**3. The impact of foreign aid on human development in Africa: a multifactorial approach/model**

and administrators) to developing countries, but this method is widely criticized. The OECD admits in its official reports that technical cooperation is "perhaps the most controversial type of aid" [15]. And the critics add the attributes of "ineffective" are based on a multitude of negative cases. Some international experts have been accused of introducing technologies and procedures that are inappropriate to developing countries' needs. Also, technical cooperation has been criticized for failing to increase the local theoretical and practical skills. For instance, many students who were trained overseas have opted to stay there, thus fueling a brain drain

Who are the donors? As we mentioned above, a large part of aid is channeled through many mul‐ tilateral agencies, such as the World Bank and United Nations. In one of the latest report, OECD [1] estimates around 200 multilateral donors and agencies involved in development assistance.

of local human capital.

140 International Development

0.00 0.02 0.04 0.06 0.08 0.10 0.12

**ODA as a per cent of GNI** 

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

**Figure 2.** Net ODA/GNI, DAC members. Source: Authors' compilation based on 2016 OECD Report [14].

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

2014

**Constant 2014 USD billion** 

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

**Figure 1.** Net official development assistance, DAC members. (a) Total ODA excludes debt forgiveness of non‐ODA claims in 1990, 1991, and 1992. (b) Preliminary data. Source: Authors' compilation based on 2016 OECD Report [14].

1987

1988

1989

1990 (a) 1991 (a) 1992 (a) 1993 1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015 (p)

Despite all efforts that have been made, the institutional framework of foreign aid still seems to function suboptimally. In September 2015, UN released a new Agenda for the next 15 years, suggestively entitled "Transforming our world: the 2030 Agenda for Sustainable Development." The purpose of the new agenda is to continue (and also to extend) the eight MDGs promoted since 2000. The main aim of this international project still remains the eradi‐ cation of poverty and promotion of sustainable development in the third world.

Although UN has solved some pressing issues (better access to medicines and healthcare technologies, better school infrastructure, etc.), it becomes increasingly evident that aid has little effect on growth in a country with an institutional framework unfavorable to economic and political freedom. As mentioned in our introduction, we included potential explanatory factors for development as, in addition to ODA, two other independent variables. The first one is Economic Freedom Index (ILE) published by Fraser Institute, a famous Canadian think‐tank. It is widely recognized in the economic literature that economic growth and development depend on (endogenous) economic, social, and political institutions. The second one is Polity score developed by Marshall in Polity IV Project [16]. This indicator reflects a spectrum of governing authority that spans from fully institutionalized autocra‐ cies through mixed, or incoherent, authority regimes (termed "anocracies") to fully institu‐ tionalized democracies.

Since foreign aid is expected to have better results in improving human life conditions (health‐ care and education) rather than promoting economic growth, we have opted in our model to decompose Human Development Index and test the impact of ODA, ILE, and Polity on some HDI subindicators. The three‐dimension index of HDI are: *Life Expectancy Index* measured by the indicator life expectancy at birth, *Education Index* based on indicators like mean years of studies and expected years of schooling, and *GNI Index* that is based on the indicator GNI per capita.

#### **3.1. The data**

African countries in the sample are Algeria, Angola, Benin, Burkina Faso, Botswana, Burundi, Cameroon, Cabo Verde, Central African Republic, Chad, Congo, Arab Rep Egypt, Ethiopia, Gabon, Gambia, Ghana, Guinea‐Bissau, Guinea, Cote d'Ivoire, Kenya, Lesotho, Libya, Mauritania, Madagascar, Mauritius, Malawi, Mali, Morocco, Mozambique, Namibia, Nigeria, Niger, Rwanda, South Africa, Senegal, Sierra Leone, Swaziland, Tanzania, Togo, Trinidad and Tobago, Tunisia, Uganda, Zambia, and Zimbabwe.

The first data series we have worked with is the Human Development Index (HDI) for African countries shown in **Figure 3**. This is an indicator that emphasizes the importance of people development, in terms of health, knowledge, and standard of living.

The data available for this indicator consists of a panel of 41 time series between 1980 and 2014 with up to 11 cross section values for the 41 African countries that were included in this research. The main challenge regarding this data is that it consists of data for every fifth year between 1980 and 2010 and then yearly data between 2000 and 2014. This makes it impossible to test for root unit in first and second difference due to the very low number of remaining observations. The ADF unit root test for HDI allows us to reject the null hypothesis that the indicator follows a root unit process in level, with individual trend, and intercept for each of the African countries with a significance level of 5.2%, below the targeted 10% considering the issues with data.

The second series of data, presented in **Figure 4** is regarding ODA received by the African states to promote economic development and welfare. Only the official grants and loans that include at least 25% grant were taken into consideration. In order to make this information comparable between the states we have chosen to divide ODA by the mid‐year population estimate and, thus, work with ODA per capita.

little effect on growth in a country with an institutional framework unfavorable to economic and political freedom. As mentioned in our introduction, we included potential explanatory factors for development as, in addition to ODA, two other independent variables. The first one is Economic Freedom Index (ILE) published by Fraser Institute, a famous Canadian think‐tank. It is widely recognized in the economic literature that economic growth and development depend on (endogenous) economic, social, and political institutions. The second one is Polity score developed by Marshall in Polity IV Project [16]. This indicator reflects a spectrum of governing authority that spans from fully institutionalized autocra‐ cies through mixed, or incoherent, authority regimes (termed "anocracies") to fully institu‐

Since foreign aid is expected to have better results in improving human life conditions (health‐ care and education) rather than promoting economic growth, we have opted in our model to decompose Human Development Index and test the impact of ODA, ILE, and Polity on some HDI subindicators. The three‐dimension index of HDI are: *Life Expectancy Index* measured by the indicator life expectancy at birth, *Education Index* based on indicators like mean years of studies and expected years of schooling, and *GNI Index* that is based on the indicator GNI per capita.

African countries in the sample are Algeria, Angola, Benin, Burkina Faso, Botswana, Burundi, Cameroon, Cabo Verde, Central African Republic, Chad, Congo, Arab Rep Egypt, Ethiopia, Gabon, Gambia, Ghana, Guinea‐Bissau, Guinea, Cote d'Ivoire, Kenya, Lesotho, Libya, Mauritania, Madagascar, Mauritius, Malawi, Mali, Morocco, Mozambique, Namibia, Nigeria, Niger, Rwanda, South Africa, Senegal, Sierra Leone, Swaziland, Tanzania, Togo, Trinidad and

The first data series we have worked with is the Human Development Index (HDI) for African countries shown in **Figure 3**. This is an indicator that emphasizes the importance of people

The data available for this indicator consists of a panel of 41 time series between 1980 and 2014 with up to 11 cross section values for the 41 African countries that were included in this research. The main challenge regarding this data is that it consists of data for every fifth year between 1980 and 2010 and then yearly data between 2000 and 2014. This makes it impossible to test for root unit in first and second difference due to the very low number of remaining observations. The ADF unit root test for HDI allows us to reject the null hypothesis that the indicator follows a root unit process in level, with individual trend, and intercept for each of the African countries with a significance level of 5.2%, below the targeted 10% considering the issues with data.

The second series of data, presented in **Figure 4** is regarding ODA received by the African states to promote economic development and welfare. Only the official grants and loans that include at least 25% grant were taken into consideration. In order to make this information comparable between the states we have chosen to divide ODA by the mid‐year population

tionalized democracies.

142 International Development

Tobago, Tunisia, Uganda, Zambia, and Zimbabwe.

estimate and, thus, work with ODA per capita.

development, in terms of health, knowledge, and standard of living.

**3.1. The data**

**Figure 3.** Evolution of HDI in African countries. Source: Authors' compilation based on UNDP Human Development Report 2015.

**Figure 4.** Net ODA for African countries. Source: Authors' compilation based on OECD and World Bank data.

ODA is a relevant indicator for the financial support received in order to develop the wellbe‐ ing of people in that country rather than poverty support. Also, ODA represents around 80% of the total development support that an African country receives. The other 20% is regularly coming from NGOs being focused on alienating poverty in specific areas.

For this indicator we will use a panel of data with 43 time series, one for each of the African countries and 40 cross sections for the period between 1976 and 2015. We have tested the series for unit root with multiple tests for both trend and difference, both for common root and for individual root unit, and the data proved to be stationary. We have rejected the null hypothesis of root unit with a significance level close to zero.

The third data series reflects the authority of the political regime in the country, measured by Polity score presented in **Figure 5**. The polity score captures this regime authority spectrum on a 21‐point scale ranging from −10 (hereditary monarchy) to +10 (consolidated democracy). According to the authors, the Polity scores can also be converted into regime categories in a suggested three part categorization of "autocracies" (−10 to −6), "anocracies" (−5 to +5 and three special values: −66, −77, and −88), and "democracies" (+6 to +10).

When testing for unit root, the Polity2 series proved to be stationary both in trend and difference, taking into account both the individual effects and the individual linear trends, just like in all the other tests. Thus, the null hypothesis can be rejected and we can use the polity data in modeling.

**Figure 5.** Polity scores in African countries. Source: Authors' compilation based on Polity IV Project data.

The last data series we work with is Economic Freedom Index (ILE), presented in **Figure 6**. In many ways, a country's economic freedom ranking is a measure of how closely its institutions and economic policies are compared with the idealized structure implied by the standard text‐ book analysis of economics. It uses 42 distinct variables to create an index, which is measured in 5 areas: size of government, legal structure and security of property rights, access to sound money, freedom to trade internationally, and regulation of credit, labor, and business [17]. The countries are ranked on a scale between 1 and 10, moving from the less free to the most free.

ODA is a relevant indicator for the financial support received in order to develop the wellbe‐ ing of people in that country rather than poverty support. Also, ODA represents around 80% of the total development support that an African country receives. The other 20% is regularly

For this indicator we will use a panel of data with 43 time series, one for each of the African countries and 40 cross sections for the period between 1976 and 2015. We have tested the series for unit root with multiple tests for both trend and difference, both for common root and for individual root unit, and the data proved to be stationary. We have rejected the null

The third data series reflects the authority of the political regime in the country, measured by Polity score presented in **Figure 5**. The polity score captures this regime authority spectrum on a 21‐point scale ranging from −10 (hereditary monarchy) to +10 (consolidated democracy). According to the authors, the Polity scores can also be converted into regime categories in a suggested three part categorization of "autocracies" (−10 to −6), "anocracies" (−5 to +5 and

When testing for unit root, the Polity2 series proved to be stationary both in trend and difference, taking into account both the individual effects and the individual linear trends, just like in all the other tests. Thus, the null hypothesis can be rejected and we can use the polity data in modeling.

**Figure 5.** Polity scores in African countries. Source: Authors' compilation based on Polity IV Project data.

coming from NGOs being focused on alienating poverty in specific areas.

hypothesis of root unit with a significance level close to zero.

144 International Development

three special values: −66, −77, and −88), and "democracies" (+6 to +10).

When testing for unit root, the series have shown that we cannot reject the null hypothesis for common unit root process according to Levin, Lin, and Chu test (548.454 statistical value) under first difference. Thus the series has been transformed to stationary by differentiation. ILE was tested for root unit with multiple tests and has proven to be nonstationary with an ADF‐Fischer value of 81.7 (9.12% confidence level, higher than the targeted 5%). Thus, we have rejected the null hypothesis and transformed the data by first level differentiation. Thus, we will test the effects of Delta ILE over the endogenous variables. We found that the new series proved to be stationary.

GNI was tested for unit root with multiple tests and has proven to be nonstationary with an ADF‐Fischer value of 84.3 (59.12% confidence level for the 44 cross sections, 1073 observa‐ tions) for individual effects and individual trend. Thus, we have rejected the null hypothesis and transformed the data by first level differentiation. Thus, we will test the effects of the exogenous variables on Delta GNI. The new series proved stationary.

**Figure 6.** Economic Freedom Index for African countries. Source: Authors' compilation based on Economic Freedom Index of the World, 2015 Annual Report.

Life expectancy at birth data was tested for root unit with multiple tests and has proven to be nonstationary with an ADF‐Fischer value of 102.5 (13.7% confidence level for 44 cross sec‐ tions, 1010 observations). Thus, we have rejected the null hypothesis and transformed the data into percentage change of life expectancy at birth (DPLE). The new series proved stationary.

#### **3.2. The methodology and findings**

In the first model, life school expectancy was regressed versus the exogenous factors ODA, DILE, and Polity.


Depending on the lags used for ODA and ILE, the model explains between 35% (DILE lagged one period) and 42.6% (no lags) of the life school expectancy. Although the model partially explains life school expectancy, the parameters associated with the three variables used can be accepted as relevant with low probabilities. ODA is the most relevant variable in all mod‐ els with 93% relevance in the model with no lags, 85% in the model with lagged DILE, and 30.5% or 45% in the two models with lagged ODA. In all models, the relation between ODA and life school expectancy proves to be negative showing that the main problems regarding the education in Africa are not improved with the development aid received by the countries.

Improvements in economic freedom appear to have a negative and weak short term effect with around 75% relevance in the models where the indicator is not lagged. With 1year lagged DILE, the impact of economic freedom becomes positive the next year but the impact is even less relevant (46% t stat relevance). This result may be commented that in the years when economic freedom improves, there is a slight incentive to give up schooling for other benefits. Political regime authority changes also prove to have a very limited effect on the life school expectancy.

Life expectancy at birth data was tested for root unit with multiple tests and has proven to be nonstationary with an ADF‐Fischer value of 102.5 (13.7% confidence level for 44 cross sec‐ tions, 1010 observations). Thus, we have rejected the null hypothesis and transformed the data into percentage change of life expectancy at birth (DPLE). The new series proved stationary.

In the first model, life school expectancy was regressed versus the exogenous factors ODA,

(0.463922) (0.596327) (0.573915) (0.479372)

(0.007355) (0.007965) (0.009202) (0.007812)

(0.852597) (0.988261) (0.883243) (0.865338)

(0.118211) (0.149517) (0.122132) (0.121545)

**Equation 1 2 3 4** Dependent variable SCH SCH SCH SCH Lags None ILE(−1) ODA(−1) ODA(−2)

(fixed cross‐section effects, white adjustment)

Constant 5.944561 5.726093 5.449028 5.530522

ODA −0.013381 −0.01159 −0.002397 −0.004658

DILE −1.011094 0.469779 −1.027069 −0.996587

Polity2 −0.093381 −0.179178 −0.122334 −0.114812

Adjusted period 2001 2014 2001 2014 2001 2014 2001 2014 Observations 472/40 472/40 472/40 472/40 Adjusted *R*<sup>2</sup> 0.426965 0.356552 0.422928 0.423342 Sum squared reside 6802.978 7640.24 6850.896 6845.982 S.E. of regression 3.98218 4.22012 3.99618 3.994747 *F*‐statistic 9.355682 7.214153 9.218801 9.232748

Prob(*F*‐statistic) 0 0 0 0

Depending on the lags used for ODA and ILE, the model explains between 35% (DILE lagged one period) and 42.6% (no lags) of the life school expectancy. Although the model partially explains life school expectancy, the parameters associated with the three variables used can be accepted as relevant with low probabilities. ODA is the most relevant variable in all mod‐ els with 93% relevance in the model with no lags, 85% in the model with lagged DILE, and 30.5% or 45% in the two models with lagged ODA. In all models, the relation between ODA and life school expectancy proves to be negative showing that the main problems regarding the education in Africa are not improved with the development aid received by the countries.

**3.2. The methodology and findings**

Method Panel least squares (unbalanced)

DILE, and Polity.

146 International Development

In the second model we have looked for the impact of ODA, ILE, and Polity over the growth of GNI. Since, after the transformation, the domestic gross national income (DGNI) has a distribution that is different than the normal one, the residuals of all the GNI models have a skewness close to zero but a high kurtosis. Since the GNI data needed to be converted to stationary leading to negative values, the log transformation cannot be applied. We have cho‐ sen to accept the nonnormal distribution of the errors instead of performing other box‐cox transformation of the data, based on the fact that when the sample size is large enough (472 observations in our case), the violation of the normality assumption does not cause major problems [18].


A sensitivity analysis was performed by modifying the lags of the data in order to see modi‐ fications past periods bring to the model.

Delta GNI was regressed versus the exogenous variables Delta ILE, ODA, and Polity2. Comparing the coefficients for the three variables with their variance (under brackets), for each specification of the model we find out that in the model with no lags, explaining 31.4% of the DGNI variance, the Polity indicator is the most relevant exogenous variable with DILE less rel‐ evant and ODA almost not relevant at all. In this case, ODA is also negatively correlated with the changes in GNI. In the second model with lagged DILE, that explains 24.2% of the variance, this lagged indicator becomes the most relevant of all showing us that changes in economic freedom take time to produce improvements in GNI. ODA still has an extremely low relevance but becomes slightly positive while Polity2 becomes the number 2 indicator in relevance, very close to lagged DILE. The third and fourth models with lagged ODA, that explain 32% and 31%, respectively, are still showing a low relevance for ODA but an even lower relevance for the nonlagged DILE. Polity2 remains the only important exogenous variable in the model.

The third model is the regression for a double differentiated series of life expectancy depend‐ ing on the same three indicators: ODA, DILE, and Polity2. Since the model has led to auto cor‐ related residuals, we have preferred using the EGLS white consistent method of estimating the parameters instead of Cochrane‐Orcutt. Since the volume of the database is high (472 observa‐ tions), EGLS will maintain similar properties for the estimators as per the regular OLS [19].


The most relevant parameter in all the regressions above is the one attached to ODA (99%) in all regressions, closely followed by Polity2 (approx. 95%) and then DILE (19–65%). Thus, between the three variables, ODA is the most important factor related to life expectancy as it is most often correlated with an inflow of know‐how and pharmaceutical innovations.
