The cost of providing electricity to Africa

Category	Articles
Sub-Categories	n/a
Year	2012
Author(s)	Haakon Vennemo Orvika Rosnes
	https://www.researchgate.net/profile/Haakon-Vennemo/publication/256967775_The_cost_of_providing_electricity_to_Africa/links/5f2bf2f192851cd302e039dc/The-cost-of-providing-electricity-to-Africa.pdf
Download	file_download (383.9 kB)
Read in browser	PDF

The cost of providing electricity to Africa
Orvika Rosnes	a, Haakon Vennemo	b,⁎	
aStatistics Norway. PO Box 8131 Dep., 0033 Oslo, NorwaybOslo & Akershus University College, PO Box 4 St. Olavs Plass, 0130 Oslo, Norway	
abstract article info	
Article history:
Received 20 February 2011
Received in revised form 14 May 2012
Accepted 3 June 2012
Available online 9 June 2012
JEL classiﬁcation:
O13
O55
Q41
Q47
Keywords:
Electricity
Sub-Saharan Africa
Modeling
Trade	
Sub-Saharan Africa lacks electricity. We estimate the cost of providing electricity to the region. To do so, we
build an optimisation model that links the electricity demand to the supply and links the supply to the gen-
eration, distribution and transmission of electricity between countries. To the best of our knowledge, such a
model is novel in the literature.
We determine that the investment cost of providing electricity to Sub-Saharan Africa over a 10-year period is
between 160 and 215 billion U.S. dollars, depending on assumptions for electricity access and the cross-
country electricity trade. Although the electricity trade increases the investment cost estimate moderately,
it provides a high return to African countries and is cost-efﬁcient overall.
© 2012 Elsevier B.V. All rights reserved.	
1. Introduction
Along with India, Sub-Saharan Africa is the region in the world
with the lowest per capita consumption of electricity and the lowest
rate of electricity access. The annual per capita electricity consumption
in Africa is 518 kWh, which is equal to the electricity consumption
over 25 days in the OECD (IEA, 2008). Two thirds of the population in
Sub-Saharan Africa—585 million people—lack access to electricity (IEA,
2011). In rural areas of Malawi, Ethiopia, Niger, Chad and several other
countries, less than 2% of the population have access to electricity.	
1	
While Sub-Saharan Africa is struggling to bring electricity to its
people, the region has ample undeveloped energy resources. Several
countries are rich in hydropower resources. For example, in the
Democratic Republic of the Congo, a single hydropower project
(Grand Inga) has a potential of 40 GW.	
2Other countries with large
hydropower resources include Cameroon, Ethiopia, Sudan, Nigeria,
Guinea, Angola and Mozambique. Several countries house large reserves
of natural gas, notably Nigeria and Angola. The Republic of South Africa
has major coal reserves.Evidently, a lack of natural resources does not explain the inade-
quate supply of electricity in Sub-Saharan Africa. But the resources
are not put adequately to use. Investment in, e.g., a hydropower
plant is a large undertaking for a poor country. The fact that the
hydropower reserves are often located far from population centres
adds further cost; there are vast areas to be covered by the grid.
Hence, cost is a deﬁnite problem, but donors are prepared to pick
up a share of the bill.	
3But surprisingly, nobody knows with any degree
of conﬁdence what supplying electricity to Sub-Saharan Africa would
actually cost. The funding implications of pursuing political targets for
household access, or environmental targets, are not known either.
In this paper, we estimate the cost of providing electricity to Sub-
Saharan Africa, as well as the costs of political interventions for trade
and electricity access. To do so, we set up an optimisation model
that, for each country of Sub-Saharan Africa, describes the electricity
demand and the opportunities for electricity generation and trade
with other countries. The model includes an extensive database of
infrastructure stocks, fuel costs and investment and refurbishment
costs for a range of technologies, including a long list of potential
hydropower projects throughout the continent. This database is com-
bined with data on electricity demand and information about targets
related to electricity access. Because several countries are equipped	
Energy Economics 34 (2012) 1318–1328	
⁎Corresponding author. Tel.: + 47 911 36 716.
E-mail addresses:orvika.rosnes@ssb.no(O. Rosnes),
haakon.vennemo@vista-analyse.no(H. Vennemo).
1When no reference is given, data references in this paper are toRosnes and Vennemo
(2009)and references therein.
21 GW is 1000 MW. See, e.g.,IEA (2008)Box 15.2 for a description of the Grand Inga
project.	3SeeEberhard and Shkaratan (2012)for detailed analysis of funding strategies.	
0140-9883/$–see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.eneco.2012.06.008	
Contents lists available atSciVerse ScienceDirect	
Energy Economics	
journal homepage: www.elsevier.com/locate/eneco

with abundant energy resources while others are not, it is also impor-
tant to model trading opportunities in the form of transmission
investments and their associated costs.
Having described the generation opportunities, the links between
countries and the cost of distributing electricity to households and
industries, the model estimates the cost of electricity supply in Sub-
Saharan Africa with supply linked to demand and countries linked
in a trading framework. An interactive version of the model is avail-
able athttp://www.infrastructureafrica.org/tools/models/electricity-
spending-needs-model.
We wish to analyse the impact of a 10-year expansion programme
for electricity in Sub-Saharan Africa. To do so, we use our model to
simulate the year 2015 and compare the results with our benchmark
year of 2005. The year 2015 is the target year for the Millennium
Development Goals, and energy access is important to achieve many
of these goals. The year 2005 was chosen as the benchmark year
due to the data availability. In a technical sense, the importance of
the 10-year label is that economies, in most cases, have increased in
size, and the population has grown by the period's end. In addition,
it takes time to commission and build power plants and networks,
and the time period allows for this. Most capacity additions that
have come online since 2005 are included in the model as planned
(exogenous) investments. Hence, the results are fairly representative
of the next 10 years ahead.
Focusing on a 10‐year investment window, we estimate the re-
quired investment for providing electricity to Sub-Saharan Africa
will be 160–215 billion U.S. dollars (USD), depending on targets for
electricity access, trade and other assumptions.	
4This amount will
provide 100–120 GW of new and refurbished electricity generation
capacity, plus associated distribution and transmission investments.
Around two thirds of the additional capacity is new (“greenﬁeld”)
investment.
Compared to the GDPs of most African states, 160–215 billion USD
is a very large amount.	
5However, it is misleading to compare a one-
off investment to annual aidﬂows or the annual GDP. Our estimated
annualisedcost of meetingadditionalpower needs in Sub-Saharan
Africa is 28–37 billion USD. This amount is similar to the annual over-
seas development assistance to Sub-Saharan Africa.	
6	
In this study, we pay particular attention to the gains from trade in
Sub-Saharan Africa. We construct a benchmark scenario of free trade,
which we compare with a scenario of the current capacity for trade.
Weﬁnd that trade requires an investment because new inter-
connectors must be built and also because a free trade scenario relies
on capital-intensive hydropower. However, the investment has an
annual expected return of 20–30% for the foreseeable future. In
other words, trade is a better investment than most physical invest-
ment opportunities. Trade brings not only economic beneﬁts but en-
vironmental beneﬁts as well: 70 million tonnes of CO	
2may be saved
annually if unrestricted electricity trade is allowed.
We also focus on the cost of scaling up electricity access. Again, we
construct two scenarios: a benchmark scenario in which national
targets for electricity access are implemented and a scenario in
which there is constant access. We estimate that providing electricity
access for an additional 300 million people will equal around a quar-
ter of the benchmark cost of the electricity supply.
This paper also reports the sensitivity of the primary results to
critical assumptions concerning the economic growth and oil prices.
These analyses indicate the power of our model to handle a range of
interesting policy questions.To the best of our knowledge, our model is theﬁrst detailed model
of the regional electricity system in Sub-Saharan Africa. However,
there have been several attempts at modelling the system from
other angles. The IEA, e.g.,IEA (2011)uses a model that estimates
the electricity provision to the whole of Africa as a by-product of a
global energy demand and supply. Hence, the estimate is aggregate
in nature, and electricity in Africa is not the main focus of the global
energy assessment. Also using aggregate assumptions, theWorld
Bank (2006)estimated the cost of providing electricity access to
households and villages in Sub-Saharan Africa. This estimate ignored
the electricity demands of industries and afﬂuent households, as well
as most of the grid cost.Nexant (2007)estimated the cost of electric-
ity provision in the Southern African Power Pool, andGnansounou
et al. (2007)estimated this cost in the Western Africa Power Pool.
These efforts also appear to ignore most of the grid cost. We compare
our work to these studies inSection 6.
The volume ofEberhard et al. (2011)is a broad discussion of the
power sector in Sub-Saharan Africa and draws on several sources,
includingRosnes and Vennemo (2009). It does not emphasise the
quantitative methodology, however.
Our model groups countries into four power pools. Currently,
several countries cooperate closely to supply electricity to the region.
For example, the SAPP (Southern African Power Pool) has a formal
organisation. In other regions, e.g., Central Africa, there are currently
no such organisations. The concept of a power pool is, nevertheless,
useful in organising the discussion of results.
To explain the results, it is necessary to describe the theoretical
model we use and the data that provides the model with its empirical
properties. These issues are discussed next.
2. Model framework
To estimate the investment needs in each country, we have built
a linear programming optimisation model that simulates electricity
generation and investment strategies in response to the demand in
the countries of Sub-Saharan Africa. The optimisation criterion is cost-
minimisation subject to constraints; hence, we present the least-cost
estimates of the electricity supply.
Our model covers the following 43 Sub-Saharan African countries:
•The Southern Africa Power Pool (SAPP) consists of Angola, Botswana,
the Democratic Republic of the Congo (DRC), Lesotho, Mozambique,
Malawi, Namibia, South Africa, Zambia and Zimbabwe.
•The Nile Basin–East Africa Power Pool (EAPP) consists of Burundi,
Djibouti, Egypt, Ethiopia, Kenya,Rwanda, Sudan, Tanzania and Uganda.
•The Western Africa Power Pool (WAPP) consists of Benin, Burkina
Faso, Côte d'Ivoire, Gambia, Ghana, Guinea, Guinea-Bissau, Liberia,
Mali, Mauritania, Niger, Nigeria, Senegal, Sierra Leone and Togo.
•The Central Africa Power Pool (CAPP) consists of Cameroon, the Central
African Republic, Chad, the Republic of the Congo (Br), Equatorial
Guinea and Gabon.
•Cape Verde, Madagascar and Mauritius are included as island states.
Each country is represented by a demand side and supply side.
The demand is formulated outside the optimisation model. The sup-
ply is determined by the optimisation model based on assumptions
concerning available technologies and capacities with the associated
costs. The model minimises the cost of producing and distributing
electricity, given the demand. The costs include production as well
as the investment and refurbishment costs for generation; the invest-
ment costs of new cross-border transmission lines; the investment,
refurbishment and operation costs of domestic distribution lines;
and the cost of new connections to the power grid.
Each country has two demand centres or nodes: urban and rural.
For the urban nodes, it is assumed that 70% of the electricity con-
sumption occurs during peak periods. The rationale for this is that
most of the industry is located in densely populated areas. For rural	
4Here and elsewhere, we use the 2005 value of the USD.5Current Gross National Income in Sub-Saharan Africa was 706 billion USD in 2006
(World Bank, 2009).
6According to the organisation One, development assistance to Sub-Saharan Africa
amounted to 27 billion USD in 2005 and 40 billion in 2010, seeOne (2012).1319 O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

nodes, there is a 50–50 split between the peak and off-peak periods.
Peak and off-peak periods are of equal length.
A mathematical formulation of the model is given in the Appendix.
The numerical model is implemented in the GAMS programming
language (Brooke et al., 1998).
2.1. Supply system
The supply system consists of power generation plants, the trans-
mission grid (between and within countries) and the distribution
grids (within countries). The transmission grid is deﬁned by the two
highest voltage levels available in a country.
2.1.1. Generation
Each generation technology is described in terms of capacity and
cost; we use the countries’existing assets (as of 2005) as a starting
point. The existing power plants (bottom-up data) are aggregated
into technology-speciﬁc categories. The available technologies for
power generation are hydropower (i.e., large-scale, including pumped
storage, and small-scale), thermal (i.e., fuelled by coal, natural gas,
diesel or heavy fuel oil), nuclear, solar photovoltaic (PV) and geother-
mal power. In addition to the grid-based electricity supply, the off-grid
supply provided by mini-hydro, diesel and solar PV is an option in
rural areas.
Large-scale hydropower differs from mini-hydro in terms of the
productionﬂexibility and investment potential. We assume that all
mini-hydro power is run-of-river and has no reservoirs, which means
that mini-hydro production remains constant day and night. For large-
scale hydro, the reservoirs provideﬂexibility: it is possible to shift
production from a low-demand (off-peak) period to a high-demand
(peak) period. Thisﬂexibility varies from country to country.
Production in each technology is limited by a maximum number
of production hours during a year. For thermal technologies, the max-
imum number of hours per year that a power plant can run is limited
by the need for maintenance and other planned outages. For hydro-
power and solar PV, the number of hours per year will be limited by
factors including the water inﬂow and available sunlight. Conse-
quently, the number of available production hours per year will be
considerably lower than for the thermal technologies.
Capacity expansion is possible through investment, but the maxi-
mum investment in each technology is exogenous for each country.
This is particularly important for hydropower generation, as the
hydropower potential varies signiﬁcantly from country to country.
Similarly, investment in thermal technologies is limited by the avail-
ability of fuel. An exogenous minimum investment describes thermal
power projects that are being developed but are not yet online.	
7	
These projects are part of our least-cost solution, regardless of whether
they have a higher cost than the alternatives.
Given costs and other constraints, the model endogenously deter-
mines which investments (both technology and location) are proﬁtable.
In addition, refurbishment can extend the lifespan of the existing capac-
ity; hence, our model has two investment categories: new (“greenﬁeld”)
investment and refurbishment.	
8	
The concept of cost includes the capital cost and the variable oper-
ating cost. The variable operating cost is the cost of fuel, operation
and maintenance.2.1.2. Transmission and distribution
Investments must be made to extend the distribution and trans-
mission grids, and maintenance of these grids is necessary to keep
them in working condition. These investments are determined en-
dogenously in the model as part of the least-cost solution. Again, we
distinguish between new capacity and refurbishment.
Tradeﬂows are also part of the optimisation. In other words, elec-
tricity trade isnotbased on an extrapolation of today's trends but is
the outcome of the optimal development of generation resources in
the region and expansion of the cross-border grid. Hence, in compar-
ison to today, the tradeﬂows in the future may be much larger or may
even change direction.
Transmission losses are accounted for by assuming an exogenous
percentage loss when electricityﬂows from one node to another.
The transmission loss depends on the nodes and the direction of
theﬂows. Distribution losses (in the low-voltage grid) are set as a
percentage of the net demand for each node.
We proceed to describe the data that form the parameters of the
model.
3. Demand side data
The demand side of each country consists of the three categories
market demand, suppressed demand and new connections.Market
demandfor electricity is deﬁned as the demand resulting from eco-
nomic growth and structural change; that is, the growth in the exis-
ting consumers’demand.Suppressed demandis that portion of the
existing customers’electricity demand that is currently not met due
to blackouts and brownouts.New connectionsrefer to social targets
for new electricity access. The three categories of demand are added
to form our country-speciﬁc demand projection:
Demand¼market demandþsuppressed demand
þnew connections
Market demand and suppressed demand are located in the urban
node, while new connections are distributed between the urban and
rural nodes.
Below, we explain the methodology for and parameters used in
projections of the market demand and new connections. We refer to
Rosnes and Vennemo (2009)for a description of suppressed demand
but note that suppressed demand is country speciﬁc. The average
downtime across the countries of WAPP (22%) and CAPP (10%) is
much higher than in EAPP (5%) and SAPP (4%).
3.1. Market demand
There are few econometric models of electricity demand in Sub-
Saharan Africa (seeKhanna and Rao, 2009for a survey), and weﬁnd
it useful to explain the model estimated for this project. We model
market demand, measured asannual per capita electricity consumption,
as a log-linear function of real gross domestic product (GDP) per capita
and of the agricultural and manufacturing shares of the economy. In ad-
dition, we followFay and Yepes (2003)and include country-speciﬁc
ﬁxed effects and time trends to proxy for prices and technological
innovation.
We implement the econometric model on a panel covering 11 Sub-
Saharan countries.	
9For the remaining countries, the data are either
unreliable or missing. The estimation is based on 30 years (1970–2000),
which, along with the data for the 11 countries, yields 330 observations.
We allow forﬁrst-order autocorrelation and panel correlation in the
residuals to control for unobserved factors that may affect the electricity	
7The OCGT plants in South Africa that came online in 2007–2009 are included in this
manner.
8Refurbishment refers to major refurbishments: either to prolong the life of an out-
dated plant for which the operating life is coming to an end by restoring it to full op-
erational status or to repair generation assets that have been seriously damaged, for
example, due to war. Refurbishments do not include costs for ordinary maintenance
and repair work.	
9Benin, Côte d'Ivoire, Cameroon, Ghana, Kenya, Mozambique, Nigeria, Senegal, the
Republic of South Africa, Tanzania and Zambia. 1320O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

demand and may be correlated over time or across countries. All
variables are in natural logarithms; in this way, the estimated coef-
ﬁcients can be interpreted as elasticities. We estimate the model
with feasible generalised least squares (FGLS), using the economet-
ric software Stata 9.
The estimation results are reported inTable 1.
The results are robust in terms of alternative speciﬁcations and
sample sizes, with two exceptions. First, the coefﬁcient on construc-
tion is sensitive to changes in speciﬁcation, the use of alternative
estimators and changes in sample size. It is close to zero in many
speciﬁcations and is frequently not robust. We include it in the
speciﬁcation below but exclude it from the projections of future
demand. Second, the coefﬁcient on income, representing income
elasticity, drops from approximately 1.1 to 0.5 if one includes urban-
isation in the regression. The coefﬁcient on urbanisation is highly
statistically signiﬁcant and slightly below unity (0.9). The reason for
this effect is the omitted variables bias. Because urbanisation is posi-
tively correlated with both real income per capita and electricity
consumption (because people in urban areas use more electricity
than those in rural areas), omitting it from the regression raises the
coefﬁcient on income. We chose to proceed with the model with
higher income elasticity because projections of future urbanisation
are highly uncertain.
3.2. New connections
Because electricity access rates are very low in many countries,
conscious interventions are needed to raise access. New connections
refer to the social targets for new electricity access. To model the dif-
ferent targets for electricity access, one needs to specify the average
consumption per household and the number of new connections.
In arriving at an estimate ofaverage consumption per connection,
we utilise an extensive data set collected in Tanzania (Econ Pöyry,
2006; see alsoRosnes and Vennemo, 2009). The weighted average
monthly consumption for rural and urban areas is 50.4 kWh and
114.7 kWh, respectively. The corresponding average daily maximum
demand is 0.29 kW for rural households and 0.64 kW for urban
households. We cross-check our estimates with those of the global
review of rural electriﬁcation programmes performed byZomers
(2001)and those of theIEA (2011). Zomers arrived at an average
monthly consumption of 81.75 kWh and an average maximum demand
of 0.4 kW. Zomers' estimate may be slightly higher because it is global,
with only a few observations from Sub-Saharan Africa. Our estimates
are somewhat higher than those of theIEA (2011), which assumes an
annual consumption of 250 kWh in rural households and 500 kWh in
urban households in theﬁrst year of connection, gradually rising to
the national average. This assumption is based on an a priori judgement
of what is needed per day to cover basic needs (IEA, 2011, Box 13.1). Our
estimates of 50.4 and 114.7 kWh correspond to 604 and 1376 kWh per
year, respectively.
Thenumber of new connectionsdepends on the targets for increas-
ing households' access to electricity. The targets are considered exog-
enous variables and discussed below. Targets must be compared tocurrent access rates. We develop country-speciﬁc rates of current
access for urban and rural areas. In total for Sub-Saharan Africa, we
ﬁnd a current access rate of 16% in rural areas and 65% in urban
areas, resulting in a weighted total of 34%. The main source for this
data is the World Bank World Development Indicators, based on the
United Nations Statistical Ofﬁ	
ce Annual Energy Questionnaire. For
countries for which the data were missing, we used other sources,
including theIEA (2006)andECOWAS (2006). SeeRosnes and
Vennemo (2009)for details and individual country estimates.
4. Supply side data
On the supply side, we collect data for variable costs, unit invest-
ment cost and potential capacity levels for generation technologies,
transmission and distribution.
4.1. Generation technologies
The starting point is the set of existing generation assets in 2005,
characterised by the generation capacity and variable costs (e.g.,
fuel, operation and maintenance). We distinguish the assets that
will need refurbishment before 2015, either because they need refur-
bishment by the benchmark year 2005 (due to war or previous mis-
management) or because they will need it in the years 2005–2015.
If no speciﬁc information about the current state is available, we
assume that the refurbishment needs are based on the asset age.
Refurbishment can extend the lifespan of existing capacity at a cost
considerably lower than that of greenﬁeld investments.
We also assess the potential for greenﬁeld investments for each
country. It is useful to examine the main generation technologies and
options in turn.
4.1.1. Hydropower
Hydropower costs are generally site-speciﬁc. We build our esti-
mate of hydropower cost and capacity from the bottom up, going
through project information country by country. In total, 227 hydro-
power projects in Sub-Saharan Africa were examined for their invest-
ment cost, capacity,ﬁrm energy production, age and, when available,
refurbishment cost. The list of projects is not exhaustive, but a subset
where good data could be obtained. We excluded projects smaller
than 10 MW.
Hydropower plant information for the EAPP, SAPP and WAPP
regions was mostly derived from the regional master plans prepared
by the WAPP and SAPP power pools, the East African Union and the
Nile Basin Initiative through the Nile Equatorial Lakes Subsidiary
Action Programme (NELSAP) and the Eastern Nile Subsidiary Action
Programme (ENSAP). Plant information for the CAPP region is mostly
derived from various internet sources and is therefore not as com-
plete and coherent as the data for the other regions.
In this model, we use one estimate of the hydro unit investment
cost per country. In the case that there are several potential hydro-
power projects, the unit investment cost is the weighted average of
the investment costs of the potential projects in the country, and
the weights reﬂect the planned capacity of the plants. Where riparian
countries share the water rights to a project, the cost and potential of
the project are allocated between the partners. The unit investment
cost estimate varies between countries, and most fall in the range of
1000–2000 USD/kW.
Refurbishment normally requires the replacement of runners,
rewinding of generators and partial replacement of transformers, as
well as the replacement and upgrade of switchgear, control and pro-
tection equipment. An increase in production capacity can be achieved
by using more modern equipment. Heavy siltation can reduce the
lifespan of dams considerably unless sufﬁcient desiltation is provided.
Several of the countries incorporated in this study have undergone
civil wars; consequently, many hydropower stations have been more	
Table 1
Econometric model for market demand: estimation results.
Regressor Coefﬁcient Standard errorZP>z
Income 1.071 0.056 19.08 0.000
Agriculture−0.255 0.047−5.42 0.000
Manufacturing 0.127 0.053 2.38 0.017
Construction−0.074 0.032−2.34 0.019
Constant−9.415 0.488−19.28 0.000
Variables in logarithms. Manufacturing includes mining sector. Residuals assumed
autoregressive and panel correlated.1321 O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

or less destroyed, requiring additional funds for reconstruction.
Other countries have been so strapped for funds that normal main-
tenance has not been performed. Such negligence leads to the
rapid deterioration of the installation, which, in turn, increases
the repair cost ten-fold. Still, refurbishment is usually much more
economical than a new installation, and typically costs approxi-
mately 300 USD/kW.
4.1.2. Thermal power plants
In contrast to hydropower, thermal power plants can be built
almost anywhere. Nevertheless, the availability of fuels limits the
feasibility of certain investments. Therefore, we allow for investments
in natural gas-ﬁred power plants in countries that have domestic gas
reserves or that are (or will be) connected to a gas pipeline (such as
the West Africa Gas Pipeline). Investments in coal-ﬁred power plants
are reserved for countries that have domestic coal resources. Oil
products (e.g., diesel and heavy fuel oil) are assumed to be available
in most countries. No new nuclear power is allowed, but the existing
nuclear capacity in South Africa continues the operation.
Thermal power plant technologies are generic. The unit cost of
investment is therefore similar across countries. Due to market pres-
sures, costs increased in recent years until the global economic down-
turn of 2008–2009. We use unit cost estimates that are in line with
the accepted international values from several sources (seeRosnes
and Vennemo, 2009): 1100 USD/kW for coal, 670 USD/kW for natural
gas and 335 USD/kW for small diesel units. Refurbishment costs are
assumed to be 15% of the investment costs, with the exception of
coal, for which the costs are relatively higher at 272–300 USD/kW
(based on estimates in the literature and a detailed analysis of South
African coal plants).
The cost of operating thermal power plants depends on the fuel
cost. For coal, a world market price of 52 USD/tonne is our bench-
mark, based on a projected price for 2015 fromEIA (2006). The prices
of diesel and heavy fuel oil (HFO) are related to the world market
price of crude oil. The price of oil has been highly volatile in recent
years: six years prior to 2005, it was 10 USD/barrel; since 2006, it
has gone from 35 USD to 150 USD, then back to 35 USD/barrel
again, then up to 125 USD…Our scenarios are based on a price of
47 USD/barrel. The sensitivity in terms of oil price assumptions is dis-
cussed inSection 5.5. The world market price is also adjusted for
transport costs. For the EAPP and SAPP regions, we use country-
speciﬁc data provided byMetschies (2005). For the CAPP and WAPP
regions, we use data provided byArthur Energy Advisors (2007).
Estimating the price of natural gas is particularly challenging be-
cause the price and availability of gas is highly dependent on the
infrastructure, such as pipelines. We assume that for countries that
have the potential to export gas (e.g., Nigeria), the price of gas is
related to the world market price of oil (i.e., gas has an alternative
value). For countries that have no infrastructure to export gas (e.g.,
Mauritania), the price of gas is related to the domestic production
costs. We adjust the price of gas delivered to the power plant to
reﬂect transportation costs; countries with domestic resources are
assumed to have a lower gas price than countries that import gas
through a pipeline (e.g., Benin, Togo and Ghana).
4.2. Transmission and distribution (T&D)
We distinguish between domestic transmission (lines with volt-
ages of 66 kilovolts (kV) or greater), domestic distribution and inter-
national transmission lines.
4.2.1. Domestic
For domestic T&D infrastructure, we assume that the stock will
grow at half the rate of market demand, an assumption commonly
used by regulatory authorities. In addition, we assume that over the
10-year period, the original asset stock will depreciate at 3% peryear.	10Combined with annual variable costs of 2%, the relevant an-
nual investment to maintain capacity over the 10-year period is 5%
of the original T&D stock.
The refurbishment requirements are based on the asset age: lines
older than 30 years are assumed to need refurbishment at a unit cost
corresponding to 60% of the replacement value. In several countries,
the transmission systems are quite old. For example, in Niger, 96% of
the system is more than 30 years old; in Zimbabwe, 85%; in Zambia,
80%; and in the DRC, 61%. Distribution systems are younger on
average.
4.2.2. International
Transmission investments are determined endogenously in the
model. The total cost of a cross-country line depends on the unit
cost per km per MW and the line length (in km). The unit costs, per
km and per MW, of investing in transmission lines have been provid-
ed for each country. For cross-country lines, we have used weighted
unit costs for which the weights have been based on the approximate
share of the line built in the two countries. The line length depends
on the distance between load centres in the two countries or between
a load centre and a possible power plant location (where locations
are known). For instance, the relevant distance for a line between
Mozambique and South Africa is the distance from Cahora Bassa
North to Johannesburg.
4.3. Supply side assumptions related to new connections
We assume that urban households connect to the grid, while 70%
of new rural connections are covered by grid extensions;Deichmann
et al. (2010)reach a similar conclusion. The rest are supplied by off-
grid mini-hydro (20%), diesel (5%) or solar PV (5%).
4.3.1. Connection cost in urban areas
The connection cost in urban areas consists of strengthening and
extending the low- and medium-voltage distribution lines as well as
the lines to dwellings with the installation of credit meters, etc.
USAID (2004)andGaunt (2005)provide estimates of the costs of
the urban electriﬁcation programmes in South Africa, which are
approximately 420 USD and 350 USD per connection, respectively.
The cost of urban electriﬁcation is, to a large extent, generic across
countries, but other countries lack South Africa's economies of scale.
We use a slightly higherﬁgure of 500 USD per connection for the
remaining countries.
Once urban households connect to the grid, they send an impulse
to generate more power. The cost of producing power delivered
through the grid is calculated by the generation module of the model.
4.3.2. Connection cost in rural areas
In rural areas the cost ofgrid extensionis similar in type but more
expensive than the cost for urban connections. Often, one requires the
installation of low- and medium-voltage sub-transmission lines (for
example, 11 kV and 33 kV, respectively). The cost per connection
equals the contribution to demand for sub-transmission lines and a
distribution network in addition to the actual connection cost.
The connection cost ofmini-hydroequals that of grid extension.
The generation costs of mini and micro-hydropower stations are in-
cluded in the connection cost estimate.
The starting point of our estimates of cost per connection to the
grid extension and mini/micro hydropower is the extensive global
review of rural electriﬁcation projects and the studies byZomers
(2001), who arrives at an estimated grid extension cost of nearly
1500 USD per connection (distribution cost of 1200 USD, transmis-
sion cost of 125 USD, and service connection of 100 USD). These	
103 % per year implies that 40 % of the asset is functioning after 30 years. 1322O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

data correspond with the estimates from Tanzania byEcon Pöyry
(2006), which provides detailed estimates of costs per connection
for several districts in 19 different regions in Tanzania, each with dif-
ferent access rates and population densities. The resulting average
unit cost per connection in the 19 different regions lies between
550 and 4300 USD. The average unit cost across all regions is 1547
USD. Studies from Zambia (Sanghvi et al., 2005) and Mozambique
(Bergman and Davies, 2005) estimate a cost per connection at ap-
proximately 2000 USD. Based on Econ Pöyry's work and theﬁndings
of Zomers, it is assumed that large-scale electriﬁcation programmes
would converge around a cost of approximately 1550 USD in most
countries. The main exceptions are South Africa, which requires 350
USD even for rural connections, and the island states (2500 USD).
Moving on to off-grid alternatives,dieselgeneration involves a
generator accompanied by a distribution network, generally with no
need for transmission or sub-transmission lines. Thus, the relevant
connection cost is simply the unit distribution and service connection
costs. Here,Zomers' (2001)cost of 1325 USD per connection is used.
The operation of diesel generation requires fuel and capital. The cap-
ital investment costs for small diesel units are derived from expert
judgement and cross-checked for this study against data from pro-
ducers of mini-diesel plants, such as Jacobsen, Caterpillar and others.
The unit capital cost used in the model is 335 USD/kW.
Thesolar PVmarkets in most Sub-Saharan African countries have
only recently been developed. As a result, prices have fallen consider-
ably in most countries in recent years.CORE (2003)reports the
results of Afrepren, which arrives at an estimated average price of
1100 USD for a 50 Watt-peak (Wp) system for the SAPP countries.
However, the work performed byEcon Pöyry (2006)in Tanzania
and Zambia, where the markets are undeveloped, revealed prices of
650–800 USD. Other more recent sources from South Africa, Ethiopia,
Sudan, Kenya and Uganda reveal similar prices. Thus, country-speciﬁc
cost data are applied for those countries for which it is available,
while a conservative estimate of 800 USD is applied to the remaining
countries.
5. Results: investment needs and costs
Having presented our methodology and data, we turn to the re-
sults of our analysis: what are the investment needs of Sub-Saharan
Africa over this 10-year period and what are the costs of these invest-
ments? Investment needs and costs depend not only on the model
and its data but also on the exogenous variables. Important exoge-
nous variables include country-speciﬁc economic growth, targets for
new connections and assumptions about the electricity trade. We
introduce a benchmark scenario for these variables. Afterwards, we
bring in the impact of alternative assumptions.
5.1. Benchmark scenario: Trade and national targets for new connections
The benchmark scenario is characterised by the following exoge-
nous assumptions:
•Best guess for economic growth (which is an important driver of
market demand for electricity, seeTable 1). National annual growth
rates range between 2% and 9%. Projections of economic growth
from 2005–2015 are compiled from theWorld Bank (2009).
•Targets for household access to electricity in urban and rural areas as
deﬁned by the stated national policy targets in each country. On aver-
age, the targets imply that access rates are lifted to 67–100%, depending
on the region. This translates to 62 million new connections.	
11	
•Fully integrated regional power pools. At present, the level of integra-
tion and cooperation varies between the different regions. We allow
further integration within the four power pools to seek out cost
savings when electricity is supplied in the most cost-efﬁcient way.
5.1.1. Demand
With these assumptions, the demand for electricity at the end of
the 10-year programme (2015) is 70% higher than before the pro-
gramme starts (Table 2). Three quarters of the additional demand is
associated with economic growth (market demand), and one quarter
is associated with new connections. The Southern Africa Power Pool is
the largest region of demand and the source of half of the increase
over time. However, the relative weight of Southern Africa declines
over time. The Western Africa Power Pool has the highest percentage
of growth, more than 200%. One reason for this is that suppressed
demand is particularly high in the WAPP region.
5.1.2. Investment needs
By model simulation, we determine the need for capacity addition
to meet demand in this scenario to be approximately 80 GW in Sub-
Saharan Africa. This entails almost a doubling of its current capacity,
which, for the whole of Sub-Saharan Africa, was 87 GW in 2005. In
addition, almost 40 GW of today's capacity will be refurbished. Al-
most half of the total available capacity at the end of the investment
programme is hydropower.
5.1.3. Investment costs
We estimate that the investment cost of the expansion pro-
gramme comes to approximately 215 billion USD (Table 3). Only
approximately half of this amount is related to generation. To carry
the new generation to market, substantial amounts of new transmis-
sion and distribution lines are required, as well as new connection
points. The cost of these items makes up the other half of the total.
Although refurbished capacity is one third of the total in physical
terms, the cost of refurbishing generation capacity is only one tenth	
11Assumingﬁve individuals per household, asRosnes and Vennemo (2009)andIEA
(2011)assumed, 62 million connections provides electricity to 310 million people. Be-
cause of population growth, however, we cannot conclude that the number of people
without electricity is cut from 600 to 300 million.	
Table 2
Demand for electricity before and after 10-year programme (TWh).
Region Total net
demand
2005Market demand
and suppressed
demand 2015Demand of new
connections
2015Total net
demand
2015
SAPP 259 383 14 397
EAPP 101 145 24 169
WAPP 31 70 25 94
CAPP 11 17 3 20
Island states 1 2 2 3
Total Sub-Saharan Africa 403 616 68 683
Note: Total gross demand is larger than net demand because it includes transmission
and distribution losses.
Table 3
Investment costs in Sub-Saharan Africa over 10 years (billion USD).
SAPP EAPP WAPP CAPP Islands Total
Generation
Investment cost, new capacity 32.2 32.7 27.0 6.6 0.4 99.0
Refurbishment cost 7.6 0.4 1.5 0.3 0.0 9.8
T&D and new connections
Investment cost, new capacity 23.7 27.4 29.8 2.3 1.3 84.5
–Cross-border transmission lines 3.1 1.0 0.9 0.3 0.0 5.3
–Distribution grid 12.7 3.1 11.9 0.3 0.1 28.1
–Connection cost (urban) 4.0 5.7 7.6 1.0 0.4 18.8
–Connection cost (rural) 4.0 17.6 9.3 0.7 0.7 32.4
Refurbishment cost 9.8 3.3 6.1 0.2 0.1 19.5
Total 73.3 63.8 64.4 9.5 1.8 212.71323 O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

of the cost of new capacity. This shows that the unit cost of refurbish-
ment is signiﬁcantly lower than the unit cost of new capacity.
It is interesting to note the differences between the regions. The
EAPP stands out due to its large expenses on connections in rural
areas and low expenses on the distribution grid. One reason for this
behaviour is that the countries of the EAPP have particularly high
access targets and aim for 14 million new connections in rural areas.
The WAPP stands out for having high expenses overall. The invest-
ment cost in the WAPP region is almost on par with the EAPP and
comparable to the SAPP, even though the WAPP region is a much
smaller power pool. However, we previously observed that capacity
expansion in the WAPP is similar to the EAPP. It adds to investment
costs that several of the main hydropower projects in the WAPP are
relatively expensive to build. We also note that the investment cost
in the CAPP is much smaller than in the other power pools, which is
due to the fact that the CAPP is a much smaller pool and the estimated
capacity additions are also smaller.
5.1.4. Annualised cost of system expansion and operation
A description of investment costs does not capture variable costs
of fuel, maintenance and operation associated with generating elec-
tricity. Moreover, the economic signiﬁcance of the investment cost
depends on the lifetime of the investment asset and the discount
rate of the capital.	
12To incorporate both the variable costs and the
annual equivalent of the capital costs, we introduce the concept of
theannualised cost of system expansion and operation. The annualised
cost of system expansion and operation is the entity that our model
seeks to minimise by combining transmission, distribution and gener-
ation technologies. The system in question consists of generation,
transmission and distribution assets that are refurbished or added
over the course of the investment period, plus the assets operating
at the initiation of the investment period. Hence, the annualised cost
of system expansion and operation consists of the following three
elements:
Annualised cost of system expansion and operation = Annualised
capital cost of new and refurbished assets + variable cost of new and
refurbished assets + variable costs of pre-existing assets not needing
refurbishment.	
13	
Note that the annualised cost of expansion and operation refers to
a cost that continues for as long as one operates the system. Hypo-
thetically, if the system is operated into eternity, the cost will stay
on into eternity as well.
We estimate that the annualised cost of system expansion and op-
eration associated with the investment programme of this scenario is
approximately 48 billion USD (Table 4). However, almost 12 billion
USD is used for operating pre-existing assets that do not need refur-
bishment. Hence, the relevant entity for assessing theadditional
strain on the economy of Sub-Saharan Africa is 36 billion USD. 80%
of this additional strain is comprised of annualised capital cost.
There are interesting differences between the regions. Variable
costs are particularly low compared to the annualised capital costs
in the WAPP and CAPP regions; speciﬁcally, it is the variable cost of
generation that is low. Thisﬁnding indicates that hydropower is par-
ticularly important for generation in the WAPP and CAPP regions. A
cross-check with the generation mix conﬁrms theﬁnding. In the
WAPP, the share of hydropower in the least-cost mix is 77%, and in
the CAPP, it is as high as 97%. In the EAPP, the variable costs of run-
ning the transmission and distribution system are particularly high.The transmission and distribution asset stock is large in this region,
due to the large distances within and between countries.
5.2. What will targets for electricity access cost?
Recall that the benchmark scenario assumes 62 million new
household connections. How much of the cost is derived from this
target? We contrast the benchmark scenario of nationally derived
access targets with theconstant access scenario. Together, the two
scenarios span a range for access developments in Sub-Saharan Africa
over the next few years.
The constant access scenario implies that access rates are kept
constant at the 2005 levels. Due to population growth, the constant
access scenario does imply 17 million new connections, mostly in
urban areas, and 85 million more people are given access to electric-
ity. Still, we consider it a scenario of low ambition.
The constant access scenario carries an investment cost of approx-
imately 160 billion USD (Table 5), which is roughly 25% lower than
the benchmark scenario. Based on this result, we believe that the
main cost of providing electricity to Africa doesnotdepend on provid-
ing electricity access to new customers. The main cost depends on
meeting the market demand associated with the economic growth.
12The annuity formula we use isannualised capital cost = investment cost * r(1−1
(1 +r)	−T)−1. Based on professional judgement, we use a real discount rate of 12
% p.a. The life-time T differs between assets. We assume a life-time of 40 years for
hydropower plants, 30 years for coal and 25 years for natural gas.	
13By“not needing refurbishment”we mean no refurbishment during the 10-year period
studied.	
Table 4
Annualised costs of system expansion and operation in Sub-Saharan Africa, benchmark
scenario (billion USD).
SAPP EAPP WAPP CAPP Islands Total
Generation
Investment cost 4.5 4.4 3.5 0.9 0.1 13.4
Refurbishment cost 1.3 0.1 0.3 0.0 0.0 1.7
Variable cost (fuel, O&M) 5.3 6.0 2.7 0.1 0.3 14.4
–New capacity 2.1 5.4 0.3 0.1 0.3 8.1
–Existing capacity 3.2 0.7 2.5 0.0 0.0 6.3
T&D and new connections
Investment cost 2.9 3.3 3.7 0.3 0.2 10.4
–Cross-border 0.4 0.1 0.1 0.0 0.0 0.7
–Distribution grid 1.6 0.4 1.5 0.0 0.0 3.5
–Urban connection 0.5 0.7 0.9 0.1 0.1 2.3
–Rural connection 0.5 2.1 1.2 0.1 0.1 4.0
Refurbishment cost 1.2 0.4 0.8 0.0 0.0 2.4
Variable cost (existing capacity) 3.1 0.8 1.3 0.1 0.0 5.3
Total
Capital cost 10.0 8.2 8.2 1.2 0.2 28.0
–Investment cost 7.5 7.7 7.2 1.2 0.2 23.8
–Refurbishment cost 2.6 0.5 1.0 0.1 0.0 4.1
Variable cost 8.4 6.8 4.0 0.2 0.3 19.7
Total 18.4 15.0 12.3 1.4 0.6 47.6
–of which existing assets 6.3 1.5 3.8 0.1 0.0 11.6
–of which new and refurbished assets 12.1 13.5 8.5 1.3 0.6 36.0
Table 5
Investment costs in Sub-Saharan Africa over 10 years in different scenarios (billion USD).
Benchmark Constant
access rateTrade
stagnationLow
growth
Generation
Investment cost 99.0 87.0 85.5 81.2
Refurbishment cost 9.8 9.6 9.8 9.6
T&D and new connections
Investment cost 84.5 42.6 79.2 69.6
–Cross-border transmission lines 5.3 5.7 0.0 5.3
–Distribution grid 28.1 28.1 28.1 13.2
–Connection cost (urban) 18.8 7.4 18.8 18.8
–Connection cost (rural) 32.4 1.4 32.4 32.4
Refurbishment cost 19.5 19.5 19.5 19.5
Total 212.7 158.7 194.0 180.7 1324O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

We also examined the total annualised cost of the system expan-
sion and operation; even this variable differs by approximately 25%
from the national targets scenario.
5.3. The beneﬁts of regional integration
In the benchmark scenario, we have allowed for imports and ex-
ports of electricity within a power pool whenever it is economically
beneﬁcial. It was determined that it is cost-efﬁcient for all power
pools to make extensive use of trading possibilities. In the SAPP, for
example, the DRC emerges as a huge exporter of electricity.	
14	The
DRC's net export is almost four times larger than its domestic con-
sumption, and it supplies hydropower through Zambia, Zimbabwe
and Botswana into South Africa. South Africa becomes an importer
of 10% of its domestic consumption, trading with all neighbouring
countries: Namibia, Botswana, Zimbabwe, Mozambique and Lesotho
(Fig. 1).
A main supplier nation emerges in each of the power pools. The
WAPP is, to a large extent, supplied by Guinea; the CAPP supply is
driven by Cameroon; and the EAPP supply is driven by Ethiopia,
followed by Sudan. These few countries, led by the DRC, are the
main supply bases in regionally integrated power pools.
In reality, of course, trade in such an essential commodity as elec-
tricity requires signiﬁcant cooperation and coordination among coun-
tries. Hence, achieving optimal power trade will require a range of
important political, legal and economic commitments. It is beyond
the scope of this paper to discuss the nature of the commitments
needed to secure trade. This paper focuses on the beneﬁts of trade,
which is represented by the cost difference between trade expansion
and trade stagnation.
Thetrade stagnation scenarioassumes that no new interconnectors
between countries are built in any power pool. The existing inter-
connectors can be used, and tradeﬂows and volumes are determined
endogenously. Hence, it is the opposite extreme to the benchmarkscenario of trade expansion and serves to illustrate the range of pos-
sibilities in terms of trade.
A vivid illustration of the difference between the two scenarios
is given inFig. 1. The trade stagnation scenario obviously foregoes ex-
tensive trading opportunities and forces countries to cover their elec-
tricity demand using their own means.
From this observation, the beneﬁts of trade emerge. It is important
to realise that trade must be viewed as an investment. Weﬁnd that
it requires approximately 19 billion USD in additional investment
costs (seeTable 5). First, trading requires an investment in inter-
connectors, which amounts to approximately 5 billion USD. Second,
trading allows hydropower to substitute for thermal power; this is
more economical overall, but the investment cost is higher. The
total generation investment is 14 billion USD higher in the trade
expansion scenario.
5.3.1. Gains from trade
What is the return from this investment? Weﬁnd that annual sav-
ings in variable cost amount to 5 billion USD. It is clear that 5 billion
USD is quite a good return on an investment of 19 billion USD. The
proﬁtability of an investment is often indicated by the internal rate
of return. The internal rate of return depends on market conditions,
such as prices, and the depreciation proﬁle of the investment. As
a benchmark, assume an eternal lifetime of the investment and
unchanged market conditions. In this case,ﬁve billion becomes a
constant annual stream, and the internal rate of return is 27%. More
cautiously, one could deduct 5% as depreciation and maintenance
and claim a rate of return of 22%, which would still remain a good
investment.
The gains from trade differ between power pools. In the SAPP, the
additional investment is repaid in less than a year, and the internal
rate of return is 168%. In this region, the required additions to trans-
mission lines are low, and hydropower investments (mostly in DRC)
have relatively low unit costs. Simultaneously, there are ample
opportunities for hydropower to replace coal-ﬁred power in South
Africa. In the other regions, the benchmark internal rate of return
varies between 20% and 33% (Table 6). By far, the largest reshufﬂing
of generation occurs in the EAPP. Trade allows the EAPP to utilise	
14Due to its large hydropower resources and central position, the DRC could, in the
future, become supplier to the EAPP and CAPP, as well as the SAPP.	
Fig. 1.Tradeﬂows with trade expansion (left) and trade stagnation (right) in the SAPP after 10-year programme (TWh). Note: In the trade stagnation scenarios, the capacity on
interconnectors equals the 2005 actuals, but the direction of tradeﬂows may differ.1325 O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

hydropower resources in the Nile Basin, mostly in Ethiopia and
Sudan. These resources replace the natural gas-ﬁred generation in
Kenya, Egypt and elsewhere. However, the demands on investments
are also particularly heavy in this power pool, and the internal rate
of return is lower than in the other pools.
It is perhaps expected that trade leads to savings in the variable
costs of new capacity. However, savings also apply to the existing
capacity. This means that some existing capacity is so inefﬁcient
that the variable and capital cost of new capacity combined is lower
than the variable cost of the existing capacity. In particular, this is
the case in the EAPP, where the import of hydropower is replacing
the thermal capacity in Egypt.
5.3.2. Environmental beneﬁts
In addition to economic gains, the electricity trade brings impor-
tant environmental beneﬁts to the table. We expect an annual saving
of 70 million tonnes of CO	
2. Of this, approximately 40 million tonnes
will be saved when coal power in South Africa is replaced by hydro-
power from the DRC. Approximately 20 million tonnes will be saved
when natural gas in Egypt and Kenya is replaced by hydropower
from Ethiopia and Sudan. The rest is divided between a number of
sources.
To put the estimate of 70 million tonnes into perspective, the IEA has
estimated emissions from power production in Africa to be 405 million
tonnes in 2009 (IEA, 2011). The savings from trade are equivalent to
17% of this volume. The numbers do not include greenhouse gas emis-
sions from hydropower in the form of methane from dams, which
could modify the estimate.
5.4. The role of economic growth
Although several countries of Sub-Saharan Africa have experi-
enced healthy economic growth in recent years, the per capita growth
forecasts are low for some countries (e.g., 0.2% for Guinea-Bissau ; 0.3%
for Niger), and it is of interest to analyse the implications of lower
economic growth for electricity supply costs. For simplicity, we as-
sume that the annual economic growth per capita is cut to half in
each country in the continent.	
15	
Because of the positive relationship between electricity demand
and economic growth, it must be expected that the investment re-
quirement for electricity is reduced when growth is lower.	
16Market
demand falls, but because the income elasticity is greater than one,
the market demand falls less than economic growth. In addition, the
suppressed demand and new connections are unaffected.
The net result of the experiment is a decrease in the additional
capacity (investments plus refurbishments) from 119 to 102 GW
and a decrease in the investment requirement from 215 to 180 billion
USD (Table 5). Of course, the lower investment requirement is noeasier toﬁnance by domestic sources because the economies that
are to shoulder the investment are smaller in this scenario. Somewhat
contrary to our expectations, the share of hydropower ishigherin this
scenario than in the baseline. In other words, the hydropower invest-
ments are, to a large degree, sheltered from lower economic growth.
Sufﬁcient trading opportunities are available, even in a 15% smaller
market, to carry the economies of scale required by hydro. This result
powerfully supports the development of hydropower in Sub-Saharan
Africa, despite uncertain market conditions.
5.5. The impact of oil prices
Few variables are surrounded with as much uncertainty as future
oil prices. When considering a long-term investment in power pro-
duction, it is the long-term average oil price that matters. In the sim-
ulations to test for the sensitivity of oil prices, we applied a crude oil
price in the range of 25–75 USD/barrel. The consequences of an even
higher oil price, for example, 125 USD/barrel, follow by extrapolation.
We have used historical data to estimate the correlation between the
crude oil price and the prices of other oil products and natural gases.
Sensitivities show that a higher oil price leads to a higher share of
hydropower in the generation mix, a higher total investment, and
higher annual costs of electricity supply. There are two mechanisms
behind this result. First, higher oil prices cause investments in ther-
mal power plants to be less attractive. Second, higher oil prices also
cause the operation of existing gas-, oil- and diesel-fuelled power
plants to be more expensive, thus contributing to the faster replace-
ment of existing plants. Because hydropower has substantially higher
investment costs than thermal power plants, the up-front investment
cost is higher.
As hydropower resources are developed, however, the potential
for replacing fossil fuels with hydropower is depleted. Further in-
creases in oil prices will therefore have a diminishing impact on
investments while the annual costs continue to increase.
Note that oil-based fuels (e.g., diesel and HFO) make up only a tiny
share of the total electricity production capacity in Sub-Saharan
Africa. The main technologies are hydropower and coal-ﬁred power,
counting for approximately one-half and one-third of total produc-
tion capacity in 2015, respectively. Natural gas is important in several
regions (i.e., Eastern and Western Africa), but we have assumed that
gas prices are related to oil prices only in countries that export gas.
Oil-related fuels have only 3% of the total production capacity and
less than 2% of the production in Sub-Saharan Africa in 2015. However,
these fuels can be important in isolated regions.
6. Comparison with the literature
As we mentioned in the introduction, there are already several
projections of the cost of electricity provision in Sub-Saharan Africa.
When comparing these projections to each other and to our work,
one should note the following:
•Demand assumptions: Does the projection cover the market demand,
new connections or both?
•Transmission and distribution assumptions: Does the projection
cover transmission, distribution or both?
•Trade assumptions: Is trade based on optimisation or extrapolation
of today's trends? Are investments in international transmission
lines based on optimisation or on existing plans?
•Unit cost assumptions: How are unit costs of generation investments
speciﬁed?
Other assumptions, such as fuel prices, are also important.
The IEA regularly publishes electricity demand forecasts for Africa
as part of its World Energy Outlook. The main tool for these forecasts
is the World Energy Model. Like our model, the World Energy Model
considers economic activity and population as exogenous factors and	
15This assumption was also chosen to imitate the impact of a shorter investment
window than ten years at the original pace of growth.
16Economic growth can be a driver of electricity demand as well as a response to
electricity demand. In our study, we have not discussed this issue, but we did note
the historical correlation between the two as represented in our econometric demand
equation.	
Table 6
Savings and costs when there is trade (billion USD).
Additional
investment costSavings in
variable costInternal rate of return
(savings over investment)
SAPP 0.6 1.1 168%
EAPP 15.1 2.9 19%
WAPP 2.1 0.7 34%
CAPP 0.9 0.3 28%
Total 18.7 5.0 27%
Note: There are no gains from trade in island states because they do not trade electricity. 1326O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

derives demand on that basis. Supply depends on technology, and the
whole energy sector is observed in context. TheIEA (2011)concludes
that 69 GW of power generating capacity will be added in the whole
of Africa over the period of 2009–2020, which is approximately 85% of
our estimate. Hence, the IEA's capacity projection for Africa is lower
than ours. The IEA builds its demand forecasts by sector but states
that economic growth is by far the most important determinant in
demand. The main difference with our demand analysis seems to be
the aggregate impact of economic growth on the electricity demand
in Africa, i.e., the income elasticity. The IEA assumes an annual eco-
nomic growth in Africa of 4.6% for the period of 2009–2020, but its
electricity demand only increases 3.1% annually. Hence, its aggregate
income elasticity in terms of the demand per capita is only approxi-
mately 0.7, while it is 1.1 in our analysis.
Another determinant of demand is household access rates. TheIEA
(2011)assumes that access rates increase from 31% in 2009 to 35% in
2015 and to 56% in 2030. In our national targets scenario, we assume
a similar increase (to 59%) by 2015. In addition, we assume higher
consumption per new connection, as discussed inSection 3.2. This
also contributes to the lower IEA demand projections for 2015.
On the supply side, the IEAﬁnds, as we did, that the investment
cost of transmission and distribution is about as large as the cost of
generation, approximately 45% of the total investment. However,
the unit investment cost is 50% higher in the IEA analysis, which is
approximately 3 USD million per MW in total compared to an average
of roughly 2 million USD per MW across the technologies and coun-
tries included in our analysis. The reason is not that the IEA requires
a higher return to capital; on the contrary, its return requirement
seems to be lower.	
17Instead, the reason may simply be that the IEA
uses PPP dollars, whereas we have used nominal dollars based on
actual cost estimates.
18In addition, we include a signiﬁcant amount
of refurbishment at a low unit cost.
TheWorld Bank (2006)estimates the cost of increasing access,
corresponding to new connections in our terminology. Theyﬁnd
that a total of 280 billion USD is required in the generation, transmis-
sion and distribution of electricity until 2030 to provide electricity to
100% of the population (i.e., 200 million new connections). While
several of the assumptions regarding new connections and consump-
tion per new connection are similar to those made in this study, the
World Bank (2006)takes a more generalised approach to estimating
costs. TheWorld Bank (2006)assumes that 75% of all households
connected by 2030 would be through grid connections. The low volt-
age distribution cost is assumed to be 500 USD per household until
35% of households are connected, 800 USD between 35% and 50%,
1000 USD between 50% and 75%, and 1500 USD thereafter. The off-
grid connection cost is assumed to be 800 USD per household. Losses
in generation and transmission are assumed to amount to 1.5 billion
USD per GW. Despite differences in methodology, the end result of
theWorld Bank (2006)exercise is comparable to our estimate of
the marginal cost of new access connections. The 280-billion USD in-
vestment to obtain 200 million new connections averages 1400 USD
per connection. Our similar estimate is 1200 USD per connection.	
19	
Gnansounou et al. (2007)estimate the cost of the increasing elec-
tricity supply in the WAPP and study the cost advantages of the electric-
ity trade. Their analysis is model-based, and the model is presented as
a“least-cost techno-economic model”. Only grid-based electricity de-
mand is considered, which emphasises the concept of market demand.
An important difference from our study is that the cost of transmission
and distribution investment in the WAPP only amounts to 3% of the
total cost. By contrast, weﬁnd transmission and distribution to be
approximately 50% of the total cost. From their presentation, it seems
thatGnansounou et al. (2007)ignore national transmission and distri-
bution, and their cost component for transmission and distribution is
referred to as the“cost of adaption and expansion of interconnection
lines”. This brings the numbers together because transmission makes
up only 3% of the investment cost in our data, as well (Table 3).
Although the method is similar in principle,Gnansounou et al.
(2007)ﬁnd far greater beneﬁts in the electricity trade in the WAPP
than we do. In their analysis, trade brings about a reduction in the
discounted total costs of electricity of no less than 38%. By compari-
son, weﬁnd a reduction in the annualised cost of system expansion
and operation of 5.1%. They claim that the electricity trade reduces
capital expenditures. Weﬁnd that electricity trade increases capital
expenditures due to more hydropower and more interconnectors.
The difference here could be due to the different assessments of
transmission and distribution costs.Gnansounou et al. (2007)com-
pare their results to earlier studies, includingSparrow et al. (1999)
andGraeber et al. (2005), which also estimate quite large beneﬁts
of the electricity trade.
Nexant (2007)is a regional generation and transmission expan-
sion plan that assesses power investment requirements in the SAPP
for the period of 2006–2015. The study is not based on an optimising
framework, but calculates the costs of speciﬁc additions to the gener-
ation capacity as well as the speciﬁc transmission reinforcements
needed to support the anticipated levels of trading (based on the in-
dividual TSO assessments). With this exception, the results of Nexant
are quite similar to those ofGnansounou et al. (2007). The cost of dis-
tribution is ignored, and the cost of transmission is found to be small
compared to the cost of generation. Trade leads to large (approxi-
mately 50%, 12 billion USD) savings in investment costs and small
(approximately 4%) savings in the variable cost. By contrast, weﬁnd
an increase in the investment cost of 0.6 billion USD in the SAPP
and large (approximately 20%) savings in the variable cost. At a meth-
odological level, the fact thatNexant (2007)does not require con-
strained optimality means that a non-optimal situation with trade is
compared to a non-optimal situation without trade.
7. Conclusions
Sub-Saharan Africa demands large amounts of electricity, both to
develop economically and to end the untenable situation in which
nearly 600 million of its people do not have access to electricity. Inter-
nationalﬁnance institutions and bilateral donors are ready to work
with the governments of Sub-Saharan Africa to increase the electrici-
ty supply, but nobody knows the cost of this endeavour. Our study
has found that a realistic estimate for the region over a 10-year period
is approximately 100–120 GW in terms of generation capacity invest-
ments and that the investment cost would amount to approximately
160
–215 billion USD. The estimate includes a signiﬁcant amount of
refurbishment, which is more economical than greenﬁeld invest-
ments. The cost of generation comprises half of the total, while trans-
mission and, in particular, distribution comprise the other half. The
annual cost of system expansion and operation, a concept that in-
cludes variable cost (e.g., fuel cost) and the annualised cost of capital,
amounts to 39–48 billion USD. Approximately 10–12 billion of this
amount refers to a cost-efﬁcient operation of the existing electricity
system. The lower end of the 39–48 billion range is associated with
moderate access targets. The higher end of the range is associated	
17TheIEA (2011)does not state its return requirement explicitly, but from the back-
ground material, it appears to be 5–10 %. An annual cost reduction (“learning curve”)is
also assumed. To us, 5–10 % seems quite low for African conditions and, as noted, we
use 12 %.
18Generally, the PPP rate of dollar against poorer countries’currencies is lower than
the nominal rate. This is to correct for a lower price level in the poorer countries. How-
ever, when equipment is imported to the poorer country, the use of PPP can be mis-
leading. Consider a country with a nominal exchange rate of 10:1 versus USD and a
PPP rate of only 1:1. If a machine that costs one USD is imported, it will cost 10 in local
currency. When translated back to USD using PPP, the price suddenly becomes 10 USD.
19We develop our estimate by comparing the constant access and benchmark sce-
narios. There are 62 million new connections in the benchmark scenario of national
targets, and 17 million new connections in the constant access scenario. The difference
is 45 million connections, which creates an investment cost difference of 55 billion
USD.1327 O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

with national access targets and trade stagnation. The range is similar
in size to current annual overseas development assistance to Sub-
Saharan Africa.
The range of investment is spanned by ambitions in terms of
providing electricity access to new households. The 160-billion USD
estimate is obtained if one maintains the current access rates. Most
of this cost is for satisfying the electricity demand from industry,
the public sector and afﬂuent households. The 215-billion USD esti-
mate is obtained if one assumes that national targets for electricity
access are met. Viewed as projections, the national targets may be
optimistic, and the current access assumption may be pessimistic.
Economic growth is an important driver of the market demand
from industry, the public sector and afﬂuent households. We estimate
an income (GDP) elasticity of 1.07. Still, cutting economic growth per
capita by one-half only reduces the investment requirement from 215
to 180 billion USD. This relatively small reduction is partly due to the
fact that the cost of new connections and eliminating suppressed
demand is independent of economic growth. Moreover, population
growth will, by itself, imply increased demand.
Even though we technically focus on the period until 2015, our es-
timates can be generalised to a later 10-year period, e.g., 2010–2020
or 2013–2023. On the demand side, there are two opposing develop-
ments. On one hand, the global economic recession starting in 2008
has reduced growth rates in Africa, although not as much as it has
in other parts of the world. On the other hand, the population growth
continues. On the supply side, many of the most advanced projects
are delayed, while others are still on the drawing table.
There are signiﬁcant economic beneﬁts to linking the power sys-
tems of different countries and encouraging trade. In contrast to ear-
lier studies, such asNexant (2007)andGnansounou et al. (2007),we
have determined that allowing electricity trade in Sub-Saharan Africa
will generate higher investment costs. The reason for our result is
partly that trade will require strengthening the transmission lines
and partly that trade will allow an expansion of capital intensive
hydropower production at the expense of thermal power production.
However, weﬁnd that the higher investment costs are counteracted
by the lower variable costs. Therefore, viewed as an investment,
trade in Sub-Saharan Africa carries an annual internal rate of return
of 20–30%. In the SAPP, the pay-back time is less than a year, and
the rate of return is above 100%. Trade also saves signiﬁcant amounts
of CO	
2, which is interesting in its own right. For example, NGOs have
sometimes claimed that trade is detrimental to the environment.
Here is an example for which that is clearly not the case, at least
for CO
2.
Rosnes and Vennemo (2009)report on a number of policy experi-
ments that we have not related here, including the higher investment
costs of critical technologies, CDM support to clean energy, and the
impact of climate change.
Although our work provides considerable detail on the cost of
electricity supply in Sub-Saharan Africa, there are aspects that
could be improved upon in later studies. Demand is not price elas-
tic; hence, there is no simultaneous determination of demand and
supply. We are not convinced that price is an important equilibrat-
ing factor in several countries, but the current dichotomy between
supply and demand is not satisfactory either. Another issue is that
the categorisation of countries into one of four power pools has a
degree of arbitrariness. In future work, we would like to analyse
the integrated trade across the continent on one hand and the in-
teraction between the physical possibilities for trading and political
constraints on the other.Acknowledgements
This work builds on a multi-year project funded by the World
Bank as part of the Africa Infrastructure Country Diagnostics (AICD)
study. Thanks to Halvor Aarrestad, Ryan Anderson, Sarah Carter,
Jonathan Hedgecock, Stephen Kennedy, Siobhan Murray, Sudeshna
Ray, Ståle Skrede and Fredrik Wilander for invaluable assistance
during various stages, as well as to the World Bank staff, led by Vivien
Foster, for constructive criticism and support. We also thank two
anonymous reviewers for valuable comments.
Appendix A. Supplementary data
Supplementary data to this article can be found online athttp://
dx.doi.org/10.1016/j.eneco.2012.06.008.
References	
Arthur Energy Advisors, 2007. WAPP Emergency Power Supply Plan to Curtail Power
Shortages Faced by ECOWAS Member States. Draft Study Report. (August 2007).
Bergman, S., Davies, I., 2005. Mainstreaming Low-Cost Innovations in Electricity Distribution
Networks. Study prepared for Energy Sector Management Assistance Program. World
Bank, Washington, DC.
Brooke, A., Kendrick, D., Meeraus, A., Raman, R., 1998. GAMS—A User's Guide. GAMS
Development Corporation.
CORE, 2003. Issues and Options for Rural Electriﬁcation in SAPP Member Countries and
Rural Electriﬁcation Planning in Lesotho. Unpublished.
Deichmann, U., Meisner, C., Murray, S., Wheeler, D., 2010. The economics of renewable
energy expansion in rural Sub-Saharan Africa. Energy Policy 39, 215–227.
Eberhard, A., Shkaratan, M., 2012. Powering Africa: meeting the reform and funding
challenges. Energy Policy 42, 9–18.
Eberhard, A., Rosnes, O., Shkaratan, M., Vennemo, H., 2011. Africa's Power Infrastruc-
ture. Improving investment, connectivity, reliability and efﬁciencies. The World
Bank Press, Washington, DC.
ECOWAS, 2006. White Paper for a regional policy geared towards increasing access to
energy services for rural and periurban populations in order to achieve the Millen-
nium Development Goals. ECOWAS/UEMOA.
EIA, 2006. Annual Energy Outlook.
Fay, M., Yepes, T., 2003. Investing in Infrastructure: What is Needed from 2000 to
2010? Policy Research Working Paper 3102. The World Bank, Washington, DC.
Gaunt, C.T., 2005. Meeting electriﬁcation's social objectives in South Africa, and impli-
cations for developing countries. Energy Policy 33, 1309–1317.
Gnansounou, E., Bayem, H., Bednyagin, D., Dong, J., 2007. Strategies for regional inte-
gration of electricity supply in West Africa. Energy Policy 37, 4142–4153.
Graeber, B., Spalding-Fecher, R., Gonah, B., 2005. Optimising transnational power gen-
eration and transmission investments: a South African example. Energy Policy 33,
2337–2349.
IEA, 2006. World Energy Outlook, 2006. International Energy Agency, Paris.
IEA, 2008. World Energy Outlook, 2008. International Energy Agency, Paris.
IEA, 2011. World Energy Outlook. International Energy Agency, Paris.
Khanna, M., Rao, N.D., 2009. Supply and demand of electricity in the developing world.
Ann. Rev. Resour. Econ. 1, 567–595.
Metschies, G., 2005. International Fuel Prices 2005, 4th edition. Deutsche Gesellschaft
für Technische Zusammenarbeit (GTZ) GmbH, Eschborn, Germany.
Nexant, 2007. SAPP Regional Generation and Transmission Expansion Plan Study. October.
One, 2012. G7. EU15 and DAC commitments to Sub-Saharan Africa. The Data Report 2011.
Econ, Pöyry, 2006. Tanesco Pre-Feasibility Grid Project. Report prepared for Tanzanian
Ministry of Energy and Mines.
Rosnes, O., Vennemo, H., 2009. Powering up. Costing power infrastructure investment
needs in Sub-Saharan Africa. Research Report 2009–034. Econ Pöyry, Oslo.
Sanghvi, A., et al., 2005. Rural electriﬁcation: costs of connection, willingness and abil-
ity to pay. Presentation at Consultative Workshop on Public Private Partnership in
Rural Electriﬁcation.
Sparrow, F.T., Masters, W.A., Yu, Z., Bowen, B.H., 1999. Modelling electricity trade in
Southern Africa. Second Year Final Report to the Southern African Power Pool.
Purdue University, USA.
USAID, 2004. Innovative Approaches to Slum Electriﬁcation. USAID, Washington, DC.
World Bank, 2006. An Investment Framework for Clean Energy and Development. The
World Bank, Washington, DC.
World Bank, 2009. World Development Indicators. World Bank, Washington DC.
Zomers, A., 2001. Rural Electriﬁcation. PhD dissertation, University of Twente. 1328O. Rosnes, H. Vennemo / Energy Economics 34 (2012) 1318–1328

Parse pdf-doc by clicking the button below. Parsed pdf-documents will be searchable.

Metadata Found
CreationDate
2012-06-25T18:25:40+08:00
ModDate
2012-06-25T18:25:46+00:00
Producer
Acrobat Distiller 10.0.0 (Windows)
Pages
11