfolio; CDM investments are highly concentrated in energy pro-
jects. Thus, the distribution of different types of projects within
each of the major host countries is not skewed toward low-cost
projects.
Fig. 3 depicts the estimated average cost curves for hydro power,
wind power, biomass energy, and, energy efficiency projects in
China, India, and Brazil. 27 The markers on the average cost curves
indicate average costs of CERs at the mean scales of the project
categories in the respective host countries. As Fig. 3 shows, the
average cost of CER by hydro power projects is the lowest in India
and the highest in China at the respective mean (and median) scales.
Yet, only 14.1% of all hydro power projects in these three countries
are hosted in India; 79.5% of the hydro power projects in these coun-
tries are hosted in China. The average cost of CER by wind power pro-
jects is the lowest in China and the highest in India at the respective
mean (and median) scales; wind power projects in Brazil are more
efficient than those in China at a scale larger than 100,000 CERs per
year and less efficient than those in India at a scale smaller than
40,000 CERs per year. However, 63.3% of the wind power projects

26
The results of one-tailed non-linear tests indicate that the estimated elasticity for
27
abatement is significant at standard critical levels for the projects in China, India, and GHG abatement cost curves for different CDM host countries as drawn by Castro
Brazil (Stata, 2012). (2012) do not show economies of scale.
 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141 139

Table 6 in these countries are hosted in China while 33.7% are situated in
Results from the host-country models — elasticities and discrete change effects. India.
China India Brazil At the respective mean and median scales of biomass energy pro-
Annual certified emission reduction (CER) 0.643*** 0.774*** 0.410***
jects in these countries, the average cost of CER is the lowest in China
Indicator variables for project categories (hydro power is the base) and the highest in Brazil. But, 61.6% of the biomass energy projects in
Wind power 0.154*** 0.948*** 0.291 these three largest host countries are situated in India while 28.6% are
Biomass energy 0.510*** 1.077*** 1.031*** located in China. At all observed scales of the energy efficiency projects
Energy efficiency −0.265** 1.169*** 1.824***
in these three countries, the projects in Brazil have the highest cost of
CH4 avoidance/reduction −0.866*** −1.007* −3.712***
Solar power 1.419*** 2.262*** 1.324*** emission reduction while the projects in China have the lowest cost.
Landfill gas 0.235** 0.880 −0.865*** However, 51.7% of the energy efficiency projects in these three coun-
HFCs, PFCs & SF6, and N2O −1.380*** −1.439*** −3.700*** tries are situated in India while 46.9% are hosted in China. The distribu-
Fossil fuel switch 2.778*** 1.982*** tion of other project categories in these three countries also does not
Cement & fugitive 0.229 2.692*** 1.341***
Afforestation & reforestation 0.095 −0.953** −1.420***
correspond to the relative average cost of CER. Thus, sector specific
Tidal & geothermal 1.653*** – – CDM investments are not concentrated in places where per unit cost
Transportation 2.888*** 4.950*** – of CER is the lowest.
Indicator variables for methodology groups (small scale is the base) The principle of comparative advantage implies that even if one
Large scale 0.649*** 0.459*** 0.833***
country is more efficient in emission reduction in all sectors, all
Indicator variables for project start years (2005 and earlier is the base)
2006 0.114 0.164 0.026 countries involved in producing CERs will still gain by concentrating
2007 0.104 0.096 0.361 in the sectors with the lowest cost (see Baumol and Blinder, 2011).
2008 0.271*** 0.284*** 0.28 Based on the coefficient estimates presented in Annex Table 3, average
2009 0.558*** 0.246** 0.041 costs of CERs by wind, hydro, biomass, and energy efficiency projects in
2010 0.654*** 0.256*** −0.183
China, India, and Brazil are compared. It appears that China has a
2011 0.615*** 0.353*** 0.53
2012 0.712*** 0.278*** 0.501 comparative advantage (i.e., lowest opportunity cost) in energy effi-
2013 0.675*** 0.265* 0.819*** ciency projects, India has a comparative advantage in hydro power pro-
Constant 4.035*** 3.278*** 5.526*** jects, and Brazil has a comparative advantage in wind power projects.
Observations 3580 1526 176
However, energy efficiency category accounts for only 7.9% of the
R-squared 0.557 0.673 0.553
RMSE 0.865 0.962 0.931
CDM projects in China, only 11.6% of the projects in India are hydro
power projects, and only 17.8% of the projects in Brazil are wind
Note: Log of annualized total CER cost is used as the dependent variable. Parameter estimates
with standard errors are given in Annex Table 3. Asterisks ***, **, and * indicate significance
power projects. Thus, the distribution of the different types of CDM pro-
at 1%, 5%, and 10% levels. The results of separate one-tailed tests indicate that the estimated jects across host countries does not follow the principle of comparative
elasticities for CER volumes are less than 1.00 at 1% significance level for both models. advantage.
Average Cost (US$/MWh)
50

50
Average Cost (US$/CER)

Hydro Power Wind Power

40

40
30

30
20

20
10

10

0 100 200 300 400 500 0 100 200 300 400 500

Emission Reduction (1,000 CERs per year) Emission Reduction (1,000 CERs per year)

China India Brazil China India Brazil

150

300
Average Cost (US$/MWh)

Average Cost (US$/MWh)

Biomass Energy Energy Efficiency

100

200
100
50
0

0 100 200 300 400 500 0 100 200 300 400 500

Emission Reduction (1,000 CERs per year) Emission Reduction (1,000 CERs per year)

China India Brazil China India Brazil

Fig. 3. Estimated average cost of CERs by hydro power, wind power, biomass energy, and energy efficiency projects in China, India, and Brazil (markers show average costs at mean project scales).
140 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141
Comparative advantage, however, is only one of several factors in
attracting CDM investments. Endowment of and access to natural re-
sources, various costs of doing business, and national policies
contribute to the competitive advantage of the host countries (see
Stutz and Warf, 2006). For example, the Chinese government's support
for renewable energy projects includes reduced corporate income
taxes, significant reductions in value added taxes, feed-in tariffs and
subsidies to operators of renewable energy projects to compensate for
their costs (KPMG International Cooperative, 2011). Furthermore,
central government policies and measures have likely attributed to
the success of the use of the CDM for energy and other types of projects
in China.
5. Conclusions
In this paper we examine the costs of certified emission reduction
(CER) by various projects under the Clean Development Mechanism
of the Kyoto Protocol. The CDM projects are located in different de-
veloping countries, and vary by underlying technology (type),
scale, and methodology for evaluating emission reductions. Many
of the projects simultaneously generate CERs and additional
byproducts. Using project data, we calculate the cost of CER by
subtracting the discounted stream of byproduct revenues from the
fixed and variable costs over the life of each project. Thus, we esti-
mate a separable cost function based on an explicit disentanglement
of cost and revenue. To analyze the CER cost structure of the CDM
projects, we control for the duration of the projects, types of the pro-
jects, types of methodology used for evaluating emission reduction,
years in which the projects began, and locations of the projects in
our empirical models. We repeat the analysis for specific types of
projects, and also for the countries that host a large number of
CDM projects. We consider both log–log and log–quadratic function-
al forms for the cost function. From a technical aspect, we find that
introducing additional flexibility in the form of the cost function
had little effect on the estimation results.
The main results of this paper are as follows. First, the CDM projects
in general have economies of scale in CERs; larger projects have lower
per unit cost of CER. Second, there is significant variation in the scale
effects by project types; for example, the average cost of CER for
solar power, biomass energy, landfill gas reduction, and industrial
gas reduction projects declines at a faster rate than aggregate
average, while the average cost of CER for energy efficiency and
methane reduction/avoidance projects declines at a slower rate than
the aggregate average. Third, the average cost of CER decreases with
project duration, even though differences in the timing of outputs
have been accounted for by discounting; the CDM projects with longer
duration have lower per unit cost of CER. Fourth, there is an upward
trend in the costs of CERs; the average costs of CER are higher for the
projects that started later. Fifth, while the average costs of CER signifi-
cantly vary across project types, the distribution of the projects in the
CDM portfolio or a given location does not quite follow the relative
cost structure. Finally, the average cost of CER significantly varies across
project locations, but the distribution of the CDM projects in different
regions or host countries does not strictly follow the principle of com-
parative advantage.
In accordance with the evidence of economies of scale, the average
size of hydro power, biomass energy, and landfill gas projects gradually
increased during the first commitment period of the Kyoto Protocol,
thus implying learning from experience. In contrast, the average
size of wind power and industrial gas reduction projects gradually
decreased during the same period while no particular trend is observed
in other categories. The upward trend in the cost of CER is contrary
to the presumption that followers are likely to improve efficiency.
This may be because of enhanced completion for access to natural
resources and stringent additionality tests adopted by the CDM
Executive Board. We also find strong evidence that the fixed costs of
CERs are lower for small-scale methodologies for evaluating emission
reductions, while some technologies are not suited for small-scale
methodologies.
Surprisingly, we find that the types of projects that attracted larger
number of investors (e.g., hydro wind, and solar power), or larger
amount of investments, were not the projects associated with lower
per unit cost of CER. Even for the three largest host countries, China,
India, and Brazil, investments are not concentrated in projects with
the lowest per unit costs (e.g., afforestation & reforestation, industrial
gas reduction, landfill gas reduction, and methane avoidance). Similarly,
we find little evidence that per unit costs of CERs were lower in
the places where investments most often took place or where most of
the investments were made. While the costs of CERs are similar in
Asia and Africa, more than 83% of the projects in the CDM pipeline are
in Asia; more than 69% of the projects are in China and India alone.
More than 75% of the projects in China and India are energy related.
While per unit cost of abatement plays an important role in the
bottom-up and top-down models to evaluate emission reduction po-
tential and analyze policy alternatives, these findings contradict the pre-
sumption of such models that project investors seek out low-cost
opportunities.
We also find that the distribution of different types of CDM projects
across host regions or countries does not follow the principle of compar-
ative advantage. It appears that China has a comparative advantage
(i.e., lowest opportunity cost) in energy efficiency projects, India has
a comparative advantage in hydro power projects, and Brazil has a
comparative advantage in wind power projects. However, energy
efficiency category accounts for only 7.9% of CDM projects in China,
only 11.6% of the projects in India are hydro power projects, and only
17.8% of the projects in Brazil are wind power projects. Comparative ad-
vantage may be only one of several factors in attracting CDM investments.
Endowment of and access to natural resources, various costs of doing
business, and national policies contribute to competitive advantage of
the host countries.
These findings are significant but perplexing, yet there are several
potential explanations that are consistent with a market where
unit costs are crucial in characterizing investment decisions. It may
be the case that the lowest-cost opportunities identified in the
analysis have been fully exploited and cannot be duplicated; as a
consequence, investors have moved on to higher-cost alternatives.
Also, it may be the case that investors consider production costs as
well as transaction costs in various stages of project development;
access to and ease of use of natural resources, ease of doing business,
and national policies of certain host countries might have motivated
investment decisions in those countries and to certain types of CDM
projects. It is worth pointing out that the unit cost of CER does not
necessarily reflect the true cost or value of the projects to investors;
investors also consider the risks and costs associated with transac-
tions at different stages of the production and marketing processes.
Future research may be able to take account of transaction costs in
the estimation of the cost of CERs and analyze competitive advan-
tages of the host regions and countries. Nonetheless, our findings
provide a basis for evaluating the incentives that the mechanism
offers as a cost effective policy instrument that balance greenhouse
gas mitigation across sectors and regions, while fulfilling the objec-
tive of the convention.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.eneco.2014.10.020.
References
Baumol, W.J., Blinder, A.S., 2011. Economics: Principles and Policy, 12th edition. Mason,
South-Western Cangage Learning.
 S.M. Rahman, G.A. Kirkman / Energy Economics 47 (2015) 129–141
Booth, W.J., 1991. Economies of time: on the idea of time in Marx's political economy. Pol.
Theory 19 (1), 7–27.
Boudreaux, D.J., 2008. Comparative advantage, In: Henderson, David R. (Ed.), Concise
Encyclopedia of Economics, 2nd ed. Library of Economics and Liberty, Indianapolis.
Boyd, G., Molburg, J., Prince, R., 1996. Alternative Methods of Marginal Abatement
Cost Estimation: Non-parametric Distance Function. United States Department
of Energy.
Bystrom, O., 1998. The nitrogen abatement cost in wetlands. Ecol. Econ. 26 (3), 321–331.
Castro, P., 2012. Does the CDM discourage emission reduction targets in advanced devel-
oping countries? Clim. Pol. 12, 189–218.
Castro, P., 2014. Discounting emissions credits and competitiveness of CDM host coun-
tries. Climate Change Mitigation in Developing Countries: A Critical Assessment of
the Clean Development Mechanism. Edward Elgar, Cheltenham, Northampton.
Considine, T.J., Larson, D.F., 2006. The environment as a factor of production. J. Environ.
Econ. Manag. 52 (3), 645–662.
Considine, T.J., Larson, D.F., 2009. Substitution and technological change under carbon cap
and trade: lessons from Europe. Policy Research Working Paper Series 4957. World
Bank, Washington.
Ferdinand, M., Melum, F., 2013. Update to our sCER Price Forecast: Phoenix or Dying Swan.
Carbon Market Research, (http://www.pointcarbon.com/research/marketsoverview/1.
2126808?date=20130107&sdtc=).
Gillenwater, M., Seres, S., 2011. The Clean Development Mechanism: A Review of The First
International Offset Program. Markets and Business Strategy SeriesPew Center on
Global Climate Change.
Hamaide, B., Boland, J.J., 2000. Benefits, costs, and cooperation in greenhouse gas
abatement. Clim. Chang. 47 (3), 239–258.
Hartman, R.S., Wheeler, D., Singh, M., 1994. The cost of air pollution abatement. Appl.
Econ. 29 (6), 759–774.
Kirkman, G.A., Haites, E., Seres, S., Spalding-Fecher, R., 2012. Benefits of the Clean
Development Mechanism 2012. UNFCCC, (http://cdm.unfccc.int/about/dev_ben/
index.html).
Kirkman, G.A., Seres, S., Haites, E., 2013. Renewable energy: comparison of CDM and
Annex I projects. Energy Pol. 63, 995–1001.
KPMG International Cooperative, 2011. Energy and natural resources: taxes and incen-
tives for renewable energy. http://www.kpmg.com/Global/en/IssuesAndInsights/
ArticlesPublications/Documents/Taxes-Incentives-Renewable-Energy-2011.pdf.
Larson, D.F., Ambrosi, P., Dinar, A., Rahman, S.M., Entler, R., 2008. A review of carbon mar-
ket policies and research. Int. Rev. Environ. Resour. Econ. 2 (3), 177–236.
141
Maradan, D., Vassiliev, A., 2005. Marginal costs of carbon dioxide abatement: empirical
evidence from cross-country analysis. Swiss J. Econ. Stat. 141 (III), 377–410.
Metz, B., Davidson, O., Bosch, P., Dave, R., Meyer, L., 2007. Climate change 2007 abate-
ment. Contribution of Working Group III to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge.
Michaelowa, A., Jotzo, F., 2005. Transaction costs, institutional rigidities and the size of the
clean development mechanism. Energy Pol. 33, 511–523.
Newell, R.G., Stavins, R.N., 2003. Cost heterogeneity and potential savings from market-
based policies. J. Regul. Econ. 23 (1), 43–59.
Newell, R.G., Pizer, W., Shih, J.-S., 2003. Estimating the gains from emission trading. Work-
ing Paper. Resources for the Future, Washington, DC.
Nordhaus, W., 1994. Managing the Global Commons: The Economics of Climate Change.
MIT Press, Cambridge, MA.
Pizer, W., Kopp, R., 2005. Calculating the cost of environmental regulation. In: Mäler, Karl-
Göran, Vincent, Jeffrey R. (Eds.), Handbook of Environmental Economics. Elsevier,
Amsterdam.
Rahman, S.M., Dinar, A., Larson, D.F., 2012. Cross-country adoption of the clean develop-
ment mechanism. Paper Presented at the 2012 EAERE Annual Conference.
Rahman, S.M., Larson, D.F., Dinar, A., 2014. Costs of Greenhouse Gas Emissions Abatement
Under the Clean Development Mechanism. Climate Change Econ. (forthcoming).
Sims, R.E.H., Rogner, H.-H., Gregory, K., 2003. Carbon emission and mitigation cost com-
parisons between fossil fuel, nuclear and renewable energy resources for electricity
generation. Energy Policy 31, 1315–1326.
Stata, 2012. Stata Reference Release 12. Stata Corporation, College Station, Texas.
Stutz, F.P., Warf, B., 2006. The World Economy: Resources, Location, Trade, and
Development, 5th edition. Prentice Hall.
Timilsina, G.R., Lefevre, T., 1999. Reducing GHG emissions from the power sector in devel-
oping Asian countries: an AIJ perspective. World Resour. Rev. 11 (1), 115–131.
USEIA, 2014. Annual Energy Outlook 2014, DOE/EIA-0383(2014). U.S. Energy Information
Administration, (available at http://www.eia.gov/forecasts/aeo/pdf/0383(2014).pdf).
Vattenfall, 2007. Global mapping of greenhouse gas abatement opportunities up to 2030.
Available at http://www.schoonschakelen.nl/library_files/1_1253885146_Power%
20Sector_Global_Mapping_of_Greenhouse_Gas_Abatement_Opportunities_up_to_
2030-Vattenfall-2007%5B1%5D.pdf (accessed 10 June 2014).
Wetzelaer, B.J.H.W., van der Linden, N.H., de Coninck, H.C., Groenenberg, H., 2007. GHG
marginal abatement cost curves for the non-Annex I region. ECN Report No. ECN-E-
06-060. Petten, Energy Research Centre of the Netherlands.