Energy Policy 87 (2015) 125–135

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

Provision of electricity to African households: The importance of

democracy and institutional quality
Helene Ahlborg a,n, Frida Boräng b,c, Sverker C. Jagers b,d, Patrik Söderholm d
a
Chalmers University of Technology, Division of Environmental Systems Analysis, 412 96 Göteborg, Sweden
b
University of Gothenburg, Department of Political Science, Box 711, 405 30 Göteborg, Sweden
c
University of Gothenburg, The Quality of Government Institute, Box 711, 405 30 Göteborg, Sweden
d
Luleå University of Technology, Division of Social Science, 971 87 Luleå, Sweden

H I G H L I G H T S

 Differences in household electricity use across African countries are studied.

 Empirical analysis is based on data from 44 countries over the period 1996–2009.
 The results display the importance of both democracy and institutional quality.
 This suggests a stronger focus on institution-building efforts in energy reforms.

art ic l e i nf o a b s t r a c t

Article history: How can differences in per capita household electricity consumption across African countries be un-
Received 17 April 2015 derstood? Based on theories that highlight the importance of democracy and institutional quality for
Received in revised form provision of public goods, the aim of the paper is to analyse the degree to which the level of per capita
1 September 2015
household electricity consumption in African countries can be attributed to the countries’ democratic
Accepted 1 September 2015
Available online 12 September 2015
status and their institutional quality. We rely on regression analysis employing a pooled data set for 44
African countries over the time period 1996–2009. The analysis shows that democracy and institutional
Keywords: quality both have significant positive effects on per capita household consumption of electricity. Our
Africa results have implications for how energy sector reforms are promoted in developing countries. At a more
Household electricity use
general level they illustrate that institution-building policy efforts are relevant also in areas where
Democracy
contemporary debates have tended to primarily centre on economic development, financial prerequisites
Institutional quality
and ownership issues.
& 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction energy carriers will be instrumental to achieving many of the

Sustainable Development Goals (SDGs) adopted in September
1.1. Background and motivation 2015 by the member states of the United Nations. Whereas the
Millennium Development Goals (MDGs) did not explicitly address
In this paper we address the question of why African countries the energy situation and its relation to poverty alleviation, Goal
differ in terms of households’ ability to benefit from electricity 7 among the SDGs is to “ensure access to affordable, reliable,
services. Electricity is a unique energy carrier that is purposefully sustainable and modern energy for all” (United Nations, 2015). The
applied in almost all aspects of daily life among the richest seg- related targets promote enhanced international cooperation to
ment of the world's population. It is essential for modern com- facilitate energy access, and express the need to expand infra-
munications, industrial development and the build-up of public structure and upgrade technology, especially in the least devel-
oped countries. Also, a recent report from ‘Sustainable Energy for
services such as streetlights, improved education and health care.
All’ (SE4All) emphasizes the importance of understanding the in-
Existing and future applications of electricity and other modern
teractions between energy and other development areas, such as
food, water, gender equality, education and health, in order to
n
Corresponding author. achieve the objectives of the SDGs and the SE4All programme (WB
E-mail address: [email protected] (H. Ahlborg). and IEA, 2015).

http://dx.doi.org/10.1016/j.enpol.2015.09.002
0301-4215/& 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
126 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135

In many African countries, the demand for electricity generally investigating if this notion also applies to the case of electricity
exceeds supply (Eberhard et al., 2011). The continent is rich in provision to households in Africa.
unexploited energy resources, and increasing the generation and
availability of electricity is a political priority in most countries. In 1.2. Objectives and scope
the public debate, electricity is often portrayed as part of ‘mod-
ernity’ – the aspirations towards improved socioeconomic stan- Even if electricity is provided in one sense, i.e., through the
dards of living and political-economic equality (e.g., Ferguson, extension of the national grid, one also needs to assess whether
2006; Winther, 2008). Evidence shows that in areas where access the electricity actually becomes a “good” for the public in the sense
to electricity has been provided, there have been substantial so- that households can benefit from it. Therefore we go beyond
cietal benefits of long-term importance for social and economic previously studied issues, such as provision of infrastructure (as
development, including lightning, education, health, leisure and when measuring access to the grid), and even more closely, yet
security (e.g., Davis, 1998; Karekezi and Majoro, 2002; Kirubi et al., distant matters, such as residential share of electricity relative to
2009; Spalding-Fecher, 2005). industry (Brown and Mobarak, 2009). We do this by trying to
However, in spite of the above, especially sub-Saharan Africa is capture realized use of electricity.
lagging behind the rest of the world in regard to electricity pro- Thereby, our aim is to provide a statistical investigation of the
vision and use in the household sector. The countries in this region determinants of per capita household electricity consumption in
suffer both economically and socially from underdevelopment of African countries with a particular emphasis on the impacts of
the electric power sector, including insufficient generation capa- democratic government and high quality institutions (e.g., rule of
city, unreliability of existing power supplies, poor transmission law and control of corruption). As will be discussed in more detail
and distribution infrastructure and very low rates of access to the below, studying per capita household electricity consumption has
electricity grid (Eberhard et al., 2011; WB and IEA, 2015). In 2012, several advantages over alternative indicators of provision of
the access rate in sub-Saharan Africa was estimated at 35 per cent. electricity to households, including access rates (see Section 2.2).
During the period 2010–2012, the access expansion in this region This article contributes to a growing literature on electrification
only kept pace with population growth, and expansion mainly in developing countries (e.g., Min, 2008; Nanka-Bruce, 2010;
took place in urban areas whereas access rates actually fell in rural Onyeji et al., 2012). This research has pointed to a number of
areas (WB and IEA, 2015). Thus, even if the African continent is drivers and barriers to electrification, including oil production,
rich in energy sources, technical solutions exist and increased economic development, population density, as well as the risk of
electricity provision is much desired, this has not been enough to conflict and violence. These variables are therefore also included
result in a clear increase in electricity access and higher con- as controls in our regression analysis. While previous quantitative
sumption levels in the majority of African countries. There is a studies often have had to rely on cross-sectional data, the present
large literature that addresses the multitude of reasons – histor- analysis builds on data from 44 African countries over the time
ical, financial, political, social and technical – behind the energy period 1996–2009.
situation in various African countries (e.g., Eberhard et al., 2011;
Khennas, 2012; Sokona et al., 2012). Institutional aspects have
1.3. Household electrification and electricity consumption in Africa
received significant attention both in terms of country-specific and
regional institutional drivers and barriers to increased electricity
According to Eberhard et al. (2011:53), “(i)nstalled capacity will
access. The importance of “effective, accountable and inclusive
need to grow by more than 10 per cent annually (…) just to meet
institutions at all levels” is also emphasized in the SDGs Goal 16 on
Africa's suppressed demand, keep pace with projected economic
peaceful and inclusive societies (United Nations, 2015). We intend
growth, and provide additional capacity to support efforts to ex-
to complement the existing literature discussing institutional as-
pand electrification.” There are often barriers related to building
pects (e.g. Foley, 1992; Haanyika, 2006; Karekezi and Kimani,
and operating the infrastructure needed in both urban and rural
2002; WB, 2009) by approaching electrification in African coun-
areas, as well as barriers to electrification resulting in socio-eco-
tries as a case of public goods provision.
nomic development (for a review of these, see Ahlborg, 2012).
This implies that the benefits of providing access to affordable
Many African governments are constrained by the substantial
electricity via national grids are non-excludable and “non-riv-
costs involved in building large-scale electricity infrastructure and
alrous” (e.g., street lights, reliability, etc.) indicating a limited in-
contributions from official development assistance (ODA)1 to
centive for individuals or the private sector to contribute to their
public investment in the electric power sector have been far below
production (e.g., Abbott, 2001). The building of large-scale trans-
the levels needed to keep pace with economic growth and/or in
mission and distribution infrastructure is hugely expensive and a
order to expand public access (World Bank, 2009).
long-term investment, thus typically of little interest to commer-
Access rates reflect the degree to which the government makes
cial investors. Rural electrification is particularly expensive in
electricity infrastructure accessible to the public, but data on
places with difficult terrain, dispersed settlement and/or low po-
connection rates need to be complemented with aspects such as
pulation density. For these reasons, national electric utilities have
capacity, affordability, reliability, quality and safety (WB and IEA,
historically been reluctant to extend the electricity services to
2015: 30). In 2008, the entire generation capacity of sub-Saharan
rural areas. Affordable rural electrification has in most countries
countries amounted to 68 gigawatts (GW), which is no more than
been achieved through special national programmes and funding
that of Spain, and 60 per cent out of this was installed in South
arrangements, including the use of subsidies (Zomers, 2003).
Africa alone. Annual electricity generation is in turn comparatively
Hence, the undertaking of electrifying the entire population is
low for a number of reasons, including aging plants, lack of
primarily politically driven, posing high demands on both the
political and the administrative system. maintenance, and seasonal lack of water in dams. Many countries
Previous research suggests that democracy and institutional are also dependent on imported oil and diesel, a situation that
quality are key prerequisites for the active support of and the creates a vulnerability to high oil prices in the world market
provision of public goods (McGuire and Olson, 1996; Deacon, (World Bank, 2009).
2003; Lake and Baum, 2001; Acemoglu and Robinson, 2006; Boix
et al., 2003; Rothstein, 2011; Rothstein and Teorell, 2008; 1
ODA is a widely used indicator of international aid flow, coined by the Or-
Schmitter and Karl, 1991). In this article, we are interested in ganization for Economic Co-operation and Development (OECD).
 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135 127

Fig. 1. Per capita household electricity consumption in African Countries, 1990 and Fig. 2. Per capita household electricity consumption (kWh): Selection of ten Afri-
2009. Source: United Nations (2013). can countries failing to provide increased household electricity access. Source:
United Nations (2013).

The lack of electricity supply manifests in low consumption

levels. In 2008 the average annual per capita consumption in sub-
Saharan Africa, South Africa excluded, was only 124 kWh (World
Bank, 2009).2 This estimate is, however, based on total electricity
consumption in all sectors of the economy. The consumption so-
lely by households is therefore considerably smaller. Fig. 1 displays
how a set of 44 African countries can be divided into six different
groups with respect to their annual per capita household elec-
tricity consumption levels, indicating also changes over the time
period 1990–2009.
In many of these countries, household electricity consumption
has been very low. Averaged over the whole populations – in-
cluding those with no access to electricity – consumption in many
cases has been lower than 50 kWh per capita. Although some
countries have experienced a shift towards higher levels of elec-
Fig. 3. Per capita household electricity consumption (kWh): Selection of ten Afri-
tricity consumption many have also remained stagnant with vir-
can countries succeeding to provide increased household electricity access Source:
tually no growth (some even experiencing a decline over time). United Nations (2013).
Figs. 2 and 3 provide an illustration of this through a more
detailed account of per capita household electricity consumption
1.4. Outline of paper
in Africa by scrutinizing a number of selected countries. Specifi-
cally, Fig. 2 shows ten examples of countries in which the per
The next section presents the methodological approach of the
capita consumption levels have been maintained at very low levels
analysis. We here address the main theoretical arguments by
or even decreased during the period 1990–2009, while Fig. 3 in-
discussing the relationship between democracy and institutional
stead displays the development in ten African countries that have
quality on the one hand and public good provision on the other.
managed to achieve a non-negligible increase in consumption. The
Two empirically testable hypotheses are derived. In this section
difference in the scale used to display per capita consumption
our data and analytical techniques are also presented. Section 3
levels (y-axis) in these two figures is profound. Overall Figs. 2 and
then outlines and discusses the results from the statistical analysis,
3, although only focusing on a selection of countries, illustrate an
while Section 4 discusses some important implications, not the
interesting variation across various African countries. This varia-
least for policy-making. Section 5 concludes the paper and points
tion deserves further investigation.
towards some directions for future research.
In many African countries, electricity supply is notoriously
unreliable. Low electric power quality, exemplified by blackouts
and power rationing in national grids, is a common problem that
2. Methods: Theoretical framework and empirical approach
incurs significant costs associated with damages to equipment and
disruption of productive activities (World Bank, 2009). Aspects of
2.1. The relationship between democracy, good governance and
capacity and quality are critical for industrial development (VPC,
household electricity consumption
2008; World Bank, 2008) but also for the development of micro
and small-scale businesses, many of which are found among the
One of the central debates in research on the drivers behind
household segment of electricity customers (Haselip et al., 2015).
public goods provision concerns what kinds of governments –
Thus, a statistical investigation of electricity provision as a case of
democratic or autocratic – that provide public goods, such as basic
public good provision requires that the dependent variable mea-
infrastructure and social services, most effectively. Clearly, demo-
sures the degree to which households not only access but actually
cratic institutions – through which the leaders of a country are
use electricity (which indirectly reflects the quality of supply since held accountable to the citizens – create a strong incentive among
consumption levels are lower when access is low and supply is leaders to deliver, for example, generally demanded public goods
very unreliable). such as affordable electricity (Acemoglu and Robinson, 2006;
Schmitter and Karl, 1991). Since elections provide citizens with the
2
A slightly different estimate for 2008, i.e., 155 kWh per capita, is provided by power to replace leaders that do not fulfil these expectations, and
Bazilian et al. (2012). since public good provision is likely to be included in the
128 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135

evaluation of political leaders, democracy can be hypothesized to rules facilitate economic and social interactions), and the presence
lead to more public good provision (McGuire and Olson, 1996; or absence of corruption, conventionally defined as the exercise of
Deacon, 2003; Lake and Baum, 2001). Moreover, generally de- public power for private gain (Gupta et al., 2000; Holmberg et al.,
mocracy can be expected to work in the interest of the majority 2009; Kaufmann and Kraay, 2002; Mauro, 1998; North, 1990; Nye,
(i.e., the rural population) since a larger portion of the citizenry is 1967).
included in the political process, whereas authoritarian regimes On the one hand, institutions can have direct effects on public
will frequently only see to a narrow elite (Bueno de Mequita et al., good provision (Barzel, 2002). On the other hand they can also
2003; Gandhi and Przeworski, 2006). Though primarily concerned have an indirect effect in that they influence the expectations that
with how the share of electricity consumption differs between the people have on the implementing agencies. In other words,
residential sector and the industry sector depending on regime trustworthy institutional systems, with a positive “history of play”
type, Brown and Mobarak (2009) nonetheless find certain support (Acemoglu and Robinson, 2006) can be expected to positively in-
for this argument. At least among the poorest countries in their fluence the ability of agencies to provide public goods, not least
data set (some of which are African), democratic governments through their impact on the ability to build consent regarding the
tend to increase the residential sector's share of electricity con- collection of taxes and other contributions.
sumption relative to industry's share, thus indicating a positive In line with this argument, the literature on rural electrification
relationship between democracy and provision of electricity to the in developing countries has identified poor organizational struc-
public. Following this line of literature, the first hypothesis to be tures and corruption as some of the barriers to successful elec-
tested is trification (e.g., Ahlborg and Hammar, 2014; Jones and Thompson,
1996; Karekezi and Majoro, 2002). In addition, earlier studies have
H1. : The more democratic a country is, the higher is per capita shown that political interference in public electric utilities impacts
household electricity consumption. negatively on their performance (Karekezi and Kimani, 2002; Zo-
Despite strong theoretical reasons to expect a positive effect of mers, 2003).
democratic institutions on public goods provision, some scholars What these findings suggest is that not only are investments
argue that democratic institutions are alone no guarantee for needed in improved generation capacity and expansion of large-
successful public good provision. The first reason is that elected scale grid infrastructure; investments in well-functioning and
leaders often work with short time horizons (Haggard, 1991; adequate institutional frameworks and organizations seem to be
Keefer, 2006) whereas public good provision – not least in the equally important prerequisites for effective electricity provision
form of investments in electric power infrastructure – is a long- at the household level. This is well in line with a growing literature
term undertaking (Min, 2008). Second, the discussion on the im- arguing that regardless of whether a country is democratic or not,
portance of democracy has a certain Western bias. The focus on the output-side of the governance system is crucial (Boix et al.,
accountability and incentive structures for political leaders tend to 2003; Rothstein, 2011; Rothstein and Teorell, 2008). Thus, ac-
cording to, for instance Rothstein (2011), the presence of corrup-
overlook situations where political leaders wish to provide public
tion and poor government effectiveness explains social perfor-
goods but are not able to do so. In African countries, democratic
mance rates, and thus not the level of democracy.
accountability may still be crucial for political leaders’ relative
Proceeding from the broad literature on the quality of institu-
interest in providing public goods, but it might not be all that
tions, the following hypothesis can be formulated:
relevant for their capacity to do so. This was highlighted by an
engineer working with rural electrification at the public utility in H2. : The higher institutional quality a country has, the higher is the
Tanzania (quote from interview by first author in 2010): “Political per capita household electricity consumption.
pressures you know. Like now in October it's the elections, so all
the politicians are running like hell [laughing]. They want elec- The strong focus on the potential roles of democracy and in-
tricity, there is no money but they want the people to hear that stitutional quality on the electrification of African countries re-
they are struggling and working.” This statement reflects that presents our main contribution to the existing empirical literature.
especially in times of elections, politicians may try to fulfil pro- Below we explain in more detail how these analytical concepts
mises to the electorate, even in cases when necessary financial have been operationalized in the quantitative analyses, along with
resources are not available. the other variables used.
Third, several studies also show how corruption and cliente-
listic practices (i.e., the exchange of goods and services for political 2.2. Statistical analysis: variable definitions and sources
support) can undermine governments’ performance also in cases
where democratic institutions are in place (Bratton and van de 2.2.1. Dependent variable
Walle, 1994, 1997; Chandra, 2004; Min, 2008). The importance of The dependent variable in our statistical analysis is annual
these types of factors should not be underestimated: “bad gov- household electricity consumption per capita (kWh per capita/
ernance” – characterized for example by corruption, patronage year). Data over household electricity consumption come from the
and favouritism – has been described as a “spectre haunting de- Energy Statistics Database provided by the United Nations Statis-
mocracy in the world today” (Diamond, 2007: 119). tics Division Database (United Nations, 2013). Using a per capita
Thus, the hypothesis on the importance of democracy must be measure – rather than measuring average consumption among the
complemented with other perspectives. In this paper we focus on electrified minority – has the advantage that we can compare
the importance of institutional quality for public goods provision. development patterns across countries of different population si-
With “institutions” we here refer to “the rules of the game” (e.g., zes. Moreover, we can also assess whether consumption levels
North, 1990), i.e., various external factors shaping and constraining have kept pace with population growth, a good indication of
human behaviour and learning including both legal rules and in- whether the government provides services of a certain quality,
formal codes of conduct (March and Olsen, 1989; Peters, 1999). reaches new parts of the population, and/or whether the popu-
Institutional characteristics that are expected to be important for lation with existing access increases its consumption over time.3
public service provision include everything from the quality of the
bureaucracy and the competence of civil servants, the extent to 3
It should also be noted that in our dependent variable will also account for
which there is effective “rule of law” (where fair and predictable any off-grid access to electricity using solar PV or other generation options in those
 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135 129

Growth in household consumption per capita should be of Table 1

great interest to policy makers since such growth could indicate Per capita household electricity consumption versus electricity access in 44 African
countries, year 2004.
that the middle class of these countries has gained access to
electricity of reasonable quality and price. The middle class can Access to electricity per capita household 0–25% 26–60% 61–100%
afford appliances and other electricity-consuming activities, why a electricity use
growing middle class will result in notable increase in electricity
0–50 kWh 21 0 0
load. Poor households that become electrified, on the other hand,
51–300 kWh 3 10 4
tend to stay at a minimum consumption level for many years also 301 kWh or higher 0 1 5
when tariffs are low (Louw et al., 2008; van der Vleuten et al.,
2007). In this paper, though, we are not able to address the often Source: International Energy Agency (2004) and United Nations (2013).
profound country heterogeneity (e.g., in terms of income disparity,
rural versus urban population shares etc.), which clearly is an advantage of our dependent variable, compared to connection
important issue for future research. rates, is that it measures how access translates into real use, rather
In order to explain the merits and limitations of our dependent than just availability of distribution infrastructure. In case there
variable, we now discuss it in comparison with two alternative are significant problems of reliability of supply, it should have a
measures that are used by other scholars as dependent variables in negative effect on the level of consumption. However, the way we
statistical analyses. The first one is the commonly used ‘elec- have constructed our dependent variable, it adequately reflects
trification rate’ or ‘access rate’ defined as the share of population both the reliability of supply and the access rate.
with access to electricity from national grids (regardless of the The second alternative measure is a newly introduced one,
level of use). The International Energy Agency (IEA) provides data showing the share of population in unlit areas based on analyses of
on access rates. We expect access rates and household consump- satellite night time images (Doll and Pachauri, 2010; Min, 2008).
tion levels to be indirectly related. Notably, our use of a per capita This measure shows the number of people in unlit villages, defined
measure is connected to access rates in the sense that it represents as places that would have detectable lights if electricity were
the consumption of the share of the population that does have present. Apart from the methodological problems that impact the
access averaged over the whole population (including those reliability of the estimates, e.g., discussed by Doll and Pachauri
without access). When comparing these two data sources, we find (2010), this method only measures stable outdoor lights, which
a clear positive relationship.4 can be a problem in African countries where supply is notoriously
As can be seen from Table 1, for countries with an electricity unreliable and load shedding is used in many areas (World Bank,
access rate between 0 and 25 per cent, 21 out of 24 countries have 2009). Another drawback is that people without direct access
a per capita household consumption rate of 50 kWh or lower. living in electrified towns and villages are counted as electrified
Moreover, all nine countries with an electricity access share of 61 (or ‘lit’). For example, people living in slums in big cities fall within
per cent or higher report a per capita consumption level higher the lit areas, but typically have restricted access to other equally
than 50 kWh (and five of these have a consumption rate exceeding important benefits from electricity.
300 kWh). Thus, although there is likely to be unresolved country
2.2.2. Independent variables
heterogeneity, this at least shows that past increases in the per
Our two independent variables are the levels of democracy and
capita household electricity consumption have not only benefited
institutional quality and we rely on existing indicators when
the rich minorities.
measuring them. To capture countries’ democracy level, we use
There are several reasons why we choose to study household
the imputed Freedom House/Polity score Freedom House (Mar-
consumption rather than access rates. First, there are well-known
shall and Jaggers, 2002; Teorell et al., 2011). This score is a com-
reliability problems with the IEA data on access rates. The defini-
bination of two widely used measures of democracy, the Freedom
tion of access is not strictly applied; data are compiled from dif-
House measures for political rights and civil liberties (including
ferent sources and self-reported by countries (e.g., Doll and Pa-
the fairness of the electoral process, the right of opposition parties
chauri, 2010; WB and IEA, 2015). It is argued that the sources of
to take part, freedom for media and organizations, the right of
errors in the data often lead to overestimated access rates (Min,
assembly, etc.) and Polity IV, which focuses on electoral matters –
2008). This is possibly the case also with the data on household
such as elements of political competition and the role of popular
electricity consumption, provided by the UN. We do not have
participation in recruiting the executive – and the distribution of
sufficient information about the sources of data to know its ac-
power, including constraints on the executive (for a critical dis-
curacy. Still, the documented problems with reliability of data on
cussion and comparison of the two indicators, see Hadenius and
access rates motivate us to use another type of indicator, and the
Teorell, 2005). Hadenius and Teorell (2005) show that a combi-
consumption data from the UN is available for more countries and
nation (average) of these two measures performs better in terms
years than the EIA data series for access rates, which has more
of validity and reliability than its constituent parts. The scale of
gaps in the data. Second, although electricity access is an im-
this variable ranges from 0 to 10, where 0 denotes ‘least demo-
portant indicator of the extent to which states have entered the
cratic’ and 10 ‘most democratic’. Where data on Polity is missing,
undertaking of providing electricity infrastructure, it usually says
data has been imputed by regressing Polity on the average Free-
little about the quality and the reliability of the supply. Awaiting
dom House measure (Hadenius and Teorell, 2005).5 The Quality of
improved metrics on electricity access (see WB and IEA, 2015), the
Government Institute provides data for this variable (Teorell et al.,
2011).
(footnote continued) For institutional quality we use a combination of the World
case where these systems are operated by the public utility. Non-governmental
systems are not included.
4 5
This cross-country comparison builds on data for the year 2004, and where Out of the countries used in our analysis, data have been imputed for the
we have categorized 44 African countries according to two dimensions: (a) three Maldives, Sao Tome and Principe, and Seychelles. We also ran sensitivity tests using
different groups based on their per capita household electricity consumption levels country samples without the imputed data for missing observations. The results
(i.e., 0–50 kWh, 51–300 kWh, and 301 kWh or higher); and (b) three groups based from these show that the overall conclusions of the analysis remain the same: the
on their household electricity access rate (i.e., 0–25%, 26–60%, and 61–100%). This effect of the institutional quality index and democracy on per capita household
categorization results in a matrix consisting of nine groups (Table 1). electricity consumption are still positive and statistically significant.
130 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135

Bank's Worldwide Governance Indicators Rule of law and Control of that electrification is largely dependent on economic resources
corruption (Kaufmann et al., 2009). The Worldwide Governance (Zomers, 2001). We thus include an estimate of the countries’ GDP
Indicators are based on several hundred individual variables per capita. Data come from the World Bank's World Development
measuring perceptions of governance, and reflecting the views of a Indicators, and are also provided by the Quality of Government
diverse range of stakeholders. The individual measures are as- Institute (Teorell et al., 2011; World Bank, 2013a). Due to its
signed to categories capturing six dimensions of governance: skewed distribution, the natural logarithm of GDP per capita is
Voice and Accountability, Political Stability and Absence of Vio- used in many other studies. However, in the African context,
lence/Terrorism, Government Effectiveness, Regulatory Quality, where the economies are very small in a global comparison, it is
Rule of Law, and Control of Corruption.6 The scores range from – not obvious that one should use the natural logarithm and by that
2.5 to 2.5, with higher scores corresponding to better outcomes.7 also expect a diminishing “utility” of GDP per capita. Because of
The dimensions that best fit with our theoretical notions about this, we do not employ the logarithm of GDP per capita in the
the role of good governance in provision of electricity are Rule of main model, but we run an additional model including this
Law and Control of Corruption. Rule of Law includes several in- variable.
dicators, which measure the extent to which fair and predictable Since it is, all else equal, more difficult to increase access to
rules form the basis for economic and social interactions as well as reliable and affordable energy when the population is spread over
the extent to which property rights are being protected. This in- vast areas, we include a variable measuring population density.
cludes the effectiveness and predictability of the judiciary, and the This measure – number of people per square kilometre of land
enforceability of contracts. Control of Corruption measures the area – is based on data from the Food and Agriculture Organization
perceptions about corruption, conventionally defined as the ex- and World Bank population estimates included among the World
ercise of public power for private gain. The specific indicators Development Indicators (World Bank, 2013b).
range from perceptions of the frequency of “additional payments When political conflicts emerge, this usually means a great
to get things done”, effects of corruption on the business en- challenge for the goal of electrification. In countries that have
vironment, and “grand corruption” or “state capture” (Teorell et al., suffered from armed conflicts, the electricity infrastructure has
often been seriously damaged or is out of order due to a lack of
2011). Our index of “institutional quality” takes the average of
maintenance. Not only do conflicts imply a great obstacle to the
these two items.8 Observations are available from 1996, with some
normal functioning of the state. It is also common that the parties
gaps in the series during earlier years.
included in the conflict are trying to cut the electricity supplies of
The item Government Effectiveness – which among other in-
its enemies by destroying their infrastructure. We therefore in-
dicators includes the competence of civil servants and the in-
clude the World Bank Worldwide Governance indicator “political
dependence of the civil service from political pressures – is the-
stability” (as previously mentioned), which combines several in-
oretically also relevant as a measure of institutional quality. The
dicators that measure perceptions of the likelihood that the gov-
problem with this dimension of governance, however, is that it
ernment in power will be destabilized or overthrown through
also includes perceptions of the quality of public service provision.
unconstitutional and/or violent means, including domestic vio-
Since such perceptions could well include provision of reliable and
lence and terrorism (Kaufmann et al., 2009; Teorell et al., 2011).9
affordable electricity, this dimension conflates the independent
Finally, we also control for domestic oil production measured in
and the dependent variable.
metric tons per capita. Previous research has identified a statisti-
We have chosen not to include the dimension Voice and Ac-
cally significant and positive impact of oil production on the share
countability, which measures various aspects of the political pro-
of the population that lives in lit areas at the global level (Min,
cess, civil liberties and political rights. The reason for this omission
2008). This finding by Min (2008) contrasts with ideas put forward
is that in our theoretical framework these aspects represent de- in the “resource curse” literature, where the argument has been
mocratic qualities; they should thus be distinct from institutional made that in the absence of good democratic institutions (e.g., rule
quality and be measured separately (and they are included in our of law, lack of corruption, conflict management etc.) exploitation
measure of democracy). The dimension Political Stability, which of natural resources could provide meagre opportunities for public
measures perceptions of the risk for the government being over- benefits (Auty, 2001; Van der Ploeg, 2011). For our purposes this
thrown by unconstitutional or violent means (including domestic suggests that high oil production rates may hinder the develop-
violence and terrorism) (Teorell et al., 2011), is treated as a sepa- ment of competing energy sources such as electricity, including
rate control variable as it may be an important explanatory factor, not the least investments in the public infrastructure needed to
but it does not belong in our theoretical definition of good access new households. In order to address these contradictory
governance. perspectives, and the possibility that oil production could be an
alternative explanation driving our results, we include oil pro-
2.2.3. Control variables duction as a control variable. Data over crude petroleum produc-
As indicated in Section 1, the statistical analysis includes a tion (in metric tons per capita) come from the Energy Statistics
number of control variables. First, previous research has shown Database (United Nations, 2013).

6
An unobserved components model has been used to construct six aggregate
2.2.4. Description of data
governance indicators. The governance estimates are normally distributed with a The statistical analyses rely on yearly country-specific ob-
mean of zero and a standard deviation of one (1) each year of measurement. Al- servations from 44 African countries over the time period 1996–
though standardized estimates can be problematic for comparisons within coun- 2009. The selection of countries and time-period is primarily an
tries over time, Kaufmann et al. (2006) find no systematic time trends in a selection
outcome of data availability. Our focus on the role of institutional
of indicators, thus suggesting that the time-series information in the World Bank's
scores can be used. quality does not permit an analysis of the situation prior to 1996.
7
The fact that the indicators are exclusively based on subjective (perceptions- In addition, due to some reporting lags and frequent revisions of
based) data is a source of criticism. Still, in many instances there are no objective
measures available, and/or the objective measures tend to capture de jure gov-
9
ernance rather than de facto governance (see Kaufmann et al., 2009). Like the other governance indicators, this variable is normally distributed
8
In order to test whether the items are sufficiently inter-related to justify their with a mean of zero and a standard deviation of one (1) each year of measurement,
combination in an index we calculated Cronbach's alpha. This test showed an alpha with scores normally ranging between –2.5 and 2.5. Higher scores correspond to
value of 0.92, thus providing a high credibility to our index. better outcomes.
 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135 131

Table 2
Descriptive statistics for dependent and independent variablesa.

Variables Definitions Mean Std. dev. Min Max

Dependent variable
Household electricity consumption Annual household electricity consumption (kWh/capita) 109.009 165.795 0.830 896.128

Independent variables
Institutional quality Average scores of Rule of law and Control of corruption 0.599 0.565 2.053 0.959
Democracy Average of Freedom House and Polity IV scores (0–10) 4.591 2.508 0.5 10
Political stability Perceptions of risk for political instability/violence 0.461 0.874 2.986 1.188
Population density People per square kilometre of land area 97.810 163.039 1.936 1039.13
Oil production per capita Annual crude petrol production (metric tons per capita) 0.850 3.439 0 31.842
GDP per capita GDP per capita, PPP adjusted (constant International USD) 3013.263 4301.083 247.889 31,518.5

a
Mean estimates, standard deviations, max and min are calculated based on the countries included in the analysis over the period 1990–2009. Observations for the
Institutional Quality Index and Political Stability, however, are missing until the year 1996.

observations for more recent years, we do not employ data after Table 3
2009. The countries included are listed in Appendix A. Table 2 Estimated regression coefficients (standard errors in parentheses).
summarizes the variables used in the quantitative analysis, and
(1) (2) (3) (4)
displays how each variable is defined as well as some important
descriptive statistics. Institutional quality 42.67*** 23.29***
(10.22) (8.879)
Democracy 3.417*** 8.746*** 3.000** 2.963**
2.3. Statistical estimation issues (1.263) (1.674) (1.249) (1.244)
Political stability 10.81** 17.34*** 18.57*** 16.80***
The use of a “pooled” data set, or Time-Series Cross-Section (5.219) (5.323) (6.265) (6.062)
Population density 0.0166 0.0275 0.0330* 0.0313
data set, compared to a cross-sectional data set has the advantage
(0.0217) (0.0212) (0.0188) (0.0195)
of increasing the number of observations (and thus the degrees of Oil production 16.11*** 9.255*** 18.10*** 17.65***
freedom). Another advantage is that pooled data sets permit the (4.465) (2.336) (4.373) (4.241)
inclusion of country-specific effects (which can be assumed to be GDP per capita 0.0322*** 0.0337*** 0.0332***
(0.00265) (0.00243) (0.00245)
fixed over time), thus avoiding bias in the estimates due to un-
GDP per capita (log) 136.2***
observed heterogeneity across countries. Still, Baltagi and Griffin (10.73)
(1984) suggest that if the variation between countries greatly ex- Rule of law 47.76*** 48.27***
ceeds the within variation (i.e., the variation over time), then or- (10.64) (9.173)
dinary least squares regression (OLS) with common intercepts (i.e., Control of corruption 3.498
(9.742)
no country-specific fixed effects) as well as slope coefficients be- Constant 31.67*** 922.4*** 36.99*** 37.05***
comes the preferred estimator. (10.36) (76.16) (9.859) (10.19)
Finally, in reporting our results we primarily focus on the sign
and the statistical significance of the estimated regression coeffi- N 477 477 477 477
cients. An important reason for is that our key independent vari- R2 0.642 0.648 0.665 0.659
ables (e.g., institutional quality) are measured along an ordinal
*
p o0.10.
scale (i.e., data are shown simply in order of magnitude as there is **
p o 0.05.
no standard of measurement of differences). It is well-known that ***
p o0.01.
in such cases the estimated coefficients do not provide any
meaningful information about the magnitudes of the respective
impacts. robustness tests. In what follows, we first devote attention to the
base model results and then revert to the remaining model
specifications.
The results in our base model are well in line with our theo-
3. Results
retical expectations; we find statistically significant positive im-
pacts of both institutional quality and democracy on per capita
The results from the panel data regression analyses are pre-
household electricity consumption. These results are thus con-
sented in Table 3, which displays the results from four different
sistent with the notion that where politicians are held accountable
model specifications.10 The first model is our base model, while
for their decisions and where they are also able to implement
the remaining models represent the outcomes of different types of
large infrastructure projects, these countries are more likely to
report relatively high levels of per capita household electricity
10
Preliminary analyses revealed that one country – Libya – clearly distorted consumption.
the overall relationships. This is especially true for the democracy variable: In the Regarding the control variables, the GDP per capita variable has
year 2000, Libya was the least democratic of all the countries in the sample, but
the expected positive effect on per capita household electricity
simultaneously had the highest per capita household consumption of electricity.
Libya has built extensive electricity infrastructure around the oil industry and is consumption, highlighting, for instance, the importance of funding
now one of the countries with the highest installed capacity on the continent and for the huge financial undertakings that investments in electricity
with the highest total electricity consumption in Northern Africa. In fact, Libya is an infrastructure and generation constitute.
outlier not only within the full set of countries but also within the group of oil- Population density does not have a statistically significant ef-
producing countries: when Libya is excluded, oil production is negatively correlated
with per capita household consumption of electricity; otherwise this is not the
fect on the per capita electricity consumption levels in the base
case. The Libyan case deserves further attention, but for the above reasons this model. This implies that it may not be population density per se,
country has been excluded from our analysis. but rather the distribution of people within a country of a given
132 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135

density, which influences household electrification patterns. In availability of funding is a necessary condition for increased elec-
other words, future research should also address the potential tricity access, it is still not sufficient. Our analysis contributes to
impacts of urbanization rates, given that it is easier to reach the previous discussions of political and institutional issues (e.g.,
larger cities, but also how settlement patterns in rural areas affect Onyeji et al., 2012) by framing public electricity provision as a case
the conditions for, and in turn are affected by, rural electrification. of provision of public goods and by testing the relevance of in-
We find a negative correlation between oil production per ca- stitutional theories that highlight the importance of democracy
pita and per capita electricity consumption levels. This result is in and institutional quality for service delivery in the context of
opposition to previous research findings using satellite images and household electricity consumption in Africa.
reporting a positive effect of oil production on the provision of Contrary to, for instance, Rothstein (2011), we find that both
electric light (Min, 2008). Our result is however consistent with democracy and good governance matter for public good provision
the notion that, in general, the oil-producing countries in Africa and not only the latter, even though we control for per capita in-
have not invested their profits from oil into the provision of come (GDP) levels. The fact that both these factors have a positive
electricity infrastructure for the public domain (Libya being the effect on per capita household electricity consumption should not
exception). come as a surprise. They capture two important aspects or di-
Finally, and clearly in contradiction to our expectations, the mensions of policy-making: the input dimension, i.e., the proce-
results in the base model suggest a negative effect of political dures preceding decision-making and the throughput dimension,
stability, basically implying that the higher the perceived like- the latter referring to the governance structures through which
lihood that the government in power will be destabilized or official decisions are being implemented (Sharpf, 1997). The way
overthrown by possibly unconstitutional and/or violent means, we have conceptualized the causal link between democracy and
the higher is the per capita electricity consumption at the house- public electricity provision, operationalized and finally measured
hold level. It should however be mentioned that in the bivariate it, we should expect that democratic institutions – i.e., competing
case, this variable displays the expected positive relationship with opposition parties, free media and public participation in the re-
household electricity consumption. Excluding political stability cruiting of executives – lead politicians to respond to public de-
from the analysis yields somewhat lower coefficients for the mand for electricity services in order to be able to demonstrate
variables of interest, however the conclusions regarding statisti- results when facing future elections.
cally significant and positive effects of democracy and institutional The importance of well-performing implementing organizations in
quality remain (results not shown here). Future studies may want the case of providing electricity can be illustrated with an example
to consider measuring also actual outbreaks of conflict besides from the time period following decolonization in Africa. In many co-
perceptions of the risk for such outbreaks. lonized countries, the national systems for the generation, transmission
Overall the results from our base model suggest that neither of and distribution of electric power were built by the colonial powers in
our two hypotheses regarding the role of democracy and good order to supply important industries and cities with fuels and elec-
governance, respectively, on the provision of household electricity tricity. At time of independence, many countries were drained of staff
services can be falsified. Before this can be properly concluded, in government and public sector as the European colonial states left. In
however, we must test how robust these results are with respect some countries, various political conflicts damaged the infrastructure,
to the use of alternative model specifications. such as in Mozambique (Hultman, 2009). The drop in organizational
In model (2) we included per capita GDP in logarithmic form. capacity and the associated decrease in institutional quality that fol-
As has been noted above, this is a standard measure motivated by lowed from the sudden loss of human capital, as well as the time lag in
the argument that a modest absolute impact should be assumed rebuilding those, probably had long-lasting negative impacts on state-
for this variable for countries with relatively high-income levels. dominated energy sectors. Thus, even if countries are scoring high on
Table 3 shows that at a general level, our results are robust with democratic indexes, the present governments most probably fall short
respect to this alternative model specification; the coefficients in their ambitions and political promises unless they can be backed-up
representing both democracy and the institutional quality index by an administrative apparatus that can ensure that the democratically
remain statistically significant and positive coefficients. founded decisions are being implemented properly.
Finally on robustness tests, we also tested two different model Our results suggest that in pursuing, for instance, foreign aid
specifications (models (3) and (4)) in which the different in- policy and energy sector reform goals, an integrated approach
dicators of the institutional quality index were included separately towards household electrification is needed. Electricity access is
in the regression equation. Here a general-to-specific modelling part of broader institutional challenges in Africa requiring both
strategy (e.g., Hendry, 1995) was used, i.e., starting with both in- well-functioning input institutions and throughput institutions,
dicators and reducing the coefficients by removing the least sta- and these must be addressed and improved jointly.
tistically significant indicator from the model. Overall Table 3 As we see it, such issues have in part been neglected in the
shows that our results are overall robust with respect to these development of foreign aid as well as in the strategies employed
alternative model specifications. The Rule of Law indicator is the by, e.g., the World Bank. In the 1980s up until the 1990s, the World
only statistically significant institutional variable, but given the Bank's strategy on rural electrification shifted from financial sector
high correlation between the different indicators (see also Section support towards the promotion of energy sector reforms. These
2.2) this result should be expected. reforms have primarily targeted formal institutional frameworks,
Finally, in order to take into account the sometimes divergent de- economic efficiency and involvement of the private sector (Kar-
velopment trajectories of Northern and Sub-Saharan Africa, respec- ekezi and Kimani, 2002; Nawaal Gratwick and Eberhard, 2008;
tively, we also ran the analysis while excluding the Northern African Weisser, 2004; World Bank, 2008).
countries (i.e., Egypt, Libya Morocco and Tunisia) from the sample. The The emphasis on economic aspects in energy sector reform
results from this sensitivity test are reported in Appendix B, and dis- efforts has often come at the cost of social aspects (i.e., equity,
play that the central conclusions hold also in this specification. public services and ‘electricity for all’) (e.g., Zomers, 2003).11 But

4. Discussion 11
The poor management of some state-owned electric utilities in part moti-
vated private sector involvement. Many African public utilities have been char-
The existing literature on household electrification trends in acterized by poor financial and technical performance, as well as by political in-
low-income countries generally concludes that while the terference from governments and urban elites (Karekezi and Kimani, 2002).
 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135 133

also, the reforms have overlooked the contextual challenges and attention to important institutional constraints and capacity-
informal aspects of implementation. For example, previous work building efforts. These efforts are thus not only about bringing in
identifies the lack of human capital, top-down organizational the necessary economic resources in terms of financing.
structures, and low salaries for staff that lead them to take jobs on While our results are overall robust in terms of country sample,
the side, as well as corruption, as important barriers to effective variable specifications, imputed values etc., the findings also
implementation (Ahlborg and Hammar, 2014; Haanyika, 2006). generate several important questions for future research efforts.
Such informal and human aspects of organizational capacity and First, we have only been concerned with explaining the variation
culture have been left largely unaddressed by previous reforms,
found in data over household electricity consumption. Equally
which instead have, arguably, been driven by a top-down logic.
important to investigate is if our findings are valid also when it
Thus, rather than making any claims that the solution to suc-
comes to countries’ total provision of electricity, i.e., also for in-
cessful electrification in African countries solely lies in the pre-
dustry and public institutions but also studies of the role that
sence of democratic and institutional reform, we rather argue that
our findings contribute to the recognition that the provision of democracy and institutional quality play in enabling electricity
public goods is a complex task—electrification being no exception access to various groups in society (e.g., income groups, urban
—and that any success is most likely dependent upon multiple versus rural populations etc.). Another issue for future research is
factors, including political regime type and institutional quality. to analyse how per capita consumption is distributed across dif-
We are also aware that capturing the full complexity of the issue ferent groups of households within the same country. Second, to
clearly requires additional factors, data and research methods. assess whether our findings are unique for the African continent,
On a final note, although rural electrification has hitherto been studies of household consumption in low-income countries at
a responsibility of the state, due to the huge investments needed other continents would be required. Third, while we show that
for building large-scale grid infrastructure, there is now an in- both democracy and institutional quality appear to play significant
creased interest in decentralized development paths. A combina- roles in enabling household electrification Africa, it is also clear
tion of factors – e.g., technological innovation, environmental that these issues deserve further attention in future work.
concerns, change in political priorities and new forms of financing There are also limitations with respect to the statistical analy-
– has led many countries to consider an alternative, or at least a
sis, e.g., in this paper we have implicitly assumed that democracy
complementary, policy approach to rural electrification. Small-
and institutional quality are additive and unrelated, thus implying
scale decentralized generation and distribution, e.g., making use of
that a given improvement in the quality of institutions will have
locally available renewable energy sources, offers a range of so-
the same effect on household electrification regardless of the level
lutions that can meet household and community energy needs at
lower cost than an expansion of the national grid. If this trend of democracy. In practice, the institutional dynamics and linkages
continues it may well alter the role of government policy in between government, institutions and electrification are more
electricity service provision. Our understanding of the historical complex and it would be useful to study these dynamics from a
relationships between democratic government, institutional socio-technical systems perspective. Qualitative analysis would be
quality and public goods provision may therefore not be relevant helpful in order to grasp the importance of informal institutions
for future energy systems where actors presumably may take on and the ‘through-put’ dimension of governance. Finally, the role of
new roles, technical systems are configured differently and cen- states in electrifying the population is not necessarily going to
tralized governance is replaced by polycentric forms of governance remain the same in the coming decades. Entirely new forms of
(Goldthau, 2014). governance may be expected to develop with the necessary global
transition to environmentally sustainable energy systems. Under-
standing such development will require different kinds of analyses
5. Conclusion and policy implications
as well as data other than national statistics on large-scale elec-
tricity provision.
In this paper, we have addressed the overall question of what
explains popular access to and use of affordable and reliable
electricity in African households. Founded in theories that high-
Acknowledgement
light the importance of democracy and institutional quality for
public good provision, we hypothesize that these factors can help
An earlier version of this paper was presented at the Centre for
us better understand the variance in average household con-
sumption of electricity across African countries over the time Environmental Political Studies (CEPS) conference at Tjärnö, 29–31
period 1996–2009. According to our findings, both qualitative in- May 2013. We gratefully acknowledge financial support from the
put aspects (i.e., political and legal procedures) and throughput two institutions that have funded the research programme of
aspects of the state (regulating the behaviour of the implementing which this study is part: the Swedish International Development
bureaucracy) contribute to explaining the observed variation in Cooperation Agency (SIDA) and the, providing grants of approx.
household electricity use. Specifically, the results from our statis- 359 000 Euro and 645 000 Euro respectively during the period
tical analyses cannot falsify our two hypotheses suggesting a po- 2012-2014. We also appreciate valuable comments from our Po-
sitive relationship between both the presence of democratic in- litical Science colleagues at Tjärnö and three anonymous re-
stitutions and institutional quality on the one hand and per capita viewers. Any remaining errors, however, reside solely with the
household electricity consumption on the other. authors.
This illustrates that institutional theories are relevant also in
areas where contemporary debates have tended to centre on
economic development, financial prerequisites and ownership is-
sues. An important policy implication is that energy sector reform Appendix A
efforts in general and efforts to promote household electrification
(e.g., as part of foreign aid) in particular should increasingly pay See Table A1.
134 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135

Table A1 Democracy. Cambridge University Press, New York.

African countries Included in the statistical Ahlborg, H., 2012. Electricity for Better Lives in Rural Tanzania and Mozambique.
analysis. Understanding and Addressing the Challenges. Department of Energy and En-
vironment, Chalmers University of Technology, Göteborg.
Angola Madagascar Ahlborg, H., Hammar, L., 2014. Drivers and barriers to rural electrification in Tan-
zania and Mozambique – grid-extension, off-grid, and renewable energy
Benin Malawi technologies. Renew. Energy 61, 117–124.
Auty, R.M., 2001. The political economy of resource-driven growth. Eur. Econ. Rev.
Botswana Maldives
45, 839–846.
Burkina Faso Mali
Baltagi, B.H., Griffin, J.M., 1984. Short and long-run effects in pooled models. Int.
Burundi Mauritania Econ. Rev. 25, 631–645.
Cameroon Mauritius Barzel, Y., 2002. A Theory of the State: Economic Rights, Legal Rights, and the Scope
Central African Republic Morocco of the State. Cambridge University Press, Cambridge.
Chad Mozambique Bazilian, M., et al., 2012. Energy access scenarios to 2030 for the power sector in
Comoros Niger sub-Saharan Africa. Util. Policy 20, 1–16.
Congo Nigeria Boix, C., Adserà, A., Payne, M., 2003. Are you being served? Political accountability
Congo, Democratic Republic Rwanda and quality of government. J. Law Econ. Organ. 19 (2), 445–490.
Cote d'Ivoire Sao Tome and Principe Bratton, M., van de Walle, N., 1994. Neopatrimonial regimes and political transi-
Djibouti Senegal tions in Africa. World Politics 46 (4), 453–489.
Egypt Seychelles Bratton, M., van de Walle, N., 1997. Democratic Experiments in Africa: Regime
Equatorial Guinea Sierra Leone Transitions in Comparative Perspective. Cambridge University Press,
Eritrea South Africa Cambridge.
Brown, D.S., Mobarak, A.M., 2009. The transforming power of democracy: regime
Ethiopia Swaziland
type and the distribution of electricity. Am. Polit. Sci. Rev. 103, 193–214.
Gabon Tanzania
Bueno de Mequita, B., Smith, A., Siverson, R.M., Morrow, J.D., 2003. The Logic of
Ghana Togo Political Survival. MIT Press, Cambridge, MA.
Guinea Tunisia Chandra, K., 2004. Why Ethnic Parties Succeed. Cambridge University Press, Cam-
Kenya Uganda bridge, UK; New York, NY.
Lesotho Zambia Davis, M., 1998. Rural household energy consumption: the effects of access to
electricity—evidence from South Africa. Energy Policy 26 (3), 207–217.
Deacon, R., 2003. Dictatorship, Democracy and the Provision of Public Goods.
Economics Working. University of California at Santa Barbara, pp. 11–99.
Appendix B Diamond, L., 2007. A quarter-century of promoting democracy. J. Democr. 18,
118–120.
See Table B1. Doll, C., Pachauri, S., 2010. Estimating rural populations without access to electricity
in developing countries through night-time light satellite imagery. Energy
Policy 38, 5661–5670.
Eberhard, A., Rosnes, O., Shkaratan, M., Vennemo, H., 2011. Africa's Power Infra-
structure. The World Bank, Washington, DC.
Table B1
Ferguson, J., 2006. Global Shadows: Africa in the Neoliberal World Order. Duke
Regression model estimates when excluding Northern African countries (i.e., Egypt,
University Press, Durham.
Libya, Morocco, Tunisia) from the sample. Estimated regression coefficients (stan- Foley, G., 1992. Rural electrification. The institutional dimension. Utilities Policy,
dard errors in parentheses). 283–289.
Freedom House. 〈http://www.freedomhouse.org〉.
(1) (2) (3) (4) Gandhi, J., Przeworski, A., 2006. Cooperation, cooptation, and rebellion under dic-
tatorships. Econ. Politics 18 (1), 1–26.
Institutional quality 33.93*** 25.48*** Goldthau, A., 2014. Rethinking the governance of energy infrastructure: scale, de-
(8.781) (8.350) centralization and polycentrism. Energy Res. Soc. Sci. 1, 134–140.
Democracy 6.192*** 9.103*** 6.080*** 6.113*** Gupta, S., Davoodi, H., Tiongson, E., 2000. Corruption and the provision of health
(1.541) (1.806) (1.531) (1.498) care and education services. Working Paper. International Monetary Fund,
Political stability 9.634* 16.53*** 10.67** 11.23** Washington, DC.
(4.972) (5.135) (5.170) (5.276) Haanyika, C.M., 2006. Rural electrification policy and institutional linkages. Energy
Population density 0.00310 0.0236 0.0000500 0.00383 Policy 34 (17), 2977–2993.
Hadenius, A., Teorell, J., 2005. Assessing alternative indices of democracy, C&M
(0.0232) (0.0213) (0.0225) (0.0222)
Working Paper 6, IPSA, August 2005. Available at 〈www.conceptsmethods.org/
Oil production 14.33*** 8.508*** 14.53*** 15.29***
working_papers/20050812_16_PC%206%20Hadenius%20&%20Teorell.pdf〉 (ac-
(4.427) (2.230) (4.420) (4.385)
cessed 10.05.12).
GDP per capita 0.0306*** 0.0308*** 0.0313*** Haggard, S., 1991. Inflation and stabilization. Politics and policy making in devel-
(0.00281) (0.00279) (0.00279) oping countries. In: Meier, Gerald M. (Ed.), Perspectives on the New Political
GDP per capita (log) 131.8*** Economy. ICS Press, San Francisco, pp. 233–249.
(10.37) Haselip, J., Desgain, D., Mackenzie, G., 2015. Non-financial constraints to the es-
Rule of law 22.39*** 31.19*** tablishment and scaling up of energy SMEs in Ghana, Senegal, Tanzania and
(8.348) (7.296) Zambia. Energy Res. and Social Sci 5, 78–89.
Control of corruption 13.07 Hendry, D.F., 1995. Dynamic Econometrics. Oxford University Press, Oxford.
(9.043) Holmberg, S., Rothstein, B., Nasiritousi, N., 2009. Quality of government: what you
Constant 8.982 892.5*** 10.68 8.596 get. Annu. Rev. Polit. Sci. 12, 135–161.
(12.01) (74.67) (11.58) (11.59) Hultman, L., 2009. The power to hurt in civil war: the strategic aim of RENAMO
violence. J. South Afr. Stud. 35 (4), 821–834.
International Energy Agency (IEA), 2004. World Energy Outlook. OECD, Paris.
N 445 445 445 445 Jones, G.J., Thompson, G., 1996. Renewable energy for African development. Sol.
R2 0.646 0.645 0.650 0.651 Energy 58 (1–3), 103–109.
Karekezi, S., Kimani, J., 2002. Status of power sector reform in Africa: impact on the
*
p o 0.10. poor. Energy Policy 30 (11–12), 923–945.
** Karekezi, S., Majoro, L., 2002. Improving modern energy services for Africa's urban
po 0.05.
*** poor. Energy Policy 30 (11–12), 1015–1028.
p o 0.01.
Kaufmann, D., Kraay, A., 2002. Growth Without Governance. World Bank Policy
Research Working Paper no. 2928 (November), Washington, DC.
Kaufmann, D., Kraay, A., Mastruzzi, M., 2006. Governance Matters V: Aggregate and
Individual Governance Indicators for 1996–2005. The World Bank.
Kaufmann, D., Kraay, A., Mastruzzi, M., 2009. Governance Matters VIII: Aggregate
References and Individual Governance Indicators for 1996–2008. World Bank Policy Re-
search Paper No. 4978.
Keefer, P., 2006. Clientelism, credibility and the policy choices of young democ-
Abbott, M., 2001. Is the security of electricity supply a public good? Electr. J. 14 (7), racies. Am. J. Polit. Sci. 51 (October), 804–821.
31–33. Khennas, S., 2012. Understanding the political economy and key drivers of energy
Acemoglu, D., Robinson, J.A., 2006. Economic Origins of Dictatorship and access in addressing national energy access priorities and policies: African
 H. Ahlborg et al. / Energy Policy 87 (2015) 125–135 135

perspective. Energy Policy 47 (Suppl. 1), S21–S26. Sharpf, F.W., 1997. Games Real Actors Play: Actor-centered Institutionalism in
Kirubi, C., Jacobson, A., Kammen, D.M., Mills, A., 2009. Community-based electric Policy Research. Westview Press, Boulder.
micro-grids can contribute to rural development: evidence from Kenya. World Sokona, Y., Mulugetta, Y., Gujba, H., 2012. Widening energy access in Africa: to-
Dev. 37 (7), 1208–1221. wards energy transition. Energy Policy 47 (Suppl. 1), S3–S10.
Lake, D.A., Baum, M., 2001. The invisible hand of democracy: political control and Spalding-Fecher, R., 2005. Health benefits of electrification in developing countries:
the provision of public services. Comp. Polit. Stud. 34 (6), 587–621. a quantitative assessment in South Africa. Energy Sust. Dev. 9 (1), 53–62.
Louw, K., Conradie, B., Howells, M., Dekenah, M., 2008. Determinants of electricity Teorell, J., Samanni, M., Holmberg, S., Rothstein, B., 2011. The QoG Standard Dataset
demand for newly electrified low-income African households. Energy Policy 36 version 6Apr11, University of Gothenburg: The Quality of Government Institute.
(8), 2812–2818. 〈http://www.qog.pol.gu.se〉.
March, J.G., Olsen, J.P., 1989. Rediscovering Institutions: the Organizational Basis of United Nations, 2013. Energy Statistics Database. United Nations Statistic Division
Politics. Free Press, New York. (downloaded May 2013)〈http://data.un.org/〉.
Marshall, M.G., Jaggers, K., 2002. Polity IV Project: Political Regime Characteristics United Nations, 2015. Sustainable development goals. 〈http://www.un.org/sustai
and Transitions, 1800–2002: Dataset Users’ Manual. University of Maryland, nabledevelopment/sustainable-development-goals/〉 (accessed 20.08.15).
Maryland. Van der Ploeg, F., 2011. Natural resources: curse or blessing? J. Econ. Lit. 49 (2),
Mauro, P., 1998. Corruption and composition of government expenditure. J. Public 366–420.
Econ. 69, 263–279. van der Vleuten, F., Stam, N., van der Plas, R., 2007. Putting solar home system
McGuire, M.C., Olson, M., 1996. The economics of autocracy and majority rule: the programmes into perspective: what lessons are relevant? Energy Policy 35 (3),
invisible hand and the use of force. J. Econ. Lit. 34 (1), 72–96. 1439–1451.
Min, B., 2008. Democracy and Light: electoral accountability and the provision of VPC, 2008. Productive Uses of Energy. Vattenfall Power Consultant, Maputo.
public goods. Working paper. Weisser, D., 2004. Power sector reform in small island developing states: what role
Nanka-Bruce, O., 2010. The Socio Economic Drivers of Rural Electrification in Sub- for renewable energy technologies? Renew. Sust. Energy Rev. 8 (2), 101–102.
Saharan Africa, Surrey Energy Economics Discussion Paper Series 128. Surrey Winther, T., 2008. The Impact of Electricity. Development, Desires and Dilemmas,.
Energy Economics Centre, University of Surrey, UK. Berghahn Books, New York.
Nawaal Gratwick, K., Eberhard, A., 2008. An analysis of independent power projects World Bank, 2008. The Welfare Impact of Rural Electrification: A Reassessment of
in Africa: understanding development and investment outcomes. Dev. Policy the Cost and Benefits. World Bank Independent Evaluation Group, Washington,
Rev. 26 (3), 309–338. DC.
North, D., 1990. Institutions, Institutional Change and Economic Performance. World Bank, 2009. Underpowered: The State of the Power Sector in Sub-Saharan
Cambridge University Press, Cambridge. Africa. Africa Infrastructure Country Diagnosis. Background paper 6. The World
Nye, J.S., 1967. Corruption and political development: a cost–benefit analysis. Am. Bank, Washington.
Polit. Sci. Rev. 61, 417–427. World Bank, 2013a. World Development Indicators. 〈http://data.worldbank.org/
Onyeji, I., Bazilian, M., Nussbaumer, P., 2012. Contextualizing electricity access in data-catalog/world-development-indicators〉.
sub-Saharan Africa. Energy Sust. Dev. 16, 520–527. World Bank, 2013b. World Development Indicators. 〈http://data.worldbank.org/
Peters, B.G., 1999. Institutional Theory in Political Science. The ‘New In- data-catalog/world-development-indicators〉 (downloaded May 2013).
stitutionalism’. Continuum, London, New York. WB and IEA, 2015. Sustainable Energy for All 2015: Progress Toward Sustainable
Rothstein, B., 2011. The Quality of Government: Corruption, Social Trust and In- Energy. World Bank, Washington, DC.
equality in International Perspective. University of Chicago Press, Chicago. Zomers, A., 2001. Rural Electrification. Utilities’ Chafe or Challenge?. Twente Uni-
Rothstein, B., Teorell, J., 2008. What is quality of government: a theory of impartial versity Press, Enchede.
political institutions. Governance 21 (2), 165–190. Zomers, A., 2003. The challenge of rural electrification. Energy Sust. Dev. 7 (1),
Schmitter, P.C., Karl, T.L., 1991. What democracy is… and is not. J. Dem. 2 (3), 75–88. 69–76.