This study has examined sectoral analysis of the impact of foreign aid on aggregate and sectoral economic growth in Ethiopia over the period 1981 to 2012 using multivariate Vector auto regression analysis. All the necessary time series tests such as stationary test, co-integration test, weak exiguity test, vector error correction, and causality test in vector error correction model and the like are conducted. The empirical result from the growth equation shows that aid has a significant positive impact on educational sector GDP in the long run. On the other hand, foreign aid has positive but insignificant impact on real GDP, agriculture GDP, and health sector GDP of Ethiopia. Foreign aid is effective in enhancing growth at aggregate level of the economy in general and education sector of the economy of Ethiopia in particular. The test result of the study result reveals that there is a bi-directional causal relationship between educational GDP and educational foreign aid in Ethiopia. However, the agricultural and health sector does not show any bi-directional causality with their respective sector aid. This implies that all aid allocated for sectors is ineffective all in all in achieving its objectives of economic development. Therefore, aid recipient country like Ethiopia has to work how to enhance the domestic revenue raising capacity of the country which is at the heart of the mechanism to meet the capital required for the economy in times of short falls and ineffectiveness of external resources.
Table of Content
Acknowledgment
Acronyms
List of Table
List of Figure
Abstract
CHAPTER ONE: INTRODUCTION
1.1. Background
1.2. Statement of the Problem
1.3. Objective of the Study
1.3.1. General Objectives
1.3.2. Specific Objectives
1.4. Research Questions
1.5. Hypothesis to be Tested
1.6. Significance of the Study
1.7. Scope and Limitation of the Study
1.7.1. Scope of the Study
1.7.2. Limitation of the Study
1.8. Organization of the Paper
CHAPTER TWO: METHODOLOGY OF THE STUDY
2.1. Data Types and Sources
2.2. Description of Variables
2.3. Model Specification
2.4. Econometric Estimation Techniques
2.4.1. Testing for Unit Root
2.4.2. Test for Cointegration
2.4.3 Vector Error Correction Model
2.4.4 Causality Test in Vector Error Correction Model
CHAPTER TREE: RESULT AND DISCUSSION
3.1. Descriptive Analysis of Macroeconomic Performance in Ethiopia
3.1.1. Analysis of Aggregate Economic Growth in Ethiopia
3.1.2. Sector Wise Analysis of Economic Growth in Ethiopia
3.1.3. Trends of Domestic Savings and Investment in Ethiopia
3.1.4. Trends of External Sector Economic Growth in Ethiopia
3.1.5. Government Expenditure and Economic Growth
3.1.6. Foreign Aid (ODA) and Economic Growth in Ethiopia
3.1.7. Sectorial Composition of Foreign Aid (ODA) in Ethiopia
3.2. Empirical Results Analysis and Interpretations
3.2.1. Results for Unit-Root Test
3.2.2. Co-integration and Long run Growth Equation Analysis
3.2.3. Short Run VECM Estimation of Economic Growth
3.2.4. Causality Analysis of Aggregate and Sectoral Economic Growth
CHAPTER FOUR: SUMMARY AND CONCLUSION
4.1. Summary
4.2. Conclusion
4.3. Policy Implication
Reference
Appendixes
Acknowledgment
First of all I would thank the Almighty God for his an invaluable helps for all things that human beings can perform with help of his supreme power.
Then, I would like to express my sincere gratitude to my major advisor Dr Amsalu Bedemo (PHD) and Mr Wondimu Saketa (MSC) for their genuine and constructive ideas of providing their comments and their patience to go through and forward comments with respect to the work of research work on the title of sectorial analysis of the effect of foreign aid in Ethiopia.
My special thanks would be extended to Mekonnen Bersisa (PhD Candidate), who has given me a genuine and constructive comments starting from the modification of the title by sharing from the meager time he has for his personal study.
Last but not list, I have a great full thanks to my wife S/r Deme Chemeda, our sister Adanech Iteffa and others for their support and taking care of our children Firomsa Fikadu and Bilise Fikadu during my thesis work and reviewing of different sources for the interesting work.
Acronyms
illustration not visible in this excerpt
List of Table
Table 1 : Average growth rates of real GDP, per capita GDP and population
Table 2: Annual growth of industrial sectors’ value added (% of GDP)
Table 3: Domestic saving and capital formation annual growth (% of GDP)
Table 4: Annual growth rate of import and export (% of GDP)
Table 5: Current, capital, & external assistance share of GDP and total expenditure
Table 6: Percentage shares of net ODA to economic sectors
Table 7: Three years average share of sectoral distribution of aid (% of total aid)
Table 8: ADF unit root test result for variables in growth equation
Table 9: Johansen co-integration tests result for aggregate Ethiopian economy
Table 10: Johansen co-integration tests result for agricultural sector
Table 11: Johansen co-integration tests result for educational sector
Table 12: Johansen co-integration tests for health sector
Table 13: Normalized long run β Coefficients of aggregate & sectorial level
Table 14: Adjustment (α) coefficients of aggregate and sectorial level
Table 15: Results of weak exogeneity test for aggregate and sectorial levels
Table 16: Result of zero restriction test on β coefficients for aggregate and sectors
Table 17: Stationary of error correction terms in aggregate and sectors
Table 18: Result for dynamic growth equation for real GDP
Table 19: Result for dynamic growth equation for educational and health GDP
Table 20: Johansen cointegration tests for the causality test detection
Table 21: Causality test between GDP and aid at aggregate and sectoral level
List of Figure
Figure 1: Ethiopian and Sub-Sahara African GDP annual growth
Figure 2: Share of major industrial sectors value added (% of GDP)
Figure 3: Export, import, and trade balance trends (% of GDP)
Figure 4: Export, import, and trade balance trends (% of GDP)
Figure 5: Flow of Bilateral, multilateral and Non DAC aid to Ethiopia
Abstract
This study has examined sectoral analysis of the impact of foreign aid on aggregate and sectoral economic growth in Ethiopia over the period 1981 to 2012 using multivariate Vector auto regression analysis. All the necessary time series tests such as stationary test, co-integration test, weak exiguity test, vector error correction, and causality test in vector error correction model and the like are conducted. The empirical result from the growth equation shows that aid has a significant positive impact on educational sector GDP in the long run. On the other hand, foreign aid has positive but insignificant impact on real GDP, agriculture GDP, and health sector GDP of Ethiopia. Foreign aid is effective in enhancing growth at aggregate level of the economy in general and education sector of the economy of Ethiopia in particular. The test result of the study result reveals that there is a bi-directional causal relationship between educational GDP and educational foreign aid in Ethiopia. However, the agricultural and health sector does not show any bi-directional causality with their respective sector aid. This implies that all aid allocated for sectors is ineffective all in all in achieving its objectives of economic development. Therefore, aid recipient country like Ethiopia has to work how to enhance the domestic revenue raising capacity of the country which is at the heart of the mechanism to meet the capital required for the economy in times of short falls and ineffectiveness of external resources.
Key words: foreign aid, economic growth, sector, aggregate Ethiopia, impact
CHAPTER ONE: INTRODUCTION
1.1. Background
Developing countries face challenges of massive poverty, slow Gross Domestic Product (GDP) growth, high mortality rates from illnesses, and low levels of education. More than 20 percent of the developing world (1.1 billion) people subsist on less than $1 a day. In some developing countries, such as those of Sub-Saharan Africa (SSA), more than 70 percent of the population lives in extreme poverty (i.e. the proportion of population earning less than $2 a day) (Kumler'07, 2007; Leeson, 2008). The governments in these countries do not have sufficient financial resources to fight these challenges effectively. Foreign aid has played an instrumental role in the implementation of development programs to combat the challenges (Lohani'04, 2004; Pattillo, Polak, & Roy, 2007).
Foreign aid is defined as any flow of capital to a developing countries for the objective that should be non commercial from the point of view of the donor on development, poverty reduction, or income distribution grounds and it should be characterized by concessional terms; that is, the interest rate and repayment period for borrowed capital should be softer (less stringent) than commercial terms (Bakare, 2011; Randhawa, 2012; Todaro & Smith, 2012). Official Development Assistance (ODA), commonly known as foreign aid is a flow of financial resources from developed countries to developing countries on development grounds (Eroglu & Yavaz, 2005; Moreira, 2005; Leeson, 2008; Paul & Pistor, 2009). It is an international transfer of public funds in the form of loans or grants either directly from one government to another (bilateral) or indirectly through multilateral assistance agency such as International Monetary Fund and World Bank (Radelet, 2006; UNCTAD, 2006; Abuzeid, 2009; OECD, 2009).
Similarly, Development Assistance Committee (DAC) defines ODA as a grants or loans to developing countries and multilateral agencies active in development that are undertaken by the official sector at concessional terms (if a loan, having a grant element of at least 25%), with the promotion of economic development and welfare as the main objective. Technical cooperation is also included in aid. Grants, loans, and credits to be used in military purposes are excluded (OECD, 2009; Paul & Pistor, 2009).
Despite the volume of ODA from bilateral and multilateral agency flows has grown from an annual rate of under $5 billion in 1960 to $50 billion in 2000 and then to over $128 billion in 2008, the percentage of developed country Gross National Income (GNI) allocated to ODA declined from 0.51 percent in 1960 to 0.23 percent in 2002 before improving to 0.33 percent by 2005 and then to 0.45 percent in 2008 as part of a campaign to increase assistance in the wake of the continued lag in human development in Sub-Saharan Africa (OECD, 2009; Paul & Pistor, 2009; Todaro & Smith, 2012).
Foreign aid is important source of finance in most countries in SSA where it supplements low savings, narrow export earnings and thin tax bases (Bhattarai, 2007; Arellano, Bulíř, Lane, & Lipschitz, 2009). Africa in general and SSA in particular receives a greater share of global aid than any other region in the world with East Africa receiving approximately 25 percent of all ODA to SSA. Within East Africa, Ethiopia receives the largest percentage (7%) of total ODA from all donors, followed by Tanzania (6%). According to OECD DAC statistics, aid to Ethiopia increased from US$1.1 billion in 1995 to US$3.5 billion in 2010 and is concentrated on core social sectors and infrastructure (Nganwa, 2013; Prizzon & Rogerson, 2013).
Most foreign aid is designed to meet one or more of broad economic and development objectives: to stimulate economic growth through building infrastructure, supporting productive sectors such as agriculture, or bringing new ideas and technologies; to strengthen education, health, environmental, or political systems; to support subsistence consumption of food and other commodities, especially during relief operations or humanitarian crises; or to help stabilize an economy following economic shocks (WB, 1998; Radelet, 2006; Bakare, 2011).
1.2. Statement of the Problem
Despite these broader objectives of foreign aid as well as the tremendous increases in the follow of foreign aid to developing countries like Ethiopia (under $5 billion in 1960 to $50 billion in 2000 and to over $128 billion in 2008) (OECD, 2009) from time to time, there is controversies about aid effectiveness which go back to decades. There are two general but contending schools of thought with regard to the effectiveness of aid in spurring economic growth in the third world(Gebhard, Kitterman, Mitchell, Nielson, & Wilson, 2008). Research results on the macroeconomic impacts of foreign aid in developing countries are ambiguous. Some studies concluded that this relationship was negative; others concluded it was positive, and others found no relationship at all. Several studies have noted that foreign aid has a positive effect on the economic growth of poor countries (Chenery & Strout, 1966; Burnside & Dollar, 2000; Lohani'04, 2004; Bhattarai, 2007).
Foreign aid is advocated the promotion of economic development and could play a vital role in promoting economic development in the Less Developed Countries (LDCs) which is explained in term of concepts such as the savings gap and the foreign exchange gap (Eroglu & Yavaz, 2005; Shah, Ahmad, & Zahid, 2005). Critics of foreign aid argue that foreign aid discourages domestic saving and domestic tax revenue in developing countries and is simply diverted into consumption instead of investment. They find that grant have a negative effect on domestic revenue. The decline domestic revenue is almost as much as the increase in grant (Benedek, Crivelli, Gupta, & Muthoora, 2012; Bwire, Morrissey, & Lloyda, 2013). Since these countries do not have the technical ability to use the aid effectively, it gets spent on nonproductive activities and poorly conceived projects concluding that there exists no significant correlation between aid and GDP growth because the majority of foreign aid is spent on consumption (Lohani'04, 2004).
The study of foreign aid on a sectoral basis that is conducted by cross country analysis doesn’t allow us to clearly examine the sectoral impact of aid on sector’s spending. In this respect, different countries have different result for the impact of aid on sectors spending (Paul & Pistor, 2009). Projects financed by foreign aid are often highly visible and important successes such as roads and highways, schools and health clinics, irrigation infrastructure, power plants and so on. But success can be assessed at two levels that are at the micro or project level, which typically shows high rates of success or at the macro level of economy wide growth and poverty reduction, where there is less visible success (WB, 1998; Pettersson, 2007). Generally, there seemed to be a micro–macro paradox, with evidence suggested that aid clearly worked at the micro level but evidence, on balance, relating to the macro level (McGillivray, Feeny, Hermes, & Lensink, 2005; Odusanya, Abidemi, & Akanni, 2011).
Odusanya et al (2011) carried out studies on the impact of foreign aid and public expenditure on economic growth in Nigeria. Their result reveals that foreign aid and public expenditure impact positively on the economic growth with foreign aid indicating a very significant impact on growth. They conclude that foreign aid can have positive effect on economic growth, through public expenditure if properly channeled to the productive sectors of the economy. But Odusanya et al (2011) carried out their studies using static framework. Their conclusions may be incorrect because the economic agents associated with aid administration are dynamic. Also, they did not disaggregate the GDP into different sector levels for the sectorial level study they have conducted. Therefore, this research tends to use dynamic model that will be able to capture the dynamic nature of these agencies. Also, for the sectorial level analysis, this study will use sectorial level GDP instead of aggregate GDP. Hence, such work can usually be conducted in the context of country specific, to capture the different impact of earmarked aid on the sectors. It is therefore pertinent to check whether and to what extent has foreign aid might have caused or contributed to economic growth in Ethiopia.
1.3. Objective of the Study
1.3.1. General Objectives
The major objective of the study is to investigate sectoral analysis of the impact of foreign aid on economic growth in Ethiopia
1.3.2. Specific Objectives
The specific objectives of the study are to:
- Analyze the trend of macroeconomic performance Ethiopia under two regime so as to draw implications for economic growth;
- Analysis the causality between foreign aid and economic growth in Ethiopia;
- Analyze the impact of foreign aid on agriculture, education and health sector of economic growth in Ethiopia;
1.4. Research Questions
The above specific issues can be met by answering the following research questions.
- What is the causal relationship that exists between foreign aid and economic growth in Ethiopia?
- What is the impact of foreign aid on economic growth of developmental sectors?
- In which sectors of the economy does foreign aid has significant effect?
1.5. Hypothesis to be Tested
This study is inspired by a new stream of research supports aid works better at sectoral level (micro level) in an individual country based study than at cross country level (macro-level). Specifically the main hypotheses to be empirically tested are:
1. There is no a positive relationship between economic growth and foreign aid in Ethiopia.
2. Foreign aid does not Granger-cause GDP growth in Ethiopia.
3. Sectoral foreign aid does not have impact on sectoral economic growth of agriculture, health and education.
1.6. Significance of the Study
The study of sectoral analysis of the impact of foreign aid in Ethiopia is important due to the fact that foreign aid may have differentiated impacts at the sector levels as compared to the result obtained when analyzed at an aggregate level. It is noteworthy that the impact of foreign aid at the sectoral level has not been given consideration in analysis of impact of foreign aid in Ethiopia at large. Therefore, the result of this study is useful for improving policy design, institutional set up, implementation, monitoring and evaluation in the area of foreign aid allocation to public spending in general and sector wise in particular for the sake of economic growth. Finally, the result of the study with respect to sectoral analysis of foreign aid becomes stepping stone for academicians, researchers, students, policy makers and other organization that are in need to use its result as an input in their organization.
1.7. Scope and Limitation of the Study
1.7.1. Scope of the Study
This study covers a period of thirty two years for a country level analysis in Ethiopia. The study uses aggregate aid, sectoral aid, Government spending, GDP, and other time series data for the period of 1981 to 2012 is used.
1.7.2. Limitation of the Study
The model analysis of the impact of foreign aid in Ethiopia assumes that all aid is allocated through government sector and none of it will be spend though other ways. This is due to the fact that in developing countries like Ethiopia, aid may have come through Non-governmental organizations or other sources that are understate the whole figure of aid that may be used in the study.
The other critical limitation of this study is that in the analysis of aid flows to sectors with recipient’s country wise in the context of the total aid flow during the past three or more decades is greatly hampered by the lack of complete, reliable, and consistent data. The only systematic data available and easily accessible for researcher is foreign aid flows that have been compiled by the OECD/DAC. However, the easily accessible OECD/DAC website does not provide a consistent and comparable breakdown of aid flows from both multilateral and bilateral donors from the early 1970s to the present. This source provides a sectoral classification of aid flows from the 1973 onward only for the bilateral aid and not for the rest of the donors.
Furthermore, the sectoral classification of the total aid flows to individual country like Ethiopia is available only from 1995 to 2012 that the researcher used in the study. Moreover, the data available regarding aid to the sector from the OCED/DAC database has a category of “multi-sectoral” aid which includes aid to more than one sector. Therefore, it is not clear for researcher how much aid from multi-sectoral aid is directly or indirectly goes to which sector of the economy in recipient countries. This also makes us to underestimate the aid allocated to agriculture, education and health sectors which are the main concerns of the study.
1.8. Organization of the Paper
The study of research is organized under five chapters. The first chapter deals with the problems and its approaches, which include background of the study, statement of the problem, research objectives, research questions to be answered, significance of the study, hypothesis to be tested, scope and limitation of the study, and organization of the research paper. The second chapter deals with methodology of the study that includes source and type of data, description of variables, model specification, and estimation techniques .Chapter three deals with macroeconomic performance of the country which includes: GDP & its growth trend including per capita GDP, and the share of main economic sectors (agriculture, industry and service) in GDP; trends of gross fixed capital formation and saving; the trends in external market (export-import) and its annual growth; and finally trends and sectoral compositions of ODA flows to Ethiopia. Chapter four presents the estimation results and time series characteristics of the data and tests for long-run relationships and short run model and finally chapter five, the last chapter gives summary, conclusions and policy implication of the study. All the reference materials used in the study are listed under reference.
CHAPTER TWO: METHODOLOGY OF THE STUDY
2.1. Data Types and Sources
Concerns of analyzing the impact of foreign aid are important for Ethiopia due to an increasing per capita of foreign aid and the country’s dependence on it. The study used time series data. For analyzing the impact of foreign aid on economic growth of Ethiopia, the study used country level macro-data covering the period from 1981 to 2012. The choice of the period is based on the availability of relevant data for the study. The relevant data was collected from various sources: National Bank of Ethiopia, Ministry of Finance and Economic Development (MoFED), Ethiopian Economic Association, World Bank, World Development Indicator database, and OECD/CRS websites.
2.2. Description of Variables
The major variables that are used in the sectoral analysis foreign aid in Ethiopia include: Real GDP which is the real GDP of an economy over the period 1981-2012 for Ethiopia both at aggregate and at sectoral level for agriculture, health and education which are dependent variables. Furthermore, AGDP is agricultural GDP, EDGDP is educational GDP; and HEGDP is health sector GDP of the country that are used in the analysis. FAID is the ODA as defined by the OECD as percentage of GDP. Also EDUCAID, AGRIAID, and HEALAID are official development assistance allocated for educational, agriculture and health sector as percentage of GDP of the country respectively.
PRIV is the total net private capital flows as a percentage of GDP (international remittance can also be classified as part of net private capital flows). Private capital flows consist of net foreign direct investment and portfolio investment. SAV is the domestic savings as a percentage of GDP. TRADE is the openness to trade, which is defined as (X + M) /GDP. That is the addition of export and import divided by GDP. Openness to trade is often hypothesized to raise growth through several channels, such as access to advanced technology from abroad, possibilities of catch-up, greater access to a variety of inputs for production, and access to broader markets that raise the efficiency of domestic production through increased specialization. Various measures of openness have been proposed and tested, with no single ‘best’ measure emerging. For instance, (Durbarry, Gemmell, & Greenaway, 1998) used to measure trade openness by the ratio of total trade to GDP and changes in the terms of trade. X is total value of goods and services exported and M is total value of goods and services imported. Moreover, GOV represents the total amount of government expenditure (developmental plus non developmental) as a percentage of GDP for the period under consideration.
D is dummy variable for major political changes from Derg regime to Ethiopian peoples’ Revolutionary Democratic Front (EPRDF) that takes in to account to see the effect of major shifts in political environment on the performance of economic growth and foreign aid in the short run. The dummies are incorporated in to the model for growth and D took 1 for 1992 to 2012 and 0 otherwise.
2.3. Model Specification
Different researcher uses different models. This study employs Vector Autoregressive (VAR) model. One of the main advantages of VAR model over single equation model is that its ability to deal with several endogenous variables(Hayakawa, 2011) and cointegrating vectors, the ability to test for weak exogeneity and parameter restrictions, and to handle both I(1) and I(0) variables in one system. The VAR approach is data based and little economic theory is imposed directly (M’Amanja, Lloyd, & Morrissey, 2005). VAR assumes that all the variables are endogenous.
The model used in this study was rely on the framework of Durbarry, et al (1998) as modified by Odusanya et al (2011) for the case of Nigeria. However, in order to meet the objectives of study in terms of its concentration, as well as on the bases of availability of data on the appropriate variables, foreign aid, total private capital flows, and domestic saving mobilized as a vector of capital sources (domestic and foreign), and openness to trade and total government expenditure as is a vector of control variables from economic growth components over the other variables to capture potential side effects of foreign aid such as ‘Dutch-Disease’ effects and other policy variables that are hypothesized to affect growth are incorporated (Odusanya, Abidemi, & Akanni, 2011). It was hypothesized that economy growth can be related to foreign aid as:
illustration not visible in this excerpt
However, in order to concentrate on the dynamic effect of foreign aid on the economic growth at the sectoral levels, equation 1 is first modified to VAR model as:
illustration not visible in this excerpt
Where FA and AG are foreign aid and annual GDP growth respectively, while and represent their lagged values in j years, p is the maximum lag length. That is, the Lag Exclusion Wald Tests will be used to select the most appropriate lag length, and l and μ are error terms and, β’s and a’ s are parameters to be estimated.
Differencing away the fixed effects in equation 2 and 3 by including TRADE, GOV and D (D is dummy variable) it was transformed to equations 4 and 5 respectively which will be used to examine the impact of foreign aid on the Ethiopian economy at the aggregate level. Where Δ denotes the first difference operator, are random error terms.
illustration not visible in this excerpt
In the sectoral analysis, the impact of foreign aid on agriculture, health and education will be analyzed. Going by equations 4 and 5, FA will be the foreign aid for agriculture, health and education respectively, while AG will be the agriculture, health, and educational sector GDP growth respectively for the period under consideration in the sectoral analysis.
2.4. Econometric Estimation Techniques
The method employed in the study is based on recent advancements in the theoretical and empirical aid-growth relationships. As the data used is time series, testing for stationarity (unit root test), co-integration test, Vector Error Correction Model (VECM), and causality test is conducted in the model. The standard estimation and hypothesis testing assumed that all, in particular regression, variables are stationary. A data series is said to be stationary if its error term has zero mean, constant variance and the covariance between any two time periods depends only on the distance or lag between the two periods and not on the actual time which it is computed.
However in reality most macroeconomic variables are non stationary. If variables entering into the estimation are non stationary, then the result obtained using OLS technique would be spurious in the sense that variables would seem to have promising diagnostic test (high R2 and low Durbin Watson test) result just because they have common trend over time rather than actual causation(Harris, 1995). Therefore hypothesis testing and inference using such results will be invalid. To avoid such wrong inferences from the non stationary regressions, the time series property of the data should be checked prior to the estimation of the long run model.
2.4.1. Testing for Unit Root
Unit root test has become a widely popular approach to test for stationary. It has been argued that if a variable has deterministic trend, including trend variable in the regression removes the trend component and makes it stationary. Such process is called trend stationary since the deviation from the trend is stationary. However, most time series data have a characteristic of stochastic trend (that is, the trend is variable cannot be predicted with certainty). In such cases, in order to avoid the problem associated with spurious regression, pre-testing the variables for the existence of unit roots (i.e. non-stationary) becomes compulsory. In general if a variable has stochastic trend, it needs to be differenced in order to obtain stationary. Such process is called difference stationary process (Gujarati D. , 1995). The number of unit roots a given variable posses determines how many times that variable should be differenced in order to attain stationary.
A commonly applied formal test for existence of a unit root in the data is the Dickey-Fuller (DF) tests. It’s simple extension being the Augmented Dickey Fuller (ADF) test. The augmentation is adding lagged values (p) of first differences of the dependent variable as additional regressors which are required to account for possible occurrence of autocorrelation. The Dickey-Fuller test starts with the following first order autoregressive model:
illustration not visible in this excerpt
Subtracting Yt-1 from both sides gives
illustration not visible in this excerpt
Then the test for stationarity is conducted on the parameter γ. If g=0 or δ=1 it implies that the variable Y is not stationary. The hypothesis to be tested is formulated as follows:
illustration not visible in this excerpt
Furthermore, the use of equation (7) is appropriate only when the series Yt has a zero mean and no trend term (Harris, 1995). If a variable has a zero mean, it implies that Yt= 0 when t=0 implying no constant term. A constant (drift) is included to the regression since it is difficult to know whether the true value of Y0 is zero or not. Including a constant (α) to equation (7) gives:
However, testing for stationarity using equation (8) is invalid if a series contains a deterministic trend. Because if g=0, the null hypothesis will be accepted that the series contains a stochastic trend when there exists deterministic trend. Thus to avoid such results, it is important to incorporate time trend in the equation above and gives:
illustration not visible in this excerpt
Where, T the trend element
For the above equations (equation 8 and 9), the parameter g is used while testing for stationarity and the decision is made using t-statistics that is used for the regression without drift and trend. If the calculated value of t is less than the critical value the null hypothesis is accepted and not if otherwise. Accepting the null hypothesis implies the presence of unit root-i.e. the series is non stationary. However, the DF test has a series limitation in that it suffers from residual autocorrelation. Therefore to overcome this problem, the DF model is augmented with additional lagged first differences of the dependent variable. This is called Augmented Dickey-Fuller model.
The advantage of using ADF over that of DF model is that ADF model avoids the autocorrelation among the residuals. Therefore incorporating lagged first differences of the dependent variable to the above three equations (equations 7, 8 and 9) gives the corresponding ADF model as follows:
illustration not visible in this excerpt
Where, Yt is any variable in the model to be tested for stationarity, a is a constant (drift), T is a trend element, k is the lag length, et is an error term, and D is the first difference operator.
The null hypothesis of ADF is d=0 against alternative hypothesis that d<0. Where d=g-1. A rejection of this hypothesis means that the time series is stationary or it does not contains a unit root while not rejecting means that the time series is non stationary(Enders, 1995). If the variable that is not stationary in levels appear to be stationary after dth difference, then the variable is said to be integrated of order d I(d).
2.4.2. Test for Cointegration
Before going to examine the causal relationship between aid and economic growth, it is better to check whether the two variables are cointegrating or not. Cointegration means the regressions of one variable over the other is no meaning full (spurious). Economically speaking, two variables will be cointegrated if they have a long-term, or equilibrium, relationship between them (Gujarati D. , 2004). Despite variables are being individually non stationary, a linear combination of two or more time series can be stationary. Cointegration among the variables reflects the presence of long run relationship among non stationary variables in the system.
Testing for cointegration is important because differencing the variables to attain stationarity generates a model that does not show long run behavior of the variables. Thus testing for cointegration is the same as testing for long run relationship. In general, if variables that are integrated of order‘d’ produce a linear combination which is integrated of order less than ‘d’ say ‘b’, then the variables are cointegrated and hence have long run relationship(Gujarati D. , 2004). In order to determine whether or not a long-run equilibrium relationship exists among the unit root variables in a given model, we need to test empirically that the series in the model are cointegrated. To conduct test for co-integration, the study used Johanson (1988) maximum likelihood estimation procedure.
To conduct a test for co-integration in a multivariate framework using Johansen’s maximum likelihood procedure, first the general VAR model of relationship between the variables should have to be formulated. Thus a general VAR (p) of the following form is formulated:
illustration not visible in this excerpt
Where Xt is a (mx1) vector of stochastic I(1) variables, Wt is a (qx1) vector of deterministic variables (for instance trend and dummy variables), and each i (i=1….p) and Ψ are (mxm) and (mxq) matrices of parameters. et is a (mx1) vector of normally and independently distributed disturbances with zero mean and non-diagonal covariance matrix (vector of white noise disturbance terms), and t=1….T (T is the number of observation).
Providing the variables are (at most) integrated of order one i.e. I(1) and co-integrated also has an equilibrium error correction representation that is observationally equivalent but which facilitates estimation and hypothesis testing, as all terms are stationary. The vector error correction model (VECM) is:
illustration not visible in this excerpt
Simplifying equation (14) gives:
illustration not visible in this excerpt
From the above equation, the long run relationship among the variables is captured by the term pXt-p . The i coefficients estimate the short run effects of shocks on DXt and thereby allow the short and long run responses to differ. In the Johansen (1988) procedure, determining the rank of p (i.e. the maximum number of linearly independent stationary columns in p) provides the number of cointegrating vector between the elements in x. In this connection, there are three cases worth mentioning. (i), If the rank of p is zero it points that the matrix is null which means that the variables are not co-integrated. In such case the above model is used in first difference, with no long run information,( ii), If the rank of p equals the number of variables in the system (say n) then p has full rank which implies that the vector process is stationary. Therefore the VAR can be tested in levels, (iii), If p has a reduced rank i. e. 1<r(p)<n it suggests that there exists r<(n-1) cointegrating vector where r is the number of co-integration in the system.
The matrix p is given by (p=αβ'), where β coefficients show the long run relationship between the variables in the system(Cointegration parameters) and α coefficients show the amount of changes in the variables to bring the system back to equilibrium i.e. it shows the speed with which disequilibrium from the long run path is adjusted. To identify the number of cointegrating vectors, the Johansen procedure provides n eigenvalues (λ) characteristic roots whose magnitude measures the degree of correlation of the cointegration relations with the stationary elements in the model.
Two test statistics (ltrace and lmax) are used to test the number of cointegrating vectors, based on the characteristic roots. The statistics are calculated from the following formula:
illustration not visible in this excerpt
Where, T the sample size, li is the estimated eigenvalues
In Johanson procedure, the likelihood ratio (LR) test is used to test the significance of estimates of λi eigenvalues. The λtrace tests the null that the number of co-integrating vectors is less than or equal to r against an alternative of (r+1). The λmax statistics, on the other hand, tests the null that the number of co-integrating vectors is r against an alternative of (r+1). The distribution of both test statistics follows chi-square distribution.
Since the VAR approach assumes that all variables in the system are potentially endogenous, it is important to identify the endogenous and exogenous variables in the system. M’Amanja, Lloyd & Morrissey (2005) pointed that the weak exogeneity test gives an indication of the variables in the system with feedback effects on the long run levels of other variables but themselves are not influenced by these long run variables(M’Amanja, Lloyd, & Morrissey, 2005). This implies that if a variable is weakly exogenous its error correction term doesn’t enter the error correction model. As a result the dynamic growth equation for that variable depicts no information concerning the long run relationship in the system. Thus such variables should appear in the right hand side of the VECM. For this reason test for weak exogeneity is conducted by imposing zero restriction on the relevant adjustment parameters.
2.4.3 Vector Error Correction Model
Obtaining long-run estimates of cointegration relationships is only a first step to estimating the complete model. The short-run structure of the model is also important in terms of the information it conveys on the short run adjustment behavior of economic variables. The analysis of short-run dynamics is often done by first eliminating trends in the variables, usually by differencing. This procedure, however, throws away potential valuable information about long-run relationships about which economic theories have a lot to say.
Vector error correction model enables to capture the short run dynamics of the model and formulated based on the identified long run relationships. The VECM has cointegration relation built into the specification so that it restricts the long run behavior of the endogenous variable to converge to their cointegrating relationships while allowing for short run adjustment dynamics. The cointegrating term is known as the error correction term since the deviation from long run equilibrium is corrected gradually through a series of partial short run adjustments. Thus cointegration implies the presence of error correcting representation and any deviation from equilibrium will revert back to its long run path.
The existence of co-integration allows for the analysis of the short run dynamic model that identifies adjustment to the long run equilibrium relationship through the error correction model representation. If the number of co-integrating vector(s) is/are determined and once the endogenous and exogenous variables are identified in the system, it is possible to formulate a VECM. Using the variables of our interest in the model a system of equations is developed that portray the VECM. Hence, assuming that Yt is endogenous variable(s) and Xjt representing weakly exogenous variables in the model, we can model Yt. Yt is modeled using the lagged first difference of Yt itself, the lagged first differences of the explanatory variables and the error correcting term which is designed to capture the speed of adjustment to the long run equilibrium. The equation is represented as:
Where ECTt-1 is the error correcting term, DXjt-1 is a vector of first differences of explanatory variables, DYt is a vector of first differences of endogenous variable(s) and D is a dummy variable for major political changes. The general VECM model for growth is represented below using the respective variables used in the estimation of the long run equilibrium equation. The general VECM model for economic growth equation is specified as:
Where lag length of two is determined by Akakie Information Criterion (AIC), D and ECT represents a dummy for major political changes and error correction term respectively.
Using the above VECM specifications, a short run dynamic equation is estimated for growth. Dropping insignificant regressors from the specification (i.e. step-by-step elimination of insignificant regressors and lags from the general VECM model) following the general to specific modeling strategy, a parsimonious result for growth is estimated. In the estimation of the dynamic equation for growth, a dummy variable is incorporated to capture the influence of major political (government) changes on growth in the short run. In other words, dummy is used to see the immediate impact of major shifts in government on economic growth.
2.4.4 Causality Test in Vector Error Correction Model
The idea of Granger causality as described in (Granger, 1969) is based on the principle that a cause cannot come after its effect. When variable x is affecting a variable y, addition of the past values of the former variable should increase the precision of prediction of the latter variable (Marc., 2014). A test for causality is performed on variables of interest to detect the presence and direction of causality between pairs of variables. The variables of interest are to test causality between foreign aid and economic growth at aggregate level and foreign aid and economic growth at sectorial level by estimating a VECM for each pairs of variables. Following the VECM, causality test is made to identify the presence and direction of causality. The VECM to analyze the causal relationship economic growth and foreign aid is specified as follows:
Where (βgi, Өgi) and (βai, Өai) are coefficients of the differenced (lagged) terms of real GDP and foreign aid respectively, (Xt-1, Yt-1) is the one period lagged error correcting term for real GDP and foreign aid respectively. And (et, gt) are white noise error terms.
Causality inferences among the pairs of variables in the above models are based upon estimating the parameters of the model, subject to the predetermined number of co-integrating vectors in the system. Then hypothesis are formulated: for the real GDP equation (20A) the null hypothesis is “foreign aid does not cause GDP growth” whereas “GDP growth does not cause foreign aid” is the null for the foreign aid equation (20B). Rejection of the null of the GDP growth equation indicates the presence of causality from foreign aid to real GDP growth, or alternatively foreign aid causes Real GDP growth. Similarly, rejection of the null for the foreign aid equation points that it is GDP growth which causes foreign aid. Furthermore, the short run and long run causality can be discriminated for each equation. Absence of causality in the short run implies that the lagged coefficient values of the first difference terms of the relevant causal variable in the VECM are jointly insignificant. Whereas long run causality test is made by imposing zero restriction on the respective adjustment parameters of each equation.
Similarly, the VECM used to examine the causal relation between sectorial GDP and foreign aid allocated to sectors. The causal relation between agricultural GDP growth and agricultural sector foreign aid was examined as follow:
Where (agi,dgi) and (bai,cai) are coefficients of the difference (lagged) terms of agricultural GDP and foreign aid agricultural sector respectively, (Wt-1, St-1) is the one period lagged error correcting term for agricultural GDP and agricultural foreign aid respectively, and p and k are optimal lag lengths determined by information criteria. And (Zt, Ut) are white noise error terms.
The null hypothesis to be tested is that there is no causality between the variables in each equation whereas rejecting the null implies the presence of causality between the variables. Absence of short run causality requires that bai to be insignificant for agricultural aid not to cause agricultural GDP and similarly, dai to be insignificant for agricultural GDP not to cause agricultural sector aid for equations (21A) and (21B) respectively. On the other hand, absence of long run causality necessitates the coefficients (W and Ψ) of the error correcting term to be zero for the respective equations. Similarly, the causal relation between health and educational GDP and foreign aid allocated for health and educational sector done by VECM respectively. The causality test between the foreign aid and economic growth i determined by using F-test to test whether lagged information on foreign aid provides any statistically significant information about GDP growth in the presence of lagged GDP. If not, foreign aid does not Granger-cause economic growth.
In the following chapter, descriptive analysis of the macroeconomic performance of Ethiopia is described followed by estimation results and discussion of econometric variables. All the estimation of the empirical results is made by the use of STATA 10 software packages.
[...]
-
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X.