autónoma.
Entre 14 Entre 18 Entre 22 Entre 26 Total y 17 años y 21 años y 25 años y 30 años de
There have been various initiatives to help promote the development of the African continent for example the Priority Programme of Economic Redressing of Africa (PPREA), the UN programme for the economic redressing and development of Africa and the New Partnership for African Development (NEPAD) (de la Croix Nkurayija,
2011). Very few of these initiatives targeted investment in Africa explicitly, as increased investment was usually a by-product of development initiatives. If more African countries can attain formal sovereign credit ratings, increased investment in the continent will follow as by-product.
In 2002, the United Nations Development Plan (UNDP) launched an initiative to promote the attainment of sovereign credit ratings to sub-Saharan Africa and other developing countries (Standard and Poor’s, 2003). The aim of the project was to give support to countries in order to gather funds from private capital markets (Standard and Poor’s, 2003). The initiative did not only explain the potential benefits to sovereigns, but also provided technical and financial support to countries who requested ratings (African Development Bank, 2011a). Before this initiative, only six African countries had been rated by Standard and Poor’s, namely, Botswana, South Africa, Tunisia, Egypt, Morocco and Senegal (Panapress, 2004). Ghana was the first country to benefit from this initiative when it was assigned a rating of ‘B+’ in September 2003 (Standard and Poor’s, 2003). Thereafter, Cameroon and Benin were rated in November and December 2003 (Panapress, 2004). In 2004, Standard and Poor’s added Burkina Faso, Kenya, Madagascar, Mali and Mozambique to their rated list (Panapress, 2004).
A similar initiative was also launched in 2002 by the US Department of State, Bureau of African Affairs (US Department of State, 2009). The initiative commenced with a conference to sub-Saharan countries in order to educate them about the benefits and processes involved in obtaining a credit rating (US Department of State, 2009). In addition to the conference, a project was announced to fund a number of sub-Saharan African countries in their initial attainment of sovereign credit ratings (US Department of State, 2009). Fitch was awarded the contract to conduct the ratings for 12 sub- Saharan countries over the period 2002 to 2006 (US Department of State, 2009). Before the project was launched, only four sub-Saharan African countries had a formal sovereign credit rating. At the end of 2006, there were 19 sub-Saharan African countries with a formal rating (US Department of State, 2009). According to the US Department of State (2009), a few countries decided to pursue formal ratings of their own.
These initiatives ended years ago and were never revived by another international body that was in the position to do so (African Development Bank, 2011a). By 2015, just over 43.0 percent of African countries had obtained at least one formal credit rating from one of the three major rating agencies. There is no doubt that African countries who are not rated formally yet can benefit from a similar initiative in the future. Sovereign credit ratings are an indication of the strengths and weakness of African countries. Although there are costs involved in obtaining credit ratings, the benefits of these ratings far outweigh the costs.
African countries were initially very sceptical about the benefits of investment (Dupasquier and Osakwe, 2005). However sub-Saharan Africa took the lead in embracing the opportunities that investment brings to the continent by working on more favourable environments for investment on the continent (Fischer et al. (1998). Subsequently, the flows of investment into the continent has been mixed in the past. Most recently, FDI flows decreased in North, West and Southern Africa, whereas there was an increase in East and Central Africa (UNCTAD, 2015a). There are endless opportunities in different regions of the continent, but the sub-Saharan region stands out. As a net exporter of primary commodities, the sub-Saharan region still managed to grow, even after the end of the commodities boom due to sound macroeconomic policies and strong sectoral reforms (World Bank, 2015a). Although the continent is plagued with various weaknesses, the weaknesses that create excessive uncertainty for investors are limited to only a few regions on the continent. For example, political instability activities like military interventions and ethnic and religious conflicts occur mainly in Central, West and East African regions (ACLED, 2015).
4.4 RATIONALE FOR THE REGIONAL AND INCOME MODELS
A comprehensive literature review on the determinants of sovereign credit ratings was presented in chapter 3 (see section 3.2). The mentioned literature is also relevant to the regional and income models that will be covered in this chapter. There are, however, two studies from this literature review that specifically serve as rationale for the regional and income models estimated in this study.
The study by Bissoondoyal-Bheenick (2005), which was one of the first researchers to make use of a panel ordered probit model, concluded that the importance of specific variables varies according to the development of a country. Bissoondoyal-Bheenick (2005) found that GNP per capita and inflation were the most significant determinants when modelling ratings and that the weighting of variables for high rated countries is different to those of low rated countries. Additional variables that were important in low rated countries were the current account balance and the level of foreign reserves (Bissoondoyal-Bheenick, 2005).
Another study that incorporated a similar ordered response model as the model Bissoondoyal-Bheenick (2005) was that of Teker et al. (2013). Teker et al. (2013) eliminated less important determinants by identifying homogenous factor groups as determinants of ratings by using factor analysis. The most important factor identified in their study was the ‘level of development and resources’ (Teker et al., 2013:5). This factor explained 59% of the total variance in sovereign credit ratings. The factor consists of GDP per capita, the corruption perception index, the economic freedom index and foreign reserves to GDP.
It is expected that the regional and income level models in this study will capture the different levels of development amongst African countries in the same way that Bissoondoyal-Bheenick (2005) and Teker et al. (2013) captured the effect for other developing countries.
4.5 DATA AND METHODS
Various important determinants that influence sovereign credit ratings of the African continent were identified in chapter 2. The same variables will be utilised once more in the regional setting of the African continent in order to see if different determinants are significant between the different regions. The African continent will also be divided into the different income levels, as classified by the World Bank, to shed further light on the dynamics of significant determinants of African sovereigns.
A panel of 27 African sovereigns will be used for the time period between 2007 and 2014 on an annual basis for the regional and income classification models. Only the ratings from NKC African Economics will be utilised since the entity rates the most African countries and it will be difficult to compare regions with each other if the ratings from the major rating agencies are used because they do not rate enough countries in each regional and income category. The countries that NKC rated from 2007 to 2014 were: Algeria, Angola, Benin, Botswana, Cameroon, DRC, Egypt, Ethiopia, Gabon, Kenya, Lesotho, Malawi, Mauritius, Morocco, Mozambique, Namibia, Nigeria, Rwanda, Senegal, South Africa, Swaziland, Tanzania, Tunisia, Uganda, Zambia and Zimbabwe. The countries were divided into their geographical regions for the first model as identified in section 4.2 and displayed in table 4.2.
Table 4.2: Regional Classification of Included African Countries in the NKC model
North Southern Central West East
Algeria Angola Cameroon Benin Ethiopia
Egypt Botswana DRC Ghana Kenya
Morocco Lesotho Gabon Nigeria Uganda
Tunisia Malawi Rwanda Senegal
Mauritius Mozambique Namibia South Africa Swaziland Tanzania Zambia Zimbabwe Source: Wikitravel (2015).
In addition, the African countries were also categorised into their level of income as classified according to the World Bank for the second model. The 27 countries included in this study fall into three categories, namely: low income (GNI per capita of $1 035 or less), lower middle income (GNI per capita between $1 036 and $4 085) and upper middle income (GNI per capita between $4 086 and $12 615) (World Bank, 2013c). The income classification is presented in table 4.3.
Table 4.3: Income Classification of Included African countries in the NKC model
Low Lower Middle Upper Middle
Benin Cameroon Algeria
DRC Egypt Angola
Ethiopia Ghana Botswana
Malawi Kenya Gabon
Mozambique Lesotho Mauritius
Rwanda Morocco Namibia
Tanzania Nigeria South Africa
Uganda Senegal Tunisia
Zimbabwe Swaziland
Zambia
Source: World Bank (2013c).
This geographical and income division is done in order to identify if there are any differences in the significant determinants of African countries in specific regions and income levels of countries. As stated previously, the different countries in the African continent are at different stages of development and this division could also aid in shedding some light on the importance of certain determinants that are linked to development stages.
The data for the NKC qualitative ratings and the determinants identified from literature were sourced from the NKC, World Bank and Bloomberg databases. A linear transformation was used to quantify the rating categories of NKC. Therefore, the rating category of D was assigned a 1, through to AAA, which was assigned a value of 26. The end-of-year ratings were used in the study. The explanatory variables in the two models will be the same determinants identified in chapter 3. The variables are summarised in table 4.4
Table 4.4: List of Determinants Included in the Regional and Income Models Variable Name Definition
Macroeconomic Indicators GDP growth Annual real growth, year-on-year. Investment FDI to GDP.
Inflation Annual consumer price inflation. Government Performance Fiscal balance Budget balance to GDP.
External debt External debt to GDP.
External Balance External balance Current account to GDP. Foreign reserves Foreign reserves to GDP.
Variable Name Definition
Developmental Explanatory Variables Per capita income GDP per capita.
Corruption Transparency international – corruption perceptions index. Regulatory quality Ability of government to form and implement sound policies
and regulations that promote private sector development. Internet users Individuals who have made use of the internet via computer,
mobile phone or any other electronic device in the past 12 months.
Regional Binary Variables
North D=1 if sovereign is located in North Africa, D=0 otherwise. Southern D=1 if sovereign is located in Southern Africa, D=0 otherwise. West D=1 if sovereign is located in West Africa, D=0 otherwise. East D=1 if sovereign is located in East Africa, D=0 otherwise. Central D=1 if sovereign is located in Central Africa, D=0 otherwise.
Income Binary Variables
Low D=1 if sovereign is classified as a low income country, D=0 otherwise.
Lower Middle D=1 if sovereign is classified as a lower middle income country, D=0 otherwise.
Upper Middle D=1 if sovereign is classified as a upper middle income country, D=0 otherwise.
Source: Author’s own representation.
4.5.2 Econometric Framework
The regional and income determinants models will tested by using an ordered probit panel data model. The panel data method is used because it increases the number of observations and is appropriate for the nature of the data, being a combination of cross sectional and time series data. Due to the ordinal nature of the sovereign credit ratings data, the ordered probit model is used since it is most suited to a dependent variable that is ordinal in nature. The ordered probit model is specified as (Teker et al., 2013):
𝑦𝑖𝑡∗ = 𝑥
𝑖𝑡𝛽 + 𝜖𝑖𝑡 (1)
where 𝑦𝑖𝑡∗ is an unobservable latent variable that represents the sovereign credit rating of country 𝑖 in period 𝑡; 𝑥𝑖𝑡 is a vector of time-varying explanatory variables; 𝛽 is a vector of unknown parameters; and 𝜖𝑖𝑡 is a random disturbance term.
According to Teker et al. (2013), if 𝜖𝑖𝑡 s normally distributed, equation (1) delivers an ordered probit model. It is assumed that 𝑦𝑖∗ is related to the observed variable, 𝑦𝑖, the sovereign credit rating, in the following way (Long and Freese, 2006):
𝑦𝑖 = { 1 𝑖𝑓 𝑦𝑖∗ < 𝜏 1 2 𝑖𝑓 𝜏1 ≤ 𝑦𝑖∗ < 𝜏2 3 𝑖𝑓 𝜏2 ≤ 𝑦𝑖∗< 𝜏3 4 𝑖𝑓 𝜏3 ≤ 𝑦𝑖∗< 𝜏 4 ⋮ 26 𝑖𝑓 𝑦𝑖∗ > 𝜏 26 (2)
where 𝜏𝑚 is known as cutpoints or threshold parameters and are estimated through maximum likelihood estimation (MLE).
The regional model is specified as follows:
𝑦𝑖𝑡∗ = 𝑅𝐷𝑖𝑡𝛽 + (𝑅𝐷𝑖𝑡𝑥𝑖𝑡)𝛽 + 𝜖𝑖𝑡 (3) where 𝑦𝑖𝑡∗ is an unobservable latent variable that represents the sovereign credit rating of country 𝑖 in period 𝑡; 𝑅𝐷𝑖𝑡 is a vector of regional dummy variables;𝑥𝑖𝑡 is a vector of time-varying explanatory variables; 𝛽 is a vector of unknown parameters; and 𝜖𝑖𝑡 is a random disturbance term. The East Africa region will serve as benchmark in the model because it included the minimum amount of countries in the sample. Each determinant (as shown in table 4.4) will thus interact with the respective regional binary variables. The interactive binary variables will highlight the differences in the significant determinants between different regions in the model.
The income classification model will be specified as follows:
𝑦𝑖𝑡∗ = 𝐼𝐷
𝑖𝑡𝛽 + (𝑅𝐷𝑖𝑡𝑥𝑖𝑡)𝛽 + 𝜖𝑖𝑡 (4)
where 𝑦𝑖𝑡∗ is an unobservable latent variable that represents the sovereign credit rating of country 𝑖 in period 𝑡; 𝐼𝐷𝑖𝑡 is a vector of income classification dummy variables; 𝑥𝑖𝑡 is a vector of time-varying explanatory variables; 𝛽 is a vector of unknown parameters; and 𝜖𝑖𝑡 is a random disturbance term. The low income countries will serve as benchmark in the model. Each determinant (as shown in table 4.4) will thus interact with respective income classification binary variables. The interactive binary variables will highlight the differences in the significant determinants between different income classifications in the model.
According to Livingston et al. (2008), marginal effects have to be calculated for an ordered probit model in order to analyse the economic significance of the independent variables. The marginal effects in ordinal outcome models are estimated as follows (Long and Freese, 2006):
𝜕Pr (𝑦=𝑚|𝑥) 𝜕𝑥𝑘
=
𝜕𝐹(𝜏𝑚−𝑥𝛽) 𝜕𝑥𝑘−
𝜕𝐹(𝜏𝑚−1−𝑥𝛽) 𝜕𝑥𝑘 (5)which is the slope of the curve connecting 𝑥𝑘 to Pr (𝑦 = 𝑚|𝑥)ceteris paribus. Marginal effects illustrate the effects of change in an interaction variable (for example GDP growth in North Africa) on the probability of changes in the ratings of NKC.
4.6 RESULTS
The regional interactive model will be presented and analysed first in table 4.5. Thereafter, the income classification interactive model will be presented.
Table 4.5: Results of the NKC Regional Panel Data Model
Variable Region Coefficient Marginal Effect %
Regional Binary Variables (Benchmark = East Africa) North Africa 5.8783 3.1774 Southern Africa 0.2831 0.0335 Central Africa 3.9115 2.1624 West Africa -2.1048 -1.149 GDP Growth North Africa -0.1045 -0.0504 Southern Africa 0.0092 0.0285 Central Africa -0.1994 ** -0.1092 West Africa 0.0522 0.0281 East Africa -0.0613 -0.0334 Fiscal Balance North Africa -0.0346 -0.0208 Southern Africa -0.0676 * -0.0341 Central Africa -0.1464 -0.0815 West Africa -0.0299 -0.0162 East Africa 0.0930 0.0506 External Balance North Africa -0.0508 -0.0262 Southern Africa 0.0945 *** 0.0446 Central Africa -0.0302 -0.0168 West Africa 0.0437 0.0237 East Africa -0.0191 -0.0104 External Debt North Africa -0.0146 ** -0.0074 Southern Africa -0.0010 -0.0016 Central Africa -0.0019 -0.0011 West Africa -0.0047 -0.0025 East Africa -0.0041 -0.0022
Variable Region Coefficient Marginal Effect % Southern Africa 0.1248 *** 0.0651 Central Africa -0.1257 -0.0694 West Africa -0.2220 -0.1204 East Africa 0.4354 0.2371 Inflation North Africa 0.0452 0.0247 Southern Africa 0.0543 0.0293 Central Africa 0.1007 0.0557 West Africa 0.0112 0.0059 East Africa -0.0245 -0.0134 Foreign Reserves North Africa 0.0746 *** 0.0374 Southern Africa 0.0349 *** 0.0205 Central Africa -0.0947 -0.0521 West Africa -0.0583 -0.0317 East Africa -0.1677 -0.0914
Per Capita Income
North Africa 0.0005 0.0004 Southern Africa 0.0012 *** 0.0007 Central Africa 0.0006 ** 0.0004 West Africa 0.0051 0.0028 East Africa 0.0031 0.0017 Corruption North Africa -0.0567 -0.0382 Southern Africa -0.0640 ** -0.0348 Central Africa -0.0759 -0.0419 West Africa 0.0080 0.0044 East Africa -0.0389 -0.0212 Regulatory Quality North Africa 8.7385 *** 4.6344 Southern Africa 2.3505 *** 1.3159 Central Africa 4.6316 *** 2.5449 West Africa 0.5002 0.2753 East Africa 1.2889 0.7007 Internet Users North Africa 0.0150 0.0068 Southern Africa -0.0935 *** -0.0468 Central Africa 0.1503 0.082 West Africa -0.0503 -0.0274 East Africa 0.0674 0.0367 Total Panel Observations 206 Log Likelihood -250.7632
*, **, *** 10%, 5%, 1% level of significance, respectively. Source: Model estimations.
It can be seen from table 4.5 that none of the regional binary variable are statistically significant, meaning that the various regions are not taken into account when ratings are conducted. However some of the regional interaction dummy variables are significant in the model showing that different determinants per region are important in terms of ratings.
From table 4.6, it is striking that none of the explanatory interaction variables are statistically significant for East and West Africa, that is, none of the included variables have any statistical power in explaining their specific credit ratings. For Central Africa, three explanatory variables turned out to be statistically significant: economic growth, GDP per capita and the regulatory quality of their respective governments. The economic growth variable is significant and it again exhibits the negative sign that was revealed in chapter 3 in the Fitch model. It is obvious that the macroeconomic indicator GDP plays an important role in determining the credit rating of countries in Central Africa. For the North Africa region the significant variables are the external debt, foreign reserves and the regulatory quality variable.
It can be deduced that this model is best geared towards Southern Africa. A total of eight of the 11 included variables are significant for this region. The fiscal balance, external balance, investment, foreign reserves, GDP per capita, corruption, regulatory quality and internet users are all significant in this model. If the significant variables for Southern Africa are compared to those of the general ordered probit model for NKC from chapter 3 (see table 3.7), it can be seen that all the variables that were significant in the original NKC model are also significant for Southern Africa in the regional model. The fiscal balance is the only variable that is significant in the regional model for Southern Africa, but not in the original NKC model.
The sign for the internet variable for Southern Africa is puzzling. A positive relationship is expected of this proxy for the technological advancement of countries. Therefore the more internet users there are in a country, the higher the credit rating is expected to be; however, in the regional model above, the sign is negative.
If the results for the regional model are considered, in addition to table 4.6, which is a combination of table 4.2 and table 4.3, it can be seen that the regional model is more geared toward the countries that fall under the lower middle and upper middle income classes.
Table 4.6: Country Classification According to Region and Income Class
North Southern Central West East
Low Malawi Mozambique Tanzania Zimbabwe DRC Rwanda Benin Ethiopia Uganda Lower Middle Egypt Morocco Lesotho Swaziland Zambia Cameroon Ghana Nigeria Senegal Kenya Upper Middle Algeria Tunisia Angola Botswana Mauritius Namibia South Africa Gabon
Source: Wikitravel (2015) and World Bank (2013c).
Table 4.6 shows that the countries of West and East Africa fall into the low and lower middle income categories with no countries in the upper middle income category. North Africa only has countries in the lower middle and upper middle income brackets. Central and Southern Africa have countries spread over all the income categories. Southern Africa has the most countries in the upper middle income category. The income model performs better in the lower middle and upper middle categories. This is confirmed in table 4.6, where the ordered probit model was conducted again with the income classification groups. This could be the reason why none of the included explanatory variables are significant for the West and East Africa regions. This could also explain why so many variables are statistically significant for the Southern Africa region.
The results of the income classification model are presented in table 4.7.
Table 4.7: Results of the NKC Income Classification Panel Data Model