Table of Content
List of Figures
List of Tables
List of Abbreviations
II. Literature Overview
A. The Aid-Growth Relationship
B. The Disaggregation of Aid Flows
C. The Electricity Sector Worldwide: An Overview
D. The Nexuses of Electricity Supply
III. Empirical Analysis
A. Data Sources and Units
B. Descriptive Statistics
C. Estimation Strategy
E. Robustness Tests
Abstract: This bachelor thesis examines whether foreign aid to the energy sector has a significant impact on the development of a country’s electricity sector by analysing the impact of aid disbursements to the energy sector on three different sector-specific outcome indicators in a fixed effects model. Besides, as heterogeneity of aid recipients might influence the effect, different regressions are run for low income, lower middle income and upper middle income countries. Results indicate that foreign aid to the energy sector affects electricity output positively and significantly, while not having an impact on generation capacity. The impact on electricity consumption per capita remains unclear. Furthermore, it seems to work more effectively in richer countries. This is robust to a set of sensitivity tests. Thus, the results emphasize the importance of disaggregating aid flows and the use of sector specific outcome variables in order to measure the impact of foreign aid.
List of Figures
Figure 1: Foreign Aid to the Energy Sector 1973-2010
Figure 2: Energy foreign aid by subsector (1991-2010)
Figure 4: Control Variables (1991=100)
Figure 3: Output, Capacity and Consumption (1991=100)
Figure 5: Foreign aid to the energy sector 1973-2010, without Iraq A
Figure 6: Average foreign aid to the energy sector as share of GDP by continent, 1990-2010 A
Figure 7: Histogram (percent) of capacity factor, 1991-2010, yearly data A
Figure 8: Average capacity factor, 1991-2010, yearly data A
Figure 9: Histogram (percent) of difference between total consumption and production in percent A
Figure 10: Average renewable electricity production in percent of total production, 1991-2010, weighted by population A
List of Tables
Table 1: Summary statistics, yearly data 1991-2010 20
Table 2: Fixed Effects Regression 26
Table 3: Fixed Effects Regression by Income Group, Dependent Variable: Log of Total Output 28
Table 4: Included countries, sorted by income group A3
Table 5: Total energy foreign aid by country, 1973-2010 A5
Table 6: Robustness tests I A6
Table 7: Robustness tests II A7
Table 8: Robustness tests III A8
Table 9: List of included donors A9
List of Abbreviations
Abbildung in dieser Leseprobe nicht enthalten
Imagining a modern economy without electricity is impossible. Since the ability of electricity to transmit energy was discovered in the 19th century, it has become a crucial part of everyday life, at least in industrial countries. Industry, agriculture and most services can only function if stable and reliable electricity supply is secured. But also non-commercial parts of society rely heavily on electricity: Health institutions, education, security or communication – electricity is a basic need of a modern society. It is obvious that the challenges that many developing countries are facing in those mentioned areas can only be solved if electricity supply is secured as well.
One possible way to improve electricity supply in developing countries might be through foreign aid to the energy sector. This aid is typically granted for the construction or rehabilitation of power plants, the construction of transmission lines, rural off-grid electrification or technical assistance and capacity building (OECD 2012).
Since 2015, Sustainable Development Goal (SDG) No. 7 for the first time in history defines what the international community wants to achieve until 2030 with regard to electricity supply, which is “ensure access to affordable, reliable, sustainable and modern energy for all” (United Nations 2015b, 19). New initiatives, such as the “Sustainable Energy for All (SE4ALL)” initiative of the United Nations, the “Energy Sector Management Assistance Program (ESMAP ) “ of the World Bank or, on a national level, the „Deutsche Klima- und Technologieinitiative (DKTI)“ in Germany were launched to achieve this SDG and combine the goals of electrification and climate protection.
While the existing literature on the effectiveness of foreign aid to promote economic growth is vast, the literature on sector specific aid is, compared to the aid-growth relationship, marginal.
As foreign aid to the energy sector is an important part of aid flows, a better understanding of the effects that it can have on the development of a country’s electricity sector is needed.
Thus, this bachelor thesis shall answer the question, whether foreign aid to the energy sector fosters development in this same sector.
It does so by testing the hypothesis that this aid does improve electricity supply empirically in a data panel. The remainder of this paper is organized as follows:
Subsequently, in section II, I will first give an overview of the existing literature on general aid effectiveness and explain findings of previous studies that analyse sector-specific foreign aid to other sectors such as education, health or democratization. Afterwards, I will provide a brief overview of the development and the current situation of electrification in developing countries and explain the nexuses of electricity supply with other goals of human development and economic growth.
Section III contains the empirical analysis of a data panel, which aims to discover if or to what extend foreign aid to the energy sector can improve electricity supply in developing countries. Section IV contains the conclusion and a short outlook.
II. Literature Overview
A. The Aid-Growth Relationship
The impact and effectiveness of foreign aid on the development of countries has been an issue for researchers for over 40 years. However, there is still no clear answer to the question whether foreign aid does or does not foster economic growth or under which circumstances it could do so.
In line with the recent literature, e.g. Arndt, Jones, and Tarp (2010), Doucouliagos and Paldam (2009) or Mekasha and Tarp (2013), I categorise the existing studies on the aid-growth nexus into four generations:
The first generation, which represents literature in the 1970s and early 1980s was based on the idea that foreign aid should accumulate capital in receiving countries and that growth would be achieved by investing this capital. Being based on neoclassical growth theory, a linear relationship between growth on the one hand and investment in physical capital on the other hand was assumed (Arndt, Jones, and Tarp 2010, 2), but the actual impact on growth was hardly investigated in these studies, as they focus on the impact of aid on investment. Papanek (1973, 129) for example comes to the result that “Savings and the components of foreign inflows ... explain over a third of the growth rate. Foreign aid ... has a more significant effect on growth than savings or the other forms of foreign resource inflows.“ Altogether, most of these studies find a positive effect of aid on investment but also show that most aid is consumed (often by the government) and not invested (Arndt, Jones, and Tarp 2010, 2). In their detailed meta-analysis of the aid effectiveness literature (AEL), Doucouliagos and Paldam (2009) analyse the results of 97 AEL papers, which include a total of 182 different models, among those 58 models that belong to the first generation of AEL. From the first generation papers, they conclude, that aid increases capital accumulation by 25 % of the actual given aid, while the rest is consumed. Whether aid can lead to growth, however, depends on how the remaining 75 % are used. It is also important to mention, that available methodology, lack of data and lack of modern computing technologies made research difficult during the first generation of AEL. Gupta (1970), for example, uses a dataset of only 50 observations and Papanek (1973) uses 85 observations, compared to hundreds or even thousands of observations in modern panel-data approaches. Besides, first-generation AEL did not take into account endogeneity of aid, meaning the fact that countries could explicitly receive more aid because of low growth rates.
The second-generation AEL focused on the link between foreign aid and growth directly. Most of this literature, which was mainly published during the 1980s and early 1990s, is based on regressions that included aid as well as a set of control variables. Doucouliagos and Paldam (2009, 451) find about 60 different controls that have been used in this literature, on average five per model. The meta-analysis of 543 different estimates from this generation reveals, that the impact of aid on growth, found by these studies altogether, is insignificant but on average, they find slightly positive estimates. One important finding of this time is the so-called micro-macro paradox (Mosley 1986), which points out the differences between micro and macro evidence of aid effectiveness: „The World Bank ... reports average ex-post rates of return of over 10 % ... yet the macroeconomic data from regressions of aid on growth across a cross-section of developing countries are discouraging“ (Mosley 1986, 129). Besides the still existing lack of long-term data, it was still a main issue for this literature that hardly any of the studies considered the possible endogeneity of aid (Arndt, Jones, and Tarp 2010, 2).
The third generation of AEL, which developed in the late 1990s, can be recognised by more advanced methodological approaches such as the use of panel data, new theories that influenced economic research and the availability of modern computing technologies. This allowed not only to analyse huge datasets, but also to include more complex models such as two-stage least squares (2SLS) that can control for endogeneity. This literature also finds a plausible solution for the so-called „zero correlation result“ (Doucouliagos and Paldam 2009, 437), which states that simple regressions fail to show a significant impact of aid on growth. Third-generation AEL solves this issue by stating that the effectiveness of aid depends on other circumstances, such as good policy in receiving countries. This idea was first published by Burnside and Dollar (1997) and has greatly influenced the literature and politics, especially with regard to conditional lending, afterwards, when even the President of the United States concluded: „We must tie aid greater to political and legal and economic reforms“ (Bush 2002).
Doucouliagos and Paldam (2009) find 23 studies and 232 estimates that follow this idea, however, in their meta-analysis, the results turn insignificant. They recall that „How much this message has actually affected World Bank lending over the last decade is not known, but it has probably had some effect“ (Doucouliagos and Paldam 2009, 453-454).
The fact that good policy and functioning institutions are vital for a country’s development is widely accepted throughout the literature (see for example Acemoglu, Johnson, and Robinson 2000), but the findings of Burnside and Dollar remain unclear until today: Easterly, Levine, and Roodman (2003) expanded the original dataset and then applied the exact same model as Burnside and Dollar, leading to the collapse of any significant result. In some of their regressions, the coefficients even show the wrong sign (Easterly, Levine, and Roodman 2003, 5). Consequently, even though third-generation AEL is recognized as an improvement compared to previous literature, it could not produce consistent results.
The latest, or fourth-generation, AEL is also the most critical about granting aid in general. As no study in the past could undoubtedly and reproducibly show that aid works, the literature has moved on, on the one hand to a more pessimistic view. On the other hand, the methodological borders are becoming clearer and the necessity to have a more disaggregate view on aid is mentioned often in recent literature (e.g. Mavrotas and Nunnenkamp 2007, 586).
One of the most cited studies of this kind is Rajan and Subramanian (2008). Their extensive study is above all impressive because of the generality of their analysis. Instead of focussing on one aspect that might influence aid effectiveness, such as institutions or geography; or focussing on the endogeneity of aid, they provide answers to nearly all underlying questions: By generating a donor-related instrument of aid, based on donors’ interest, all their instruments are significant and together account for around 40 % of the variance in aid (Rajan and Subramanian 2008, 649). With this approach, they analyse the impact of aid on growth using different time horizons, controlling for possibly decreasing returns to aid, including interaction terms with institutions and geography and taking into account different kinds of aid such as social aid vs. economic aid or bilateral aid vs. multilateral aid. They do this for cross-country data as well as for panel data, using different econometrical approaches. Their finding, however, is simple: The aid coefficient is not significant anywhere, nor are the interaction terms; the coefficient comes even out negative in a not ignorable share of the regressions they run.
From this result, two possible conclusions can be drawn. Firstly, aid might really be ineffective in every way and should just be stopped as soon as possible, as also proposed by some popular literature, for example Moyo (2009).
Another interpretation is, that it may just not be possible to find evidence of aid an impact of aid on growth through cross-country or panel evidence because of the vast amount of variables that influence both, aid and growth; meaning because of the huge noise within the data. Consequently, no matter how good the methodology and the data might be, aid-growth regressions will not serve its purpose. This would mean having to „move away from the traditional cross-sectional analysis, and focus on more direct evidence of the channels through which aid might help or hinder growth“ (Rajan and Subramanian 2008, 660). This approach, which is called disaggregation of aid, is also the motivation for this bachelor thesis and will be explained more detailed in the following section.
B. The Disaggregation of Aid Flows
The obvious challenge of the mentioned literature is the complexity of the aid-growth relationship and the vast number of variables that influence both, economic growth and the flow of foreign aid. The relationship between aid and growth is tremendously complex and some argue that “it is virtually impossible to acknowledge all factors with potential impacts on the link between aid and economic development” (Michaelowa and Weber 2007, 2). It is more and more acknowledged in the current literature that foreign aid and economic growth are not linked close enough to find a significant impact of the former on the latter (Mishra and Newhouse 2009, 855).
Besides that, using aggregates of aid disbursements is misleading, as it assumes that all kinds of aid are granted to foster economic growth and, moreover, in the same quantity.
From a donor point of view, however, aid flows are not granted in general to achieve growth in receiving countries, in contrast, donors always see (besides of potential strategic interest) sector-specific goals when choosing where to grant money. This is best shown by the common goals set by the international community such as the SDGs: While there are goals set for different sectors (e.g. no. 7 for the energy sector), there is no general goal for economic growth (United Nations 2015b). Aid does not even have to have long-term growth impacts: Donors give aid for example to protect the climate and the environment (which does not have a link to growth at all), to health projects (which may or may not have a link to growth ) or to projects that support democracy building, which probably do not cause growth either (Tavares and Wacziarg 2001). This aid is given in order to improve human living conditions, and economic growth does not play a major role in this context. National aid agencies also publish their goals, and it is not surprising that they have multiple objects, apart from economic growth
If donors often use sector-specific indicators in order to set the goals and measure the impact of their financing as shown by the SDGs, it is only consequent that macroeconomic econometrical approaches should at least consider using sector-specific indicators as well.
A growing number of researchers has become aware of this issue in recent years, among others Harms and Lutz (2004, 23) who’s statement was that “In fact, given that ODA comprises of such diverse components as emergency food aid, the building of wells, the construction of airports and the salaries of teachers, it is surprising that some researchers obtained any results at all.” That these different types of aid will have different impacts on economic growth is quite intuitive (Asiedu and Nandwa 2007, 633), however, it is hardly covered by the aid-growth literature yet.
Results from a disaggregated analysis of aid effectiveness could, among other things, help donors to get a better understanding of which kind of aid is effective and which is not.
Disaggregation is possible in different ways: On the one hand, aid can be divided into categories such as general budget support, project aid or emergency aid. Finding differences in the effectiveness of these categories can have important implications for donors, especially with regards to potential fungibility of aid, meaning the danger that aid receiving countries spend the money that they save because of foreign aid inflows for other purposes (e.g. arms).
This approach was undertaken by Mavrotas (2002) who divided aid into different categories such as emergency aid, programme aid, project aid and food aid. By analysing these different kinds of aid, he finds significant results which prove that project aid is less fungible than programme aid and thus concludes that “disaggregating aid flows should take a decisive role in evaluating aid effectiveness” (Mavrotas 2002, 553).
Even though fungibility and potential solutions to it is an important issue, I will focus on disaggregating aid by sectors in this bachelor thesis.
This can be done at three different levels, (1) by analysing the impact of aggregate aid on a sector specific outcome (e.g. aid on democracy), (2) by analysing the impact of sector-specific aid on economic growth (e.g. education aid on growth) and (3) by analysing aid flows as well as outcomes on a sector specific level (e.g. health aid on life expectancy).
An example of (1) is Knack (2004) who investigated the impact of total aid flows on democratisation in recipient countries. The main reason that the study chose aggregate aid as an independent variable (instead of only using aid specifically granted for democratisation) is that through conditionality, every aid flow could potentially contribute to democratisation and that improved education and higher income, achieved through aid, might contribute to it as well (Knack 2004, 251). Independent of the estimation method, Ordinary Least Squares (OLS) or Instrumental Variables (IV), the study fails to find any significant impact of aid on democratisation and concludes that aid in general has not contributed to democratisation, which also indirectly criticises conditional aid as ineffective.
For an author to consider level (2) of disaggregation, there must be a clear link from the sector where aid is invested towards economic growth. This clearly is the case for education, where economic theory, such as neoclassical and new growth models, are in line with the empirical literature (e.g. Barro 2001) and show that education clearly affects growth positively. It is thus interesting to assess both, the impact of aid to the education sector on educational outcomes as well as on economic growth.
The former has been done by Dreher, Nunnenkamp, and Thiele (2008) using a panel of nearly 100 countries over 34 years. By using a social production function approach, their model implies that educational outcomes, in this case net enrolment rates, depend on supply as well as demand side variables. The supply side includes government spending, aid for education measured in commitments and other variables such as the pupil-teacher ratio. Demand variables, among others, are income, adult literacy and the level of urbanization (Dreher, Nunnenkamp, and Thiele 2008, 293).
Independent of the estimation method and robust to several sensitivity tests, they find that aid to the education sector significantly increases net enrolment rates: In the short-term, one additional dollar per capita of aid to the education sector increase enrolment rates by 0.26-0.29 %, depending on the estimation method. In the long-run, using lagged terms, enrolment rates even rise by 0.8 % per additional USD per capita of education aid (Dreher, Nunnenkamp, and Thiele 2008, 302). They include instrumental variables to control for endogeneity and run different robustness tests, such as excluding outliers, using levels instead of logs, changing some control variables and controlling for institutions using interaction terms. They conclude, that aid to the education sector indeed is effective and “modestly but not negligibly contributes to achieving universal education in developing countries” (Dreher, Nunnenkamp, and Thiele 2008, 308). However, there do remain weaknesses, such as the use of commitment data, which often overestimates aid flows and where the delay of actual disbursements is not known (see section III.A).
A similar study which addresses this issue is by Michaelowa and Weber (2007), who analyse the impact of aid to the education sector for a larger panel, using 5-year averages as well as annual data of disbursements. Interestingly, they find significant results only for the annual panel, where “an increase in aid for education by 1 % of GDP implies an increase in primary completion rates by 2.5 percentage points” (Michaelowa and Weber 2007, 16).
For the relationship between aid to the education sector and economic growth, the few existing studies are encouraging as well:
By dividing up aid even more, into primary, secondary and higher education and running different regressions by income group, Asiedu and Nandwa (2007) obtain detailed results about the mentioned relationship: They find, interestingly, that when controlling neither for the type of education nor for a country’s income level, education aid has no significant impact on growth. The same accounts if they only control for the type of education. However, by controlling for countries’ income levels, results change: It turns out that aid for primary education does significantly increase growth in low-income countries, but aid for secondary or higher education does not. In contrast, in middle-income countries, aid for primary or secondary education has significant negative impacts in growth (potentially due to already high enrolment rates), while aid for higher education has significant positive impacts. This leads to the conclusion, that aid needs to be addressed properly and that its effects can only be measured correctly if heterogeneity of aid is not ignored. (Asiedu and Nandwa 2007, 644). These findings are approved by a later study, focussing on 38 countries in Sub-Sahara Africa (Asiedu 2014)
For other sectors, the effects that sector-specific improvements have on economic growth are more controversial:
As already mentioned, the effect of improved health on economic growth is unclear in current literature.
This is also why the two existing studies on the effect of foreign aid to the health sector use health indicators such as infant mortality (Mishra and Newhouse 2009) or life expectancy (Williamson 2008) instead of economic growth as dependent variables.
Williamson (2008) runs a fixed effects model as well as an instrumental variable model, using lagged health aid as an instrument. In her regressions, she includes the typical control variables such as GDP per capita, a proxy for institutions, the level of urbanisation and sector-specific control variables such as the number of physicians. Quite untypically, she includes all 208 countries for which the World Bank collected data, independently of if they do get health aid or not (Williamson 2008, 190–193).
Using different estimation methods and running some robustness checks, the study does not find links from health aid to any of the health indicators. The aid variable even turns negative in some regressions and never is significant (Williamson 2008, 194). However, it must be mentioned that in most of her models, neither GDP per capita nor the institutional level enter significantly, which is inconsistent with most of the literature on the relationship between economic development and health.
Interestingly, Mishra and Newhouse (2009) had the same idea at the same time (as neither of the studies cite the other one, it seems they were published without knowing of each other). They used the same time horizon, but a limited sample of only 118 countries and tried to eliminate endogeneity through lags as well. Their results, when analysing the impact of lagged health aid on infant mortality and controlling for war, population size, GDP per capita, fertility and HIV prevalence (all variables lagged), suggest that health aid does in fact lower infant mortality significantly. From the regression it follows that health aid per capita will lead to about 1.5 fewer infant deaths per 1,000 birth and the authors conclude that worldwide, “doubling health aid would save approximately 170,000 infants per year” (Mishra and Newhouse 2009, 862). It does, however, not affect life expectancy at all, leading to the authors’ interpretation that health aid does reduce infant mortality more than adult mortality (Mishra and Newhouse 2009, 866).
The two studies are interesting examples for the weaknesses of regression analysis, as two similar studies come to completely different results. This also shows that disaggregating aid does not guarantee robust results and that independent reproduction is always needed to approve empirical work.
Other sectors, such as climate protection or infrastructure are not yet covered by the literature. This bachelor thesis is the first one that specifically analyses the impact of foreign aid to the energy sector on sector-specific outcome indicators. I chose level (3) of disaggregation, because the link between electricity supply and economic growth still is unclear, as it will be explained in section II.D.
C. The Electricity Sector Worldwide: An Overview
In 2012, 15 % of the world population, meaning about 1.1 billion people, did not have access to electricity (IEA and World Bank 2015, 2). Above all in Sub-Saharan Africa and Oceania, the electrification rate (Percentage of population that has access to electricity) was still tremendously low, with 35 % and 29 % respectively. Oceania has the lowest rate, however, only represents seven million people without electricity in contrast to 589 million people in Sub-Saharan Africa. There is progress though: The global electrification rate rose from 76 % in 1990 to the mentioned 85 % in 2012 and the remaining deficit is mainly concentrated on 20 countries which account for 83 % of the global access deficit (IEA and World Bank 2015, 42–45), with India (263 million people without access) being the most important of those countries. India also is the country where most people (nearly 55 million between 2010 und 2012) gained access to electricity for the first time recently.
But not only the rate of access to electricity is improving, also the total production of electricity in Gigawatt hours (GWh) worldwide has been growing over the last decades from 5,267,771 GWh in 1971 to 11,873,248 GWh in 1990 and 23,391,299 GWh in 2013, representing an average annual growth rate of 3.4 %. Besides, in this period, the share of production of non-OECD countries in this number has more than doubled from 1,420,154 GWh (26 %) in 1971 to 12,532,785 GWh (53 %) in 2013. And while the total figure has tripled for OECD countries from 1971-2013, it is about nine times higher for non-OECD countries in 2013 than it was in 1971 (OECD 2015b, III, 33). As electricity production grew on average 5.2 % per year in non-OECD countries and only 2.2 % in OECD countries, non-OECD countries produced, for the first time in history, more electricity than OECD countries in 2012 and continue to do so since then. (OECD 2015a, II.3)
However, regional differences are high as well throughout non-OECD countries: While total production was nearly 20 times higher in Asia in 2013 compared to 1971, the figure was only about eight times higher for Africa. (OECD 2015b, III.31–III.33). This is mainly due to the fact that electricity consumption grew at an annual rate of 9.1 % in People’s Republic of China between 1973 and 2013 (OECD 2015a, II.14).
In spite of this progress, there still is a lack in electricity access and there still exist huge differences in electricity production between OECD and non-OECD countries: In 2013, OECD countries still produced nearly the same amount as non-OECD countries (OECD 2015b, III.31–III.33), in spite of the fact that OECD countries only represent about 1.2 billion people or less than 18 % of the world’s population (OECD 2016c, 8).
Taking a look at electricity consumption per capita also reveals huge global differences: Sub-Saharan Africa above all has extremely low figures and if South Africa is excluded, they look even worse: Consumption in Sub-Sahara Africa then averages around 162 kilowatt-hours (kWh) per capita per year, compared to the global average of 7,000 kWh and “it would take the average Tanzanian around eight years to consume as much electricity as an American uses in one month.“ (Africa Progress Panel 2015, 16)
Worldwide electricity production mainly comes from fossil fuels, which accounted for 67.2 % of total production in 2013, while hydroelectric plants provided 16.6 %, nuclear plants 10.6 % and biofuels, waste, geothermal, solar, wind and other sources accounted for the remaining 5.7 %. (OECD 2015a, II.3)
In conclusion, it can be said that in spite of the progress throughout recent years, the challenges in improving electricity supply remain huge and the disparities with regard to production and consumption of electricity are only disappearing slowly.
D. The Nexuses of Electricity Supply
That electricity supply is one of the most basic needs of any modern society is undoubted, as mentioned in section I.A and acknowledged by the United Nations through SDG No. 7 (United Nations 2015b).
Supplying people with electricity is not a purpose of its own, but electricity supply is related to many other issues that countries are facing worldwide. In this section, I want to display and explore some of this so called nexuses that could possibly exist.
Obviously, having access to electricity already increases personal welfare because it allows the use of lights, machines, communication devices and many other. But through various other channels, electricity supply can improve human welfare as well.
Firstly, health institution such as clinics need reliable electricity supply to function properly (IEA and World Bank 2015, 28). Medical devices, machines but also refrigerators for the storage of medication need electricity. Adair-Rohani et al. (2013) found in a cross-country study of 11 countries in Sub-Sahara Africa that only 34 % of health facilities had reliable access to electricity and 36 % did not have access to electricity at all.
Furthermore, if people use electricity instead of biomass for cooking and heating, local air pollution is reduced which can also improve people’s health conditions. Another channel could be, that by replacing kerosene lamps or solid fuel cook stoves, electricity might decrease the number of domestic injuries (IEA and World Bank 2015, 261).
Electricity can thus be seen as a prerequisite for the improvement of human health.
The education sector can also profit from electricity supply, as it allows schools to have modern services such as computers. In addition, lights in household make studying possible even if it is dark outside. Through the channel of improved lighting, such as the availability of street lights, electricity might also improve overall security, as dark streets are reduced. (IEA and World Bank 2015, 267). In a case study for Tanzania, Cecelski et al. (2005, 13) find that “the rate of female enrolment caught up with that of boys after households, schools, and public streets were electrified.“
This directly leads to another possibly nexus: Gender.
Increased security due to street lights especially improves living conditions of women and girls, as it can protect them from gender-based violence and thus improves their mobility as mentioned. But also the other electricity nexuses mentioned have a special relation to gender: According to IEA and World Bank (2015, 267–270), women and children suffer the most from indoor air pollution, women profit most from modern cooking solutions as they are the main wood-collectors and they profit most from washing machines and other household devices as it gives them more time for leisure. Improved health services furthermore increase the number of successful childbirth deliveries or sterilizations and electric water pumps reduce the time that women spend to fetch water. The background of this is that woman are seen as particularly time poor, because „they must systematically add up domestic and care duties (reproductive work) to their ... productive work so that this double time-budget makes of time a resource which is more scarce for women than for men“ (Charmes 2006, 67). By reducing the amount of time that women spend for these activities every day, electricity can actively improve women’s lives.
This, however, has to be handled with caution, as the final decision on how to use the newly gained free time relies on the household and the specific circumstances: For example, “electric light may improve the quality of life for some by allowing reading, entertainment or education via radio or television, while for others it may simply expand the working day“ (IEA and World Bank 2015, 267).
From an economic point of view, the obviously most interesting nexus of electricity supply is the one with economic growth. This is also, why the majority of the literature on the nexuses of electricity supply focuses on this relationship. By using different methods, such as time-series or panel data and different datasets, researchers are trying to find the link between electricity consumption per capita in kilowatt hours (kWh) and economic growth. As the direction of causality is unclear, data availability is limited and both, electricity consumption as well as economic growth might be influenced by many different aspects, finding a clear link has found to be difficult.
The possible causality is commonly categorized as follows: (1) growth hypothesis, which states that higher electricity consumption causally leads to economic growth, (2) conservation hypothesis, saying that the links goes unidirectional from economic growth to electricity consumption, (3) neutrality hypothesis, meaning that neither of the two variables affects the other, and (4) feedback hypothesis, which comes to the conclusion that there is a bidirectional causality (Payne 2010, 723).
If there is causality from electricity consumption to economic growth, it might work through different channels: The most important one probably is productivity for companies, as improved supply of electricity enables them to have better planning of their production. In a survey of more than 130,000 companies, the World Bank (2016a) finds that, on average, managers in Sub-Sahara Africa report a loss of sales of 5.5 % yearly due to power outages, in South Asia this number even is 6.6 %, compared to 0.1 % in OECD countries. Besides, reliable electricity supply makes expensive generators unnecessary and thus improves efficiency and increase profitability. But also through some of the channels already mentioned such as reducing domestic work and enabling people to focus more on professional work, empowering women or through improved education, improved electricity supply could at least in the long term foster economic growth.
In a literature survey of 34 studies, which analyses the relationship between electricity consumption and economic growth, Payne (2010) finds that 23% of the assessed studies favoured growth hypothesis, 31 % supported neutrality, 28 % conservation hypothesis and 18 % feedback hypothesis; meaning that only a minority of 41 % of the studies sees a causal link running from electricity consumption to economic growth. Most of the literature in this survey uses causality tests such as the Granger or the Johansen-Juselius test in order to find the relevant causal links.
These results have to be interpreted with care, however, as the survey simply aggregates studies of single countries as well as cross-country studies, ignoring regional and economic differences across countries. That these differences exist, was firstly shown by Ferguson, Wilkinson, and Hill (2000), using simple correlation analysis: While the correlation between the two variables was above 0.9 for all OECD countries apart from Mexico and Czech Republic, three quarters of non-OECD countries did not show such a strong relationship.
In one of the most recent studies on this topic (Karanfil and Li 2015), this issue is approached by dividing a data panel of 160 countries from 1990-2010 into various subsamples, categorized by region, income level and OECD membership. By also controlling for electricity imports and the level of urbanization, the authors find that neutrality hypothesis does not hold for any of their nine panels in the long run, whereas proof for feedback hypothesis is found for most panels. The results in the short run vary largely depending on the panel, but they mostly support conservation hypothesis, stating that there is a unidirectional causality from economic growth to electricity consumption. From these heterogenic results, it is difficult to draw a clear conclusion with regard to the direction of causality. The authors themselves conclude with the remark that while in the long run, feedback hypothesis is found in most panels, in the short run conservation hypothesis seems most appropriate (Karanfil and Li 2015, 272–276).
From the literature, it can thus be summarized that there is an important and most probably bidirectional link between electricity consumption and economic growth, but it seems to vary largely depending on time-horizon, region and income level.
Due to this unclear link, my following empirical analysis will focus on sector specific indicators of the electricity sector in order to get an insight in the effectiveness of foreign aid to the energy sector. Whether improved electricity supply will also foster economic growth, however, is still open for discussion.
 Throughout this text, “electricity supply” is used to describe the general ability of a country to supply its population with electricity. It must not be confused with any of the variables in the empirical model.
 The SDGs succeed the Millennium Development Goals since 2015 and are the most important international framework on foreign aid: 17 goals define what shall be achieved in international development until 2030.
 See for example Lorentzen, McMillan, and Wacziarg (2008) and Acemoglu and Johnson (2007) for different points of view on this matter
 According to Svensson (2006, 81), Swedish International Development Agency (SIDA), for example, lists eight goals of development cooperation: human rights, democracy, gender equality, protection of the environment, economic growth, social progress, security and global initiatives to protect the environment and combat contagious diseases
 In the same period, world population grew at 1.2 % per year on average (United Nations 2015a), this must be subtracted to get net production growth rates