Đưa bài này lên theo đề nghị của học viên cao học
(1) EXTERNAL DEBT, DEBT BURDEN AND ECONOMIC GROWTH NEXUS: EMPIRICAL EVIDENCE AND POLICY LESSONS FROM SELECTED WEST AFRICAN STATES
BY
CHINEDU SAMUEL OKONKWO
SCHOOL OF ECONOMICS, UNIVERSITY OF NOTTINGHAM, UK
&
GBADEBO OLUSEGUN ODULARU
REGIONAL POLICIES AND MARKETS ANALYST,
FORUM FOR AGRICULTURAL RESEARCH IN AFRICA (FARA), ACCRA, GHANA.
ABSTRACT
There exist copious and mixed evidential outcomes on the effect of debt on output growth. With large investment/savings gaps in these countries, it seems plausible that external borrowing can influence growth positively if well utilized or negatively as the debt becomes a burden. Amidst few country-based studies, we employ annual time series data from to investigate the effect of external debt on output growth of selected West African Countries for the period 1970 to 2007.
This study initially adopts a systems approach in the form of a vector autoregressive model to identify the existent relationship. The finding from this approach suggests that the debt-growth relationship is difficult to evaluate based on the information contained in our variables and hence, inferring a possible contemporaneous effect. Thus, this study employs a single line equation based on the Engel Granger technique to cointegration in order to investigate the direct effect of external debt.
In general, the findings appeared to support that external debt stock and the burden (induced by debt servicing) had a direct negative effect on output growth for the period covered. Contrary to the debt overhang hypothesis through the effect on investment, a cursory investigation of the debt-investment relationship suggests that external debt spurs rather than impede investment. However, impact of debt servicing seems mixed as it may be expected that debt servicing ‘crowds out’ investment.
1.0 Introduction
External borrowing in itself should not be a problem at sustainable levels. However, most Sub-Saharan Africa (SSA) countries have accumulated large debt stocks since the early 70s which, compounded by structural weaknesses, may have made it difficult for advancement of growth and development. Nakatani and Herara (2007) would not agree any less that external debt may not be a solution but a problem itself. This debt burden problem has generally been observed to lead to debt overhang[1] and crowding out effect in various nations. That is, large debt accumulation of the developing nations acts as a deterrent to growth process since benefits obtained from growth are constrained by huge debt serving requirements as well as creating a disincentive effect for investment (especially private).
For the past three decades, a number of studies have been carried out to establish the nexus between external debt and economic development. Further, since early 1980’s, debt crisis has been a major issue for many nations especially developing nations. By conventional propositions, it is expected that external borrowing will serve as a source of capital formation which spurs economic growth. However, economic performance of many debtor countries has been undermined by huge debt accumulation. Given the increasingly growing concern of the debilitating impact of debt on growth, especially among developing countries, this paper attempts to investigate the pandora of mixed findings on the external debt and growth nexus. In the midst of mixed findings, it may not be totally clear of the impact of external debt on economic growth.
Consequently, there has been a call for debt relief (or debt forgiveness) as a bailout package for indebted poor countries of which SSA countries constitute approximately 81%. Many of these calls have come out of claims that debt forgiveness is pertinent to reducing the debt burden and reposition heavily indebted poor countries for significant and sustainable growth. In response to this, the Highly Indebted Poor Countries (HIPC) Initiative was flagged off in 1996 with an objective of bringing the debt stock of these nations to sustainable levels. Some of the major criteria for receiving the debt relief include the establishment of a track record of good performance under the supported IMF and World Bank programmes, as well as the satisfactory adoption and implementation of key economic reforms and Poverty Reduction Strategy Papers (PRSP), IMF (2010).
Objectives of the Study
The main focus of this study is to examine the link between external debt on economic growth for selected West African countries. These countries include Ivory Coast, The Gambia, Ghana and Senegal. The choice of these four West African countries is due to the macroeconomic growth being recorded, except the recent political crisis in Ivory Coast.
For many years now, developing nations have amassed foreign borrowings in expectation of transforming them into rapid growth and development. On the contrary, they have succeeded in getting into a debt trap which is taking its toll on their economic development plans, usually in the form of depleted resources (through massive debt repayment) which could have been gainfully invested.
Over time the debt stock of most developing countries have generally increased despite its inconsistent movements with their growth rates. This largely suggests dependence on external finance which is assumedly utilized for productive economic activities. Therefore, this paper examines this external indebtedness of the selected four West African countries, the impact of the burden and its implication for individual country economic growth. More specifically, the core objective of this study is to empirically examine the effect of debt accumulation on the economic growth in Ivory Coast, The Gambia, Ghana & Senegal.
Scope of Study
This study employs a modified endogenous growth model where debt variables are primary determinants of the real growth. The model allows for an examination of the effect of debt accumulation on economic growth. First, the paper adopts a cointegration technique under a vector autoregressive (VAR) framework to examine this effect and also employ the error correction model (ECM) based on Engel Granger 2 Step to capture the short run contemporaneous effect. The paper utilizes an annual time series macroeconomic data for the period 1970-2007.
In order to ensure that less degree of freedom is lost, especially given the sample size, some variables which are not relevant to the study objectives were jettisoned, thereby, strengthening the regression results.
we leave out variables which are not necessary for the focus of the study. Including them could weaken the results of the model. More importantly, this study restricts its analysis to four West African countries, namely Cote d’Ivoire, Gambia, Ghana and Senegal,[2] partly because such external debt issues have not been individually analysed in literature.
Organisation of the Paper
This paper follows a six-section arrangement. While the first section introduces the paper, section two briefly reviews the theoretical and empirical literature on external debt-growth nexus. Section three discusses the methodological framework and data employed in the study. While section four presents the economic results of the model, the final section articulates the conclusion and policy implications.
2.0 LITERATURE REVIEW
The External Debt-Growth Relationship
The empirical findings of Afxentiou and Serletis (1996), for developing countries, shows that there exists a negative relationship between indebtedness and national productivity from 1980-1990. This was attributed to excess debt accumulation from 1970-1980 when foreign loans were taken to cushion the shock from oil price increases in early 1970. Earlier findings of Geiger (1990) also asserts this using some highly indebted South American countries. The result of the study showed existence of a statistically significant inverse relationship between debt and economic growth from 1974 to 1986.
Using Ordinary Least Square (OLS) technique, Fosu (1996) examined the degree to which debt had a negative impact on economic growth in Sub-Saharan African countries. The result confirmed that debt directly[3] and negatively affects growth by reducing productivity and, on average, a high debt country experiences almost 1 percent of reduction in GDP growth rate annually. His findings seemed to be consistent with the ‘direct effect of debt hypothesis’ which theoretically states that for countries facing large debt repayment, debt outstanding and servicing will directly and negatively impede growth even if it does not affect investment. Fosu (1999) study reaffirmed his earlier findings that external debt directly affects Sub-Saharan African Countries negatively. Further evidence from his work also showed a weak negative effect of debt on investment levels. On the contrary, there have been few studies like Cohen (1993) who for a large dataset of developing countries found no implicative evidence of a negative effect of debt on economic growth for the period 1965 - 1989.
The degree to which external debt affects an economy varies by country. Chowdhury (1994) investigated the extent of external debt impact on GDP and vice versa using a system of simultaneous regressions. The study employed panel data for the period 1970-1988 on selected Asian and Pacific countries which include Bangladesh, Indonesia, Malaysia, Philippines, South Korea, Sri Lanka and Thailand. Results obtained from the standard simultaneous equation model showed that external debt (private and public) had only small effects on the GNP. Hence, by his findings, it could be summarised that external debt has no significant effect on economic growth.
Metwally and Tamaschke (1994) investigated the interaction between debt servicing, capital inflows and growth for 3 North African Countries (Algeria, Egypt and Morocco) for the period of 1975-1992. Using standard OLS and the Two Stage Least Square (2SLS) methods, they examined simultaneous models.[4] Their result suggests that there was a two way relationship between debt servicing and growth. Furthermore, they discovered that debt servicing affected economic growth negatively. High growth rate was also found to accelerate capital inflow which again enhances economic growth. This was observed to have a positive effect on productivity as it leads to reduction in overdependence on external borrowings as well as reducing adverse effects of debt servicing on an economy.
Furthermore, Savvides (1992) claimed that debtor nations who were unable to pay their external debts would have any debt payment to be negatively linked to economic performance. Their finding is suggestive that economic benefits that accrue to the debtor nation in terms of increments in output or exports is minimized due to debt servicing requirements.
Some findings also suggest external debt and economic growth to be linearly related. However, some researchers have found the existent relationship to be nonlinear. Among these is Patillo et al (2002) whose study empirically investigates the relationship between total external debt and growth rate of GDP for developing countries[5] over a period of 29 years, starting from 1969. They keenly conclude that the relationship between external debt and economic growth is nonlinear in the form of an inverted U shaped curve. By implication, at low levels of external debt, growth is affected positively but at higher levels of total debt, the relationship becomes negative. The authors were able to determine the exact turning point which was put at 35-40 percent of debt to GDP ratio and between 160-170 percent for debt export ratio. Besides, Patillo et al (2004) paper which establish a nonlinear relationship between debt and growth, other studies which find the existence of a nonlinear effect include Cohen (1997), and Elbadawi et al (1997).
However, Schclarek (2004) conducted a similar study like that of Patillo (2002) but using 9 developing and 24 industrial countries with datasets obtained from World Bank Development Indicators (WDI) dataset. For developing countries, the study found lower levels of external debt to be related to higher growth rates. Notwithstanding, the study did not find existence of an inverted U shape relationship between total external debt and economic growth as claimed by Patillo (2002). In the case of industrial countries, the study found no significant relationship between total government debt and economic growth. Adegbite et al (2008) was also unable to find any significant nonlinear relationship between external debt and economic growth for Nigeria.
The Debt Overhang Issue
Many studies in the literature have blamed the slow growth in developing economies on the debt overhang effect. Debt overhang is believed to create a disincentive effect which inversely affects growth. The major argument of debt overhang theory is that indebted nations have limited investment in their productive capacity. This is seen to serve as disincentives to investment (especially private) as a result of expectation about the consequential economic policies (like increased taxation) required to service debts.
Studies like those of Morisset (1991) argue that if a private sector is credit rationed then higher external debt stocks will have adverse effects on productive investment through a disincentive effect. The disincentive effect is such that where government in debtor countries are unable or unwilling to make debt repayments then private sector investors anticipate higher taxation on both financial and real assets. This seems to tie in nicely with some versions of the Ricardian equivalence which states that an increase in government debt constitutes deferred taxation. This may be seen as a first debt overhang effect. A second debt overhang effect is the creation of macroeconomic instability including anticipated inflation, increased fiscal deficit, exchange rate depreciation etc. Therefore, debt overhang effect also takes its toll on the private sector by providing disincentives to private investments[6] in productive activities.
Iyoha (1999) investigated the effect of external debt on economic growth in Sub Saharan African countries using a small macro-econometrics model estimated for the period 1970-1994. He adopts a simulation approach and finds debt overhang and crowding out effects to be significantly present. Debt overhand and crowding out were found to be negatively related to investment and hence, investment and potential economic growth is suppressed through both disincentive and crowding out effects. Were (2001) examined the structure of Kenya’s external debt and resultant implications on economic growth. Using time series data from 1970 to 1995 the study finds external debt accumulation to contribute negatively to economic growth and private investment, hence, confirming existence of debt overhang problems in Kenya. Further findings of the investigation indicated that debt servicing does not have a negative impact on growth but had some crowding out effects on private investment.
IMF (1989) claimed that debt overhang problem existed in mainly developing debtor countries in 1980’s. The study finds two evidence supporting the debt overhang proposition. Firstly, savings ratio decreased when external finance also decreased. Secondly, the study compares a group of countries with debt problems against another group of heavily indebted countries without debt servicing problems and found a significant drop in savings ratio of the former. By implication, it is deducible that where foreign finance begins to dry up, savings and investment ratio follow suit since savings is utilized to service debt (directly or indirectly through taxation), hence, potential investment from savings is also limited.
Literature also has it that debt overhang may be exaggerated and hence, may not be the true cause of economic slowdown as observed by some researchers. Hofman and Reisen (1991) rejected findings of IMF (1989) based on two pieces of evidence relating to debtor countries. Firstly, they noted that investment in debtor countries were financed by foreign savings in 1978-1981 hence, the period of study was highly exceptional. Secondly, they believed that the IMF paper considered only a group of middle income debtor countries which were arbitrarily selected, wrongly classified as ‘indebted’ countries and also having never faced a “serious debt servicing problem”. The paper claimed that there was no debt overhang in debtor countries. However, their findings suggest that transfer of financial resources from debtor countries to other countries would more convincingly pass as a better explanation for reduction in investment than levels of outstanding debt.
As a backup for Hofman et al (1991), Chowdhury (1994) findings do not support the debt overhang argument. On the debt overhang issue, the paper concludes that external debt of developing countries is not a primary cause of slowing economic growth. Another study which asserts to this finding was carried out by Bullow Rogoff (1990) whose findings suggest that there was no need for establishing institutions for organizing debt relief since debt overhang found in many studies was an exaggeration. This finding has not been taking seriously as it seemed more widely acceptable that the massive debt stock of developing nations were impeding their growth and hence, in 1996 the Highly Indebted Poor Countries (HIPC) Initiative was established by the World Bank and other donors to provide debt relief to debt ridden poor countries.
For an empirical justification using simulation study, Iyoha (1999) found that debt stock reduction would have a significant positive effect by increasing investment and rapid growth process. His finding suggests that if the debt stock were to be reduced by 20%, on average, investment would rise 18% and economic output will increase by 1% for the period 1987-1994 for SSA countries. Therefore, the finding was also a call for debt relief which acts as a stimulus for resurgence of investment and growth in developing countries. Boyce et al (2002) also supports this but suggests that a realistic and effective approach to end the debt crisis in African countries would require complete cancellation of debt. The suggestion comes due to the large setbacks of HIPC debt relief initiatives since inception. Additionally, it may be as result of fear that many of these nations at the receiving end could misuse the relief efforts. It is possible that they may also see it as licence to expand their debt stock in anticipation of further debt relief.
Evidence of Causality
There has been a compendium of findings on the causality flow between external debt and economic growth in many studies. Most of these studies conducted for both developing and developed countries have proved to be a controversial one. Some studies in literature have also found no causal relationship between external debt and economic growth. An example is Afxentiou and Serlitis (1996a) paper which employed Granger causality test. Data for the period 1970-1980 were utilized and 55 countries, classified according to the World Bank debtor country classification system, were examined. For there to be Granger causality then the expectation is such that changes in indebtedness will cause changes in per capita income. The test showed absence of any causal relationship between debt and income for the 55 developing countries investigated. Therefore, implying that the debt overhang problem largely reported in literature is only an exaggeration.
Amoateng et al (1996) utilize a sample of African countries and investigated the relationship between debt servicing, economic growth and exports. They employ the Granger’s causality test to analyze the interrelationship between exports, GNP growth and foreign debt servicing during 1971-1990. The empirical finding suggests that there is a unidirectional and negative causal link between foreign debt service and GDP growth for middle income African countries for the period 1983-1990. In essence, external debt ‘Granger causes’ decrease in economic growth. For low income African countries, they find a positive and unidirectional causality between GDP growth and foreign debt service for the 1971-1982 sub periods after excluding export revenue growth. Therefore, this implies that external debt had a positive impact on economic growth prior to 1982. Furthermore, the post 1982 sub period (1983-1990) result suggests existence of a bidirectional and positive causality between foreign debt service and GDP growth.
In an attempt to resolve the causal relationship between external debt and economic slowdown Chowdhury (1994) investigates if external debt of developing countries is “a symptom rather than a cause of economic slowdown.” He examines the effect of external debt on GNP growth rate in seven selected Asian and Pacific countries. Interestingly, foreign debt accumulation rate was found to have a positive long run effect on GNP growth rate in 3 countries. Only one country (Philippines) was found to have its external debt accumulation affected by GNP growth rate such that a 1% increase in GNP would lead to a 1.25% increase in external debt in the long run. Additionally, by means of the Granger causality test, the study firmly rejects the Bullow- Rogoff (1990) proposition that “external debt of developing countries are a symptom rather than a cause of economic slowdown”. However, they accept the proposition that external debts of developing nations are not a primary cause of economic downturn. Chowdhury (1994) was also able to show that a bidirectional (or feedback) relationship existed between foreign debt accumulation rate and GNP growth rate for two countries. On the whole, he finds that any increase in GNP will bring about a rise in external debt but external debt on the other hand has no negative effect on economic growth. What this finding has successfully done is to prove that the debt-growth relationship is highly country specific.
In a more recent study, Karogol (2002) utilized cointegration technique and employed standard production function to investigate short and long run relationship between economic growth and external debt service for Turkey for the period 1956-1996. The variables were found to be negatively related in the long run and the Granger causality test results show a unidirectional causal flow form debt service to economic growth. More recently, Butt (2009) also examined the causal relationship between economic growth and short term external debt for 27 Latin American and Caribbean countries from 1970-2003. Out of a total of 13 countries found to exhibit Granger causality, several were also found to have bidirectional causal relationships.
3.0 METHODOLOGY AND DATA
Introduction
Despite the controversial discoveries on the relationship between external debt and output growth, most findings, in general, support a negative effect of foreign debt accumulation on economic growth especially in developing countries. Many of these studies have employed different empirical approaches to uncover the relationship for different time frames. However, most studies use neoclassical growth models augmented with other factors such as debt and export which are not captured in the traditional neoclassical growth model. Additional variables used to augment the model are dependent on the motivation of the study.
Model Specification
Not many similar studies have employed a vector autoregressive (VAR) technique in examining the relationship between external debt and economic growth. However, some studies like Frimpong (2006) have employed this technique. Our model initially follows a VAR framework. VAR models were generally popularized by Sims (1980). The model is a multiple variable system with flexibility and ease of generalisation. In the system, each variable is a potential endogenous variable which is explained only by its own lags as well as those of other variables. In essence it includes no contemporaneous terms which may be a slight drawback depending on the information contained in the data. Furthermore, VAR models are known to be ‘atheoretical’ since they ignore economic restrictions. Compared to univariate time series model, VAR models provides several advantages such as having the ability of incorporating both endogenous and exogenous variables into a single system, testing for and applying restrictions as well as providing a flexible and “rich structure, implying that it can capture more features of the data” Brooks (2008).
Furthermore, we observe that the VAR may be weak in providing estimates for contemporaneous terms. Studies like Sen et al (2007) and Were (2001) employ both contemporaneous terms to investigate the growth- debt relationship and found this to exhibit significant estimates. Hence, we also employ the Engel-Granger (EG) 2 step method to cointegration in explaining our short run model. By doing so, we get investigate the direct debt effect by observing the contemporaneous effect of debt indicators on the output growth.
This paper proceeds with a modified specification of the growth model similar to that estimated by Sen (2007). Also similar to Butt (2009) and Schclarek (2004) full log model , our model takes a semi log linear specification[7]. It is useful to specify the model in natural logarithms because it expresses estimated regression coefficient in elasticities hence, making it easy for discussion. Also, first differences of our model represent growth rates[8]. The model estimated for our four countries also incorporates debt variables to access the effect of external debt on the economic growth. Without recourse to the formulation of conventional growth model, we focus on the following growth model:
(4.1)
RYPC = real GDP per capita
FDIY = foreign direct investment as a percentage of GDP
EXDY = total external debt as a percentage of GDP
GDIY = gross domestic investment as a percentage of GDP
TDSE = total debt service as a percentage of exports
Where α0 is the constant term and α1,s to α4,s are coefficients of the explanatory variables. ε is the stochastic error term. The subscript ‘it-s’ refers to country and lagged time period in equation (1) above.
In the model, real GDP per capita is used as dependent variable since this measure has extensively been used as proxy for real growth. External debt as a percentage of GDP will be used to show the impact of external debt on output growth. This can also be used to make inference on possibility of problems of debt overhang. Among other variables, total debt service to export ratio[9] is a conventional variable used as a reflection of debt burden. Foreign direct investment to GDP ratio captures the net external capital inflow. Though we could not obtain individual data for public and private investment on economic growth, we examine the effect of the gross domestic investment on output growth especially with reference to external debt accumulation.
To examine the debt-growth relationship, many studies have used a cross-country regression which lumps up data from various countries that are structurally heterogeneous and size variant. For this reason, we follow a co-integration technique for analysis of the model taking into consideration country specific structures. The advantage of the co-integration technique is that it enables us test for a long run relationship among the selected variables. Estimating a simple OLS regression may be inappropriate if the variables are not integrated of the same order. In essence, the regression may become spurious hence, yielding inconsistent estimates. Therefore, to ensure we do not have a spurious regression, unit root tests are conducted before the co-integration test after which the regression estimates could be meaningfully discussed.
Methods of Analysis
Working with most time series econometric datasets can be very challenging especially for African countries. Where data used are not integrated of the same order then results may be unrevealing. To ensure that our estimates can be effectively evaluated, it is necessary to employ tests to check that our variables are useable in our as specified in our models. we start by testing for unit root in variables before proceeding to the cointegration test.
1. Unit Root Test
Using macroeconomic time series data in a co-integration analysis requires all the variables to be integrated of the same order. Therefore, the first step is to determine the order of integration or the non stationarity property of the variables. Given that a series, say Zt is integrated of order d, that is Zt ~ I(d) then it must be the case that the series contains d unit roots and can only be stationary if differenced d times. Where the series is I(0) then it is void of unit roots and hence, stationary. To determine the order of integration, the Augmented Dickey Fuller (ADF)[10] and the Phillips Perron (PP) tests are employed.
For the ADF test, the Akaike Information Criterion (AIC) and Schwarz Criterion (SC) are employed for selection of the appropriate lag length of the model. Selection of the appropriate lag length is necessary to ‘whiten the residual’ Asteriou et al (2007). When the lag length selected to run the test is small then there are possibilities of choosing the wrong model where as if the lags are too many then the power of the test may be weakened. In the case of mixed results from the two criteria then the Hannan-Quinn Criterion (HQC) is employed as a check as well as scrutinizing the model estimates to determine which model minimizes the AIC and SC values. Unlike the ADF tests, PP unit root tests allow for autocorrelated residuals and hence, it is employed for further check of order of integration of variables.
Since we use annual time series data, we run our unit root tests automatically for individual series up to a maximum of 2 lags after which the model with the best lag length is tested for unit root. The null hypothesis of the unit root test states that the variable is non stationary and hence, contains unit root. This null is rejected if the calculated t-statistic is greater, in absolute terms, than the critical values of the test statistics.
2. Cointegration Test
The test for cointegration involves two major steps, the first which is the unit root test. The second step is the cointegration test which is employed to determine if the variables in the system have a long run relationship. Different methods can be used in performing the cointegration test. A first is the residual based test. Before this is done, it must be the case that all individual variables are integrated of the same order. A cointegrating regression is ran and the residuals are tested for unit root. Where there is absence of unit root then the null hypothesis that there is no cointegration is rejected.
We employ the EG 2-step and compare it to a second method for testing for cointegration-Johansen’s test for cointegration. However, there is possibility of our variables forming a endogenous system, the Johansen’s method to cointegration can effectively avoid simultaneous equation bias. Since we use more than two variables, we may be able to obtain more than one cointegration relationships. This means that variables in the model form several equilibrium relationships which may account for joint movement of all the variables in the model.
Therefore, we also perform the Johansen’s test for cointegration[11] based on vector autoregressive (VAR) framework which is more appropriate. This approach makes use of a maximum likelihood procedure in estimating and determining the rank of cointegrating vectors. Algebraically, if we assume that a vector of p variables, Yt = (Y1t,...,Ypt), generated by a VAR of order k then this can be written as
Yt = b0 + Π1Yt-1 + . . . + ΠkYt-k + µt (4.2)
where Yt is an nx1 vector of I(1) variables, b0 is an nx1 constant vector, Πi is an nxn matrix of unknown parameters to be estimated (with i=1,2,3,…k ) and µt is an nx1 independent and Identically distributed (i.i.d) vector of error terms assumed to be white noise. To use Johansen test, the VAR model must be transformed into a vector error correction model (VECM) which is specified as equation (4.3) below:
∆Yt = Г1∆Yt-1 + . . . + Гk-1Yt-k-1 + ΠYt-k + µt (4.3)
Where ∆Yt is now I(0)
The square matrix which is to be estimated (Π) is also known as the long run coefficient or impact matrix. The rank of this matrix indicates the number of cointegrating vectors in the system. If rank (Π) = r then we reject the null of no cointegration if 0 < r ≤ P. If this is the case, then the P variables have a long run relationship with r cointegrating vectors. Furthermore, Π= αβ’ where α measures the average speed of adjustment and β is the matrix containing cointegrating vectors.
The Johansen’s method suggests use of two test statistic measures to determine the number of cointegrating vectors in the VAR system. These measures include the trace statistic (λtrace) and the maximum eigenvalue statistic (λmax). The λtrace is a joint test with the null that the number of cointegrating vectors is either less than or equal to rank (r) whereas the alternative states that the ranks are more. Separate tests are performed by λmax on individual eigenvalues with the null that the rank of cointegrating vectors is r against an alternative of r +1. Where r = 0 then there is no cointegration.
3. Error Correction Model (ECM)
Once cointegration has been established, then it usually is the case that all variables are integrated of the same order and hence, the possibility of having a spurious regression is null. The Engel and Granger (1987) show that if variables are cointegrated then there must be an error correction representation. For the VAR case, we get a VECM which acts as a special type of restrictive VAR and shows changes in the dependent variable as a function of long run speed of adjustment parameters, captured by the Error Correction Term (ECT), and short run dynamics. Since VEC models are specified in first difference, this means all variables in the system are stationary. Hence, there is data consistency such that the conventional t-statistic can be employed to analyse the model parameters.
An error correction model in a single line equation will be employed to capture the short run direct of our regressors on the responding variable (dependent variable). Though it does not produce a system of equations as presented in VECM, its celebrated merit lies in the fact that it allows for examination of individual equations and permits control of the equation by adding desirable variables (at contemporaneous and lagged terms) that is specific to each single line equation. For this reason, the advantages of the VECM are not obvious especially when it has only one cointegrating vector.
Data
The empirical investigation of this paper employs data for four West African countries. The data retrieved were all annual time series datasets covering the period 1970-2007 for all countries. Real GDP per capita (2000 constant prices), total external debt, total debt service, exports of goods and services, and current GDP were all drawn from the World Bank Development Indicators (WDI) database and measured in units of US dollars. Foreign direct investment as a percentage of GDP was also obtained from the WDI database. Gross domestic investment as a percentage of GDP was obtained from the African Development Indicator (ADI). All the variables for the model are expressed in natural logarithm with the exception of foreign direct investment as a percentage of GDP. This is because, as a net inflow[12] of investment to GDP ratio, it contains negative terms and hence, it could not be logged. Additionally, all relevant series were expressed in real terms[13].
[1] The debt overhang can be loosely defined a situation in which the debt stock of a county is so high that it cannot repay. See Krugman (1988) for more exposition.
[3] The result does not support any adverse indirect effect where economic growth is affected indirectly by debt.
[4] simultaneous model was examined because they believed that debt growth relationship was not explained by a one way relationship.
[7] This is because a variable was found to have negative numbers and hence, we could not log it.
[8] This is so done because using growth rates in our model may not provide us with the relationship we seek to investigate. This is because when the growth rates are differenced then what we have is the growth of the growth and no longer the growth.
[9] Note that total debt service to export ratio is at times used interchangeably with total debt service as a percentage of exports where the later is only in percentages.
[13] We initially deflate the current series into real terms using the GDP deflator but discovered that taking the ratio gives off the deflator and hence, presenting similar results as when the ratio were derived in current terms. This seems to be the reason why literature are flooded with ratios in current terms.
Không có nhận xét nào:
Đăng nhận xét