Determinants of long run economic growth : a story of African lions

Determinants of Long Run Economic Growth:

A Story of African Lions

Abstract:

This paper examines the most important drivers of long run economic growth for a sample of 50 African countries, covering a time period from 1990 to 2013. Using an OLS regression, variables are selected which could have a significant impact on growth. These variables are

added to an augmented Solow model incorporating a human capital factor, and estimated using a fixed effects regression. The findings indicate that the most important drivers for economic growth in Africa over the researched period, apart from the conventional drivers

of the Solow model, were: political freedom and FDI. These drivers were found to have a positive effect. Additionally, inflation, being a former French colony and the debt to GDP

ratio had a negative effect.

Ruben Stukart June 2015

10445943

Statement of Originality

This document is written by Ruben Stukart who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document

is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for

the supervision of completion of the work, not for the contents.

Index

3. Data and Methodology 10

3.1 Dependent Variable 10

3.2 Independent Variables 10

3.3 OLS Cross-Country Model 12

3.4 Panel Data Fixed Effects Model 13

3.5 Assumptions, Hypotheses and Tests 14

4. Empirical Results 15

4.1 OLS Results 15

4.2 Panel Data Results 18

4.3 Discussion 21

5. Conclusion 21

References 23

Appendix 27

Appendix A: List of Countries 27

Appendix B: Descriptive Statistics 28

Appendix C: Data Sources 30

Appendix D: Graphs 31

1. Introduction

Countless research papers have been written about the rise of Asia and the Asian Tigers. Yet these Asian Tigers face a new formidable adversary: the African Lions. According to The Economist, the first decade of this century has seen a remarkable pattern. Although much of the attention has been given to the BRIC countries and the Asian tigers, a staggering 6 out of 10 countries with the highest economic growth were African countries (The Economist, 2011). Only one out of ten, China, was an Asian country. The forecasts are also in favour of the African Lions, beating the Asian Tigers in the expected top ten with 7 out of 10 places. These forecasts could not be more different from the 1990s, when Africa was just recovering from a bust period in the 1980s. Also the perception in the media changed. At the beginning of this century Africa was labelled: ‘’The Hopeless Continent’’ by The Economist while its issue of March 2013 was entitled: ‘’Africa: A Hopeful Continent’’ (Busse et al., 2015, p.3). In twenty years’ time the situation turned completely.

This changing pattern can also be seen in Figure 1. Average GDP per Capita growth in Africa has slumped in the eighties and nineties, but has recovered miraculously in the first decade of the 21st century. When Africa was performing poorly, the region attracted the interest from developmental economists, and now with the recent growth miracles Africa is becoming interesting once again.

Figure 1: GDP per Capita Growth rates

Source: World Bank 2015 (WDI) -8 -6 -4 -2 0 2 4 6 8 10 12 19 80 19 81 19 82 19 83 19 84 19 85 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13

Although Africa is certainly receiving attention from scholars studying single effects, there are not that many studies about the drivers of long run economic growth in Africa compared to the libraries written about the growth in Asia. Almost 20 years ago, in 1995 Savvides analysed the drivers of long run economic growth of a panel of 28 African countries between 1960 and 1987 (Savvides, 1995). More recently, Hoeffler investigated the need for a specific growth model for Africa because of the negative and significant dummies for Africa in other studies. He found that normal growth models could be applied to Africa, as Savvides did, provided that country specific effects and the endogeneity of investments are accounted for (Hoeffler, 2002). Building on these results, a very recent, yet still unpublished article, constructed an augmented Solow model for Africa (Busse et al., 2015).

While most studies use data from the bust period in Africa, it would be interesting to see whether these results still hold when both the data for the bust, and the boom period after the mid-90s are incorporated. Furthermore, due to data limitations most studies have used only a limited number of countries. After 1990, an increasing number of datasets has become available at the World Bank and the IMF. Now it is possible to get data for 50 out of 54 African countries for the period after 1990.

Applying part the methodology of Savvides, the research in this paper will investigate whether his results still hold twenty years after publication, using a more recent time period and incorporating more countries. This leads to the following research question:

‘’What are the most important determinants of long run economic growth in Africa?’’

To answer this question, an extensive review of the existing literature will be given in section 2 to determine which variables could have an effect on economic growth in Africa. In section 3 the methodology used to asses these variables and the datasets will be described, whereas in section 4 the results will be presented. Finally, the last section will contain a conclusion and a discussion of the possible improvements that could be made for future research.

2. Literature Review

In this section a thorough review of the existing literature about economic growth in Africa will be presented. From previous research, variables of interest will be selected, along with other variables which, according to general economic theory could have a significant influence on economic growth in Africa.

2.1 Definition of growth

There are many different definitions of growth but most researchers and empirical models of economic growth use the gross domestic product (GDP) per capita as the definition of economic growth. Both the original Solow model (Romer, 2012, p10) and the augmented Solow model (Mankiw, Romer and Weil, 1992) use the output per worker as definition of economic growth. Although this definition could be different from GDP per capita growth, Hoeffler showed that regressions models for growth in Africa were not sensitive to the use of per worker or per capita data (2002, p143.), so using per capita data even though the model is meant for per worker data, is not a problem. He explained that these results might be the result of data quality. Population data are gathered by means of a census which is held, on average, every ten years. Population data for the years in between censuses are obtained by interpolation from these data points. On the other hand, data about the workforce are calculated from the population data, so there are two distinct sets of calculations needed instead of one. Due to the measurement error that might be more distinct for workforce data and in order to compare his results to other studies, which used mainly per capita data, Hoeffler used per capita instead of per worker data for the estimation of a Solow model for Africa. Due to the availability and the reliability of the used data, the GDP per capita growth rates will be used as a proxy for economic growth and as the main research object.

2.2 Variables of interest

A lot of research has been conducted to study the effect of different factors on the economic growth in Africa. From economic factors like the debt ratio to non-economic factors like AIDS, all have been included before. While there have been case studies, studying individual African nations, most research has focussed on the Sub Saharan region of Africa or the continent as a whole. Descriptive statistics for the different factors in Africa can be found in Appendix B.

Government Debt

The average debt ratio in Africa has been substantial since the bust period in the 1980s, reaching a peak of 120 per cent in 1995 (IMF, 2012). Since the beginning of the 21st century, with increasing growth rates, the average debt ratio went down steadily from 110 per cent in 2000 to only 43 per cent in 2002. At the same time the debt ratios rose in Europe and the

rest of the developed world. With the recent financial crisis, debt has become a constant topic of concern in the developed world. Ludvigson (1996) found early evidence that debt financing mattered for GDP growth. A later study by Checherita-Westphal (2012) showed that in the Euro Area a negative effect of government debt on GDP per capita growth, starting at a 90 per cent debt ratio, was found. In a follow up study a positive effect was found until a tipping point of 67 per cent was reached (Baum, Checherita-Westphal & Rother, 2013). In another worldwide study a similar trend was found of positive effects for low debt ratios (below 30 per cent) and a negative effect above 90 per cent (Afonso & Jalles, 2013). On the other hand, when controlling for endogeneity, Panizza and Presbitero (2014) found no significant effect of government debt on GDP per capita. These results could be interesting for the African case. Although the current average debt to GDP ratio is low, there are huge differences between the countries, with current debt ratios ranging from as low as 10 per cent to more than 100 per cent.

Inflation

Besides the debt ratio, another factor that could have an effect on economic growth is inflation. In 1973, Dornbush & Frenkel developed new ways to look at the effects of inflation on growth. They found ambiguous effects of inflation on real money balances, consumption, and the capital labour ratio if the yield was a function of real money balances and if consumption was an increasing function of the rate of inflation (Dornbush & Frenkel, 1973). In a later study focussing on developing countries, De Gregorio (1992) found a negative relationship between high inflation rates and GDP per capita growth. A follow up study that focussed on the (long run) transmission channels of inflation found that in developing countries in South America the main channel was through the reduction of the productivity of capital (De Gregorio, 1993). Another study also found that inflation had a negative effect above a certain threshold (8 per cent) (Sarel, 1995). In a more recent study Gillman & Kejak (2005) examined the effects of a selection of different models used in empirical literature and found that all of them could generate significant negative effects of high inflation rates on GDP growth rates per capita. While the average inflation rate in Africa is currently down to about 5 per cent, it has been above the 8 per cent threshold found by Sarel for the largest part of the last 20 years (World Bank WDI, 2015).

Openness to trade

In a study conducted by Yanikkaya (2003) a positive and significant relationship was found between trade barriers and GDP growth per capita for some developing countries. This observation contrasts with the main view on the importance of exports for promoting economic growth. However, these significant results were highly dependent on the characteristics of the developing country in question. Evidence on the traditional importance of exports on growth was found by Feder (1982). He found that economic growth could be promoted by allocating more resources from the less efficient non-export sectors to the more productive export sectors. Furthermore in a study conducted by Savvides (1995) a

positive relationship was found between the growth of openness defined as imports and exports as a percentage of GDP and the growth of per capita GDP in Africa. The growth of openness would represent a more outward looking policy that could promote growth in African countries.

Investments

In his 1995 study, Savvides also studied other drivers of economic growth in Africa. Along with openness, investments, in particular FDI (full definition on p.11) had a positive and significant effect on GDP per capita growth rates in Africa. In another paper positive effects for FDI were also found, but only when the countries in question had enough human capital to do something with the FDI (Borensztein, De Gregorio & Lee, 1998). Markusen and Venables (1999) theorized that FDI, although it could crowd out local industries, could also benefit local supplier industries even up to a point where local industries get strong enough to crowd out the original foreign investors. More recent studies also found positive effects for FDI (Li & Liu, 2004). On top of that, there is also evidence that countries with a more developed financial sector benefit more from FDI. This insight could be interesting for the African case because there are big differences among African countries regarding financial development (Alfaro et al., 2004).

Financial Aid

According to research conducted almost 30 years ago, financial aid has a positive impact on GDP per capita growth rates in Africa (Levy, 1988). Other research also found a positive effect, but it was less for low income countries or countries with a small amount of aid (less than 13 per cent of GDP) (Durbarry et al., 1998). Hansen & Tarp found that aid had decreasing returns and that the effectiveness was dependent on the used empirical model (2001). But they confirmed previous results that aid affects growth through the investments channel. Finally, when financial aid was split up into developmental and non-developmental aid, significant results were found for a positive effect of developmental aid (Minoiu & Reddy, 2010). Another interesting result was that they found that aid from different donors had different effects, especially aid from Scandinavian countries promoted GDP growth. These differences could be caused by differences in efficiency and strategic interests of the donor countries. For instance, when countries give financial aid not out of strategic interests but out of humanitarian or other interests they are more likely to encourage economic reforms that are not that popular. They could stop the aid if the reforms or conditions are not met, because they have no strategic interests (see also: Bearce & Tirone, 2008). Minoiu and Reddy concluded that more research is needed, because there is still too little known about why some kinds of aid promote growth while others are less effective.

Stability

One factor that could ensure the effectiveness of aid could be stability. In situations of war or civil unrest it might be difficult to get the aid where it is needed, instead of ending up in

the pockets of local warlords. The relationship between stability and economic growth is a topic that is studied quite a lot in the 90s. A strong negative relationship was found between stability measured as political stability, the propensity of a government collapse, and GDP growth (Alesina et al., 1996). Even when they account for the possible reverse relationship of economic growth on the possibility of a governmental collapse, they still found significant effects. Countries with a governments that are less likely to be overthrown, grow faster. Further evidence of a negative relationship between political instability and GDP growth in Sub Saharan Africa (SSA) was found by Kwasi Fosu (1992). He classified countries from stable to high instability based on a political instability (PI) measure, which measured the probability that the government was overthrown. The estimated effect of being a high instability country in the SSA region was an approximate 1.1 percentage point reduction of the annual GDP growth over the period.

Regime Type

Regime type and stability share some attributes but do not necessarily move together. In the same study as quoted before, Alesina et al. (1996) found that adding a ‘’democracy’’ variable did not change the results, it was not a significant driver of economic growth. Another study, conducted by Tiruneh (2006) found that regime type has some, but not a strong influence in Africa. There is stronger evidence that regime type contributes to a bigger variation in growth rates (Weede, 1996). In his 1995 paper, Savvides found a significant effect of political freedom on GDP per capita growth rates in Africa. The weak regime type outcome, in contrast to the strong results for political freedom as a proxy for regime type, could be explained by the arbitrary division into democracy or another regime type. Considering that there are also variations among democracies with respect to the actual political freedom experienced by the citizens, a measure of political freedom could be a good proxy for regime type, for instance the freedom house score of political freedom.

Colonisation

The African continent has an extensive history of colonisation, almost every country has at some point in history been colonised by a foreign power. The only country which has been relatively untouched was the former kingdom of Ethiopia, although there were Italian influences when the Italians occupied the region in the period before and during the Second World War. Acemoglu, Johnson and Robinson (2001) used a very unusual and interesting instrument to estimate the effect of institutions on economic performance. They used the European mortality rates in the colonies, because in the countries where the Europeans faced high mortality rates, they founded other institutions. Interestingly these institutions are still present today and influence the growth of African countries. Nunn (2005) provided an example of why historic influences still matter today in Africa. When colonisation extracts enough resources, a society could be driven to a low production equilibrium. These unfavourable equilibria are stable and therefore hard to change, consequently the society will be trapped in the low production equilibrium. In other research the effect of

colonisation was also determined to be significant in Africa. For a selection of African countries, post colonisation average growth was larger than during colonisation, providing evidence for a significant structural break in the growth pattern after independence. Another outcome was that the identity of the colonizer also mattered for growth (Bertocchi & Canova, 2002).

Oil

Stijns (2005) found that abundance of natural resources does not necessarily have to mean high growth rates. It depends critically on the country and the nature of the learning process involved. Other studies found that energy consumption for oil-exporting countries, where the prices of energy are often kept below market rates, does have a relation with the growth rates. However, the causality runs backwards, from higher growth to higher consumption (Mehrara, 2007). Gylfason (2001) found that natural wealth has a negative impact on human capital, for instance through schooling. These negative effects of natural resources are found with the ‘’Dutch Disease’’. This ‘’curse of natural resources’’ causes some countries with natural resources to grow slower than countries without. It could be the case that these resource abundant economies are high price economies and that they therefore export less other products because of the high prices caused by the export of natural resources (Sachs & Warner, 2001). A study conducted in Africa stated that the export of natural resources in Africa has a positive effect on GDP, but these figures do not take the natural resource depletion into account, which could alter long term results (Winter-Nelson, 1995).

AIDS

A study conducted in Mozambique found that AIDS, and particularly AIDS mortality rates affected economic growth through a deterioration of human capital accumulation (Arndt, 2006). In a global study, an insignificant effect of AIDS on economic growth was found (Bloom & Mahal, 1996). In a more recent study they found a large effect of an AIDS epidemic on economic growth, and some small positive effects of anti-AIDS policies (Corrigan, Glomm & Mendez, 2005). McDonald & Roberts (2006) also found significant effects of AIDS on per capita growth in Africa. A 1 per cent increase in the HIV prevalence rate decreased the GDP per capita with 0.59 per cent. Dixon, McDonald & Roberts (2001) found that for countries in Africa with low AIDS prevalence rates it was fairly easy to continue with ‘’normal’’ economic expectations. But for some east African countries with ‘’catastrophic’’ AIDS prevalence rates of up to 20 per cent, the macroeconomic consequences were severe. Summarising the existing literature, it can be said that AIDS has an ambiguous effect. If the rates are low there is no real impact on economic growth, but if the rates are high there can be a significant negative impact.

2.3 Control Variables

Besides the above mentioned focus variables, most researchers also use variables to control for effects caused by factors other than the focus variables. Factors that are prescribed by

general economic theory include: consumption, investments and government expenditure. Savvides used control variables for investments and government expenditure (1995). He used investments to GDP and the growth of government expenditures to GDP as control variables in his regressions. The augmented Solow model he applied also used the logarithm of initial GDP per capita to control for unconditional convergence. Furthermore, the same investment and government expenditure ratios were used in research about the effects of colonisation on growth by Bertocchi and Canova (2002). Depending on the research method used, some dummy variables capturing time or country specific effects have to be applied.

3. Data and Methodology

In this section the datasets used and data sources will be explained. Furthermore, a thorough description of the empirical models and hypotheses will be given. Finally a subsection will be devoted to the definition of the tests used to assess the validity and significance of the models used.

3.1 Dependent Variable

As stated before, the dependent variable in the models will be the annual growth rate of the gross domestic product (GDP) of 50 African countries over the time period 1990-2013 (a full list of the countries can be found in Appendix A). In the analysis GDP growth and growth are used interchangeable for GDP per capita growth. Data can be found at the World Bank World Development Indicators database, the aggregates are based on constant 2005 Dollars to make comparison and averaging between the countries possible (World Bank WDI, 2015).

3.2 Independent Variables

The independent variables will be the focus variables discussed in the previous section. Apart from these focus variables, some control variables needed for the validity of the model in question are added. These variables will be explained in the model specification section and sources can be found in Appendix C.

For the debt ratio, data from the IMF is used. Abbas et al. have compiled an extensive historical public debt database with data for the debt to GDP ratio of 46 African countries for the period 1990-2012 (see also: Abbas et al., 2010). The target variable is gross central government debt which was then scaled to nominal GDP.

Inflation data is obtained from the World Bank (World Bank WDI, 2015). Inflation is defined as the GDP deflator, computed as the ratio of GDP in current local currency to the GDP in constant local currency, These data are available for all 50 countries. A more appropriate measure could be a consumer price index (CPI). However these are not available for many African countries. Although 45 countries have data about the CPI many gaps appear in the dataset. Nevertheless it will be checked if using another inflation measure will

alter the results. This is done because Sarel stated that an implicit GDP deflator could be negatively correlated with growth by design (Sarel, 1995 p2.).

Data for the trade (commercial) openness is not readily available for Africa, but data about imports and exports are available. The trade openness measure is constructed with data from the World Bank and is defined as imports plus exports as a percentage of GDP (World Bank WDI, 2015). For 48 African countries data could be constructed in this way.

FDI is defined as Foreign Direct Investment, net inflows of investment aimed to acquire a lasting interest (over 10 per cent in equity) in an enterprise operating in a foreign economy. It is the sum of equity capital, reinvested earnings, other long-term capital, and short-term capital. It is regarded as a percentage of GDP and is available for 49 countries in Africa (World Bank WDI, 2015).

Official Developmental Aid (ODA) as a percentage of GNI, is defined as a grant or a loan with at least a 25 per cent grant element aimed at promoting economic development and welfare in the receiving countries. This variable is available for 48 receiving countries at the World Bank (WDI, 2015). This definition of ODA is taken because it is a form of ‘’developmental aid’’, and research has shown that this form could have an effect on growth in contrast to ‘’non-developmental aid’’ (Minoiu & Reddy, 2010).

Stability is measured as a political stability measure similar to the one used by Kwasi Fosu (1992). It is available at the World Governance Indicators database of the World Bank (World Bank WGI, 2015). Political stability is defined as Political Stability and Absence of Violence/Terrorism and captures the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means. The estimate produces a normal distribution score that captures the aforementioned probability and has a continuous range between -2.5 and 2.5 (see also: Kaufman et al., 2010). The measure is available for all 50 countries.

A variable for regime type could be constructed in several ways. One of them would be to use a dummy variable, for instance, for democracy or dictatorship. However, this would ask for a lot of arbitrary decisions for ‘’border cases’’ and it might not capture all the effects of a regime type on economic growth. A better way would be to use a score, like the Freedom House score of political freedom to classify the regime type and freedom of a country. This is in line with the method that Savvides used to measure the effect of freedom (1995). The Freedom House score of political freedom ranges from 1 to 7 (7 categories) and classifies the political system in a country from Free, Partly Free to Not Free at all (Freedom House, 2015). The Freedom House score is available for all 50 countries.

Dummy variables will be constructed capturing the effects of colonisation in Africa, which will be one if the country was colonised and zero otherwise. There were two main colonial powers in Africa: the French and the British. Apart from these powers there were other countries with colonies in Africa that will be captured in the category entitled ‘’other’’. Finally there were two countries that stayed relatively untouched, Ethiopia and Liberia, which was founded as a colony of former slaves by the United States but gained its independence in the 19th century.

To capture the importance of natural resources, in particular oil, on a local economy the term ‘’Oil Rents’’ will be used. The variable is calculated as the profits above the extraction costs (so world prices minus the extraction costs) as a percentage of GDP. The data are available for all 50 countries at the World Bank, of which 21 are actually oil producing countries (World Bank WDI, 2015).

Data about the AIDS and HIV prevalence rates are available at the UNESCO Institute for Statistics and the World Bank. More data are available for HIV prevalence rates at the World Bank, so this measure is taken for 47 countries. It is defined as the percentage of the working population (15-49 in Africa) suffering from HIV (World Bank WDI, 2015).

3.3 OLS Cross-Country Model

To check which variables could have a significant influence on the GDP per capita growth rate, an Ordinary Least Squares (OLS) regression model will be estimated. The variables that turn out to have an influence will later be tested in further detail, using part of the methodology of Savvides (1995).

Gaps in the dataset are not uncommon when working with African countries. Some countries have experienced multiple revolutions or have been in a period of civil unrest, which made data collection impossible. When the gaps in the data are not too large (most gaps are one or two years wide) averaging the data can solve this problem. A cross-country setting is used to make advantage of averaging and is estimated with an OLS regression model. In total four different models will be estimated, three of them to see which variables should be analysed further as well as to account for the correlations between FDI, Investments and Openness. An F-Value will be presented for each model to test whether using the CPI for inflation instead of the GDP deflator could alter the results, like Sarel proposed (1995, p.2).

The models will be based on the Solow model, augmented by Mankiw, Romer and Weil to consider human capital accumulation (1992). In the original model, the percentage of the working population that is still in secondary school was used as the human capital accumulation factor. However, this variable is not available for African countries. For that reason the percentage of the population that is still in secondary school is used instead. This is in line with the methodology of Savvides and most other researchers (Savvides, 1995, p.3). The model will be based on the regressions of Savvides and the specification is as follows:

where the GDP per capita growth rate ( ) is explained by the intercept, the natural logarithm of initial GDP per capita, the average investment to GDP ratio, the average population growth, the average secondary school enrolment and an error term. This model will then be expanded to incorporate the effects of the focus variables and estimated using OLS. The variables will be added to the model to form the following equation:

In this equation and form a vector of coefficients and variables containing the focus variables: the debt ratio (DEBT), the GDP deflator (INFLATION), Official Developmental Aid (ODA), Political Stability (STABILITY), Freedom (FREEDOM), a dummy variable indicating a former British colony (BC), a dummy variable indicating a former French colony (FC), oil rents (OIL) and the AIDS prevalence rate (AIDS), all variables are averages. For model 3 and 4 the investment ratio is swapped for, respectively FDI (FDI) and the level of openness (OPEN) to control for the correlation between those variables. Furthermore, the F-statistic of the model with CPI instead of the GDP deflator is included for each model.

3.4 Panel Data Fixed Effects Model

The variables that turn out to be significant in the OLS regression will be included in a panel data model (Savvides, 1995). The dataset is converted to panel data and divided into 4 six-year periods: 1990-1996, 1997-2002, 2003-2008 and 2009-2013. In the first model the unconditional convergence hypothesis is tested, in which poorer countries (lower initial GDP per capita) grow faster. The second model will test the conditional convergence hypothesis (see also: Solow, 1956 and Mankiw, Romer & Weil, 1992).

Both models will be estimated using a fixed effects regression model incorporating both country and time fixed effects. This regression model eliminates omitted variables bias that is caused by unobserved variables that are constant, both over time and over the countries (Stock & Watson, 2012, p.402). Time specific effects are incorporated by adding 3 binary variables indicating the time period (T96, T02, T08). Country specific effects are incorporated by adding country dummies (not shown in the equation due to space constraints). Model 1 will be specified as follows:

In this equation the GDP per capita growth rate ( ) is only explained by the natural logarithm of initial GDP per capita, 3 time dummy variables and an error term. The subscript ‘’i’’ is the index for the countries and ‘’t’’ is the time index. To test the conditional convergence theory, two variables will be added: the investment to GDP ratio and the population growth rate. With these variables the equation is transformed into model 2:

The final models will test the effect of the variables that turned out to be significant in the OLS models. Each variable is added to the augmented Solow model, incorporating some variables of new growth theories (Savvides, 1995). The models will be estimated using a panel data regression model incorporating country fixed effects to control for omitted variable bias. The model specification is as follows:

In this model the GDP per capita growth rate is explained by variables that are derived from the augmented Solow model incorporating human capital, the inflation rate, the growth of the openness ratio, and the growth rate of the government consumption. Added to the model will be, for each separate model, one of the significant variables of the OLS regression ( .

3.5 Assumptions, Hypotheses and Tests

For the regression techniques used several assumptions have to be made (Stock & Watson, 2012, p.405). The same assumptions will have to hold for both techniques. Applied to the panel data approach these assumptions are:

1. has conditional mean zero: ( |

2. All variables and error terms are independent identically distributed draws from their joint distribution (i.i.d.).

3. Large outliers are unlikely.

4. There is no perfect multicollinearity.

The first assumption is analysed by plotting the residuals, which shows no clear pattern and also comparing them to a normal distribution, which indicated that the errors were normally distributed (graphs can be found in Appendix D). Furthermore the regression data showed that the fixed effects were strongly correlated with the explanatory variables, so it is useful to control for them with a fixed effects regression model. The same variables are drawn for each separate country, and they are not perfectly correlated with each other so assumption 2 and 4 hold. By using robust standard errors, the effect of large outliers is reduced and assumption 3 is met. Using robust standard errors also mitigates the effects when the variables and errors are not i.i.d. draws, so this eliminates any issues with assumption 2 as well. Finally a test-statistic will be provided in Appendix D to test whether using time specific effects is needed, and is justified in the first two equations. This is an F-test to test the hypothesis that all the time dummies are 0 (for all tests, see: Torres-Reyna, 2007). They indicated that the dummies were not zero, so in the first two panel data models (convergence models) time specific effects need to be controlled for. For the last models they were inconclusive, so those models only control for country specific effects.

Many researchers only include the Sub Saharan part of Africa (SSA). In the sample used in this paper all African countries are considered, including the countries ‘’above’’ the Sahara desert. To test whether this alters the results a Chow-Test of structural change is constructed (all the calculations can be found in Appendix E):

( (

Where and are, respectively the Sum of Squares Error of the combined model and only the SSA sample, q is the number of restrictions and k is the number of variables. The test will produce an F value to see if the hypothesis that the coefficients of the pooled and restricted model are the same can be rejected. If this is true, restricting the model to only SSA countries would provide a better model. Taking into account the results of previous research, the following hypotheses are constructed for the focus variables:

4. Empirical Results

In this section the results of the analysis will be summarised into two tables. The first table summarises the results for the OLS models and the second table for the Fixed Effects models.

4.1 OLS Results

The results obtained from the aforementioned regressions are summarised in the table on the next page. The t-values are given in parenthesis below the estimated coefficients. Also two F-values are included: the first tests the significance of the whole model and the second of an alternative model where the CPI measure is taken for inflation (degrees of freedom are in parenthesis). Furthermore, a coefficient of determination is included (R2, an indicator of the percentage of the total variance explained by the model).

The first model estimates the augmented Solow model incorporating a human capital factor. The results confirm the conditional convergence hypothesis in which poor countries grow faster. The coefficient is negative and significant (although only at the 0.10 level), a higher initial GDP per capita has a negative effect on the growth rate of GDP per capita. Furthermore, in line with the original theory, the investment to GDP ratio has a highly significant positive impact on growth. Although the model as a whole is significant (F-value of 6.74), the lack of significance of the other explanatory variables could indicate one of the shortcomings of the OLS analysis. That is, the lack of significance could be caused by the low number of observations, because due to the cross-country design the number of observations is limited to the number of countries.

For the second regression the first model is expanded to incorporate the effects of the focus variables. When these are added, the initial GDP per capita loses significance but the investment to GDP ratio, although it decreases in magnitude, stays highly significant. Furthermore, the population growth and schooling effects stay the same.

Explanatory Variables (1) (2) (3) (4) CONSTANT -0.847 2.496 8.399 12.34 (-0.29) (0.33) (1.68) (1.69) LNGDPIN -0.820* -0.794 -0.688 -1.873** (-1.96) (-0.87) (-1.04) (-2.24) INV 0.235*** 0.184*** (8.89) (3.72) POPGR 0.538 0.264 -0.487 0.133 (1.01) (0.40) (-0.85) (0.16) SCHOOL 0.037 0.022 0.013 0.039 (1.63) (0.90) (0.56) (1.14) DEBT -0.007 -0.018* -0.018 (-0.75) (-1.81) (-1.09) INFLATION -0.002*** -0.002*** -0.002* (-2.94) (-3.13) (-1.86) ODA -0.030 -0.039 -0.045 (-0.37) (-0.57) (-0.52) STABILITY 1.124 0.764 1.035 (1.33) (1.12) (1.09) FREEDOM 0.494* 0.286 0.344 (1.96) (1.12) (1.12) BC -0.583 -0.834 -1.079 (-1.03) (-1.00) (-1.34) FC -2.529*** -2.539*** -2.759*** (-4.37) (-3.10) (-3.14) OIL 0.021 0.004 0.045 (0.79) (0.12) (1.25) AIDS -0.112* -0.137* -0.128 (-1.72) (-1.98) (-1.68) FDI 0.464*** (4.76) OPEN 0.037* (2.05) F-ratio 6.74*** 29.87*** 21.27*** 6.47*** df (4, 42) (13, 27) (13, 27) (13, 26) R2 0.677 0.834 0.846 0.759 F-ratio CPI 4.55*** 8.67*** 3.20*** df (13, 24) (13, 24) (13, 23) N 47 41 41 40

t-values are in parentheses below the estimated coefficients, the first F-ratio gives the significance of the estimated model and the second F-ratio gives the significance of the model with CPI as inflation measure. * = significant at 0.10 level, ** = significant at 0.05 level, *** = significant at 0.01 level.

The high coefficient of determination and F-value indicate a highly significant model. Of the aforementioned hypotheses, only the results for political freedom (FREEDOM) and aid (ODA) are not in line with previous research. Political freedom was thought to have a negative effect on growth, because the variable is measured as a freedom house score in which 1 indicates free and 7 indicates not free. The higher the score, the lower the growth, but the estimates seem to indicate a positive relationship, although barely significant at a 0.10 level. This counter intuitive result, that less freedom is good for economic growth, could be due to the limitations of the OLS analysis. To see whether this is the case, and to compare the results to the negative relationship that Savvides found, political freedom is also included in the panel data regression (Savvides, 1995). Official developmental aid has a small negative effect where a positive effect was expected. Although the same results were found in a recent study, this effect is not significantly different from zero (Busse et al., 2015). This might be explained by the high variability of aid in Africa, there are countries that receive almost no aid at all, and some countries of which 56 per cent of the GDP is dependent on developmental aid (Appendix B: Descriptive Statistics). The average ODA to GDP is also quite low, 10 per cent, combined with the fact that almost all countries in Africa are low-income countries this could explain the insignificant results. Previous research has shown that for low-income countries that receive less than 13 per cent of their GDP as aid there is less effect (Durbarry et al., 1998). Furthermore it can be assumed that ODA affects GDP only indirectly and after a longer period (Busse et al., 2015).

Moreover the debt ratio has a small but insignificant negative effect, contrary to the inflation rate that has a very small but significantly different from zero negative effect. More stability is also good for growth, although the results are insignificant, which could be caused by the low number of observations and the fact that almost the whole sample is ‘’unstable’’. The average stability is negative, and the most stable countries have a rating of only just above 1. Recalling that the scores were normal distribution scores, this still gives a probability of government collapse within a reasonable period of time of almost 15 per cent. The effect of stability on economic growth might be more significant when researched in a sample of countries with a wider range of stability scores. The effect of colonisation is large and negative, and the results indicate that former French colonies belong to the slowest growing countries. Finally, the effect of oil was small and positive, and AIDS had a significant negative effect. All of these results are in line with the aforementioned literature and stated hypotheses.

The F-ratio of the model in which the CPI was taken as inflation measure can be found in the table. In all of the cases this value was lower than that of the original model (as was the R2). This is due to the lower degrees of freedom caused by the lower number of observations available for the CPI. The correlation between the two measures was 0.99, because of this high correlation and the fact that there were more observations available for the GDP deflator, this measure was believed to be a better representative of inflation.

The last two regressions swapped the investment ratio for the openness measure and the FDI measure. Although trade openness was less correlated with investment (0.35),

than FDI (more than 0.55), both variables had a significant impact on growth, as could be expected from the existing literature. The variables that were selected to be added to the panel data regression are: political freedom, the debt to GDP ratio, the AIDS prevalence rate and FDI. Inflation and the growth of the openness measure are part of the original model so they will be added as well.

4.2 Panel Data Results

As described in section 3, the 6 variables which were found to be significant drivers of growth in the OLS regression are translated into panel data and estimated with a Fixed Effects regression model, summarised on the next page. The first two models test the unconditional and conditional convergence hypotheses. The results of the first model indicate evidence for the unconditional convergence hypothesis, and the second model indicates also a weaker conditional convergence, although not significant. This significance problem is interesting and requires further analysis. In the final results, robust standard deviations are used, mitigating the effects of large outliers. When they are not used, the model changes significantly, all variables become highly significant (except population growth). These results indicate the importance of outliers for these models. Especially the first two models suffer from these outliers because they are highly dependent on just a few variables. For instance, the GDP growth suffers from large outliers, when taken as 6 year average values the maximum values were above 50 per cent while the minimum values were below -25 per cent. One country that stands out is Equatorial Guinea, in 1997 for instance, the economy grew with almost 200 per cent after the discovery of oil (World Bank WDI, 2015). Equatorial Guinea contributes significantly to the large outliers, when this country is omitted, the results for the first two models change significantly: evidence for both the conditional and unconditional convergence theory turn out to be highly significant. These results highlight the importance of outliers in these convergence models. In models three to six, leaving out these outliers did not change the results significantly, because growth is explained by more variables in those models.

The third model indicates a strong negative and highly significant effect of political freedom. This is directly opposed to the results of the OLS regression that indicated a positive relationship. Because a fixed effects panel data analysis better analyses the time dimension and controls for omitted fixed effects, it is well-suited to analyse growth. Combined with the larger number of observations (due to panel data versus OLS) this contributes to the larger significance. Comparing these results to the results of Savvides (1995), it can be said that some variables did not change much, considering the 20 year time gap between his research and this analysis. The most notable findings are for the growth of government consumption, where the coefficient is 0.041 and where Savvides found a coefficient of 0.044. Also for the political freedom measure, although the magnitude is different, the sign and significance of the variables are the same. The most notable difference is the population growth rate, which had almost no effect in model 3, whereas Savvides found a large negative effect.

Explanatory Variables (1) (2) (3) (4) (5) (6) CONSTANT 51.28* 23.78 24.39* 35.08** 21.88 28.01** (1.80) (0.98) (1.72) (2.17) (1.33) (2.07) LNGDPIN -8.039* -4.555 -3.249 -5.174** -3.64 -4.45** (-1.85) (-1.27) (-1.53) (-2.17) (-1.44) (-2.19) INV 0.259* 0.0722 0.018 0.036 (1.83) (1.31) (0.32) (0.62) POPGR 0.005 0.090 0.214 0.238 0.231 (0.03) (0.17) (0.40) (0.48) (0.53) INSCHOOL 0.049** 0.059** 0.048** 0.048** (2.15) (2.34) (2.08) (2.49) INFLATION -0.001** -0.001** -0.001* -0.001 (-2.45) (-2.49) (-1.94) (-1.60) OPENGR 0.034 0.021 0.027 0.014 (0.48) (0.28) (0.36) (0.21) GOVGR 0.041** 0.054*** 0.054*** 0.057*** (2.36) (2.67) (2.72) (2.75) FREEDOM -1.061*** (-2.74) DEBT -0.025** (-2.50) AIDS 0.188 (1.59) FDI 0.204*** (2.83) T96 3.697*** 2.799*** (3.14) (3.72) T02 4.532*** 3.679*** (5.12) (6.40) T08 5.103*** 2.740*** (4.40) (2.86) F-ratio 7.04*** 9.34*** 9.75*** 13.57*** 6.39*** 10.45*** df (4, 145) (6, 137) (8, 56) (8, 49) (8, 56) (7, 57) R2 0.509 0.667 0.763 0.788 0.740 0.797 F-ratio SSA 0.114 0.305 2.279* 1.264 1.717 1.689 df (5, 193) (5, 186) (5, 97) (5, 88) (5, 97) (5, 99) N 199 192 107 98 107 108 Countries 50 49 43 41 43 44

Table 2: Panel Data Fixed Effects Model of African GDP per capita growth

t-values are in parentheses below the estimated coefficients, the first F-ratio gives the significance of the estimated model and the second F-ratio gives the value of a Chow-test on structural change, analysing whether taking only the SSA region alters the results significantly. * = significant at 0.10 level, ** = significant at 0.05 level, *** = significant at 0.01 level.

However, in a more recent study for the period 1975 to 2010 population growth was found to have a positive effect on growth (Busse et al., 2015). This changing pattern could be related to the differences in population growth between the two samples, between 1960-1987 the average population growth was higher specification reasons the growth of openness was taken for and less volatile than between 1990-2013 (World Bank WDI, 2015). Additionally, the growth of openness decreased in significance (note: this is the growth and not the level of openness as studied before, which had a significant impact, but for comparison and model this model). This could be the case because some African economies might have reached a more ‘’mature’’ level of openness where they were not growing as fast as they could in the 60s and 70s. Finally, the level of inflation was found to have a small but significant negative effect on economic growth, which is in line with the results of Sarel, who found negative effects of high inflation (1995). While some countries in Africa have experienced steady and low inflation, there are some countries that suffered from very high inflation. This resulted in a high average inflation rate over the period of over 50 per cent. Furthermore, the effect is almost identical in magnitude to the recent study of Busse et al. (2015) they also found a very small but very significant negative effect.

The fourth model estimates the effect of the debt to GDP ratio on growth, which is found to have a significant negative effect on growth. The average debt to GDP ratio is 85 per cent. As mentioned before, some studies found positive effects until a tipping point of 67 per cent was reached (Checherita-Westphal, 2012), while others found negative effects above a 90 per cent ratio (Afonso & Jalles, 2013). Both ratios are close to the average debt ratios of most African countries. The remainder of the model does not change much, although the effect of convergence turns out to be significant again (this might be due to controlling for other factors in this regression).

The effect of AIDS turns out to be not significantly different from zero. The average AIDS prevalence rate in Africa is just 4.7 per cent. These low rates will not influence the economy, while very high rates could have an impact. There are just 5 countries in Africa with ‘’catastrophic’’ rates above 20 per cent as mentioned by Dixon, McDonald & Roberts (2001). These could suffer from negative effects, while the rest of the continent might be unaffected, which could cause the lack of evidence for an effect of AIDS. The effect of net FDI inflows as a percentage of GDP is highly significant and positive. This is in line with previous research that found that FDI had a positive influence on growth through the investment channel (see also: Li & Liu, 2004).

All models were highly significant, as indicted by the first F-value. Furthermore the high coefficient of determination indicated a good fitted model. The second F-statistic provided above, analysed whether excluding the North-African countries, like many researchers do, altered the model significantly. The Chow-test failed to reject the hypothesis of equality of coefficients for all models at the conventional significance level, indicating that the countries can be pooled together. Only for the model that considered political freedom, the test rejected the hypothesis at the 0.10 level. This could be due to the difference in

political freedom characteristics: all North-African countries, except Tunisia, are classified as not free, which could cause the structural difference.

Summarising, of the focus variables, inflation, (lack of) political freedom and the debt to GDP ratio were found to have a significant negative effect, whereas investments and FDI were found to have a positive effect. Lastly, the findings from the OLS model indicated that former French colonies belonged to the slowest growing African nations (the colony dummy was omitted from the panel data regression to prevent multicollinearity with the country dummies, because both of these dummies do not change over time).

4.3 Discussion

Although all of the presented results can be explained by existing literature and economic theory, the reader has to be cautious with interpreting these results. Unreliable and lacking data will always be a problem when investigating data produced in Africa. These problems are for the most part mitigated by the use of the fixed effects model and robust standard errors, which caused that the reader can at least infer the sign of the coefficient and the underlying effect with statistical certainty. A hopeful note for future research is that more and more reliable data are becoming available for African countries, which could increase the precision and significance of further research. Finally, a more advanced econometric method like the Generalised Method of Moments (GMM) estimator could improve the results. This was proved by Hoefler (2002) to be a good method for estimated growth, using the augmented Solow model in Africa. When Busse et al. (2015) also estimated a GMM model, besides their fixed effects model, they concluded that it increased significance, but that the main results actually did not change much. In conclusion, the results obtained in this analysis can still be interpreted as valid.

5. Conclusion

With the recent economic performance of Africa, the academic world has shown a revival of interest in drivers of economic growth in Africa. This paper examined the most important drivers of long run per capita GDP growth in Africa. This was done for a panel of 50 African countries covering a time period from 1990 until 2013. Improvements were made on existing research by expanding the number of countries and by including more variables in the used growth models to explain the economic growth in Africa in a more detailed way.

To answer the main research question, ‘’what are the most important determinants of long run economic growth in Africa?’’, first a thorough review of the existing literature was done. This produced 10 focus variables that could be important for economic growth. In addition to these 10 focus variables, there were also variables added that mattered for economic growth according to the augmented Solow model, which incorporated a human capital factor and some variables of the new growth theories. Two empirical approaches were constructed to investigate the effect of the focus variables on growth. In the first

approach, an OLS cross-country model was used to see which of these 10 focus variables could have a significant impact on growth. This analysis produced 7 variables that had a significant impact on growth: the debt to GDP ratio, the AIDS prevalence rate, political freedom, FDI, a dummy variable for a former French colony, the inflation rate, and the trade openness. Of these variables, the first 4 were added to a fixed effects regression model, whereas the inflation rate and the growth of openness were already part of the augmented Solow growth models used to estimate the effects on growth. The dummy for a former French colony was omitted from this model to prevent multicollinearity with the country dummies present in the fixed effects estimation technique. However, from the first OLS analysis and existing literature it could be concluded that former French colonies are among the slowest growing economies in Africa. Finally, swapping the debt deflator for the CPI as a measure for inflation did not improve the results.

For the second empirical method, the data were divided into 4 six-year sub periods for all of the 50 countries. Six different models were estimated to study the effects of the selected focus variables. The first two analysed the unconditional and conditional convergence hypotheses. The findings produce evidence for the unconditional convergence hypothesis. The conditional convergence hypothesis was insignificant, which was caused by the presence of large outliers in the GDP per capita sample. When these outliers were omitted there was strong evidence for both the conditional and unconditional convergence hypothesis. The third to sixth model estimated the effects of the 4 selected focus variables. Of these variables, political freedom had a strong positive effect: political freedom promoted economic growth in Africa during the sample period. The debt to GDP ratio had a significant negative effect, indicating that countries with a high debt to GDP ratio experienced lower growth rates. The AIDS prevalence rates were found to have no effect on growth at all. Existing literature indicated that the majority of countries in Africa with low rates would experience no effects, while some countries with high rates would experience a negative effect (Dixon, McDonald & Roberts, 2002). Additionally, the net inflows of FDI were found to have a highly significant effect on economic growth in Africa. Finally, the inflation rate was found to have a small negative but highly significant effect on growth, which was in line with existing recent research (Busse et al., 2015). Also, the growth of openness was found to have no significant impact on growth, opposed to research by Savvidess (1995), which could be an indicator that some African countries are on a more mature growth path for openness.

Furthermore, a Chow-test on structural change was constructed, which indicated that the Sub Saharan part of Africa (SSA) could be pooled with the North African countries. All of the findings are consistent with existing literature. Nevertheless, it would be interesting for future research to study the effects on growth with better data, and more advanced econometric methods like the GMM method.

In conclusion, the findings of this research indicate that the most important drivers for economic growth in Africa over the researched period, apart from the conventional drivers of the Solow model, were: political freedom and FDI, which were found to have a positive effect. Additionally, inflation, being a former French colony and the debt to GDP

ratio were found to have a negative effect. Finally, the findings pose an improvement over existing literature by expanding the research into the 21th century. Time will tell whether the contemporary theories will still be valid for the ever changing African landscape. It is to be seen whether the African Lions can defend their top spots in the top 10 list against the (new) Asian Tigers for the years to come.

References

Abbas, S. Ali, Nazim Belhocine, Asmaa El Ganainy, and Mark Horton. "A Historical Public Debt Database." IMF Working Papers 10, no. 245 (2010)

Acemoglu, Daron, Simon Johnson, and James Robinson. "The Colonial Origins of Comparative Development: An Empirical Investigation." The American Economic Review 91, no. 5 (2001): 1369-401.

Afonso, Antonio, and João Tovar Jalles. "Growth and Productivity: The Role of Government Debt." International Review of Economics and Finance 25 (2013): 384-407.

Alfaro, Laura, Areendam Chanda, Sebnem Kalemli-Ozcan, and Selin Sayek. "FDI and Economic Growth: The Role of Local Financial Markets." Journal of International Economics 64 (2004): 89-112.

Alesina, Alberto, Sule Ozler, Nouriel Roubini, and Phillip Swagel. "Political Instability and Economic Growth." Journal of Economic Growth 1 (1996): 189. doi:211.

Arndt, Channing. "HIV/AIDS, Human Capital, and Economic Growth Prospects for Mozambique." Journal of Policy Modeling 28 (2006): 477-89.

Baum, Anja, Cristina Checherita-Westphal, and Philipp Rother. "Debt and Growth: New Evidence for the Euro Area." Journal of International Money and Finance 32 (2013): 809-21.

Bertocchi, Graziella, and Fabio Canova. "Did Colonisation Matter for Growth?" European Economic Review 46 (2002): 1851-871.

Bloom, David, and Ajay Mahal. "Does the AIDS Epidemic Really Threaten Economic Growth?" Journal of Econometrics 77 (1997): 105-24.

Borensztein, Eduardo, Jose De Gregorio, and Jong-Wha Lee. "How Does Foreign Direct Investment Affect Economic Growth?" 45 (1998): 115-35.

Busse, Matthias, Erdogan, Ceren and Mühlen Henning. “Temporary Acceleration or Permanent Growth Take-Off? An Analysis of the Drivers of Economic Growth in Sub-Saharan Africa’’ . Working Paper (January 2015).

Checherita-Westphal, Cristina, and Philipp Rother. "The Impact of High Government Debt on Economic Growth and Its Channels: An Empirical Investigation for the Euro Area." European Economic Review 56 (2012): 1392-405.

Corrigan, Paul, Gerhard Glomm, and Fabio Mendez. "AIDS Crisis and Growth." Journal of Development Economics 77 (2005): 107-24.

Dixon, Simon, Scott Mcdonald, and Jennifer Roberts. "AIDS and Economic Growth in Africa: A Panel Data Analysis." Journal of International Development 13 (2001): 411-26.

Dornbusch, Rudiger, and Jacob A. Frenkel. "Inflation and Growth: Alternative Approaches." Journal of Money, Credit and Banking 5, no. 1 (1973): 141.

Durbarry, Ramesh, Gemmell, Norman and Greenaway David. “New Evidence on the Impact of Foreign Aid on Economic Growth.’’ CREDIT Research Paper 98, no. 8 (1998).

The Economist. “”The Lionkings?” The Economist Newspaper Limited: London, England (Jan 6, 2011). Feder, Gershon. "On Exports and Economic Growth." Journal of Development Economics 12 (1982):

59-73.

Fosu, Augustin Kwasi. "Political Instability and Economic Growth: Evidence from Sub-Saharan Africa." Economic Development and Cultural Change 40, no. 4 (1992): 829-41.

Freedom House. ‘’Freedom House Political Freedom Database’’. Freedom House Organisation (retrieved: May 2015).

Gillman, Max, and Michal Kejak. "Contrasting Models of the Effect of Inflation on Growth." Journal of Economic Surveys 19, no. 1 (2005): 113-36.

Gregorio, José De. "Inflation, Taxation, and Long-run Growth." Journal of Monetary Economics 31 (1993): 271-98.

Gregorio, José De. "The Effects of Inflation on Economic Growth." European Economic Review 36 (1992): 417-25.

Gylfason, Thorvaldur. "Natural Resources, Education, and Economic Development." European Economic Review 45 (2001): 847-59.

Hansen, Henrik, and Finn Tarp. "Aid and Growth Regressions." Journal of Development Economics 64 (2001): 547-70.

Hoeffler, Anke E. "The Augmented Solow Model and the African Growth Debate." Oxford Bulletin of Economics and Statistics 64, no. 2 (2002): 135-58.

IMF. “ World Economic Outlook Database’’. The International Monetary Fund (October 2013). Kaufmann, Daniel, Aart Kraay and Massimo Mastruzzi (2010). "The Worldwide Governance

Indicators: Methodology and Analytical Issues". World Bank Policy Research Working Paper No. 5430

Levy, Victor. "Aid and Growth in Sub-Saharan Africa: The Recent Experience." European Economic Review 32 (1988): 1777-795.

Li, Xiaoying, and Xiaming Liu. "Foreign Direct Investment and Economic Growth: An Increasingly Endogenous Relationship." World Development 33, no. 3 (2005): 393-407.

Ludvigson, Sydney. "The Macroeconomic Effects of Government Debt in a Stochastic Growth Model." Journal of Monetary Economics 38 (1996): 25-45.

Mankiw, N. Gregory, Romer, David and Weil, David N. ‘’A contribution to the empirics of economic growth “. The Quarterly Journal of Economic, Vol. 107, No. 2 (May 1992), pp. 407-437. Markusen, James, and Anthony Venables. "Foreign Direct Investment as a Catalyst for Industrial

Development." European Economic Review 43 (1999): 335-56.

Mcdonald, Scott, and Jennifer Roberts. "AIDS and Economic Growth: A Human Capital Approach." Journal of Development Economics 80 (2006): 228-50.

Mehrara, Mohsen. "Energy Consumption and Economic Growth: The Case of Oil Exporting Countries." Energy Policy 35 (2007): 2939-945.

Nunn, Nathan. "Historical Legacies: A Model Linking Africa's past to Its Current Underdevelopment." Journal of Development Economics 83 (2007): 157-75.

Panizza, Ugo, and Andrea F. Presbitero. "Public Debt and Economic Growth: Is There a Causal Effect?" Journal of Macroeconomics 41 (2014): 21-41.

Reddy, Sanjay, and Camelia Minoiu. "Development Aid and Economic Growth: A Positive Long-Run Relation." The Quarterly Review of Economics and Finance 50 (2010): 27-39.

Romer, David. ‘’Advanced Macroeconomics (4th edition)’’. McGraw-Hill: New York, USA (2012). Sachs, Jeffrey D., and Andrew M. Warner. "The Curse of Natural Resources." European Economic

Sarel, Michael. "Nonlinear Effects of Inflation on Economic Growth." Staff Papers – International Monetary Fund, 1995, 199.

Savvides, Andreas. "Economic Growth in Africa." World Development 23, no. 3 (1995): 449-58. Stijns, Jean-Philippe C. "Natural Resource Abundance and Economic Growth Revisited." Resources

Policy 30 (2005): 107-30.

Stock, J. H., & Watson , M. M..’’ Introduction to Econometrics’’. Pearson Education Limited:Harlow, England (2012).

Tiruneh, Gizachew. "Regime Type and Economic Growth in Africa: A Cross-national Analysis." The Social Science Journal 43 (2006): 3-18.

Torres-Reyna, Oscar. ‘’Panel Data Analysis Fixed and Random Effects using Stata’’. Data & Statistical Services, Princeton University (2007).

UNESCO. ‘’Education Indicators Database’’. UNESCO Institute for Statistics (retrieved: May 2015). Weede, Erich. "Political Regime Type and Variation in Economic Growth Rates." Constitutional

Political Economy 7 (1996): 167-76.

Winter-Nelson, Alex. "Natural Resources, National Income, and Economic Growth in Africa." World Development 23, no. 9 (1995): 1507-519.

Yanikkaya, Halit. "Trade Openness and Economic Growth: A Cross-country Empirical Investigation." Journal of Development Economics 72 (2003): 57-89.

The World Bank. ‘’World Development Indicators Database’’. The World Bank, Washington, DC, USA. (retrieved: May 2015).

The World Bank. ‘’World Governance Indicators Database’’. The World Bank, Washington, DC, USA. (retrieved: May 2015).

Appendix

Appendix A: List of Countries

North Africa

Algeria

Egypt, Arab Rep. Morocco

Sudan Tunisia

All Countries

Algeria Eritrea Namibia

Angola Ethiopia Niger

Benin Gabon Nigeria

Botswana Gambia, The Rwanda

Burkina Faso Ghana Senegal

Burundi Guinea Seychelles

Cabo Verde Guinea-Bissau Sierra Leone

Cameroon Kenya South Africa

Central African Republic Lesotho Sudan

Chad Liberia Swaziland

Comoros Madagascar Tanzania

Congo, Dem. Rep. Malawi Togo

Congo, Rep. Mali Tunisia

Cote d'Ivoire Mauritania Uganda

Djibouti Mauritius Zambia

Egypt, Arab Rep. Morocco Zimbabwe

Appendix B: Descriptive Statistics

Table: Descriptive Statistics

Variable N Mean Std. Dev Min Max

GDP 200 1.754 5.485 -26.396 52.985 LNGDPIN 199 6.583 1.089 3.972 9.622 POPGROW 200 2.398 1.057 -4.055 7.268 SCHOOL 174 37.410 23.503 5.331 120.661 DEBT 186 84.981 65.412 5.92 427.72 INFLATION 200 57.001 430.912 -4.413 5880.734 ODA 192 10.588 9.512 0.136 56.668 STABILITY 150 -0.5434 0.879 -2.805 1.0493 FREEDOM 200 4.662 1.684 1 7 BRITISH 200 0.54 0.499 0 1 FRENCH 200 0.40 0.491 0 1 OIL 198 6.141 14.964 0 78.195 AIDS 200 4.697 6.315 0.1 27.65 FDI 200 4.136 6.675 58.671 OPEN 195 75.792 44.676 14.446 422.328 INSCHOOL 127 36.085 24.618 5.199 121.612 CPI 183 51.992 414.069 -0.805 5444.475 GOVGR 165 7.585 19.411 -55.055 153.621 INV 193 22.178 11.486 5.098 109.046

Table: Correlations

Table: Correlations

GDP LNGDPIN POPGROW SCHOOL DEBT INFLATION ODA STABILITY

GDP 1 LNGDPIN 0.0975 1 POPGROW -0.0699 -0.7293 1 SCHOOL -0.0459 0.5395 -0.5535 1 DEBT -0.3930 -0.2176 0.0876 -0.0124 1 INFLATION -0.3734 -0.1365 0.0308 0.0221 0.4299 1 ODA -0.0329 -0.6908 0.5263 -0.4409 0.3175 -0.0811 1 STABILITY 0.1126 0.5678 -0.4811 0.2776 -0.2326 -0.3592 -0.1718 1 FREEDOM -0.1593 -0.3441 0.3217 -0.1449 0.3296 0.1831 0.0720 -0.5434

FREEDOM BRITISH FRENCH OIL AIDS FDI OPEN INSCHOOL CPI GGDP INV FREEDOM 1 BRITISH -0.1934 1 FRENCH 0.1934 -1 1 OIL 0.3095 0.0400 -0.0400 1 AIDS -0.0481 0.1513 -0.1513 -0.2064 1 FDI -0.1294 0.1160 -0.1160 -0.0422 -0.1101 1 OPEN -0.2193 0.1532 -0.1532 0.0686 0.1491 0.4692 1 INSCHOOL -0.2995 0.2284 -0.2284 -0.1722 0.2214 -0.0389 0.2753 1 CPI 0.1888 0.1791 -0.1791 0.0654 0.0026 -0.0799 -0.1054 0.0009 1 GGDP 0.0221 0.2007 -0.2007 0.3516 -0.1578 0.0084 -0.0137 -0.1483 -0.1814 1 INV -0.275 -0.088 0.088 -0.041 0.026 0.553 0.356 0.0253 -0.121 0.074 1