ABSTRACT
1. Introduction
it is shocking to discover that roughly a third of the countries in Sub‐Saharan Africa (16 countries in all) had higher per capita GDPs in the early 1960s then they do three and a half decades later.” (Rodrik, 1998, p. 9) Economically, Sub‐Saharan Africa is seriously underdeveloped. Average GDP per capita is only 2,466.531 dollar a year, as opposed to 13,405.26 in the rest of the world in 2012.Why is this so? The question of Sub‐Saharan Africa's slow economic development has puzzled many economists. In general, it is thought that Sub‐Saharan countries lag economically because of the lack of state capacity and institutional equality (Englebert, 2000). Ineffective governments cannot create the proper conditions for economic growth.
Nowhere can this be seen more clearly than in the level of infrastructure in Africa. The 47 countries of Sub‐Saharan Africa (with a combined population of 800 million) together generate roughly the same amount of electrical power as Spain (with a population of 45 million). One‐third of Africans living in rural areas are within two kilometers of an all‐season road, as opposed to two‐third of the population in other developing regions across the world.
With regard to the ICT sector, Africa is staying closer to developments in the rest of the world. The percentage of population living within range of a GSM signal rose dramatically from 5 percent in 1999 to 57 percent in 2006 (Foster, 2008).
In itself, infrastructure is an important determinant of economical development (Easterly and Levine, 1997). Infrastructure is seen as a major constraint on doing business in African countries (Foster 2008). , Foster concludes that poor infrastructure depresses firm productivity by around 40 percent. The power sector was deemed the most limiting factor for business.
1 http://www.imf.org/external/pubs/ft/weo/2013/01/weodata/index.aspx, retrieved
Why is African infrastructure so poorly developed? In this research, I will research the link between the level of infrastructure and the ethnic composition of Sub‐Saharan countries. Africa is the most ethnically diverse continent in the world. Problems that have arisen due to this high ethnic diversity are social polarization and nepotism among ethnic groups in Sub‐Saharan Africa.
Recent research in the United States has found that agreement on public policies and investment is negatively correlated with ethnic fragmentation (Cutler and Glaeser 1997; Alesina, Baqir, Easterly, 1999; and Lutmer, 2001). In Africa, ethnically the most diverse continent in the world, this correlation is supposedly even more pronounced. I hypothesize that amount and sizes of ethnic groups present in an area are important factors in the provision of infrastructure in Africa. This is partly based on the work of Cheikbossian (2007), who found that the probability of civil war – the epitome of ethnic conflict‐ is maximized when the ethnic composition of a country consists out of two or three ethnic groups of roughly the same size.
In this study I classified and subdivided three different ethnic compositions to investigate if this new ethnic measurement provides more answers on the relation between ethnic tension and the level of infrastructure in Sub‐Saharan Africa.
Research question
Does the ethnic composition of a country in Sub‐Saharan Africa influence the level of infrastructure?
2. Literature review
I start the literature review with a discussion on why Sub‐Saharan Africa has lagged behind, focussing on the quality of governance and the determinants of quality. Afterwards I zoom in on the role of infrastructure, an important task of the government which is essential in generating economical growth: what factors can explain the quality of infrastructure. Finally, I establish relation between ethnicity and level of infrastructure, subsequently, I argue that not a high amount of ethnic diversity is of influence but the ethnic composition of country influences the level of infrastructure.
The development of the continent Sub‐Saharan Africa has lagged behind in comparison to other regions since the end of the colonial period. There is a wide body of literature that explain this perfomance. Consensus is reached that determining factors in the lack of economical development in Sub‐Saharan Africa are the lack of state capacity and institutional quality (North 1981, Ndulu and van der Walle 1996, Lewis 1996). This lack of state capacity to refers to a ”state’s capacity to design and implement policies,
implement policies, make credible comments, run an efficient bureaucracy and provide constraints to opporunitisic behavior” (Englebert 2000:8). Institutional quality is defined
by Borrmann, Busse and Neuhaus (2006:2) as the ability to “decrease information
assymetries by channeling information about market conditions, goods and participants”.
The empirical research on links between economical growth and institutional efficiency of Mauro (1995), Knack and Kneefer (1995) support that the policies and quality of institutions matter for the development of a country.
Moving to the determinants of Africa’s weak state capacity. There are three different theories on the determinants.
The first theory use a political theory to explain Africa’s weak state‐building (Clapham 1982, Boone 1994, van de Walle 1994 and Lewis 1996). African leaders rule with a divide‐and‐rule strategy with a focus on extortion (Illorah, 2009). This weakens the capacity of state and neglects growth‐enhancing policies.
social capital and can be translated to “the compound of trust, norms of reciprocity,
participation and equality and of associative life” (Englebert 2000:9). Translating this
theory to Africa’s stagnation would mean that poor governance and weak state building would come from a lack of civic culture and thus low social capital due to three things (Seragaldin and Taboroff (1994): weight of tradition blocks a nation‐wide mindset, strong ethnic differences block a sense of trust and distance to elite blocks effective political participation and equality.
The final theory of weak state capacity can be found in the proposition that African countries have weak institutions and adopt poor policies due to the ethnic diversity in this continent. This theorisation finds its offspring in the work of Easterly and Levine (1997). They found that ethnic diverisity in a country has led to social polarisation. This degree of social conflict increased the sub‐optimal policies as ethnic representatives in government fail to serve the country’s needs instead of their own ethnic groups needs.
One of the consequences of weak state building or economical development in the past, where Sub‐Saharan Africa struggles with today, is the poor state of infrastructure (Rodrik, 2008). A well developed infrastructure in a country is a product of both good governance (De, 2000) and an important determinant of economical growth (Easterly and Levine 1997) in Sub‐Saharan Africa.
Past research on developing infrastructure has demonstrated that for a country, demographic factors, geographic factors as well as finances may influence the level of infrastructure in a country (Bell, 2011).
The demographic factors that influence infrastructure are population size (Heller, 2010) “The larger a population in a country, the greater the need for capacity” Heller, 2010, p.5 and the age structure of the population. The age structure influences the demand for different things in infrastructure. As an example Heller (2010, p5) gives “A young
population implies, all other things equal, a greater demand for education services versus a largely elderly population calls for things like: availability for long‐term facility, elderly friendly transport etcetera.”
A different study on infrastructure: McDaniels, Chang, Cole, Mikawoz and Longstaff (2008) found that one geographic factor is influencing the level of infrastructure: the exposure to extreme events. One important point they made in their study is that infrastructure of a country often exhibits extreme levels of vulnerability to non‐planned events. These non‐planned events can be events like floods, earthquakes, a period of extreme drought, extreme temperatures etcetera. Bell (2011) states that the finances of a country may influence the level of infrastructure. The ability of a country to self‐generate finances is one of the key elements in developing infrastructure according to a United Nations (2009) research. A country in a developing region has several methods to be able to self‐generate finances (United Nations, 2009). In Africa two main strategies can be roughly identified2. The first strategy is based on
the amount of natural resources present and to what extent they can be exploited to finance infrastructure development. The second strategy is the use of public‐private agreements (Pessoa, 2008), by relying on Official Development Assistance (ODA) to enhance the quality of infrastructure. Now that the determinants of a quality infrastructure are given, I will provide the link between level of infrastructure and ethnicity. Political economy models (Alesina and Rodrik 2004, Alesina and Spolaore 1995) suggest that ethnic diverse societies are prone to competitive rent‐seeking and have difficulty agreeing on public goods and services. Easterly and Levine find evidence for these models in that high amount of ethnic diversity has a negative influence on level of infrastructure. When looking at the level of infrastructure of a country these two suggestions : competitive rent‐seeking and difficulty on agreeing has several implications. Competitive rent‐seeking is explained by Osborn (2000:509) as “members
of ethnic groups devote their resource endowments to engage in activity with other members of the ethnic group”. The results of this rent‐seeking in Sub‐Saharan Africa is
found to be distorting investment in public sector projects (Mauro, 1995) and thus
2http://www.africaminingvision.org/amv_resources/AMV/Financing%20mining%20re
lowers level of infrastructure (Montalvo and Reynal‐Querol, 2005). This relation on infrastructure is found to be even stronger in the case of an ethnic conflict (Osborn, 2000). Kenny (2006) finds evidence for rent‐seeking activities in infrastructure in his analysis on corruption in infrastructure Africa by looking at surveys. The difficulty of agreeing between ethnic groups has an impact on level of infrastructure due to slow decision processes. A large body of research is done in why in some cases the provision of public goods and services, such as infrastructure, is not effective in the United States. In the study of Alesina and Laferrara (2000) the results show that each ethnic group places significantly different values on a potential set of infrastructure, and in line with this, Vigdor (2001) finds that each ethnic groups places little value on the potential needs of a different ethnic group. These two factors have two implications for the decision process of where to invest public money: If the ethnic groups present differ in preferences the results of these studies indicate that less money will be invested in infrastructure. Next to this factor the investment in infrastructure is also found to be done sooner when there the ethnic composition of the area consists either out of multiple small ethnic groups present (found to be more likely to cooperate) or out one ethnic group that is dominant is size (less need for mutual agreement). Results of empirical studies of Cutler and Glaeser (1997), Lutmer (2001) and Alesina et al (1993) in areas within United States confirm that the shares of public spending devoted to infrastructure is negatively influenced by ethnic fragmentation.
Combining the elements of the previous study of ethnic diversity on infrastructure in Sub‐Saharan Africa (Easterly and Levine 1997) and new insights adopted from research out of the United States that are proved by Cheikbossian and Collier and Hoefller I suspect that ethnic composition has an influence on the level of infrastructure in Sub‐ Saharan Africa. The research done in the United States shows that there are three ethnic composition to be defined in the world: • Multiple small ethnic groups, • Two or three ethnic groups of roughly the same size • One ethnic group dominating in size. In these three categories I expect, given the outcomes of the studies in the United States, that in countries where the ethnic composition consists out of two or three ethnic groups of roughly the same size the level of infrastructure is lower due to slower decision process and more ethnic tension than in countries with either an ethnic composition of multiple small ethnic groups (who are in nature more cooperative) and countries where one ethnic group is dominating (less need for mutual agreement). Therefore the following schematic overview is set up. 2.1 Schematic overview relationship In the figure 1 infrastructure is translated to four elements: communication, electricity, medical services and transport. Ethnic composition is categorised in three categories: ‘Low’ meaning: ethnic composition consists out of multiple small ethnic groups, ‘Medium’ group: ethnic composition consists out of 2 or 3 large ethnic groups of similar size and the ‘High’ group: ethnic composition consitst out of one ethnic group in power. The red square indicates the potential ethnic conflict / tension zone, where the infrastructure spending is lower due to the ethnic tension and difficulty agreeing on provision of infrastructure.
figure 1: schematic overview (author) The relationship between ethnic composition and the level of infrastructure in Sub‐ Saharan Africa is u‐shaped. Hypotheses
1. When the ethnic composition of a country in Sub‐Sarahan Africa consists of multiple small ethnic groups there should be ‘normal’ infrastructure spending since ethnic tension lower and more cooperative nature between small ethnic groups. 2. When the ethnic composition of a country in Sub‐Saharan Africa consists of 2 or 3
large ethnic groups of roughly the same size there should be less developed infrastructure since ethnic tension is the highest and slower decision process.
3. Methodology
The previous section has outlined the theory which has been written on the topic of the ethnic composition and infrastructure. This section will describe the methodology which will be used to test the hypotheses. 3.1. Measures and Data collection This section will explain the measures of all the variables used in the statistical model described above. It will also explain the data sources from which the data is collected. 3.1.1. Ethnic composition3
The ethnic composition in a country is translated into the Political Relevant Ethnic Group index made by Daniel Posner (2004). This index provides data about the ethnic composition of the Sub‐Saharan continent per country. Posner (2004) constructed the Political Relevant Ethnic Group (PREG) index for every country in Sub‐Saharan Africa in several steps. He used the ethnic breakdown provided in the Atlas counting (which is done in 1960 by the Ethno‐ logical Institute at the Department of Geodesy and Cartography of the State Geological Committee of the Soviet Union) as a baseline. Posner then conducted an exhaustive search for books, academic articles and news reports that described the ethnic politics, hereby specially focusing on sources that described the dynamics of competition over resources and political power among ethnic groups in a country. Posner and his research assistants followed the practice of continuing to consult additional references until they reached a point:
“where all the sources seemed to be mentioning the same ethnic groups as significant
participants in the competition over the country’s economic policies.” (Posner, 2004, p.
854)
After this point was reached Posner found it worthwhile to put the ethnic group in the dataset of political relevant ethnic groups.
The PREG index is an improved version of the Ethnic Fractionalization Index that is used by Easterly and Levine (1997). Posner (2004) identified a problem with the Ethnic fractionalization index. The ELF measure suffers from “grouping problems” according to Posner (2004). In the ELF count there are some groups that combine different ethnic
3 http://www.nsd.uib.no/macrodataguide/set.html?id=16&sub=1, retrieved on
groups into one ethnic group that are clearly different in ethnicity. For example, Kasfir (1976) states in his research that in Uganda the ‘Acholi’ and ‘Lango’ are combined in one ethnic group while there is a history of political rivalry going on between the groups. Numerous examples of this grouping problem can be given. The most remarkable grouping problem in the ELF data is the ‘Barundi’ ethnic group which consists out of Hutus and Tutsis, who fought each other in a bloody ethnic war in Rwanda.
Posner tested his new ethnic index by re‐doing the Easterly and Levine (1997) study and found that his PREG ethnic measurements are better suited for testing propositions about ethnic diversity than the ELF measurements. Posner finds the same claim as Easterly and Levine, that ethnic fractionalization is strongly negatively correlated to economic growth, but on much firmer theoretical and methodological footing than the results generated with ELF. Therefore I chose the PREG measure over the ELF measure. Group subdivision in the PREG measure In my research I used a new ethnic composition measure based on recent insights. For this measure it is needed to create three categories: ‘Low’: multiple small ethnic groups), ‘Medium‘: 2 or 3 ethnic groups of roughly similar size and ‘High’: one ethnic group dominating in size. Unfortunately there is no previous study on this new ethnic composition classification that has made an absolute distinction or formula to classify this ethnic variation in countries. To overcome this difficulty in subdividing the 47 countries of the Sub‐Saharan African continent into these categories, I looked at the Herfindahl concentration formula used in the Herfindahl‐Hirschmanindex. Posner (2004) calculated his PREG index by using the Herfindahl concentration formula.
small firms to a single monopolist. The Herfindahl‐Hirschman index is used as a screening tool to evaluate mergers by the United States Federal anti‐trust authorities.
It is because of the absence of existing ethnic classification that I decided to follow the guidelines of these United States Federal anti‐trust authorities for categorising the countries into the three groups (low, medium and high). The department of Justice considers Herfindahl indices below 0.15 to be not concentrated, between 0.15 and 0.25 to be moderately concentrated and indices above 0.25 to be highly concentrated.4
Translating this to the ethnic composition of a country’s population: when the PREG index is low: there are multiple ethnic groups of the same size (below .15). When the PREG index is medium (between 0.15‐0.25) the ethnic composition of a country consists out of 2 or 3 ethnic groups of roughly the same size. Whereas the PREG index is high (>0.25) the ethnic composition of a country consists out of one large ethnic group controlling the country. The PREG value of the country determines in what category the country is subdivided. If the PREG value lies below 0.15 the country is placed in the ‘Low’ group, when the PREG value lies between 0.15‐0.25 the country is placed in the ‘Medium’ group and final if the PREG value is above 0.25 the country is placed in the ‘High’ category.
As an example: Qatar consists of the following ethnic groups5: Arab (40%), Indian
(18%), Pakistani (18%), Iranian (10%) and other (14%). The Herfindahl formula (see
appendix I) is calculated as follows: .402 + .182 + .182 + .102 + .142= .2544, thus falls in the
medium category. The United States consists of White (79.96%), Black (12.95%), Asian (4.43%), Amerindian and Alaska natives (0.97%), native and other Pacific islander (0.18%) and two or more races (1.61%)6. The ethnic composition there is .79962 +
3.1.2. Level of infrastructure
To determine the level of infrastructure of a country I have to define what infrastructure is. According to the World Bank: “Infrastructure helps to determine the success of
manufacturing and agricultural activities. Investments in water, energy, housing, transport and investments in communication also support economic growth7”. Quite a wide definition, in this study I chose to narrow it down to four categories: ‘Communication’, ‘Electricity’, ‘Health services’ and ‘Roads’. Communication consists out of indicators: mobile cellular subscriptions and internetusers per 100 inhabitants. ‘Electricity’ is translated to electricity access as percentage of population. ‘Health services’ is translated to health expenditure as percentage of GDP and final ‘Roads’ is translated to three indicators: Road density, paved road in good condition and unpaved road in good condition. There are more indicators to establish the level of infrastructure but in Sub‐Saharan Africa little data can be found for these additional measures. This selection of infrastructure indicators is picked because data was available for most countries in Sub‐Saharan Africa.
3.2. Sample
The sample group is divided in three categories of ethnic composition: low, medium and high. The ‘Low’ group consists out of countries that have an ethnic composition that consists out of multiple small ethnic groups with roughly the same size. In the ‘Medium’ group countries are placed with an ethnic composition that consists out of 2 or 3 ethnic groups of roughly the same size. In the ‘High’ group countries are placed with an ethnic composition consisting of one large ethnic group and possible small minority ethnic groups.
In table 1 there is a subdivision of ethnic composition in Sub‐Saharan Africa. The sample group is out of balance. In the low and medium group there are both 9 countries while in the high group there are 24 countries.
Frequency Percent Valid Percent Cumulative percent
Subdivision Ethnic composition PREG Sub‐Saharan Africa Low Medium High 9 9 24 19.1 19.1 51.1 21.4 21. 57.1 21.4 42.9 100 Total 42 89.4 100 Missing 5 10.6 Total 47 100 Subdivision Ethnic composition worldwide Low Medium High 9 16 161 4.8 8.6 86.6 4.8 8.6 86.6 4.8 13.4 100 Total 186 100 Table 1: Subdivision ethnic composition In order to create a more balanced dataset I tried to include more countries since I am testing a structural relation by calculating and adding the ethnic composition of 186 countries worldwide. Since the Political Relevant Ethnic Group Index is only focused on the Sub‐Saharan African continent I retrieved the ethnic data of these 186 countries via CIA world factbook8 and calculated the ethnic composition with the Herfindahl
concentration formula.
In the overview of table 1 there is a subdivision of the categories of ethnic composition worlwide, showing an even larger imbalance in the sample group. There are only 9 countries in the world with an ethnic composition that consists of multiple small ethnic groups, 16 countries that have an ethnic composition of 2 or 3 large ethnic groups of
8 https://www.cia.gov/library/publications/the‐world‐factbook/, retrieved at April 26,
roughly the same size and 161 countries that have one large ethnic group and one or multiple small ethnic groups.
I left out the additional 7 countries in the medium group: (Afghanistan, Bolivia, Fiji, Libya, Yugoslavia (pre 1991), Suriname and Qatar. It turns out the only additional country in the medium group for which data is available is Bolivia. Only adding Bolivia to the medium group does not expand the sample group in a meaningful manner therefore I solely focus on the Sub‐Saharan Africa continent.
3.3 Limitations dataset
There are limitations in the dataset. First, due to data availability issues of the continent Sub‐Saharan Africa: Somalia was dropped from the sample because there was no data available about Somalia. In the cases of ‘Electricity’ and ‘Health services’ only one indicator could be found that had data of most countries in the Sub‐Saharan continent. In the cases of ‘Electricity’ and ‘Transport’ it was not possible to calculate a 10‐year average because the Electricity indicator was only published in the year 2009 and the Roads indicators only published in 2008. Not all countries in Sub‐Saharan Africa published the data about Electricity and Transport. The rest of the indicators had most data but in some cases there were data gaps. In this case I took the average of the numbers of years in which data was published.
The missing data entry points have an impact on the generalizability of this study. In
Summary of the variables described above
Indicators (all Interval data) Available Missing Data source
3.4 Control variable
In order to test for these factors I constructed several measurements to control the alternative influences on the level of infrastructure. The alternative influences can be subdivided into two categories: the demographic events and the financial side of a country.
It is important to control for the demographic events that have a large impact on the level of infrastructure. In the literature review I mentioned three demographic factors that influence the level of infrastructure: population size, age structure and extreme events. In this study I decided to solely focus on the resistance to extreme weather events due to the indicators used in my study, that are interlinked with either population size (per 100 inhabitants) or economy size (as a percentage of GDP).
By controlling for these non‐planned events in the form of ‘Extreme events’ I eliminate the possibility that in the years collected, a non‐planned event happened that had a devastating effect on roads, medical services, access to electricity or communication. This way I exclude influences on the level of infrastructure which are not related to the ethnical composition in a country. The indicator ‘extreme events’ is translated as droughts, floods and extreme temperatures happened in the period 1990‐2009. The indicator used is derived from World Bank, called “droughts, floods, extreme
temperatures (% of population affected)”.
4. Empirical results
This section will describe the empirical results of the previously determined methods.
4.1 Descriptive statististics
The table below shows the descriptive statistics of the data. The standard deviation of the indicators ‘Electricity access (% of population)’, ‘Classified paved road in good condition’, ‘Classified unpaved road network in good condition’ and ‘Mobile cellular subscriptions’ is quite large, indicating that the level of infrastructure amongst sub‐ Saharan Africa countries vary. The total valid N (listwise) is 18, which unfortunately is low but due to data availability issues it cannot be improved.
Table 3. Descriptive statistics
Methodological problems low amount of cases
In this study there are only 18 complete cases due to data availability issues. The indicators ‘Extreme events’, ‘Health expenditure’, Internetusers’, ‘Mobile cellular subscriptions’ and ‘Net ODA per capita’ each contain more than 30 cases. The indicator ‘Electricity access’ contains 25 cases while the indicators: ‘Paved road’, ‘Road density’ and ‘Unpaved road’ each contain 18 cases. This sets the study for some methodological problems. The first problem that arises is a selection bias (Keller, 2008). A selection bias in this case refers to the distortion of a statistical analysis due to low amount of complete cases. There are 18 complete cases while there are 42 countries in the Sub‐ Saharan African continent. This impacts the outcomes of the statistical tests performed, which has to be noted. Another problem that arises due to the low amount of cases is how to test the alternative explanations of the results. A general rule of thumb is that there have to be 15 cases for every independent variable. I bypassed this by creating a model per every control variable.
Variables N Min Max Mean Sd
Subdivisions PREG score
To test my hypotheses I made a subdivision within the PREG index. As noted before I have made three different categories; 1, 2 and 3. The following table shows the number of countries in the three categories and the five excluded countries: Cape Verde, Comoros, Djibouti, Eritrea, Mauritania and South Sudan.
Frequency Percent Valid Percent Cumulative percent
1 2 3 9 9 24 19.10 19.10 51.10 21.40 21.40 57.10 21.4 42.9 100 Total 42 89.40 100 Missing 5 10.60 Total 47 100
Table 4. Subdivision PREG score
4.2. Inferential statistics
The data with more than 30 data entry points is tested using the ANOVA (Analysis Of Variance) since inferences have to be made about three categories; low, medium and high. The data with less than 30 data entry points is tested using a non‐parametric test: the Kruskal‐Wallis test.
The ANOVA test tells us whether there are significant differences in the mean scores on the dependent variable across the three groups (Keller, 2008). In the following I will test as a dependent variable: Mobile cellular subscriptions, Internet users and Health expenditure.
ANOVA results
Table 5. ANOVA descriptives
Low Medium High Health expenditure Mobile cellular subscriptions per 100 inhabitants Internetusers per 100 inhabitants
In figure 2, a graph is plotted to clear up the exact steepness of the u‐shape in these three cases.
figure 2. Steepness u-shape
All the indicators tested in this ANOVA test are not significant: ‘Internetusers per 100 inhabitants’: 0.53, ‘Mobile cellular subscriptions’: 0.70 and ‘Health expenditure’: 0.39 (all are α>0.05) and thus it has no use looking at the post‐hoc tests generated (Pallant, 2007:, p. 246).
Next, the data indicators with below 30 data point entries are tested with a non‐ parametric test: Kruskal‐Wallis test. The Kruskal‐Wallis test is the non‐parametric alternative to an ANOVA test (Pallant, 2007, p.226).
Table 6. Descriptives Krusal Wallis test
Figure 3. Steepness u-shape
The Kruskal‐Wallis test showed that none of the indicators is significant: Electricity access .55, Roadnetwork density: 0.69, Paved road: 0.22 and Unpaved road: 0.18 (>α =0.05). So there are no statistical differences between groups in electricity access (% of population) or in any indicator of transport. So concluding, all hypotheses have to be rejected since there are no significant differences in the mean across groups.
Low Medium High
4.3 Control variables
The previous section indicates that in the cases of ‘Internet users per 100 inhabitants’, ‘Health expenditure’ and ‘Transport there is a curvilinear relation, albeit not significant. To control for the different background of every country in the Sub‐Saharan continent I regressed the control variables ‘Extreme events’, ‘Natural resources rent’, ‘Net ODA per capita’ separately (due to the limitation of one control variable is 15 cases) with the dependent variable. Resulting in the table 7:
Table 7. Control variables regressed with WTS factor: PREG indicator score
5. Discussion
The results do not show significant statistical evidence that would suggest that there is a difference among the three groups; low, medium and high. However, it should be noted that by every indicator except ‘Electricity access’ and ‘Mobile cellular subscriptions’ the mean plot showed a u‐shaped relation. Although the relationship is weak and not significant the results do suggest that in some parts of infrastructure spending (‘Internet users’, ‘Health services’ and ‘Transport’) there is a curvilinear relationship. The countries with an ethnic composition of 2 or 3 large ethnic groups show a lower level of internetusers, health services and transport than in the countries where the ethnic composition of a country consists of multiple small ethnic groups or one ethnic group of a large size is controlling the rest. In the other two cases ‘Electricity’ and ‘Mobile cellular subscriptions’ the results suggest (although not significantly) that in countries where the ethnic composition consists out of multiple small ethnic groups the access to electricity is lower than when the ethnic composition consists out of 2 or 3 ethnic groups of the same size battling for power or one ethnic group of a large size controlling the rest. In countries where there are multiple small ethnic groups there are more mobile cellular subscriptions than in the countries with either 2 or 3 ethnic groups of the same size or with one ethnic group dominating. The indicators: ‘Internet users’, ‘Health services’ and ‘Transport’ show a u‐shaped figure when plotting the means of every group, indicating that there is a relation. None of the indicators however are statistical significant. The variance in group sizes in my sample: low: 9, medium group: 9 and high group: 24 can be seen as an explanation of why no indicator is significant. In an attempt to expand the dataset with more countries to prove the structural analysis that ethnic composition influences the level of infrastructure, I collected ethnic data10 from 186 countries worldwide and calculated the ethnic
composition with the Herfindahl formula. When categorising these 186 countries worldwide I found out that only 9 countries worldwide have an ethnic composition that would fit into my ‘low’ category. And all these 9 countries are in my data sample. So I
10 https://www.cia.gov/library/publications/the‐world‐factbook/, retrieved at April 26,
was unable to expand my dataset to statistically prove this relationship. The ‘medium’ group has 16 countries and the high group has 161 groups showing that 86% of the countries in the world have an ethnic composition of one ethnic group dominating the rest.
6. Conclusion
16 years after Easterly and Levine found a relation between high amount of ethnic diversity and low provision of infrastructure in Sub‐Saharan Africa is a good moment to determine if groups and sizes of ethnic groups present in a country influence the level of economical development. In this study I researched whether the ethnic composition of a country has played a role in the level of infrastructure This paper stated the following research question: Does the ethnic composition of a country influence the level of infrastructure?
To answer this question a dataset was developed containing 42 countries in the Sub‐ Saharan continent. Whenever possible a 10‐year average was created from the years 2002‐2011 to overcome yearly effects and improve the quality and generalisility of the study. From theory the following control variables are selected: ‘Extreme events’ to control for non‐planned events, ‘Net ODA per capita’ to control for loans and grants for economic development and final the total ‘Resources rents’ to control for a country’s ability to generate finances.
The level of infrastructure is not significant influenced by the ethnic composition in a country in this study. In the cases: ‘Internet users’, ‘Medical services’ and ‘Transport’ a u‐ shaped relationship is found, indicating that my hypotheses for these make sense, although ultimately they have to be rejected. In the cases of ‘Mobile cellular subscriptions’ and ‘Access to electricity’ there was no u‐shaped relationship when plotting the means.
little data available online about the level of infrastructure in the Sub‐Saharan African continent. The final reason that can explain the lack of significance is the use of the Herfindahl concentration formula as a tool to calculate a score for the ethnic composition of a country.
Two attempts to find additional evidence to support the curvilinear relation by looking at cities instead of countries in Sub‐Saharan Africa or religion instead of ethnicity in Sub‐ Saharan Africa were unsuccessful. There is no data available for infrastructure in cities
in Sub‐Saharan Africa. The groups and sizes of religion I retrieved from CIA world
factbook11 and calculated the religious composition with the Herfindahl concentration
formula. As an example: Gambia consists of12 Muslim (90%), Christian (8%) and
Indigenious belief (2%). The religious composition there is .902 + .082 + 0.022 = .82, thus falls in the high category. Resulting in an even more imbalanced dataset, as can be seen in table 8. Frequency Percent Low Medium High 0 0 45 0 0 95.74 Total 45 95.74 Missing 2 4.26 Total 47 100
Table 8. Subdivision religious composition SSA
Showing that in 45 countries in the Sub‐Saharan African continent there is a religious composition where one religion is dominating the rest, 2 countries did not publish exact figures about religion (Cape Verde and Mauritania). The imbalance in the dataset makes it impossible to find additional support for the curvilinear relation.
6.1 Limitations As with all scientific research, there is no perfect way of conducting research. Therefore, this study has also some methodological and theoretical limitations that might have an influence on the result. Firstly, there are the limitations to the sample. This quantitative study was conducted on an imbalanced sample size per group (low group: 9, medium group: 9, and high group: 24) and the amount of data gaps is high due to little data captured about Sub‐Saharan Africa. In retrospect, the dataset of this study is more suitable for a qualitative study of Sub‐Saharan Africa than for a quantitative study. In the case of Sub‐Saharan Africa the quantitative cross‐country study is a rough way to summarise complex histories of politics, economies and ethnic groups where qualitative research would be more suitable to address the specific country related sensitivities in these matters.
Secondly, the PREG index used has a methodological weakness since at the base of the Political Relevant Ethnic Group index lays the Atlas base count, which is done in 1960 by the Ethno‐ logical Institute at the Department of Geodesy and Cartography of the State Geological Committee of the Soviet Union and therefore is out of date. Next to this weakness, the PREG index is calculated on Herfindahl index, not 100 % accurate on ethnic composition since this measure is developed to evaluate mergers not on indicating the various ethnic group sizes and their relation amongst each other. Therefore it does not take explicit account of either the degree of concentration of the ethnic groups in the country or the exact depth of the division, which can possible bias the subdivision in the PREG index and thus the subdivision of the ethnic composition.
7. Further research and Managerial implications
Although much of the statistical evidence is absent it does provide some support for re‐ examining the current literature.
7.1 Further research
The study of Easterly and Levine (1997) was the first to point out ethnic diversity within a country as a cause of lagging growth rates. Others followed and included the ethnic diversity issues or the homogenous versus heterogeneous countries to expand the literature. In my research I have tried to find an explanation, by using a new classification of ethnic composition, in why some African countries (Botswana, South Africa etcetera) are successful and others (Somalia, CAD etcetera) still are fighting extreme poverty. What I suspected before conducting the research came true. There are so many factors in play in Africa that it is hard to indicate the most influential factor, and although my research indicated very small evidence that in most indicators there is a curvilinear relation between the ethnic composition and the level of infrastructure the results show that there are more factors at play. For example: the preferences of the ethnic groups in relation to the battle for power need to be examined to expand the literature about ethnic composition and the level of economic development in a country. The ethnic composition index (PREG) that I used is based on the Atlas index created by the Soviets in 1960. By taking this base count as a starting point for every ethnic fractionalization index, every researcher has to accept a limitation and the possible biasing of the results. By recounting this index, the results would be up to date and would generate more precise predictions about the relationship. Besides this index, in this study I used the Herfindahl formula to calculate the different ethnic compositions and used the scales of the United States Federal anti‐trust authorities to index the three categories: low, medium, high. These scales are used as a screening tool to evaluate mergers. Future research is needed to create a formula that exactly determines the ethnic composition of a country. This formula has to give a value that is representable for the exact degree of concentration of ethnic groups in the country and the exact depth of every ethnic group.
The combination of unavailability of data in Sub‐Saharan Africa and the complex nature of histories of politics and economics in Sub‐Saharan Africa call for a more qualitative approach in this topic. In most countries in Sub‐Saharan Africa there are so many factors at play regarding economical development that it is a challenge to do a complete cross‐ country quantitative analysis that does not miss any relevant point. To illustrate the complexity of this, I will give a brief example of the most successful countries in Sub‐ Saharan Africa: Botswana. In Botswana there is an ethnic composition that consists out of multiple small ethnic groups according to the PREG index. Botswana is one of the countries in this study that scores one of the highest scores on communication, electricity, health services and roads and thus has one of the highest levels of infrastructure in Sub‐Saharan Africa. One could conclude that Botswana is succesful due to the ethnic composition of Botswana but from the study of the African success story: Botswana of Acemoglu, Johnson and Robinson (2001) we know that the tribes in Botswana have a long history of cooperation among themselves before independence. This indicates that quantative cross‐country comparison is too rough in determing whether ethnic composition influences the level of infrastructure but asks for a more qualitative study. A study that does not only include ethnic composition as a influence in a country but in Botswana’s case, for starters, look at the history between ethnic groups and the mentality of the ethnic group present. 7.2. Managerial implications The findings of this study can have important implications for policy makers. Indicating that ethnic composition of a country can determine the level of infrastructure, although not statistically significant, can help politicians or ODA donors to become more aware of possible tensions among ethnic groups and the impact of that tension on a country’s economic development. This ethnic tension originates from the fact that these ethnic groups choose to strive for their own ethnic group instead of serving all citizens without special political favors or prejudice to any group on the basis of ethnicity. When overcoming this prejudice, countries in Africa will grow faster economically because they share one important unifying goal.
List of References
Foster, V. (2008) Africa Infrastructure Country Diagnostic, Worldbank, http://siteresources.worldbank.org/INTAFRICA/Resources/AICD_exec_summ_9‐30‐ 08a.pdf, retrieved at January 11, 2013. Acemoglu, D.S., Johnson, S and Robinson, J. (2001) An African success story: Botswana, in Analytical Country narratives of Economic Growth.
Alesina, A., Baqir, R., Easterly, W. (1999) Public goods and ethnic divisions. Quarterly Journal of Economics 114, p1243 – 1284 Alesina, A., La Ferrara, E. (2000) Participation in heterogeneous communities. Quarterly Journal of Economics 115, p847 – 904. Alesina, A., La Ferrara, E. (2002) The Determinants of Trust. Journal of Public Economics, in press. Alesina, A., Rodrik, (1994) Distributive Politics and Economic Growth, Quarterly Journal of Economics, p 465‐490
Alesina, A. Spoarlare, E. (1995) “On the Number and Size of Nations, The Quarterly
Journal of Economics, 112 (4), p1027 – 1056.
Atlas Narodov Mira (1964) Moscow: Miklukho‐Maklai Ethno‐ logical Institute at the Department of Geodesy and Cartography of the State Geological Committee of the Soviet Union Baliamoune‐Lutz, M. Ndikumana, L. (2008) Corruption and Growth: Exploring The Investment Channel, Department of Economics, Working paper 2008‐08 Bell, G. (2011) Infrastructure and nation‐building: the regulation and financing of Spain’s infrastructure 1720 – 2010, Business history, 53 (5), p688‐705
Borrmann, A., Busse, M., Neuhaus, S. (2006) Institutional Quality and the Gains from Trade, HWWA Discussion paper 341,
Cutler, D.M., Glaeser, E.L. (1997) Are Ghetto’s Good or Bad?, Quarterly Journal of
Economics, 112, p827‐872
De, P. (2010) Governance, Institutions, and Regional Infrastructure in Asia. ADBI
Heller, P. (2010) People and Places: Can they align to Bring Growth to Africa, Center for
Global Development Essay, September 2010, P1‐28
Kasfir, N. (1976) The Shrinking Political Arena: Participation and Ethnicity in African Politics . Berkeley and Los Angeles: University of California Press.
Keller, G. (2008) Managerial Statistics, Toronto, Canada:Nelson education, Ltd.
Kelly, W.A. (1981) A generalized interpretation of the Herfindahl index, Southern
Economic Association, p50‐57
Kenny, C. (2006) Measuring and Reducing the Impact of Corruption in Infrastructure,
World Bank Policy Research Working paper, 4099.
Khan, M. (2008) Governance and development: The prospect of growth‐enhancing
governance, Diversity and Complementarity in Development Aid: East Asian Lessons for
African Growth. Tokyo: GRIPS Development Forum/National Graduate Institute for
Policy Studies, pp. 107‐152.
Knack, S. Keefer, P. (1995) Institutions and Economic Performance: Cross‐Country Tests Using Alternative Institutional Measures. Economics and Politics 7(3) p207‐227
La Porta, R., Lopez‐de‐Silanes, F., Shleifer, A., Vishny, R. (1997) Trust in Large Organizations, American Economic Review Papers and Proceedings, 87 (2) 333‐338 Lewis, P. M. (1996) Economic Reform and Political Transition in Africa: The Quest for a Politics of Development. World Politics 49 (1) p92:129 Lutmer, E.F.P. (2001) Group Loyalty and the Taste for Redistribution, Journal of Political Economy, 109 (3), p500‐528 Mauro, P. (1995) Corruption and Growth, Quarterly Journal of Economics, 110(3), p 681 ‐ 712 McDaniels, T., Chang, S., Cole, D., Mikawoz, J. and Longstaff, H. (2008) Fostering resilence to extreme events within infrastructure systems: Characterizing decision contexts for mitigation and adaption, Global Environmental Change 18 (2), p310‐318.
Montalvo, J.G., Reynal‐Querol, M. (2005) Ethnic Diversity and economic development,
Palland, J. (2007) SPSS survival manual, McGrawHill,
Pessoa, A. (2008) Public‐private partnerships in developing countries; are infrastructures responding to the new ODA strategy, Journal of International
Development, 20 (3), p311‐325
Posner, D.N. (2004) Measuring Ethnic Fractionalization in Africa, American Journal of
Political Science, 48 (4), p849‐863
Putnam, R.D. (2007) E pluribus unum: Diversity and community in the twenty‐first century, Scandinavian Political Studies 30 (2), p137‐174
Rodrik, D. (1998) Trade policy and Economic performance in Sub‐Saharan Africa, Expert
group on Development issues,
Illorah, R. (2009) Ethnic bias, favoritism and development in Africa, Development
Southern Africa, 26 (5), p695‐707 United Nations Economic Commision for Africa (2009), Exploiting Natural Resources for Financing Infrastructure Development. AU Conference, p7‐61 Seragaldin, I., Taboroff, J. (1994) Culture and Development in Africa: Proceedings of an International Conference held at the World Bank. Washtington DC.
Vigdor, J.L. (2001) Community Heterogeneity and Collective Action: Analyzing Initial Mail Response to the 2000 Census, Duke University mimeo.
Appendix II: Definitions variables
Variables Definition Data Source
PREGscore (low,medium, high) ‐ Political Relevant Ethnic Groups Posner’s website Communication ‐ Internetusers per 100 inhabitants, 10 year average, 2002‐2011 ‐ Mobile cellular subscriptions per 100 inhabitants, 10 year average, 2002‐2011 Worldbank Electricity ‐ Access to electricity (% of population), 2009 Worldbank Medical services ‐ Health expenditure as % of GDP, 10 year average 2002‐2011 Worldbank Transport ‐ Road network density per population 2008 ‐ Paved and unpaved road network in good condition (% of total network) 2008 Worldbank Control variable: ‐ Extreme events ‐ GDP per capita
